Up and down the Cline - The Nature of Grammaticalization
Typological Studies in Language (TSL) A companion series t...
50 downloads
1255 Views
2MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Up and down the Cline - The Nature of Grammaticalization
Typological Studies in Language (TSL) A companion series to the journal Studies in Language
General Editor
Michael Noonan
Assistant Editors
Spike Gildea, Suzanne Kemmer
Editorial Board Wallace Chafe (Santa Barbara) Bernard Comrie (Leipzig) R. M. W. Dixon (Melbourne) Matthew Dryer (Buffalo) John Haiman (St Paul) Bernd Heine (Köln) Paul Hopper (Pittsburgh) Andrej Kibrik (Moscow) Ronald Langacker (San Diego)
Charles Li (Santa Barbara) Edith Moravcsik (Milwaukee) Andrew Pawley (Canberra) Doris Payne (Eugene, OR) Frans Plank (Konstanz) Jerrold Sadock (Chicago) Dan Slobin (Berkeley) Sandra Thompson (Santa Barbara)
Volumes in this series will be functionally and typologically oriented, covering specific topics in language by collecting together data from a wide variety of languages and language typologies. The orientation of the volumes will be substantive rather than formal, with the aim of investigating universals of human language via as broadly defined a data base as possible, leaning toward cross-linguistic, diachronic, developmental and live-discourse data.
Volume 59 Up and down the Cline - The Nature of Grammaticalization Edited by Olga Fischer, Muriel Norde and Harry Perridon
Up and down the Cline The Nature of Grammaticalization Edited by
Olga Fischer Muriel Norde Harry Perridon University of Amsterdam
John Benjamins Publishing Company Amsterdam/Philadelphia
8
TM
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 New Reflections on Grammaticalization II Conference (2002) : University of Amsterdam) Up and down the Cline - The nature of grammaticalization / edited by Olga Fischer, Muriel Norde and Harry Perridon. p. cm. (Typological Studies in Language, issn 0167–7373 ; v. 59) Papers presented at the New Reflections on Grammaticalization II Conference held at the University of Amsterdam on April 4-6, 2002. Includes bibliographical references and indexes. 1. Grammar, Comparative and general--Grammaticalization-Congresses. I. Fischer, Olga. II. Norde, Muriel, 1968- III. Perridon, Harry. IV. Title. V. Series. P299. G73N484 2004 415-dc22 isbn 90 272 2968 6 (Eur.) / 1 58811 504 6 (US) (Hb; alk. paper) isbn 90 272 2969 4 (Eur.) / 1 58811 505 4 (US) (Pb; alk. paper)
2004041137
© 2004 – 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 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
TSL[v.20020404] Prn:7/05/2004; 13:22
F: TSL59CO.tex / p.1 (v)
Table of contents
Preface Introduction: In search of grammaticalization Olga Fischer, Muriel Norde, and Harry Perridon
vii 1
On directionality in language change with particular reference to grammaticalization Martin Haspelmath
17
Rescuing traditional (historical) linguistics from grammaticalization theory Brian D. Joseph
45
The English s-genitive: A case of degrammaticalization? Anette Rosenbach An investigation into the marginal modals dare and need in British present-day English: A corpus-based approach Martine Taeymans Redefining unidirectionality: Is there life after modality? Debra Ziegeler
73
97 115
From pronominalizer to pragmatic marker: Implications for unidirectionality from a crosslinguistic perspective Foong Ha Yap, Stephen Matthews, and Kaoru Horie
137
Conditionals and subjectification: Implications for a theory of semantic change Jacqueline Visconti
169
Unidirectionality in the grammaticalization of modality in Greek Anastasios Tsangalidis How cognitive is grammaticalization? The history of the Catalan perfet perifràstic Ulrich Detges
193
211
TSL[v.20020404] Prn:7/05/2004; 13:22
F: TSL59CO.tex / p.2 (vi)
Table of contents
Perfect and resultative constructions in spoken and non-standard English 229 Jim Miller Grammaticalization and standardization Lea Laitinen
247
External factors behind cross-linguistic similarities Ilona Herlin and Lari Kotilainen
263
What constitutes a case of grammaticalization? Evidence from the development of copulas from demonstratives in Passamaquoddy Eve Ng
281
Multi-categorial items as underspecified lexical entries: The case of Kambera wàngu Marian Klamer
299
The acquisition of polysemous forms: The case of bei2 (“give”) in Cantonese Kwok-shing Wong
325
Phonetic absence as syntactic prominence: Grammaticalization in isolating tonal languages Umberto Ansaldo and Lisa Lim
345
Grammaticalization of word order: Evidence from Lithuanian Sergey Say
363
Language index
385
Name index
387
Subject index
393
TSL[v.20020404] Prn:13/04/2004; 15:27
F: TSL59PR.tex / p.1 (vii)
Preface
The 1980s and 1990s witnessed a rapidly increasing interest in grammaticalization phenomena. Such was the number of people engaged in this branch of historical linguistics that an entire conference was dedicated to it – the New Reflections on Grammaticalization Conference, held at the University of Potsdam in 1999. The proceedings of this conference, edited by Ilse Wischer and Gabriela Diewald, appeared in the Typological Studies in Language series in 2002 with John Benjamins. Due to its success, a follow-up to this conference was organized by the editors of this volume at the University of Amsterdam, April 4–6, 2002, which was attended by over a hundred participants. Judging by the vast range of theoretical issues and languages covered, as well as the liveliness of the discussions, the popularity of grammaticalization studies and its significance for the field of historical linguistics looks unabated. This volume contains a selection of the papers presented at the New Reflections on Grammaticalization II Conference. Rather than presenting a random compilation, we have chosen to make this a thematically coherent volume, and invited papers that specifically addressed topics relating to a modification of the ‘theory’ of grammaticalization, such as the principle of unidirectionality, alternative grammaticalization pathways (or chains), the synchronic (internal or external) factors that may interfere with what is seen as ‘usual’ in grammaticalization processes, and internal or universal factors which make grammaticalization a truly separate mechanism in language change. We have selected seventeen papers, written by leading theoreticians and specialists covering a number of different language families. A large number of people have assisted and supported us in the organization of the conference and in editing this volume. We gratefully acknowledge the financial support of the Royal Netherlands Academy of Arts and Sciences (KNAW). We would also like to thank all the contributors and participants, who by their expertise and enthusiasm made the conference such a success. Special thanks are due to the students who helped us during the conference: Margot van den Berg, Annerieke Boland, Corrien Blom, Pieter Claeys, Marion Elenbaas, Tessa Liebman and especially Ernie Ramaker, who also took care of the conference website. We would also like to thank Anke de Looper and Kees Vaes of Benjamins Publishers and the editor of the series for their help towards the publication of this volume, and Jetty Peterse for her help with the index. Finally, we are particu-
TSL[v.20020404] Prn:13/04/2004; 15:27
F: TSL59PR.tex / p.2 (viii)
Preface
larly indebted to all the referees who generously commented on the first round of papers and made innumerable suggestions for improvement: Johan van der Auwera, Adrie Barentsen, Lyle Campbell, Holger Diessel, Marcel Erdal, Bernd Heine, Kees Hengeveld, Brian Joseph, Ans van Kemenade, Marian Klamer, Roger Lass, Jouko Lindstedt, Bettelou Los, Helena Raumolin-Brunberg, Eva SchultzeBerndt, Elizabeth Traugott, Seongha Rhee, Sali Tagliamonte, Jeroen Wiedenhof and Debra Ziegeler.
Amsterdam, November 2003 Olga Fischer Muriel Norde Harry Perridon
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.1 (1)
Introduction In search of grammaticalization Olga Fischer, Muriel Norde, and Harry Perridon University of Amsterdam
.
Introduction
During the past two decades, the revived interest in grammaticalization phenomena has firmly established itself as one of the most flourishing branches of historical linguistics. It boasts some of the most widely used textbooks in the field (Heine et al. 1991; Hopper & Traugott 2003 [1993]), as well as a veritable flood of case-studies and now even its own lexicon (Heine & Kuteva 2002). Judging by the sheer volume of pages devoted to it, it would seem that grammaticalization ‘theory’ provides the principal framework to account for the origin of and changes within grammar and hence a forceful tool for reconstruction. And indeed, in the early days of grammaticalization studies it was generally assumed that it had generated (virtually) exceptionless laws of grammatical change (e.g. Lehmann 1995 [1982]; Traugott & Heine 1991; Haspelmath 1999). In present-day historical linguistics there are no indications of a fading interest in grammaticalization – our notion of grammaticalization is continuously being expanded and refined (see e.g. Giacalone Ramat & Hopper 1998 and Wischer & Diewald 2002). But perhaps more importantly, its very nature as a ‘theory’, and some of its seemingly unchallengeable principles, including the principle of unidirectionality, have increasingly become the object of critical examination. In 2001 the journal Language Sciences devoted a special issue to grammaticalization in order to provide “a critical assessment”. The editor of this issue (Lyle Campbell) along with the contributors (Richard Janda, Brian Joseph, Frederic Newmeyer and Muriel Norde) pointed to a number of problems that the theory raises. These pertain to some widely accepted principles of (the theory of) grammaticalization, such as the principle of unidirectionality and the idea that grammaticalization is primarily driven by semantic or pragmatic factors, but the
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.2 (2)
Olga Fischer, Muriel Norde, and Harry Perridon
contributors to this issue also took stock of the phenomenon of grammaticalization itself. They critically evaluated the claim that grammaticalization is a unitary, continuous process and that it has explanatory value. They also considered the mechanisms that are said to play a role in grammaticalization (such as semantic bleaching, phonetic reduction, analogy and reanalysis), to see in how far they are connected or can be seen to work independently. The issues raised were not settled in this volume, which threw up quite a bit of dust in the grammaticalization field. It seemed appropriate therefore to take up some of these issues again by means of case studies, paying special attention to both the philological details involved in any development that could be seen as a case of grammaticalization as well as the theoretical implications of each case. This has resulted in the present collection of papers, some of them pursuing the points raised in the Language Sciences volume, some of them, on the other hand, criticizing these points and the examples that had been brought up to support them. What all the papers have in common, however, is that they present truly new reflections on grammaticalization, enriching our notion of what (de)grammaticalization entails, and what it does not entail.
. The principle of unidirectionality One of the most pressing themes in recent research on grammaticalization is whether or not it is possible for a grammatical item to become less grammatical, in other words, to move ‘up’ on Hopper and Traugott’s (2003 [1993]) ‘cline of grammaticality’ (content item > grammatical word > clitic > inflectional affix). In earlier work it was generally believed that such degrammaticalization is not possible, see e.g. Lehmann (1995 [1982]: 19): “[. . . ] no cogent examples of degrammaticalization have been found. This result is important because it allows us to recognize grammaticalization at the synchronic level. Given two variants which are related by the parameters of grammaticalization [. . . ] we can always tell which way the grammaticalization goes, or must have gone. The significance of this for the purposes of internal reconstruction is obvious [. . . ].” In recent years, however, the unidirectionality hypothesis has been criticized both on theoretical (e.g. by Newmeyer 1998) and methodological grounds (e.g. Lass 2000) as well as with reference to a number of well-described changes in which the directionality of grammatical change appears to be reversed. Martin Haspelmath’s paper in the present volume acknowledges that some of this criticism is justified and that some counterexamples do exist, which implies that an absolute universal has to be weakened to a statistical universal, but he emphasizes at the same time that the number of exceptions is very small. Referring to ‘true’ exceptions to unidirectionality as cases of ‘antigrammaticalization’, he lists eight such
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.3 (3)
Introduction
cases, including the English and Continental Scandinavian s-genitive (from affix to clitic) and Irish muid (from 1st person plural subject suffix to independent pronoun). For all other cases that have been put forward as exceptions Haspelmath identifies seven types of change that according to him are not instances of antigrammaticalization, such as ‘delocutive’ word-formation (i.e. a metalinguistic use of grammatical words, as in ifs and buts, isms), back-formations (teens, burger) or adverb-to-verb/noun conversions (to up, to off ). Another type of change which is not antigrammaticalization in Haspelmath’s view is a process he terms ‘retraction’, which means that less grammaticalized items regain ground at the expense of their more grammaticalized counterparts. While accepting the existence of counterexamples, Haspelmath concludes that unidirectionality is still a useful generalization, as the main goal of historical linguistics ought to be an understanding of language change, not merely describing individual changes, and that for this reason it is important to look for universals of language change. Many of these universals take the form of directionality constraints, including the assumed unidirectionality of grammaticalization. Looking at the phenomenon of grammaticalization from a traditional historical linguistic point of view Brian Joseph doubts that grammaticalization theory has added much to our understanding of the processes and mechanisms that play a role in language change. Joseph shows, first of all, that traditional, Neogrammarian methods were not only concerned with the phonetic/phonological level, as is often thought, but also with the grammatical level. In other words, it is not the case that the theory of grammaticalization fills a gap in the traditional historical methods. Similarly, he indicates that grammaticalization theory makes use of notions, such as phonetic reduction, analogy and reanalysis, that were well understood in earlier theories. The only difference is that within grammaticalization theory these notions or mechanisms are linked together so as to form one continuous process of change rather than a series of separate steps, not necessarily connected, within a particular development. Joseph’s main point, however, is that this combining of mechanisms, which is so typical for grammaticalization theory, often leads to an ahistorical way of looking at changes in that there is a tendency not to take full account of all the different steps present in any particular development. There is a real danger within grammaticalization studies to give greater weight to typological or cross-linguistic tendencies than to the details of the language undergoing the development in question. This may result in a neglect of details, both of a historical and a synchronic nature. Concerning the latter, Joseph shows that grammaticalization studies are indeed often also asynchronic in that they do not take into consideration the synchronic system within which a particular construction functions and develops. In her paper on the historical development of the English s-genitive, which is often quoted as a counterexample to the unidirectionality hypothesis (among
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.4 (4)
Olga Fischer, Muriel Norde, and Harry Perridon
others, by Haspelmath in this volume), Anette Rosenbach seems to pay heed to Joseph’s admonitions. Following Allen (1997), she assumes that this development proceeded in several steps and that it is closely interrelated with other developments within the NP, in particular the loss of NP-internal agreement and the rise of phrasal N+N constructions such as brother-in-law. She furthermore shows that the development towards a phrasal possessive marker is closely interwoven with the new function of -s as a definite determiner. As to the status of the Modern English s-genitive, Rosenbach argues that -s is a ‘clitic-like’ element – it is clearly no longer an inflectional suffix, as it was in Old English, but not (yet) a full clitic either. In order to answer the question of whether this development ought to be regarded as a case of degrammaticalization, Rosenbach introduces the distinction between ‘type grammaticalization’ and ‘token grammaticalization’ (based on Andersen’s 2001 distinction between ‘change schema’s’ and ‘historical changes’). As far as the s-genitive is concerned, the relevant change schema is Hopper and Traugott’s (2003[1993]) ‘cline of grammaticality’. According to Rosenbach (contra Norde 2001), English -s did not move back along this cline from affix to clitic, but instead became part of the newly emerging article system. Following the type-token distinction, Rosenbach argues that -s looks like a case of degrammaticalization on the token level, but since she defines unidirectionality as a property of types (i.e. clines), and since a token can form part of several clines, she concludes that unidirectionality on the type level is preserved. She stresses, however, that the interpretation of the history of the s-genitive depends heavily on the definition of (de)grammaticalization, and that therefore priority should always be given to the data, which may give rise to different interpretations. Another frequently discussed possible counterexample to the unidirectionality of grammatical change is the topic of Martine Taeymans’ paper on the English marginal modals dare and need and the semi-modals dare to and need to. Drawing data from the British National Corpus World Edition (BNC), which contains approximately 10 million words of spoken and 90 million words of written presentday British English, she shows how DARE and NEED oscillate between main verb and modal verb morphology and syntax. Their histories are totally different, however – whereas NEED is originally a main verb which acquired modal characteristics, DARE derives from an Old English preterite-present which acquired main verb characteristics at a later stage. For this reason, the history of DARE is commonly regarded as an example of degrammaticalization (e.g. in Beths 1999). From the data on NEED it becomes evident that semi-modal need to has become the predominant form, at the expense of modal need. Though this appears to violate the unidirectionality hypothesis, Taeymans argues that unidirectionality is observed on the semantic level, in that need to changes its meaning from internal to external necessity (previously a function of modal need). According to van der Auwera and Plungian (1998) such a shift is in accordance with claims about semantic unidi-
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.5 (5)
Introduction
rectionality. The author’s frequency analysis of DARE, on the other hand, does not yield conclusive results. Assuming that a difference between written and spoken PDE reflects a diachronic development, it would seem that modal use is increasing (52 per cent in the spoken parts of the BNC versus 35.6 per cent in the written ones, excluding the petrified expressions How dare you and I dare say). Taeymans points out, however, that an increase in modal dare seems counter-intuitive in the light of two facts. First, modal verbs tend to be replaced by periphrastic expressions and secondly, modal dare has become very rare in American English, which is generally considered less conservative than British English. Taeymans is therefore unable to identify a clear direction in which DARE may be considered to be heading. As a result, it cannot be established whether DARE is grammaticalizing or degrammaticalizing.
. Semantic change In her article on a seemingly counterdirectional change in the meaning and use of the Chinese modal dei/dé Debra Ziegeler arrives at the same conclusion as Taeymans, viz. that the hypothesis is not at risk if unidirectionality is defined at the semantic level. In Mandarin Chinese dé was originally a main verb, meaning ‘to obtain’, later it developed a permissive meaning when followed by a verb (phrase), which in its turn gave rise to the meaning of obligation. In present-day Mandarin the modal can be used epistemically, but, what is more surprising, also as a (new) main verb meaning ‘to need’, which need not be followed by another verb but can also take a noun (phrase) as its complement. If category change is automatically defined as grammar change we are faced here with a clear counter-example to the unidirectionality hypothesis, since one and the same item turns from a main verb (‘to obtain’) into an auxiliary (‘be allowed’, ‘have to’) back into a main verb (‘to need/require’) again. Assuming that grammaticalization begins with conceptual changes Ziegeler suggests that semantic change is unidirectional in the sense that a meaning A can change into meaning B, but that this meaning B cannot change back to meaning A again. These changes may occasionally lead to less specific meanings, in which case they may or may not be accompanied by changes in form. A series of such changes may give rise to a grammaticalization chain. But the endpoints of such chains may be subject to new changes, which eventually may lead to a new lexification of the item in question.1 Unidirectionality of semantic change as defined by Ziegeler is also assumed by Foong HaYap, Stephen Matthews and Kaoru Horie in their paper on the ways in which (pro)nominalizers in three unrelated Asian languages (Japanese, Mandarin Chinese and Malay) turn into pragmatic markers. Japanese no is originally
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.6 (6)
Olga Fischer, Muriel Norde, and Harry Perridon
a genitive marker: Taroo no hon ‘Taro’s book’, which turns into a kind of pronoun when the head noun is not expressed: Taroo no ‘Taro’s one, ‘the one that is Taro’s; in a later stage a clause may take the place of the possessor noun: Taroo-ga kat-ta no ‘[Taro bought]’s one, the one Taro bought’. It developed further into a complementizer, which presents the nominalized sentence as a fact. Used without a matrix verb, no functions as a stance marker, expressing the unchallengeable status of the information encoded in the sentence preceding no. Chinese de has approximately the same range of meanings, but parts of its history are different. Originally a noun meaning ‘bottom, end’, it came to be used as a demonstrative and an interrogative, and later on as a relative, and as a nominalizing element. From this pronominal use evolved its function as a genitive marker. As a stance marker de has the same value as Japanese no. Malay empu-nya was originally a noun phrase consisting of the noun empu ‘master’ and a third person clitic, which became a verb (‘to own’), and later on a genitive marker, a nominalizing pronoun and finally a stance marker. The various paths taken by Japanese no, Chinese de and Malay (em)punya all lead to the same goal, which raises some doubts as to the universality of the direction of semantic change. Yap et al. try to save the notion of unidirectionality in semantic change by claiming that it holds for changes across ontological domains, but not for changes within one and the same domain. Since both the relative, the pronominal and the genitive uses of the elements in question belong to the same domain of first-order entities (persons, things and the like, cf. Lyons 1977) grammaticalization may (and does) proceed in one way or the other. Changes from uses with third order entities (propositions) to uses with second or first order entities, e.g. from complementizer to relative pronoun, or genitive marker, are on the other hand not to be expected. A more modest view on unidirectionality in semantic change is expressed by Jacqueline Visconti. In a detailed description of the development of the verb suppose from an objective, lexical verb via a subjective, evaluative verb into a conditional connective (or conjunction), she shows how the path of development is an instance of subjectification similar to the process described for modals, deictic elements, discourse markers etc. in such works as Sweetser (1990), Kuteva (2001) and Traugott and Dasher (2002). In each case the original lexical element first operates in the propositional domain, and moves via the textual domain into the epistemic domain. Visconti shows that in such a subjectification process, it is the semantic pragmatic change that leads the way, while the grammatical developments follow. The process is made visible by indicating the shifts in contexts the verb undergoes (suppose appears more and more often in interrogative and imperative contexts, and contexts containing ‘related free adjuncts’), the way in which the verb may be coordinated with other verbs and adverbs, and the changes in ‘control’ and in the operation of scope from the narrative subject to the Speaker/Writer (who need not be the syntactic subject of the clause). Visconti indicates the generality of this
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.7 (7)
Introduction
development by comparing the process in English with developments involving the cognates of the verb suppose in French and Italian. In all cases the stages are the same. Visconti argues, like Ziegeler, that certain types of semantic change (in this case subjectification) may be unidirectional. This semantic unidirectionality should not be seen as a principle of this type of grammaticalization but rather as a strong tendency that may serve as a useful heuristic tool. Serious doubts as to the assumed unidirectionality of semantic change are expressed by Anastasios Tsangalidis, who shows that the diachronic development of modal markers in Modern Greek poses a number of problems for the ‘unidirectionality’ principle. Applying Bybee et al.’s (1994) path of development suggested for grammaticalization within the domain of modality, to the Greek case, he comes to the conclusion that the suggested pathway, which involves a conceptual cline (from ‘agent-oriented’ via ‘speaker-oriented’ and ‘epistemic’ to ‘subordinate’) as well as a formal one (from lexical modal elements to particles and affixes) is not always followed in Greek. The first problem encountered is the imperative mood, which is expressed inflexionally but does not seem to have been derived from earlier lexical material, as predicted by the pathway. Also problematic is the development of tha from a lexical verb expressing ‘desire’ into a pure future marker, where the formal reduction of thela to tha does not neatly follow the conceptual development. The same is true for the development of na, which grammaticalized from a purposive conjunction (hina ‘so that’) to a general purpose subordinator (‘that’) and to a particle expressing various modalities, and was even reduced to semantically zero. Again the formal changes do not historically correlate with the conceptual changes. Especially troublesome is the fact that subordinating uses of na are historically earlier than the main clause uses, which according to Bybee et al.’s pathway should be the other way around. Like Tsangalides, Ulrich Detges questions the idea that the forces which trigger grammaticalization are primarily cognitive in nature, and more especially the suggestion of theorists like Bernd Heine and Tania Kuteva that grammaticalized structures for e.g. tense and aspect derive from a small number of basic ‘source concepts’, and that the grammaticalization development from source to target is pretty similar across languages.2 In his article on the development of the Catalan perfet perifràstic (va cantar, ‘he goes sing’, i.e. ‘he sang’), Detges first of all shows that in Catalan the go (+to) infinitive does not follow the ‘normal’ route from sourcepath-goal to ‘future’ but instead develops into a past tense marker. Secondly, he argues that the driving forces are discourse-pragmatic rather than cognitive, showing how the perfet peripfàstic (the past tense) in Catalan developed in narrative texts.3 In his discussion, Detges indicates that a number of general narrative principles play an important role in Catalan, but also in other Romance languages, such as the desire to use inchoative verbs in order to mark an event as more dynamic and spectacular and thus to increase listeners’ involvement. In a later stage
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.8 (8)
Olga Fischer, Muriel Norde, and Harry Perridon
this discourse marker may come to be used to mark the ‘turning point’ in narrative sequences, and indeed any major (sub)episode boundary. At this stage the marker still occurs with a restricted number of verbs. When its use becomes more frequent, it tends to lead to ‘overuse’; the marker being used with more and more verbs, so that it becomes conventionalized. This last stage was reached only in Catalan, and not in the other Romance languages. Detges suggests that this may well be due to the absence of a written standard in Catalan (between the 16th and 19th centuries) so that a more specifically oral device (the marker was characteristic in story-telling) was able to grammaticalize before the standard was developed. Detges also shows how the same device – i.e. the use of an incipient marker to enhance the narrative event – could lead to lexicalization in the case of the inchoative suffix -sc- in the Romance languages. In the case of -sc- the device became eventually restricted to a small number of verbs, while in the case of Catalan anar ‘go’ + infinitive, it became applied to any kind of event.
. Standardization The suggestion that the development of a standard (written) language may play a role in grammaticalization cases, is an interesting one, which needs further investigation. Jim Miller’s contribution on the grammaticalization of the English perfect also looks at the influence of the standard language but he is especially interested in this aspect from a methodological point of view. He shows in his analysis of perfect and perfect-resultative constructions in spoken and non-standard English that the grammaticalization path described for the perfect that appears in written standard English is very much an idealization. The standard written perfect looks perfectly fixed, protected as it is by grammars of standard English and editorial practices. The situation is very different in spoken English and non-standard forms of English, where the perfect still competes with the past tense and with the resultative construction out of which it arose (i.e. I have my work done). Even more interesting are some developments of the perfect with adverbs, which in fact show a kind of anti-grammaticalization in that these constructions are again becoming more specific rather than more general (the latter would be usual on the grammaticalization cline). Miller’s article indeed calls into question the status of standard English as a yardstick against which to measure grammaticalization processes. Miller’s analysis is different from that of linguists like Tagliamonte (2000) and Winford (1994), who see the variety of perfect types present in non-standard varieties of English as a range of variations available before the perfect grammaticalizes, and indeed see these variants as stages towards the ‘idealized’ perfect of the standard. Miller, on the other hand, treats the perfect-variants in non-standard varieties not as part of
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.9 (9)
Introduction
a ‘perfect’-pathway or cline but as independent structures, which may each go their own way. In other words, it is not the ideal pathway that gives shape to the development of these variants, but the system of the non-standard or spoken languages themselves, in which these variants are used. Lea Laitinen in her contribution also emphasizes the importance of language external factors such as standardization and codification. Such factors are particularly relevant in Finnish historical linguistics, since Finnish has a relatively short written history and its standard form was not created until the late 19th century. Laitinen hypothesizes that in the process of standardization the more context-bound, indexical functions tend to be rejected. The changes discussed are the grammaticalization of modal necessity verbs (NEC constructions), the emergence of discourse particles from logophoric pronouns and the development of the non-agreeing negation word (NEG). In the case of NEC-constructions, the more grammaticalized variant is avoided in standard Finnish. The change from logophoric pronoun to discourse particle, on the other hand, had been overlooked by normative grammarians and hence the more grammaticalized variant has not been excluded from the written standard. The negation marker is discussed in most detail. Laitinen shows that the development of a non-agreeing negation particle was interrupted in the written language (but not in certain contemporary dialects) because the less grammaticalized, agreeing negative auxiliary was felt to be more genuinely Finnish. Ilona Herlin and Lari Kotilainen stress, as do Miller and Laitinen, the significance of language-external factors. Their main aim is to show that seemingly straightforward examples of grammaticalization may be far more complex than grammaticalization textbooks with their emphasis on universal grammaticalization paths would lead us to believe. One of these paths is the development from temporal to causal meaning, of which the Finnish subordinating conjunction kun ‘as, when, while’ is said to be a typical example (discussed e.g. in Traugott & König 1991; Heine & Kuteva 2002). Herlin and Kotilainen outline the historical development of this conjunction and the converb essa. They argue that changes in grammatical meaning are not necessarily the result of universal cognitive processes and pragmatic principles. In the course of standardization, kun was assigned temporal and causal functions while its one-time comparative and similaritive functions were transferred to kuin, originally a variant form of kun. The authors believe that this artificial distinction was made because Indo-European languages usually have different conjunctions for these functions as well. Spoken Finnish still has kun in all four functions. In addition, only written Finnish may use kun in contrastive constructions, most probably under the influence of Indo-European cognates (English while, Swedish medan) as well. A further point they note is that the converb essa developed causal, conditional and contrastive uses in written standard Finnish only. The history of kun and essa thus shows that synchronic polysemy does not
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.10 (10)
Olga Fischer, Muriel Norde, and Harry Perridon
necessarily reflect language-internal diachronic developments, but may be due to language-external factors such as standardization and language contact.4
. Reanalysis, polysemy and homonymy An important question in grammaticalization studies has always been whether grammaticalization leads to polysemy (one and the same item has different functions) or to homophony (different items share the same form). In the latter case a lexical split must be assumed to have taken place that separated the evolving grammatical item from its lexical mother (i.e. what Hopper 1991 terms ‘divergence’). In most cases the grammaticalizing item will then belong to another word class than the lexical item it derives from. In English, for example, the indefinite article and the numeral one go back to one and the same word, viz. OE a¯ n. In unaccented position the vowel was weakened, the nasal was lost before a consonant: a¯ n > a(n), and the meaning was ‘bleached’ to that of an indefinite article. The original item a¯ n split in this way into what is commonly assumed to be two different items, viz. a numeral (which it always had been) and an indefinite article (a new item in the lexicon). Others would, however, argue that the article and the numeral are still one and the same word, with different forms in different (stress) positions. In that case the meanings of the polysemous item would be distributed in a complementary way: article when unstressed, numeral when stressed. In her paper on demonstratives in copula-like constructions in the NorthAmerican language Passamaquoddy, Eve Ng assumes that a lexical split has taken place. The language possesses a large number of demonstratives, which are marked for three distances, two numbers, animacy, obviation (only animate forms), and absentativity. In verbless constructions with nominal predicates, distal demonstrative forms seem to function as copulas, which raises the question whether such forms should be analysed as copulas or as occurrences of the distal demonstrative. A careful analysis of the agreement phenomena in the various verbless constructions in which the demonstrative form seems to function as a copula, shows according to Ng that it is almost impossible to argue that the forms used in the verbless constructions belong to the same word class as the demonstratives they derive from. In her paper on the uses and meanings of the Kambera word wàngu Marian Klamer defends a more radical solution to the problem of homonymy or polysemy. Her view on the matter is that ‘bleaching’ not only affects the semantics of a given lexical item, but also its word class membership. The outcome of a grammaticalization process may hence be an underspecified, category neutral lexical item. The different interpretations of the word, then, are induced by the context in which it
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.11 (11)
Introduction
occurs. In this way the unity between form and meaning is restored. One of the main arguments Klamer adduces for her theory is that in a large number of cases it is not evident (or rather: it does not matter) in exactly what function wàngu is used. In the less ambiguous cases it is the context alone that determines how the word is to be interpreted (as a main verb, meaning ‘to use’, as a preposition (‘with’), as part of a compound verb, or as a matrix verb). Whether or not the term ‘multifunctional item’ as used by Richard Wong (and also by Umberto Ansaldo and Lisa Lim, see below) is synonymous with what Klamer calls an underspecified item is not entirely clear, but at least Wong’s paper on the acquisition of the various uses of bei2 by Cantonese children seems to suggest that this indeed might be the case. Wong addresses in this paper the question whether the various stages in the diachronic development of this morpheme are recapitulated in language acquisition. The lexical meaning of the word bei2 is ‘to give’, but it is also used as a dative marker (‘to’, recipient), as a matrix verb in permissive constructions (‘to let/allow’) and in passive constructions. The first three uses of bei2 share according to Wong the same core meaning ‘transfer’, whereas the passive is related to the permissive by way of a shared syntax. The lexical use of bei2 is also basic in the sense that it does not require (or even permit) another verb to be present in the construction, whereas the three derived uses all require a second verb. In order to investigate the hypothesis that there are at least three different stages in the acquisition of bei2 (main verb (‘transfer’) > dative, permissive > passive) Wong looked at occurrences of bei2 in (a modified version of) CANCORP, consisting of the transcripts of recordings from a longitudinal study of eight Cantonese-speaking children, aged between 22 and 42 months. The findings corroborate the hypothesis that syntactically simpler constructions (bei2 as a main verb) are acquired first, followed by those constructions that are semantically closest to the original meaning. Passives are rare in Cantonese child language, and appear relatively late. It seems thus safe to conclude that the order of acquisition of the various uses of a lexical item that has been grammaticalized mirrors the diachronic development of the item in question, at least in those cases where the lexical item can still be used in its original meaning.
. Form Another question that is still in need of a definite answer is what constitutes a change in form. How, for instance, is phonetic reduction, which is often one of the main indicators that a grammaticalization process has occurred, to be asserted in isolating tonal languages, which do not allow the kind of reduced syllables that are found elsewhere? Umberto Ansaldo and Lisa Lim claim that phonetic reduction
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.12 (12)
Olga Fischer, Muriel Norde, and Harry Perridon
in fact does take place in these languages, but is to be found in the suprasegmental features. In order to prove this hypothesis they investigated the intonational behaviour of some multifunctional items in two Sinitic languages, Cantonese (C) and Hokkien (H), in both non-grammaticalized and grammaticalized contexts, viz.: (C) gwo33 ‘to pass/cross’ > comparative marker ‘-er than’; (C) dou33 ‘to arrive’ > resultative marker; (H) khi21 ‘to go’ > aspectual (perfective) marker; and (H) ho33>21 ‘to give’, which like the words for ‘to give’ in many other Sinitic languages can be used as a causative marker, as a permissive marker and as a passive. When used in their original lexical sense, the morphemes had in all cases a significantly longer duration than when used as function words. There were, moreover, clear differences in vowel quality between the lexical and (more or less) grammatical uses of the items in question. Finally, the grammaticalized items tended to have lower fundamental frequency and intensity. Ansaldo and Lim conclude from these results that semantic bleaching is accompanied by phonetic reduction in isolating tonal languages. The phonetic reductions are probably due to the fact that when grammaticalized, the morpheme tends to be in a prosodically weak position, most often between two content morphemes. Another question pertaining to the reduction of form is posed by Sergey Say who wonders whether the fixing of word order constitutes a case of grammaticalization. He makes a distinction between word order shift (such as SOV to SVO) and word order fixation in that the former cannot be considered a case of grammaticalization since there is no reduction in variety (as is typical of grammaticalization processes) nor is there an increase in the abstractness of the rule. The case he looks at is the fixing of the genitive-noun word order within the NP in modern Lithuanian. He investigates the history of this construction and finds evidence that there used to be more variety in the position of the genitive in older Lithuanian texts. However, some of the evidence is obscured by the fact that many Old Lithuanian texts are rather literal translations from Polish. Say also suggests that there does not seem to be a semantic difference between pre- and postposed genitives in these texts. In other words, if this case represents an instance of grammaticalization, it did not lead to a loss of semantic meaning (this in contrast to Fischer (2000), who, when investigating a similar case of grammaticalization involving fixing of word order, found that the fixing of adjective-noun order within the NP in the history of English involved a loss of meaning and hence presented some sort of semantic bleaching). Say explains the direction of the change from pre- and postposition to preposition only (with one, clearly defined exception) to the fact that the Lithuanian genitive covers a wide range of functions (wider than in most languages) and that the genitive may have become preposed since one of the important functions of the genitive in Lithuanian is adjectival, and adjectives were generally preposed in Lithuanian. A harmonic word order tendency may also have been at issue but Say
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.13 (13)
Introduction
stresses that it is the synchronic factors rather than general diachronic processes that play an important role here.
. Outlook The seventeen papers in this book show that there is not yet a body of knowledge that might be a called a ‘theory’ of grammaticalization. The initial optimism created by the renewed interest in grammaticalization in the eighties and nineties of the last century has somewhat vanished, and it has become clear that the laws that one liked to think were discovered in processes of language change are not laws, but tendencies at best. Not all changes go in the same direction, there are exceptions to the assumed unidirectionality of change in grammar, which remain exceptions even if one rejects the majority of the suggested counterexamples as mere ‘lexifications’. Not all changes have to originate in cognition, formal changes do not need to follow changes in conception, in some cases it might even be the other way round, as perhaps is the case with the indefinite article in English.5 Borrowing, standardization, competition between various forms and constructions of different dialects and/or styles, all these factors may play a role in the history of grammatical forms and constructions in a given language. When confronted with a ‘grammatical’ and a ‘lexical’ use of a given form or construction, one cannot be absolutely certain that the grammatical meaning derives from the lexical meaning. It is still a probability, and as such any theory of grammatical change would have to account for this. Similarly, the circumstance that the mechanisms of grammaticalization (e.g. phonetic reduction, semantic bleaching, category change etc.) so often occur together is something that a theory of grammatical change needs to explain.6 We believe that certainty about or a better understanding of the nature of grammaticalization can only be obtained by a careful analysis in each particular instance of the history of the forms or constructions in question. Grammaticalization studies thus find themselves in the same position as other branches of historical linguistics: tendencies may be discovered but each word and each construction also has its own history.
Notes . It seems to us that, if unidirectionality is redefined in this purely semantic way, the relexification of the Mandarin modal dé/dei, or indeed any semantic change, does not then constitute a threat to the unidirectionality hypothesis.
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.14 (14)
Olga Fischer, Muriel Norde, and Harry Perridon . Indeed the grammaticalization Lexicon provided by Heine and Kuteva (2002) is based on this idea, i.e. that there are regular paths between source concepts (the initial lexical stage) and targets (grammaticalized structures), and that the range of both is restricted. . Note, however, that Kuteva (2001) emphasizes both the general cognitive forces as well as the discourse-pragmatic ones. . Note that this has serious implications for diachronic claims based on polysemies in languages without a written history, such as most (all?) African languages, on which the work of Heine and others crucially rests. . It has also been suggested in Fischer (1994) that in the grammaticalization of the English have + to infinitive, the formal changes forced the semantic changes and not the other way around. . Plank’s (1995) study on the properties of grammaticalization and degrammaticalization is very interesting in this respect because he shows that grammaticalization involves “ein schrittweise[s] Ablauf ” (p. 200, ‘a stepwise decline’), i.e. a gradual, orderly loss of thematically linked properties along a cline of properties which are part of the construction in question. In cases of degrammaticalization, this cline of properties has been disturbed by other (synchronic) factors (‘Systemstörung’), with the result that it is not the properties themselves that get re-ordered (they cannot be because the cline itself has been disturbed) but it is the formal expression that gets re-ordered or changed in such a way that it fits the left-over, disturbed properties. What his study clearly implies is that there are implicational relations between the various mechanisms involved in grammaticalization but not in degrammaticalization. Building on this important distinction between grammaticalization and degrammaticalization, Norde (2003) concludes that, unlike grammaticalization, degrammaticalization is not an autonomous process but the result of other changes, usually of the kind that Plank had identified as ‘Systemstörung’. Crucially, degrammaticalization always involves a single shift from right to left on the cline of grammaticality, in other words, there are no known cases in which a grammatical item gradually moves all the way up the cline, passing though the same intermediate stages a grammaticalizing item passes through, in the reverse order.
References Allen, Cynthia (1997). “The Origins of the ‘Group Genitive’ in English”. Transactions of the Philological Society, 95, 111–131. Beths, Frank (1999). “The History of dare and the Status of Unidirectionality”. Linguistics, 37, 1069–1110. Bybee, Joan L., William Pagliuca, & Revere D. Perkins (1994). The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: University of Chicago Press. Campbell, Lyle (Ed.). (2001). Grammaticalization. A Critical Assessment (Special issue of Language Sciences, 23, 2–3).
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.15 (15)
Introduction
Fischer, Olga (1994). “The Development of Quasi-Auxiliaries in English and Changes in Word Order”. Neophilologus, 78, 137–164. Fischer, Olga (2000). “The Position of the Adjective in Old English”. In Ricardo BermúdezOtero, David Denison, Richard M. Hogg, & C. B. McCully (Eds.), Generative Theory and Corpus Studies. A Dialogue from 10 ICEHL (pp. 153–181). Berlin: Mouton de Gruyter. Fischer, Olga, Anette Rosenbach, & Dieter Stein (Eds.). (2000). Pathways of Change. Grammaticalization in English. Amsterdam: John Benjamins. Giacalone Ramat, Anna & Paul J. Hopper (Eds.). (1998). The Limits of Grammaticalization. Amsterdam: John Benjamins. Haspelmath, Martin (1999). “Why is Grammaticalization Irreversible?” Linguistics, 37, 1043–1068. Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer (1991). Grammaticalization. A Conceptual Framework. Chicago: University of Chicago press. Heine, Bernd & Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hopper, Paul J. (1991). “On Some Principles of Grammaticization”. In Traugott & Heine (Eds.), Vol. I (pp. 17–35). Hopper, Paul J. & Elizabeth Closs Traugott (2003[1993]). Grammaticalization. Cambridge: Cambridge University Press. Kuteva, Tania (2001). Auxiliation: An Enquiry into the Nature of Grammaticalization. Oxford: Oxford University Press. Lass, Roger (2000). “Remarks on (Uni)Directionality”. In Fischer et al. (Eds.), 207–227. Lehmann, Christian (1995[1982]). Thoughts on Grammaticalization. München & Newcastle: Lincom Europa. Lyons, John (1977). Semantics. Volume 2. Cambridge: Cambridge University Press. Newmeyer, Frederick J. (1998). Language Form and Language Function. Cambridge, MA: MIT Press. Norde, Muriel (2001). “Deflexion as a Counterdirectional Factor in Grammatical Change”. Language Sciences, 23, 231–64. Norde, Muriel (2003). “Degrammaticalization – Process or Result?” Paper presented at the XVIth International Conference on Historical Linguistics, University of Copenhagen, August 11–15 2003. Plank, Frans (1995). “Entgrammatisierung – Spiegelbild der Grammatisierung?” In Norbert Boretzky, Wolfgang Dressler, J. Orešnik, K. Teržan, & Wolfgang U. Wurzel (Eds.), Natürlichkeitstheorie und Sprachwandel (Essener Beiträge zur Sprachwandelforschung.) (pp. 199–219). Bochum: Brochmeyer. Sweetser, Eve E. (1990). From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Tagliamonte, Sali (2000). “The Grammaticalization of the Present Perfect in English. Tracks of Change and Continuity in a Linguistic Enclave”. In Fischer et al. (Eds.), 329–354. Traugott, Elizabeth C. & Richard B. Dasher (2002). Regularity in Semantic Change. Cambridge: Cambridge University Press. Traugott, Elizabeth C. & Bernd Heine (Eds.). (1991). Approaches to Grammaticalization. Vols. I, II. Amsterdam: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 15:31
F: TSL5901.tex / p.16 (16)
Olga Fischer, Muriel Norde, and Harry Perridon
Traugott, Elizabeth C. & König, Ekkehard (1991). “The Semantics-pragmatics of Grammaticalization Revisited”. In Traugott & Heine (Eds.), Vol. I (pp. 189–218). van der Auwera, Johan & Vladimir A. Plungian (1998). “Modality’s Semantic Map”. Linguistic Typology, 2, 79–124. Winford, Donald (1994). “Variability in the Use of Perfect have in Trinidadian English: A Problem of Categorial and Semantic Mismatch”. Language Variation and Change, 5, 141–188. Wischer, Ilse & Gabriela Diewald (Eds.). (2002). New Reflections on Grammaticalization. Amsterdam: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.1 (17)
On directionality in language change with particular reference to grammaticalization* Martin Haspelmath Max-Planck-Institut für evolutionäre Anthropologie, Leipzig
.
Introduction
The issue of the directionality of grammaticalization has attracted considerable attention in recent years, as illustrated by works such as as Plank (1995), Frajzyngier (1997), Newmeyer (1998: Ch. 5), Haspelmath (1999a) (and the ensuing debate: Geurts 2000a, b; Haspelmath 2000b), the papers in the recent special issue of Language Sciences (Campbell (Ed.) 2001), Traugott (2001), Kim (2001), van der Auwera (2002), Heine (2003), and quite a few others. In this paper, I would like to put this issue in a somewhat broader perspective, discussing not only the unidirectionality of grammaticalization, but also unidirectionality in other areas of language change. But the main focus will be on grammaticalization, and after defending the unidirectionality claim for grammaticalization against several criticisms, I will examine a substantial number of alleged exceptions to the unidirectionality and show that only very few of them can be accepted as real exceptions. The paper is divided into three main sections, in which I will make the following larger points: i.
If we want to understand language change, we need to identify universals of language change. Directionality constraints are among the strongest universals of language change (Section 2). ii The unidirectionality of grammaticalization is the most important constraint on morphosyntactic change (Section 3). iii. Most cases of “degrammaticalization” that are cited in the literature do not show the reversal of grammaticalization (or ‘antigrammaticalization’), but something else (Section 4).
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.2 (18)
Martin Haspelmath
. Universals of change and directionality constraints Most of the time, historical linguists are occupied with the business of describing language change, which is quite a challenging task in itself, given that change is so difficult to observe. But ultimately we would also like to understand language change to the extent possible, or in other words, we want to answer why-questions. In particular, we want to know why language structure changes in the way it does. There are other why-questions about language change, such as the question why languages change at all, the question why the social propagation of an initial innovation can often be described by an S-curve, and so on. These will not be addressed here. Linguists working on particular languages are also often interested in particular why-questions such as the question why the Romance languages lost the Latin case inflections. But unfortunately, particular why-questions of this kind are for most practical purposes unanswerable. The number of factors affecting language change is so enormous and we can control only so few of them that most change events must appear to us as historical accidents. Latin could have kept its cases, even with all the phonological erosion that made them difficult to distinguish, simply by applying morphological changes serving to preserve the case contrasts. Or Latin speakers could even have developed more cases the way Hungarian and Finnish speakers did. It so happened that Latin lost its cases, and trying to understand this unique historical event typically leads to frustration. In general, understanding requires that we identify non-accidental phenomena, and for understanding language change, this means that we have to find universals of language change. To illustrate what I mean by this, a few random examples of proposed universals of language change (of different degrees of generality) are given in (1). (1) a.
Survival of the Frequent (‘Unmarked’) (e.g. Winter 1971; Wurzel 1994) When a grammatical distinction is given up, it is the more frequent category that survives. (E.g. plural forms survive when dual/plural distinction is lost.) b. Sound Alternations Result from Sound Change (phonetics > phonology; *morphology > phonology) c. From Space to Time (e.g. Haspelmath 1997b) (spatial > temporal marker; *temporal > spatial marker) d. From ‘Something’ to ‘Nothing’ (Haspelmath 1997a: 230) indefinite pronouns ‘something’ > ‘nothing’ (*‘nothing’ > ‘something’) e. From Esses to Aitches: s > h (*h > s) (Ferguson 1990)
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.3 (19)
On directionality in language change
These are all general laws which we can potentially explain,1 and when we have such an explanation, we can apply it to individual instances of these universals. For example, we might want to say that the universal ‘Survival of the Frequent’ is explained with reference to the cognitive notion of frequency-induced entrenchment (cf. Bybee 1985: 119): A frequent linguistic unit is remembered better because frequency of exposure leads to greater memory strength. When a distinction is given up, only the most entrenched category survives. Now let us take an individual instance of the Survival of the Frequent, say, the fact that when the Classical Greek dual/plural distinction was given up, only the plural forms survived. The plural was more frequent than the dual (cf. Greenberg 1966: 31–37), so this change is in line with the universal, and if we want to know why the plural rather than the dual survived in Greek, we can appeal to the explanation that we just gave. So in this sense we can say that a particular change was explained after all; but of course the explanation of the particular change has nothing particular about it. We cannot explain why this changed happened in Greek but not, say, in Slovene (where the old dual survived), and we cannot explain why it happened two and a half millennia ago rather than a thousand years later or a thousand years earlier. So wherever we can understand structural change, it is really universals of structural change that we understand. But unless we know whether a given instance of change is part of a larger trend, we do not know whether there is anything to explain. Now when we look at reasonably robust universals of language change, we see that many of them take the form of directionality constraints. Of the five examples in (1) four have the form “X can change into Y, but Y cannot change into X”. Especially in phonology, it is easy to find cases of this type, and I list a few more in (2). (2) a. b. c. d. e. f.
[k] > [tw] (*[tw] > [k]) [p] > [f] (*[f] > [p]) [u] > [y] (*[y] > [u]) [z] > [r] (*[r] > [z]) [ts] > [s] (*[s] > [ts]) [l] > [w] (*[w] > [l])
So quite a few sound changes appear to be unidirectional, but there are of course also bidirectional sound changes, such as those in (3). Some of these changes are more likely in some positions than in others, and maybe a more fine-grained description of the type of change would reveal a directionality tendency in some of these cases as well. (3) a. [t] > [θ] and [θ] > [t] b. [o] > [a] and [a] > [o] c. [i] > [6] and [6] > [i]
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.4 (20)
Martin Haspelmath
d. [au] > [o] and [o] > [au] e. [b] > [v] and [v] > [b] Thus, it is an empirical question whether a type of sound change is unidirectional or not. I am not aware of any extensive discussion of this issue in the theoretical literature on phonological change, but as Ferguson (1990) observes, every linguist with some experience in diachronic phonology has the intuition that there are often directionality constraints at work: One of the most powerful tools in the armamentarium of linguists engaged in the study of diachronic phonology is the often implicit notion that some changes are phonetically more likely than others. Thus if a linguist finds a systematic correspondence between [g] and [dŠ] in two related language varieties, it will be reasonable to assume that the stop is the older variant and the affricate the younger one until strong counter evidence is found. The linguist makes such an assumption because experience with many languages has shown that the change of [g] to [dŠ] is fairly common and tends to occur under certain well-documented conditions whereas the reverse change is unusual and problematic. (Ferguson 1990: 59–60)
Ferguson goes on to observe that this powerful tool of directionality constraints is not generally covered in textbooks or handbooks of phonology or historical linguistics. These typically include taxonomies of attested sound changes and introduce technical terms like lenition, assimilation, syncope and epenthesis, but they usually do not say what an impossible change is, or which changes are ubiquitous and which ones are exceedingly rare. For synchronic universals in phoneme systems, we have Maddieson’s (1984) handbook with inventories of 317 languages. Diachronic phonology, whether theoretically oriented or primarily interested in reconstructing particular protolanguages, would profit enormously from having a handbook of attested sound changes in the world’s languages. Such a handbook would make it possible to identify constraints on possible sound changes, and many of the most interesting constraints will no doubt be directionality constraints. After all, that [u] presumably never changes to [a] in one step, or that [l] never changes to [b], is not surprising, whereas the unidirectionality of the [u] > [y] change and the [l] > [w] change is much harder to explain. There are also some clear tendencies of lexical semantic change (e.g. ‘cup’ can change to ‘head’ and ‘head’ can change to ‘chief ’, but the opposite changes are extremely unlikely). Once we are confident that we have a universal directionality constraint in some domain, the question arises as to how it should be explained. If the source structure and the target structure are similar enough so that one can change into the other gradually and often imperceptibly, why can’t they change in either direction? The historical-linguistic literature is full of proposals accounting for specific
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.5 (21)
On directionality in language change
cases, appealing to a variety of structural and external factors. For phonology, some authors have proposed that sound change is by and large reductive in nature (Mowrey & Pagliuca 1995; Bybee 2001), and that the unidirectionality of changes like [p] > [f] and [ts] > [s] can be accounted for in this way. The explanation for unidirectionality is also beginning to be addressed by researchers working in the area of grammaticalization (e.g. Lehmann 1993; Haspelmath 1999a), and this discussion could profit from analogous discussions in the other subfields of linguistics. In this paper, I will not say anything about the correct explanation of unidirectionality in grammaticalization, because at present I have nothing to add to my earlier proposals. I will, however, address a number of criticisms and counterexamples that can be found in the recent literature.
. Unidirectionality of grammaticalization . How important is unidirectionality? Although it is very difficult to quantify language change, it seems to me that it is undeniable that the unidirectionality of grammaticalization is by far the most important constraint on morphosyntactic change, simply because grammaticalization changes are so ubiquitous. As far as I can see, the only serious competitor of unidirectionality is the diachronic universal ‘Survival of the Frequent’ (see (1a)). This universal seems to hold not only when categorial distinctions break down, but also in analogical leveling in inflectional morphology. For example, when a stem alternation such as dream/dreamt is leveled, it is the more frequent present-tense stem that survives (so that we get dream/dreamed, not *drem/dremmed). This is a fairly important universal for morphological change, but it seems to be much less important for syntactic change. Grammaticalization, by contrast, is of paramount importance both for syntactic change and for morphological change. A rough estimate is that two thirds of the papers on diachronic syntax published in recent volumes such as van Kemenade and Vincent (1997) and Pintzuk et al. (2000) deal with grammaticalization changes (even if they rarely mention the term ‘grammaticalization’). The relatively high number of non-grammaticalization papers in these volumes has to do with the fact that word order change is so salient in some European languages, especially of course word-order change having to do with verb-second phenomena. But we know that verb-second word order and the changes related to it are highly unusual phenomena that are hardly found outside of Europe. My guess is that if we were able to study syntactic change on all continents, grammaticalization would play an even greater role in diachronic syntax. Of course, this is not more than an impres-
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.6 (22)
Martin Haspelmath
sionistic statement, but I challenge anyone to come up with a long list of interesting syntactic changes that are unrelated to grammaticalization. Like unidirectionality in sound change, unidirectionality in grammaticalization is very important in practical terms for the historical-comparative linguist. Suppose we have two related languages with no historical documentation, and one of them has a future-tense affix that looks similar to a future-tense auxiliary of the other language. If both directions of change were equally likely, we would not know what to reconstruct for the ancestor language. But because grammaticalization is overwhelmingly irreversible, the historical linguist can safely reconstruct the future auxiliary for the protolanguage in this case. Moreover, unidirectionality helps us assess the likelihood of competing etymologies even if older stages are attested. For instance, historical linguists of Indo-Aryan have long debated the etymology of the Hindi/Urdu ergative-case clitic =ne. Quite a few linguists in the 20th century traced this element back to Sanskrit -ina, an instrumental suffix that would be a very plausible source from a semantic point of view. In a recent contribution to this debate, Butt (2001: 114) has pointed out that such a change would constitute a counterexample to unidirectionality and is hence very unlikely (one would have to postulate phonological expansion from [na] to [ne:] and a change from affix to clitic). This, among other reasons, leads Butt to reject this etymology and look for some other possible source of =ne in a full lexical item.2 Now despite the theoretical importance of grammaticalization studies for understanding language change and their practical importance for historical linguistics, there have been a number of critical voices in recent years. In the remainder of this section I would like to address some of these points of criticism and show that while some are well-taken, others are quite unfounded.
. Partially valid criticism Three points made by grammaticalization critics that I regard as partially justified are listed in (4). (4) a. Unidirectionality is not exceptionless b. ‘Grammaticalization theory’ is not one theory c. ‘Pathways of morphemes’ must be linked to speakers’ actions Unidirectionality was apparently first stated explicitly as an important universal property of grammaticalization in Lehmann (1995a [1982]: 16–19). Lehmann coined the term degrammaticalization for a phenomenon that he believed did not exist, the reverse of grammaticalization. But now the phenomenon had a name, and it seems that Lehmann’s strong initial claim plus his nice neologism spurred
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.7 (23)
On directionality in language change
linguists to look for actual examples of degrammaticalization. And indeed, a number of good exceptions were found (see Newmeyer 1998; Campbell 2001; Janda 2001; Norde 2001), and this has generally been acknowledged by grammaticalization researchers. The counterexamples did not pose a serious threat to the original generalization, but a presumed absolute universal had to be weakened to a statistical universal. What are the consequences of this for the theory of grammaticalization? One might say that now that we know that unidirectionality has exceptions, it has become somewhat less interesting. This is true insofar as stronger generalizations are more surprising than weaker generalizations, but the reverse is also true: Because unidirectionality is so interesting, we know about the exceptions. If someone proposes an uninteresting universal, we may never discover the exceptions because nobody bothers to look for them. The basic generalization of unidirectionality stands unchallenged as long as nobody shows that degrammaticalization is as common as grammaticalization. If one is interested in generalizations rather than arbitrary facts, one must put aside the exceptions, because unless they can be subsumed under some further generalization, they cannot be explained. Harris and Campbell (1995: 338) say in this context: “An adequate theory must account for infrequent phenomena, not merely for the most common patterns.” This is of course right if by ‘theory’ they mean ‘descriptive framework’: We need terminology also for rare phenomena. But if by ‘theory’ we mean understanding and explanation, this is not right, because exceptions cannot be understood by definition; they are the residue that resists explanation.3 Thus, although it is true that unidirectionality is not exceptionless, this does not make it any less intriguing and important for our understanding of language change. The second point of criticism that I find partially justified is Newmeyer’s (1998) claim that what linguists commonly call ‘grammaticalization theory’ is not a theory in the sense of a well-defined system of interconnected falsifiable hypotheses. What unites researchers in the area of grammaticalization is not that they subscribe to a single monolithic theory, but that they see a large class of semantic and morphosyntactic changes as sharing similarities and as theoretically interesting. There are a fair amount of quite different theoretical ideas and hypotheses concerning grammaticalization changes, and some of them are probably not compatible. Thus, ‘grammaticalization theory’ is more like ‘evolutionary theory’, which is not one single monolithic system either, but describes a range of related approaches and basic issues in the area of historical biology. It would perhaps be more accurate to say ‘theorizing about grammaticalization’ (instead of ‘grammaticalization theory’), and to some extent the use of the term ‘theory’ may be motivated by its prestige. The prestigious term ‘theory’ has experienced a rather inflationary development in recent decades in linguistics. Like grammaticalization, inflationary processes are generally irreversible (cf. Dahl 2002), so it seems unlikely that
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.8 (24)
Martin Haspelmath
the term ‘grammaticalization theory’ will be abandoned. But if we want to be clear about what we are doing, the term ‘theory’ is not particularly useful. I prefer to talk about the goal of understanding, or explaining, or answering why-questions. These are terms from our everyday language that everyone understands, and our endeavors can be accurately characterized with them. The third point of criticism is Janda’s (2001) reminder that it is impossible to understand language change phenomena if we see them as divorced from the speakers.4 If we talk about a morpheme traveling along a pathway, we should be aware that this is a very abstract metaphor that may invite all kinds of unwarranted inferences. We need to be careful with metaphors, and we should make more efforts to go down to the micro-level of individual speakers and derive the observed constraints on structural changes from known constraints on speakers’ linguistic behavior. But at the same time it is clear that we cannot do linguistics without abstract metaphors, and so far at least concepts like ‘grammaticalization path’ have done far more good than damage. We would know far less about possible and impossible changes if we had not started drawing diagrams of grammaticalization paths and semantic maps. Recently, some linguists have stated precise rules for interpreting a diagram showing a semantic map with grammaticalization paths (see, e.g., van der Auwera & Plungian 1998; Croft 2001; Haspelmath 2003). Consider, for example, the semantic map in Figure 1. This figure embodies the following two claims: i.
Synchronic: Polysemous forms cover adjacent nodes (i.e. nodes linked by a line or arrow); ii. Diachronic: A linguistic form may extend its range of functions on the map in any direction, but not against the direction of an arrow. Thus, it is predicted that a direction marker such as Latin ad ‘to’ can become a recipient marker (as in French à), but it cannot then go on to become a beneficiary marker. Thus, we can see the path metaphor and the diagrams based on it as analogous to tree diagrams in syntax: These are probably not literally in people’s mental grampredicative
external
possessor
possessor
direction
recipient
beneficiary
purpose
experiencer
judicantis
Figure 1. A semantic map of typical dative functions (from Haspelmath 2003: 234)
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.9 (25)
On directionality in language change
mars either, but if syntacticians drew no tree diagrams, we would know much less about syntax.
. Invalid criticism Three points made by grammaticalization critics that I regard as unjustified are listed in (5). (5) a.
Unidirectionality implies a fully isolating prehistoric state and thus contradicts uniformitarianism. b. Unidirectionality is built into the definition of grammaticalization and is hence not an empirical claim. c. There is nothing unique about the kinds of changes that are associated with grammaticalization. (Newmeyer 1998; Campbell 2001; Janda 2001)
The first point, about the contradiction to uniformitaranism, was recently brought up by Roger Lass (see also Hoenigswald 1991: 25): The claim that all grammatical material is ultimately lexical means that there was a time when all human languages were ‘isolating’ (in the days of Homo erectus or whatever everybody spoke Vietnamese). . . [This] is counteruniformitarian, and so methodologically inadmissible. (Lass 2000: 216)
But first of all, the claim that “all grammatical material is ultimately lexical” does not follow from unidirectionality, because it may well be that some elements such as demonstratives or interrogative pronouns are never created by grammaticalization from full lexical items, and have simply always been demonstratives or interrogative pronouns. Moreover, at least since Meillet (1912) it has generally been recognized that analogy is another important source of grammatical items, besides grammaticalization. But even if one were to make the speculative claim that all grammatical material is ultimately lexical, there would be no methodological problem because the principle of uniformitarianism does not require the assumption that early hominids such as Homo erectus (if they already had some kind of language) had languages of the same type as modern humans. If we allow ourselves speculation about the distant past, we can easily imagine that the first modern humans inherited part of their lexicon from the cruder languages of earlier hominids and added more lexical differentiation and grammatical elaboration. But since language has been around for tens of thousands of years and we know next to nothing about its origin, we really do not have to worry about the consequences of diachronic universals for prehistory.5 The second point of criticism is that unidirectionality is built into the definition of grammaticalization and hence represents a tautology (Campbell 2001: 124;
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.10 (26)
Martin Haspelmath
Janda 2001: 294). This is very easy to counter: Yes, it is true that unidirectionality is built into the definition of grammaticalization. My current definition of grammaticalization is given in (6). (6) A grammaticalization is a diachronic change by which the parts of a constructional schema come to have stronger internal dependencies.6 This describes a unidirectional process, so saying that “grammaticalization is unidirectional” is strictly speaking tautologous. The point is, of course, that the easily imaginable reverse of this process does not occur (apart from a few exceptional instances). So this is not a substantive point at all, and one wonders why one hears it repeated so often. The third point of criticism is that there is nothing special or unique about grammaticalization changes. Campbell (2001) expresses it as follows: Grammaticalization has no independent status of its own; it merely involves other kinds of changes and mechanisms of change which are well understood and are not limited to cases involving grammaticalization: sound change, semantic change, and reanalysis. (Campbell 2001: 117)
And Janda (2001: 266) maintains that grammaticalization “is actually an epiphenomenon which results from the intersection and interaction of other, independently motivated domains” (see also Newmeyer 1998: 237ff.). Somehow these authors seem to think that grammaticalization is wrongly regarded as a primitive concept, although I know of no claim to this effect. On the contrary, studies of grammaticalization such as Lehmann (1995a [1982]), Heine and Reh (1984), Hopper and Traugott (1993) are quite explicit in listing the various low-level changes that are associated with grammaticalization, such as phonological erosion, desemanticization, reanalysis, decategorialization, and so on. The claim that these authors and other have made is that grammaticalization is a particularly interesting concept, because it is largely irreversible and because we observe strong correlations between phonological, syntactic and semanticpragmatic changes. It is a macro-level phenomenon which cannot be reduced to the properties of the corresponding micro-level phenomena. Campbell’s, Janda’s and Newmeyer’s criticism is similar to an objection against sociological studies of social classes on the grounds that social class is not a primitive concept, but an epiphenomenon which results from the interaction of human individuals. Most of the subject matters studied by linguists are epiphenomenal in the sense that they are complex higher-level phenomena involving the interaction of a multiplicity of lower-level phenomena.7 Campbell says in the above quotation that sound change, semantic change and reanalysis are “well understood”, but unless he refers to the terminology and really means “well defined”, I find this far too optimistic. Diachronic phonologists and di-
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.11 (27)
On directionality in language change
achronic semanticists have not even begun collecting the systematic cross-linguistic data that would allow us to arrive at empirically well-founded universals of sound change and universals of lexical semantic change. Whereas for grammaticalization we now have Heine and Kuteva’s (2002) World Lexicon of Grammaticalization, we are still waiting for a World Lexicon of Sound Change and a World Lexicon of LexicalSemantic Change. In diachronic syntax, all we have is the handbook by Harris and Campbell (1995) with an exhaustive classification of syntactic changes and ample cross-linguistic exemplification, but few constraints and thus little explanation. In all these areas we are far from really understanding language change. Newmeyer (1998: 259) urges linguists not to “invite the conclusion that some dynamic is at work in grammaticalization that cannot be understood as a product of [certain independently occurring semantic and phonetic changes].” But this is exactly what is claimed by grammaticalizationists. Even if there is no single universally accepted explanatory architecture for grammaticalization yet, we have made a lot of progress in understanding the dynamic of grammaticalization. At the very least we have thorough cross-linguistic documentation, and a strong generalization, unidirectionality.
. Antigrammaticalization and “degrammaticalization” My third main point is that most cases of “degrammaticalization” that are cited in the literature do not show the reversal of grammaticalization, but something else. I will discuss a fairly large number of changes that have been mentioned in the literature, and I will classify them into various types. There is no space here to describe the changes in any detail, so I must refer the reader to the earlier literature. The purpose of this section is twofold. On the one hand, I want to show that exceptions to the unidirectionality universal are not “rampant” (as Newmeyer 1998: 263 claims), but are quite rare. Although probably around a hundred cases of degrammaticalization have been mentioned in the literature, the number of real exceptions is much lower. On the other hand, since the phenomena called “degrammaticalization” are so heterogeneous, it seems useful to identify various subclasses of “degrammaticalization”. I do not think that these cases have anything in common, so that we do not really need a term like “degrammaticalization” for them, and I only use this term in quotes.
. Antigrammaticalization: The reversal of grammaticalization One important new term that I want to introduce here is antigrammaticalization. By this I mean a change that leads from the endpoint to the starting point of a
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.12 (28)
Martin Haspelmath
potential grammaticalization and also shows the same intermediate stages. For instance, a change from a case suffix to a free postposition with the intermediate stage of a postpositional clitic would be an antigrammaticalization. This implies that the change occurs in a construction which can be seen as preserving its identity before and after the change, as in grammaticalization, where we also have a gradual change of the properties of a construction, but we do not get a new construction. In this characterization of grammaticalization and antigrammaticalization I presuppose that grammaticalization changes modify the constructions they affect but preserve their identity. Admittedly this makes my characterization somewhat vague, because there is not always a consensus on which changes just modify a construction and which changes lead to totally new constructions. In Haspelmath (1998), I have argued at some length that grammaticalization changes are gradual, preserve the identity of the construction, and occur in ordinary language use, whereas reanalysis (and likewise analogy) is abrupt, leads to new constructions (or subsumes an existing unit under a completely different construction), and occurs in language acquisition. For example, a change whereby an erstwhile relational noun turns into a preposition (e.g. German wegen “because of ”, from the dative plural of Weg “way”) is said to preserve its identity because it consists of numerous imperceptible changes with no radical break at any single point. In particular, there is no reason to postulate an abrupt reanalysis of the noun Wegen as a preposition wegen (unless one assumes that speakers only have a small fixed set of innately specified categories at their disposal and cannot internalize a grammar with items that are intermediate between nouns and adpositions). It should be noted that my definition of antigrammaticalization is intended to cover types of changes, not tokens. Janda (2001: 295) and Norde (2001: 236) seem to interpret the term ‘reversal’ as ‘token reversal’, so that irreversibility would only mean that once a structure A has changed into a structure B, it does not change back to A. This claim, that token reversal does not occur (or is very unlikely), is of course not particularly interesting. My term antigrammaticalization is intended to cover any type of change that goes against the general direction of grammaticalization (i.e. discourse > syntax > morphology). Armed with this new term, we can now say that only antigrammaticalizations are exceptions to unidirectionality (cf. Lehmann 1995b: 1256), whereas other kinds of “degrammaticalization” are not necessarily expected to be rare or exceptional. This is not a weakening of the unidirectionality claim, because at least the way it was originally formulated (in Lehmann 1995a [1982]: 16–19), it is clear that only antigrammaticalizations were supposed to be ruled out, not any kind of change from grammar to lexicon. Janda (2001) has made a similar terminological distinction between reversibility of grammaticalization and counterability (however, I am not aware that anyone ever claimed that grammaticalization should be not only irreversible, but also ‘non-counterable’).
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.13 (29)
On directionality in language change
Let us now look at some antigrammaticalizations in this sense. Eight cases are listed in (7). (7) attested antigrammaticalizations a. English and Mainland Scandinavian genitive suffix -s > clitic =s (Janda 1980; Norde 1997; Newmeyer 1998: 266, 256; etc.; but see Börjars 2003) b. Irish 1st person plural subject suffix -muid > independent pronoun muid (Bybee et al. 1994: 13–4; Roma 1999) c. Japanese adverbial subordinator -ga “although” > free linker ga “but” (Matsumoto 1988) d. Saame abessive suffix *-ptaken > clitic =taga > free postposition taga (Nevis 1986a) e. Estonian question marker -s > =es > free particle es (Nevis 1986b; Campbell 1991: 290–292) f. English infinitive prefix to- > proclitic to= (Fischer 2000; Fitzmaurice 2000) g. Modern Greek prefix ksana- “again” > free adverb ksana “again” (Méndez Dosuna 1997) h. Latin rigid prefix re- “again” > Italian flexible prefix ri- (e.g. ridevo fare “I must do again”) For me, these cases are real exceptions, which means that they do not fall under any other generalization, and I cannot say more about them. This does not mean that more could not be said about them in the future. For instance, Bybee et al. (1994: 13–14) say about the case of the Irish personal pronoun muid that there was “strong paradigmatic pressure” that facilitated the change. It could be that we will eventually be able to identify further factors such as ‘paradigmatic pressure’ that make antigrammaticalization possible, but until we have a solid generalization, any attempt at explaining these cases away seems premature. All other cases of “degrammaticalization” that I have found in the literature are not antigrammaticalizations, as I will now show.
. Delocutive word-formation from function words and affixes A first type of change that has been called “degrammaticalization” is delocutive word-formation from function words and affixes. A delocutive lexeme is one that was derived by some regular word-formation process from another lexeme whose use in speech somehow determines the meaning of the derived lexeme. For instance, Latin negare “deny” is said to derive from the negative marker nec, so it literally means “say not”, and French tutoyer is derived from the pronoun form tu
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.14 (30)
Martin Haspelmath
and means “use tu as address form”. An example of a delocutive noun would be a hello in English (meaning “an act of saying hello”), as in I heard many hellos and few good-byes, or the noun yes in I never know whether her yes is really a no. Now my claim is that expressions such as ifs and buts are delocutive nouns of the same type, and iffy is a delocutive adjective. A few further cases are listed in (8). (The symbol “<:” should be read as “was formed on the basis of ”.) (8) some delocutive formations a. Latin negare “deny, say no” <: nec “and not; neither” (Benveniste 1966 [1958]: 279) b. French tutoyer “use tu as address form” <: tu “you.familiar” (Norde 2001: 235; Ramat 2001: 396) c. ifs and buts (van der Auwera 2002: 22), iffy (Newmeyer 1998: 274), must (from auxiliary to noun; van der Auwera & Plungian 1998: 117) d. Dutch Is het een hij of een zij? “Is it a he (male) or a she (female)?” (Norde 2001: 235) e. French le pour et le contre, German das Für und Wider (Hagège 2001: 1622) f. Chinese s¯an tóng “the three withs”, sì huà “the four -izations” (Hagège 1993: 210) g. ism “doctrine ending in -ism”, itis “disease ending in -itis” (Ramat 1992: 549) The original sense of ifs must have been “situations in which one uses the word if ”, and the relevant sense of Dutch hij is “person for which one uses the pronoun hij”. Delocutive word-formation presupposes some kind of reflection about linguistic expressions; it is a metalinguistic act that is probably very conscious. It is therefore not surprising that function words can be the basis of delocutive word-formation processes.8 It is probably not an accident that such changes have primarily been reported from languages used by highly literate societies. If speakers know their language also in a written form, even affixes can become sufficiently salient to serve as the basis for a delocutive conversion process, as illustrated in (8f, g). When a word such as ism is coined, this can be regarded as a change which has a grammatical item (the suffix -ism) as its input and a lexical item (the noun ism) as its output, and in this sense it is “degrammaticalization”, defined as a change “from grammar to lexicon” (van der Auwera 2002). But clearly this is not antigrammaticalization as defined above, because this is not the reverse of a potential grammaticalization change. There is no intermediate stage at which ism is a clitic, and there is no sense in which we would say that the two items occur in the same construction.
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.15 (31)
On directionality in language change
Especially when the affix that has been turned into a noun is a prefix and when this happens in English, a good alternative explanation is that we are dealing with a clipping (ad-hoc shortening), because English uses this device so freely. In cases like ex “ex-partner”, pro “in favor”, anti “against” (cf. Crowley 1997: 148; Lazzeroni 1998: 277), I find an explanation in terms of clipping more plausible than an explanation in terms of delocutive word-formation.
. Back-formation of bound compound members Next let us look at back-formation, which is a kind of reanalysis or analogical change (cf. Becker 1993). Cases like English burger, which was evidently backformed from compounds like ham-burger (Crowley 1997: 148), can be explained in this way. I would also regard numerals such as English seventeen, German siebzig “seventy” and Italian settanta as compounds, so that free forms like teens, German zig “dozens” and Italian anta “age from forty upwards” (Ramat 1992: 550; Norde 1997: 3) can be explained as back-formations.9 Again this could not be antigrammaticalization because we do not have the same construction after the change, and there are no intermediate clitic stages.
. Adverb-to-verb/noun conversion Another set of examples that can be described by ordinary word-formation is adverb-to-verb conversion as in English upv <: upadv , downv <: downadv , off v <: off adv , and so on (Hopper & Traugott 1993: 127; Newmeyer 1998: 273), or the analogous Spanish examples sobrarv “be extra” <: sobreadv “above”; dentrarv “insert” <: dentroadv “inside” (Harris & Campbell 1995: 432, n. 23). We also find an example of an abstract noun derived from an adverb (Finnish pääll-ysn “upper part” <: päälläadv “above”; Hagège 1993: 209). Sometimes it has been claimed in the literature that these verbs and adverbs were formed from prepositions and not from adverbs, so that they would be potential cases of “degrammaticalization”. But even if they were derived from prepositions, i.e. grammatical items, they would not constitute antigrammaticalizations. It should be uncontroversial that these verbs and nouns were created by regular word-formation, not by the gradual modification of a construction.
. Phonogenesis So far we have seen instances of word-formation which represent instantaneous changes, and for this reason alone they are evidently not antigrammaticalizations. But there are also several kinds of gradual “degrammaticalization” changes that are
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.16 (32)
Martin Haspelmath
not exceptions to unidirectionality. One is what Hopper (1994) calls phonogenesis, i.e. the loss of structure in a polymorphemic lexeme which thereby becomes monomorphemic. Hopper (1990: 155) mentions examples like modern German bleiben, whose initial consonant is a former prefix (older German be-lîben). Ramat (1992: 551) cites English drench (from the Proto-Germanic causative *drank-jan), and Hopper and Traugott (1993: 127) discuss English tomorrow, which is no longer analyzed as to + morrow (cf. also Traugott 1994: 1485). Van der Auwera (2002: 21) gives English twit (from Old English æt-w¯ıtan “at + blame”). In all these cases one can say that former grammatical constituents became purely phonological constituents, so we have phonogenesis. This change is also known as demotivation, and its reverse is called remotivation or folk etymology. This is “degrammaticalization” in the sense that grammatical elements lose their grammatical status, but it is of course not the reverse of grammaticalization.
. Loss of an inflectional category with traces In demotivation we are mostly talking about derivational morphology being lost, but entire inflectional categories may disappear in languages as well, and linguists have sometimes called this “degrammaticalization”. The older Indo-European dual no longer exists in Latin, and there are individual lexical items like ambo “both” which have an ending that goes back to the old dual ending but synchronically no longer has grammatical status (Ramat 1992: 551). Similarly, Wichmann (1996) calls the loss of the inflectional category of agentivity in Tlapanec “degrammaticalization”. These cases are quite similar to derivational demotivation. But it may also happen that a disappearing inflectional category leaves so many traces in surviving lexical items that the morphological pattern remains productive, but only as a derivational pattern. For instance, the Latin present participle has disappeared in many Romance languages as an inflectional category of the verb, but it survives as a derivational pattern in Spanish and Italian (-ante/-ente; cf. Luraghi 1998; Newmeyer 1998: 264). A fairly similar case is the Swedish propertybearer suffix -er cited by Norde (1997: 230) (e.g. dummer “stupid person” <: dum “stupid”), which goes back to the Old Norse inflectional nominative-case suffix. So this is a change from inflection to derivation, and Kuryłowicz (1965: 69) had said that changes from inflection to derivation show the reversal of grammaticalization.10 However, such changes, which are clearly attested, should not be lumped together with grammaticalization. Inflectional patterns do not show stronger internal dependencies than derivational patterns. While changes from discourse to syntax and from syntax to inflection do form a natural class, changes between inflection and derivation should not be put in this class.
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.17 (33)
On directionality in language change
Sometimes we also observe changes from a semantically empty stem extender to a meaning-bearing morphological affix or vice versa. Thus, the Old High German stem-extender -ir became the Modern German plural suffix -er (as in Kalb/Kälb-er “calf/calves”; Harris & Campbell 1995: 338), and the Latin derivational inchoative suffix -¯esc(¯o) somehow became the Romance stem-extender -isc(o) (as in Italian finisco “I finish”; Ramat 1992:552; Allen 1995). Again, I would say that these are changes internal to the morphology which are unrelated to grammaticalization and cannot be regarded as counterexamples to unidirectionality.
. Retraction Next I will consider a change type that I would like to call retraction. This is in some sense the opposite of expansion in grammaticalization, but it is not antigrammaticalization. Expansion is the development of new constructions or meanings that exhibit a greater degree of grammaticalization. Figure 2 shows a prototypical case of grammaticalization. As an item expands to the right and forms a grammaticalization chain, some of its earlier manifestations on the left typically disappear, so that the chain loses on the left what it gains on the right. Now we know that the older members of the chain do not have to be lost (this is often described as layering; Hopper 1991; Bybee et al. 1994: 21), so that, for instance, A1 does not have to be lost at stage 3 in Figure 2 (cf. stage 4, where A2 is still there despite the further expansion to A4 ). We also know that further expansion need not occur: an item may get lost from the language before it expands further (Hopper & Traugott 1993: 95). Now one additional possibility is a grammaticalization chain in which a right-hand member becomes obsolete. Everything in language can become obsolete, independently of its degree of grammaticalization, so there is no surprise here. Schematically such a situation is depicted in Figure 3. Until stage 4, the element B expands rightward, but then the members B4 and B3 are lost. Here I say that B has retracted to B2 . degree of grammaticalization
degree of grammaticalization
1. A1 2. A1 – A2 3. A2 – A3 4. A2 – A3 – A4 5. A4 – A5 A5 – A6 6.
1. B1 2. B1 – B2 3. B2 – B3 4. B2 – B3 – B4 5. B2 – B3 B2 6.
t
t
Figure 2. Rightward expansion
Figure 3. Retraction
(= Grammaticalization)
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.18 (34)
Martin Haspelmath degree of grammaticalization 1. C1 2. C1 – C2 3. C2 – C3 4. C3 – C4 5. C2 – C3 6. C1 – C2
Figure 4. Leftward expansion (= Antigrammaticalization)
But crucially, it has not expanded leftward. Such an example is seen in Figure 4, where the element C first grammaticalizes and then (at stage 5 and 6) antigrammaticalizes. The crucial difference between retraction and antigrammaticalization should be clear. However, retraction has sometimes been cited as counterevidence to unidirectionality. I will mention three cases. First, Newmeyer (1998: 273) discusses the English word man and notes: . . . the history of the word man also presents challenges for any sweeping claims about unidirectionality. In Old English, its predominant use was as an indefinite pronoun (cf. German man). Subsequently it seems to have swung back and forth from pronoun to full lexical noun and back again. In any event, it is the less grammaticalized use that has survived into Modern English.
But of course the non-grammaticalized word man always existed, so this is a case of retraction rather than leftward expansion. Second, Newmeyer (1998: 270–271) (citing Kroch et al. 1982: 287–291) mentions the case of English postverbal subject clitics in Early Modern English (e.g. Where dwellyth=she?). The pronouns allegedly underwent decliticization after 1550, so that in Modern English no subject clitics are found. But of course the independent subject pronouns always existed side by side with the subject clitics, so again this is not antigrammaticalization but simply retraction. The third example is the development of the English verb dare. According to Beths (1999), this goes against unidirectionality because dare was a semi-lexical verb in Old English, then became an auxiliary in Middle English and has reverted to a lexical verb in Modern English. But again, this is a case of retraction, not of leftward expansion. Traugott (2001: 9) observes: . . . this is not a conclusive counterexample to unidirectionality, because main verb dare to uses were always attested in the data. The best we can say is that the earlier main verb use was marginalized in the early periods and then the grammaticalized one was marginalized in turn and then lost in the later periods.
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.19 (35)
On directionality in language change
Thus, these changes do not provide counterevidence to the unidirectionality claim.11
. Conclusion I have made three major points in this article: That directionality constraints and other constraints on language change are an important prerequisite for understanding language change; that the unidirectionality of grammaticalization is one of the most important constraints on morphosyntactic change, despite various general criticisms; and that many of the alleged counterexamples to unidirectionality are not antigrammaticalizations and hence do not provide evidence against it. I will end the paper by making a few more general remarks.
. Broader agendas The first point concerns the usefulness of directionality constraints in the study of language change in general and in grammaticalization in particular. As I made clear in Section 2, I find them very useful, because only when we have a universal generalization do we have anything to explain. Thus, identifying and refining generalizations about directionality is high on my agenda. But is there also an opposite agenda? In other words, is there a theoretical perspective on language change that would want to ignore or deny directionality constraints because they do not fit into its general goals and assumptions? The answer is yes: If one thinks of language change as occurring exclusively in language acquisition, and if one thinks of crosslinguistic variation in terms of different settings of innate parameters, then one expects language change to be “essentially a random “walk” through the space of possible parameter settings” (Battye & Roberts 1995: 11). So it is in particular linguists with a Chomskyan perspective on language and language change that should see unidirectionality as a challenge, and they should try to discredit it. And indeed, at least David Lightfoot has argued against the general notion of directionality constraints (Lightfoot 1999: 34ff.) and against unidirectionality of grammaticalization in particular (Lightfoot 2002: 125–127).12 Also, Newmeyer (1998) contains a very critical chapter on grammaticalization, in which the author discusses many alleged counterexamples to unidirectionality. However, Newmeyer does not seem to be interested in defending a Lightfoot-style or Roberts-style approach to language change; his main concern seems to be to show that the evidence from grammaticalization is not incompatible with generative grammar. Moreover, he recognizes himself that there is a strong quantitative asymmetry favoring grammaticalization over its reverse, and he even proposes his own explanation of this asymmetry.13
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.20 (36)
Martin Haspelmath
But counterexamples to unidirectionality have also been highlighted by nonChomskyans who do not seem to have a broader agenda (Paolo Ramat, Richard Janda, Brian Joseph, Lyle Campbell, Alice Harris, Muriel Norde, Johan van der Auwera, Roger Lass). In the writings of some of these linguists, one senses a frustration with theoreticians who make broad sweeping claims but do not back them up with solid and careful historical linguistic work. Clearly, once one starts asking larger questions, there is the danger that one pays less attention to the data and more attention to the ideas, but there is also the opposite danger of missing the generalizations and the big picture because one sees too many details. Progress in linguistics will depend on finding a proper balance between these two ways of viewing the world of language change.
. Terminology The history of the term degrammaticalization shows how important it is to pay attention to terminological ambiguities. As pointed out in Section 3.2, coining this term was useful because it made people look for actual instances of the phenomenon. But it is also easy to take the term too literally, for instance in the sense of ‘loss of grammar’, or ‘lexeme creation on the basis of a grammatical item’. The potential for misunderstanding is particularly well illustrated by the term demorphologization, which has been used in two totally different senses: In Joseph and Janda (1988), it refers to a change from morphology to phonology, while in Hopper (1990), it refers to a change from a morphologically complex word to a simple lexical item (cf. also Ramat 2001: 394).14 These two change types are clearly unrelated, and similarly I have made the point that the various uses of degrammaticalization have rather little to do with each other.15 So when encountering the term “degrammaticalization”, one should first make sure to understand what exactly the author means by it before drawing conclusions from it. My own practice is to avoid the term entirely, and to use it only in scare quotes when talking about others’ terminological usage.
. Broad agreement As I made clear in Section 4.1, I accept the existence of exceptions to unidirectionality, and in this respect I find myself in agreement not only with grammaticalization critics, but also with other grammaticalization enthusiasts (Traugott 2001; Heine 2003). Thus, where is the disagreement? Do the detractors of grammaticalization studies claim that grammaticalization and antigrammaticalization are equally common? I am not aware of such a claim in the literature, and some of those linguists who have emphasized “degrammaticalization” have simultaneously made
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.21 (37)
On directionality in language change
it clear that grammaticalization occurs much more often (e.g. Ramat 1992: 549; Harris & Campbell 1995: 338; Janda 2001: 270). Newmeyer (1998) even offers an estimate of the relative proportions of both types of changes: “a rough impression is that downgradings have occurred at least ten times as often as upgradings” (Newmeyer 1998: 275–276). I have said elsewhere that my impression is that they are rather a hundred times as frequent (Haspelmath 2000a: 249), and it would be interesting to try to assess these impressionistic estimates in a systematic empirical way. Unfortunately, such an enterprise encounters some serious difficulties, not only the problem of defining a reasonable sample of languages and language changes (and of getting reliable diachronic data from different families), but also the problem of defining discrete events of grammaticalization (cf. Lass 2000: 214– 215). When a free adposition first becomes a clitic and then an affix, is this one change of grammaticalization or two? Eventually these difficulties could perhaps be overcome, but I still wonder whether such a quantitative study would be worth the trouble, in view of the fact that there is really broad agreement: Grammaticalization is far more common than antigrammaticalization, and this is a surprising fact. Anyone who is interested in understanding language change should be interested in why this is the case. I have proposed an explanation in Haspelmath (1999a), and I look forward to further contributions to this discussion.
Notes * I am grateful to the audience at the Amsterdam grammaticalization conference for their observations on my talk, and to Bernard Comrie, Muriel Norde and an anonymous referee for useful comments on an earlier version of this paper. . Like many other general laws, these laws probably have some exceptions (see, e.g., Haspelmath (1997b: 141–142) for possible exceptions to (1c)), but this does not mean that they are not laws, or that a general explanation is not possible or necessary. . Interestingly, Butt observes that this conclusion had already been reached by 19th century linguists (such as Beames 1872–1879), presumably because at that time most linguists were familiar with the concepts of grammaticalization and (implicitly) unidirectionality. It must be remembered that it was only the structuralism of Saussure and Bloomfield that made linguists forget about grammaticalization, until it was rediscovered toward the end of the 20th century. Incidentally, Butt’s alternative suggestion is that ne has a nominal origin, perhaps the Sanskrit locative form janye “for the sake of, because of ” (Butt 2001: 116). . Of course, what is an exception with respect to one generalization may be completely in line with a different generalization, and in that case it may be explainable after all. For instance, the English verb be shows exceptional inflectional properties that fall under no generalization of English grammar, but it is not an accident that it is this verb that shows the least regular inflection: If a language has any verbal suppletion, it almost always has
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.22 (38)
Martin Haspelmath
suppletion in its ‘be’ verb (Veselinova 2003). In a similar way, it might be that the exceptions to unidirectionality fall under a different generalization and can thus be understood in some way, but to my knowledge, nobody has proposed such a generalization. . This point was made by Osthoff and Brugmann (1878) in their Neogrammarian Manifesto: “. . . dass die sprache kein ding ist, das außer und über dem menschen steht und ein leben für sich führt, sondern nur im individuum ihre wahre existenz hat.” [. . . that language is not a thing which stands outside or above people and leads a life of its own, but has its true existence only in the individual.] (Osthoff & Brugmann 1878; cited after Ahrens 1969: 344). . It is worth remembering that two hundred years ago, the situation was very different. At that time, it was perfectly reasonable to speculate that our reconstructions of protolanguages brought us close to the first human languages. Until well into the 19th century, it was widely believed that life on earth was no more than six or seven thousand years old, and it was only the great discoveries of historical geology and evolutionary biology that made it clear that the biblical creation stories were way off the mark. . Thus, word-order change consisting of a change from freer to more fixed word order falls under grammaticalization as well (cf. Lehmann 1992), not just changes involving free words becoming dependent elements (clitics, affixes). . Noam Chomsky and his followers have repeatedly pointed out that language itself is an epiphenomenon (of internalized grammars), but they have generally failed to note that grammars, too, are epiphenomenal in the sense that they do not simply grow in individuals, but presuppose a community of speakers that the individual is part of (cf. Haspelmath 1999b). It is thus difficult to see which linguistic phenomena are not epiphenomenal. . In Haspelmath (1999a) I argue that function words do not in general replace content words in unconscious changes, because function words are produced more automatically than content words, and that this in part explains the irreversibility of grammaticalization. . In Dutch, there is a pronunciation difference between tig “dozens” (pronounced with a full vowel) and -tig as in veertig “forty” (pronounced with a schwa). This is not so easy to explain by back-formation, and I must assume that the Dutch form tig was created under the influence of the spelling and/or language contact. . Cf. the famous definition of grammaticalization from Kuryłowicz (1965: 69), which is often cited without the last part (starting with “e.g.”): Grammaticalization consists in the increase of the range of a morpheme advancing from a lexical to a grammatical or from a less grammatical to a more grammatical status, e.g. from a derivative formant to an inflectional one. In the next paragraph, Kuryłowicz asserts: “A reverse process is the lexicalization of a morpheme”, and his examples make it clear that he thinks of cases like Italian -ante/-ente. . A somewhat similar case is the Swedish verb må (cognate with English may, German mag), which originally meant only “may”, but now has acquired the meaning “feel” as well (van der Auwera & Plungian 1998: 105). When it means “feel”, it follows a different inflection class. Van der Auwera and Plungian (1998: 116–117) regard it as a counterexample to unidirectionality, but the morphological change in itself does not make the word less grammaticalized, and the semantic change does not go against any well-defined semantic
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.23 (39)
On directionality in language change
grammaticalization path. So it is certainly not a good exception to unidirectionality (see also Burridge 1995 for a somewhat similar case in Pennsylvania German). . See Haspelmath (1999a) for a critical review of Lightfoot (1999). Unfortunately, Lightfoot (2002) does not respond to these criticisms but mostly limits himself to simply repeating some of the points from his (1999) book. . See Haspelmath (2000a: 247–250) for further discussion of Newmeyer’s views on grammaticalization. . Another term that is used in multiple senses is ‘lexicalization’ (see Himmelmann (to appear) for some recent lucid discussion). Here, too, the terminological polysemy seems to derive from the fact that linguists have not necessarily followed other linguists’ usage, but have used lexicalization for diverse phenomena that can be interpreted as ‘putting in the lexicon’ or ‘making lexical’. . The term degrammaticalization has even been used in a totally different sense, to refer to the tendency in informal e-mail communication to violate normative grammatical rules (cf. Pansegrau 1997).
References Ahrens, Hans (1969). Geschichte der Sprachwissenschaft, Bd. 1. Frankfurt am Main: Athenäum Fischer. Allen, Andrew S. (1995). “Regrammaticalization and Degrammaticalization of the Inchoative Suffix.” In Henning Andersen (Ed.), Historical linguistics 1993: Selected papers from the 11th ICHL, Los Angeles, 16–20 August 1993 (pp. 1–8). Amsterdam & Philadelphia: John Benjamins. Battye, Adrian & Ian Roberts (Eds.). (1995). Clause Structure and Language Change. New York: Oxford University Press. Beames, John (1872–1879). A Comparative Grammar of the Modern Aryan Languages of India. Reprinted 1966. Delhi: Munshiram Manoharlal. Becker, Thomas (1993). “Back-Formation, Cross-Formation, and ‘Bracketing Paradoxes’ in Paradigmatic Morphology”. In Geert Booij & Jaap van Marle (Eds.), Yearbook of Morphology 1993 (pp. 1–25). Dordrecht: Kluwer. Benveniste, Émile (1958/1966). “Les Verbes Délocutifs.” Problèmes de Linguistique Générale, Vol. 1, 277–285 (1966). (Originally in Mélanges Spitzer, 57–63.) Beths, Frank (1999). “The History of Dare and the Status of Unidirectionality.” Linguistics, 37, 1069–1110. Börjars, Kersti (2003). “Morphological Status and (De)grammaticalisation: the Swedish Possessive.” Nordic Journal of Linguistics, 26, 133–163. Burridge, Kate (1995). “From Modal Auxiliary to Lexical Verb: The Curious Case of Pennsylvania German Wotte”. In Richard M. Hogg & Linda van Bergen (Eds.), Historical linguistics 1995: Germanic linguistics, Vol. II (pp. 19–33). (=Amsterdam Studies in the Theory and History of Linguistic Science, 162.) Amsterdam & Philadelphia: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.24 (40)
Martin Haspelmath
Butt, Miriam (2001). “A Reexamination of the Accusative to Ergative Shift in Indo-Aryan.” In Miriam Butt & Tracy Holloway King (Eds.), Time over matter (pp. 105–141). Stanford: CSLI Publications. Bybee, Joan L. (1985). Morphology: A Study of the Relation between Meaning and Form. Amsterdam & Philadelphia: John Benjamins. Bybee, Joan L. (2001). Phonology and Language Use. Cambridge: Cambridge University Press. Bybee, Joan L., William Pagliuca, & Revere D. Perkins (1994). The Evolution of Grammar. Chicago: The University of Chicago Press. Campbell, Lyle (1991). “Some Grammaticalization Changes in Estonian and their Implications”. In Elizabeth Closs Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization, Vol. I (pp. 285–299). Amsterdam & Philadelphia: John Benjamins. Campbell, Lyle (2001). “What’s Wrong with Grammaticalization?” Language Sciences, 23, 113–161. Campbell, Lyle (Ed.). Grammaticalization: a Critical Assessment. Special issue of Language Sciences, 23, 2–3. Croft, William (2001). Radical Construction Grammar. Oxford: Oxford University Press. Crowley, Terry (1997). An Introduction to Historical Linguistics. Oxford: Oxford University Press. Dahl, Östen (2002). “Inflationary Effects in Language and Elsewhere.” In Joan L. Bybee & Paul J. Hopper (Eds.), Frequency and the Emergence of Linguistic Structure (pp. 471– 480). Amsterdam & Philadephia: John Benjamins. Ferguson, Charles A. (1990). “From Esses to Aitches: Identifying Pathways of Diachronic Change.” In William Croft, Keith Denning, & Suzanne Kemmer (Eds.), Studies in Typology and Diachrony (pp. 59–78). Amsterdam & Philadelphia: John Benjamins. Fischer, Olga (2000). “Grammaticalisation: Unidirectional, Non-Reversable? The Case of to before the Infinitive in English”. In Fischer et al. (Eds.), 149–169. Fischer, Olga, Anette Rosenbach, & Dieter Stein (Eds.). (2000). Pathways of Change: Grammaticalization in English. (= Studies in Language Companion Series, 53.) Amsterdam & Philadelphia: John Benjamins. Fitzmaurice, Susan (2000). “Remarks on De-grammaticalization of Infinitival to in PresentDay American English.” In Fischer et al. (Eds.), 171–186. Frajzyngier, Zygmunt (1997). “Bidirectionality in Grammaticalization.” In Robert K. Herbert (Ed.), African Linguistics at the Crossroads: Papers from Kwaluseni (1st World Congress of African Linguistics, Swaziland 1994) (pp. 17–38). Köln: Köppe. Geurts, Bart (2000a). “Explaining Grammaticalization (the Standard Way).” Linguistics, 38, 781–788. Geurts, Bart (2000b). “Function or Fashion? Reply to Martin Haspelmath.” Linguistics, 38, 1175–1180. Greenberg, Joseph (1966). Language Universals, with Special Reference to Feature Hierarchies. The Hague: Mouton. Hagège, Claude (1993). The Language Builder: An Essay on the Human Signature in Linguistic Morphogenesis. Amsterdam & Philadelphia: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.25 (41)
On directionality in language change
Hagège, Claude (2001). “Les Processus de Grammaticalisation.” In Martin Haspelmath et al. (Eds.), Language typology and language universals: An International Handbook (= HSK, 20–2) (pp. 1609–1623). Berlin: de Gruyter. Harris, Alice & Lyle Campbell (1995). Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Haspelmath, Martin (1997a). Indefinite Pronouns. (= Oxford Studies in Typology and Linguistic Theory.) Oxford: Oxford University Press. Haspelmath, Martin (1997b). From space to Time: Temporal Adverbials in the World’s Languages. (= Lincom Studies in Theoretical Linguistics, 3.) München & Newcastle: Lincom Europa. Haspelmath, Martin (1998). “Does Grammaticalization need Reanalysis?” Studies in Language, 22, 49–85. Haspelmath, Martin (1999a). “Why is Grammaticalization Irreversible?” Linguistics, 37, 1043–1068. Haspelmath, Martin (1999b). “Are there Principles of Grammatical Change? A Review Article of [Lightfoot, David (1999). The development of language: Acquisition, change, and evolution. Oxford: Blackwell.]” Journal of Linguistics, 35, 579–595. Haspelmath, Martin (2000a). “Why Can’t we Talk to each other? A Review Article of [Newmeyer, Frederick (1998). Language form and language function. Cambridge: MIT Press.]” Lingua, 110, 235–255. Haspelmath, Martin (2000b). “The relevance of Extravagance: a Reply to Bart Geurts.” Linguistics, 38, 789–798. Haspelmath, Martin (2003). “The Geometry of Grammatical Meaning: Semantic Maps and Cross-Linguistic Comparison.” In Michael Tomasello (Ed.), The New Psychology of Language, Vol. II (pp. 211–242). Mahwah, NJ: Lawrence Erlbaum. Heine, Bernd (2003). “On Degrammaticalization.” In Barry J. Blake & Kate Burridge (Eds.), Historical Linguistics 2001 (pp. 163–179). Amsterdam & Philadelphia: John Benjamins. Heine, Bernd & Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd & Mechthild Reh (1984). Grammaticalization and Reanalysis in African Languages. Hamburg: Buske. Himmelmann, Nikolaus (to appear). “Lexicalization and Grammaticalization: Opposite or Orthogonal?” In Walter Bisang, Nikolaus Himmelmann, & Björn Wiemer (Eds.), Grammaticalization and Lexicalization. Berlin: Mouton de Gruyter. Hoenigswald, Henry M. (1991). “Morphemic Change, Typology, and Uniformitarianism: a Study in Reconstruction.” In Winfred P. Lehmann & Helen-Jo Jakusz Hewitt (Eds.), Language Typology 1988: Typological Models in Reconstruction (pp. 17–26). Amsterdam & Philadelphia: John Benjamins. Hopper, Paul (1990). “Where do Words Come from?” In William Croft, Keith Denning, & Suzanne Kemmer (Eds.), Studies in Typology and Diachrony (pp. 151–160). Amsterdam & Philadelphia: John Benjamins. Hopper, Paul (1991). “On some Principles of Grammaticalization.” In Elizabeth Closs Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization, Vol. I (pp. 17–35). Amsterdam & Philadelphia: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.26 (42)
Martin Haspelmath
Hopper, Paul (1994). “Phonogenesis.” In William Pagliuca (Ed.), Perspectives on Grammaticalization (pp. 29–45). Amsterdam & Philadelphia: John Benjamins. Hopper, Paul & Elizabeth C. Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Janda, R. (1980). “On the Decline of Declensional Systems: the Overall Loss of OE Nominal Case Inflections and the ME Reanalysis of -es as his”. In E. C. Traugott, R. Labrum, & S. Shepherd (Eds.), Papers from the 4th International Conference on Historical Linguistics (pp. 243–252). Amsterdam & Philadelphia: John Benjamins. Janda, Richard D. (2001). “Beyond “Pathways” and “Unidirectionality”: on the Discontinuity of Language Transmission and the Counterability of Grammaticalization.” Language Sciences, 23, 265–340. Joseph, Brian & Richard Janda (1988). “The How and Why of Diachronic Morphologization and Demorphologization”. In Michael Hammond et al. (Eds.), Theoretical Morphology: Approaches in Modern Linguistics (pp. 193–210). San Diego: Academic Press. Kim, Hyeree (2001). “Remarks on the Unidirectionality Principle in Grammaticalization.” Folia Linguistica Historica, 22, 49–65. Kroch, Anthony et al. (1982). “Understanding Do”. In K. Tuite et al. (Eds.), Papers from the 18th Regional Meeting, Chicago Linguistic Society (April 15–16, 1982) (pp. 282–294). Chicago: Chicago Linguistic Society. Kuryłowicz, Jerzy (1965/1975). “The Evolution of Grammatical Categories”. Diogenes, 51, 55–71. (Reprint in: Esquisses linguistiques, Vol. II, 38–54. München: Fink, 1975.) Lass, Roger (2000). “Remarks on (Uni)directionality.” In Fischer et al. (Eds.), 207–227. Lazzeroni, Romano (1998). “Divagazioni sulla Degrammaticalizzazione.” In Giuliano Bernini et al. (Eds.), Ars Linguistica: Studi offerti a Paolo Ramat (pp. 275–283). Roma: Bulzoni. Lehmann, Christian (1993). “Theoretical Implications of Processes of Grammaticalization.” In W. A. Foley (Ed.), The role of Theory in Language Description (pp. 315–340). Berlin: Mouton. Lehmann, Christian (1995a [1982]). Thoughts on Grammaticalization. Munich: LINCOM Europa. (Originally: Institute für Sprachwissenschaft, Universität zu Köln, 1982.) Lehmann, Christian (1995b). “Synsemantika.” In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld, & Theo Vennemann (Eds.), Syntax, Vol. II (pp. 1251–1266). Berlin: de Gruyter. Lightfoot, David (1999). The Development of Language: Acquisition, Change, and Evolution. Oxford: Blackwell. Lightfoot, David W. (2002). “Myths and the Prehistory of Grammars.” Journal of Linguistics, 38, 13–136. Luraghi, Silvia (1998). “On the Directionality of Grammaticalization.” Sprachtypologie und Universalienforschung, 51, 355–365. Maddieson, Ian (1984). Patterns of Sounds. Cambridge: Cambridge University Press. Matsumoto, Y. (1988). “From Bound Grammatical markers to Free Discourse Markers: History of some Japanese Connectives”. Berkeley Linguistics Society, 14, 340–351. Meillet, Antione (1912). “L’Évolution des Formes Grammaticales.” Scientia, 12. (Reprinted in Linguistique historique et linguistique générale. Paris: Champion.)
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.27 (43)
On directionality in language change
Méndez Dosuna, J. (1997). “Fusion, Fission, and Relevance in Language Change: Deuniverbation in Greek Verb Morphology”. Studies in Language, 21, 577–612. Mowrey, Richard & William Pagliuca (1995). “The Reductive Character of Articulatory Evolution.” Rivista di Linguistica, 7, 37–124. Nevis, J. (1986a). “Decliticization and Deaffixation in Saame: Abessive Taga”. In B. Joseph (Ed.), Studies on Language Change (= Working papers in Linguistics, 34) (pp. 1–9). Ohio: Ohio State University, Department of Linguistics. [summarized in Joseph & Janda 1988: 200.] Nevis, J. (1986b). “Decliticization in Old Estonian”. In B. Joseph (Ed.), Studies on Language Change (= Working Papers in Linguistics, 33) (pp. 10–27). Ohio: Ohio State University, Department of Linguistics. Newmeyer, Frederick J. (1998). Language Form and Language Function. Cambridge, MA: MIT Press. Norde, Muriel (1997). The History of the Genitive in Swedish: A Case Study in Degrammaticalization. Amsterdam: Vakgroep Skandinavische taal- en letterkunde. Norde, Muriel (2001). “Deflexion as a Counterdirectional Factor in Grammatical Change”. Language Sciences, 23, 231–264. Osthoff, Hermann & Karl Brugmann (1878). “Vorwort”. Morphologische Untersuchungen auf dem Gebiet der indogermanischen Sprachen. Leipzig: Hirzel. Pansegrau, Petra (1997). “Dialogizität und Degrammatikalisierung in E-mails.” In R. Weingarten (Ed.), Sprachwandel durch Computer (pp. 86–104). Opladen: Westdeutscher Verlag. Pintzuk, Susan, George Tsoulas, & Anthony Warner (Eds.). (2000). Diachronic Syntax: Models and Mechanisms. Oxford: Oxford University Press. Plank, Frans (1995). “Entgrammatisierung – Spiegelbild der Grammatisierung?” In Norbert Boretzky et al. (Eds.), Natürlichkeitstheorie und Sprachwandel (pp. 199–219). Bochum: Brockmeyer. Ramat, Paolo (1992). “Thoughts on Degrammaticalization”. Linguistics, 30, 549–560. Ramat, Paolo (2001). “Degrammaticalization or Transcategorization?” In Chris SchanerWolles, John Rennison, & Friedrich Neubarth (Eds.), Naturally! Linguistic studies in honour of Wolfgang Ulrich Dressler presented on the occasion of his 60th birthday (pp. 393–401). Turin: Rosenberg & Sellier. Roma, Elisa (1999). “Grammaticalizzazione di Pronomi Soggetto: una Strada Alternativa.” Archivio Glottologico Italiano, 84, 1–35. Traugott, E. C. (1994). “Grammaticalization and Lexicalization”. In R. Asher & J. Simpson (Eds.), Encyclopedia of Language and Linguistics, Vol. III (pp. 1481–1486). Oxford: Pergamon Press. Traugott, Elizabeth C. (2001). “Legitimate Counterexamples to Unidirectionality.” Paper presented at the University of Freiburg, 17 October 2001. (http://www.stanford. edu/∼traugott/papers/Freiburg Unidirect.pdf accessed 9 January 2004) van der Auwera, Johan (2002). “More Thoughts on Degrammaticalization.” In Ilse Wischer & Gabriele Diewald (Eds.), New Reflections on Grammaticalization (pp. 19–29). Amsterdam & Philadelphia: John Benjamins. van der Auwera, Johan & Vladimir A. Plungian (1998). “Modality’s Semantic Map.” Linguistic Typology, 2, 79–124.
TSL[v.20020404] Prn:13/04/2004; 15:41
F: TSL5902.tex / p.28 (44)
Martin Haspelmath
van Kemenade, Ans & Nigel Vincent (Eds.). (1997). Parameters of Morphosyntactic Change. Oxford: Oxford University Press. Veselinova, Ljuba (2003). Suppletion in Verb Paradigms: Bits and Pieces of a Puzzle. Ph.D. dissertation, University of Stockholm. Wichmann, Søren (1996). “The Degrammaticalization of Agentivity in Tlapanec.” In Elisabeth Engberg-Pedersen, Michael Fortescue, Peter Harder, Lars Heltoft, & Lisbeth Falster Jakobsen (Eds.), Content, expression and structure: Studies in Danish Functional Grammar (pp. 343–360). Amsterdam & Philadelphia: John Benjamins. Winter, Werner (1971). “Formal Frequency and Linguistic Change.” Folia Linguistica, 5, 55– 61. Wurzel, Wolfgang Ullrich (1994). Grammatisch initiierter Wandel. Bochum: Brockmeyer.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.1 (45)
Rescuing traditional (historical) linguistics from grammaticalization theory* Brian D. Joseph The Ohio State University
.
Introduction
There is a rich history, dating back, according to Heine (2003), at least to observations in the 18th century by Condillac (1746) and Horne Tooke (1857) (first published 1786), if not earlier, associated with the study of the phenomenon of grammaticalization, i.e. how grammatical elements arise out of less (or non-) grammatical elements.1 Still, only in the past twenty-five or so years has it attracted considerable attention among professional linguists. Indeed, what began as an interesting observation – one might think of Givon’s (1971) dictum that “today’s morphology is yesterday’s syntax” as the starting point for the ‘modern’ era of grammaticalization studies, though Meillet (1912) and Kurylowicz (1964) are also possible benchmark studies in this regard – has turned into a ‘cottage industry’ of sorts, and is now to the point where it can be considered a ‘movement’. Among the evidence that points towards such a ‘movement’ are the following indicators. First, there is a huge amount of relevant literature now, with textbooks and surveys (e.g. Diewald 1997; Heine, Claudi, & Hünnemeyer 1991; Hopper & Traugott 1993; Lehmann 1982/1995), two dictionaries or similar compendia (Lessau 1994; Heine & Kuteva 2002), singly-authored studies (e.g. Bowden 1992; Heine 1993, 1997; Kuteva 2001; Ziegeler 2000, to name a few), and numerous edited volumes (such as Pagliuca 1994; Ramat & Hopper 1998b; Traugott & Heine 1991b; Wischer & Diewald 2002, among many others) all dedicated to different aspects of the study of grammaticalization phenomena. Second, there are now many conferences devoted to aspects of grammaticalization.2 Third, ‘grammaticalization’ is mentioned specifically as a separate category in that part of a recent annual report of the editor of Language enumerating the different sub-areas for submitted papers (see Aronoff 2002) and it is a subheading in Library of Congress subject
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.2 (46)
Brian D. Joseph
headings for classifying the content of books.3 Finally, a less traditional but no less indicative fact of relevance here is that searches of the World-Wide Web using a search engine such as ‘Google’ (www.google.com) yield the results noted in (1):4 (1) ‘Grammaticalization’ searches on google.com using search word grammaticalization: 6460 hits using search word grammaticalisation: 1770 hits using search word grammaticization: 783 hits using search word grammaticisation: 104 hits These numbers are fairly robust, to be sure, though admittedly other frameworks for the study of grammar yielded higher numbers; for instance, HPSG (HeadDriven Phrase Structure Grammar) yielded some 17,000 hits, Optimality Theory 88,000 hits, and Minimalist Syntax 19,300. Nonetheless, in absolute terms, the results in (1) suggest a high level of scholarly activity. All these indicators suggest that ‘grammaticalization studies’, generally referred to now as ‘grammaticalization theory’ (cf., e.g., Heine et al. 1991; Heine & Kuteva 2002: 2),5 have ‘made it’, so to speak, and command the attention of serious linguists. These various pointers are telling, as is the fact that some of the more exuberant and ardent practitioners of grammaticalization have put forth very strong claims about the relationship of grammaticalization studies to other areas within linguistics and to the very field itself. For instance, Hopper (1991: 33) suggested that the study of grammaticalization is coterminous with the study of language change itself, saying that “since grammaticization is always a question of degree, not an absolute, the criteria which control this gradation are not restricted to grammaticization, but are simply general criteria of change”, and this viewpoint was picked up by Ramat and Hopper (1998a: 3-4) in their statements that “an extreme formulation . . . is that ultimately all grammaticalization is not separately definable from the concept of change in general” and that “we may still speculate on whether there are cases of language change that are not part of the phenomenon of grammaticalization”. Moreover, it is echoed as well in the assertion by Katz (1998: 95) that “the process of grammaticalization is a constant force that drives language onward, and it can account for all change that moves from iconicity to formalism”.6 To be fair, there is disagreement expressed in the grammaticalization literature on this very point: for Traugott and Heine (1991a: 3), for instance, grammaticalization is merely “a kind of language change”, and so does not encompass all change. Furthermore, some have taken an expanded view of grammaticalization theory in which it essentially points the way to a new theory of language itself. This view is implicit, for example, in Hopper’s ‘emergent grammar’ model (Hopper 1987) in which “there is . . . no ‘grammar’, but only grammaticalization – movements
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.3 (47)
Rescuing traditional (historical) linguistics
towards structure” (148). And, as Newmeyer (2001: 188) describes this particular stance, for “Heine et al. (1991: 1), ‘grammaticalization theory’ challenges what they see as the predominant conception of theoretical linguists since Saussure. Indeed, they feel that it calls for a ‘new theoretical paradigm’”. Given such evidence of acceptance for grammaticalization theory and such viewpoints as those quoted here,7 a question that naturally arises is what the relation is between grammaticalization and the more traditional areas of study within linguistics, and in particular, given the importance of diachrony in grammaticalization studies, historical linguistics, the subfield upon which the modern scientific study of language was built in the 19th and into the 20th centuries. Accordingly, in this paper, I explore various facets of this question. For instance, many linguists trained in traditional historical linguistics seem somewhat reluctant to jump on this grammaticalization bandwagon and embrace the new theory;8 is it because their methods and results are so at odds with those of grammaticalization? In view of grammaticalization being a relatively new paradigm for analyzing language change,9 is it the case, as is often so with newly emerging paradigms and theories, that historical linguists have been doing things so wrong in the past that everything needs to be redone?10 I suggest that this reluctance is not because those with traditional training are hidebound dinosaurs unwilling to give up on putatively obsolete methods and retool for a new era of historical linguistics – rather, I would say, it is the result of there not being any distinct advantage to this new ‘paradigm’. In particular, grammaticalization theory, it seems to me, perhaps inadvertently, often takes stances that are quite at odds with constructs and notions about language and language change that have long been held and upheld within traditional historical linguistic frameworks; for those schooled traditionally, therefore, grammaticalization comes across as just flat out wrong. For instance, just to give a taste of what is to come, certain ways in which phonetic reduction is invoked in discussions of grammaticalization fly in the face of what is known about the regularity of sound change and the sorts of conditioning that can hold on sound changes. Similarly, well-understood processes of analogy and reanalysis are often abandoned in favor of claims about grammaticalization as a process of change. I argue here, moreover, that although purporting to be a theory of language change, grammaticalization theory – at least as it is practiced by many – is often ahistorical, not giving due consideration to the full range of information about the steps in a particular development and attempting to work out the history of various phenomena from synchrony alone. Also, far from being a theory of language in general, it is often asynchronic as well, with only vague synchronic analyses given, even though understanding the historical development generally depends crucially on understanding the synchronic status of a given element at various stages. These properties make grammaticalization as a framework problematic if it were to serve as a replacement
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.4 (48)
Brian D. Joseph
for – or even simply as a complement to – traditional historical linguistics, that is if its status as a new paradigm for analyzing language change is taken seriously. My remarks here, therefore, are intended to stand as somewhat of a corrective to the onrush of enthusiasm shown in recent years for grammaticalization.
. Grammaticalization as anti-neogrammarian (or: a-historical linguistics as well as a-historical) I regret to say that it seems to me that much of what passes for historical linguistics in a grammaticalization framework is ill-informed and quite simply seems not to really know or understand traditional historical linguistics; as a result, claims that the proponents of grammaticalization make about how it relates to traditional historical linguistics really carry no weight at all. This fact in turn would explain why traditional historical linguists generally see no benefit to working within this new paradigm. The ‘School’ most associated with traditional historical linguistics is that of the Neogrammarians, so I consider here how grammaticalization as a guiding principle for investigating language change measures up against results and methods associated with Neogrammarian practices.11 One telling remark made within the grammaticalization literature is the following on the Comparative Method, taken from the Cambridge University Press advertising blurb for Heine and Kuteva (2002): “While the comparative method is concerned with regularities in phonological change, grammaticalization theory deals with regularities of grammatical change”. Admittedly, these may not be Heine and Kuteva’s own words, but they are consistent with the statement found in Heine (2003: 596) that the Comparative Method looks to “regularities . . . for example in sound correspondences”. They seem to be indicative of a claim concerning the Comparative Method that is hard to justify, however. Statements of this sort display a misunderstanding of the Comparative Method on the part of grammaticalization adherents. The implication of such a statement is that the Comparative Method deals only with phonology, in contrast to grammaticalization theory, but in fact it has never been the case that the comparative method is concerned just with phonology. The essence of the comparative method is the determination of comparable features across related languages, with the goal being to use that comparison as a basis for understanding what the starting point for the related languages was like, i.e. for reconstructing aspects of the proto-language for the languages in question regardless of the level of structure for the feature at issue. Therefore, it is unclear why, for purposes of contrasting it with grammaticalization, one would focus just on the Comparative Method as it pertains to phonological de-
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.5 (49)
Rescuing traditional (historical) linguistics
velopments. Indeed, the Comparative Method has been applied in domains other than phonology, with considerable success, where success is measured by the degree of acceptance the results have enjoyed among specialists.12 In contrast to the view that the Comparative Method is limited to phonology, it can be noted that Franz Bopp’s famous first application of the comparative method in 1816 was his study On the Conjugation System of Sanskrit in Comparison with that of the Greek, Latin, Persian, and Germanic Languages (emphasis added); that is, his goal was morphological reconstruction. Similarly, the comparative method has been used successfully to reconstruct many aspects of linguistic (sub)systems other than the phonological – for example, for Proto-Indo-European (PIE) alone the following striking (and generally agreed upon) results can be cited as evidence of what the Comparative Method has yielded (with similar results well-known for other languages and language families): (2) Non-phonological successes from traditional application of Comparative Method in IE a. forms, besides roots and derivational affixes such as *-ti- for abstract nouns, also inflectional elements such as person/number endings, e.g. *-mi for 1SG present, or case endings, e.g. *-s for nominative singular, *-m for accusative singular b. categories, such as present tense, aoristic aspect, nominal cases (nominative, accusative, locative), etc. c. morphological processes, e.g. e/o/Ø ablaut, both for derivation and for inflection d. morphophonemic rules, e.g. the shared degemination of /*-ss-/ to [-s-] as in Sanskrit 2SG asi “you are” from /as-si/ paralleled by Greek eî “you are” from /es-si/ (admittedly, one reaches these results in each language by internal reconstruction, but then in putting the results of the internal reconstructions in each language in juxtaposition with one another, one is using the comparative method to project back to the common ancestor of Greek and Sanskrit, that is, PIE) e. syntax, e.g. the agreement pattern by which neuter plural nouns trigger singular agreement, found in Hittite, Vedic Sanskrit, Avestan, and Ancient Greek; thus presumably a PIE pattern f. semantics, e.g. Sanskrit vadhris “castrated”, Greek éthris “castrated ram”, so that if from a root *wedh- “strike” (thus, *wedh-ri-), then presumably the derivative had the specialized semantics referring to castration already in PIE g. poetic formulas, e.g. Adalbert Kuhn’s (1853) demonstration that Sanskrit svravah. aks. itam and Homeric Greek kléos áphthiton, both mean-
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.6 (50)
Brian D. Joseph
ing “imperishable fame”, are to be equated and reflect a PIE poetic formula; themes of poetic diction (cf. Watkins 1995 on IE dragonslaying myths) h. legal practice as well as legal diction (cf. Watkins 1987). This listing in (2) is a fairly impressive catalogue of successful results – successful in that, as noted above, they are generally accepted by most practicing IndoEuropeanists – from the comparative method, so that it is fair to ask the question of whether any of those results could have been achieved with ‘grammaticalization’ alone as the guiding methodology, even concerned as it is with ‘regularities of grammatical change’. That is, is something more really needed to allow reconstruction beyond what the comparative method can achieve? Similar sorts of questions can be raised with regard to statements made about Internal Reconstruction and grammaticalization; for instance, Heine (2003: 596– 597) has said that grammaticalization “may as well concern language-internal analysis. In this respect grammaticalization resembles internal reconstruction . . . . Compared to the latter, however, which concentrates on unproductive/irregular alternations, grammaticalization studies are not restricted in such a way.” This alleged ‘restriction’ of language-internal analysis, however, is not really substantiated by the facts of what internal reconstruction is and how it has been used: in essence, internal reconstruction is just a method of generating hypotheses, by any means available, about otherwise unreachable earlier stages. While there are a number of textbook examples of internal reconstruction that do work from irregularities (e.g. the classic analysis and reconstruction of PIE laryngeals given by de Saussure (1878), based on irregularities in vowel-alternation (‘ablaut’) patterns), they need not, and studies have not been limited to these types of phenomena. For instance, the regular but generally morphologically conditioned palatalizations that are quite widespread in Slavic, as in Bulgarian 1SG present plaˇca “I cry” vs. 1SG aorist plakax “I cried, or Russian vostok “east” vs. vostoˇcnyj “eastern”, coupled with what is known generally about possible phonological triggers for palatalization, lead to a reasonable, internally arrived at reconstruction of some [+front] segment after the velar in cases like these, e.g. *plak-j- or *vostok-i.13 Of course, one could say that the Slavic case does involve an irregularity (in the form of an unmotivated [ˇc] in the present stem or in the adjective), but the basic idea is always that of drawing inferences from gaps in patterns and from what is known about language change in general. Viewed in this way, internal reconstruction is not supplanted or complemented even by grammaticalization; rather, in this sense, grammaticalization is rather just another instantiation of the basic methodology of internal reconstruction. This last point is made clear by many of the cases reported on in Heine and Kuteva (2002), where, for example, the co-existence of a noun with a given meaning and of an adverbial element with a similar form and a related meaning is taken to justify a
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.7 (51)
Rescuing traditional (historical) linguistics
claim of historical relationship leading (generally) to a hypothesis of the historical derivation of the adverbial from the noun. Such cases can involve synchronic polysemy, as with Colonial Quiché chi “lip” and chi “in, into, out of ” reported by Dürr (1988: 52; see Heine & Kuteva 2002: 195) or synchronic alternations, as with Kono kùn “head” alternating with kù- in kùma “over”, where -ma is a synchronic postposition meaning “on” so that kùma is quite plausibly derived synchronically (and thus historically, by inference) from /kùn-ma/, as reported in Heine and Kuteva (2002: 69–70, drawing on information provided to them by Lessau). The historical inferences here are quite reasonable and may well be right, but the important methodological point is that the basis in such cases is entirely synchronic in nature; internal reconstruction is an inherently ahistorical method, working as it does from synchronic facts and patterns and drawing inferences therefrom.14 More generally, besides these traditionally recognized methods of studying change, there are the various well-recognized mechanisms of change to wonder about in regard to grammaticalization, e.g., sound change, analogy, metaphoric extension, etc. In particular, where do they fit in within a grammaticalization framework, and more to the point, if they are needed independently, is grammaticalization needed as something yet again beyond them?15 Among the argumentation that is relevant here is the discussion in Joseph (2001a) (see also Joseph 2003) regarding the Modern Greek future marked with θa from earlier thélei na “it-wants that” and the development of the Modern Greek weak nominative pronoun tos (cf. earlier Greek a(u)tos). I claimed that although these look like textbook cases of grammaticalization, in that, for instance, the emergence of θa involves semantic bleaching (‘want’ → FUTURE), phonetic reduction, and a shift of lexical status to affixal status, in fact the bleaching took place very early on, at a point at which the full verb was intact, and to account for the phonetic reduction and shift in status, all one needs to posit is regular sound change and analogy as the processes by which the changes leading to a future marker with the form θa were effected. Based on that, I suggested that grammaticalization is not a process in and of itself but rather just a label given to a particular type of outcome of independently needed mechanisms of change.16 I offer a further speculation here. This result is reached in a situation where there is access to a significant amount of detail about the historical development, as in fact all the relevant stages in the emergence of θa are attested generally with abundant documentation.17 Perhaps, then, this sort of scenario, in which wellunderstood processes of change give results that one might interpret as showing a development towards a greater degree of embedding in grammatical structure for some element, might well be so for all cases of grammaticalization, especially those based on considerably less direct evidence and more a matter of internally arrived at inferential reconstruction. At the very least, I would argue, putative cases of grammaticalization should be re-examined in light of the process vs. outcome
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.8 (52)
Brian D. Joseph
interpretation of just what sort of phenomena are subsumed under grammaticalization. To some extent, a basic problem here, as I see it, is that either by working from synchronic data and drawing even well-reasoned historical inferences (i.e., making a claim of grammaticalization based on internal reconstruction) or by comparing two stages somewhat distant in time and trying to infer what the pathways from one stage to the other were, one is falling into the trap discussed by Andersen (1989) of confounding a diachronic correspondence with an actual innovation or change. While this is not a problem that is restricted to grammaticalization studies, it is a real one, and, speaking just impressionistically here, it is one that proponents of grammaticalization, with its appeal to universally applicable pathways of change that in essence do the work of historical investigation for one, seem rather prone to fall victim to. These musings suggest that a thorough re-examination of all putative cases of grammaticalization in this way would be revealing, but a task of that sort is well beyond the scope of this paper. Still, I mention here one instance that is consistent with the sort of approach that I advocate and thus relevant to the concerns I raise here. In particular, Rosén (2002) has provided an extremely detailed account of the emergence within historical Latin – thus involving stages of the language for which there is ample documentation – of the generalizing indefinite relative pronouns quiuis and quilibet, both meaning “any one whatever (i.e. any one you wish) of three of more”. Their development shows some characteristics that are reminiscent of what happens in grammaticalization, in that clausally linked elements (e.g. relative pronoun qui and a form of velle “to want”18 occurring in a full relative clause) come to be compacted into a single word, with the result that without the details of the history one might be inclined to simply refer this as a case of grammaticalization and then move on. However, Rosén concludes (2002: 104) that it “was a gradual process . . . [and] no delexicalization, or semantic depletion, was involved, nor morphophonemic attrition, but eventually fossilization of verb-form and – in certain patterns – syntactic metanalysis of sentence constituents”. Moreover, she suggests, ultimately (111), that “the inevitable, but not unexpected conclusion is not to take for granted, when encountering one materialization, any other of the facets of grammaticalization”. Thus here too, an apparent instance of grammaticalization of a seemingly classical sort evaporates upon closer examination of the facts. Another area in which assumptions and practices of grammaticalization are at odds with traditional wisdom about aspects of language change is sound change. The traditional, i.e. Neogrammarian, understanding of sound change together with what, to my thinking, is the best available evidence, points to sound change as being purely phonetically based; apparent cases of nonphonetic conditioning usually represent results of reanalysis or reinterpretation of once phonetically con-
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.9 (53)
Rescuing traditional (historical) linguistics
ditioned phenomena (see Hock 1976, and Section 6 below). Such a view, among other things, allows one to derive regularity of sound change since phonetic factors can be thought of as the most generally applicable sorts of conditions, being defined entirely in terms of aspects of production of sound and thus not linked to particular lexical items or morphological categories (see Labov 1981, 1994; Joseph 1999; Janda & Joseph 2003). Nonetheless, some of the discussion in the grammaticalization literature seems to take a view that is counter to this Neogrammarian stance on sound change, in that phonetic ‘erosion’ of grammatical forms (e.g. English not → n’t) is often talked about as if there were nonphonetic conditioning of sound change. Thus, Hopper and Traugott (1993: 150), while acknowledging the value of recognizing regular sound change, nonetheless talk about ‘special phonological changes’ in particular collocations of grammatical elements. Cases like this seem to be saying, contra the Neogrammarians, that the fact of being an (incipient) grammatical element alone conditions the phonetic reduction; relevant here also is the fact that frequency is often referred to in the grammaticalization literature as a conditioning factor (as if phonetic erosion were like geological erosion).19 Both grammatical status and frequency would violate Neogrammarian principles, as they are nonphonetic conditioning factors, being in the first instance a matter of a higher level of analysis beyond phonetics alone and in the second a matter of usage. Still, in keeping with the Neogrammarians, one could say that phonetic ‘erosion’ or ‘reduction’ would not directly be a result of grammatical status; rather it could reflect the effects of low prosodic prominence that function words often show. Under such a view, the change of nòt → n’t would be due to low salience in phrasal prosody or to stresslessness, not to grammatical status or frequency per se; indeed, Hopper and Traugott do recognize the relevance of prosody and low stress in this particular change. It can be noted here, in a somewhat similar case, that in the case of the development of the Greek future marker alluded to above, some of the relevant steps were simply ordinary regular sound changes and not any sort of special developments that affected only grammatical forms; e.g., the apocope of -a from the intermediate stage thana (from thélei na, as noted above) was regular before certain vowels, and in the resulting form than, the loss of -n was regular before (most) consonants, and so on. Thus there is no reason a priori to say that grammatical forms have to be treated differently by sound change from non-grammatical forms;20 sound change can indeed operate in the mechanical way in which the Neogrammarians conceived of it, even with regard to grammaticalized elements.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.10 (54)
Brian D. Joseph
. Grammaticalization as ahistorical Many practitioners of grammaticalization make claims about language change and historical linguistics, even if, as argued above, they do not follow the tenets and practices of traditional historical linguistics. That is perhaps reasonable enough, as it is conceivable that the grammaticalization perspective offers something new to the study of language change (even if the previous section casts doubt on such an assertion). Still, it can be argued further that despite a desire to contribute to an understanding of language change and language history, certain aspects of the way in which proponents of grammaticalization arrive at these contributions show instead that they take an ahistorical approach that often bypasses crucial considerations needed to make historical accounts work. In particular, a key concern in any historical investigation is what may be called the issue of ‘direct lineal descent’. That is, given an element X in Stage n and an apparent X’ in Stage n + m, is there a direct lineal connection between the two so that we can legitimately say that X’ is an altered form of X at a later stage of the language? Oftentimes the connection is obvious and no one has any objections, as for the most part with the examination of languages with relatively long historical records, such as English or Greek or the Romance languages. Thus, no one seriously disputes that Old English nu: is the direct lineal ancestor of Modern English now or hu:s of house, and so on, and with good reason – there is the principle of regularity of sound change to make this connection work. But a similar connection between earlier English pi:pen “to make a high-pitched noise” and Modern English peep, with roughly the same meaning, is problematic because it runs afoul of regularity of sound change; rather, the Modern English verb may well have been created independently, most likely as an onomatope, and thus may not truly be in a direct lineal descent connection with earlier pi:pen. This same concern arises with grammatical forms, even those generally talked about with regard to grammaticalization. Murtoff (1999) provides considerable detail on the development of the Spanish adverbial marker -mente, often cited21 as a classic case of grammaticalization in that a Latin lexical item – the ablative singular of ment- “mind” – has come to serve a purely grammatical function. She shows that there really cannot be a straight line of descent from lexical item in a phrase, e.g. Latin clara mente “with a clear mind” to word-level marker, e.g. Spanish claramente “clearly”.22 In particular, these formations show developments in Medieval Spanish, involving variants with mientre and miente, that suggest that the introduction of the specific form mente (which, importantly, does not show the regular sound change development of medial -e-) into later Spanish adverbials represents a learnèd borrowing and not a direct development out of a Latin antecedent. Borrowing therefore interrupted the ‘pathway’ of development and this additional information suggests that this case is not, in fact, as straightforward as
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.11 (55)
Rescuing traditional (historical) linguistics
it first seems, and significantly is not a matter of direct lineal descent from Latin ADJ + mente to modern Spanish ADJ + mente. In this case, only a more detailed look at the relevant historical documentation brings to light the separation of the seeming Latin predecessor from the apparent later Spanish instantiation, meaning that caution is called for when a full historical record is not available. The same kind of caveat holds of course when the connections are made between elements that are in the same synchronic stage that differ in some respect from one another, as, for instance, if someone were to connect the English future marker will with the noun will meaning ‘volitional force’ and draw conclusions from that linkage about the possible development of a future marker out of a volitional element, or, to take a case from the existing literature, when Heine and Reh (1984: 120) argue for a development from an adverb meaning ‘then’ to a future marker in Bari based on the homophony between the two forms (dé in each case). While all practicing historical linguists do exactly this sort of hypothesis-formation in their work, and oftentimes the connections and the inferences are perfectly reasonable (as in both examples just given), without the establishment of historical or synchronic connections but rather with just an assumption of them, all one really has is speculation.23 It is essential, therefore, to approach any such connections with a healthy dose of skepticism, especially when they are based on obvious-looking pairs of items, because looks can often be deceiving. This holds whether one is looking across different stages of a language or within a single stage, and anyone interested in the truth about historical developments must constantly keep that in mind. A few examples should make this clear where the most obvious pathway is most likely not the correct one. I present each of these, which to my knowledge have not been discussed previously in the relevant literature, not to counter some specific extant account of them, but to point out the perils of jumping to conclusions, as they offer the kind of situation in which one could easily be led to the wrong results by not exercising suitable caution.24 For instance, in contemporary American English there is a noun ultimate that refers to a game known also as Ultimate Frisbee (an extreme form of sport involving a flying disc called a Frisbee that is very popular in the US now). The word ultimate has been in use in English for some time as a noun meaning ‘final point, conclusion’, so it is fair to ask if the new usage meaning ‘Ultimate Frisbee’ is perhaps simply an extension of the already existing noun, the result of a semantic shift. Most likely, however, this is not the case, but rather ultimate in the new sense seems to result from a ‘clipping’ (a type of change sometimes also called ‘beheading’ or simply ‘ellipsis’) and reanalysis of the noun phrase ultimate frisbee , as schematized in (3): (3) [[ultimate]ADJ [frisbee]N ]NP → [[ultimate]ADJ [Ø]N ]NP → [[ultimate]N ]NP
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.12 (56)
Brian D. Joseph
Thus linking the two meanings of the noun ultimate directly involves actually falling into the trap of treating a diachronic correspondence (the result of various processes/mechanisms of change) as being a diachronic process/mechanism; the link between the two is indirect (non-lineal) at best, and what is really crucial is the mediation of the NP containing the adjective ultimate. A similar case is the Contemporary American English slang use of Mc- as a prefix in words like McPaper, as a derogatory characterization of the newspaper USA Today, or McJob, for a low-paying job with no future, based on the corporate name McDonald’s and accompanying connotations of less than satisfactory quality, and of products that are fast but not satisfying, ephemeral, with high turnover, etc. Mc- in McDonald’s derives from Proto-Celtic *mac- “son” (cf. Old Irish macc), but one must ask whether it would make sense here to say that Mc- of Mcpaper is just a case of grammaticalization of a word (early Celtic mac) to an affix, a change of a generally somewhat familiar type? I would say not, since what should be at issue here is the determination of the proximate source of the Mc- prefix, not its ultimate source.25 More particularly, the creation of McPaper is rather like a case of truncation than any direct ‘grammaticalization’ per se; that is, it seems to represent an extreme reduction of an associative phrase ‘Newspaper that is to other (real) newspapers as McDonald’s is to (real) food’, as if from Mc(Donald’s-esque) paper. Such developments can give the appearance of direct lineal descent and thus also of ‘grammaticalization’, but there is no grammar-creating process at work here per se, only an ordinary and well-recognized type of change that in its outcome gives a form that looks as if it could be related directly to a piece of an existing word. A similar concern – thus giving a reason not to simply accept ‘obvious’ connections at face value – arises with what we might call ‘false’ positives etymologically speaking, as with ear (of corn) and ear (the body part) in English. These two words are etymologically and thus historically distinct, reflecting Germanic *ahuz and *awz- (PIE *ak’-u- / *ows-) respectively. Nevertheless, they can be plausibly connected synchronically, inasmuch as both are types of appendages, and one might be tempted to draw historical inferences from such a connection, looking to metaphor or the like as the mechanism of semantic change. Given what is known about the etymology, however, caution about leaping to metaphor-driven conclusions here seems essential. Such caution is also needed for grammatical material. The strong plural pronoun them in modern English has a corresponding weak form ‘em. An apparent synchronic connection can be made, as the weak form seemingly derives via reduction from the strong form, not unlike the relationship seen in the masculine singular ‘im versus the strong form him. It turns out, though, that such a connection and the accompanying inference of change are counterfactual: ‘em is the inherited form (cf. OE oblique him) while them shows effects of Norse contact
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.13 (57)
Rescuing traditional (historical) linguistics
(borrowed, replacing OE hie). Thus the ‘reduction’ process, while plausible synchronically for relating them/’em, has no historical basis, and indeed the borrowing account allows one to avoid the embarrassing stipulation of getting rid via ‘reduction’ of a relatively robust consonant like the initial th- (whereas reducing him to ‘em involves only the loss of a relatively weakly articulated and acoustically depleted h). Another such case involves two Modern Persian morphemes: be “to” marking indirect object and direction towards with nouns, and be marking subjunctive mood on verbs. These can be plausibly connected in much the same way that English to is polyfunctional (cf. go to school, live to fight again, etc.), especially if one takes subjunctive as an extension from subordination marking in general – in somewhat archaic usage, be is an infinitival marker too – and draws as well on claims in which infinitives are derived from purpose expressions, themselves deriving from directionality marking (as with English to). However, etymologically these two be’s are distinct, the nominal one deriving from an adposition as seen in Old Persian patiy/Avestan paiti, and the verbal one deriving from a verbal prefix as seen in Avestan vi- (cf. Joseph 1975 for discussion and references); among other things, these etymologies give the modern nominal vs. verbal usage directly, without any further assumptions about cross-categorial generalizations or alterations of distribution. Thus caution is certainly in order, but one can ask whether, more generally, it is the case that any forms that can be connected should be connected in order to see what historical inferences are possible. Admittedly, this is how etymological conjectures are made and oftentimes eventually proven (to the extent that any etymology can be considered proven) but it is also the case that ‘obvious’ connections often prove to be wrong, as the above cases show, whereas ‘true’ etymologies are often anything but obvious. For instance, within English, from the synchronic evidence of the modern language only, it would not be at all obvious that the -s of yes and the verbal form is are connected, yet in fact they are, the former being from the optative of *(H)es- “be” and the latter also from *(H)es-. Another such case is the reflexes of earlier English ri:c- “realm”, found today only in the -ric of bishopric “bishop’s realm” and in the -ritch of eldritch “other-worldly”, two pieces that are probably not connected synchronically by any speakers of the language even if they happen to know the rare word eldritch. Such caution is necessary of course when looking for connections across languages, as in the case of the set of Albanian dirsë, Armenian khirtn, and Greek hidro:s, all cognate with (and translation equivalents of) English sweat (from Proto-Indo-European *swid-ro-, under the most widely accepted formulation of PIE phonology) or the well-known example of Armenian erku, Sanskrit dva:, and Latin duo, cognate with (and translationally equivalent to) English two (from Proto-Indo-European *dwo:)
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.14 (58)
Brian D. Joseph
In fact, based on these examples, which are quite illustrative of a general problem in working out language history, it would seem that a useful rule of thumb in making historical connections perhaps is that the less obvious the connection the better. In a sense, in the less obvious cases, one has to work harder to be sure the connection is valid, and that additional work is what pays off in a more secure etymology. Too often, though, it seems that in applying the tenets of grammaticalization, linguists just ‘eyeball’ forms and go with what looks obvious; a goodly number of the cases catalogued in Heine and Kuteva (2002: passim) are of that sort – all reasonable hypotheses to be sure but based often on little more than similarity of form and a possibility of a linkage.26
. Unidirectionality and traditional historical linguistics A key tenet of grammaticalization theory is the principle of unidirectionality, the claim that movement involving grammatical elements is never in the direction of a given element developing a less grammatical status, e.g. never, on the assumption that affixes are always more grammatical in a sense than free words, from an (bound) affixal element to an independent word-level unit. This principle has guided much of the research within the framework of grammaticalization. It differentiates work in grammaticalization from earlier work in historical linguistics, and thus it bears on the general question of the relationship between grammaticalization and more traditional approaches. Actually, though, the claim inherent in this principle can be taken in at least two related but slightly different ways, as sketched in (4): (4) Two ways (at least) of interpreting notion of unidirectionality: UNIDIRECTIONALITY-A = there exists only movement towards greater grammatical status UNIDIRECTIONALITY-B = there exists no movement towards less grammatical status In each case, the claim would be that one never finds a more grammatical element becoming less grammatical in nature, but Unidirectionality-A makes the stronger claim that there can be no ‘lateral’ moves, ones in which the degree of grammatical status does not change even though the element changes, whereas Unidirectionality-B permits lateral moves. Unidirectionality-B, being the weaker claim, is subsumed under the A interpretation, and so it could be true even if A is not. Related to these claims is what Hopper (1994) has referred to as “phonogenesis”, a way in which elements gain phonological ‘bulk’ through the ‘erosion’ and ‘semantic bleaching’ of elements that at one time were full-fledged morphemes;
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.15 (59)
Rescuing traditional (historical) linguistics
indeed, for Hopper, adding a “functionally empty [set of] phonological elements [is] an advanced stage of grammaticalization” (p. 31).27 The question of the proximate source of the Mc- in McPaper mentioned earlier is relevant here since if it is from the Mc- that occurs in McDonald’s, as seems to be the case, it represents a case of an affix deriving from another affix (if the Mc- of McDonald’s is synchronically analyzable as an affix) or from an arbitrary piece of a word (if McDonald’s as a proper name is treated as being monomorphemic28 ). In the first case, it would be a lateral move, and thus a counter-example to Unidirectionality-A, and in the second case, it would be an instance of ‘counter-erosion’ in that an empty string of segments would have taken on grammatical status as an affix. In this latter case, an element would have added to its ‘bulk’ moreover but not via the devolution of a word to affixal status, contrary to Hopper’s claims. Such lateral moves have long been recognized within historical linguistics, so it does not take the recent case of Mc- to disprove Unidirectionality-A; for instance, Latin offers the example of the denominal adjectival suffix -nus becoming -":nus via the resegmentation (or reanalysis) of forms such as Rom":-nus “pertaining to Rome” as Rom-":nus, and numerous others can be found.29 Moreover, it has long been recognized that affixes can be stable over thousands of years; Fortescue (2003), for instance, states that evidentiality affixes in present-day Eskimo languages are to be reconstructed as suffixes in Proto-Eskimo, over a span of some three thousand years, and the 1SG -m of many modern Slavic languages (e.g. Macedonian) has been an affix since Proto-Indo-European (cf. Janda 1996 on this affix). If affixes can show long-term stability, and can move laterally, as it were, how can we be confident that a given affix must have been a lexical item in some stage prior to the period of stability or prior to a lateral move? Why not treat the default position here as being that the element in question has been an affix for as long as one can tell? Thus, traditional historical linguistics is really silent on the issue of unidirectionality and in essence has let the chips fall where they may; if an affix is the most reasonable source of another affix, so be it. If an affix is created out of originally non-affixal material, so be it. If an affix is the starting point for the creation of a word,30 so be it. While the lack of any constraints here might seem to some to be problematic, especially as opposed to the constrained system that grammaticalization offers (only movement from less grammatical to more grammatical), one answer would be that the traditional approach, while less constrained, is more realistic, in that there are cases of counter-directionality, lateral movement, and movement from empty phonological strings to meaningful affixal status. Under this view, then, it is not clear that grammaticalization offers much in the way of an advance over traditional viewpoints.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.16 (60)
Brian D. Joseph
. Grammaticalization as asynchronic I have argued here that in a sense, grammaticalization proponents, while purporting to be interested in the history of particular forms and constructions, actually in practice are often ahistorical, not paying enough attention to the full range of details of the development and unfolding of a form through time and not paying careful enough attention to matters of etymology. At the same time, though, there are some indications that advocates of grammaticalization are somewhat ‘asynchronic’ as well, in that they do not pay as much attention to details of synchronic analysis as might be desirable too and in a sense eschew definitive accounts of the synchronic accounts of given elements. For instance, much has been said about the way in which the form to, universally assumed to have been nothing more than a preposition in earlier stages of English, e.g. Old English, came to serve a grammatical function in marking an infinitive in Modern English, as in I live to do linguistics or I yearn to study historical linguistics. And, in terms of what it means to be a marker of the infinitive, many studies (see, e.g., Fischer 2000) simply (but quite reasonably) follow the characterization given by Jespersen (1927) that to in this use is “a mere empty” marker. But, being able to chart the development of a grammatical morpheme presupposes that we have a clear idea of just what the element was to start with and what it has ended up as. Thus, it is fair to ask what being “a mere empty” marker actually means in terms of a formal characterization of infinitival to in an explicit grammatical account. For instance, Pullum (1982: 181) argued that infinitival to is “not an affix, particle, complementizer, preposition, . . . or tense morpheme . . . [but rather is a member of] the category V . . . [and specifically] is an auxiliary verb and head of a verb phrase”; Pullum has since given up that specific claim (see Huddleston & Pullum 2001: 1185–1186) in favor of treating it as a subordinator, but if the verbal analysis were to be maintained,31 then the view of the historical development of to would look quite different, and would involve a drastic reanalysis of the element from a non-verb to a verb. Admittedly one could just focus on the function and say that whatever infinitival to is in Modern English from a formal structural standpoint, its function is as a mere marker of infinitival status. However, since at least one thread within grammaticalization studies has always been on such issues as categorial status (especially, e.g., word vs. affix), it seems that it is indeed important to ask about the categorial status of to and make an explicit statement of that a part of the diachronic account. To fail to do so leads to claims and accounts that I would call ‘asynchronic’.32 My plea here, therefore, would be that before claims are made about diachronic developments, it behooves us to pay serious attention to what the best synchronic analysis is at the various stages being examined. There is no doubt in my mind that early English to was a preposition and that it ultimately gave rise to the infiniti-
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.17 (61)
Rescuing traditional (historical) linguistics
val marker to (whatever its formal charactierization is to be) in Modern English. However, if the modern to is a verb, then is this an instance of a ‘universal pathway’ of grammaticalization or simply a reanalysis – admittedly perhaps an unusual one but a reanalysis nonetheless – brought on by the role that the element plays in the overall system, that is, in comparison with other combinations of verbal elements such as must/may/can + VERB, as Pullum originally suggested?
. Conclusion – being fair to grammaticalization but at the same time (and overridingly) being fair to language history I have been somewhat critical of grammaticalization in my remarks so far, but so as to be fair, I would like in this conclusion to be absolutely clear about my position here. In particular, I do not for a moment deny that there is a ‘phenomenon’ of grammaticalization; as I have argued elsewhere (Joseph 2001a, 2003), however, I see it as a result, not a process, an epiphenomenon perhaps. I would say that we could just do with grammaticalization as a result, a product, and reserve the designation ‘process’ or ‘mechanism’ for the traditionally recognized sound change, analogy, reanalysis, and metaphorical extension; moreover, once social dimensions are taken into consideration (e.g. in contact situations, whether between dialects or languages), then borrowing (and all this last entails, like calquing) and hypercorrection also need to be added in, the latter being a powerful process/mechanism in change due to dialect contact. In this regard, grammaticalization theory adds little to the insights of traditional historical linguistics despite purporting to offer a new way of looking at data concerning grammatical forms. Still, one thing that grammaticalization definitely has gotten right in recent years is the emphasis on constructions and on forms in actual use, and not in the abstract. That is, it has been realized that it is not enough to simply say, for instance, that a body part has become a preposition (e.g. head → on-top-of) but rather one must recognize that it is head in a particular collocation, e.g. at-the-head-of that has yielded a preposition, or that have turning into exist is not necessarily just a random semantic shift but rather is one that happens in the context of adverbials such as in Late Latin habet ibi → Spanish hay (Fr. Il y a) or that the shift of since from temporal to causal meaning came from in contexts in which causality can be safely inferred from temporal priority, and so on. This is a big step forward, since it takes semantic change especially out of the realm of the purely lexical and places it into the pragmatic domain, deriving changes from inferencing and the like that are possible for words in constructions with other words and in actual, contextually keyed usage.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.18 (62)
Brian D. Joseph
What is particular interesting about this view, in the light of the concern herein for the relation between traditional historical linguistics and grammaticalization, is that it accords with an insight that the Neogrammarians had concerning sound change. Even though sound change is often talked about as happening to individual sounds, in fact, the sounds that undergo changes are always embedded in morphemes and words and even intonational phrases and utterances. In order to understand sound developments oftentimes one must look beyond the immediately adjacent sounds in the same morpheme or word and look instead to the broader surrounding context, including constructional aspects such as prosody or adjacent phrasal elements. For instance, the seemingly unusual voicing of voiceless stops that occurred in word-final position between PIE and Latin, as e.g. in PIE *Hs-ye:-t “it would be” → (Old) Latin sied, was understood by the Neogrammarians not in terms of word-final position, i.e. triggered by a word-boundary following the *-t, but rather in terms of the possibility of voiced sounds following it; that is, it actually involved an original phrasal sandhi (combinatory) process of -t # [+voice] → -d # [+voice] (and, by the same token, -t # [–voice] staying as such) with the ultimate generalization of one of the variants.33 This view of sound change thus involves essentially a construction-based perspective over what can affect individual sounds, and in that way is rather similar to what grammaticalizationists have been doing with regard to the emergence of grammatical elements. In this way, grammaticalization theory and practice converge with Neogrammarian practice in recognizing that the key to understanding language change is not to look at elements atomistically but to see them in their connections with other elements in actual use. Further, though, as noted in Section 4, grammaticalizationists part company with traditional historical linguistics with regard to ‘unidirectionality’. My own position here is that while at best any controversy concerning unidirectionality may reflect nothing more than a definitional matter (see Janda 2001) and at worst the notion might be just completely wrong, I do not and cannot deny that movement from less grammatical items such as words to more grammatical items such as affixes is highly common and widespread; it is just I do not see this movement as an absolute and therefore feel that a more realistic approach has to allow for movement in different directions other than the direction of greater grammatical status. That is, for me, ‘counter-directionality’ (the inverse of ‘unidirectionality’) is not only allowed but actually occurs. A question that needs to be asked of grammaticalization adherents is the following: if unidirectionality is the bedrock principle of grammaticalization theory that many proponents make it out to be, then how can it permit even any exceptions without vitiating the theory?34 In this regard, it is important to note how ‘unidirectionality’ fits in with other models of change and especially what is recognized in traditional approaches. In particular, there are documentable cases where over a long stretch there is ‘mono-
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.19 (63)
Rescuing traditional (historical) linguistics
tonicity’, in the sense that there is a steady slope to the movement overall, but not in the short run, in that occasional ‘blips’ get in the way of there being movement only in one direction.35 A case in point is the Greek loss of the infinitive, as described in Joseph (1978/1990, 1983). While the abundance of infinitival usage and forms in Ancient Greek is not lost altogether until after Medieval Greek of c. 1500AD, there is clear reduction in use and in forms from century to century, beginning with the period of the Hellenistic Koine, with the category eventually disappearing altogether. Importantly, though, the loss was not a steady-state (monotonic) decline, for there were a few renascences of infinitives functionally and formally even as the infinitive was being replaced all over, i.e. at a stage when the category was rapidly sinking into moribundity. These counter-directional developments include the creation of a new infinitive for the verb ‘be’ (eîsthai, as opposed to Ancient Greek eînai) that arose in the 12th century, and three new uses for remaining infinitives that emerged in the 12th–15th centuries: (i) Circumstantial Infinitive (the infinitive as an absolute clausal modifier setting the circumstances for the action of the main verb) (ii) a future formation with thélo: “want”, and (iii) a perfect formation with ékho: “have”. Rather than being troublesome, it seems that accepting multi-directionality/ counter-directionality in change in general36 is a far more realistic view, as this seems simply to be a key characteristic of change in language. In analogy, for instance, it is quite common to find levelling in different directions (cf., Tiersma 1978) as is the case in Modern English where the irregular plural wharves (to singular wharf ) is being replaced for many speakers by the regular wharfs while at the same time the (now) regular plural37 dwarfs (to singular dwarf ) is being replaced by the irregular dwarves. In sound change, too, one finds cases in which perception is the determining factor and cases in which production governs the outcome – both forces/directions must therefore be taken into account since each can govern how a change ends up. Thus a reasonable question, it would seem, is why grammaticalization should be any different. Thus, as a final plea, I suggest that the following is what we should aim for in doing historical linguistics, within any framework: (5) a.
the best analysis of the starting point of developments in question and the best analysis of the end point of developments of interest to us; only then can we really know what changes have occurred b. the best account of the transitions between the points we (arbitrarily) select for our analysis, keeping in mind the axiom that as linguists trying to understand change as something that speakers do (not something that happens to a language), we should not take a perspective on language change which a speaker cannot take (thus no trans-generational ‘diachronic processes’ – cf. Janda 2001)
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.20 (64)
Brian D. Joseph
c.
the best coverage of the relevant historical facts, so as not to fall into the false-positive and non-direct-lineal descent trap.
Only with these aims and only when they are achieved, I would argue, can we ever hope to have a truly explanatory ‘theory’ of language change.38 And, to a large extent, these are exactly the goals and methods that traditional historical linguistics have long offered the discipline.
Notes * I would like to thank Hope Dawson, Olga Fischer, Bernd Heine, and Rich Janda for their insightful and helpful comments on an earlier version of this paper. Needless to say, they are not to be held responsible for the views expressed herein. . See Campbell and Janda (2001) for a compendium of various definitions of grammaticalization to be found in the literature; they conclude their survey by saying that “we are left with a notion of grammaticalization which minimally includes, at its core, some linguistic element > some more grammatical element” (107). . Many of the edited volumes noted herein stemmed from organized conferences, and indeed, this paper has its origins in a presentation given at a conference, held in April 2002 at the Universiteit van Amsterdam. . For instance, Bowden 1992 is listed under “Oceanic Languages – Grammaticalization”. . These searches were done on January 11, 2003. . I mention this without taking sides on whether the label of ‘theory’ is justified or not; see Newmeyer (1998: Ch. 5; 2001) for some argumentation against the use of this label and Heine and Kuteva 2002 for some counter-discussion. . Note also Ramat (1998: 123), who states that “grammaticalization theory [without a principle of] ‘unidirectionality’ would be left with a too vague definition of its field, including almost every instance of change”; as discussed below, unidirectionality probably should be given up in at least some form, thus leaving grammaticalization in the position Ramat cautions against. . In fairness, I point out that Heine (2003: 575) has explicitly stated that “grammaticalization theory is neither a theory of language nor of language change”; thus not all practitioners take an expanded view of what grammaticalization theory is and what it can do. The variety of viewpoints expressed by those who work within a grammaticalization framework admittedly makes it hard to know which to take as representative. By giving some actual statements that have been expressed in print, I am trying to at least document the viewpoints I discuss, though it should be clear that not all viewpoints are held by all. . Without naming names, I note that I count myself among this group, and would include as well the historical linguists represented in Campbell (2001a), among many others.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.21 (65)
Rescuing traditional (historical) linguistics . As noted at the outset, there are pre-contemporary-era antecedents to grammaticalization studies, but the recognition and formulation of governing principles, as in Lehmann (1982/1995) or Hopper (1991), are only fairly recent phenomena. . No one working in the grammaticalization paradigm has actually said anything to this effect, but paradigm shifts are often accompanied by revisionist views of earlier work, so this question is a reasonable one to ponder. . Just by way of comparison, a Google search for ‘Neogrammarian’ yielded 573 hits. . Admittedly, the Comparative Method is illustrated in most textbooks with phonological examples and application of the method in domains other than phonology generally gets short shrift, but that is still no reason to act as if the scope of the method must be restricted just to phonology. . Or conceivably *plak-i- or the like; the reconstruction with *-j- for the verb is most assuredly the historically correct one (see Scatton 1993: 194) but the point is that one brings to bear whatever tools one can muster in order to reach reasonable internally arrived at reconstructions. . This point has been made repeatedly by Raimo Anttila; see, for instance, Anttila (1972/1989: Ch. 10, 12, 13, passim) and note his statement (1989: 264) that “internal reconstruction is already known to the reader, as it is exactly the same as [synchronic] morphophonemic analysis”. . As Bernd Heine has pointed out to me, proponents of grammaticalization have not ignored these mechanisms of change, and in fact have incorporated recognition of them into the framework of grammaticalization theory. Still, I would say that the real question that needs to be addressed here is if a notion of grammaticalization – and especially a separate process of change so labelled – is needed on top of these other processes. This was a focal point for papers such as Campbell (2001b), Joseph (2001a), and Newmeyer (2001). . Heine (2003: 585) raises an objection to this point, saying that such a view misses the fact that the Greek development involves “the same lexical verb (‘want’) and the same structural characteristics as observed in Swahili and many other languages” and fails to answer the questions of why the development “necessarily [led] from lexical verb to tense prefix; that is, why . . . it [is] unlikely that the process could have proceeded in the opposite direction from verbal prefix to lexical verb”. My response here is that since ‘want’ is not the only verb to do this, even though it is a common one, it is not clear that any theory of internally driven change can explain why ‘want’ is selected in a given language and not some other verb. And, as to the latter objection, the matter of ‘likelihood’ of directionality seems to be assuming a status for unidirectionality that is not warranted (see discussion in Section 4); in a sense, anyway, it is irrelevant as far as any single example is concerned as any one change can only go in one direction at a time (see Note 36). . See Pappas and Joseph (2002) for details and discussion. . Note that uis in quiuis is the second person singular present indicative of ‘want’, and libet is the third person singular present indicative of ‘please’ (thus the forms literally are ‘who you want’, ‘who (it) pleases (you)’). . See, for instance, the articles and discussion in Bybee and Hopper (2001).
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.22 (66)
Brian D. Joseph . Bernd Heine (p.c.) has raised a question here as to why grammatical forms typically are shorter than lexical forms – this observation, I would say, is easily derivable from the typically low prosodic prominence found with grammatical items and the effects, such as assimilations, syncopes, and various reductions, that are often associated with that prosodic characteristic. . Indeed, I have cited this example myself by way of illustrating the phenomenon, e.g. in Joseph (2001b). . It may well be that in Spanish, -mente forms are best treated as compounds and not simple stem-plus-affix word-level units, due to facts of conjunction reduction (as in rapida y claramente ‘rapidly and clearly’) for instance, but the finer details of the analysis in modern Spanish, while interesting and important in their own right, are beside the point here. . Part of what is at work here in such speculation is an assumption that the principle of unidirectionality is correct and that given a relation between ‘then’ and ‘future’ the pathway could only be that of ‘then’ developing into a future marker and not the reverse. However, if the principle of unidirectionality is too strong (see discussion below), then (among other things) its value in determining the historical sequencing in such cases is severely diminished. . My impression, moreover, is that throughout much of the grammaticalization literature, one can find far too many instances of just these types presented as if they were robustly established, though I readily acknowledge that quantifying this impression for the purposes of an explicit comparison with work done in other frameworks would be tricky at best. . It can be noted as a sidebar that the proximate source of the Mc- affix in McPaper/McJob is an element that is at best another affix (if McDonald, McPherson etc. allow for identification in English of a prefix Mc-) but could just as well be a meaningless piece of a word (since it is part of a proper name and thus not really analyzable in the usual sense that lexical items can be). This would thus run counter to the claim made in Hopper and Traugott (1993: 128– 129) that all grammatical elements, e.g. affixes, have a “prior lexical history”; admittedly, if one goes back far enough, Mc- does have a lexical source but its most immediate source is not a lexical item at all but either an affix or just a string of segments in a lexical item. . It is instructive to note the somewhat similar problems associated with an ‘eyeballingthe-obvious’ methodology in comparative linguistics, as exemplified by the multilateral comparison strategy espoused by Greenberg (1987) and claims of relatedness derived therefrom; see Campbell (1988), Greenberg (1989), Matisoff (1990) for some discussion of the issues. I reiterate here that the problems I note are not necessarily found only in the grammaticalization literature, and in many cases, practitioners are simply trying to use any available evidence to see what historical inferences are possible; however, that does not mean that putative connections must be taken to be established, hence my plea for caution here. . Though see Joseph (2003: 477, 490, n. 21) for some critical discussion concerning the notion of ‘phonogenesis’. . As pointed out in Note 25, the status of Mc- in names like McDonald or McPherson in modern English is tricky and I offer no definitive judgment on it here, though either analysis would be problematic for some version of unidirectionality.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.23 (67)
Rescuing traditional (historical) linguistics . For instance, the prehistory of the -ness suffix in English, coming from the resegmentation of the piling up of several affixes, in Indo-European terms, *-n-ot-tu- (with *-t-t- giving Germanic -ss-); cf. Gothic ibnassu- ‘evenness’, where the Germanic suffixal complex *-assuwas added to an n-stem form. . See Janda (2001) for a catalogue of literally dozens and dozens of cases of this type in the literature, despite the frequent assertions in the grammaticalization literature that such a development is either impossible or so rare as to be not worthy of notice. . And note that Huddleston and Pullum (2001: 1185) do consider it defensible, saying that “it would not be impossible to maintain . . . [though] the case for to being a VP subordinator is stronger”. . It may well be that Hopper’s (1987) notion of “emergent grammar”, which, as mentioned in the Introduction, has been embraced by many practitioners of grammaticalization, leads some to the position that an explicit synchronic analysis is not necessary, inasmuch as language, in this view, involves only “movements towards structure” not the development of structure per se. . This view is in keeping with the position regarding the nature of sound change expressed in Hock (1976) and Joseph (1999), as noted in Section 2. . See Campbell (2001b: 134ff., and especially 140–141) on this point. . Indeed, ‘monotonicity’ might be a better term than ‘unidirectionality’ for the insight that proponents of grammaticalization are trying to capture. . Individual changes, of course, can only go in one direction; a given change is thus not like Canadian humorist Stephen Leacock’s Lord Ronald, who (in “Nonsense Novels”, 1911) “flung himself upon his horse and rode madly off in all directions”. I am indebted to my mother Harriet Joseph for bringing this quote to my attention. . I say “now regular” as the -v- plural, which seems to be the innovative form here, is a replacement for the -f- plural, which itself seems to be an innovation. That is, the OE plural dweorgas developed regularly into later dwerwes or dwerows, so that the plural in -fs is a replacement, based on the singular dwarf (from OE dweorg) for the plural in -wes or -ows. Neither of those would have yielded a form in -v-, so the -v- plural must be based on the -f-/-v- alternation of forms such as wharf/wharves, the point which is most relevant to the discussion here. . The rub here, of course, is that it is not always obvious what the ‘best’ is in a given case or who is to decide; we must strive to determine it but must always as well be prepared to reevaluate in the light of new data and new perspectives.
References Andersen, Henning (1989). “Understanding Linguistic Innovations”. In Leiv Egil Breivik & Ernst Håkon Jahr (Eds.), Language Change: Contributions to the Study of its Causes (pp. 5–27). Berlin: Mouton de Gruyter.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.24 (68)
Brian D. Joseph
Anttila, Raimo (1972/1989). Historical and Comparative Linguistics. New York: MacMillan. (2nd edn., 1989, Amsterdam: John Benjamins.) Aronoff, Mark (2002). “The Editor’s Report”. Language, 78, 394–397. Bowden, John (1992). Behind the Preposition: Grammaticalisation of Locatives in Oceanic Languages. Canberra: Dept. of Linguistics, Research School of Pacific Studies, Australian National University. Bybee, Joan & Paul J. Hopper (Eds.). (2001). Frequency and the Emergence of Linguistic Structure. Amsterdam: John Benjamins. Campbell, Lyle (1988). Review article on Greenberg 1987. Language, 64, 591–615. Campbell, Lyle (Ed.). (2001a). Grammaticalization: A Critical Assessment (Special Issue of Language Sciences, Vol. 23, 2–3). London: Pergamon. Campbell, Lyle (2001b). “What’s Wrong with Grammaticalization?” Language Sciences, 23, 113–161. Campbell, Lyle & Richard D. Janda (2001). “Introduction: Conceptions of Grammaticalization and their Problems”. Language Sciences, 23, 93–112. Condillac, Etienne Bonnot de (1746). “Essai sur l’origine des connaissances humaines”. Paris. Diewald, Gabrielle (1997). Grammatikalisierung: Eine Einführung in Sein und Werden grammatischer Formen. Tübingen: Niemeyer. Dürr, Michael (1988). “Reference to Space in Colonial Quiché”. Función, 8, 47–78. Fischer, Olga (2000). “Grammaticalisation: Unidirectional, Non-Reversable? The Case of to before the Infinitive in English”. In Olga Fischer, Anette Rosenbach, & Dieter Stein (Eds.), Pathways of Change. Grammaticalization in English (pp. 149–169). Amsterdam: John Benjamins. Fortescue, Michael (2003). “Evidentiality in West Greenlandic: A Case of Scattered Coding”. In Alexandra Aikhenvald & R. M. W. Dixon (Eds.), Studies in Evidentiality (pp. 291– 306). Amsterdam: John Benjamins. Givón, Talmy (1971). “Historical Syntax and Synchronic Morphology: An Archaeologist’s Field Trip”. CLS, 7, no. 1, 394–415. Greenberg, Joseph (1987). Language in the Americas. Stanford: Stanford University Press. Greenberg, Joseph (1989). “Classification of the American Indian Languages: A Reply to Campbell”. Language, 65, 107–114. Heine, Bernd (1993). Auxiliaries: Cognitive Forces and Grammaticalization. New York: Oxford University Press. Heine, Bernd (1997). Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd (2003). “Grammaticalization”. In Joseph & Janda (Eds.), 575–601. Heine, Bernd & Mechthild Reh (1984). Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske Verlag. Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer (1991). Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Heine, Bernd & Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hock, Hans Henrich (1976). Review article on Anttila (1972). Language, 52, 202–220. Hopper, Paul J. (1987). “Emergent Grammar”. BLS, 13, 139–157.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.25 (69)
Rescuing traditional (historical) linguistics
Hopper, Paul J. (1991). “On Some Principles of Grammaticalization”. In Traugott & Heine (Eds.), Vol. 1. (pp. 17–35). Hopper, Paul J. (1994). “Phonogenesis”. In Pagliuca (Ed.), 29–45. Hopper, Paul J. & Elizabeth C. Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Horne Tooke, John (1857). Epea pteroenta or the diversions of Purley. 2 Vols. London (first published 1786). Huddleston, Rodney & Geoffrey K. Pullum (2001). The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Janda, Laura (1996). Back from the Brink: A Study of How Relic Forms in Languages Serve as Source Material for Analogical Extension. Munich/Newcastle: LINCOM EUROPA. Janda, Richard D. (2001). “Beyond ‘Pathways’ and ‘Unidirectionality’: On the Discontinuity of Language Transmission and the Reversibility of Grammaticalization”. Language Sciences, 23, 265–340. Janda, Richard D. & Brian D. Joseph (2003). “Reconsidering the Canons of Sound Change: Towards a Big Bang Theory” (Paper presented at 15th International Conference on Historical Linguistics, Melbourne 2001). To appear in Barry Blake & Kate Burridge (Eds.), Selected Papers from the ICHL 15. Amsterdam: John Benjamins. Jespersen, Otto (1927). A Modern English Grammar on Historical Principles. Part III. Copenhagen: Munksgaard. Joseph, Brian D. (1975). “Case Marking and Complementizers in Persian”. Stanford Working Papers on Language Universals, 17, 141–144. Joseph, Brian D. (1978/1990). Morphology and Universals in Syntactic Change: Evidence from Medieval and Modern Greek. New York: Garland Publishers (Outstanding Dissertations in Linguistics Series; expanded and updated version of 1978 Harvard University Ph.D. Dissertation). Joseph, Brian D. (1983). The Synchrony and Diachrony of the Balkan Infinitive: A Study in Areal, General, and Historical Linguistics. Cambridge: Cambridge University Press. Joseph, Brian D. (1999). “Utterance-Finality: Framing the Issues”. In Bohumil Palek, Osamu Fujimura, & Brian Joseph (Eds.), Proceedings of LP ’98 (4th Linguistics and Phonetics Conference), Vol. 2 (pp. 3–13). Prague: Charles University Press. Joseph, Brian D. (2001a). “Is There Such a Thing as ‘Grammaticalization’?” Language Sciences, 23, 163–186. Joseph, Brian D. (2001b). “Historical Linguistics”. In Mark Aronoff & Janie Rees-Miller (Eds.), Handbook of Linguistics (pp. 105–129). Oxford: Blackwell. Joseph, Brian D. (2003). “Morphologization from Syntax”. In Joseph & Janda (Eds.), 472– 492. Joseph, Brian D. & Richard D. Janda (Eds.). (2003). Handbook of Historical Linguistics. Oxford: Blackwell. Katz, Aya (1998). “Recycled Morphemes and Grammaticalization: The Hebrew Copula and Pronoun. Southwest Journal of Linguistics, 17, 59–97. Kuhn, Adalbert (1853). “Ueber die durch nasale erweiterte verbalstämme”. Zeitschrift für Vergleichende Sprachforschung, 2, 455–471. Kuryłowicz, Jerzy (1964). The Inflectional Categories of Indo-European. Heidelberg: Winter.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.26 (70)
Brian D. Joseph
Kuteva, Tania (2001). Auxiliation: An Enquiry into the Nature of Grammaticalization. Oxford: Oxford University Press. Labov, William (1981). “Resolving the Neogrammarian Controversy”. Language, 57, 267– 309. Labov, William (1994). Principles of Linguistic Change. Volume I: Internal Factors. Oxford: Blackwell. Lehmann, Christian (1982/1995). Thoughts on Grammaticalization. München: LINCOM Europa. [Originally published as Thoughts on grammaticalization: a programmatic sketch. V. 1. Köln: Institut für Sprachwissenschaft der Universität, 1982. (Arbeiten des Kölner Universalienprojekts; 48).] Lessau, Donald A. (1994). A Dictionary of Grammaticalization (3 volumes). Bochum: Universitätsverlag Dr. N. Brockmeyer. Matisoff, James (1990). “On Megalocomparison”. Language, 66, 106–120. Meillet, Antoine (1912). “L’évolution des formes grammaticales”. Scientia (Rivista di Scienze), 12, no. 26, 6. Murtoff, Jennifer (1999). “Spanish Adverbs in -mente”. Paper presented in Diachronic Morphology (Linguistics 802), The Ohio State University, Spring 1999. Newmeyer, Frederick J. (1998). Language Form and Language Function. Cambridge: MIT Press. Newmeyer, Frederick J. (2001). “Deconstructing Grammaticalization”. Language Sciences, 23, 187–229. Pagliuca, William (Ed.). (1994). Perspectives on Grammaticalization. Amsterdam: John Benjamins. Pappas, Panayiotis A. & Brian D. Joseph (2002). “On Some Recent Views Concerning the Development of the Greek Future System”. Byzantine and Modern Greek Studies, 26, 247–273. Pullum, Geoffrey K. (1982). “Syncategorematicity and English Infinitival to”. Glossa, 16, 181–215 Ramat, Anna Giacalone (1998). “Testing the Boundaries of Grammaticalization”. In Ramat & Hopper (Eds., 1998b), 107–127. Ramat, Anna Giacalone & Paul J. Hopper (1998a). “Introduction”. Ramat & Hopper (Eds., 1998b), 1–11. Ramat, Anna Giacalone & Paul J. Hopper (Eds.). (1998b). The Limits of Grammaticalization. Amsterdam: John Benjamins. Rosén, Hannah (2002). “Grammaticalization in Latin?” Glotta, 76, 94–112. Saussure, Ferdinand de (1878). Memoire sur le système primitif des voyelles en indo-européen. Recueil des publications scientifiques de F. de Saussure. Geneva: Droz. Scatton, Ernest (1993). “Bulgarian”. In Bernard Comrie & Greville Corbett (Eds.), The Slavonic Languages (pp. 188–248). London: Routledge. Tiersma, Peter (1978). “Bidirectional Leveling as Evidence for Relational Rules”. Lingua, 45, 65–77. Traugott, Elizabeth C. & Bernd Heine (1991a). “Introduction”. In Traugott & Heine (Eds., 1991), Vol. 1 (pp. 1–14). Traugott, Elizabeth C. & Bernd Heine (Eds.). (1991b). Approaches to Grammaticalization (2 Vols.). Amsterdam: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 15:47
F: TSL5903.tex / p.27 (71)
Rescuing traditional (historical) linguistics
Watkins, Calvert (1987). “In the Interstices of Procedure: Indo-European Legal Language and Comparative Law”. In Wolfgang Meid (Ed.), Studien zum Indogermanischen Wortschatz (pp. 305–314). Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft. Watkins, Calvert (1995). How to Kill a Dragon. Aspects of Indo-European Poetics. Oxford: Oxford University Press. Wischer, Ilse & Gabriele Diewald (Eds.). (2002). New Reflections on Grammaticalization. Amsterdam & Philadelphia: John Benjamins. Ziegeler, Debra (2000). Hypothetical Modality: Grammaticalisation in an L2 Dialect. Amsterdam/Philadelphia: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.1 (73)
The English s-genitive A case of degrammaticalization?* Anette Rosenbach Heinrich-Heine University Düsseldorf
.
Introduction
Cases of (alleged) degrammaticalization are particularly interesting, and in fact, challenging for grammaticalization research, since they are counter-examples to the presumed unidirectionality of grammaticalization processes. Unidirectionality is taken to be the cornerstone of grammaticalization, and all success – and failure – seems to lie on its shoulders. Recently, however, the validity of the unidirectionality hypothesis has been increasingly questioned. Two types of arguments can be distinguished. According to Janda (2001, 2002) it is hard to imagine how unidirectionality should come about when taking the speaker-perspective, since at any point in the history of a language speakers do not have any awareness about the past stages of a form/construction, nor do they know what is the ‘supposed’ pathway a form should take along a grammaticalization cline.1 Second, it is sometimes argued that unidirectionality is not true because counter-examples to it (i.e. cases of degrammaticalization) do exist. Here basically two positions can be discerned: while some linguists maintain that one counter-example is sufficient to prove unidirectionality wrong (cf. e.g. Newmeyer 1998: 263), others argue that, in light of the overwhelming evidence in favour of unidirectionality, the statistically very rare counter-examples reported are to be neglected (cf. e.g. Hopper & Traugott 1993: 126–128; Haspelmath, this volume).2 Whatever the conceptual, theoretical and empirical stance one is inclined to take on this question, cases of (alleged) degrammaticalization are important and can in fact serve a ‘constructive’ function in that they may help sharpen our definition of what prototypical grammaticalization is (cf. e.g. Hopper & Traugott 1993: 126; Norde 2001: 238). In the present paper I will be looking at the historical development of the English possessive ’s (POSS ’s), which is one of the very few cases discussed in the
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.2 (74)
Anette Rosenbach
grammaticalization literature as a genuine case of degrammaticalization. First I will give an overview of the various ways in which the s-genitive has been discussed so far in the literature, and which result in quite different, in fact often opposing interpretations as to the question of degrammaticalization. I will then lay out a sketch of the history of the s-genitive, which will offer yet another perspective on this development. My main argument will be that in order to understand this development (more fully), it is necessary to look at the broader context of this change, in particular, at what else happened in the English NP. Adopting such a broader perspective, I will finally argue that it is questionable that the development of the s-genitive represents a genuine counter-example to the unidirectionality of grammaticalization processes, if we conceive of unidirectionality as a property of change types rather than of single tokens of change.
. The English s-genitive as a case of degrammaticalization . The grammatical status of possessive ’s (POSS ’s) By now, we can encounter a series of accounts and interpretations as to the question whether, and in which respect the development of POSS ’s should indeed be treated as a case of degrammaticalization. Let us first have a brief look at the structural facts. In Old English (OE), -s was clearly an inflectional ending, marking genitive case. There was still a whole genitive paradigm with various inflectional endings (not only -s), according to number, gender, and noun class; it is only at the end of the Middle English (ME) period that -s became generalized as the only genitive ending. In Modern English (ModE), there are, above all, two reasons which make an analysis as an inflectional ending doubtful.3 First, as we can see in (1), POSS ’s does not attach to the head but rather to the end of the whole possessor NP. A particularly ‘creative’ case can be seen in (2), which illustrates how wide the scope of POSS ’s can be in such constructions. Secondly, example (3) shows that POSS ’s can today also attach to just any non-nominal element, in this case a verb. It has therefore been suggested that ModE POSS ’s should be treated as a clitic, i.e. a syntactic element which is phonologically realized on the preceding word (e.g. Carstairs 1987; Bauer 1994 [1988]: 99).4 (1) [the king of England]’s daughter (2) It’s not mine it’s a [person who went to Cambridge and got a first in engineering]’s (3) I walked away with [whoever it was]’s pen
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.3 (75)
The English s-genitive
. Why a case of degrammaticalization (if at all)? – Or: ‘Anything goes’ If we now consider this development of POSS ’s from inflectional affix in Old English to a clitic in Modern English and compare this to the morphosyntactic cline (4) as proposed in Hopper and Traugott (1993: 7), (4) content item > grammatical word > clitic > inflectional affix we can see that this is a development back down this cline and therefore, at least at first glance, qualifies as a case of degrammaticalization (affix > clitic). At second glance, however, things are much more complicated. According to an account made popular by Janda (1980), though in fact going back at least to the end of the 19th century (cf. Jespersen 1894 and 1918; Wyld 1936 [1920]: 314–316), the ModE POSS ’s is derived from the so-called his-genitive (John his book). In Middle English, both the syllabically pronounced -(e/i/y)s and the phonetically reduced (h)is could be homophonous. Due to this homophony, Janda argues, the OE inflectional -(e/i/y)s became reanalysed as his. In a second step then, his become cliticized to ModE POSS ’s (cf. Figure 1 below). In this account, it is the development from inflectional affix to the possessive pronoun his which can be regarded as degrammaticalization (back along the cline in (4) above, from affix to grammatical word); the cliticization of his to ’s is fully in accordance with a typical grammaticalization process.5 Janda’s account is controversial though. In particular, Allen (1997) shows that the ModE POSS ’s cannot be causally derived from such his-genitives. She argues instead that his, at least until the 16th century, has precisely the same distribution as the inflectional -(e/i/y)s and therefore is simply an orthographical variant of it.6 Recently, Janda (2001, 2002) has slightly modified his 1980 account, assuming now that the “ME possessivemarking involved an invariant possessive Det that was homophonous but not identical with the masc[uline] [s[in]g[ular] possessive his” (Janda 2001: 303). In this scenario, the old inflectional genitive was analysed by speakers as a shortened word, and, by means of hypercorrection ‘undone’ into the alleged full form Middle English
early Modern English
early Modern English/ Modern English
-(e)s (inflectional affix)
his
’s (clitic)
Johnys book
John his book
John’s book
degrammaticalization
grammaticalization
Figure 1. The his-genitive as a case of (de)grammaticalization
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.4 (76)
Anette Rosenbach
his. This modification has the advantage that Janda now does not have to assume that the his in the his-genitive resulted from the reanalysed inflectional genitive, which is quite controversial since it presupposes that the his-genitive is a late Middle English innovation.7 As major evidence for the existence of an invariant clitic determiner Janda (2001: 302) gives examples such as (5) in which ys/his attaches to a feminine noun. Interestingly, the very same type of example is taken by Allen (1997) as evidence that at that time ys/his was only an orthographical variant of inflectional -s. (5) Margere ys dowghter
(Cely 188.3; 1482; cf. Allen 1997: 118)
In fact, both interpretations are possible and we cannot decide on the basis of such examples which is the correct one.8 It would be interesting to see some further evidence which could support Janda’s recent (and interesting) claim that a reanalysis of -(e/i/y)s to an invariant clitic his is due to hypercorrection. Although I don’t think that the last word has yet been spoken on this issue (particularly inasmuch as the link between his and the new function of POSS ’s as a possessive determiner is concerned; see also Section 3.3 below), we need to conclude so far that the empirical support for this hypothesis is still too weak to warrant a strong position here.9 Therefore, in the following I will focus on the development ‘affix > clitic’. Assuming such a development, however, crucially rests on the assumption that ModE POSS ’s is indeed a clitic. In morphological theory, this is not an uncontroversial assumption. Zwicky (1987), for example, regards POSS ’s as an inflectional affix, albeit of a special type, i.e. a phrasal affix.10 Analysed like this, there would not be a development from affix to clitic and hence, as, for example, argued by Eythórsson et al. (2002), also no case of degrammaticalization. It should, however, be kept in mind that there would be still a change from head inflection to phrasal inflection. That is, according to the parameters given in Lehmann (1995 [1982]) there would have been an increase in scope, which on this ground would run counter to a typical grammaticalization process. However, as argued recently by Tabor and Traugott (1998), Lehmann’s criterion of scope reduction may be not correct; instead, Tabor and Traugott (1998) propose scope increase as typically accompanying grammaticalization, and they accordingly analyse the development of POSS ’s towards a phrase marker as a case of grammaticalization (and not degrammaticalization). This proposal is however not uncontroversial either. Although this redefinition of structural scope as a defining property of grammaticalization makes the development of POSS ’s conform to grammaticalization (and hence ‘rescues’ it from the label ‘degrammaticalization’), this new definition of scope now leaves other developments to be explained, as argued by Campbell (2001: 140). Another (and to my mind the most plausible) alternative is to abandon a strict categorical distinction between affixes and clitics and analyse POSS ’s as somewhere
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.5 (77)
The English s-genitive
intermediate between affix and clitic. This is the stance taken by Plank (1992, 1995), who proposes a series of criteria for affixhood and adpositionhood, on the basis of which he locates POSS ’s somewhere in between the two. According to Plank, the synchronically categorically indeterminate status of POSS ’s (= no longer an inflectional affix, though not a full clitic either) is due to its historical development: it used to be a fully-fledged inflectional affix in Old English but has not (yet) acquired full clitic status.11 Note that although Joseph (this volume) emphasizes the importance of synchronic analysis for grammaticalization research (instead of a mere ‘asynchronic’ approach), the case of POSS ’s shows how crucial the particular analysis applied is, leading to very different, in fact opposing results. So, which analysis is to be preferred? There cannot be any consensus since there is no single theoretical framework for grammatical analysis underlying grammaticalization research. Also, Plank’s approach shows how misleading it can be to adopt a strict synchronic, or say at least a strict categorical approach to the question of whether POSS ’s should be regarded as an affix or a clitic, since this obscures the fact that POSS ’s is indeed somewhere in between the two.12 This again links to another question: How much does it take to qualify as a case of degrammaticalization? Does POSS ’s need to make the complete (categorical) ‘jump’ from affix to clitic,13 or is it sufficient that it changes its grammatical status gradually from morphological (i.e. inflectional) to more syntactic (i.e. clitic-like) properties? Some even argue that a grammatical element needs to develop all the way back into some lexical element again to qualify as a genuine case of degrammaticalization (cf. Eythórssson et al. 2002), in which case the s-genitive certainly does not qualify as a case of degrammaticalization (as is indeed Eythórssson et al.’s conclusion). Norde (2001: 236–237) further distinguishes between degrammaticalization as a mirror image of grammaticalization in the sense of a form/construction literally reversing back into a prior state and degrammaticalization more abstractly as the reversal in the status of a form/construction back along a grammaticalization cline. I will address the question of whether the s-genitive can be regarded as a case of degrammaticalization in these two senses at the end of this paper (Section 4). The various ways in which the development of POSS ’s may (or not) be viewed as a case of degrammaticalization are summarized in Figure 2 below. As we can see, there is (too?) much room for interpretation here. This shows that grammaticalization as an empirical theory is not well-defined yet, since there is no consensus as to the question of what should indeed qualify as a genuine counter-example to the hypothesis of unidirectionality (cf. also Lass 2000: 208 for such criticism). In the following I will adopt a strong position and, as a first approximation, regard the development of POSS ’s as a (potential) case of degrammaticalization back along the cline of grammaticality. In the next section I will then sketch a historical scenario for this development, on the basis of which I will
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.6 (78)
Anette Rosenbach POSS ’s synchronic analysis of POSS ’s phrasal affix
in-between affix and clitic
defining criteria for grammaticalization? scope scope reduction increase
de-GR
GR
affix > possessive pronoun
clitic
de-GR
defining criteria for degrammaticalization?
whole in general less ‘jump’ grammatical back along cline no de-GR
de-GR
grammatical > mirror image of lexical GR no de-GR
de-GR?
abstract reversal along cline de-GR?
Figure 2. Ways of interpreting the development of POSS ’s as degrammaticalization (de-GR) or grammaticalization (GR)
finally argue that this development is, strictly speaking, not a counterexample to the unidirectionality of the grammaticality cline in (4).
. The development of POSS ’s – a historical scenario In the following I will propose a historical scenario, assuming that the development of POSS ’s from affix to phrase marker proceeded in several intermediate steps, from smaller to bigger units (following and extending a suggestion made by Allen 1997), and that it has been influenced by other developments within the same structural domain, i.e. the NP. That it is indeed beneficial to adopt such a broader perspective has also been argued by Norde (1997, 2001), who regards deflexion as the driving force in this development. In addition, recently also Demske (2001) offers a detailed analysis of the diachronic development of the German NP, showing how developments within the article system, the adjective inflection, genitives and possessive pronouns are all highly interrelated. I will focus here on the loss of NPinternal agreement, and the rise of phrasal lexicalization and (phrasal) noun+noun (N+N) constructions as possible factors in the development of POSS ’s towards a phrase marker. In addition, I will draw attention to the fact that there is not only a change of POSS ’s towards a phrase marker but that the s-genitive has also acquired a new function as a (definite) determiner.
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.7 (79)
The English s-genitive
. Stages towards POSS ’s as a phrase marker As a very first step, I assume that POSS ’s became potentially ambiguous in premodified prenominal genitives due to the loss of NP-internal agreement (cf. also Seppänen 1997: 200–201; Norde 1997, 2001). While in Old English each element of the possessor NP was still marked for genitive case (as in (6)), at the end of the Middle English period, genitive was only marked once, namely on the possessor head, as in (7c) below. (6) þæs cyninges þegnas the-gen king-gen thanes “the king’s thanes” (Chron A 755; cited from Mitchell 1985: 555) (7) Seinte Katerine (late 12th c; cf. Millar 2000: 300) a. B 695 þet wes þes deofles budel b. R 882-3 þ wes þes deoules budel the-gen devil-gen door-man c. T þ was te deoules budel the-∅ devil-gen door-man “who was the devil’s door-man” Due to the use of the uninflected article in (7c), the genitive suffix can be both interpreted as attaching either to the possessor head king(-es), hence inflectional, or to the whole possessor NP [þe king]es, hence phrasal. Example (7) is particularly illustrative because here we can see variation in the use of the inflected (7a–b) versus uninflected article (7c) in the various manuscripts of Seinte Katerine, which Millar (2000: 300) dates to the late 12th century. Such variation points to the late 12th century as the crucial transition period towards such ambiguous constructions, which now potentially (though not necessarily so) allow for a phrasal interpretation of POSS ’s. As the next two steps I assume that POSS ’s could attach to the last element in appositional and coordinated possessors, following a suggestion made by Allen (1997: 122–123), in which she gives (8) as evidence for first phrasal POSS ’s with appositions (cf. also Fischer 1992: 229–230), and (9) as first evidence for coordinated NPs. (8) Uppon Herode kingess daŠŠ “In King Herod’s day” (9) god and þe virgynes sone Marie “God and the Virgin-gen Mary’s son”
(13thc, Orm 257) (14thc, EDVERN, 249.376)
All these constructions in which POSS ’s can already be interpreted as a phrase marker were possible in English before the first group genitives are attested, which
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.8 (80)
Anette Rosenbach
traditionally are regarded as first evidence for the phrasal status of POSS ’s. It is not until the late 14th century that the first group genitives make their appearance in English documents, as the one in (10). In group genitives the possessor is postmodified and therefore phrasal POSS ’s does no longer attach to the possessor but to the last element of the postmodifier.14 There appear to be two stages in the development of the group genitive, one in which the postmodifier is a noun (and POSS ’s therefore still attaches to a noun as in (10), albeit not the possessor noun), and a second stage in which the postmodifier does not end in a noun as in (11). I could not find any examples for the latter, more ‘progressive’ type of group genitive before the 17th century, as the one in (11a) from Altenberg’s (1982) corpus. (10) a.
God of Loves servantz (late 14th c, Chaucer, Troilus I.15; cited from Fischer 1992: 230) b. the king of Portingales ship (16th c, [HC] Richard Madox: 133)
(11) a. a year or two’s Intrigue (17th c, cf. Altenberg 1982: 62) b. they took some of the trouble out of you-know-who’s head (19th c, Dickens, cf. Jespersen 1961 [1949], vi: 293) While the loss of NP-internal agreement certainly opens up for POSS ’s to get phrasal scope with premodified possessor-NP, this alone can not explain the phrasal attachment of POSS ’s with postmodified possessors, i.e. group genitives.15 In the following I will argue that the development towards the group genitive has been facilitated by the rise of phrasal lexicalization and (phrasal) N+N constructions (cf. also Rosenbach 2002: 218–224).
. Phrasal lexicalization and (phrasal) N+N constructions Phrasal lexemes such as pen-and-ink or brother-in-law are first attested in Middle English (cf. Koziol 1972), and so are N+N constructions such as stone wall (cf. e.g. Jespersen 1961 [1949]: 310–330). Note that phrasal lexemes can be conjoined NPs (pen-and-ink, bread-and-butter) as well as postmodified NPs ending in a noun (e.g. brother-in-law) or in a non-nominal element (e.g. you-know-who, what-d’yecall-‘em). Such phrasal lexemes, I assume, could easily serve as a kind of ‘bridging context’ for the phrasal attachment of POSS ’s with coordinated and postmodified possessor NPs (i.e. group genitives):16 they are analysed as one lexical unit, though their syntactic nature is still transparent. Accordingly, should we regard POSS ’s in (12) as a phrasal or as a head (inflectional) marker?17 (12) On his Brother-in-Law’s behalf
(OED, 1700 TYRELL Hist. Eng. II. 901)
Note also that the first group genitives occur with possessors that are fixed expressions, predominantly in constructions with the pattern [TITLE of PLACE]’s X (e.g.
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.9 (81)
The English s-genitive
the king of England’s daughter; the earl of Warwick’s men). That such titles can easily be perceived of as one name can be seen from example (13) where we see variation between an s-genitive with the postmodified possessor the Earl of Crawfurd and a corresponding appositional possessor the Earl Crawfurd, even within the same line. (13) a. the Erle of Craufurdis cruelltie b. The Erle Craufurdis godlie desire (1570, Pittscottie, Historie & Chronicles of Scotland, 97) It has been pointed out by Mutt (1967) that phrasal lexemes can enter as the first part of such N+N constructions from the second half of the 14th century onwards.18 If Mutt is right with this timing, then it is from this time onwards that constructions as in (14) become possible in English. It is likely that this development started with phrasal lexemes, as in (14a) and only later extended to ad hoc formations such as (14c), which today are very common in English, particularly in journalistic language and fiction. (14) a. a pie in the sky promise b. the allocation of finance committee c. a-less-than-thirty-second event (John Irving, The Fourth Hand, 3) I do not think it is a coincidence that both phrasal N+N constructions (a pie in the sky promise) as well as the first group genitives with name-like possessors (the king of England’s daughter) would have come about at precisely the same time in English, i.e. the late 14th century. Apart from the chronological correlation, there are some further interesting semantic and structural overlaps between s-genitives and N+N constructions, as illustrated in Table 1 below. When talking about the sgenitive we in fact have to distinguish between two different types of constructions. In ‘specifying genitives’ the possessor is referential, referring to a specific referent, i.e. John, identifying whose book it is, i.e. John’s.19 Therefore, in these constructions the possessor functions like a determiner.20 In ‘classifying genitives’ however, the possessor is non-referential, used generically, not referring to a specific referent and thereby not identifying whose cottage it is but rather what type of cottage, i.e. a fisherman’s cottage. In these constructions therefore the possessor functions as a modifier. Similarly, the first element in N+N constructions is non-referential and in a modifier relation to the head noun as well.22 Interestingly, in Middle English and early Modern English (and in northern dialects well until the 20th century, cf. Klemola 1997) we can also find an s-less variant of the specifying genitive, as in Jack wife or the man name.23 Such s-less genitives can also be found as group genitives as shown in (15). (15) Rafe of Raby time
(16thc, [HC] Leland: 76)
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.10 (82)
Anette Rosenbach
Table 1. (Types of) s-genitives versus N+N constructions specifying s-genitive s-genitive s-less genitive
classifying s-genitive21
N+N constructions
examples
John’s book
POSS ’s
+
Jack wife, the man name –
a fisherman’s cottage +
stone wall, city traffic –
referentiality of possessor/first element
+
+
–
–
phrasal possessor/first element
[the king of England]’s daughter
[Rafe of Raby] time
(+)
+
[the Earl of Northumberland] son
in [[twenty years’] time] an old-person’s life
the [[allocation of finance committee]]
There is both structural variation (absence of presence of POSS ’s) and semantic variation ([±referential] possessor), which leads to a mismatch between form and meaning. It is particularly in two constructions that the isomorphism between form and meaning is ‘disturbed’, i.e. in classifying genitives and s-less genitives. They therefore provide a possible ‘bridge’ between s-genitives and N+N constructions (see also Figure 3 below). I assume that such form/meaning mismatches have given rise to some kind of analogy, with hearers having two possible interpretations ([±referential] possessor/first noun) for one surface construction: both the s-genitive as well as a N+N juxtaposition can entail a referential possessor/first noun (Jack’s wife ∼ Jack wife) or a non-referential one (a fisherman’s cottage vs. stone wall, city traffic). Likewise, speakers have two constructions (with or without POSS ’s) at their disposal to convey the same meaning (either a referential or a non-referential possessor/first noun). For the scenario I am suggesting it is not necessary that there is always true variation in each single case. Rather, what is crucial is that on a more abstract level the distinction between the constructions becomes blurred, which makes it possible for the properties of one construction to ‘jump’ to another one. Given such an analogy, it is plausible that the constructions have mutually affected each other in the rise of phrasal first elements in the s-genitive and N+N constructions. A crucial precondition in this process was the rise of phrasal lexemes, which provide a good bridge for phrasal elements to slip into a position usually reserved for words only. All this certainly still needs to be looked at in more detail in the textual evidence; the crucial point in the present
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.11 (83)
The English s-genitive +POSS ’s specif. s-genitives (Jack’s wife)
[–ref.] possessor/first N N+N constructions classif. s-genitives (stone wall) (a fisherman’s cottage)
[+ref.] possessor
–POSS ’s
(specif.) s-less genitive (Jack wife)
Figure 3. ‘Bridging contexts’ between the s-genitive and N+N constructions
argumentation is, however, that such an analogy is possible, due to the similarities and affinities between the constructions.24
. The new function of the s-genitive as a (definite) determiner What has so far been largely neglected in the literature is that POSS ’s did not only become a phrase marker but that the s-genitive also turned into a new construction.25 It is quite uncontroversial that the possessor in the ModE (specifying) s-genitive functions as a definite determiner (cf. Quirk et al. 1985: 326ff.; Taylor 1996; Huddleston & Pullum 2002: 467–468). It is in complementary distribution and cannot co-occur with other determiners (16), and it forces a definite interpretation of the whole possessive NP even if the possessor itself is indefinite (17). (16) the daughter this daughter John’s daughter (17) a king’s daughter → the daughter of a king Looking at Old English, we see that here the inflected genitive does not exhibit the properties of a definite determiner (yet). The genitive-marked possessor can still co-occur with other determiners (18), and it does not necessarily render the whole NP definite (19). (18) on Godes þa gehalgodan cyricean in God-gen that hallowed church “in that hallowed church of God” (BlHom X.111.8–9, cited from Traugott 1992: 175)
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.12 (84)
Anette Rosenbach
(19) sume wæstmas godra weorca some fruits good-gen works-gen “some fruits of good works” (ÆlfHom 3.1829, cited from Traugott 1992: 175) Another possible indication for the determiner status of posssessives in Old English is the strong/weak adjective inflection. In Old English, the strong adjective declension is generally connected to the indefiniteness of the whole NP. Therefore, if the possessor in the inflected genitive had already the properties of a definite determiner, we would expect any adjectival modifier of the possessum to be in the weak declension. According to the handbooks this is indeed the case (cf. e.g. Mitchell 1985: 65; Traugott 1992: 171). Note, however, that Thomas (1931: 140–151) provides a list of cases from his extensive Old English corpus where the possessum can be modified by a strong adjective, as in (20). (20) sylfrene Cristes mael silver-strg Christ-gen mark/cross “(a) silver mark/cross of Christ” (Bede 58,24, cf. Thomas 1931: 141) It should be noted, however, that most of the examples given in Thomas (1931) are genitive constructions which can very easily be perceived of as (syntactically opaque) possessive compounds as in (20), and therefore cannot provide conclusive evidence against a definite interpretation of possessive NPs in Old English.26 All in all, we can conclude, however, that in Old English possessives were not categorically connected to the definiteness of the whole possessive NP. So at some point in time there must have been a change to the new determiner function, though there has not been any discussion of this yet in the literature. The question of how this change took place is beyond the scope of the present paper; here it is sufficient to note that such a change must have taken place. As the crucial transition period I assume the Middle English period. By the 14th century the sgenitive has become restricted to prenominal position (GEN-N; *N-GEN), which corresponds well to the structural position of determiners in English.27 At that time also the new article system made its way into English. According to Fischer (1992: 217–218) the evolution of the new uninflected definite article the and the indefinite article a/an is closely related to the expression of definiteness. Possessives as such are closely related to definiteness (cf. e.g. Löbner 1985; Haspelmath 1999b; Fraurud 2001), and the possessive meaning is the central (albeit not the exclusive) meaning expressed by genitive case. As shown in Rosenbach and Vezzosi (2000), the semantic-pragmatic restrictions of the possessor in late Middle and early Modern English to [+human] and [+topical] possessors correspond well to those of a referential anchor, i.e. a definite determiner. Moreover, it is interesting that the sgenitive drops out completely from those functions which are not compatible with
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.13 (85)
The English s-genitive
those of a referential anchor, e.g. partitives (OE sumne dæl þæs(gen) felles(gen) > some (part) of that skin) and descriptive genitives (OE haliges(gen) lifes(gen) mann > a man of holy life); see also Rosenbach (2002: 226–227).28 It therefore seems likely that the (specifying) prenominal s-genitive becomes part of the newly emerging article system in late Middle English, very much in the sense as recently proposed by Demske (2001) for the German NP. While it looks as if the two developments, i.e. the development towards phrasal POSS ’s and the new determiner function of the s-genitive are closely interwoven, they yet appear to be logically independent, when taking a brief look at the development of prenominal possessives in other Germanic languages (cf. Table 2). While in English and Swedish the s-genitive has become both a determiner and a phrase marker, in German and Dutch prenominal possessives have turned into determiners but are phrase markers, which shows that the two developments can be dissociated. Table 2. Prenominal possessives in Germanic languages prenominal possessives phrase marker determiner English Swedish (cf. Norde 2001) German (cf. Demske 2001) Dutch (cf. Weerman & de Wit 1999)
+ + – –
+ + + +
Note that the acquisition of this new determiner function opens up yet another possibility to interpret the development of the s-genitive as a potential case of degrammaticalization so far not mentioned in the literature. If Payne (2002) is right with his pathway ‘definiteness > genitive case’ (but see Fraurud 2001 for suggesting a pathway ‘possessive’ > ‘definite article’), then the development of the s-genitive from an inflectional construction (marking genitive case) to that of a definite determiner would also constitute a development back along this grammaticalization pathway.29 The whole development of POSS ’s towards a phrasal marker is now summarized and illustrated in Figure 4 below.
. A case of degrammaticalization? What now are the implications of such a scenario for the question of degrammaticalization? First, the development of POSS ’s can certainly not be a case of degrammaticalization as a mirror image of grammaticalization – it did not develop
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.14 (86)
Anette Rosenbach POSS ’s
potentially ambiguous (inflection ~ phrase marker) on possessor on possessor
on part of possessor
phrase marker
not on possessor
on noun
on non-nominal elements
Art. (Adj.) N appositions
conjoined NPs
group genitive (I) group genitive (II)
the devil’s door man
Herode kingess day
god and þe virgynes soul
the God of Loves servantz
a year or two’s Intrigue, you-know-who’s head, anybody else’s
13thc
14thc
late 14thc
17thc
(12thc)
loss of NP-internal agreement
phrasal lexicalization & phrasal N+N constructions reanalysis of possessor as possessive determiner
evolution of new article system
restriction to prenominal genitive
loss of inflection in article and adjective
Figure 4. Historical scenario: development of POSS ’s towards a phrase marker
back into a previous historical stage but the prenominal s-genitive acquired a new function as a definite determiner (and see also Norde 2001: 236 for arguing that such a mirror image reversal is conceptually quite unlikely to occur). So, it can only be regarded as a case of degrammaticalization in the more abstract sense of reversing its grammatical status back along the cline from affix to clitic. In the following I will now argue that it is even questionable that it is a genuine case of degrammaticalization in this more abstract sense. For this reason, I will refer to Andersen’s (2001) distinction between ‘change schemas’ and ‘historical changes’.30 Change schemas are abstract generalizations of classes of change (i.e. types of change) – these can be, though not necessarily have to be, grammaticalization clines. Historical changes, in contrast, are instantiations of change, i.e. concrete change(s) of specific forms/constructions (i.e. tokens of change). Applied to the English sgenitive, the relevant change schema in question is the grammaticality cline in (4), while the historical change is the concrete development of POSS ’s. Adopting this distinction we can now argue that POSS ’s is originally located on the morphosyn-
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.15 (87)
The English s-genitive
tactic cline as the relevant change schema, and it would have run the full course, presumably to zero, if nothing else had happened. And note, that there is indeed evidence for stage ‘zero’ as attested in the s-less genitive (Jack wife) in Middle and early Modern English (and in northern dialects even beyond). But something else did happen which made POSS ’s change track and leave the grammaticality cline and become part of the newly emerging article system, allocating a new function as a definite determiner to the s-genitive. This seems to have ‘liberated’ POSS ’s to become a phrase marker too, and this development was presumably facilitated by the loss of NP-internal agreement and the rise of phrasal lexicalization and N+N constructions (for illustration, see Figure 5 below). So it appears that the s-genitive has been subject to what Plank (1995) called a ‘Systemstörung’ (cf. also Norde 2001).31 Alternatively, the new function of POSS ’s might also fall under the term ‘exaptation’, which is defined by Lass (1997: 316) as “a kind of conceptual renovation, as it were, of material that is already there, but either serving some other purpose, or serving no purpose at all” (cf. also Norde 2001, 2002; Traugott 2001).32 Whatever we wish to call the reason for this switch, what is important for the present argumentation is that when adopting the conceptual distinction between type and token, we can view POSS ’s as being part of several change schemas/types, which makes it possible in principle for it to change its track of change.33 If we look at the concrete development of POSS ’s, i.e. at the token level (see the bold line), it certainly looks as if it were a case of degrammaticalization when comparing its (synchronic) status as an affix in Old English with that of a phrase marker in Modern English. If unidirectionality is however a property of types (i.e. grammaticalization clines), and if a token of change can be part of several change types/schemas, we can now argue that the development of POSS ’s is, strictly speaking, not a counterdirectional change. That is, unidirectionality on the type level of the morphosyntactic cline remains unaffected by the concrete development of the
change schema (type) token
CLITIC > AFFIX > POSS ’s (infl.)>
ZERO s-less genitive
evolution of ModE article system change schema (type) token
s-genitive (possessor) as DET case loss & loss of NP-agreement
change schema (type) token change schema (type)
POSS ’s = phrase marker development towards phrasal lexemes & N+N constructions
Figure 5. Type (change schema) – token distinction and the development of POSS ’s
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.16 (88)
Anette Rosenbach
s-genitive. Where does this leave us empirically? Does that mean we can now ‘explain away’ any potential case of degrammaticalization by simply letting it become part of some other (hypothesized) development? I don’t think so. It does mean, however, that in every single case of reported degrammaticalization we should also take a look at the broader context of the change (as in the present case, the broader development of the English NP). And only if we cannot see any ‘kick-out’ factor being at work should we accept it as a real counter-example to unidirectionality. Note that the importance of ‘other developments’ in the grammar has already been stressed by Plank (1995) and Norde (2001) for the s-genitive and Fischer (2000b) for the development of English to. In Plank’s and Fischer’s argumentation it is precisely the very presence of such ‘kick-out’ factors that make degrammaticalization possible in principle. And if we wish to consider unidirectionality to be a property of token development, this is certainly the right interpretation which does not, I think, clash with the one presented here. That is, the interpretation of whether the presence of ‘kick-out’ factors speaks for or against unidirectionality rests on our assumption on which level to locate unidirectionality, the token or the type level. If, however, at any stage in the history of a form/construction (i.e. a token) disturbing factors can change the track of its ‘predestined’ development (i.e. its change schema/type), then it is no longer a very reliable tool for reconstruction. Norde (2001) comes yet to another solution. She suggests that the process of grammaticalization is qualitatively so different from degrammaticalization that one should a priori exclude the latter from the definition of what comprises prototypical grammaticalization: . . . , the most workable model is one in which the Strong Hypothesis of Unidirectionality is rejected and in which grammaticalization, non-directional and counterdirectional changes are strictly separated. This implies that the unidirectionality of grammaticalization is indeed tautological, and that counterdirectional changes are not counterexamples to grammaticalization, but simply another (albeit less common) type of change. (Norde 2001: 238)
While I agree with Norde that the development of the s-genitive from affix to phrase marker does not challenge the directionality of grammaticalization, for epistemological reasons I don’t think it is desirable to rule out counterdirectional changes in principle from the definition of grammaticalization, since then unidirectionality would no longer be an empirical hypothesis but part of the definition, hence not falsifiable (cf. also Cowie 1995: 188–189; Campbell 2001: 124; and Newmeyer 1998: 263 for similar arguments and discussion). Rather, I think it is sufficient to restrict unidirectionality as a property to single types of grammaticalization paths. In every single case it is then always an empirical question whether what appears to be a case of degrammaticalization, i.e. a potential falsifier of unidirectionality, stands up to closer inspection and really turns out to be a development back
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.17 (89)
The English s-genitive
down that same path, or whether it simply becomes a part of yet another (regular) development. Finally a word of caution: It should be stressed that the assessment of whether the s-genitive is a case of degrammaticalization is interpretation which heavily depends on how (de)grammaticalization is defined in the first place. As the discussion in Section 2.2 has shown, however, the defining properties of what should constitute a genuine case of (de)grammaticalization are not yet agreed on, and various interpretations are possible. Priority should therefore always be given to the history of the s-genitive – future (re)definitions of (de)grammaticalization may well come up with other interpretations.
Notes * I would like to thank Muriel Norde for engaging into an extensive and stimulating discussion on the s-genitive, and Olga Fischer for valuable and inspiring comments on an earlier version of this paper. . In the same vein, though from a very different ontological position, Lass (2000: 223) argues that “[i]nformation-loss processes have no memory”. Therefore, there is nothing inherent in grammaticalization clines that would make an upgrading of a form/construction impossible in principle. . Note that Martin Haspelmath used to maintain a stronger position in the past (cf. e.g. Haspelmath 1999a). . I owe the examples to Monika Schmid (2) and Sheila Watts (3). Note that in the following, if not indicated otherwise, the examples used are from the corpus investigated in Rosenbach and Vezzosi (1999), where also a full list of the texts used is provided. [HC] = Helsinki Corpus. . And see also Rosenbach (2002) for showing that once POSS ’s has turned into a clitic, the s-genitive increases both in its relative frequency and in the possible contexts in which it can be realized, most notably to inanimate possessors in Modern English. There I argue that in this extension of the (new) s-genitive, grammaticalization processes can be discerned. . For further discussion of Allen’s (1997) and Janda’s (1980, 2001) arguments, see e.g. Norde (2001: 254–256) and Rosenbach (2002: 212–217). . See e.g. Jespersen (1960 [1918]: 306); Mustanoja (1960: 159–161); Brunner (1984 [1962]: 30–32), and Fischer (1992: 230) for tracing such his-genitives back to Old English times; see however Allen (1997) for arguing that there have not been any genuine cases of his-genitives before the 16th century. . Actually, there appears to be yet another possible interpretation for the use of the masc. possessive pronoun ys (“his”) with a feminine possessor. In some dialects of German (including my own), it is very common to refer to women with the neuter article, even with proper names (das Maria, “the-neuter Mary”), and accordingly do the construction with a possessive linking pronoun look like: dem Maria sein Buch, “the-dat-neuter Mary its-
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.18 (90)
Anette Rosenbach
neuter book”). Note also den Besten (2002) for giving examples for a similar use of masc./neuter possessive pronouns in Dutch dialects. If we consider that his used to be the form for both the masculine as well as the neuter possessive pronouns until the 17th century, when the new possessive pronoun its emerged in the English language (cf. Jespersen 1961 [1949]: 307), example (5) can also be interpreted as a case where the feminine possessor is (quite naturally) referred to by the neuter his. . To my mind, the most plausible position is to assign neither a strong causal role to the his-genitive (as Janda does) nor to reject it at all for the emergence of the clitic -s (as is Allen’s 1997 position), but rather to regard the his-genitive as a kind of catalyst in this development (for such a position see e.g. Jespersen 1960 [1918]; Plank 1992). . For some arguments for treating POSS ’s still as an affix, see also Huddleston and Pullum (2002: 480–481). . I will therefore myself, in the following, refer to POSS ’s as a ‘phrase marker’ rather than a ‘clitic’. . See also Allen (1997: 122) for suggesting to adopt a theoretical approach towards the categorization of POSS ’s which incorporates gradualness; in particular she refers to Sadock’s (1985) autolexical-syntax approach, which assumes affixes and clitics being located on a continuum. . This also begs the question of what justifies those clines or defines the stages on them in the first place; see e.g. Lass (2000) for discussion. . As with appositional and coordinated possessors there is a transition period expanding well into the 16th century, in which the older inflectional construction co-exists next to the new phrasal POSS ’s (cf. e.g. Allen 1997; Rosenbach & Vezzosi 1999), as evident from split genitives, in which the -s in postmodified possessors does not attach to the last element of the postmodifier but to the possessor head, as in the dukis brothir of Bretayn (15th c, [HC] Capgrave: 246). . As in English, in Swedish loss of NP-internal agreement was the first stage towards a phrasal interpretation of the s-genitive, and group genitives only appear at a later stage (cf. Norde 2002: 56–58). This suggests that loss of NP-internal agreement with premodified possessors is indeed the first step towards a phrasal interpretation of the prenominal genitive, with group genitives, in which the possessor is postmodified being a later add-on. . The term ‘bridging context’ is simply used here to refer to specific contexts which are ambiguous (for details, see further below) and not as more specifically defined by Heine (2002). . Interestingly, also plural attachment was variable on such phrasal lexemes in early Modern English. According to Nevalainen (1999: 422) there are three different variants for the plural of son-in-law in Shakespeare’s King Lear, one in which the plural -s attaches to the end of the phrasal lexeme (son-in-laws), one in which it attaches to the first noun (sons-inlaw), and finally one in which plural is marked on both nouns (sons-in-laws). According to Quirk et al. (1985: 313) plural-marking is still variable in such cases, with the phrasal attachment of -s, as in mother-in-laws regarded as being more informal. This indicates that lexicalised phrases such as son-in-law could (and still can) be both perceived as syntactic
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.19 (91)
The English s-genitive
phrases and phrasal lexemes and shows, in general, how fluid the borderline between syntax and morphology is in such cases. . N+N compounds were a common feature of Old English word formation. What is meant here is the ability of nouns to serve as premodifiers. Note that there is some controversy in the literature whether to treat such constructions in Modern English as compounds or as syntactic phrases (cf. e.g. Bauer 1998). What is essential for the present argument is that no matter how to categorize them, the first element is always non-referential. . For the terms ‘specifying genitives’ and ‘classifying genitives’ see Biber et al. (1999: 294– 295). . Compare also the term ‘subject-determiner’ as used in Huddleston and Pullum (2002: 467–468); see also Section 3.3 for the determiner-function of the s-genitive. . Huddleston and Pullum (2002: 469–470) distinguish two different types of classifying genitives, or, in their terminology ‘attributive genitives’, i.e. (a) descriptive genitives (fishman’s cottage) and (b) measure genitives (a second [one hour’s delay]). According to them, it is only in the latter type that the possessor can easily be phrasal, while in the former this is only a very marginal option (e.g. an [all girls’ school]). . As with N+N constructions (cf. Note 17 above) there is a theoretical controversy as to the status of such constructions as compounds or syntactic phrases. Again, what is important for the present argumentation is that both share a non-referential possessor/first element. . S-less genitives could already be found in Old English with certain noun-classes, but in Middle English and early Modern English they also occur with nouns originally not lacking the -s as in the examples given above. . However, it is questionable that phrasal lexicalization and phrasal N+N constructions are a necessary criterion for the emergence of phrasal POSS ’s. Muriel Norde informs me that in Swedish, which similarly to English also developed a phrasal s-genitive (cf. particularly Norde 1997 for an exhaustive account of this development), phrasal lexemes seem to be restricted to certain loan-translations (such as forget-me-nots) and phrasal N+N constructions do not exist. On the other hand, German, which does not permit phrasal possessors in the prenominal genitive does have phrasal first elements in such constructions (dieser ich-fasses-nicht Blick, “this I-don’t-believe-it expression”) though I don’t think such constructions are as productive as in English. . But see brief notes in Janda (2001: 302) and for the Swedish s-genitive in Norde (1997: 227–229, 2001: 253). . Note also that it is possible that Old English had a system similar to Modern German, where a prenominal genitive (analysed as a definite determiner, cf. Demske 2001) does not require a following adjective to be weak, as in (cf. Annas kleiner Bruder; “Anne’s (gen/+def) little-strg brother”). According to Demske (2001: 40) it is only a morphologically inflected determiner carrying information about case, number and gender that rules out the strong adjective inflection. In this case, the type of adjective declension used with possessives cannot provide any evidence for or against the definiteness of possessive NPs in the first place. . Thomas (1931: 105ff.) links this restriction to prenominal position to the loss of inflection in the article and the strong adjective; his data clearly show that there is a striking
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.20 (92)
Anette Rosenbach
chronological correlation between the two: in the 14th century inflection on the article and strong adjectives is lost completely, and at that time also the prenominal s-genitive is the only option. That is, in his vast corpus there is no single occurrence of constructions such as: *þe sune þe mannes (the son the man-gen). Note also that the loss of postnominal position has most likely facilitated the reanalysis of POSS ’s as a definite determiner (the precise chronology, however, is still unclear, it could as well be said that the loss of the postnominal genitive is a result of this reanalysis). For adjectives, Fischer (2000a) also observes a strong correlation between definiteness for the prenominal and indefiniteness for postnominal position. . See also Fischer (1992: 228–229) for discussing further contexts in which the genitive disappears in Middle English, as e.g. with numerals (OE þara(gen) fiftig > fifty of them), superlatives (OE husa(gen) selest > the best of houses) or measure words (OE fotes(gen) trym > space of a foot/foot space). Note that in all these genitive functions it is not the possessor that functions as a referential anchor, but the anchoring relation is precisely the other way round with the possessum/head noun individuating the possessor. . Regarding the new determiner function of the s-genitive from (supposedly) the late Middle English period onwards may also bring Janda’s his-account back into the discussion, since such possessive constructions with possessive linking pronouns are particularly apt for entering into a specification (i.e. determiner) relation (see Rosenbach 2002: 227– 228 for further discussion). It should also be noted that all the Germanic languages which have developed a possessive determiner from a prenominal possessive do/did have such constructions with possessive linking pronouns (cf. also Koptjevskaja-Tamm 2003: 665–676); although only in English this possessive pronoun could be completely homophonous with the genitive marker. . I am grateful to Elizabeth Traugott for having drawn my attention to this distinction and for discussion. . In both Plank’s and Norde’s account it is above all the breakdown of the case system that paves the way for POSS ’s to become a phrase marker. . It is questionable, however, whether the s-genitive should qualify as a genuine case of exaptation. Note that there seem to be strong conceptual links between possessives on the one hand and the expression of definiteness (as typical of definite determiners) on the other (cf. e.g. Löbner 1985; Haspelmath 1999b; Fraurud 2001), which at least calls into question that the s-genitive acquired a new function completely unconnected to the old. . Alternatively, we might also consider the possibility of competing change schemas/types, which can be in competition with each other, just as the notion of ‘competing motivation’ in functional linguistics (cf. DuBois 1985) or as formalized in Optimality Theory (OT) as constraint interaction.
References Allen, Cynthia (1997). “The Origins of the ‘Group Genitive’ in English”. Transactions of the Philological Society, 95, 111–131.
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.21 (93)
The English s-genitive
Altenberg, Bengt (1982). The Genitive v. the Of-Construction. A Study of Syntactic Variation in 17th Century English. Malmö: CWK Gleerup. Andersen, Henning (2001). “Actualization and the (Uni)directionality of Change”. In Henning Andersen (Ed.), Actualization. Linguistic Change in Progress (pp. 226–248). Amsterdam & Philadelphia: John Benjamins. Bauer, Laurie (1994 [1988]). Introducing Linguistic Morphology. Edinburgh: Edinburgh University Press (corrected 1992, reprinted 1994). Bauer, Laurie (1998). “When is a Sequence of Two Nouns a Compound in English?” English Language and Linguistics, 2, 65–86. den Besten, Hans (2002). “The Afrikaans Possessive Constructions Revisited“. Paper presented at New Reflections on Grammaticalization 2, April 4–6, 2002, University of Amsterdam. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, & Edward Finegan (1999). Longman Grammar of Spoken and Written English. London: Longman. Brunner, Karl (1962). Die englische Sprache. Ihre geschichtliche Entwicklung, Vol. 2 (2nd revised edition 1984). Tübingen: Niemeyer. Campbell, Lyle (2001). “What’s Wrong with Grammaticalization?” Language Sciences, 23, 113–161. Carstairs, Andrew (1987). “Diachronic Evidence and the Affix-Clitic Distinction”. In Anna G. Ramat, Onofrio Carruba, & Giuliano Bernini (Eds.), Papers from the 7th International Conference on Historical Linguistics (pp. 151–162). Amsterdam & Philadelphia: John Benjamins. Cowie, Claire (1995). “Grammaticalization and the Snowball Effect”. Language & Communication, 15, 181–193. Demske, Ulrike (2001). Merkmale und Relationen. Diachrone Studien zur Nominalphrase des Deutschen. Berlin & New York: Mouton de Gruyter. DuBois, John W. (1985). “Competing Motivations”. In John Haiman (Ed.), Iconicity in Syntax (pp. 343–365). Amsterdam & Philadelphia: John Benjamins. Eythórsson, Thórhallur, Kersti Börjars, & Nigel Vincent (2002). “On Defining Degrammaticalization”. Paper presented at New Reflections on Grammaticalization 2, April 4–6, 2002, University of Amsterdam. Fischer, Olga (1992). “Syntax”. In Norman Blake (Ed.), The Cambridge History of the English Language, Vol. II (1066–1476) (pp. 207–408). Cambridge: Cambridge University Press. Fischer, Olga (2000a). “The Position of the Adjective in Old English”. In Ricardo BermúdezOtero, David Denison, Richard M. Hogg, & C. B. McCully (Eds.), Generative Theory and Corpus Studies. A Dialogue from 10 ICEHL (pp. 153–182). Berlin & New York: Mouton de Gruyter. Fischer, Olga (2000b). “Grammaticalisation: Unidirectional, non-reversable? The case of to before the infinitive in English”. In Olga Fischer, Anette Rosenbach, & Dieter Stein (Eds.), Pathways of Change. Grammaticalisation in English (pp. 149–169). Amsterdam & Philadelphia: John Benjamins. Fraurud, Kari (2001). “Possessives with Extensive Use: A Source for Definite Articles?” In Irène Baron, Michael Herslund, & Finn Sørensen (Eds.), Dimensions of Possession (pp. 243–267). Amsterdam & Philadelphia: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.22 (94)
Anette Rosenbach
Haspelmath, Martin (1999a). “Why is Grammaticalization Irreversible?” Linguistics, 37, 1043–1068. Haspelmath, Martin (1999b). “Explaining Article-Possessor Complementarity: Economic motivation in noun phrase syntax.” Language, 75, 227–243. Heine, Bernd (2002). “On the role of context in grammaticalization.” In Ilse Wischer & Gabriele Diewald (Eds.), New Reflections on Grammaticalization (pp. 83–101). Amsterdam & Philadelphia: John Benjamins. Hopper, Paul & Elizabeth C. Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Huddleston, Rodney & Geoffrey K. Pullum (2002). The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Janda, Richard D. (1980). “On the Decline of Declensional Systems: the Overall Loss of OE Nominal Case Inflections and the ME Reanalysis of -es as his”. In Elizabeth C. Traugott, Rebecca Labrum, & Susan Sheperd (Eds.), Papers from the 4th International Conference on Historical Linguistics (pp. 243–253). Amsterdam & Philadelphia: John Benjamins. Janda, Richard D. (2001). “Beyond ‘Pathways’ and ‘Unidirectionality’: On the Discontinuity of Language Transmission and the Counterability of Grammaticalization”. Language Sciences, 23, 265–340. Janda, Richard D. (2002). “(De)Grammaticalization with a Human Face: Or, how a Socially Informed Can Be both Hyper(-) & Correct”. Paper presented at New Reflections on Grammaticalization 2, April 4–6, 2002, University of Amsterdam. Jespersen, Otto (1894). Progress in Language. With Special Reference to English. London: Swan Sonnenschein & Co. Jespersen, Otto (1960 [1918]). “Chapters on English”, reprinted 1960 in Selected Writings of Otto Jespersen (pp. 153–345). London: Allen & Unwin. Jespersen, Otto (1961 [1949]). A Modern English Grammar on Historical Principles. London: Allen & Unwin. Klemola, Juhani (1997). “Dialect Evidence for the Loss of Genitive Inflection in English”. English Language and Linguistics, 1(2), 350–353. Koptjevskaja-Tamm, Maria (2003). “Possessive NPs in the Languages of Europe”. In Frans Plank (Ed.), The Noun Phrase in the Languages of Europe (pp. 621–722). Berlin & New York: Mouton de Gruyter. Koziol, Herbert (1972). Handbuch der Englischen Wortbildungslehre. Heidelberg: Carl Winter. Lass, Roger (1997). Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Lass, Roger (2000). “Remarks on (Uni)directionality”. In Olga Fischer, Anette Rosenbach, & Dieter Stein (Eds.), Pathways of Change. Grammaticalization in English (pp. 207–227). Amsterdam & Philadelphia: John Benjamins. Lehmann, Christian (1995 [1982]). Thoughts on Grammaticalization. A Programmatic Sketch. University of Cologne. Revised an expanded version 1995 (LINCOM Studies in Theoretical Linguistics 01). München/Newcastle: Lincom. Löbner, Sebastian (1985). “Definites”. Journal of Semantics, 4, 279–326.
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.23 (95)
The English s-genitive
Millar, Robert McColl (2000). “Some Suggestions for Explaining the Origin and Development of the Definite Article in English”. In Olga Fischer, Anette Rosenbach, & Dieter Stein (Eds.), Pathways of Change. Grammaticalization in English (pp. 275–310). Amsterdam & Philadelphia: John Benjamins. Mitchell, Bruce (1985). Old English Syntax. Vol. 1. Oxford: Clarendon. Mustanoja, Tauno (1960). A Middle English Syntax. Helsinki: Société Néophilologique. Mutt, O. (1967). “On Some Recent Developments in the Use of Premodifiers in English”. Zeitschrift für Anglistik und Amerikanistik, 4, 401–408. Nevalainen, Terrtu (1999). “Early Modern English Lexis and Semantics”. In Roger Lass (Ed.), The Cambridge History of the English Language, Vol. III (1476–1776) (pp. 332–458). Cambridge: Cambridge University Press. Newmeyer, Frederick (1998). Language Form and Language Function. Cambridge, MA: MIT Press. Norde, Muriel (1997). The History of the genitive in Swedish. A Case Study in Degrammaticalization. Ph.D. dissertation. Vakgroep Skandinavische taal-en letterkunde, Amsterdam. Norde, Muriel (2001). “Deflexion as a Counterdirectional Factor in Grammatical Change”. Language Sciences, 23, 231–264. Norde, Muriel (2002). “The Final stages of Grammaticalization: Affixhood and Beyond”. In Ilse Wischer & Gabriele Diewald (Eds.), New Reflections on Grammaticalization (pp. 45–65). Amsterdam & Philadelphia: John Benjamins. Payne, John (2002). “From Definiteness to Genitive Case”. Paper presented at New Reflections on Grammaticalization 2, April 4–6, 2002, University of Amsterdam. Plank, Frans (1992). “From Cases to Adpositions”. In N. Pantaleo (Ed.), Aspects of English Diachronic Linguistics: Papers read at the Second National Conference on History of English, Naples, 28–29 April 1989 (pp. 19–61). Fasano: Schena. Plank, Frans (1995). “Entgrammatisierung – Spiegelbild der Grammatisierung?” In Norbert Boretzky, Wolfgang Dressler, Jan Orešnik, Karmen Terzan, & Wolfgang Wurzel (Eds.), Natürlichkeitstheorie und Sprachwandel (pp. 199–219). Bochum: Universitätsverlag Dr. N. Brockmeyer. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, & Jan Svartvik (1985). A Comprehensive Grammar of the English Language. London & New York: Longman. Rosenbach, Anette (2002). Genitive Variation in English. Conceptual Factors in Synchronic and Diachronic Studies (=Topics in English Linguistics, 42). Berlin & New York: Mouton de Gruyter. Rosenbach, Anette & Letizia Vezzosi (1999). “Was the s-genitive a Traveller through England?” In Alexander Bergs & Dieter Stein (Eds.), LANA-Düsseldorf Working Papers on Linguistics, Vol. 1 (pp. 35–55). (http://ang3-11.phil-fak.uni-duesseldorf.de/∼ ang3/LANA/LANA.html) Rosenbach, Anette & Letizia Vezzosi (2000). “Genitive Constructions in Early Modern English: New evidence from a corpus analysis”. In Rosanna Sornicola, Erich Poppe, & Ariel Shisha-Halevy (Eds.), Stability, Variation and Change in Word-Order Patterns over Time (pp. 285–307). Amsterdam & Philadelphia: John Benjamins. Sadock, Jerrold (1985). Autolexical Syntax. A Proposal for the Treatment of Noun Incoporation and Similar Phenomena. Chicago: University of Chicago Press.
TSL[v.20020404] Prn:13/04/2004; 15:48
F: TSL5904.tex / p.24 (96)
Anette Rosenbach
Seppänen, Aimo (1997). “The Genitive and the Category of Case in the History of English”. In Raymond Hickey & Stanislaw Puppel (Eds.), Language History and Linguistic Modelling. A Festschrift for Jacek Fisiak on his 60th Birthday, Vol. 1 (pp. 193–214). Berlin & New York: Mouton de Gruyter. Tabor, Whitney & Elizabeth C. Traugott (1998). “Structural Scope Expansion and Grammaticalization”. In Anna G. Ramat & Paul Hopper (Eds.), The Limits of Grammaticalization (pp. 229–272). Amsterdam & Philadelphia: John Benjamins. Taylor, John (1996). Possessives in English. Oxford: Clarendon. Thomas, Russel (1931). Syntactical Processes Involved in the Development of the Adnomonal Periphrastic Genitive in the English Language. Ph.D. dissertation. Ann Arbor: University of Michigan. Traugott, Elizabeth D. (1992). “Syntax”. In Richard Hogg (Ed.), The Cambridge History of the English Language, Vol. I (The Beginnings – 1066) (pp. 168–289). Cambridge: Cambridge University Press. Traugott, Elizabeth C. (2001). “Legitimate Counterexamples to Unidirectionality”. Paper presented at the University of Freiburg, October 2001. Weerman, Fred & Petra de Wit (1999). “The Decline of the Genitive in Dutch”. Linguistics, 37, 155–1192. Wyld, Henry Cecil (1936 [1920]). A History of Modern Colloquial English (3rd edition). Oxford: Basil Blackwell. Zwicky, Arnold M. (1987). “Suppressing the Zs”. Journal of Linguistics, 23, 133–148.
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.1 (97)
An investigation into the marginal modals dare and need in British present-day English A corpus-based approach Martine Taeymans University of Antwerp & Limburg University Centre
.
Introduction
This paper investigates the current behaviour and recent changes in the marginal modals dare and need on the one hand, and the semi-modals dare to and need to on the other, in the spoken and written British National Corpus. Through a comparison of written and spoken data, I will try to reveal whether there have been any significant shifts in frequency of the modal and semi-modal variants, assuming that discourse reflects the latest trends in language change. Whenever necessary or relevant, synchronic data will be backed by diachronic findings available in the literature. Changes in frequency will tell us more about how dare and need have fared with respect to (de)grammaticalization, as grammaticalization goes hand in hand with increasing text frequency (Krug 2000: 21; Hopper & Traugott 2003: 103). Since Early Modern English dare and need have oscillated between main verb (dare to, need to) and modal morphosyntax (dare, need). need, originating from the OE impersonal verb (ge)neodan, occurred in a range of personal and impersonal constructions in ME. But in the 16th century it started to show modal features: “[. . . ] it develops preterite-present need alongside needs, and instead of disappearing, the plain infinitive continues and even strengthens its position: it seems to be the less favoured infinitive in late Middle English [. . . ] but it is the commoner in Shakespeare [. . . ]” (Warner 1993: 203). dare, on the other hand, was an OE preterite-present verb taking mostly as its complement a plain infinitive or directional phrase.1 In the course of the ME period, it developed non-finite forms. The forms and constructions typical of a modal verb continued in ENE,
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.2 (98)
Martine Taeymans
but at the same time dare acquired the inflected forms dareth/dares and the tensed form dared. Beside a bare infinitive, it could also have a to-infinitive and in the 17th century dare to starts to occur with periphrastic do (Beths 1999: 1098; Visser 1970: §1364; Warner 1993: 202–203). So, while need was not a fully-fledged modal in the past, it developed modal behaviour (need) without completely discarding its more lexical counterpart (need to). This property is called divergence (Hopper & Traugott 2003: 116–120) and points toward grammaticalization. dare also shows divergence from the ENE period onwards, but has followed the opposite track: it remained to be used as a modal, but at the same time it developed main verb characteristics it had not displayed before (Beths 1999: 1069). As the unidirectionality claim central to grammaticalization seems to be countered here, dare’s development has been considered an example of degrammaticalization (Beths 1999). Variable usage with dare and need lasts until today and is the natural outcome of change in progress. Examples (1) and (2) show that need and need to both express notions of necessity and obligation with no noteworthy difference in meaning, while examples (3) and (4) illustrate the semantic closeness of dare and dare to: (1) But we needn’t worry too much about staff. I always thought NCPR had employed too many people. (Written BNC, GUU 4044) (2) “There’s no need to worry”, she said huskily, “I promise you’ll make everything all right so you don’t need to worry again.” (Written BNC, JYD 1289) (3) When he played for Sheffield United he told a local newspaper that he dared not go into city nightspots in case he ran into any Wednesday supporters. (Written BNC, CH1 8878) (4) The smell of paint was very strong. Soon Jasper said he couldn’t sleep in it and went upstairs. His tone was such that Alice did not dare to go with him. (Written BNC, EV1 1574) However, their distribution is very different: while modal dare and need are restricted to non-assertive contexts which include not only negative (cf. (1) and (3)) and interrogative sentences (cf. (5)), but also semi-negatives such as hardly and scarcely (cf. (6)), only (cf. (7)), conditional clauses (cf. (8)) and universal quantifiers (cf. (9)), the semi-modals dare to and need to are unrestricted in their use. Examples (10) and (11) show uses of dare to and need to in affirmative sentences: (5) What task may the Douglas ask of me, Dunbar and March? What task dare he ask? (Written BNC, CD8 1329)
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.3 (99)
An investigation into the marginal modals dare and need
(6) I hardly dare say this in this context, but if we . . . well now the Council has taken a decision to close this elderly person’s home, there is actually a logic to it, which I’d like to talk about. (Spoken BNC, KRL 2656) (7) The fact that Governors need only pay £20 flabbergasted me . . . anyone desiring the privilege of that title should certainly be asked to contribute £50 pa. (Written BNC, EDR 679) (8) Life is precious! And life is sweet. But there is no life, if you dare not test yourself, if you dare not feel free to live it to the full. (Written BNC, GUE 3268) (9) All eager students need do is to plough through its well-presented 600 pages and they will emerge as masters too. (Written BNC, B7G 1691) (10) Saddam is seen as the Third World leader who dared to challenge the established order imposed by the West, [. . . ] (Written BNC, G2J 1204) (11) Well in really warm weather a T-shirt and shorts may be enough but as it gets colder so you need to add layers of sweaters, trousers and top the whole thing up with a wind and waterproof spray suit [. . . ] (Spoken BNC, J3X 30)
. Aim In the following sections an attempt will be made to uncover how both dare and need have fared with respect to (de)grammaticalization. As for need, which apparently strengthened its modal and thus more grammatical use in the ENE period (Krug 2000: 202; Warner 1993: 203), it will be interesting to see whether modal use is still frequent. For dare, which started to feature main-verb like morphosyntax in the ENE period, I would like to find out whether and to what extent the use of semimodal dare to has spread over time. By classifying British PDE occurrences of need and dare as auxiliary, semi-modal or – in the case of dare – blend constructions and by looking at the frequencies of these respective variants, dare and need’s predominant use in British PDE will become apparent. Identifying their predominant use is important to uncover whether the apparent grammaticalization of need and degrammaticalization of dare have set through. Since grammaticalization has its source in language use, that is, in real discourse, spoken data on dare and need will be considered indicative of language change (Krug 2000: 7–11) and will be compared with more conservative written data. In doing so, we get a good view on recent developments in the behaviour of both verbs.
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.4 (100)
Martine Taeymans
It was shown in the introduction that context plays an important role in the distribution of dare and need: whereas modal dare and need are restricted to non-assertive contexts, dare to and need to can always be used. Therefore, any remarkable changes in frequency will be further explored in relation to context: what is particularly of interest is where and to what extent growth is noted.
. The classification of dare and need: Formal features Before we move to the frequency analysis, it is necessary, first, to set out criteria for the classification of the modals dare and need, and then go to deal with items that are in various ways intermediate between modal and full verb status, i.e. the semi-modals dare to and need to, and the blend constructions with dare. Quirk et al. (1985: 121–128) have identified a set of criteria to define (modal) auxiliaries. This set defines the primary auxiliaries be, have and do and, allowing for a few anomalies, the central modal auxiliaries. In Table 1, the criteria identified by Quirk et al. are summarized and applied to dare and need.2 An example is supplied whenever a criterion is met. When we examine Table 1, it becomes clear that both modal dare and need live up to almost all of the criteria typical of (modal) auxiliaries. Only criteria (d) ‘emphatic positive’ and (e) ‘operator in reduced clauses’ are not completely satisfied: since modal dare and need mainly occur in non-assertive contexts, they are unsuitable to mark a finite clause as positive rather than negative, hence a minus for (d); for the same reasons, only the negative forms daren’t and needn’t appear as operators in reduced clauses, hence an approximation sign for criterion (e). Dare is also anomalous when it comes to criteria (h) ‘semantic independence of subject’ and (m) ‘abnormal time reference’: it fails the independence of subject criterion as “[t]here is no semantic equivalence between the active The boy daren’t contact her and the passive She daren’t be contacted by the boy” (Quirk et al. 1985: 139). As to criterion (m), dare in auxiliary use does have a modal past form ‘dared’, unlike the other central modals. Beyond the modal variants of DARE and NEED, there are the semi-modals dare to and need to which are semantically related to the central modals, but pattern entirely like main verbs in: a.
showing regular verb morphology (i.e. 3sg present -s, and past forms) as in (12) and (13); (12) Somebody needs to be a contact for getting all the information to by a certain date. (Spoken BNC, D97 880) (13) “No one dared to say anything.”
(Written BNC, ADD 711)
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.5 (101)
An investigation into the marginal modals dare and need
Table 1. Criteria for (modal) auxiliary verbs (adapted from Quirk et al. 1985: 137) need dare Examples from the BNC AUXILIARY CRITERIA (a) Operator in negation with not (b) Negative contraction
+
+
You daren’t touch it. (Spoken BNC, KBE 6240)
+
+
+
+
– +/–
– +/–
+
+
+
+
I needn’t go to school on Thursday so I can have a lie in. (Spoken BNC, KC2 1309) Dared she stop and wish him a Happy Christmas? (Written BNC, BM4 808) – “Well, need I point out?” “No, you needn’t, ma’am.” (Written BNC, CK9 2367) But knowledge is something which has achieved the certainty of being known and proved, and therefore need never be doubted. (Written BNC, C8V 431) Not attested
+
–
[N]o contribution need be made by the defendant to his or her own costs. (Written BNC, GVH 965)
MODAL AUXILIARY CRITERIA (j) Bare infinitive
+
+
(k) No non-finite forms (l) No -s form
+ +
+ +
(m) Abnormal time reference
+
–
[S]he knew there was an image of someone, but she dare not look too closely. (Written BNC, JYE 2319) Cf. (1), (3), (5), (6), (7), (8) and (9) If the Home Secretary is worried about cautioning he need look no further than Folkestone, his own constituency, and the rest of Kent. (Written BNC, CDA 1109) Tears started to trickle down Donna’s face and Juliet had to swallow hard. She dared not cry on duty. (Written BNC, JY0 6387)
(c) Operator in inversion (d) Emphatic positive (e) Operator in reduced clauses (f) Pre-adverb position
(g) Post-position of quantifier (h) Semantic independence of subject
b. having non-finite forms (i.e. an infinitive and present/past participles) as in (14)–(16); (14) Why do I need to practice that with you counting numbers up to thirtyone? (Spoken BNC, JA8 49) (15) He shivered, not daring to move in case they noticed and dragged him out. (Written BNC, C85 2189)
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.6 (102)
Martine Taeymans
(16) Mr. Grant boldly went where no leadership representative had dared to go. (Written BNC, C95 933) c.
taking do-support in interrogative, negative and positive emphatic constructions, exemplified in (14), (17) and (18) respectively; (17) You do not need to be very rich;
(Written BNC, CMK 452)
(18) Smoke alarms need very little maintenance, but they do need to be tested regularly to make sure they are in good working order. (Written BNC, A0J 173) d. taking a to-infinitival complement (cf. examples (12)–(19)); e. co-occurring with central modals, as in example (19). (19) And they said within fifty years false teeth will be a thing of the past. Nobody will need to wear them any more. (Spoken BNC, KBK 1375) For dare a third variant needs to be distinguished, i.e. blend dare. It blurs the definitional criteria above by blending features typical of auxiliaries (e.g. a bare infinitival complement) and main verbs (e.g. 3sg inflection or do-support). Two types of blend constructions are attested in the corpus. In the first type the bare infinitive is found with dares and non-finites of dare, as in examples (20) and (21): (20) The government has failed to add any convincing arguments to its feeble justifications for the changes, and dares not come clean about the real reasons. (Spoken BNC, AAP 74) (21) She was everything he had dared hope for, and more. (Written BNC, EWH 458) The second and most common type of blend use combines do-support or auxiliary negation with a bare infinitival complement in negative and interrogative clauses. This is illustrated in (22) and (23): (22) [D]on’t you dare come over smoking!
(Spoken BNC, KC0 1429)
(23) [Y]ou wouldn’t dare answer the teacher back like they did now and call them by their Christian names. (Spoken BNC, H4B 604) The blending of properties from more than one class is not unusual: it is indicative of forms in the process of being re-assigned to a different category.
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.7 (103)
An investigation into the marginal modals dare and need
. Material The data studied are provided by the British National Corpus World Edition (BNC), which contains approximately 10m words of spoken and 90m words of written British PDE. Given the size of the corpus, only parts of the BNC were studied. For need, random samples of approximately 2000 occurrences were taken from the written and spoken parts of the BNC respectively; then, nominal and adjectival uses were filtered out. Table 2 shows the amount of verbal occurrences found for need. Transitive uses of need + nominal complement are included in the table for reasons of completeness. However, only modal and semi-modal constructions will be taken into account in the following sections, as they are competing for overlapping territories. The same sampling procedure was followed for dare in the written BNC. Since it appeared to be a rather infrequent token in the spoken corpus, I looked at all available occurrences. Table 3 shows the verbal uses studied for dare in the written and spoken BNC samples. Here, too, the transitive uses of dare + nominal complement are given, but will further be ignored in the analysis. When sampling, I did not make a selection as to text types or register: the random samples obtained from the written BNC contain a variety of text types and media; the spoken sample contains monologues as well as dialogues, and informal as well as formal encounters.
Table 2. Verbal uses of need in the written and spoken BNC samples3 modal % of total semi-modal % of total need + NP % of total total need need to Written BNC 84 Spoken BNC 40
4.8% 1.9%
701 852
39.8% 40.9%
965 1186
54.8% 57%
1750 2078
Table 3. Verbal uses of dare in the written and spoken BNC samples modal % of dare total Written BNC 853 Spoken BNC 192
blend % of dare total
47.8% 321 59.1% 77
semi-modal % of dare to total
17.9% 433 23.7% 28
dare + NP % of total total
24.3% 178 8.6% 28
10% 1785 8.6% 325
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.8 (104)
Martine Taeymans
. The status of need and dare in British PDE . The status of need in the BNC Both the data from the written and spoken samples in Table 4 prove that semimodal need to clearly is the dominant form in British PDE (701 in the written sample; 852 in the spoken sample). Modal use appears to be rare, even more so in the spoken than in the written sample (40 vs. 84 occurrences respectively). If discourse is indeed indicative of language change, it seems probable that need will further decline, while need to will further strengthen its position. Note that all modal occurrences appear in non-assertive contexts, with a remarkable preference for direct not-negation: 23 out of 40 modal occurrences attested in the spoken sample take the form needn’t and 6 take the form need not; in the written sample 51 out of 84 are of the need not type, while only 9 show contraction with not (needn’t). While an investigation into the use of need in Shakespeare by Krug (2000: 202) reveals that, by the end of the 16th century, “[. . . ] modal constructions by far outnumber main verb constructions [. . . ]”, we are now confronted with a situation where modal usage has almost disappeared in favour of the semi-modal construction. This movement from auxiliary-like to main verb-like morphosyntax seems to contradict the unidirectionality claim of grammaticalization. However, quite a few scholars studying grammaticalization phenomena have argued that developments such as the one noted for need are not legitimate counterexamples to grammaticalization (Haspelmath, this volume; Traugott 2001; Traugott & Dasher 2001). The main argument put forward is that we do not witness the rise of a more lexical variant, but that we have a situation where “[. . . ] the earlier main verb use was marginalized in the early periods and then the grammaticalized one was marginalized in turn and then lost in the later periods (Traugott 2001: 9)”. In this view, the apparent rise of modal need in the Early Modern Period was just ‘a short-lived innovation’. Haspelmath (this volume) has recently coined the term ‘retraction’ to refer to this type of development. Another argument was formulated by Traugott and Dasher (2001: 87), who acknowledge that unidirectionality has been challenged, most especially in the domain of morphosyntax. However, they remark that “[e]ven if structural unidirectionality is violated, semantic unidirectionality is not.” It will be shown in Section 5.1.1 that this also seems to be the case with need. Table 4. need +(to)infinitive in the written and spoken BNC samples
Written BNC Spoken BNC
modal need
% of total
semi-modal need to
% of total
total
84 40
10.7% 4.5%
701 852
89.3% 95.5%
785 892
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.9 (105)
An investigation into the marginal modals dare and need
.. The rise of need to in context In this section we take a closer look at the rise of need to in context. It was shown in the introduction that the distribution of need and need to is by no means uniform: whereas the use of need is restricted to non-assertive contexts such as negative and interrogative sentences, need to is free to appear in both non-assertive and affirmative contexts. Particularly of interest are the contexts where need to has become most frequent, and the semantic change enabling the rise of the semi-modal construction. Since need and need to only truly compete with one another in the simple present and simple past tense, all modally premodified constructions with need to (e.g. will need to, would need to, might need to) are excluded from the analysis in this section. ... (Implicitly) negative contexts. Tables 5–7 provide frequencies for need and need to in negative, interrogative and other non-assertive contexts. A comparison of written and spoken data convincingly shows that need to has largely taken over the role of need in all of these (implicitly) negative contexts. This development implies that semantic differences between need to and need have been neutralized, so that need to, having the literal sense of internal necessity (Smith 2003: 244; Traugott & Dasher 2001: 110; van der Auwera & Plungian 1998: 80), has also come to function as a modal marker denoting external necessity.4 This is best illustrated in cases with overt negation: like needn’t, don’t need to can be used when an external authority or external circumstances do not require the performance of an action. As a result, it is possible to substitute the semi-modal construction don’t need to for external necessity markers such as needn’t and don’t have to in PDE. This is neatly shown in examples (24) through (26), where needn’t, don’t need to and don’t have to are used interchangeably: (24) “Cathy do we have to draw these pictures?” [. . . ] “No, you needn’t [. . . ]” (Spoken BNC, F8L 477) (25) There are lots of things to attack the Conservatives on and it is a question of how that is actually done. We don’t need to make things up. We don’t have to. All we have to do is tell it as it actually is and then trust the people themselves to decide. (Spoken BNC, JJD 271) (26) In order to believe in the Devil we must rid ourselves of unhelpful images of him (though we do not need to wax philosophical about him as I did for the doubting lady!). We do not have to define him, or even believe that he is a ‘person’, in order to believe that he is real. (Written BNC, CCE 202) Needn’t, which used to fill a gap in the modal paradigm denoting absence of necessity, has largely been replaced by the semi-modals need to and have to in British PDE. This is not surprising: layering, i.e. the fact that more than one technique is
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.10 (106)
Martine Taeymans
Table 5. need +(to)infinitive under negation in the written and spoken BNC samples
Written BNC Spoken BNC
needn’t
% of total
don’t need to present past
% of total
total
47 29
61% 27.1%
26 74
39% 72.9%
77 107
4 4
Table 6. need+(to)infinitive in interrogative constructions in the written and spoken BNC samples
Written BNC Spoken BNC
inversion (Need X . . . ?)
% of total
do-support (Do X need to . . . ?) present past
% of total
total
3 2
27.3% 5.4%
5 32
72.7% 94.6%
11 37
3 3
Table 7. need +(to)infinitive in other non-assertive contexts in the written and spoken BNC samples5 modal need in other non-ass. contexts Written BNC Spoken BNC
34 9
% of total
60.7% 13.4%
need to in other non-ass. contexts present
past
16 53
6 5
% of total
total
39.3% 86.6%
56 67
available in the language to serve similar or identical functions (Hopper 1991: 22– 23), and the subsequent gradual replacement of older forms by (more) periphrastic expressions is the norm, especially in the domain of English modality: consider for example will vs. going to; must vs. have to; should vs. had better (cf. Biber et al. 1998: 205–210; Krug 2000; Smith 2003: 263–264). From external necessity it is but a short step to compulsion or obligation; it will be shown in 5.1.1.2 that need to is actually found in affirmative sentences as a somewhat weakened must or have to.
... Affirmative sentences. Table 8 shows the use of need to in affirmative sentences. A comparison with modal need is not possible because it simply does not occur in truly affirmative sentences. Frequencies show that affirmative need to is very common in the written and spoken language. Although the number of occurrences attested in the spoken sample is not remarkably higher than the ones noted in the written sample, a short-term diachronic study conducted by Smith (2003: 260) revealed that “[. . . ] most of the increased uses of NEED TO are in affir-
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.11 (107)
An investigation into the marginal modals dare and need
Table 8. need to in affirmative sentences in the written and spoken BNC samples
Written BNC Spoken BNC
affirmative need to present past
total
377 491
430 504
53 13
mative contexts. I believe that sometimes in these contexts it competes with MUST and HAVE TO [. . . ].” Whereas strong deontic markers such as must are mostly associated with direct speaker authority, Smith (2003: 260) claims that need to can function as a strong deontic marker in disguise: [Need to] can acquire the force of an imposed obligation, but – something which does not apply to the other markers – the writer or speaker can claim that the required action is merely being recommended for the doer’s own sake. This ambiguity allows the writer or speaker to appear more cognizant of individuals’ requirements, and at the same time downplay his or her own authority.
Examples (27) through (31) indeed suggest that the necessity or obligation expressed by need to is in the addressee’s best interest and at the same time in accordance with the speaker’s wish: (27) “[T]he practice manager will do his or her . . . her best to list out as many businesses which with they have links or are local. Most important thing when you receive this on . . . on day one of the assignment . . . er Douglas, is to not just accept that list as it is. You need to sit down with the practice manager or whomever and say right now you’ve listed all these, why, what’s peculiar about . . . what’s peculiar and peculiar about these [. . . ]” (Spoken BNC, J9Y 406) (28) “[I]f he’s gonna need the grommets or something in his ear then he really needs to get it sorted out this year.” “Yeah.” “I don’t want him to start in the . . . January and then have to be off school for three weeks because he’s got to go to hospital to have his ears done and things like this, you know?” (Spoken BNC, KBG 3334) (29) “[. . . ] four of you will arrive at the beginning of the session, and four of you will arrive half way through. And those of you who ar- who arrive half way through will need to wait outside. All the times that you need to arrive at you will find on the timetable. (Spoken BNC, HUU 117) (30) Now I think that the whole school meals issue, not just this, but also the way in which cost savings are being achieved at the moment in schools.
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.12 (108)
Martine Taeymans
Erm needs looking at by members. We need . . . members need to get a grip on what is happening and my wife suggests this, that the panel erm rather than setting up a new panel, look at the whole issue of school meals, this and other issues. (Spoken BNC, JWA 282) (31) “We already keep watch on all the air and sea ports”, he reminded his boss. “And I study the lists of known visitors.” “That’s the point, Frank. We need to be on the look-out for unknown visitors. Use our travel company’s film unit. We have reached a sensitive stage in the operation. So get pictures taken of everyone coming in. Then study the films they take.” (Written BNC, CN3 3325) Pragmatic inferencing plays an important role in the above account. As Heine et al. (1991: 113) note, “[. . . ] [inferencing] leads to the emergence of overlapping senses [. . . ]” and this adequately explains what is going on with need to: in its literal sense, need to expresses necessity or obligation that is internally motivated, that is, for the agent’s or doer’s sake. This has been exemplified by sentences such as “Boris needs to sleep ten hours every night for him to function properly” (Traugott & Dasher 2001: 110; van der Auwera & Plungian 1998: 80), where need to indeed focuses on the internal or personal motivation for the requirement (Westney 1995: 98). In examples (27)–(31), this internal quality is exploited on the part of the speaker: while the requirement expressed appears to be in the addressee’s best interest, it really is an indirect order or instruction given by the speaker. In this sense, need to combines its internal necessity meaning with an external necessity or obligation reading and comes to be associated with functions similar to have to and must in affirmative contexts. This is best exemplified in (32), where the speaker uses have to and need to in the same context: (32) “No, th. . . the till man sai[d] we have to keep erm you know the receipt one . . . ” “Yeah, but we never put that on do we?” “I know. But we need to keep er a a receipt in it.” (Spoken BNC, KBD 8624) According to van der Auwera and Plungian’s (1998: 111) semantic map of modality, a move from internal to external necessity is in accordance with the unidirectionality claim of grammaticalization. While the development of need violates the unidirectionality hypothesis by adopting the morphosyntax typical of main verbs (need to), it does not show a reverse semantic shift. Need to has become more frequent in all types of syntactic environment and has acquired senses of external necessity so that it comes to compete with need, must and have to. Pragmatic inferencing is the mechanism behind the move from internal to external necessity. While need is obsolescent, need to has become functionally important in the modal domain of necessity and obligation.
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.13 (109)
An investigation into the marginal modals dare and need
. The status of dare in the BNC It emerges from Table 9 that modal dare is still going strong in British PDE. Especially in the spoken sample, the modal is far more frequent than any of its competing variants, accounting for 64.6% of the total of occurrences. If we compare the more conservative written data with the more innovative spoken ones, it seems that dare to is losing ground, especially in favour of the modal construction. The modal predominance of dare in British PDE was earlier noted by Hundt (1998: 65) and Krug (2000: 201), whose corpus-based research on dare “[. . . ] yielded no trend towards main verb usage on the diachronic axis.” Although overwhelming evidence points to the fact that dare in British PDE is very modal indeed, it remains somewhat puzzling, given the fact that dare has been adopting unequivocal lexical characteristics since the Early Modern English period. A partial explanation for the modal predominance of dare in British PDE may lie in the relatively frequently occurring frozen constructions I dare say’ / ‘dare I say (it), and the idiomatic construction How dare you (+ bare infinitive) as shown in examples (33)–(35): (33) I dare say somebody’s already said it to you but the family were very comforted and grateful for your service. (Spoken BNC, KB0 2410) (34) And maybe, dare I say it? Maybe if you were a clever rivet boy you you could get more wages. (Spoken BNC, G63 116) (35) How dare you use such a northern expression in the middle of the east? (Spoken BNC, KBG 3774) Although I dare say is clearly modal in structure, it enjoys idiomatic status due to repetition, and is frequent enough to qualify as being marginally grammatical, or even lexicalized, meaning ‘I assume/presume that; I think it likely that’ (cf. (33)– (34)). How dare you (+ bare infinitive) is similar to I dare say in that it also is a frequent, entrenched phrase, meaning that “you are very angry and shocked about what someone has done or said (Summers 2003: 395)” (cf. (35)). Although the context or structure is interrogative, the meaning is not: the speaker merely expresses his annoyance at something. Traugott and Heine (1991: 8) note that frozen idioms or lexicalizations may block or delay the process of total loss. Applied to dare, this Table 9. dare +(to)infinitive in the written and spoken BNC samples modal % of total blend dare dare Written BNC 853 Spoken BNC 192
53.1% 64.6%
321 77
% of total semi-modal % of dare to total 20% 26 %
433 28
total
26.9% 1607 9.4% 297
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.14 (110)
Martine Taeymans
Table 10. dare +(to)infinitive in the written and spoken BNC samples, excluding I dare say / dare I say it and How dare you (+ bare infinitive) modal % of dare total Written BNC 418 Spoken BNC 114
blend dare
35.6% 321 52% 77
% of total semi-modal % of total total dare to 27.4% 35.2%
433 28
37% 12.8%
1172 219
would mean that the phrases I dare say and How dare you tend to resist change and are thus largely responsible for the conservation of modal dare. But even if we leave these frozen constructions out of account, modal dare still accounts for 52% of the spoken occurrences and 35.6% of the written occurrences. This is shown in Table 10. The remaining modal occurrences are exclusively found in non-assertive contexts, with a strong preference for direct not-negation: 84 out of 114 occurrences in the spoken sample are negative auxiliaries of the form daren’t, 5 take the form dare not and 2 occurrences combine the tensed form dared with not; in the written sample 60 out of 418 occurrences are of the daren’t type, 74 of the dare not type and 84 of the type dared not. The importance of blend constructions with dare in British PDE should not be underestimated either, accounting for 27.4% and 35.2% of the total occurrences in the written and spoken BNC respectively. The fact that dare moved away from the central modals in Early Modern English by gradually adopting the morphosyntactic properties of main verbs has led to the emergence of these linguistic hybrids or blends. According to Lichtenberk (1991: 38–39), tokens exhibiting properties characteristic of different categories may be in the process of being re-assigned to a different category. The problem with dare is that the present corpus data do not clearly indicate to what category it is actually shifting. The discourse frequency data in Table 10 seem to predict that modal use is growing (52% in the spoken vs. 35.6% in the written sample) and that semi-modal dare to is on its way out (12.8% in the spoken vs. 37% in the written sample). They suggest that we have little reason to believe that modal dare – like need – will become obsolete anywhere in the near future. However, the total number of spoken occurrences studied is probably too small to implicitly rely on discourse frequency as an indicator of change. Further growth of modal dare also seems counter-intuitive in the light of the following two facts: a.
older vs. periphrastic forms in the domain of English modality If one looks at the domain of English modality in general, it seems doubtful that modal dare is heading for another strong revival at the cost of the more recent semi-modal dare to. Layering and the subsequent gradual replacement of older forms by (more) periphrastic expressions is the norm in the domain
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.15 (111)
An investigation into the marginal modals dare and need
of English modality (Biber et al. 1998: 205–210; Krug 2000; Smith 2003: 263– 264). Bybee, Perkins and Pagliuca (1994: 8) argue along the same line that “[i]t is also typical of grammatical or closed classes to reduce further in size. Individual members are lost, usually by one member generalizing to take over the functions of other members.” b. the status of dare in American English In American English dare has largely acquired main-verb features such as dosupport and to-infinitives, while modal use with dare has become very rare (cf. Quirk et al. 1985: 138). It is a common assumption that American English is more advanced and open to change than British English (Krug 2000: 135). Therefore, it is not unthinkable that British English will follow the American trend. The frozen constructions discussed above are also interesting in this respect: a brief search in the American corpora BROWN (1961) and FROWN (1991) revealed that I dare say / dare I say it are no longer attested in American writing. It would be interesting to investigate whether I dare say / dare I say it was ever functional in American English and, if so, whether there is any link between the disappearance of the frozen idiom and the retreat of modal dare in general. The assumption that frozen idioms delay the process of loss in British English is an interesting one and deserves further looking into. We can conclude from this section that dare is not evidently moving in a clearly identifiable direction in British PDE: on the one hand, discourse frequencies indicate that modal dare is firmly established in British PDE and might even increase, but on the other hand we do have reasons to believe that the modal variant will become obsolete. As it is now, dare and dare to represent older and newer layers in the language that serve similar functions. Since the frequency analysis was inconclusive, I refrain from making any definite claims about the (de)grammaticalization of dare.
. Conclusion In this paper I described the current behaviour and recent changes in the marginal modals dare and need by looking at frequency data taken from the spoken and written British National Corpus. Data on need revealed that semi-modal need to has become the dominant form in British English. While modal need had become very frequent by Shakespeare’s time, it is obsolete in British PDE. This movement from auxiliary-like to main verb-like morphosyntax breaches the unidirectionality claim of grammaticalization. However, while need structurally violates the unidirectionality hypothesis, it does not seem to show a reverse semantic shift. It was shown that need to can take advantage of its literal meaning of internal
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.16 (112)
Martine Taeymans
necessity to operate as an indirect marker of external necessity and obligation in non-assertive and affirmative contexts. In doing so, it has largely taken over from modal need, but also comes to compete with have to and must. According to van der Auwera and Plungian’s (1998: 111) semantic map of modality, a shift from internal to external necessity is in accordance with the unidirectionality claim of grammaticalization. The mechanism enabling this move from participant-internal to participant-external necessity is pragmatic inferencing. It has proven difficult to make any definite claims about the development of dare. Although discourse frequencies indicate that modal dare is firmly established in British PDE, it seems doubtful that modal dare is heading for a strong revival at the cost of semi-modal dare to: both the status of dare in American English and the general tendency to replace central by periphrastic modals were invoked to call the rise of modal dare into question. More data and research are needed to shed light on the (de)grammaticalization of dare.
Abbreviations BNC = British National Corpus BROWN = Brown Corpus of Standard American English ENE = Early Modern English FROWN = Freiburg Version of the Brown Corpus ME = Middle English OE = Old English PDE = present-day English
Notes . Preterite present verbs are a group of originally strong Old English verbs whose original preterite forms came to be used as present tense forms, while new weak preterite forms and past participles were introduced. The most important verbs in this category are witan, agan, *dugan, unnan, *durran, *sulan, *motan and magan. Most of them eventually developed into central modal verbs. . For a full account of the different criteria the reader is referred to Quirk et al. (1985: 121– 127). . In the written sample, 3 occurrences of blend need where encountered, showing a mixture of auxiliary and main verb syntax. In the spoken sample, one example of blend need was found. Since these uses appear to be extremely marginal, they will be ignored in this paper. . Modal need can be used as an epistemic modal, but the majority of modal occurrences with need have deontic meaning.
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.17 (113)
An investigation into the marginal modals dare and need . Conditional clauses, clauses with shifted negation, universal quantifiers, only, seminegatives hardly or scarcely, comparative and superlative clauses and indirect questions of the type ‘I wonder/doubt wheter . . . ’ are considered non-assertive and are included in the count.
Text sources The British National Corpus (BNC), current British English, comprising 10m words of spoken and 90m words of written British PDE. Brown Corpus of Standard American English (BROWN), texts from 1961, 1m words of written American English. Freiburg Version of the Brown Corpus (FROWN), texts from 1991, 1m words of written American English.
References Beths, Frank (1999). “The History of dare and the Status of Unidirectionality.” Linguistics, 37, 1069–1110. Biber, Douglas, Susan Conrad, & Randi Reppen (1998). Corpus Linguistics: Investigating Language Structure and Use. Cambridge: Cambridge University Press. Bybee, Joan L., William Pagliuca, & Revere D. Perkins (1994). The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: University of Chicago Press. Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer (1991). Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Hopper, Paul J. (1991). “On some Principles of Grammaticization”. In Traugott & Heine (Eds.), Vol. I (pp. 17–34). Hopper, Paul J. & Elizabeth C. Traugott (2003, 2nd ed.). Grammaticalization. Cambridge: Cambridge University Press. Hundt, M. (1998). New Zealand English Grammar – Fact or Fiction? A Corpus-Based Study in Morphosyntactic Variation. Amsterdam & Philadelphia: John Benjamins. Krug, Manfred (2000). Emerging English Modals. A Corpus-Based Study of Grammaticalization. Berlin & New York: Mouton de Gruyter. Lichtenberk, Frantisek (1991). “On the Gradualness of Grammaticalization”. In Traugott & Heine (Eds.), Vol. I (pp. 37–80). Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, & Jan Svartvik (1985). A Comprehensive Grammar of the English Language. London: Longman. Smith, Nicholas (2003). “Changes in the Modals and Semi-Modals of Strong Obligation and Epistemic Necessity in Recent British English”. In Roberta Facchinetti, Manfred Krug, & Frank R. Palmer (Eds.), Modality in Contemporary English (pp. 241–267). Berlin & New York: Mouton de Gruyter.
TSL[v.20020404] Prn:13/04/2004; 15:57
F: TSL5905.tex / p.18 (114)
Martine Taeymans
Summers, Della (Ed.). (2003). Longman Dictionary of Contemporary English. Longman: Harlow. Traugott, Elizabeth Closs (2001). “Legitimate Counterexamples to Unidirectionality”. Paper presented at Freiburg University, Germany, October 2001. http://www.stanford.edu/∼ traugott/papers/Freiburg.Unidirect.pdf Traugott, Elizabeth Closs & Bernd Heine (1991). “Introduction”. In Traugott & Heine (Eds.), Vol. I (pp. 1–35). Traugott, Elizabeth Closs & Bernd Heine (1991). Approaches to Grammaticalization. (=Typological Studies in Language, 19 (2 Vols.)). Amsterdam & Philadelphia: John Benjamins. Traugott, Elizabeth Closs & Richard B. Dasher (2001). Regularity in Semantic Change. Cambridge: Cambridge University Press. van der Auwera, Johan & Vladimir A. Plungian (1998). “Modality’s Semantic Map”. Linguistic Typology, 2(1), 79–124. Visser, Frederikus Th. (1970). An Historical Syntax of the English Language (3 Vols.). Leiden: Brill. Warner, Anthony R. (1993). English Auxiliaries: Structure and History (=Cambridge Studies in Linguistics, 66). Cambridge: Cambridge University Press. Westney, Paul (1995). Modals and Periphrastics in English: An Investigation into the Semantic Correspondence between certain English Modal Verbs and their Periphrastic Equivalents. Tübingen: Niemeyer.
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.1 (115)
Redefining unidirectionality Is there life after modality? Debra Ziegeler University of Manchester
.
Introduction
The hypothesis of unidirectionality (which states that in any given grammaticalisation path, stage A always precedes stage B, and not vice versa) has been seen in earlier literature, e.g. Heine, Claudi and Hünnemeyer (1991), Hopper and Traugott (1993), as fundamental to the existence of grammaticalisation as an independent entity. Recent studies, however, are laced with instances of purported counterexamples to unidirectionality (i.e. illustrating degrammaticalisation), attempting to undermine the theoretical base upon which most grammaticalisation studies are built (e.g. Beths 1999; Janda 2001; Newmeyer 1998; Norde 2001; Ramat 1992; and van der Auwera 2002). In general, there has been little rapid response from those engaged in research in grammaticalisation, apart from, for example, Haspelmath (1999), and perhaps Klausenburger (2000), both of whom, though offering commendable accounts, are still not tackling the tougher issues such as how to accommodate the numerous counterexamples to unidirectionality and yet maintain that grammaticalisation is an autonomous chain of events, distinct from other processes of language change. The present paper will consider the problems of bidirectionality and its manifestation in the examples given in the recent literature, and endeavour to present an alternative viewpoint wherein unidirectionality is seen only as an underlying semantic phenomenon. Using historical data, an instance of demodalisation in Mandarin Chinese will be investigated as a case study, in order to support an extended hypothesis by which semantic unidirectionality in grammaticalisation may be regarded as originating in a series of unidirectional metonymic shifts framed within the wider limits of conceptual category expansion. Within such expansion, a specific sub-series of changes is identifiable as grammat-
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.2 (116)
Debra Ziegeler
icalisation. In this respect, morphological changes must be considered secondary to semantic forces.
. Defining unidirectionality . The semantic underpinning of unidirectionality In much of the earlier literature, the unidirectionality hypothesis in grammaticalisation has been held to be a strong one (Hopper & Traugott 1993: 128), due to the fact that the number of bi-directional counterexamples found were considered to be so few by comparison to the number of unidirectional examples, and since the cases that demonstrate that grammatical morphology may arise without undergoing development through earlier stages in which it had lexical functions were also thought to be relatively rare. Heine, Claudi and Hünnemeyer (1991: 5) found that the number of counter-instances (demonstrating either degrammaticalisation or regrammaticalisation) were not significant enough to be worthy of much discussion. In more recent studies, though, there has been an upsurge of studies of bidirectionality cases which are worthy of much closer scrutiny. Heine Claudi and Hünnemeyer focus attention on the unidirectional tendencies in the conceptual shifts associated with grammaticalisation processes, which move from concrete to abstract, and not in the opposite direction (1991: 31). In describing the development of grammatical morphemes as a metaphorical process, they maintained that the same unidirectionality is not found in metaphorical conceptual shifts within the lexicon (1991: 50). However, Levinson (1989: 160) notes that in general, ‘good’ metaphors usually employ a concrete term to describe an abstract one, as illustrated in (1a) (1989: 160): (1) a.
Love is a flame.
in which a concrete concept, flame, is used to describe to an abstract one, love. The reversal of such terms in (1b), in which an abstract term, love, is made to stand for a concrete one, a flame, is less successful: b. A flame is love. Although the infelicity of such sentences may be brought about by the presence of an indefinite subject in (1b),1 it could also due to the expectation that wherever concrete concepts are recruited to serve abstract functions, the same unidirectionality might apply, and this type of unidirectional semantic change should therefore not be taken as an exclusive, defining feature of grammaticalisation alone (Traugott 1994: 1483).2 Whether the same unidirectionality applies to metonymic shifts in the lexicon is another matter for discussion, but when metonymy is recruited as
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.3 (117)
Redefining unidirectionality
part of the mechanism motivating grammaticalisation changes, it is only onedirectional metonymic shifts that are relevant. Semantic unidirectionality, though, does not always need to be described in terms of movement towards more abstract conceptual levels, as will be seen below. In examining the semantic level of unidirectionality, it must be assumed from the start that grammaticalisation begins with conceptual changes, and that these are prior to and primary to other changes taking place. Heine, Claudi and Hünnemeyer (1991: 15) discussed the fact that chronologically, functional change will be more likely to precede structural change, and desemanticisation will precede cliticisation, if cliticisation is involved. Traugott (2000) discusses the necessity for a pragmatic polysemy (a Generalised or language-specific Invited Inference) to take place as a minimal requirement before grammaticalisation can begin. One possible counter-argument for semantic priority in grammaticalisation comes from Fischer (1994) in which the development of the semi-modal verb of obligation, have to, in English, initiates via juxtaposition with the infinitive in diachronic word-order shifts. However, while syntactic change may often coincide with semantic change, it is not yet clear whether there is sufficient evidence for the claims for semantic precedence to be challenged. In the present study, it will be shown that the semantic level is perhaps the only reliable level at which unidirectionality in grammaticalisation may be discussed, and that there may be stages after grammaticalisation (post-grammaticalisation stages which could be labelled degrammaticalisations by contenders for bidirectionality),3 which are transgressed by semantic unidirectionality.
. Examples of degrammaticalisation Many of the counterexamples to unidirectionality have been found in the re-use of grammatical material for lexical purposes, such as if and but as nouns and adjectives, up and down as verbs. Hopper and Traugott (1993: 49) list such cases as simple lexicalisations, e.g. to up the ante (others include to up the prices, or to down a glass of beer), suggesting that they represent a type of reanalysis in which grammatical material can be used for lexical purposes. The literature is replete with other examples of lexicalisations justifying bidirectionality, though van der Auwera (2002) considers lexicalisation as only inclusive within a broad definition of degrammaticalisation. Examples of the non-lexicalising, narrow definition of degrammaticalisation include the genitive suffix -s in English and Scandinavian languages (Norde 2001) and Rosenbach (this volume), in which an affix is said to return to clitic status (shown to be an instance of grammaticalisation by Tabor & Traugott 1998), or the Spanish modal auxiliary of necessity, which may be expressed as deber (+ infinitive) or deber de (+ infinitive), the latter use being the
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.4 (118)
Debra Ziegeler
more recent,4 and the reinforcement of the English and German infinitive markers, which illustrates a degrammaticalisation of the infinitive (van der Auwera (2002).5 Van der Auwera goes further than most accounts to suggest that reinforcement in grammaticalisation is also a subtype of degrammaticalisation; such examples, though, must be viewed only over a period of time: what we have to observe is whether, in the long term, the entire process of which it is a part is still moving along a path toward the creation of grammatical structure, and not conceptual material. Other examples of lexical innovations deriving from grammatical material include, e.g. Ramat (1992), who refers to forms such as ism, gate, and ade as lexicalisations and an ‘aspect’ of grammaticalisation, proving that grammar is not the endpoint of language development (1992: 557). Benveniste (1958) has referred to them as delocutives, as noted by Eythórssen et al. (2002), and Haspelmath (this volume). Lehrer (1996) identifies such forms as ‘splinters’ – parts of words which break off and form a life of their own, becoming productive through semantic association with their original hosts. Newmeyer (1998: 269–270) adds such instances to the list of lexicalisations, describing them as ‘rampant upgrading’, earlier stating that any type of upgrading should count as a counterexample to the unidirectionality hypothesis (p. 263). However, Hagège (1993: 209–210) discusses similar examples, maintaining though that they are not the reverse of grammaticalisation, and that the unidirectionality of grammaticalisation does not mean that morphemes cannot re-lexicalise. His definition of degrammaticalisation, therefore, extends only as far as suggesting that degrammaticalisation as a mirror-image reversal of grammaticalisation cannot be predicted with any specific formal or semantic critieria; i.e. re-lexicalisation is possible, but not in the same direction from which the morpheme first developed. Traugott (2001) also disagrees with the claims that such examples are counterexamples to grammaticalisation, having more to do with the processes of word-formation, since the processes involved are unlike those of grammaticalisation, e.g. they are instantaneous (it remains to be explored how instantaneous such processes might be). Thus, although such forms clearly seem to exemplify a simple instance of lexicalisation, it cannot be assumed that they are degrammaticalisations unless word-formation processes may also count as degrammaticalisation. The much-discussed example of the genitive inflection in English and Scandinavian languages is justified by apparent backward progression along the cline of grammaticality referred to by Hopper and Traugott (1993: 7): content word > grammatical item > clitic > inflectional affix. Although clitics and affixes represent adjacent stages on a cline of grammaticality, the differences between them in terms of levels of semantic integrity are few. Other cases of clitics developing out of affixes include that of the Irish 1st person plural subject suffix muid developing into an independent pronoun, as cited by Bybee, Pagliuca and Perkins (1994: 14),
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.5 (119)
Redefining unidirectionality
probably by analogy with the remainder of the paradigm, which has already dispensed with person suffixes and reverted to a system of independent pronouns from other sources. Haspelmath (this volume) describes such mirror-image reversals as anti-grammaticalisations, conceding that although they do exist, they are outnumbered by grammaticalisation examples 100 to 1. Another example is provided by AG Ramat (1998: 117) in the purported reanalysis of infinitive person/number inflections in Portuguese – these are assumed to be of clitic status for the reason that they are not obligatory – the impersonal, uninflected infinitive may alternate with them. It is doubtful, though, whether this constitutes an ideal example: the alternation might simply be a stage on the way to their ultimate redundancy. Many of the cited examples may be explained by some other means; for the remainder, if there is a motivation for degrammaticalisation, it should be investigated further; according to Traugott (2001: 6) “the jury is still out on this one”. However, on the surface, it appears that the changes are there because of the restrictions of functional classification. There is certainly little evidence of an increase in semantic complexity in such cases which would be expected in the reverse of a grammaticalisation process, whether or not it proceeded back along the same cline that it came from.6 From that point of view, it remains to determine whether such functional shifts which involve little or no alteration of semantic complexity, e.g., between an affix and a clitic, are important enough to determine reduced levels of grammaticality in every case.
. Further discussions Janda (2001) points out that counterdirectional changes may be allowed to occur because of the fact that speakers do not ‘know’ the lexical/grammatical or free/bound morphological status of grammaticalising morphemes from previous, non-contiguous generations (2001: 299). He (justifiably) deplores the lack of crossgenerational sociolinguistic research in the field of grammaticalisation studies; however, he fails to take into account the possibilities presented in a hypothesis of ontogenetic grammaticalisation, referred to as the Lexical Memory Traces hypothesis (Ziegeler 1997, 2000b), in which the individual speaker’s ‘record’ of the earlier stages of development of grammaticalising morphemes can be found in the stages the speaker has undergone in the acquisition of the language, stages which, only coincidentally, are seen often to parallel historical developments. These stages are represented in the speaker’s intuitive knowledge of later semantic co-occurrence restrictions on grammaticalising items, e.g. the restriction on the use of the modal verb would in such complements as *I wish the Porsche would belong to me, and such constraints are the only evidence of the speaker’s ‘record’ of the historical develop-
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.6 (120)
Debra Ziegeler
ment of the grammaticalising pathway. But unidirectionality is not a conscious phenomenon over which speakers have control. That there do exist counterexamples in one form or another must therefore be related to another cognitive process co-occurring with grammaticalisation. Haspelmath (1999) demonstrates that the functional motivation for renewal in grammaticalisation (i.e. when items of greater transparency are drafted to renew increasingly opaque grammaticalising material) would not be found in reverse: a speaker is not likely ever to return to using an expression of greater opacity to renew one of greater transparency because the old form had become too ‘lexical’ and needed refreshing with more abstract grammatical material. In grammaticalisation, the problem-solving, functional goals include those of greater automation of the code for ease of processing and fluidity of performance; the only available source material is derived from elementary form-meaning conceptual pairs. Degrammaticalisation has no similar functional goals operating since lexical items can be created out of anything, not only grammaticalised items, and thus degrammaticalisation is not a unique path to lexicalisation. Even the short-term, ‘backward’ step of renewal carries with it the long-term goal of expressing an utterance with greater clarity. For such reasons, some of the functional principles inherent in the hypothesis of unidirectionality itself might seem to rule out any need for counterdirectionality. However, Norde (2002: 59–60), in a discussion of the degrammaticalisation of inflectional affixes in Swedish, cites Plank (1995) as referring to degrammaticalisation changes as disturbances in the system. She points to one functional principle that could well provide the motivation for degrammaticalisation as well as grammaticalisation: the principle of least effort, which she attributes to Newmeyer (1998) as an explanation for grammaticalisation (Hopper & Traugott 1993: 20 cite Gabelentz 1891 as the originator of a similar notion which he believed to be in constant competition with the need for greater distinctiveness of expression). In its application to grammaticalisation, while such a principle may motivate processing ease and faster automation of the code, in degrammaticalisation the motivation is equally interesting. It may be questioned how the principle of least effort may affect degrammaticalisation; a phonological reduction as a result of degrammaticalisation unarguably involves less effort, but why should a speaker search for functional items to express lexical concepts? The reason might be that in rapid speech, the preferred forms of expression must be those most readily available to short-term memory, and the advantage of functional items is their ready accessibility due to frequency of use: it is likely that less memory effort will be involved in retrieving for lexical purposes a phonologically simplified, high-frequency grammatical form such as ifs or buts than in retrieving from the lexicon a more formally-complex, less frequently used lexical form, such as doubts or uncertainties.7 The problem of accessing (ostensibly) semantically-depleted functional material for lexical use
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.7 (121)
Redefining unidirectionality
still requires further study and it is beyond the scope of this paper to enter into any lengthy discussion on the matter: the possibility for residual semantic retention in the grammatical morpheme cannot be overlooked (see Bybee & Pagliuca 1987), nor can the likelihood that the grammatical environment is feeding the functional item with semantic information relevant to the context (hypoanalysis, according to Croft 2000), e.g. for ifs the non-factuality of the clause it introduces. Whatever the source of the contributing semantic element, it is by such means that degrammaticalisations in the form of lexicalisations can be seen to be functionally motivated. A problem of the back-tracking, anti-grammaticalisation type of degrammaticalisation is the pragmatic argument, given that grammaticalisation involves a shift towards already entailed senses. In the approach taken by Traugott (e.g. 1988, 1989, and 2000) and also in Bybee et al. (1994), the semantic shifts in grammaticalisation involve the conventionalisation of conversational implicatures, and a reversal would suggest movement towards semantic senses which have been previously conventionalised into the grammaticalised item, e.g. if volition senses in the modal will were again re-created out of a non-lexical, grammatical meaning of future time reference. This type of semantic retrogression would imply a narrowing of meaning, and a subsequent restriction of distribution, in that volition senses cannot apply to inanimate subjects. Most of the cited cases of degrammaticalisation appear to demonstrate either a gradual morphological change which involves little, if any, semantic redevelopment at all, or a lexicalisation to new meanings quite unlike the source meanings of the original grammaticalising item. A definition of degrammaticalisation based on backward progression along the grammaticalisation route would need to take the pragmatic consequences into consideration.
. Demodalisation as degrammaticalisation Van der Auwera and Plungian (1998) list a number of cases in which a grammaticalising modal verb may have meanings which extend beyond its stages of modality, often, but not always, resuming main verb functions with noun phrase objects or attributive complements. Amongst such post-modal forms they include the Russian verb moˇc, cognate to “may”, which perfectivised to form s- moˇc, marking the actualisation of possibility, “manage” (van der Auwera & Plungian 1998: 106); the affixation results in the verb resuming a full verb status again, i.e. pre-voz-mog “overcame”. Some of the other examples cited by van der Auwera and Plungian, though, are less easy to write off as degrammaticalisations. Examples of demodalisations cited by van der Auwera and Plungian may appear to be traceable to the source meanings of the modals, e.g. the German verb mögen, which is cognate with
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.8 (122)
Debra Ziegeler
“may”, have developed new lexical meanings equivalent to “to like”.8 However, in such cases, the meaning of ‘like’ developed in Old High German (c. 750–1100) as a polysemy of the original lexical source meaning of physical tolerance used in negative contexts in which the noun object referred to food or drink (see Diewald 1999: 316, who cites an example from Lühr 1987 dated at c. 1000). Such contexts could easily give rise to meanings of mental dislike or pleasure, and should be considered divergences (Hopper 1991) or ‘splits’ (Heine & Reh 1984), as they continue developing alongside the grammaticalising senses as independent lexical items, and do not affect the grammaticalisation path of the emerging modal verb. Not mentioned by van der Auwera and Plungian, though, is an interesting example of possible demodalisation from Mandarin Chinese, the development of the modal auxiliary d˘ei, a verb with the present day modal meaning of obligation. The demodalised use appears to be related to a later grammaticalised stage, and is held by native speakers to be a relatively marginal use. However, it does appear as one of the senses listed in dictionaries, and therefore requires further consideration. The history of this modal verb is summarised in Sun’s comprehensive (1996) account, in which he analyses the development of d˘ei as a grammaticalisation route. Sun (1996) discusses early uses of d˘ei/dé dating to Old Chinese (500 BC–200 AD), in which the form originally was a full verb meaning “get” or “obtain”.9 An example of its use in Old Chinese is the following (1996: 112): (2) ér dé ti¯anxià and obtain world “And have the kingdom.”
(Mèngz˘ı G¯ongs¯un Ch˘ou shàng – 372–289 BC)
The sense of obtaining could refer to the acquisition of a physical object, or of an abstract object, as in the following example (1996: 143), also taken from Mèngz˘ı (G¯ongs¯un Ch˘ou xià): (3) bì y˘ou dé ti¯an shí zhe y˘e must have obtain heaven time nom asp “. . . there must have been people who obtained a timing from heaven.” An additional function in Old Chinese was as a preverbal modal of permission in negatives or in rhetorical senses (1996: 113): (4) z˘ıkuài bu dé y˘u rén yàn Zi Kuai neg permit give other Yan “Zi Kuai is not permitted to give other(s) the state of Yan!”
(ibid.)
In this respect, the meaning of “obtain an object” (dé + NP) in (2) and (3) extends to that of “obtain an act” (dé + V), in (4) in which a full clausal complement may follow, just as in a modal function.10
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.9 (123)
Redefining unidirectionality
It should be noted at this stage that the predominance of negative constructions within which the first permission modal uses of d˘ei appeared (as noted by Sun 1996: 122) might have contributed to the development of deontic obligation senses, as shown in the following example from Old Chinese, in which the modal has a meaning roughly equivalent to “should” in English: (5) Dé gèng qiú h˘ao n˘u hòu-rì sòng zh¯ı should again seek nice woman other-day present him (Sh˘e Jì – 135–187 BC)11 “(We) should find another nice woman and present her to him some other day.” The progression from permission to obligation is summarised in Goossens (1987: 35), Traugott (1999), and Traugott and Dasher (2000: 124) for the English modal must, also once a modal of permission in Old English times: a denial of permission to do X is logically equivalent to obligation not to do X, the negative environment thus providing a channel for the future meanings of obligation to emerge. The development of the modal senses then is illustrated in a semantic pathway leading from a verb expressing acquisitional possession as follows: (i) obtain a physical object > (ii) obtain an act (permission) > (iii) negative permission > (iv) obligation-not > (v) general or positive obligation.12 All such stages appear to have been present in Old Chinese, and in the 15th century the main modal meaning was deontic obligation, or deontic necessity according to Sun (1996: 152). Deontic obligation is the main preverbal modal meaning of d˘ei today, as illustrated below (Sun 1996: 160): (6) hái d˘ei ch¯ı ròu still should eat meat “(One) still has to eat meat.” Meanwhile, in Middle Chinese, the verb-final use of dé arose, in which its pronunciation was more likely to be unstressed /d6/. The verb-final uses are retained in the present-day potential complements, e.g. mài de dào “managed to buy (it)” and degree complements, e.g. sh¯uo de h˘en kuài “speaks (very) fast”, in which the use of dé is as a complementiser, but it is debateable whether they are related to the grammaticalisation path of the pre-verbal uses (Enfield 2001 maintains that they are cognate). The only examples Sun gives for Early Mandarin (1001–1900 AD) are of verb-final uses, though he notes that all four forms, the lexical full verb, the modal, and the two types of post-verbal uses were found throughout Early Mandarin. Sun also finds epistemic meanings emerging in the present-day modal, though Chao (1968) does not list any. Lü (1999: 166) describes such uses as expressing inevitability, as illustrated below:
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.10 (124)
Debra Ziegeler
(7) zhème w˘an cái húiqù, m¯a yòu d˘ei sh¯uo n˘ı le this late then return.go mother again must speak you crs13 “If you go back this late now, mother will surely have words to say to you.” Lü (1999: 166) maintains that the use illustrated above is found only in the spoken language, and it should also be noted that such uses co-occur with non-stative verbs, unlike English epistemic must (d˘ei cannot, for example, co-occur with the stative verb shì, the copula). In this way, what is described here as epistemic necessity may coincide, semantically, with meanings of strong prediction (inevitability). It is not known how early such uses appeared, but a similar example appears in Hónglóumèng, which is dated at approximately 1771:14 (8) N˘ı shì w˘ode m¯a ah, n˘ı bù jiù w˘o, w˘o bù d˘ei húo You are my mother emp you not save me I not must live le! crs “You are my mother, if you don’t save me, I surely cannot live!” It is also observed that, in restricted present-day uses, a new main verb d˘ei is appearing, with the meaning “need”, “take” or “require”. This has been noted in dictionary examples,15 as well as in Lü (1999), as appearing in either of two contexts, one with a numeral classifier following d˘ei, and another with a noun clause following (Lü 1999: 166): (9) a. (d˘ei + numeral classifier + NP): zhè ge g¯ongzuò d˘ei s¯an ge rén this cl work needs three cl people “this work needs three people.” b. (d˘ei + noun clause): bié rén qù bù xíng, d˘ei n˘ı q¯ınzì qù other people go not ok need you in-person go “It’s not ok for other people to go, it requires that you go in person.” The opinions of native speakers vary with regard to such uses; some claim that the use in (9a) only stands if another clause is added after the object; e.g. . . . cái néng wán – “. . . in order to finish” (Lili Chang; p.c.); others find it is quite idiomatic, and conveys the meaning “should need”, “must” (Bao Zhi-Ming, p.c.), or that the use is restricted to northern dialects, including that of Beijing, and not commonly used in written genres, being mainly confined to spoken usage (Shi Ding Xu, p.c.). Lü (1999: 166) also notes the restriction to spoken usage. Chao (1968: 743) lists the use illustrated in (9) as less common than other present-day uses, and believes that it is a short form of d˘ei yào, or d˘ei yòng, “must want”, “must need”. However, Lü (1999) makes no mention of any possible ellipsis as the source for such uses. Whether
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.11 (125)
Redefining unidirectionality
it stands for an ellipted form or not, it still represents a demodalisation, since the source for the new lexical development is related only to the later stages of its modal development, i.e. it is at a post-modal stage, similar to the examples discussed by van der Auwera and Plungian (1998). That the development is a new function and not related to its original source meaning lies in the fact that the lexical source meaning of a full verb meaning “obtain” is still available for use today alongside its modalised senses, as Sun (1996: 160) illustrates: (10) dé le z¯ıliào obtain asp material “(One) obtained the data.” This is another indication of a split or divergence, in which the original lexical form continues to coexist with the new grammaticalising uses, and is unrelated to the new development, in which a full lexical verb with a meaning of concrete necessity has re-emerged from a modal meaning of deontic or epistemic necessity. That the new lexical verb is not related to the original source form is clear from its phonological form – the modal and the post-modal forms have the later diphthong variant of pronunciation, while the lexical source verb retains the original monophthong until the present-day. Exactly when the secondary lexical meanings started to emerge, or how, is not clear. The examples appearing in Lü (1999), including the examples of epistemic certainty, are all listed as occurring in spoken usage only, suggesting that they may be quite recent innovations and have not yet been conventionalised in standard Mandarin (no similar examples are provided by Sun (1996) in his diachronic study.) There is no chronological ordering assumed, although examples of the type illustrated in (9b) can be found as early as the 18th c., in Hónglóumèng, Ch. 94, as shown below: (11) Zhè jiàn shì hái d˘ei n˘ı qù cái nòng de míngbai this class matter still requires you go then make comp clear “This matter still requires that you go in order for it to be clarified.” In this type, the scope of the modality extends outside the clause in which it might have occurred if the meaning were deontic and the obligation imposed at the time of utterance: the modal element has an entire noun clause in its scope. What this construction then produces is a main verb with a general lexical sense of ‘need’ or ‘requirement’. Whether there is a subject or not is not important (zhè jiàn shì could be taken to be a topic), but even if there were no subject, the sense ‘it is necessary that’ [+ noun clause], would be likely to emerge from such examples, placing the scope of the necessity meanings outside the entire proposition, and altering the auxiliary status of the modal element to become a main verb again. This scope extension is similar to that labelled ‘wide-scope’ deontic necessity by Traugott and
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.12 (126)
Debra Ziegeler
Nordlinger (1997) except that It is hypothesised that d˘ei is no longer an auxiliary. Such uses can be hypothesised to be a stage prior to the even more lexicalised sense in (9a), since the noun clause has been replaced with a noun phrase object, s¯an ge rén, “three people”, in the Modern Chinese example. If the form is re-lexicalising in the opposite direction to a grammaticalisation path, one would expect a more concrete noun phrase complement referring to a physical object to appear after a more abstract noun clause describing an event, thus exhibiting gradual counterdirectional tendencies (the reverse of grammaticalisation in which a more abstract nominal argument might occur at a later stage than a more concrete one). The order of appearance of the two constructions, though, cannot be proven, without a more exhaustive search of all the available diachronic sources. However, there were no examples of the type shown in (9a) in the six chapters of Hónglóumèng surveyed, and (11) was not the only example of the type illustrated in (9b). Figure 1 illustrates the patterns thus described for the modalisation and demodalisation of d˘ei. The situation illustrated is typical of grammaticalisation patterns in Chinese, which supply strong evidence for panchronic grammaticalisation: in the present use of dˇei, it appears that almost every stage of its development can be seen to co-exist, the older lexical source functions overlapping for the duration of the grammaticalisation. Heine, Claudi and Hünnemeyer (1991: 258–261) discuss the term panchrony as serving a theory of grammaticalisation in which it is “ . . . both unjustified and impractical to maintain a distinction between synchrony and diachrony.” In other words, new senses are acquired without completely losing the old ones in the same grammaticalisation process. Thus, even at a post-modal stage, as demonstrated, there is a retention of a previous stage of development in the use Lexical source meaning ‘obtain’
Primary split
Secondary split ‘need’
Grammaticalisation path (modal of necessity, obligation)
Figure 1. Representation of the grammaticalisation of the modal d˘ei in Chinese, illustrating the way in which the relexicalisation of the modal can be described in terms of a secondary ‘split’ from the continuing main path of grammaticalisation.
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.13 (127)
Redefining unidirectionality
of the former senses of grammatical necessity to reconstruct a new lexical item. The only difference between the pathway demonstrated by dˇei and that of the German modals discussed earlier is that in the case of the German modal, what split off in the beginning was a new lexical path, developing as a polysemy of the original lexical source meaning in specific contexts of use, and continuing to generalise as an unrelated lexical item alongside the grammaticalising modal. However, in the case of the Chinese modal, the (primary) split left the original lexical source meaning to co-exist unchanged alongside the grammaticalising modal, retaining its original lexical functions for the duration of the grammaticalisation process. The latter case is the more usually-described (e.g. with going to); however, it does not necessarily occur with every case of grammaticalisation, since not every case of grammaticalisation involves the continued retention of the source as a lexical item, e.g. the Mandarin pre-transitive, b˘a is no longer used as a lexical item, though the present-day restrictions on its distribution reflect its lexical origins (see Ziegeler 2000a) and blur the distinction between diachrony and synchrony, as the term panchrony suggests.
. Discussion The case illustrated in the above example of Mandarin d˘ei raises the central question how it is possible to accommodate such lexicalisations within a general framework of grammaticalisation without undermining unidirectionality. It is hypothesised in the present study that the answer may be found in the theories of conceptual networks and chains of meaning often used to discuss classes of lexical items in languages, such as those united by classifier terms. Lakoff (1987: Ch. 6) describes such networks as radial categories, and provides examples from two unrelated languages in which classifiers are used, Dyirbal, an Australian language, and Japanese. In such languages, a number of members may be classed together under one conceptual category though they do not share attributes with every other member of the same class. They may be linked by chains of resemblance to no more than one other member (and by only one property or attribute), and that member may be only peripheral to the group as a whole. The concept of semantic category membership by resemblance in this way has been described in earlier work as family resemblance theory (Wittgenstein 1953) and later Rosch and Mervis (1975), with application to lexical categories. However, a more recent application of the theory has been proposed by Heine (1992, 1993) to account for the progressive stages of grammaticalisation he terms grammaticalisation chains. In such chains, family resemblance relations link adjacent stages of the chain, though there is no sense of there being a focal prototype, or core
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.14 (128)
Debra Ziegeler
sense, to which the peripheral senses are linked. In every other way, the structural relationships of a grammaticalisation chain closely match a family resemblance structure. What is implicit in Heine’s use of the family resemblance theory is a pathway leading to a determinate endpoint; in fact, Heine describes a minimal grammaticalisation chain as having a source overlapping with a target (1993: 53), in the idealisation: source > source/target > target, as three separate stages. This approach may serve to justify the nature of structural change, but in order to accommodate grammaticalisation chains within such a theory, it could be hypothesised that grammaticalisation is unidirectional across a wider network of form-function links which have no determinate end-point, and therefore the model represented in a grammaticalisation chain is only contained within such a network, it does not comprise the entire network itself. That is, it is only a sub-part of a much larger conceptual network of metonymic and metaphorical links, of which it holds salience. A transformation from a lexical item to a grammatical item takes place within the network due to the type of environmental developments taking place within that sub-part, the methods used by speakers to increase the frequency and routinisation of a lexical form in a functional role, and the nature of the source forms (usually basic level categories, or the most unmarked, frequently-used representatives of a class of items). All this means is that the metaphoric or metonymic meaning shifts which occur between neighbouring items in a conceptual network will always be there, it is just that in some cases, given the right conditions of routinisation, generalisation, frequency of use, and the socio-pragmatic effects, the shifts take off and an erstwhile metonymic shift within the lexicon develops into a grammaticalisation chain. Heine (1992: 350) describes family resemblance patterns as an outcome of grammaticalisation; however, this is only what is immediately apparent: in actual fact, it seems more the case that grammaticalisation is an outcome of shifts across family resemblance patterns which provided the input for the initial sources of grammaticalisation chains (in basic level categories), and which continue by extracting the output of grammaticalisation chains in the form of new lexicalisations long after the grammaticalisation stages have run their course, or while they are still in progress. Figure 2 illustrates an idealised model of the way in which a grammaticalisation chain may be shown to interact with other chains of meaning shifts within the lexicon across a conceptual network. Post-grammaticalisation stages, or stages which follow grammaticalisation such as the lexicalisation illustrated, have few options: either they contain renewals, regrammaticalisations or continuation in functional directions (see Greenberg 1991), or lexicalisations, which is what appears to be happening in the case of preverbal d˘ei. The category extensions are represented in Figure 2 as overlapping stages (A-G) and run from a lexical sense, as follows:
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.15 (129)
Redefining unidirectionality
(12) ‘obtain a physical object (A) > obtain an abstract notion’ (A/B); > modal sense where the obtaining of an act implies ‘permission’ (B/C); > ‘negative permission implying obligation-not (neg.)’ (C/D); > negative obligation implying social obligation (D/E); > social obligation implying deontic necessity (E/F); > deontic necessity implying wide-scope deontic necessity (F/G) > wide-scope deontic necessity yielding lexical ‘need’ (G/X) and epistemic necessity (G/H). The stage of lexical necessity is marked (G/X) to indicate that although, like the stage of epistemic necessity (G/H), it contains an element of the former stage of
GH
G/X FG
EF
DE
CD BC AB
Figure 2. Idealised model illustrating the way in which a unidirectional grammaticalisation chain may interact with and be integrated into other links in a ‘family resemblances’ conceptual network. The stages A-H are represented as overlapping; G/X is shown as linked by a broken line to illustrate a relexicalisation at a later stage of grammaticalisation.
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.16 (130)
Debra Ziegeler
wide-scope deontic necessity, it does not belong to the grammaticalisation path highlighted in bold, but branches off as a split, and it is not known at this stage whether it will form the origin of yet another new grammaticalisation path itself. Although the stages shown are represented as occurring in a chronological order, it would not be wise to propose a time sequence of development without more data, as all the stages from A to F are shown in Old Chinese. However, for the phonological reasons noted earlier, and because it appears only in spoken usage in present-day Mandarin, the degrammaticalisation found in (9a) is judged to be a recent development, and not a polysemy of its lexical source. The example of demodalisation thus outlined supports the fact that morphosyntactic changes and category shifts are only the outcome, and thus the visible evidence, of longer-term shifts in the conceptual nature of the process. It is easy to consider only the morphosyntactic evidence of underlying semantic changes without considering the functional motivation for such evidence, as many languages outside of the Indo-European family are not so easily identifiable in terms of morphosyntactic marking for category distinctiveness, and an account of unidirectionality as a category-based definition would be inadequate to describe such languages as Chinese, which frequently display ambiguity in category distinctions, e.g. full verb vs. auxiliary verb, and in which earlier stages of development may be seen to overlap and coexist with later ones for unusually lengthy periods of time. For such reasons, a hypothesis of unidirectionality based on semantic grounds is far more appropriate to account for languages in which grammaticalisation is not so frequently perceived in formal changes. The post-modal changes observed in d˘ei, then, do not violate semantic unidirectionality since the meaning of lexical necessity is a natural progression from the meaning of deontic modal necessity, as is epistemic modal necessity. Furthermore, the relexicalisation does not inhibit the further grammaticalisation of d˘ei to epistemic uses in present-day Mandarin, as demonstrated; it merely creates a new pathway in a wider conceptual network of meaning links.
. Conclusion The account presented, in fact, aims to question the traditional view of grammaticalisation as an isolated series of unidirectional changes, considering instead one of a participating series of changes in a larger cognitive sphere, something alluded to by Campbell (2001: 116) (but not explored for its semantic possibilities), in which the postgrammaticalisation stages continue to display the same semantic unidirectionality as other elements within the network. In positing such an account, it may be seen that outside grammaticalisation, a chain of contiguous meaning shifts can
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.17 (131)
Redefining unidirectionality
continue indefinitely throughout a network of family resemblances – this is permitted because of the constant motivation for metonymic shifts and increments of meaning to emerge in language use, and speakers will always resort to some sense of an old meaning to develop a wholly new meaning. The only unique features of grammaticalisation are the formal manifestations which accompany the semantic changes; however, in the example illustrated, the pathway of grammaticalisation is manifested as an increased range of grammatical functions, at the end of which there is a return to a (new) lexical function. The presence of a post-modal lexical function may therefore threaten traditional views of grammaticalisation as functionally unidirectional, but it does not threaten the prospect of a unidirectional hypothesis as long as it can be maintained that semantic change is prior to all other changes, and it indicates that semantic unidirectionality can transgress the bidirectionality of functional change (lexical verb > auxiliary verb > lexical verb). However, the present study has barely scratched the surface; much further work is needed to aim for a more comprehensive overview of the unidirectionality debate if it is to have any theoretical value in future grammaticalisation studies.
Notes . This, and many more useful suggestions were made by various members of the audience at the NRG2 Symposium, Amsterdam, 4 April, 2002. I am grateful for all such contributions, as well as those of the editors and an anonymous referee, and to Cindy Allen, Delia Bentley, Philippe Bourdin, Lili Chang, Hilary Chappell, Bernd Heine, Paul Hopper, Violet Lai, Stephen Matthews, Bao Zhi Ming, Richard Wong, Shi Ding Xu, Foong Ha Yap, and colleagues at the University of Manchester Advanced Syntax Seminar series for much interesting discussion. Any shortcomings remain my own. . I use the term semantic throughout to refer generally to both semantic and pragmatic characteristics. . I use the term ‘post-grammaticalisation’ in the general sense of any stages which follow grammaticalisation. . It is possible that there is a meaning difference between deber and deber de – the former denoting obligation and the latter form denoting epistemic possibility, a contrast which is believed to be being lost, especially in the present tense. I am grateful to Delia Bentley for pointing this out to me. . This raises the (notional) question, of course, of whether infinitives can be described as ‘grammaticalised’ in the first place – if they are not even grammaticalised, it is pointless to discuss them in terms of degrammaticalisation. . Fischer (2000) discusses the increase in semantic complexity associated with the degrammaticalisation of the infinitive marker to in English. . Naturally, psychological experiments would be necessary to verify this possibility.
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.18 (132)
Debra Ziegeler . The Dutch modal moeten “must, have to” has developed similar lexical meanings, as suggested by the editors. I do not have any equivalent evidence of the development of such meanings in this case, though Traugott and Dasher (2000: 122) maintain that the origins of the Gothic and Early Germanic pre-modal form mot- had source meanings of participantinternal (ability) semantics, like mögen. . The monophthong variant dé is related to the original pronunciation of the form, after the glottal finals disappeared due to dialectal interference in the 13th and 14th centuries (see Sun 1996: 153–155 for a detailed account). This was not affected by the diphthongisation of the form which became the modal variant (d˘ei). (The fact that both forms are written using the same character indicates that they are from the same source.) . Some native speakers are of the opinion that negative permission is no longer one of the functions of dˇei, this function being taken up by modal verbs such as kˇey˘ı and néng. . Ackowledgements are humbly due to Violet Lai for assisting me in finding examples from Sh˘e Jì and from Sh˘uih˘u Zhuàn. . As pointed out by the editors, this pathway is not a necessary one for the development of obligation meanings out of permissions senses, e.g. Dutch mogen “may” can be used in positive contexts with the meaning of deontic necessity. It remains for a more intensive study to explore what are the alternative routes by which deontic obligation may emerge from permission modalities. . crs = ‘current relevant state’, here, one of a number of functions including linking the given information in the conditional clause to the future predicted situation (see Li & Thompson 1989). Other abbreviations used include class = classifier, comp = complementiser (the post-verbal function of d˘ei), and emp = emphatic marker. . The examples from Hónglóumèng were extracted from the following electronic website: http://cls.admin.yzu.edu.tw/HLM/home/htm. The concordance used is to be found on the Academica Sinica web-site, Taipei: http://www.dmpo.sinica.edu.tw:8000/∼ words/sou/ sou.html. I acknowledge the help of the Yuan Zhe University and Academica Sinica, Taiwan, and Violet Lai for assistance and access to these web-sites. . The Times Chinese-English Dictionary, Singapore: Federal Publications.
References Benveniste, Emile (1958). “Les Verbes Délocutifs”. Reprinted in: Problèmes de Linguistique Générale, 277–285. Paris: Gallimard, 1966. Beths, Frank (1999). “The History of Dare and the Status of Unidirectionality”. Linguistics, 37, 1069–1110. Bybee, Joan & William Pagliuca (1987). “The Evolution of Future Meaning”. In A. G. Ramat, O. Carruba, & G. Bernini (Eds.), Papers from the Seventh International Conference on Historical Linguistics (pp. 109–122). Amsterdam & Philadelphia: John Benjamins. Bybee, Joan, Pagliuca William, & Revere D. Perkins (1994). The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: University of Chicago Press.
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.19 (133)
Redefining unidirectionality
Campbell, Lyle (2001). “What’s Wrong with Grammaticalization”? Language Sciences, 23, 113–161. Chao, Yuen Ren (1968). A Grammar of Spoken Chinese. Berkeley and Los Angeles: University of California Press. Croft, William (2000). Explaining Language Change. An Evolutionary Approach. Harlow: Longman. Diewald, Gabriele (1999). Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalität. Tübingen: Max Niemeyer. Enfield, Nick J. (2001). “On Genetic and Areal Linguistics in Mainland South-East Asia: Parallel Polyfunctionality of ‘acquire”’. In A. Y. Aikhenvald & R. M. Dixon (Eds.), Areal Diffusion and Genetic Inheritance (pp. 255–290). Oxford: Oxford University Press. Eythórssen, Thórhallur, Kersti Börjars, & Nigel Vincent (2002). “On Defining Grammaticalization”. Paper presented at the New Reflections on Grammaticalization 2 Symposium, University of Amsterdam, 4–6 April, 2002. Fischer, Olga (1994). “The Development of Quasi-auxiliaries in English and Changes in Word Order”. Neophilologus, 78, 137–164. Fischer, Olga (2000). “Grammaticalisation: Unidirectional, non-reversable?” In Fischer et al. (Eds.), 149–169. Fischer Olga, Anette Rosenbach, & Dieter Stein (Eds.). (2000). Pathways of Change. Grammaticalization in English. Amsterdam and Philadelphia: John Benjamins. Gabelentz, Georg von der (1891). Die Sprachwissenschaft. Ihre Aufgaben, Methoden, und bisherigen Ergebnisse. Leipzig: Weigel. Goossens, Louis (1987). “Modal Shifts and Predication Types”. In Johan van der Auwera & Louis Goossens (Eds.), Ins and Outs of Predication (pp. 21–37). Dordrecht: Foris. Greenberg, Joseph (1991). “The Last Stages of Grammatical Elements: Contractive and Expansive Desemanticization”. In Traugott & Heine (Eds.), Vol. 1 (pp. 301–314). Hagège, Claude (1993). The Language Builder: An Essay on the Human Signature in Linguistic Morphogenesis. Amsterdam: Benjamins. Haspelmath, Martin (1999). “Why is Grammaticalization Irreversible”? Linguistics, 37, 1043–1068. Heine, Bernd (1992). “Grammaticalization Chains”. Studies in Language, 16, 335–368. Heine, Bernd (1993). Auxiliaries. Cognitive Forces and Grammaticalization. New York: Oxford University Press. Heine, Bernd & Mechthild Reh (1984). Grammaticalization and Reanalysis in African Languages. Hamburg: HelmutBuske. Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer (1991). Grammaticalization. A Conceptual Framework. Chicago: University of Chicago Press. Hopper, Paul J. (1991). “On Some Principles of Grammaticization”. In Traugott & Heine (Eds.), Vol. 1 (pp. 17–35). Hopper, Paul J. & Elizabeth Closs Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Janda, Richard (2001). “Beyond ‘Pathways’ and ‘Unidirectionality’: On the Discontinuity of Language Transmission and the Counterability of Grammaticalization”. Language Sciences, 23, 265–340.
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.20 (134)
Debra Ziegeler
Klausenburger, Jurgen (2000). Grammaticalization. Studies in Latin and Romance Morphosyntax. Amsterdam and Philadelphia: John Benjamins. Lakoff, George (1987). Women, Fire, and Dangerous Things. Chicago: University of Chicago Press. Lehrer, Adrienne (1996). “Identifying and Interpreting Blends: An Experimental Approach”. Cognitive Linguistics, 7, 359–390. Levinson, Stephen (1989). Pragmatics. Cambridge: Cambridge University Press. Li, Charles N. & Sandra A. Thompson (1991). Mandarin Chinese. A Functional Reference Grammar. Berkeley and Los Angeles: University of California Press. Lü, Shuxiang (Ed.). (1999). Xiandai Hanyu Babai Ci [‘Eight Hundred Words of Modern Chinese’] (Expanded edition). Beijing: Commercial Press. Newmeyer, Frederick J. (1998). Language Form and Language Function. Cambridge, MA: MIT Press. Norde, Muriel (2001). “Deflexion as a Counterdirectional Factor in Grammatical Change”. Language Sciences, 23, 231–264. Norde, Muriel (2002). “The Final Stages of Grammaticalization: Affixhood and beyond”. In Wischer & Diewald (Eds.), 45–65. Plank, Frans (1995). “Entgrammatisierung – Spiegelbild der Grammatisierung?” In N. Boretzky et al. (Eds.), Natürlichkeitstheorie und Sprachwandel (pp. 199–219). Bochum: Universitätsverlag Brockmeyer. Ramat, Paolo (1992). “Thoughts on Degrammaticalization.” Linguistics, 30, 549–560. Ramat, Anna Giacalone (1998). “Testing the Boundaries of Grammaticalization”. In Ramat & Hopper (Eds.), 107–127. Ramat, Anna Giacalone & Paul J. Hopper (Eds.). (1998). The Limits of Grammaticalization. Amsterdam and Philadelphia: John Benjamins. Rosch, Eleanor H. & C. B. Mervis (1975). “Family Resemblances: Studies in the Internal Structure of Categories”. Cognitive Psychology, 7, 573–605. Sun, Chao-Fen (1996). Word Order Change and Grammaticalization in the History of Chinese. Stanford: Stanford University Press. Tabor, Whitney & Elizabeth Closs Traugott (1998). “Structural Scope Expansion and Grammaticalization”. In Ramat & Hopper (Eds.), 229–272. Traugott, Elizabeth Closs (1988). “Pragmatic Strengthening and Grammaticalization”. Proceedings of the 14th Annual Meeting of the Berkeley Linguistics Society, 406–416. Traugott, Elizabeth Closs (1989). “On the Rise of Epistemic Meanings in English.” Language, 65, 31–55. Traugott, Elizabeth Closs (1994). “Grammaticalization and Lexicalization”. In R. E. Asher (Ed.), The Encyclopedia of Language and Linguistics, Vol. 3 (pp. 1481–1485). Oxford: Pergamon. Traugott, Elizabeth Closs (1999). “Why must is not moot”. Paper presented at the 14th International Conference on Historical Linguistics, Vancouver, August 1999. Traugott, Elizabeth Closs (2000). “From Etymology to Historical Pragmatics”. Plenary paper presented at the conference on Studies on English Historical Linguistics, UCLA, 27 May, 2000. Traugott, Elizabeth Closs (2001). “Legitimate Counterexamples to Unidirectionality”. Paper presented at Freiburg University, 17 October, 2001.
TSL[v.20020404] Prn:24/02/2004; 14:18
F: TSL5906.tex / p.21 (135)
Redefining unidirectionality
Traugott, Elizabeth Closs & Bernd Heine (Eds.). (1991). Approaches to Grammaticalization (2 Vols.). Amsterdam and Philadelphia: John Benjamins. Traugott, Elizabeth Closs & Rachel Nordlinger (1997). “Scope and the Development of Epistemic Modality: Evidence from Ought To”. English Language and Linguistics, 1, 295–317. Van der Auwera, Johan (2002). “More Thoughts on Degrammaticalization”. In Wischer & Diewald (Eds.), 19–29. Van der Auwera, Johan & Vladimir A. Plungian (1998). “Modality’s Semantic Map”. Linguistic Typology, 2, 79–124. Wischer, Ilse & Gabriele Diewald (Eds.). (2002). New Reflections on Grammaticalization. Amsterdam and Philadelphia: John Benjamins. Wittgenstein, Ludwig (1953). Philosophical Investigations. New York: Macmillan. Ziegeler, Debra P. (1997). “Retention in Ontogenetic and Diachronic Grammaticalization”. Cognitive Linguistics, 8, 207–241. Ziegeler, Debra P. (2000a). “A Possession-Based Analysis of the Ba-construction in Mandarin Chinese”. Lingua, 110, 807–842. Ziegeler, Debra P. (2000b). Hypothetical Modality. Grammaticalisation in an L2 Dialect. Amsterdam and Philadelphia: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.1 (137)
From pronominalizer to pragmatic marker Implications for unidirectionality from a crosslinguistic perspective Foong Ha Yap, Stephen Matthews, and Kaoru Horie Chinese University of Hong Kong / University of Hong Kong / Tohoku University, Japan
.
Introduction
In a number of East and Southeast Asian languages, grammatical markers in the referential domain are in some cases recruited to also serve in the evidential domain. In this paper, we will examine the development of Japanese no, Mandarin de, and Malay (em)punya using historical data. More specifically, we will focus on identifying the early uses of each of these morphemes, and we then proceed to trace the pathway(s) leading to the emergence of the pragmatic/evidential function which we here refer to as ‘stance’. To illustrate with examples from Japanese, the morpheme no is well known for its genitive and (pro)nominalizing functions, as illustrated in (1), (2) and (3) below: (1) watashi no tsukue 1sg gen table “my table” (2) watashi no 1sg pronominal “mine” (3) tsuku-tteiru no cook-prog-nonpast (pro)nominalizer “the one (s/he) is cooking”
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.2 (138)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
What is less well understood is how this range of functions has extended to include stance or perspective marking uses of no, which often convey a degree of assertion, especially when accompanied by final falling intonation, as illustrated in (4). (4) asita wa hareru no tomorrow top become.fine.weather stance “(I assure you) the weather will be fine tomorrow.” In what follows, we will chart the grammaticalization pathways for Japanese no and its counterparts in other languages, more specifically Mandarin de and Malay (em)punya, then compare the similarities and differences in their development patterns, with special attention to the issue of unidirectionality. Our discussion will be organized as follows. In Section 2, we briefly review some mechanisms of grammaticalization, then summarize the current debate on unidirectionality as a defining characteristic of grammaticalization phenomena. In Section 3, we focus on the grammaticalization of (pro)nominalizers into stance markers in Japanese, Chinese, and Malay – with emphasis on tracking diachronic developments, semantic-pragmatic motivations, and language-specific syntactic facilitations. In Section 4, we examine instances of possible non-unidirectionality, and resolve the issue by distinguishing between semantic extensions within the same ontological domains and semantic extensions across ontological domains. More specifically, we will suggest that the direction of grammaticalization can be variable within ontological domains, while unidirectionality is more robust when traced across ontological domains.
. Defining grammaticalization and (uni)directionality . On the mechanisms of grammaticalization We begin by outlining some relevant mechanisms that underlie the grammaticalization processes being examined in this paper, in particular the role of ‘bridging contexts’ and ‘switch contexts’, which have been clarified in recent works. Grammaticalization is a process in which a linguistic item acquires new (and often more abstract) grammatical functions. Semantic changes are thus necessarily involved, and these changes are often argued to be gradient in nature, arising from extensions that are contiguous and overlapping, and hence in many instances ambiguous (Heine et al. 1991; Heine 1992, 1993). The pivotal role of semantic ambiguity is often reflected in claims such as the following: A linguistic item with function A (whether lexical or grammatical) is unlikely to evolve into a new (and more grammatical) function B without first passing through an intermediate stage where both interpretations, A and B, coexist. That is, we expect to see: A > (A∼B)
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.3 (139)
From pronominalizer to pragmatic marker
> B (e.g. Hopper & Traugott 1993: 36; Heine 1997: 82). Recent studies have more precisely identified the mechanisms that give rise to the overlapping stage (A∼B) that eventually leads to the establishment of a new function B. Evans and Wilkins (2000), for example, highlight the role of ‘bridging contexts’ in the overlapping (A∼B) stage. Crucially, in bridging contexts, “meaning B is only contextually implicated” (p. 549, emphasis added) and the item concerned has yet to acquire a meaning distinct from A. Heine (2002: 85) adds a further stage, which he refers to as a ‘switch context’ stage. Switch contexts are those in which not only is the target meaning (B) supported, but “the source meaning [i.e. A] is ruled out.” Heine (2002: 90–91) provides examples from Swahili for the initial lexical stage, followed by both the ‘bridging context’ and ‘switch context’ stages, reproduced as (5a–c) below, where the morpheme taka can be used as a desiderative/volitional verb (A), or a proximative future marker (B), or both (A∼B), depending on the context: (5) a.
a- na- taka ku- ni- ita he- pres- want inf- me- call “he wants to call me” b. a- na- taka ku- fa he- pres- want inf- die (i) “he wants to die” (ii) “he’s about to die” c. M- ti u- na- taka ku- anguka C3- tree C3- pres- want inf- fall “the tree is about to fall”
Stage I (initial stage): volitional meaning Stage II (bridging context): volitional meaning possible, but proximative often foregrounded Stage III (switch context): volition backgrounded or no longer available
Crucially, in switch contexts, meaning B still requires contextual support; once this requirement is relaxed, meaning B can be said to have moved on to the stage of conventionalized interpretation. Heine notes that while the Swahili desiderative/ proximative morpheme taka has not proceeded beyond the ‘switch context’ stage, a conventionalization stage (where only the proximative meaning is possible) is attested in another Bantu language, Chamus. Heine’s (2002) model thus reflects the following stages: Stage I: Stage II: Stage III: Stage IV:
Initial stage Bridging context Switch context Conventionalization stage
In the present study, the terms ‘bridging context’ and ‘switch context’ are used to highlight the contiguous nature of semantic extensions in the grammaticalization processes being examined for Japanese no, Mandarin de and Malay (em)punya.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.4 (140)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
. (Uni)directionality in grammaticalization Until recently, most works on grammaticalization tacitly accepted unidirectionality as a defining characteristic of grammaticalization phenomena (e.g. Heine et al. 1991; Traugott & Heine 1991). Indeed, a number of strong assertions in earlier years gave rise to what has come to be known as the unidirectionality hypothesis (e.g. Lehmann 1982: 19; Haspelmath 1989: 302). Challenges to this hypothesis (particularly Newmeyer 1998: 263–276) have recently triggered vigorous debate in the literature on the status and role of unidirectionality in grammaticalization (e.g. Haspelmath 1999; Janda 2001; Norde 2001), and much current discussion focuses on the types of phenomena that qualify as valid examples of counterdirectionality. While the debate is still not resolved, van der Auwera (2002: 22) has observed that there is a consensus that counterdirectionality or ‘degrammaticalization’, while overwhelmingly less common, is more than just an exception. Several other views have also emerged that are worth further consideration. One view is that degrammaticalization phenomena are valid and important areas of investigation within the rubric of grammaticalization studies. As Norde (2002) points out, this is because understanding the asymmetry between grammaticalization and degrammaticalization changes “requires not only an explanation for the overwhelming unidirectionality of grammaticalization change, but also a survey of the circumstances that do result in less grammatical forms” (p. 58). Of special relevance to the current debate on (uni)directionality is Ziegeler’s recent proposal (this volume), where she outlines a conceptual network model of semantic change that accounts for both grammaticalization and degrammaticalization phenomena. Her formulation allows for unidirectionality to be preserved within each grammaticalization pathway, while at the same time making room for semantic splits along the pathway that could give rise to new grammaticalization trajectories. In cases where an abstract grammatical item has become semantically bleached, reinforcement from other semantic links – including other grammaticalization pathways – is often possible and sometimes even desirable. In some other cases, a grammatical(ized) item, while semantically abstract, can nevertheless serve as the source for (re)lexicalization, otherwise also referred to as degrammaticalization. In this paper, we will examine three non-Indo-European languages to address the question of whether unidirectionality is always preserved, and where our findings reveal possible examples of counterdirectionality, to then specify some of the constraints on unidirectionality that appear to be crosslinguistically robust. More specifically, we will focus especially on grammaticalization phenomena in which (pro)nominalizers evolve into pragmatic markers in Japanese, Chinese and Malay.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.5 (141)
From pronominalizer to pragmatic marker
. The grammaticalization of pronominalizers into pragmatic markers . Japanese no As noted earlier, the Japanese morpheme no serves a wide range of functions. Within the referential domain, no has been identified as a genitive marker, a pronominal marker, a relative clause marker (particularly the headless type), and a complementizer (e.g. Horie 1998). Within the evidential domain, no has also been identified as a stance marker, denoting assertion (Aoki 1986) or group authority (Cook 1990). From a diachronic perspective, Horie (1998) has further proposed a grammaticalization pathway for the referential uses of no, without however linking this pathway to the evidential (i.e. stance or perspective marking) uses. In the subsections that follow, we will examine this link between the referential and evidential uses of no, reviewing currently available data from several converging perspectives – morphosyntactic, semantic, and diachronic.
.. Referential and evidential functions of no Within the referential domain, no is used as a genitive marker, linking two noun phrases in a possessor-possessee relationship, as illustrated in (6) below. (6) [NP Taroo] no hon Taroo gen book “Taro’s book” In a variant where the head noun is elided, no may be said to take on a pronominal function, as in (7): (7) [NP Taroo] no Taroo pronominal “Taro’s” Such a pronominal function also extends to clausal constructions, as in (8), where no head noun is present or permitted. (8) [S Taroo-ga kat-ta] no takakatta Taroo-nom buy-past (pro)nominalizer expensive-past “the one Taro bought was expensive” Both genitive no and pronominal no share a common function: they each introduce information that identifies, or specifies, the referent. In the genitive construction, the referent is overtly expressed as the possessee NP (e.g. hon “book” in (6) above). In the pronominal construction, as exemplified in both (7) and (8), the referent is not overtly expressed. This often happens in discourse contexts where the identity of the referent is already given, or known, or at least assumed to be shared knowl-
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.6 (142)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
edge between speaker and hearer. In the absence of an overtly expressed referent, no assumes a pronominal function. In structural terms, the difference between the genitive and pronominal uses of no thus involves the overt or covert expression of the head NP, as highlighted in (9) to (11) below. (9) genitive: [[NPpossessor] no [NPpossessee]]. (10) NP-pronominal: [[NPpossessor][PN nopossessee]]. (11) S-pronominal: [[Smodifier ][PN nomodified ]]. As can be seen in (7) and (8) above, and highlighted in (10) and (11), pronominal no constructions can be further distinguished semantically and syntactically. In (7), we can still see a strong genitive link through the presence of a possessor NP – in this case Taroo. Thus, as highlighted in (10), this type of constructions are here referred to as ‘NP-pronominal no constructions’. In (8), instead of a possessor NP preceding the morpheme no, we see a clausal construction – albeit reduced, and in this case encoding the event Taroo-ga katta “Taroo bought (a retrievable, hence elided, object)” – and the effect is that of an event being reified (i.e. nominalized) for the purpose of providing additional information to identify the referent. We here refer to such cases as ‘S-pronominal no constructions’ since the modifier is a clause (S), as highlighted in (11). Note that S-pronominal no constructions such as (8) above are in fact analyzable as headless relative clause constructions, as highlighted in (12) below where the elided head noun is shown as Ø. (12) [S Taroo-ga kat-ta] no Ø takakatta Taroo-nom buy-past pn/rel(elided NP) expensive-past “the one Taro bought was expensive” While headed relative clauses with no are considered ungrammatical in contemporary Japanese, they are attested in child Japanese, as illustrated in (13), and were also found in government documents in early Modern Japanese (late 19th to early 20th centuries), as shown in (14) below. (13) [aoi] no buubuu blue rel car “a blue car” (or, “a car which is blue”) (14) [seken-o odorokasu] no enzetu world-acc surprise rel speech “the speech which surprised the world”
(Clancy 1985: 459)
(Soga & Fujimura 1978)
It is not altogether surprising that Japanese children sometimes overgeneralize and produce headed relative clauses, given the availability of associative no phrases such
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.7 (143)
From pronominalizer to pragmatic marker
as (15) and (16) – in both Classical and Modern Japanese – which are semantically and structurally closely related to genitive no constructions. (15) Mukasi-yori ooku-no sirabyoosi old.times-from many-gen sirabyoosi.dance “a lot of Sirabyoosi dances from old times” (Heike monogatari, 13th C; Konishi 1969: 107) (16) Kin no tamago gold gen egg “golden eggs, eggs of gold”
(contemporary Japanese; Le 1988: 98)
The morpheme no also appears in head-internal relative clauses, as seen in (17) below, in contrast to ordinary relative clause constructions, as shown in (17’). (17) John-wa [doroboo-ga heya-kara detekita] no-o tukamaeta. John-top thief-nom room-from came.out noml-acc caught “John caught the thief who was coming out of the room.” (Horie 1998: 185) (17’) John-wa [heya-kara detekita] doroboo-o tukamaeta. John-top room-from came.out thief-acc caught “John caught the thief who was coming out of the room.” (Horie 1998: 185) As noted in Horie (1998: 185), whereas the ordinary relative clause (17’) encodes the appearance of the thief independently of the unfolding event of the thief ’s capture by John, the head-internal relative clause (17) encodes both situations simultaneously. Arguably, this ‘2-in-1’ compression becomes possible through the role of no as a “sentential nominalizer” (Horie 1998: 172). The morpheme no is also used in another type of sentential nominalizer construction, namely, a complementizer construction, as shown in (18) below. (18) [S Kinoo kyozin-ga make-ta] no sit-tei-ru. yesterday Giants-nom lose-past comp know-gerund-nonpast “(I/We/He/She/They) know that the Giants lost (the game) yesterday.” In his analysis of the referential uses of no, Horie (1998) concludes that the pronominal uses of no appear to be the crucial link – or the ‘bridging context’, in our current terminology – that captures the semantic relationship between the morpheme’s genitive and complementizing functions. In mapping the functions of no with its corresponding structural patterns, we see the following: Genitive Pronominal Pronominal Complementizer NP no NP NP no Ø S no Ø (NP) [S no] VP
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.8 (144)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
Horie (1998: 172) notes that the four types of no constructions above are related to each other through “family resemblance” characteristics as follows: Genitive no and NP-pronominal no are both “preceded by a dependent noun in dependent-head construction . . . [they] are, however, distinguishable in terms of the presence and absence of the lexical head noun.” NP-pronominal no and S-pronominal no “both function as pronouns” but the former is modified by a noun phrase, the latter by a sentential clause. S-pronominal no and complementizer no are both preceded by a clause; however, “the former no serves as an external head to the preceding relative clause, while the latter no serves as a sentential nominalizer preceded by a complement clause.”
Headless relative clause constructions fall under the category of S-pronominal no constructions, i.e. [S no Ø]. Headed relative clauses, as attested in child Japanese and pre-Modern Japanese, on the other hand, involve a related but more extended structure, namely [S no NP]. Thus, in terms of syntactic patterns, we observe the following family resemblance relationships for no, where headed relative clauses can be interpreted as an (over)extension from headless relative clauses, i.e. from S-pronominal no constructions: Genitive Pronominal Pronominal Complementizer NP no NP NP no Ø S no Ø (NP) [S no] VP | Headed Relative Clause Marker [S no NP] In the evidential domain, no has been identified as a stance marker that may denote assertion (Aoki 1986), as illustrated in (4) earlier, or group authority (Cook 1990), as illustrated in the Mother-Child interaction in (19) below.1 Cook (1990: 432) emphasizes that in using no as a positive politeness marker to mitigate her reprimand, the Mother at the same time evokes a sense of group authority, which in this case serves to subordinate the individual desire and intention of the child to those of the group. (19) Mother:
Sonna fukanoo na koto iwanai no. that impossible assoc thing say.not stance “Don’t say such an impossible thing.” (Cook 1990: 432; gloss added)
Such stance no constructions share overlapping structures with complementizer no constructions, as highlighted in (20)–(21) below, as well as with headless relative clause constructions such as (12), reproduced below as (22). Note that all three uses
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.9 (145)
From pronominalizer to pragmatic marker
of no (stance marker, complementizer, and pronominalizer in a headless relative clause construction) involve S-pronominal no structures. (20) [Asserted S [S Kinoo kyozin-ga make-ta] no.]] yesterday Giants-nom lose-past stance “The Giants lost (the game) yesterday.” (assertion) (21) [COMP [S Kinoo kyozin-ga make-ta] no]] yesterday Giants-nom lose-past comp sit-tei-ru. know-gerund-nonpast “(I/We/He/She/They) know that the Giants lost (the game) yesterday.” (22) [PN [S Taroo-ga kat-ta] no]] takakatta Taroo-nom buy-past (pro)nominalizer expensive-past “the one Taro bought was expensive” From a functional perspective, all three constructions share a common characteristic: the unchallengeable status of the information encoded in [S no]. As noted in Iwasaki (1985: 127), information in the relative clause is typically understood to be “non-challengeable”. Thus, with respect to the headless relative clause construction in (22) above, the hearer could retort with Ie, takakunakatta “No, it wasn’t expensive”, but not Ie, Taroo kawanakatta “No, Taroo didn’t buy it”. That is, one could disagree with the main proposition, but it would be more awkward to challenge the information within the relative clause. Crucially, Iwasaki goes on to show that, in Japanese, information within no desu constructions (here referred to as stance no constructions) likewise attains a non-challengeable status. This observation holds true for the assertion made in the stance no construction in (20) above. In this paper, we extend Iwasaki’s argument and note that the information presented in the complementizer no construction, as illustrated in (21) above, is likewise usually treated as non-challengeable. That is, while the speaker’s perception or assessment can be challenged, in the sense that the speaker may or may not actually know the stated fact, the fact itself (marked in Japanese by nominalizer no) is not normally open to challenge. What is worth noting here is that the common link between the headless relative clause and the stance and complementizer constructions – namely, the unchallengeable status of the information marked by no – is more than mere coincidence.2 To briefly sum up, based on their family resemblance characteristics, we see the following morphosyntactic relationships for the no constructions discussed thus far, and it is worth noting that the functionally-related S-pronominal, stance and complementizer no constructions are structurally similar/contiguous to each other:
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.10 (146)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
Genitive Pronominal Pronominal Complementizer NP no NP NP no Ø S no Ø (NP) [S no] VP Stance [S no] From a semantic perspective, particularly in terms of Lyons’ (1977) hierarchy of ontological entitities – where a thing/animal/person is analyzed as a first order entity, an event as a second order entity, and a proposition as a third order entity – Horie (1998) notes that usage of genitive no and NP-pronominal no is restricted to modifying (i.e. specifying) first order entities, S-pronominal no to either first or second order entities, and complementizer no to third order entities. Here we add that stance no constructions, being evaluative or attitudinal, and hence more abstract, also involve third order entities. The relevant semantic and syntactic mappings are highlighted in Figure 1 below. First Order Entities
Second Order Entities
Third Order Entities
(NP-possessee = thing/animal/person)
(Ø = thing / Ø = event)
(Ø = proposition, attitude, etc.)
Genitive NP no NP
S-Pronominal S no Ø
Complementizer (NP) [S no] VP
NP-Pronominal NP no Ø
Stance [S no]
Figure 1. A mapping highlighting the ontological entities encoded by the grammatical functions of Japanese no constructions
In the next section, we will examine how well the family resemblance network and the ontological hierarchy seen above correlate with diachronic evidence for the uses of no.
.. Diachronic development of no The genitive and associative uses of no are attested as early as the 8th century, as illustrated in (23) below. (23) otome no toko no be ni young.girl gen bed assoc side loc “on the side of the bed of a young girl” (Koziki [Anthology of Old Tales], 749 AD; cited in Nihon Kokugo Daiziten, vol. 10: 754)
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.11 (147)
From pronominalizer to pragmatic marker
Certain types of pronominal no constructions – more specifically, the NPpronominal type – are also attested as early as the 8th century, as shown in (24). (24) yakusi wa tune no mo are do . . . pharmacist top regularity pn also exist but “though there exists a pharmacist of the regular kind” (lit. “As for a pharmacist, there exists (a pharmacist) of the regular kind but . . . ”) (Bussokusekika, c. 753 AD; cited in Nihon Kokugo Daiziten, vol. 10: 754) S-pronominal no constructions – such as the headless relative clause shown in (25) below – are found much later, in texts from the 16th and 17th centuries. (25) soregasi ga suite yomu no wa 1sg nom like read pn top “the thing I like to read is . . . ” (Kyoogenki, 1660 AD; cited in Nihon Kokugo Daiziten, vol. 10: 753) Complementizer no constructions also appear around the late 16th to early 17th century, as illustrated in (26). (26) uta tatematu-re to ooserare-keru song give=Humble-Imperative Quotative say=Honorific-past toki-to aru no-wa uta-no tehon-ni time-quotative exist noml-top song-gen example-loc tatematu-re to aru no nari. give=humble-imperative quotative exist noml to.be “It is (cited) that when (X) respectfully said “Humbly present a poem,” that (actually) means “Humbly present as an example of a poem.”” (Lit. “. . . it (actually) means that (X) should humbly present an example of a poem”) (Ziteiki, late 16th to early 17th C; cited in Zidaibetu Kokugo Daiziten, Muromati Zidai Hen vol. 4: 549) Uses of sentence-final particle no as stance markers were also evident around the end of the 16th to the beginning of the 17th century, as shown in (27). (27) . . . mairi tai no come=Humble want stance “. . . (I) would like to humbly visit” (Toraakirabon Kyoogen, c. 1600/1700 AD; cited in Nihon Kokugo Daiziten, vol. 10: 754)
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.12 (148)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie 8th c. 10th c. 12th c. 14th c. 16th c. 18th c. 20th c. Genitive no NP-Pronominal no S-pronominal (e.g. headless relative clause) no ~Headed relative clause no Complementizer no Sentence-final particle (i.e. stance) no
Figure 2. Diachronic development of no based on written records reported in Nihon Kokugo Daiziten
Figure 2 captures the diachronic development of no based on written records reported in Nihon Kokugo Daiziten and Zidaibetu Kokugo Daiziten, two authoritative classical Japanese dictionaries. Diachronic evidence thus suggests the following grammaticalization path for no: Genitive no/ NP-pronominal no > S-pronominal no > Complementizer no / Stance no | Headed relative clause marker no (brief phenomenon) Significantly, the above grammaticalization path conforms not only to diachronic evidence, but also to a semantically-based ontological hierarchy and family resemblance linkage patterns that reflect overlapping semantic and syntactic chains.
. Mandarin de and other dialectal equivalents Most Chinese dialects have a morpheme comparable to Japanese no, with a range of functions including possessive marker, relativizer and nominalizer (e.g. Mandarin de, Cantonese ge, Hakka ke, and Hokkien e). Several studies have pointed out the parallels between Japanese no and Mandarin de (Kitagawa & Ross 1982; Simpson & Wu 2001). A further extension to the stance marking function is also fairly widespread among the Chinese dialects. The Mandarin particle de is of particular interest to us since the continuous history of written texts allows the diachronic pathway of development to be reconstructed with some confidence. The resulting pattern differs from that we have outlined for Japanese, thus raising challenging questions concerning directionality. The lexical source of de appears to be the word di meaning “end” or “bottom”, as illustrated in (28) below:3
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.13 (149)
From pronominalizer to pragmatic marker
(28) Mu hao shi dang-di jiao-fan. mother like eat pan-bottom crust-rice “His mother likes to eat the rice crust from the bottom of the pan.” (Shi shuo xin yu-de xing, AD 500; adapted from Shi & Li 2002: 13) In 5th century texts, di already had demonstrative and interrogative as well as lexical functions: (29) di nu yao dang jie ru tou dem slave want to cut 2sg head “This slave wanted to cut your head off.” (Bei zhai yan zhi tui , 5th C) (30) ge ren du di? this man read what “What is this man reading?”
(Bei zhai shu xu zhi cai zhuan, 5th C)
The modern grammatical functions of de begin to appear in 7th century texts such as Dunhuang Bianwen, where it is found in the position of relativizer (Huang 1989; Shi & Li 2002). (31) ru deng zuo ye jian di guang . . . you pl last night see rel light “The light you saw last night, . . . ” (Dun Huang Bian Wen, 7th C; Jiang 1997) The origin of the relative use appears to lie in reanalysis of a demonstrative following a relative clause (Shi & Li 2002: 13), as highlighted in stages (i) and (ii) in (32) below; the headless relative clause then arises by ellipsis of the head noun, as in stage (iii). (32) (i) [relative clause [deDEM N]] (ii) [relative clause de] N] (iii) [relative clause] de Ø During the Song period (960–1279) pronominal uses appear, including NPpronominal (33) and S-pronominal (34) constructions: (33) Zhe ge shi lao zhen di dem cl be old monk pn “This one is the old monk’s”
(Jing de zhuan deng lu; Lu 1943: 123)
(34) Nian shun lao zhen jiao di roll hurt old monk foot pn “(The thing) that ran over the old monk’s feet.”
(ibid.)
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.14 (150)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
In contrast to Japanese no, the genitive use of de is clearly a later development, attested only from the 12th century onwards, as shown in (35). (35) Yao zuo Zhou gong di shi-ye must do Zhou Sir gen business “One must carry out Sir Zhou’s business.” (Zhu zi yu lei; Zhu 1982: 195) The order of development as represented in texts thus appears to be: Lexical > Demonstrative/Interrogative > Relative > Pronominal > Genitive We thus have strong textual evidence that the diachronic developmental pathway in Chinese differs from that in Japanese. In Section 4, we take up the challenges to directionality which these differences present. In modern Mandarin, de extends to stance usages as in (36). (36) Ta bu hui lai de 3sg not will come stance “S/he won’t come.” Li, Thompson and Zhang (1998) describe such usages as evidential. Due to the nature of the written records, the extension to stance marker, which even today is restricted to colloquial language, cannot be dated so easily. Contemporary spoken Cantonese enables us to reconstruct a likely scenario for how it arose. As shown in (37) to (40) below, the particle ge has a range of functions corresponding closely to Mandarin de: (37) gaaze maai ge je sister buy rel thing “thing(s) that Sister bought” (38) gaaze maai ge sister buy pn “one that Sister bought”
(relative ge)
(pronominal ge)
(39) hai gaaze maai ge is sister buy noml “it was Sister who bought it”
(cleft with nominalizer ge)
(40) gaaze wui maai ge sister will buy stance “Sister will buy it (for sure)”
(stance with sentence final particle ge)
Potential bridging contexts include constructions such as (41), which can be taken as a pronominal and/or cleft construction:
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.15 (151)
From pronominalizer to pragmatic marker
(41) hai gaaze maai ge is sister buy pn/noml “It’s one that Sister bought” or “It was Sister who bought it”
(pronominal ge) (cleft with nominalizer ge)
Then, given that the copula hai can regularly be omitted (Matthews & Yip 1994), dropping it results in a configuration in which ge can be taken either as marking a cleft sentence, or as a final particle encoding the speaker’s stance, as illustrated in (42) below. (42) (hai) gaaze maai ge is sister buy noml/stance “It was Sister who bought it” (cleft with nominalizer ge) or “Sister bought it (for sure)” (stance with sentence final particle ge)
. Malay (em)punya Malay also has a morpheme that serves genitive, pronominal and stance marking functions. This morpheme, (em)punya, is derived from a noun empu (“master”) accompanied by a third person enclitic -nya, and is used to express meanings related to “ownership”, “lordship”, “master-craftsmanship”, etc.4 The use of empunya was attested in literary texts as early as the 14th century, when it was already used as a noun and as a verb. As a noun, it preceded a possessee NP and was often accompanied by the nominalizer yang, as shown in (43), while as a verb, it occurred in simple clauses such as NP1 empunya NP2 , as illustrated in (44). (43) (yang) empunya negeri itu noml master country dem:dist “the ruler (of the country)” (Hikayat Bayan Budiman 165: 11, 14th C) (44) Aku empunya salam doa kepada sahabat-ku itu I have greeting prayer to friend-1sg that “I bring ( verbal empunya development. (45) . . . mahligai itu Puteri Kamariah yang empunya dia palace dem:dist Princess Kamariah noml/rel own(er) 3sg
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.16 (152)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
“. . . that palace, Princess Kamariah (is) its owner” (nominal reading) “. . . that palace, Princess Kamariah (is) the one that owns it” (verbal reading) (Hikayat Bayan Budiman 157: 24, 14th C) In structural terms, (45) above can be interpreted either as (46) or (47) below, the former yielding a lexical noun reading for empunya, while the latter gives rise to a lexical verb reading. (46) [S [NP Puteri Kamariah] [VP zero-copula [NP (NOML yang) [NPpossessor empunya] [NPpossessee dia]]]] (47) [S [NP Puteri Kamariah] [VP zero-copula [REL yang [VP empunya [NPpossessee dia]]]]] In natural discourse, under real-time constraints, speakers and hearers may not need to choose between the two interpretations. Both nominal and verbal readings of empunya in (46) and (47) respectively activate the same set of words, and in the same processing sequence. Underspecification with respect to the noun/verb distinction in instances such as (45) yields no interpretative problems. In fact, cognitive processing efficiency often motivates such underspecification. This facilitates a general acceptance of lexical verb interpretations of empunya through structural patterns of the type seen in (47), which in turn paves the way for the emergence of ‘switch contexts’ that yield clear lexical verb interpretations of empunya, as shown in (44) above. The example is reproduced in (48) below to highlight the awkwardness of a lexical noun interpretation for empunya in the context given. (48) Aku empunya salam doa kepada sahabat-ku itu I have greeting prayer to friend-1sg that “I bring (
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.17 (153)
From pronominalizer to pragmatic marker
sive marker.5 To some extent, variations in the stress and intonation patterns could render either a nominal, verbal or possessive reading. In many contexts, however, an indeterminate reading works just fine. In fact, for most practical purposes, the phonological differences are often too subtle to make a significant difference in real processing time. The resulting indeterminacy or underspecification readily paves the way for the emergence of genitive uses of empunya. (49) Bukannya hamba empunya kerajaan neg.3sg 1sg(
Thus, in structural terms, empunya in (49) above can be interpreted either as lexical noun (50), or a lexical verb (51), or a grammatical possessive marker (52), as shown below. (50) . . . [S [NP hamba] [VP zero-copula [NP [NPpossessor empunya] [NPpossessee kerajaan]]]] (51) . . . [S [NP hamba] [VP empunya [NPpossessee kerajaan]]]] (52) . . . [NP [NPpossessor hamba] [GEN empunya] [NPpossessee kerajaan]] Clear uses of empunya as a genitive (or grammatical possessive marker) become evident in ‘switch contexts’ such as (53) below, where a verbal or nominal reading would be quite illogical. The scenario for (53) involves the famous Malay warrior Hang Jebat, who had got into mischief, and is here being blamed for the mess he created. The most appropriate reading of the empunya construction in this particular context would be “this is his work, or doing”, rather than “he has this work”, or “he is the possessor of this work”, hence the ‘switch’ to a genitive interpretation. (53) . . . kerana ia empunya pekerjaan ini . . . because 3sg gen work dem:prox
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.18 (154)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
“. . . because this is his doing.” (possessive reading) * “. . . because he has this work” (verbal reading not logical) * “. . . because he is the possessor of this work” (nominal reading not logical) (Hikayat Hang Tuah 218: 16, 18th C) As often happens elsewhere, empunya is also susceptible to phonological reduction. The reduced form punya was first attested in Taj al-Salatin (original text 1603; available manuscript around late 18th century), with both nominal and verbal usages that closely parallel empunya. In this particular text, empunya was still very productive, particularly when used lexically as a noun meaning “owner” (e.g. empunya rumah “owner of the house”; empunya dusun “owner of the orchard”; empunya kambing “owner of the goat”; empunya unta “owner of the camel”). Lexical verb uses were also still productive in this text, but by this time were already being more clearly marked with verbal affixes mem- and -i. Interestingly, the verb mempunyai is ambiguous between m(e)-empunya-i and mem-punya-i. Evidence of maampunyai in Hikayat Banjar dan Kota Waringin (original text ca. 1663; available manuscript 19th century), as reproduced in (54) below, suggests that phonological reduction from empunya to punya has most likely been facilitated via the lexical verb usages.6 (54) Ia itu maampunyai sagala parentah hukum dalam nagri. 3sg def pref-possess-suf all order law inside country “He has absolute authority within the country.” (Hikayat Banjar dan Kota Waringin 3867, 19th C) The lexical noun usage of punya is also highly productive and very prominent in Taj al-Salatin. More interestingly, in this particular text, the lexical noun usage of punya is semantically different from that of empunya. As shown in (55) below, punya is being used to refer to ‘possession’, unlike empunya, which refers to ‘the possessor, or owner’. This is indicative of a semantic split that gives rise to a new lexical noun usage, once the reduced form punya has emerged. (55) . . . harta itu punya segala hamba Allah juga. inheritance def possession all servant(s) God also “. . . the inheritance is the possession of (i.e. belongs to) all of God’s servants, too.” (Taj al-Salatin 76: 7) The semantic split raises an interesting question: How does the possessor NP represented in empunya get reinterpreted as a possessed NP in punya? Bridging contexts from Taj al-Salatin, such as (56) and (57) below, offer us some clues. In both examples, punya could be read as ‘possessor’ if topicalized – that is, if followed by falling intonation and a long pause – or it could be read as ‘possession’ if no
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.19 (155)
From pronominalizer to pragmatic marker
extended pause intervenes. In Taj al-Salatin, the latter reading appears to have gained ground. Indeed, given more context, it is the preferred or more natural interpretation. (56) Maka adalah orang bertanyakan, conn exist-foc person pref-ask-suf “And it was that someone asked, “Punya siapa segala mahligai itu Ya Rasulullah.” ?possessor/possession who all palace def O God ? “Who is/are the possessor(s) of all those palaces, O God?” ” √ “Whose posssessions are all those palaces, O God?” (Taj al-Salatin 187: 39) (57) Maka sabda Rasulullah, and say God “And God said, “Punya segala orang yang berkata-kata ?possessor/possession all person rel pref-speak-speak lemah lembut . . . weak soft ? “The possessors (are) all those that speak gently . . . ” ” √ “The possession(s) of all those that speak gently . . . ” (Taj al-Salatin 188: 1) Other examples in the same text (i.e. Taj al-Salatin) provide evidence of bridging contexts that welcome a more abstract possessive interpretation for punya. Consider (58) below. Given that perkataan “word(s)” is ideational and hence a third order entity, the use of punya to mean ‘possessor of a word(s)’ or ‘possession in the form of a word(s)’ becomes somewhat clumsy, and the more abstract concept of ‘possessive relationship’ becomes much more dominant instead. (58) . . . dan daripada keempat perkataan itu . . . and from all.four word(s) def “. . . and of those four words,” sepatah kata punya aku one.cl word ??possessor/?possesion/genitive 1sg ?? “(for) one word, the possessor is me” ? “(for) one word, the possession is mine” √ “one word is mine” dan sepatah kata punya engkau. and one.cl word ??possessor/?possession/genitive 2sg
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.20 (156)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
?? “(for) one word, the possessor is you” ? “(for) one word, the possession is yours” √ “one word is yours” (Taj al-Salatin 68: 22) Preferred reading: “. . . and of the four words, one is mine, and one is yours.” Clear uses of punya as a genitive (i.e. possessive marker) emerge later, through switch contexts where nominal and verbal readings of punya are no longer possible, as illustrated in (59). (59) sengaja ia tiada mahu bayar sahaya punya wang purposely 3sg neg want pay 1sg gen money * “he purposely didn’t want to repay, I’m the one with money” (*nominal reading) * “he purposely didn’t want to repay, I have money” (*verbal reading) √ “he purposely didn’t want to repay my money” (possessive reading) (Hikayat Abdullah 173: 10, 19th C) Note that a lexical noun or lexical verb reading of punya in (59) above would force a biclausal analysis, as highlighted in (60) and (61) below. (60) [S . . . ia tiada mahu bayar Øi ] [S [NP sahaya] [VP [zero-copula] [[NP punya] [NP wangi ]]]] (61) [S . . . ia tiada mahu bayar Øi ] [S [NP sahaya] [VP [V punya] [NP wangi ]]] The resulting biclausal analysis, however, renders an unacceptable interpretation of (59) within the discourse context, since the intended focus of the utterance is on the repayment of a debt that the addressee owes to the speaker, and not on the speaker having money, or the speaker being the one with money. A genitive interpretation of punya, on the other hand, appears superficially to be monoclausal, as highlighted in (62) below; in fact it represents a compression of a much more complex biclausal construction, namely, the lexical noun empunya type shown in (60). This highlights an important property of genitive constructions in general: they are both syntactically and semantically compressed, making an already abstract notion of possession even more abstract and elusive (or general), but in so doing, accepting a wider range of possessor-possessee combinations. (62) [S . . . ia tiada mahu bayar [NP [NP sahaya] [GEN punya] [NP wang]]] Usage of genitive punya abounds from the 19th century onwards, particularly in texts that recount the dealings between the local Malays and the Bugis traders and British officials. It is also worth noting that the prolific use of punya as a possessive marker also coincided with an influx of immigrant workers from Southern
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.21 (157)
From pronominalizer to pragmatic marker
China to the Malay peninsula during the tin-mining boom years in the mid-19th century. It is possible then that the genitive uses of punya could have been further entrenched through the influence of similar constructions in Chinese, involving morphemes such as Cantonese ge and Hokkien e. This ‘calquing’ account would help explain why genitive punya tends to be frequently identified with Bazaar Malay.7 Pronominal uses of punya were also attested as early as the 18th century, again in Taj al-Salatin, with ample evidence of productive use in the 19th century, in texts such as Kisah Pelayaran Abdullah, Hikayat Abdullah, Hikayat Marakarma and Salasilah Melayu dan Bugis. Pronominal usage appeared first in NP-pronominal constructions such as (63) and (64) to express possessive relations – in other words, as pronominal possessive constructions. (63) . . . jika harta itu daripada segala hamba Allah punya, . . . . . . if wealth dem:dist from all servant God pn “. . . if the wealth comes from that of all of God’s servants, . . . ” (Taj al-Salatin 76: 20, 18th C) (64) Kata seorang “Ini aku punya”; kata seorang, “Aku say one.person dem:prox 1sg pn say one.person 1sg punya.” pn “Said one person, “This is mine”; said another, “(It’s) mine.”’ (Kisah Pelayaran Abdullah 111: 8, 19th C) Pronominal punya was later extended to S-pronominal constructions in contemporary colloquial speech, as illustrated in (65) and (66), to express modifications to the referent, often via associative phrases or (reduced) relative clause constructions.8 (65) (Yang) koyak punya campak saja lah. noml torn pn throw.away just sfp “That which is torn, just throw (it) away.”
(colloquial use, 20th C)
(66) Aku tak mau (yang) rosak punya. 1sg neg want (noml) spoilt pn “I don’t want spoilt ones.” “I don’t want those that are spoilt.”
(colloquial use, 20th C)
The use of punya as a stance marker was not attested in the literary texts from the 14th to 19th centuries (based on data from the Malay Concordance Project), but such usage is fairly common in contemporary colloquial Malay – as spoken, for example, in parts of West Malaysia such as the states of Perak and Selangor, with the examples from Gil (1999) collected from fieldwork in Malay villages in the Kuala
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.22 (158)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
Lumpur vicinity). As a stance marker, punya is used in sentence-final position to convey assertion, as illustrated in (67). (67) [Speaker playing billiards, aims ball; interlocutor, who is watching, says it won’t go in; speaker shoots and pockets the ball, and then says:] Masuk punya. Enter punya [=stance by our interpretation] “It went in.” (Gil 1999: 7) [Our interpretation: An assertion in the sense of “It was sure to go in.”] As in the case of Japanese no and Mandarin de, emergence of the stance-marking function of punya has most likely been facilitated by the rise of S-pronominal uses of punya. The structural similarity between the S-pronominal constructions koyak punya and rosak punya, in (65) and (66) above, and the stance construction masuk punya, in (67), is particularly revealing: both involve an [S punya Ø] constituent. What is different is that the constituent in koyak punya occurs within a headless relative clause, while the constituent in masuk punya occurs within an independent clause, as highlighted in (68) and (69) respectively. (68) (yang) [S koyak] punya (Ø=benda) noml torn pn (elided np=thing) “that (Ø=thing) which is torn” (69) [S (Itu) masuk] punya. dem enter stance “That/It was sure to go in.” Note that stance punya in colloquial Malay (much like stance one in Singapore English, e.g. It will go in one, meaning “It will surely go in”) serves as a nominalizer/sentence-final particle, transforming a factual proposition into an evaluative one through an act of reification on the part of punya. The availability of cleft/focus constructions such as (70) also facilitates the emergence of stance punya constructions.9 Crucially, the cleft/focus construction readily serves as a ‘bridging context’, where ambiguity concerning the scope of punya (or its reduced form mia) gives rise to either a pronominal or stance interpretation. For the pronominal reading, the scope of punya is necessarily narrower, extending only over the modifying elements (e.g. [NP susah mia] “a difficult one”), whereas for the stance reading, the scope of punya is much broader, extending over the whole clause or proposition (e.g. [S yang ni susah mia] “this one’s difficult”). (70) [The speaker, while watching a friend choose a new game, says:] A, yang ni susah mia. excl noml dem:prox difficult punya
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.23 (159)
From pronominalizer to pragmatic marker
“Yeah, this is a difficult one.” (pronominal reading) “Yeah, this one’s difficult.” (stance reading) (adapted from Gil 1999: 8) Many cleft/focus constructions involving punya would, when decontextualized, readily yield an indeterminate reading, with both pronominal and stance options available. With sufficient contextual information, however, disambiguation is often relatively easy. In (71), a stance interpretation emerges, since the focus is not on which clock, but rather on whether the clock was telling the right time. (71) [In response to his passenger’s surprised reaction to the time shown on the clock in his taxi, the driver comments:] Ini tak pakai mia. dem:prox neg use punya [= s-pronominal/stance] *“This is the one that’s not being used.” (pronominal reading) “This is not being used.” (stance reading) (adapted from Gil 1999: 8) In sum, diachronic evidence based on texts from the 14th century to the present (see also Table 1 below) yields the following grammaticalization path for (em)punya: > Lexical verb Lexical noun 1 (possessor) > Genitive > Pronominal > Stance > Lexical noun 2 (possession) (facilitates the genitive>pronominal>stance development)
Note that the genitive > pronominal > stance development for (em)punya is quite similar to that of Japanese no, despite significant differences in morphosyntactic structure between the two languages. Table 1. Summary of diachronic development of (em)punya10
14th c. 16th c. 17th c. 18th c. 19th c. 20th c.
Empunya: Noun N/V Verb
Punya: Noun
9 29 66 137 37 archaic
11 12 10 6 kepunyaan ?
9 2 7 3 3 ?
6 5 28 13 16 archaic
N/V Verb
Punya, mia, nya: Genitive Pronominal Others
11 1 4 1 62 87 18 mempunyai colloquial colloquial
1 Excessive Productive use of Stance (colloquial)
Note: Frequency counts for the 14th through 19th centuries are based on analyses of classical texts from the Malay Concordance Project, excluding Warisan Warkah Melayu, since this particular collection is multi-dated. We focused our analyses on prose, rather than poetry, collections.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.24 (160)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
In ontological terms, we also see a progression from first order entities to third order entities. Recall that the lexical noun uses of empunya “owner” involves animate, human (or at least sentient and divine) entities. The genitive uses of (em)punya likewise focus on human or divine possessor NPs. Pronominal uses of punya initially referred to concrete NPs, but in colloquial Malay have extended to events and propositions as well (e.g. Itu tak adil punya “That’s not fair”, referring to a financial settlement or legal decision, etc.). As noted earlier, the line between S-pronominal and stance constructions is a very fine one, but whereas pronominal punya can refer to first, second and third order entities, stance punya must necessarily evaluate propositions and hence are linked to third order entities only. What is worth emphasizing here is that the diachronic development of (em)punya reflects a sensitivity to ontological constraints, not unlike its Japanese and Mandarin counterparts.
. Implications for unidirectionality Returning to the question of directionality, we have noted that the range of functions exhibited in the various East and Southeast Asian languages considered overlaps in striking ways: the morphemes concerned cover most or all of the points in Figure 3. On the other hand the historical sequence, where the evidence is available, varies considerably, as summarised in Figure 4 below. We are therefore not looking at a straightforward unidirectional pathway.
Complementizer (NP) [S x] V Genitive NP x N
NP-Pronominal NP x ø
S-Pronominal
Stance
Sxø
Sx
Cleft/Focus construction NP (COP) S x Headed relative clause SxN
Figure 3. Grammatical functions centered on (pro)nominalizers as an implicational map. Note that x represents the (pro)nominalizer morpheme under investigation (e.g. Japanese no, Mandarin de, and Malay empunya).
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.25 (161)
From pronominalizer to pragmatic marker Japanese no: genitive / NP-pronominal
> S-pronominal > complementizer & cleft > stance > ?headed relative clause
Mandarin de: genitive < NP-pronominal < S-pronominal > (*complementizer) & cleft > stance < headed relative clause Malay punya: genitive > NP-pronominal > S-pronominal > (*complementizer) & cleft/focus > stance | *headed relative clause
Figure 4. Observed diachronic development of stance markers from (pro)nominalizers based on a crosslinguistic comparison of some East and Southeast Asian languages
. Ontological levels Recent work has suggested that unidirectionality might apply, not at the level of individual grammatical functions, but at a semantic level (e.g. Ziegeler, this volume). With respect to Japanese, Horie (1998) has proposed that, in terms of the ontological cline of Lyons (1977), development of no proceeded from first-order entities (things/persons) to second- and third- order entities (events and propositions). Here we note that this development from lower to higher ontological levels would be consistent with directionality in the semantic domain along the lines recently suggested by Ziegeler. From this perspective, let us reconsider the prima facie challenges to unidirectionality posed by the diachronic findings in the three languages discussed in this study. Although Malay (em)punya lacks a complementizer construction equivalent to complementizer no found in Japanese, the direction of grammaticalization is otherwise very similar. Chinese de, on the other hand, appears to pose a serious challenge to unidirectionality, since the relativizer and pronominal functions precede the genitive one. In ontological terms, however, this challenge can be resolved. Note first that this is not a case of degrammaticalization, in the sense of a grammatical function giving rise to a more lexical one. Rather, it involves a development from one grammatical function to another. Moreover, in terms of the ontological cline, both relative and genitive uses belong to the same ontological domain, i.e. first-order entities. Grammaticalization can proceed either from genitive to pronominal (and sometimes to relativizer), as in the case of Japanese no and Malay (em)punya, or from relativizer to pronominal and then genitive, as in the case of Mandarin de, without violating directionality at the ontological (i.e. semantic) level. Note further that semantic extensions can proceed from first-order to secondorder entities, not only across grammatical domains, but also within the same grammatical domain. For example, in Mandarin, within the S-pronominal category, i.e. headless relative clauses, semantic extensions can proceed from first-order entities in constructions such as ta chi de (“what he eats”) to second-order en-
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.26 (162)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
tities as in ta zuo de (“what he does”), where the reference shifts from concrete things to events. Such shifts within a grammatical domain can pave the way for extensions into new grammatical domains, as in the extensions from S-pronominal (i.e. headless relative clause) constructions into cleft, complementizer and stance constructions, all of which represent the emergence of grammatical categories involving third-order entities. In the case of stance constructions, the function of the grammaticalized morpheme is to indicate the speaker’s perspective on the proposition. In the three languages examined in this study a converse direction from higher to lower ontological entities does not occur, at least not across grammatical domains. More specifically, none of the three languages show evidence of reverse grammaticalization from stance/complementizer/cleft constructions to relative/pronominal/genitive constructions. This constitutes evidence in support of semantic unidirectionality.
. Conceptual networks/implicational maps Given that bidirectional extensions are possible within grammatical domains, the question arises as to whether an implicational hierarchy could be postulated for the grammaticalization of pronominalizers into stance markers. Consistent with the implicational hierarchy principle, which stipulates that the existence of functional categories A and C would implicate the existence of functional category B as an intermediate functional category, Horie (1998) posited that it is not likely that a morpheme serving both genitive and complementizer functions would fail to also manifest the existence of a pronominal function. From an ontological perspective, this contiguity predicted by the implicational hierarchy is well preserved in the grammaticalization of no, de and (em)punya. What is significant is that bidirectionality within the same ontological domain does not disrupt or invalidate the contiguity principle. This in fact is consistent with Haspelmath’s (1997) notion of implicational maps. To elaborate, an implicational map is a network such that grammatical functions of a morpheme must cover a contiguous portion of the map. While this prerequisite is essentially a synchronic constraint, it represents the outcome of diachronic processes. Thus, the diachronic evidence for the development of de – which first extends from headed relative clause constructions into the pronominal domain, and then into cleft constructions, before developing the genitive function – is consistent with the contiguity principle articulated in Haspelmath’s implicational map formulation. This mapping between diachronic evidence and the implicational map for de is highlighted in Figure 5 below. The convergence of diachronic evidence and the contiguity constraints of the implicational map also has important ramifications for the semantic conceptual
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.27 (163)
From pronominalizer to pragmatic marker Headed Relative Clause de Genitive de < NP-Pronominal de < S-Pronominal de > Cleft de > Stance de (i.e. headless relative clause) 12th c.
9th c.
7th c. or earlier
10th c.
contemporary
Figure 5. A mapping highlighting the convergence of diachronic evidence and the contiguous nature of the development of Mandarin de
network articulated in Ziegeler (this volume). Essentially, all three accounts jive, and we see more clearly how grammaticalization – as the process whereby a linguistic item extends into new contexts and gives rise to new functional categories – proceeds in contiguous steps, rather than haphazardly. And for each semantic change trajectory that emanates from any grammaticalization node, there is a strong tendency towards unidirectionality.
. Conclusion In this paper, we have addressed two main questions arising from Horie’s (1998) study on Japanese no: (i) to what extent is the directionality of grammaticalization constant across languages, and (ii) how is the further extension of pronominal markers into pragmatic particles to be accounted for? We have shown that the very similar ranges of grammatical functions exhibited by Japanese no, Mandarin de and Malay (em)punya do not always follow identical sequences of development. In particular, Mandarin de shows reverse directionality for portions of the grammaticalization pathway concerned. We have argued, however, that reverse directionality does not always constitute degrammaticalization. More specifically, we have argued that the divergent pathways respect unidirectionality at a semantic level, which we have formulated in terms of an ontological hierarchy along the lines of Lyons (1977) and Horie (1998). In terms of this hierarchy, the emergence of stance markers represents an extension to third-order entities, namely propositions. We have also shown that, synchronically, the data are consistent with the implicational maps hypothesis (Haspelmath 1997), in that the grammatical functions of each morpheme cover contiguous segments within a common network of grammatical functions observed across languages.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.28 (164)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
Abbreviations ACC ASSOC C3 CL COMP CONN DEF DEM DIST EXCL FOC GEN INF LOC
Accusative Associative Noun Class 3 (for Swahili, etc.) Classifier Complementizer Connective Definite Demonstrative Distal Exclamatory expression Focus Genitive Infinitive Locative
NEG NOM NOML PL PN PREF PRES PROG PROX REL SFP SG SUF TOP
Negative Nominative Nominalizer Plural Pronoun/Pronominal Prefix Present Progressive Proximal Relativizer Sentence Final Particle Singular Suffix Topic marker
Notes * This work has been supported by research grants from Sophia University, Tokyo, the Chinese University of Hong Kong (Direct Grant #2010250) and the Research Grants Council of the Hong Kong Special Administrative Region, China (project HKU 7205/99H). We are grateful to Michelle Li for assistance with Chinese data. We would also like to thank the editors and an anonymous reviewer for extensive and constructive feedback, and participants at the Conference on New Reflections on Grammaticalization 2 for their helpful comments, notably Martin Haspelmath. We also wish to acknowledge our indebtedness to Ian Proudfoot and his colleagues for free e-access to literary texts via the Malay Concordance Project based at the Australian National University. . Note that stance uses of no are more commonly expressed as no desu constructions, or n desu for short. . Japanese also has cleft-like no constructions, as illustrated in (i) below. (i)
(watashi-wa) kinoo kita no desu (I-top) yesterday came no be “I came yesterday/ It was yesterday that I came.”
(Simpson & Wu 2001: 252)
Note that, whereas we can challenge the truth condition for the focused constituent kinoo “yesterday” with a reply such as kinoo ja nai (no/-n desu) “(No), it wasn’t yesterday”, the information marked by no – namely, watashi kita “(the fact that) I came” – is understood to be non-challengeable, and the addressee would not normally retort with anata konakatta “(No), you didn’t come.” Pseudo-cleft no constructions are also available in Japanese, as illustrated in (ii) below.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.29 (165)
From pronominalizer to pragmatic marker
(ii) (watashi-ga) katta no wa hon desu (I-nom) bought no top book be “What I bought was a book.”
(Simpson & Wu 2001: 252)
Here, too, the truth condition involving the focused constituent hon “book” can be challenged, via an utterance such as hon ja nai (no/-n desu) “(No), it wasn’t a book.” The information marked by no – in this case, watashi-ga katta no “what I bought” – is again usually understood to be non-challengeable. Thus, the addressee would assume that something was bought, and to otherwise come up with a remark such as Nani mo kawanakatta “(You) bought nothing/(You) didn’t buy anything” is to adopt a confrontational stance. Cleft and pseudo-cleft constructions in Japanese thus share a common characteristic – namely, nonchallengeability – with headless relative clause, stance and complementizer constructions. It is worth noting that all these constructions involve the [S no] constituent. . For a detailed account of the semantic extensions of di and its eventual replacement by de, see Yap, Choi and Cheung (2004). . We are grateful to Ian Proudfoot for discussions on the etymology of empunya. Dictionaries in both Bahasa Malaysia and Bahasa Indonesia, e.g. Kamus Dewan and An IndonesianEnglish Dictionary, 3rd ed. respectively, identify empu as “master”. Kamus Dewan, in particular, equates empunya with tuannya. The latter is unequivocally a combination of tuan “master” + -nya ‘3 person clitic’; tuannya thus captures the meaning “his/her/its master”. This derivation is analogous to that of empunya (empu “master” + -nya ‘3 person clitic’). . Malay also makes use of another possessive construction, namely the NP2 NP1 construction, where NP2 refers to the possessee and NP1 the possessor. In contemporary Malay, the NP1 punya NP2 possessive construction is restricted to colloquial speech, with the pronominal NP1 punya Ø becoming more widely used. . Note that the letter e that is used in most Malay varieties elsewhere is realized as a in this particular text, perhaps due to the influence of Banjar Malay. Malay dictionaries such as Echols and Shadily’s Indonesian-English Dictionary have also identified overlapping meanings between ampu and empu. . Genitive punya is also closely identified with Baba Malay. This variety of Malay is spoken among the descendents of Chinese who were brought to the peninsula by the famous Admiral Cheng Ho during the golden age of the sea-voyaging Ming dynasty (15th century). Intercultural marriages between the Chinese and the Malays would easily facilitate calquing of Chinese genitives (e.g. NP zhi NP, NP zhe NP and NP de NP constructions) onto the Malay genitive (em)punya constructions. . Standard Malay uses yang as a relativizer (e.g. buku yang aku beli “the book that I bought”) and a (pro)nominalizer (e.g. Yang mahal untuk abang “The expensive one is for elder brother”; Yang aku suka dah dijual “That/the one which I like is already sold”). Colloquial Malay also uses yang as a relativizer and (pro)nominalizer, but also accepts punya as a pronominalizer (e.g. (Yang) mahal punya untuk abang “The expensive one is for elder brother”). . The cleft/focus construction in (44) involves an adjectival predicate, thus representing an extended use of associative punya. The cleft/focus construction in (45) involves a verbal predicate; this represents an extended use of S-pronominal punya.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.30 (166)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie . Tokens from the 14th to 19th century texts were retrieved from the database of the Malay Concordance Project at http://www.anu.edu.au/asian studies/ahcen/proudfoot/MCP/.
References Aoki, Haruo (1986). “Evidentials in Japanese”. In Wallace Chafe & Johanna Nichols (Eds.), Evidentiality: The Linguistic Coding of Epistemology (pp. 223–238). Norwood, NJ: Ablex. Clancy, Patricia (1985). The Acquisition of Japanese. In Dan Slobin (Ed.), The Crosslinguistic Study of Language Acquisition. Volume 1: Data (pp. 373–524). Hillsdale, NJ: Lawrence Erlbaum. Cook, Haruko Minegishi (1990). “An Indexical Account of the Japanese Sentence-Final Particle No”. Discourse Processes, 13, 401–439. Diewald Gabriele, Ulrike Claudi, & Frederike Hunnemeyer (1991). Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Echols, John M. & Hassan Shadily (1989). An Indonesian-English Dictionary (3rd edition). Revised and edited by John U. Wolff & James T. Collins, in cooperation with Hassan Shadily. Ithaca/London: Cornell University Press. Evans, Nicholas & David Wilkins (2000). “In the Mind’s Ear: Semantic Extensions of Perception Verbs in Australian Languages”. Language, 76, 546–592. Gil, David (1999). “The Grammaticalization of Punya in Malay/Indonesian Dialects”. Paper presented at the Ninth Annual Meeting of the South-East Asian Linguistics Society, University of California, Berkeley, USA, May 22. Haspelmath, Martin (1989). “From Purposive to Infinitive – a Universal Path of Grammaticalization”. Folia Linguistica, 10, 287–310. Haspelmath, Martin (1997). Indefinite Pronouns. Oxford & New York: Oxford University Press. Haspelmath, Martin (1999). “Why is Grammaticalization Irreversible?” Linguistics, 37, 1043–1068. Heine, Bernd (1992). “Grammaticalization Chains”. Studies in Language, 16, 335–368. Heine, Bernd (1993). Auxiliaries: Cognitive Forces and Grammaticalization. Oxford & NewYork: Oxford University Press. Heine, Bernd (1997). Cognitive Foundations of Grammar. Oxford & New York: Oxford University Press. Heine, Bernd (2002). “On the Role of Context in Grammaticalization”. In Ilse Wischer & Gabriele Diewald (Eds.), New Reflections on Grammaticalization (=Typological Studies in Language, 49) (pp. 83–101). Amsterdam & Philadelphia: John Benjamins. Hopper, Paul & Elizabeth Closs Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Horie, Kaoru (1998). “On the Polyfunctionality of the Japanese Particle No: From the Perspectives of Ontology and Grammaticalization.” In Toshio Ohori (Ed.), Studies in Japanese Grammaticalization: Cognitive and Discourse Perspectives. Tokyo: Kurosio Publishers.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.31 (167)
From pronominalizer to pragmatic marker
Huang, Chu Ren (1989). Mandarin Chinese NP De: A Comparative Study of Current Grammatical Theories. Taipei: Academica Sinica. Iwasaki, Shoichi. “Cohesion, Non-Challengeability and the -n desu Clause in Japanese Spoken Discourse.” Journal of Asian Culture, 9, 125–142. Janda, Richard D. (2001). “Beyond ‘Pathways’ and ‘Unidirectionality’: On the Discontinuity of Language Transmission and the Counterability of Grammaticalization”. Language Sciences, 23, 265–340. Jiang, Lihong (1977). Dun Huang Bian Wen Zi Yi Tong Shi. Shanghai: Shanghai Gu Ji Chu Ban She. Kitagawa, Chisato & Claudia Ross (1982). Prenominal Modification in Chinese and Japanese. Linguistic Analysis, 9, 19–53. Konishi, Jin’ichi (1969). Kihon Kogo Jiten [Basic Dictionary of Classical Japanese] (revised edition). Tokyo: Taishukan. Le, Van Cu (1988). ‘No’ ni yoru bun umekomi no koozoo to hyoogen no kinoo [Expression and function of structures embedded by no]. Tokyo: Kurosio. Lehmann, Christian (1982). Thoughts on Grammaticalization: A Programmatic Sketch. Volume I. Department of Linguistics, University of Cologne. Li, Charles, Sandra Thompson, & Zhang Bojiang (1998). “The Particle de as an Evidential Marker in Chinese. Zhongguo Yuwen, 263, 93–102. Lu, Shuxiang (1943). “Lun ‘di3’, ‘di4’ zhi bian ji ‘di4’ de youlai” [On the difference between ‘di3’ and ‘di4’ and their origins]. Jinling Qilu Huaxi Daxue ZhongguoWenhua Congkan, 3. Lyons, John (1977). Semantics, Volume 2. Cambridge: Cambridge University Press. Matthews, Stephen & Virginia Yip (1994). Cantonese: A Comprehensive Grammar. London: Routledge. Mutomati Zidaigo Ziten Hensyuuiinkai (Eds.). (2001). Zidaibetu Kokugo Daiziten, Muromati Hen, Volume 4. Tokyo: Sanseidoo. Newmeyer, F. J. (1998). Language Form and Function. Cambridge, MA: MIT. Nihon Kokugo Daiziten Dainihan Hensyuuiinkai (Eds.). (2001). Nihon Kokugo Daiziten, Volume 10. Tokyo: Syoogakkan. Norde, Muriel (2001). “Deflexion as a Counterdirectional Factor in Grammatical Change.” Language Sciences, 23, 231–264. Norde, Muriel (2002). “The Final Stages of Grammaticalization: Affixhood and Beyond”. In Ilse Wischer & Gabriele Diewald (Eds.), New Reflections on Grammaticalization (pp. 45–65). Amsterdam & Philadelphia: John Benjamins. Rosch, Eleanor & Caroline Mervis (1975). “Family Resemblances: Studies in the Internal Structure of Semantic Categories.” Cognitive Psychology, 7, 573–605. Shi, Yuzhi & Charles N. Li (2002). “The Establishment of the Classifier System and the Grammaticalization of the Morphosyntactic Particle de in Chinese”. Language Sciences, 24, 1–15. Simpson, Andrew & Xiu-Zhi Zoe Wu (2001). “The Grammaticalization of Formal Nouns and Nominalizers in Chinese, Japanese and Korean”. In T. E. McAuley (Ed.), Language Change in East Asia (pp. 250–283). Richmond, Surrey: Curzon. Sheikh Othman bin Sheikh Salim (Chief Editor). (1989). Kamus Dewan, Edisi Baru. Kuala Lumpur: Dewan Bahasa dan Pustaka.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5907.tex / p.32 (168)
Foong Ha Yap, Stephen Matthews, and Kaoru Horie
Soga, Matsuo & Fujimura Taiji (1978). “A Note on Noun Modification in Japanese”. Journal of the Association of Teachers of Japanese, 13, 41–51. Traugott, Elizabeth Closs & Bernd Heine (1991). Approaches to Grammaticalization, Volumes 1 & 2. Amsterdam & Philadelphia: Benjamins. van der Auwera, Johan (2002). “More Thoughts on Degrammaticalization”. In Ilse Wischer & Gabriele Diewald (Eds.), New Reflections on Grammaticalization (pp. 19–43). Amsterdam & Philadelphia: John Benjamins. Wittgenstein, Ludwig (1953). Philosophical Investigations. New York: Macmillan. Yap, Foong Ha, Choi Pik Ling, & Cheung Kam Siu (2004). “Mapping Chinese morphemes across millenia (I): A diachronic study of di and its cognates.” Paper presented at the 20th Scandinavian Conference of Linguistics, University of Helsinki, January 7–9.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.1 (169)
Conditionals and subjectification Implications for a theory of semantic change* Jacqueline Visconti University of Birmingham
Modern Anglo-American linguistics, logic, and philosophy of language has been dominated by the intellectualist prejudice that language is, essentially, if not solely, an instrument for the expression of propositional thought. (Lyons 1982: 103)
.
Introduction
The existence of cross-linguistic directionalities, ‘paths’, of semantic change, is one of the most challenging proposals recently advanced in historical linguistics. In this paper I focus on one of these tendencies: subjectification, “the development of a grammatically identifiable expression of the Speaker’s belief or Speaker’s attitude towards what is said” (Traugott 1995: 32). I will show that a full understanding of the semantic and pragmatic properties of Present Day English conditional supposing and its equivalents in Romance languages (e.g. Italian supponendo che) can be gained by recognizing their evolution as an instance of this phenomenon. The paper is centered around one focal point: the importance of a fine-grained analysis of the non-propositional component of language as a key for capturing generalizations that would otherwise remain unnoticed. As the quote above states: “Modern Anglo-American linguistics, logic, and philosophy of language has been dominated by the intellectualist prejudice that language is, essentially, if not solely, an instrument for the expression of propositional thought” (Lyons 1982: 103). Both subjectivity, intended as “the way in which natural languages, in their structure and normal manner of operation, provide for the locutionary agent’s expression of himself and his own attitudes and beliefs” (Lyons 1982: 102), and subjectification, seen as “the semasiological process whereby Speakers/Writers come over time
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.2 (170)
Jacqueline Visconti
to develop meanings for lexemes that encode or externalize their perspectives and attitudes as constrained by the communicative world of the speech event, rather than by the so-called ‘real-world’ characteristics of the event or situation referred to” (Traugott & Dasher 2002: 30), are here considered fundamental heuristic dimensions. Subjectification, in particular, is viewed as an intrinsically unidirectional phenomenon: a trend towards a higher degree of encoding of the Speaker’s point of view, not vice versa. As we shall see, the whole history of English supposing and its equivalents in Romance languages is an argument in favour of the hypothesis of the unidirectionality of this type of change: from their origins in Latin subpono “place under”, with a shift from concrete to abstract (see Visconti 2002), to the later reanalysis of the verb as a conjunction having the function of conveying the Speaker’s epistemic attitude towards the proposition (i.e. its evaluation as ‘hypothetical’). The paper is organized into four sections: Firstly, I shall sketch the evolution of suppose from Middle to Early Modern English1 (Section 2), which shows an increasing degree of subjectification, with a concomitant shift from a high to a low degree of the Speaker’s commitment towards the truth of the proposition; a parallel phenomenon is documented in both Italian and French. Secondly, I will examine the categorial reanalysis of English supposing / Italian supponendo che from verb to conjunction (Section 3), focusing on the role of related free adjuncts in the process. The shift from lexical to functional category will be shown to be related to the semantic change highlighted in Section 2. Finally, I will address the hypothesis of the existence of a deontic polysemy in suppose (as in ‘He wasn’t supposed to be there’), which would appear to counter the unidirectional tendency of meanings to shift from deontic to epistemic and not vice-versa (Section 4). The data suggest instead that the primary function of this construction in Present Day English is epistemic and arises later than the deontic value, both as inferences from a combination of morphosyntactic and semantic factors. The results allow us to make important predictions on the evolution of other complex conditional connectives in Romance and in the Germanic languages (e.g. Italian ammesso che or German angenommen, daß), thus providing an insight into: (i) the origins of conditionals across languages; (ii) the relationship between semantic change and grammaticalization; (iii) the notion of unidirectionality in language change (Section 5).
. From Middle to Early Modern English: The data In this section, I shall focus on the evolution of suppose from Middle to Early Modern English. The corpora used are the diachronic component of the Helsinki Corpus (–1710), the online Middle English Dictionary, the Lampeter Corpus (1640–
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.3 (171)
Conditionals and subjectification
1740), the Corpus of Early English Correspondence (1417–1681), the Chadwyck Literature Online.2 The problem of the choice and quality of the written data is crucial to diachronic research.3 In this case, given the existing databases, I was able to retrieve a much larger corpus of examples from Middle English than from Early Modern English. Confronted with the question of either discarding some of the Middle English data or including them with a discrepancy between the data from the two periods, I opted for the second choice, thus considering 280 Middle English examples as against 67 Early Modern English examples. The choice has no consequences for the argumentation, as the trends identified in this and the following section would have been further strengthened by consideration of a smaller database of Middle English examples.
. Semantic and structural trends ME suppose, like OF supposer and OItal supporre (see Visconti 2002), shows a range of uses, most of which are in the realm of modality (for the latter, see Kiefer 1994: 2515). The Middle English Dictionary lists for suppose the following senses, of which only (c) survives in Present Day English (see OED): a.
‘believe as a fact’: (1) We shuld trow, and suppose ay þat alle er save. . . þat we se here gude werkes wirk, And has þe sacramentes of halikyrk. (Hampole, Pr. Consc. 3776, 1340 [OED]) “We should believe and assume always that all are saved. . . that we see here good works work, and have the sacraments of the holy Church.”
b
‘anticipate, expect’ (this is the first attested occurrence of the lexeme): (2) Whan Seynt Ihon herde þat seye, þat Troyle supposed for to deye. (R. Brunne, Handl. Synne 6970, 1303 [OED]) “When St. John heard that said, that Troyle expected to die.”
c.
‘take for granted, assume (without reference to truth or falsehood)’, ‘posit, conjecture, hypothesize, guess’: (3) Who wolde leeue, or who wolde suppose The wo that in myn herte was? (Chaucer, Wife’s Prol. 786, 1386 [OED]) “Who would believe or who would imagine the woe that was in my heart?”
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.4 (172)
Jacqueline Visconti
d. ‘suspect, allege, feign, speak deceptively’: (4) Certein Commune of Pasture. . . whiche the said Lord. . . claymes . . . as it is supposed by the same Bille. (RParl. 3.650a, 1411 [MED]) “Certain Commune of pasture. . . which the said Lord. . . claims. . . as it is alleged by the same Bill.”
e.
‘support, place support under’: (5) Twene ioy and woo my gost supposid is, As this to thynke and this oft to advise, My witt as now so renneth this and this. (C.d’Orl., Poems 3/64, c1450 [MED]) “Between joy and woe my soul is suspended. . . ”
As I shall illustrate in this and the following section, four phenomena characterize the evolution of the lexeme from Middle to Early Modern English: I and II are of a semantic nature, III and IV are morphosyntactic: I.
A change in modal value: from ‘believe’ (6) to ‘frame as a hypothesis’ (7), i.e. a shift from a high (6) to a low (7) degree of commitment towards the truth of the proposition. II. A change in the source of the evaluation/commitment: from the subject of the verb in the matrix clause, as in (6), to the Speaker/Writer, as in (7), i.e. a process of subjectification, as defined in the Introduction. (6) The disciples, supposing that he had ben a fantasme, criden for drede. (Love Mirror, Brsn e.9 144, c1430 (a1410) [MED]) “The disciples, believing he was a ghost, cried out in dread.” (7) Or supposing that there is but one Bishop; when he is sent into perpetual Banishment, how must his Office be supply’d? (Humphry Hody, A letter. . . Oxford, 1692 [Lampeter])
III. An increase in the frequency of suppose in imperative contexts, such as (8), from Middle English 2.5% to Early Modern English 9%:4 (8) TOM: Suppose I should marry thee. IONE: Indeed Sir, I’de be the lovingest wife that ever was made of flesh and blood. (Samuel Pepys’ Penny Merriments, 1684–1685 [Helsinki])
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.5 (173)
Conditionals and subjectification
IV. An even greater increase in frequency with interrogative contexts, from Middle English 1.5% to Early Modern English 15% – comprising both interrogative clauses such as (9) and combinations of an interrogative clause with imperative suppose (10) or with supposing:5 (9) “Whan that wiche unmixt and by nature vnparted is, that humaine error partz, and from the true and right to falz and wanting brings. Dost thou suppose that nothing he wantes that powre needes? I think not so.” (Queen Elizabeth, Boethius etc., 1593 [Helsinki]) (10) “I will put you this case: Suppose the parliament wold make a lawe that god shold not be god. Wold you then, master Riche, say that god were not god?” (William Roper, The Lyfe of Sir Thomas Moore, 1556 [Helsinki]) The evidence showing development (I) is further outlined in this section, (II) will be further discussed in Section 3.2, (III) in Section 2.2, and (IV) in Section 3.1. In Middle English suppose could be paraphrased by believe in 57% of the examples, as in (6) above or (11) below: (11) “Lamech slew Cayn with ane arowe, supposing he had bene a wylde beste.” (Maundev. (Roxb.) xiii. 57, c. 1400 [OED]) “Lamech slew Cayn with an arrow, believing he was a wild beast.” Furthermore, suppose was often found coordinated with such verbs as trow, trust and leeue, meaning ‘to believe’: (12) We shuld trow and suppose ay þat alle er save. (PConsc. (Glb E.9 & Hrl 4196) 3776, a1425 (a1400) [MED]) “We should believe and assume always that all are saved. . . ” (13) Whan tydyngys cam to hys modyr fro ouyr þe see þat hir sone had weddyd, sche was ryth glad & thankyd God with al hir hert, supposing and trusting he xulde leuyn clene & chast as þe lawe of matrimony askith. (The Book of Margery Kempe, 1420–1500 [Helsinki]) “When the news came to his mother from over the sea that her son had wedded, she was right glad and thanked God with all her heart, believing and trusting he should live clean and chaste as the law of matrimony requires.” Similarly, suppose in this period still co-occurs with epistemic adverbs such as veryly, certen, sykerly “with certainty”: (14) So fair a gardyn woot I nowher noon For out of doute I verraily suppose That he that wroot the Romance of the Rose Ne koude of it the beautee wel deuyse. (Chaucer CT.Mch. (Manly-Rickert) E.2031 (c1395) [MED])
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.6 (174)
Jacqueline Visconti
“So fair a garden I never knew anywhere for out of doubt I truly believe that he that wrote the Romance of the Rose. . . ” (15) We might say this for certen & suppose it in hert. (Destr.Troy (Htrn 388) 2317, c1540 (a1400) [MED]) “We might say this for certain and believe it in heart.” (16) Troyens supposid sykerly þat Philomene. . . Had be ded. (Lydg. TB (Aug A.4) 2.8309: c1425 (a1420) [MED]) “The Troyens assumed with certainty that Philomene had died.”6 Evidence for the gradual shift in modality from Early Modern English onwards can be found in the following facts. First of all, the paraphrasis with believe is limited to 27% of the Early Modern English examples. Secondly, suppose is no longer harmonic with ‘trust’, or ‘believe’, but with ‘think’, ‘conclude’, ‘imagine’: (17) “For if we conclude the truth, of wickednes misfortune, infinite must we suppose that misery that is everlasting.” (Queen Elizabeth, Boethius etc., 1593 [Helsinki]) Thirdly, suppose no longer occurs with epistemic adverbs expressing a high degree of certainty; instead, it is used in contexts showing a low degree of commitment, as in (18) and (19): (18) “Whan I consider thy reasons,” said I, “I can suppose nothing more true. But if I turne me to mans Judgement, who is he, to whom not only these thinges will not seeme to be beleeuid but scar[c]ely to be heard?” (Queen Elizabeth, Boethius etc., 1593 [Helsinki]) (19) “I may truly denye some part of this Confession; but bycause there is nothing material greatly, I suppose the whole to be true, and what is herein deposed, sufficiente to bring me within the compass of the Enditement.” (The Trial of Sir Nicholas Throckmorton, 1554 [Helsinki]) A parallel phenomenon is identifiable in both Italian and French. The combination of supporre “suppose” and credere “believe” is unnatural in Present Day Italian (see Visconti 2000: 117): (20) Supponendo che tu venga – ??e io ci credo – . . . “Supposing you were to come – and I believe it – . . . ” Similarly, the co-occurrence of croire “believe” and supposer “suppose” is attested in Old French: (21) Aussi, pour estre miex amee, Doiz tu tant fere a la fïee, Que ton ami cree et suppose Que tu l’aimes sus toute chose.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.7 (175)
Conditionals and subjectification
(Clef d’Amors 3147 [Tobler-Lommatzsch]) “. . . That your friend believes and supposes that you love him above all.” (22) Mauvaisement on puet sçavoir conment chil se conbatirent, qui la furent mort , tels que le conte Carle d’ Alençon, le conte Lois de Blois [. . . ] et pluisseurs aultres; mais on doit croire et supposer que si grans signeurs que chil estoient, ne furent pas mors ne ocis a petit de fait. (Froissart, Chroniques, I.14.734, ?1338 ?1410 [BFM])7 “. . . but we have to believe and suppose that such great Lords did not die without honour.” while the subsequent shift is documented in historical dictionaries: Le verbe a eu le sens de ‘poser comme un fait établi’ (13e s.), courant en français classique, mais éliminé par une autre acception: ‘admettre comme probable ou plausible, sans pouvoir affirmer de façon positive’ (v. 1270), et ‘poser à titre d’hypothèse, n’impliquant aucun jugement mais servant de point de départ à un raisonnement, à une discussion’. (Rey 1994: 2051) (“The verb had the sense of ‘posit as a fact’ (13th cent.), current in classical French, but eliminated by another sense: ‘admit as probable or plausible, without being able to affirm in a positive way’ (approx. 1270) and ‘posit as a hypothesis, without any judgement, as a starting point for a reasoning, a discussion” [my translation]).
. Preliminary hypothesis The explanation I propose for the findings described in Section 2.1 is as follows. As an absolute increase in frequency of interrogative and imperative is not documented in the history of English (see e.g. Denison 1998: 245–254; Fischer 1992: 278–280; Rissanen 1999: 273–280; Visser 1963: 15–19), the increase of suppose in such contexts must be indicative of a change in the use of suppose. This change is towards an interactive I/You mode, with an increase of a “dialoguetype of discourse” (König 1992: 432). As shown by the textual evidence from this period, Speakers/Writers start to use suppose as a means of introducing a frame of reasoning, an instruction to the Addressee/Reader to add a proposition p to her/his cognitive background, in order to process a question (mostly) or a statement (see Section 5). The recruitment of suppose for argumentative uses, reflected in its increase in imperative and interrogative contexts, would explain both the gradual move from “character-subjectivity” to “speaker/hearer-subjectivity” (Verhagen 2000: 201) and the shift in the evaluation expressed (from ‘believe’ to ‘posit as a hypothesis’), due to the frequent association of the lexeme with contexts such as the imperative, creating possible worlds, and the interrogative, in which the de-
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.8 (176)
Jacqueline Visconti
gree of commitment is ‘suspended’. This general shift in the use of the lexeme and its change into a conjunction will be illustrated in Section 3.1, while Section 3.2 will focus on the change in subjectivity.
. From verb to conjunction . Categorial reanalysis In this section I shall focus on the reanalysis of English supposing and Italian supponendo che from verb to conjunction. According to the data, the process commences in Early Modern English. I shall first highlight the main features of the shift, then focus on the CEEC and Lampeter corpora in order to identify the contexts that favour the reanalysis. In both the MED and the Middle English section of the Helsinki corpus all Ving forms are ‘related free adjuncts’, that is participial clauses such as (23), in which the empty subject is controlled by the subject of the matrix clause (see Kortmann 1991: 46–90). The ‘controller’ is in almost all cases a 3rd person (see below for some statistics): (23) Vlixes. . . Made his sone. . . to be. . . shette vp in presoun. . . supposyng . . . Fro alle meschef þer-by to go quyte. (a1420 Lydg. TB 5.3063 [MED]) “Ulixes. . . made his son. . . to be. . . shut up in prison. . . believing. . . from all mischief thereby to go quit.” In Early Modern English we still find many related free-adjuncts.8 An interesting difference is the increase of 1st person control versus 2nd–3rd person: in Middle English the related free adjuncts introduced by supposing/-yng are controlled by a 3rd person subject in 86% of the cases, as in (23) above, and by a 1st person in 14% of cases. In Early Modern English the controller is a 1st person in 61% of the cases, as in (24) below, a 2nd person in 5.5% and a 3rd person in 33.5% of all cases. The relevance of this factor to the subjectification process undergone by the lexeme will be discussed in Section 3.2. (24) And although that called a Sermon be swoln hereby above its ordinary bulk, it is notwithstanding still short of what may be necessary for the clear opening of this great Point; yet do I stop there notwithstanding at present, supposing there may be an opportunity after offered for defence of that delivered by that opposition. (Henry Jones, A sermon of Antichrist, preached at Christ-Church, Dublin [1676] 1679 [Lampeter])
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.9 (177)
Conditionals and subjectification
Reanalysis is favoured by contexts such as (25), where the empty subject of the supposing-clause could be controlled by the subject of the matrix clause (‘we’), yet the subsequent interrogative clause suggests a different reading, in which the supposing-clause is not related to the preceding clause and introduces a frame of reasoning for the interrogative: (25) Now, Lady Iffida, you are to determine this Spanish bargaine or if you please, we wil make it an English controuersie: supposing you to be the Ladye and three such Gentlemen to come vnto you in a wooing, In faith who should be the speeder. (John Lyly, Euphues and his England, 1580 [Chadwyck Literature Online]) Early Modern English data of this kind show that the reanalysis is favoured by contexts containing related free adjuncts that allow both an increase in scope and a ‘loosening’ of the control between the subject of the matrix clause and the subject of the adjunct.9 The increase in scope, defined in terms of level of adjunction of the PP headed by supposing (see Visconti 2000: 56–66),10 is from VP to IP. These contexts are interrogative, such as (25) above or (26) below, and preposed adjuncts, such as (27) or (28) below: (26) Here a Second Query will follow. Will our Adversaries grant, that supposing a Bishop should conspire against the Government or Rebel, the Government has Authority to imprison him, or to banish him; but not to Deprive him of his Bishoprick, so as that another may be plac’d in his See? [. . . ] Or supposing that there is but one Bishop; when he is sent into perpetual Banishment, how must his Office be supply’d? (Humphry Hody, A letter. . . Oxford, 1692 [Lampeter]) Crucial to the transition from verb to conjunction are cases in which the control is still present, but where the preposed adjunct opens up the possibility to reanalyse supposing as a connective. Thus, in (27), supposing thou doest suggests the reading ‘if thou doest’, with IP scope: (27) Forgive me, my chyld, I cannot forbear to say this; my hart is so topfull of desyre to have thee as happy as my self. But I wod not have thee take it for persuasion, unless thou finde, as I suspect, no solid content wher thou art; but supposing thou doest, I am truly satisfied, and will beleeve, and hope, God will make thee a sayntt, wher thou art, and thats all I car for. (Winefrid Thimelby to Gertrude Aston, 23 Feb. 1672 [CEEC]) As suggested in Section 2.2 (see Section 5 for discussion), the mechanism triggering the reanalysis is grounded in use: Speakers/Writers employ suppose for an argumentative function, in a scientific or argumentative discourse type:
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.10 (178)
Jacqueline Visconti
(28) A Circle drawn equidistantly from these describeth the Equator. This don, they take a smal Steel wyer [. . . ] and applie it to anie part of the Equator, and it will precisely turn towards the North and South Poles [. . . ] But supposing a Concavitie to bee let into this Little Earth, in anie part, either about the Equator, or betwixt it and the Poles: In that case the Needle will not point directly to the Poles, but will make a Variation. (John Gregorie, The description and use of the terrestrial globe, London, 1650 [1649] [Lampeter]) Notice how supposing in Middle and Early Modern English, unlike Present Day English, co-occurs with the subjunctive mood (29) and modal verbs (30), which contribute to providing the epistemic modalization of the proposition as ‘hypothetical’.11 This modalization is at a later stage ‘absorbed’ and conveyed by the lexeme alone: (29) I, for my part, shall say nothing of it; but leave it for others to judge. But this I must add, that the Words which were alleg’d by that Gentleman to be the Transcriber’s own, supposing they were the Transcriber’s, (as we know very well they are not) do contain nothing in ‘em, that could prove the Transcriber to have been the same Man with our Author. (Humphry Hody, A letter. . . Oxford, 1692 [Lampeter]) (30) Supposing the French should offer to join with the Spaniards, and assist them to drive the Scots from Darien, as some say they have already proffered; we are not to imagine that the Spaniards will accept their Proffers in this case, when they refused them as to the driving the Moors from before Ceuta. The Reasons are obvious [. . . ]. (A. Foyer, A. Fletcher (attr.), A defense of the Scots settlement at Darien, Edinburgh, 1699 [Lampeter]) The same generalizations defined above for the English data apply to Italian supponendo che.12 Again we see the importance of preposed adjuncts and interrogative contexts in favouring the reanalysis: (31) Supponendo tuttavia contro il vero, che il Marchetti sia stato un perfettissimo versiscioltaio in quella sua traduzione, è egli una cosa da farsene le mille croci? (Baretti, La frusta letteraria, XVIII sec. [LIZ]) “Supposing however against the truth that Marchetti was a perfect blank versifier in that translation of his, is this such a terrible thing?” Secondly, the use of supponendo che in scientific and argumentative discourse functions again as a catalyst towards recategorization:
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.11 (179)
Conditionals and subjectification
(32) Ora, perché il proceder per interrogazioni mi piace assai, già che non ci é l’autore stesso, rispondami il signor Simplicio, alle domande ch’io farò, quel ch’e’ arederà che fusse per rispondere esso. E supponendo di trattar della già detta stella del 72, apparsa in Cassiopea, ditemi, signor Simplicio, se voi credete che ella potesse esser nell’istesso tempo collocata in diversi luoghi, cioè esser tra gli elementi, ed anco tra gli orbi de’ pianeti, ed anco sopra questi e tra le stelle fisse, ed anco infinitamente più alta. (Galilei, Dial. sopra i due massimi sistemi, XVII sec. [LIZ]) “And supposing to talk about the above mentioned star . . . tell me, Sir Simplicio, whether you believe that. . . ”
. Subjectification The shift from lexical to functional category illustrated in Section 3.1 appears to be strictly linked to the semantic change highlighted in Section 2.1: supposing gradually comes to be used almost exclusively to express the Speaker’s attitude towards the proposition, i.e. its evaluation as hypothetical. The increase in 1st person control from Middle English to Early Modern English, noticed in the previous section (see (24) and (27) above), is crucial to the transition, which can be analysed in three steps. In the first stage supposing is a 3rd person control related free-adjunct. The source of the evaluation is the subject of the matrix clause (‘character-subjectivity’), as in: (33) Vlixes. . . Made his sone. . . to be . . . shette vp in presoun. . . supposyng . . . Fro alle meschef þer-by to go quyte. (a1420 Lydg. TB 5.3063 [MED]) “Ulixes. . . made his son. . . to be. . . shut up in prison. . . believing. . . from all mischief thereby to go quit.” The second stage is still verbal, in the form of a 1st person control related freeadjunct. The source of the evaluation is still the subject of the matrix clause, but this is normally the Speaker-Writer, as in: (34) [. . . ] but supposing thou doest, I am truly satisfied, and will beleeve, and hope, God will make thee a sayntt, wher thou art, and thats all I car for. (Winefrid Thimelby to Gertrude Aston, 23 Feb. 1672 [CEEC]) In the third stage we see the shift towards the category of conjunctions, with loss of control relationship. The source of the evaluation is the Speaker-Writer (‘speaker/hearer-subjectivity’) and not the subject of the matrix clause: (35) But supposing (as indeed there’s no great likelihood of it) that no such Alliance as this should ever happen; yet however, if these two Nations be
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.12 (180)
Jacqueline Visconti
not more closely united, it may be of ill consequence to England, if any of their Kings at any time should be so far disgusted with their Proceedings, as to leave them, and betake themselves to us. (A. Foyer, A. Fletcher (attr.), A defense of the Scots settlement at Darien, Edinburgh, 1699 [Lampeter]) Both the grammaticalization into a conjunction and the encoding of the Speaker’s epistemic attitude are at a very advanced stage in Present Day English, as shown by a search through the Bank of English corpus for co-occurrences of 3rd person subjects with supposing.13 Out of 43 examples, 1 verb, a related free adjunct (36), and 1 ambiguous case (37) were found (less than 5%), against 41 cases of supposing as a conjunction, such as (38) to (40), with the Speaker as default source of the evaluation (of the proposition as ‘hypothetical’; see Visconti 2000: 186–191): (36) He did not go into it, supposing me, I think, to be au courant, but I believe he had made some very strong remarks from the bench. (P. O’Brian, Testimonies, HarperCollins (Marshall Pickering), 1994 [Bank of English: British Books]) (37) She’d no pretensions about her own background. But she did think about Will and his mother, supposing that Will would ever ask her to be mistress of the house, and his mother would still be alive. (M. Kide, Polymena Cove, HarperCollins, 1994 [Bank of English: British Books]) (38) “Supposing”, he resumed abruptly, “supposing it was a Mufti officer, there’s no way McGuire and Connolly would risk saying so.” (N. Packard, No Body, HarperCollins (Crime Club), 1987 [Bank of English: British Books]) (39) then he decided to speak. “I am still not sure this is a good idea. The computer report on Newman records he is shrewd. Supposing that he does find out his wife was executed here? I do not see how he could – but just supposing that did happen?” (C. Forbes, Cover Story, HarperCollins, 2000 [Bank of English: British Books]) (40) Supposing on that occasion he had done the identical things he had done this time. Supposing he had even been stung by a scorpion before. Supposing he had even met Murray-Roberts and Dr. Haydar before. (B. Alsiss, Somewhere East of Life, HarperCollins (Flamingo), 1994 [Bank of English: British Books])
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.13 (181)
Conditionals and subjectification
An interesting cross-linguistic finding is that a higher degree of grammaticalization and subjectification is found in English as compared to Italian, where supponendo che is still productive as a verb and the epistemic modalization is frequently not attributed to the Speaker, but to a ‘character’, as in (41): (41) Supponendo che il tutto gli fosse stato raccontato per chiedergli un consiglio, Macario gli disse di pregare la signora Carolina che aiutasse la signorina Francesca in quanto poteva. (Svevo, Una vita, 1892 [LIZ]) “Assuming everything had been told to him to get an advice, Macario said ...” A further sign of this cross-linguistic difference is that Italian, in contrast to English, does not manifest the loss of the complementizer che/di “that/of ”. This loss, as argued by Kortmann and König (1992: 685), indicates a further stage of grammaticalization: One important criterion for judging [the] degree of reanalysis [of a conjunction] will have to be the obligatory versus optional presence of a complementizer as final element. In languages like English or German, where the complementizer need not form part of conjunctions, it is a safe indication of an approximation to the core area of conjunctions if the complementizer has either become optional or been lost completely (loss of that observable especially in Early Modern English).
These facts place English supposing and Italian supponendo che at different stages on the same grammatical and semantic clines.14
. A deontic polysemy? In this section I shall address the hypothesis of the existence of a deontic polysemy in suppose (as in ‘He wasn’t supposed to be there’), which would appear to counter the unidirectional tendency of meanings to shift from deontic to epistemic (for the latter see Bybee & Pagliuca 1987; Traugott 1989 and Sweetser 1990). The question is addressed by Ziegeler (2002) in her study of the evolution of semi-modals such as be supposed to. She demonstrates that the early uses of the lexeme are better classified as ‘evidentials’ rather than as ‘epistemic’, as suggested by Chafe (1986), and examines the contexts that favour the development of the obligation meaning out of the evidential stage.15 Integrating her analysis with the analysis of Modern to Present Day English data, I shall argue that be supposed to is not a counterargument to the unidirectionality hypothesis, as its evolution does follow the path from deontic to epistemic which it has been thought to challenge.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.14 (182)
Jacqueline Visconti
Ziegeler (2002) reviews various proposals on the epistemic-deontic pathway for be supposed to. She examines the influence of the aspectual nature of the complement verb on the modality of the matrix verb and concludes that “The stages prior to the development of obligation cannot be said to be epistemic modal stages, as [. . . ] epistemic modal verbs cannot be converted into questions [without a change of meaning]” (Ziegeler 2002: 16). As shown by her data, the obligation meaning arises out of the evidential stage in environments with “a generic, human subject of non-specific reference in combination with a complement which is nonstative, and an appropriate semantic context in which a habitual act is understood as one of duty or custom ” (p. 16). In my view, the crucial factor in the evolution of be supposed to is the nature of the passive construction, in which a non-expressed subject creates a non-specified source of belief/expectation. It is the passive that is responsible for the evidential nature of the construction, in examples such as (42) and (43) below, as evidentials report on someone else’s belief, as in hearsay, or “reported evidence” (Kiefer 1994: 2517–2518) (‘quotative evidentials’ in Ziegeler 2002: 16): (42) Day is supposede to drawe nye at the firste crawenge of the cocke. (Higd.(2) (Hrl 2261) 1.339, a1475 (a1425) [MED]) “Day is believed to come near at the first crowing of the cock.” (43) It is supposed þat some off hem haue goton an c li. worth lande. (Fortescue Gov.E. (LdMisc 593) 136, a1475 [MED]) “It is believed that some of them have got land worth c li.” Furthermore, the passive is responsible for both the deontic inference and the epistemic inference that characterizes Present Day English uses of the construction. Let me develop this proposal more fully. To begin with, I shall address the question of the relationship between the two senses of ‘believe’ and ‘expect’ observed in suppose. According to Traugott (1989) “there is evidence that there were two coexistent paths of development from the two senses ‘expect’ and ‘hypothesize’ that were borrowed with the form suppose from Middle French [. . . ] In the meaning ‘expect’ suppose ultimately developed a stronger deontic of obligation, as in: ‘Girls are not supposed to be soldiers, are they, Margaret?’ (R. Connor, Glengarry Days iii43, 1902 [OED]). The other path, ‘believe, imagine, hypothesize’, was epistemic, as in: ‘We should trow, and suppose’ (Hampole, Pr. Consc. 3776, 1340), i.e. ‘trust and believe’ [. . . ] and it has continued to be so until the present time. There is no need to suppose that it was the origin of the obligative supposed to” (Traugott 1989: 45–46, Note 11). According to my analysis, focusing on the active form of the verb, ‘believe’/‘expect’ is not a polysemy for suppose; rather, the ‘expect’ meaning arises from the ‘believe’ meaning when the event time of the proposition is future (later
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.15 (183)
Conditionals and subjectification
than. . . ); this is reflected in the OED definition (n. 4 (obsolete): “referring to the future: ‘to expect”’) and can be seen in these early occurrences of the ‘expect’ sense: (44) At yis next ambassiat..which may be supposed shal come to none avail or conclusion. (RParl.4.493a., 1435 [MED]) “At his next ambassade. . . which may be expected shall come to no advantage or conclusion.” (45) Thentente of the makers was to ordeyne mesure and proportion the peyne contiened therein, lyke to the offence and trespasse, and may not resonably be supposed that the prelates of the Church. . . wold have involved hemself in every caas that might be comprehended in the said bitter interpretation. (Doc.N.Convocation in Sur.Soc.113.179, 1447 [MED]) “The purpose of the makers was to devise measure and proportion the punishment therein contained like to the offence and trespass, and it may not reasonably be expected that the prelates of the Church would have involved themselves in every case that might be comprehended in the said bitter interpretation.” The hypothesis is borne out by the French data, where the ‘expect’ sense is not a polysemy, but arises out of the combination of supposer meaning “believe” with a future-referenced event time in the proposition: (46) Et en ce pais de Hainnau, ce sont gens, chevaliers et esquiers, de grant adrece et qui demandent les armes. Nous la venu, je suppose que nous serons adrechié, conforté et consillié de tout ce qu’il nous besongne. (Froissart, Chroniques, 14.56, ?1338 ?1410 [BFM]) “. . . Once arrived there, I believe we shall be (expect we shall be) welcomed. . . ” (47) Qant li Espagnol orent fait lor emploite et lor marceandise, et il orent cargiet lors vassiaus de draps et de toilles et de tout ce que bon et pourfitable lor sambloit pour retourner en lor pais, bien supposoient que il seroient rencontré des Englois, mais de tout ce il ne faisoient point grant compte, puisque il estoient pourveu d’ artelerie et de chanons. (Froissart, Chroniques, 14.882, ?1338 ?1410 [BFM]) “. . . [the Spaniards]. . . believed (expected) that they would be met by the English. . . ” On the basis of these findings, I will argue that the meanings of expectation or intention imposed on the subject of be supposed to arise via invited inferences, given the combination of the following two factors: the use of the passive construction, which creates a non-expressed subject, hence a non-specified source of expecta-
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.16 (184)
Jacqueline Visconti
tion; and the futurity of the suppose ‘expect’ construction. As deontic modality typically proceeds from a source or cause and prototypically involves an event that is projected or obligated to occur later than the reference time of the utterance (in contrast to prototypical epistemics, which are associated with present reference time), an invited inference of obligation arises associated with ‘participantexternal’ uses, invoking a general cognitive or social source of expectation (van der Auwera 1999: 54).16 Notice the ambiguity of (48) between a future-referenced ‘expected’ and a present-referenced ‘believed’ reading: (48) We have passed a mighty stupid day; Sir George is civil, attentive, and dull; Emily pensive, thoughtful, and silent; and my little self as peevish as an old maid: nobody comes near us, not even your brother, because we are supposed to be settling preliminaries; for you must know Sir George has graciously condescended to change his mind, and will marry her, if she pleases, without waiting for his mother’s letter. (Brooke, Frances Moore: Emily Montague 1769 [Chadwyck Literature Online]) Two questions arise relating to this hypothesis: (i) When did the deontic inference arise? (ii) What is the degree of conventionalization of such an inference in Present Day English? In answer to (i), the earliest instances of this inference can be found in Early Modern English, as in: (49) [. . . ] therefore it may be supposed that no ship puts into Ireland but upon some extrordinary necessity, or in expectacion of more then ordinary advantages. (Letter to Lord Clifford, 1640–1710 [Helsinki]) However, meanings of expectation or intention in the 19th century amount to only around 22% according to Ziegeler (2002), while the onset of modality can be safely placed in the 20th century (see Ziegeler 2002: 8–13 for statistics). In answer to (ii), the deontic inference appears to be still defeasible in the case of be supposed to, in contrast to must: (50) Students are supposed to see their tutors at the start of term, but no one ever does. (51) ??Students must see their tutors at the start of term, but no one ever does. Let me now come to the third point of my argument: the existence of an epistemic inference in PDE be supposed to. The hypothesis arises out of the observation of the overwhelming presence of counterfactual signals in the context of use of be supposed to from Modern to Present Day English, such as but, in fact, in reality. . . :
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.17 (185)
Conditionals and subjectification
(52) In her endeavour to follow him, Aliva met with a severe fall, and was conveyed away, in a state of insensibility, by Sir Cecil. She was supposed to be lifeless; but she survived the accident. (J. Sheppard. A Romance. By W. H. Ainsworth, 1839–1840 [Chadwyck Literature Online]) (53) This style of becoming dress she was supposed to wear in memory of Mr. Kirkpatrick; in reality because it was both lady-like and economical. (E. Gaskell, Cleghorn, Wives and Daughters, 1866 [Chadwyck Literature Online]) These contexts, in which be supposed to evokes a possible world, a state of affairs which would be expected to occur but does not, represent the most frequent ones in which the construction is used in Present Day English: (54) He arched his thick eyebrows and leered at me. “Show me ze vay to ze morgue, my pretty”. Was that supposed to be Bela Lugosi? It sounded more like Maurice Chevalier. (N. Packard, No Body, HarperCollins (Crime Club), 1987 [Bank of English: British Books]) (55) It was supposed to be a dream holiday. But the tiny boat she was sailing in sank miles from land in shark-infested sea. ([Bank of English: British Magazines]) (56) With the Russians north of the 38th parallel and the Americans to the south, the partition was supposed to be temporary. The stark fact was that the Korean people north and south were not to be given the complete democratic. . . . (J. A. S. Grenville, The Collins History of the World in the 20th Century, HarperCollins, 1994 [Bank of English: British Books]) These data show that the primary function of this construction in Present Day English is to mark the Speaker’s non-identification with the source of the evaluation, hence the Speaker’s non-commitment towards the realization of the proposition, i.e. an epistemic function. The inference arises as follows: if the Speaker/Writer chooses to use the be supposed to construction (s)he evokes an unspecified source of belief/expectation, which is distinct from her/himself. This choice invites the inference that the Speaker does not identify with the source of evaluation, and also that the Speaker signals a distance between the expected world and the ‘real’ world: this counterfactual component is often co-expressed by linguistic marks such as but. . . , in reality. . . , the fact is. . . and is confirmed by the overwhelming use of the past form of the construction in Present Day English. A further argument in support of the epistemic hypothesis is the interrogability test quoted by Ziegeler (2002: 16)
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.18 (186)
Jacqueline Visconti
(see above): the convertion into questions of the Present Day English data above leads to pragmatic or textual oddity: ‘was it supposed to be a dream holiday?’ (55); ‘was the partition supposed to be temporary?’ (56). The relationship between the epistemic and the deontic inferences, both arising from a combination of morphosyntactic (the passive structure) and semantic (future versus present temporal reference) factors, can be pinned down on the chronological axis: in the data, the epistemic examples appear later than the deontic ones, as well as being the most frequent ones in Present Day English. The path appears to follow three stages: (i) it is believed by someone else (evidential); (ii) it is expected by someone else (deontic); (iii) it is expected by someone else, not me: Speaker’s non-commitment (epistemic). Shifting from evidential to deontic then epistemic, the construction is not a counterargument to the claim that epistemic meanings arise later than deontic ones.
. Implications of the results for a theory of semantic change In this section I will discuss the theoretical implications of this study, defining its position within the current debate on the question of unidirectionality in grammaticalization. Let me summarize the results achieved so far. First, a path of change has been identified for English supposing and Italian supponendo che from verb to conjunction (i.e. a shift from a lexical into a functional head, see Section 3.1) and towards the encoding of the Speaker’s attitude (i.e. subjectification, see Sections 2.1; 3.2). Second, a mechanism for the change has been suggested, in the form of a set of specific contexts favouring, first of all, the reanalysis of supposing as a conjunction (the contexts being preposed adjuncts and interrogatives, allowing increase in scope and loosening of control between the subject of the matrix clause and the subject of the adjunct, see Section 3.1), and, secondly, the process of subjectification (i.e. cases of 1st person control, see Section 3.2). A third important factor in language change concerns motivation, which I only hinted at so far (see Sections 2.2; 3.1). The conception underlying the present paper is that language change is grounded in use: it is driven by Speakers/Writers using a certain expression for a certain purpose, as in the recruitment of supposing for argumentative uses. As the change spreads from the individual to the community, this will provoke an enhancement of its use in specific contexts, which in turn will influence the expression’s syntactic and semantic properties. Crucially, in this paper, structural and semantic shift have been shown to be interlinked, as are the contexts that favour either phenomenon: thus, interrogative contexts, i.e. a dialogic, interactive mode, favour both the reanalysis and subjectification; similarly, the loosening of syntactic control has semantic consequences: if a 3rd person
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.19 (187)
Conditionals and subjectification
control is no longer available, we ‘lose’ the source of the mental process expressed by suppose, which can then be ‘filled’ by the Speaker. Hence, the importance of defining the structural features of each stage of the reanalysis, a point often neglected in functional approaches to grammaticalization (for discussion see Fischer & Rosenbach 2000: 10; Fischer et al. 2000: 290–291). Let me now focus on the much-debated issue of unidirectionality (for an overview see Fischer & Rosenbach 2000: 19–23; Haspelmath this volume; Traugott 2000; Ziegeler this volume). If “in the Popperian sense of scientific research we should always look for counter-examples and not for cases which conform to our hypothesis” (Fischer & Rosenbach 2000: 21), then the evolution of suppose is not very interesting: indeed, it has been shown to represent a case for the unidirectionality of semantic and, to some extent, grammatical change (see below), including the be supposed to construction, which shifts from evidential to deontic/epistemic, the common thread in the development being in the nature of the passive construction (as argued in Section 4). Yet, some of the peculiarities of its path place this study at the cross-road between different conceptions of the phenomenon. First of all, unidirectionality is not intended here as a deterministic process whereby a certain change has to occur; rather, it is viewed as a heuristic principle, allowing us to make predictions about probable paths of change. In this framework, counterexamples to unidirectionality such as those discussed, for instance, in Norde (2001) or van der Auwera (2002), fall naturally within a conception of unidirectionality as a working hypothesis (see Fischer & Rosenbach 2000: 21). Secondly, the evolution of supposing stresses the predictive value of the phenomenon in the semantic realm (see Ziegeler this volume on unidirectionality as an underlying semantic fact), confirming the predictions that “meanings based in the external described situation change to meanings based in the internal (evaluative/perceptual/cognitive) described situation” (Traugott 1989: 34–35; also in Traugott & Dasher 2002: 94–95), as in the shift from concrete “place under” to “assume” in Latin subpono (Visconti 2002), and that “meanings tend to become increasingly based in the speaker’s subjective belief state/attitude towards the proposition”, as discussed in the present paper, and as shown by the development of other complex conditional connectives in Romance and the Germanic languages (e.g. Italian ammesso che or German angenommen, daß) (Visconti 2000). The grammatical side of the question, though, is less clear-cut. On the one hand, the reanalysis of English supposing from a lexical into a functional head appears to be an example of “primary grammaticalization”, “the change whereby lexical material in highly constrained pragmatic and morphosyntactic contexts is assigned functional category status” (Traugott & Dasher 2002: 81). As a result, supposing (that) becomes a closed-class item, listed as such in the “inventory of adverbial subordinators in (British) English” by Kortmann (1998: 462). On the other hand, the lexical content of the decategorialized verb re-
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.20 (188)
Jacqueline Visconti
mains high, as shown by the fact that Present Day English suppose is still found with the complementizer, as in (57): (57) Frankenstein was walking to and fro now, muttering to himself. In my hesitation to step forth lay this question: supposing that this encounter revealed my unreality rather than his. . . ? (B. Aldiss, Frankenstein unbound, 1991 (Visconti 2000: 191)) while Italian supponendo che is still productive as a verb (see Section 3.2; Visconti 2000: 110–117). I shall thus place the issue of the existence of cross-linguistic paths of semantic change at the heart of my claims, and, in particular, of those concerning the propositional to non-propositional cline, which has been shown to be a tenable and predictive one. Among the different mechanisms of change, an interesting question to be investigated further is the role of analogy in the evolution of supposing, i.e. the influence of the structural and semantic overlap with other morphemes, such as the imperative suppose (see Note 14) (see Foong Ha Yap 2002 on the lack of a detailed study of the competitive outcomes and cluster effects of ‘synonymous’ morphemes).
Notes * The data for this paper were collected during my stay at Stanford University (Spring 2000). I would like to thank Elizabeth C. Traugott for her inspiring course on ‘The Role of Pragmatics in Semantic Change’ (LSA Institute 2001), Olga Fischer, Muriel Norde, Harry Perridon and an anonimous referee for their stimulating comments on an earlier version of this paper, the British Academy and the University of Birmingham for funding the project. . The periods for English are approximately as follows: Old English (450–1150); Middle English (1150–1500); Early Modern English (1500–1770); Modern English (1770–1970); Present Day English (1970–). . Full details of the corpora are provided in the references. . See for instance Fischer et al. (2000: 26–35), Traugott and Dasher (2002: 45–48). . Statistics are to be taken as mere indicators of trends (see Section 2 on the nature of the database). . Modal structures increase moderately: Middle English 4% versus Early Modern English 7%, as well as conditional: Middle English 1.5% versus Early Modern English: 3% and negative contexts: Middle English 2% versus Early Modern English 3%. On the role of modal verbs co-occurring with suppose see Section 3.1 (ex. (30)).
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.21 (189)
Conditionals and subjectification . Many of these examples have a counterexpectation value: notice the role of these adversative uses, implying alternative, polyphonic view points, in the development of the lexeme, as sketched in Sections 3 and 4. . Thanks to Céline Guillot (CNRS, Lyon) for her kind help with the Base de Français Mèdieval. . 6/6 CEEC supposing are verbs (related free adjuncts); 27/62 in the Lampeter are conjunctions. Search of Chadwyck Literature Online (1518–1670) yields 202 examples, of which 200 verbs. . On the role of control in the categorial reanalysis of verbs (participles and gerunds) as prepositions and conjunctions see Kortmann (1992: 436–437); Kortmann and König (1992: 679–680). Both studies, however, focus almost exclusively on the category P[–NP], as in (a) and (b): (a) Following a sharp decline last year, prices rose again this year. (b) Following the instructions of his boss, John examined each part carefully. . Following the tradition adopted in X’-bar theory since Emonds (1976: 172), I consider conjunctions as prepositions with a sentential complement (P[–CP]). The relevant scope opposition is given by the level of adjunction of this PP to the CP projected by the matrix clause: thus compare (a) (VP scope) with (b) (IP scope): (a) Lamech slew Cayn with ane arowe, supposing he had bene a wylde beste. (b) Supposing the French should offer to join with the Spaniards. . . we are not to imagine that the Spaniards will accept their Proffers. As illustrated in Section 3.1, (b), where supposing introduces a frame for the following proposition, is a much more likely context to favour reanalysis of the verbal adjunct as a P[–CP] (on the notion of ‘scope change’ in diachrony see Tabor & Traugott 1998: 232–235). . Suppose also co-occurs with the inherently modal to-infinitive (28), which is also no longer possible (thanks to Olga Fischer for this observation). On the subjunctive, see e.g. Denison (1998: 160); Visser (1966: 786–941); on modal verbs, see e.g. Fischer (1992: 262– 264); Visser (1966: 1632–1674; 1705–1734; 1742–1791). . The data are from the Letteratura Italiana Zanichelli CDROM 4.0. . Many thanks to Michelle Devereux for her kind help with the Bank of English. . One factor that may have a role in accounting for the different stage of development is the influence of the imperative in English, lacking in Italian (see Visconti 2000: 191). Notice the almost identical function of supposing and the imperative in (a): (a) All the cushions laid along the length of his body. One way or another, when gravity returned, he would be ready. Supposing he was the only one. Just him and his furniture. Suppose the thud he had heard from downstairs was. . . . (Various authors, A Roomful of Birds – Scottish short stories, HarperCollins 1990 [Bank of English: British Books]) This is also evident in the lack in Italian of the puzzling Present Day English structure headed by supposing but apparently unrelated to any matrix clause, as (a) above or (40)
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.22 (190)
Jacqueline Visconti
in the text, almost intermediate between hypotaxis and parataxis, the equivalent of which is agrammatical in Italian. . On the evidential potential of the (BE) Ved to pattern and the grammaticalization of some of these matrices into auxiliary-like evidentials see Noël (2001), who, however, “skips” be supposed to in his analysis “because it has (also?) undergone a different grammaticalization into an expression of ‘participant-external necessity”’ (Noël 2001: 292, f. 11). . The trend I am suggesting is confirmed by Ziegeler’s (2002: 10–11) quantitative analysis of 1,372 OED examples from 1400 to 1989, where she furthermore identifies an interesting correlation between meanings of ‘belief ’ or ‘hypothesis’ with non-human subjects versus meanings of ‘expectation’ with human subjects.
References Bybee, Joan & William Pagliuca (1987). “The Evolution of Future Meaning”. In Anna G. Ramat, O. Carruba, & G. Bernini (Eds.), Papers from the Seventh International Conference on Historical Linguistics (pp. 109–122). Amsterdam & Philadelphia: John Benjamins. Chafe, Wallace (1986). “Evidentiality in English Conversation and Writing”. In Wallace Chafe & Johanna Nichols (Eds.), Evidentiality: The Linguistic Coding of Epistemology (pp. 172–261). Norwood, NY: Ablex Publishing Corporation. Denison, David (1998). “Syntax”. In Suzanne Romaine (Ed.), The Cambridge History of the English Language, Vol. 4 (1776–1997) (pp. 92–328). Cambridge: Cambridge University Press. Emonds, Joseph E. (1976). A Transformational Approach to English Syntax. New York: Academic Press. Fischer, Olga (1992). “Syntax”. In Norman Blake (Ed.), The Cambridge History of the English Language, Vol. 2 (1066–1476) (pp. 207–408). Cambridge: Cambridge University Press. Fischer, Olga, Ans van Kemenade, Willem Koopman, & Wim van der Wurff (2000). The Syntax of Early English. Cambridge: Cambridge University Press. Fischer, Olga & Anette Rosenbach (2000). “Introduction”. In Olga Fischer, Anette Rosenbach & Dieter Stein (Eds.), Pathways of Change. Grammaticalization in English (pp. 1–37). Amsterdam & Philadelphia: John Benjamins. Foong Ha Yap (2002). “On Synonymy Effects in Grammaticalization”. Paper presented at the New Reflections on Grammaticalization Conference, University of Amsterdam, April 2002. Kiefer, Ferenc (1994). “Modality”. In R. E. Asher & J. M. Y. Simpson (Eds.), The Encyclopedia of Language and Linguistics, Vol. 5 (pp. 2515–2520). Oxford: Pergamon. König, Ekkehard (1992). “From Discourse to Syntax: The Case of Concessive Conditionals”. In Rosemary Tracy (Ed.), Who Climbs the Grammar Tree (pp. 423–434). Tübingen: Niemeyer. Kortmann, Bernd (1991). Free Adjuncts and Absolutes in English: Problems of Control and Interpretation. London & New York: Routledge.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.23 (191)
Conditionals and subjectification
Kortmann, Bernd (1992). “Reanalysis Completed and in Progress: Participles as Source of Prepositions and Conjunctions”. In G. Kellermann & M. D. Morrisey (Eds.), Diachrony within Synchrony: Language History and Cognition (pp. 429–453). Frankfurt: Lang. Kortmann, Bernd (1998). “Adverbial Subordinators in the Languages of Europe”. In Johan van der Auwera (Ed.), Adverbial Constructions in the Languages of Europe (pp. 457–561). Berlin: Mouton de Gruyter. Kortmann, Bernd & Ekkehard König (1992). “Categorial Reanalysis: the Case of Deverbal Prepositions”. Linguistics, 30, 671–697. Lyons, John (1982). “Deixis and Subjectivity: Loquor, ergo sum?” In Robert J. Jarvella & Wolfgang Klein (Eds.), Speech, Place and Action: Studies in Deixis and Related Topics (pp. 101–124). New York: Wiley. Noël, Dirk (2001). “The Passive Matrices of English Infinitival Complement Clauses. Evidentials on the Road to Auxiliarihood?” Studies in language, 25, 255–296. Norde, Muriel (2001). “Deflexion as a Counterdirectional Factor in Grammatical Change”. Language Sciences, 23, 231–264. Nordlinger, Rachel & Elizabeth C. Traugott (1997). “Scope and the Development of Epistemic Modality: Evidence from ought to”. English Language and Linguistics, 1, 295– 317. Rey, A. (1994). Dictionnaire historique de la langue française. Paris: Le Robert. Rissanen, Matti, Merja Kytö, & Minna Palander-Collin (Eds.). (1993). Early English in the Computer Age: Explorations through the Helsinki Corpus. Berlin: Mouton de Gruyter. Sweetser, Eve (1990). From Etymology to Pragmatics. Cambridge: Cambridge University Press. Tabor, Whitney & Elizabeth C. Traugott (1998). “Structural Scope Expansion and Grammaticalization”. In Anna G. Ramat & Paul J. Hopper (Eds.), The Limits of Grammaticalization (pp. 229–272). Amsterdam & Philadelphia: John Benjamins. Tobler, A. & E. Lommatzsch (1973). Alt Französisches Wörterbuch. Wiesbaden: Steiner. Traugott, Elizabeth C. (1989). “On the Rise of Epistemic Meanings in English: An Example of Subjectification in Semantic Change”. Language, 65, 31–55. Traugott, Elizabeth C. (1995). “Subjectification in Grammaticalization”. In Dieter Stein & Susan Wright (Eds.), Subjectivity and Subjectivisation (pp. 31–54). Cambridge: Cambridge University Press. Traugott, Elizabeth C. (2000). “From Etimology to Historical Pragmatics”. Plenary paper presented at the Studies in English Historical Linguistics Conference, UCLA, May 2000. Traugott, Elizabeth C. & Richard B. Dasher (2002). Regularity in Semantic Change. Cambridge: Cambridge University Press. Van der Auwera, Johan (1999). “On the Semantic and Pragmatic Polyfunctionality of Modal Verbs”. In Ken Turner (Ed.), The Semantics-Pragmatics Interface from Different Points of View (pp. 49–64). Oxford: Elsevier. Van der Auwera, Johan (2002). “More Thoughts on Degrammaticalization”. In Ilse Wisher & Gabriele Diewald (Eds.), New Reflections on Grammaticalization (pp. 19–29). Amsterdam & Philadelphia: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 16:01
F: TSL5908.tex / p.24 (192)
Jacqueline Visconti
Verhagen, Arie (2000). “The Girl that Promised to Become Something: An Exploration into Diachronic Subjectification in Dutch”. In Thomas F. Shannon & Johan P. Snapper (Eds.), The Berkeley Conference on Dutch Linguistics 1997: The Dutch Language at the Millenium (pp. 197–208). Lanham, MD: University Press of America. Visconti, Jacqueline (2000). I connettivi condizionali complessi in italiano e inglese. Uno studio contrastivo. Alessandria: Edizioni dell’Orso. Visconti, Jacqueline (2002). “From Latin subpono to English supposing: the data”. In Luis I. Rabade & S. M. Doval Suarez (Eds.), Studies in Contrastive Linguistics. Proceedings of the Second Contrastive Linguistics Conference, Santiago, October 2001 (pp. 1051–1060). Santiago: University of Santiago Press. Visser, F. Th. (1963–1966). A Historical Syntax of the English Language. Vols. 1–2. Leiden: Brill. Ziegeler, Debra. (2002). “Epistemic Modality and Aspect”. Paper presented at the ICEHL, Glasgow, August 2002.
Corpora The Bank of English, Cobuild, University of Birmingham (http://www.titania.collins. cobuild.co.uk). Base de Français Médiéval, compiled by the Laboratoire Analyses de Corpus de l’E.N.S. Lettres et Sciences Humaines, Lyon [BFM]. The Corpus of Early English Correspondence, compiled by Terttu Nevalainen, Helena Raumolin-Brunberg et al., University of Helsinki, 1998, ICAME [CEEC]. Helsinki Corpus of English Texts, Diachronic Part. See Rissanen, Kytö and Palander-Collin (Eds.), 1993 [Helsinki]. Lampeter Corpus of Early Modern English Tracts, compiled by Josef Schmied & Eva Hertel, 1991–, ICAME [Lampeter]. Letteratura Italiana Zanichelli CD-ROM 4.0, ed. by Pasquale Stoppelli & Eugenio Picchi, Zanichelli [LIZ]. Literature Online (LION) (http://lion.chadwyck.co.uk/athens) [Chadwyck Literature Online]. The Middle English Dictionary. Ann Arbor: University of Michigan Press, 1956– (http://www.hti.umich.edu/dict/med) [MED]. The Oxford English Dictionary Online. Oxford: Oxford University Press, 2000 (http://dictionary.oed.com) [OED].
Unidirectionality in the grammaticalization of modality in Greek* Anastasios Tsangalidis University of Ioannina
.
Introduction: Mood and modality
Mood and modality have been variously studied in both typology and grammaticalization. In this paper it is argued that a number of descriptive and theoretical issues involved in the discussion of mood and modality are best treated within a grammaticalization approach. However, the diachronic development of modal markers in Greek1 pose a number of questions relating to the directionality of grammaticalization. There are various difficulties in defining mood as opposed to modality and in characterizing the relation between them in precise terms.2 One of the advantages of a grammaticalization approach to modality is that it can provide a quite straightforward definition of this relation. Regular inflection is undeniably a case of grammatical marking and a definitional property of “mood”: as Bybee et al. put it (1994: 181), “modality is the conceptual domain, and mood is its inflectional expression”. If all other forms of expression of modal concepts can also be associated with the same notional category of Modality, then the only difference will be one in “degree of grammaticalization”. In Palmer’s terms (2001: 4), “there are two ways in which languages deal grammatically with the overall category of modality. These are to be distinguished in terms of (i) modal system and (ii) mood”. This view can bring together both ‘moods’ (like the subjunctive in Latin, Classical Greek or the Romance languages) and ‘modal systems’ of ‘modal (auxiliary) verbs’ (such as those found in English or German). To the extent that we are dealing with differences in degree rather than in kind, the next question involves Palmer’s (2001: 104) typological observation (and prediction) that inflectional moods and periphrastic modalities are not likely to co-exist and that, if they do, “one will, in time, replace the other.”
Anastasios Tsangalidis
Thus, three points relate the study of modality to that of grammaticalization: First, the extent to which both moods and modal systems (in the sense of Palmer 2001) can be seen as members of a single category. Second, the possible correspondence between the “degree of grammaticalization” of a modal marker and the types of modality that it can express (e.g. Bybee et al. (1994: 241) report that “agent-oriented modalities tend to have non-bound, periphrastic expression while the other modality types tend to have bound expression”). Third, the possible historical connection between the two: Bybee et al. 1994 predict that a lexical marker of agent-oriented modality may develop into a subordinating mood – but not vice versa. Moreover, they suggest a diachronic pattern whereby the grammaticalization of agent-oriented modalities leads to other types of modalities while gradually assuming inflectional expression. The general schema involving their four types of modality is presented in Table 1 and will be relevant to the overall discussion below: Table 1. Paths of development for modalities (Bybee et al. 1994: 241) SPEAKER-ORIENTED AGENT-ORIENTED
SUBORDINATE EPISTEMIC
Thus, assuming that we are dealing with a single category, which may be expressed in a number of ways, some of these will be ‘more’ grammatical than others. Moreover, the thesis of unidirectionality (along with the strong version of the “form-meaning co-variation hypothesis” – Bybee et al. 1994: 20) predicts that ‘more’ grammaticalized items will center around ‘later’ uses on a grammaticalization path, whereas ‘earlier’ uses will be found with ‘less’ grammaticalized formations.
. Mood and modality in Greek In view of the questions raised above, the situation in Greek should be particularly interesting, in that in addition to a morphological [+/–imperative] distinction, a set of modal particles and a set of modal verbs can be identified in the language – as summarized in (1):3 (1) Modal sub-systems in Greek: a. inflectional [+/–imperative] b. periphrastic [+/–subjunctive] through the use of “modal particles” c. modal verbs of necessity and possibility
Unidirectionality in the grammaticalization of modality in Greek
Of the three subsystems, only the first is generally accepted and analysed as a grammatically marked modal distinction in all accounts of the Greek verb. This is not surprising in view of our commonly held views on the expression of grammatical categories in Indo-European languages plus the tradition of grammatical analysis in the Classical languages (cf. Palmer 1986, 2001 with particular reference to modal categories).
. A morphological imperative Imperative forms have their own endings, which distinguish between active and non-active voice and may be attached to both perfective and imperfective stems. They only appear in second person, singular and plural, and they do not inflect for tense.4 The Greek imperatives seem to conform to general descriptions of the imperative in that they do not appear in subordinate clauses or interrogative sentences. Furthermore, imperatives do not negate, and the periphrastic ‘subjunctive’ is used instead – as also attested in a number of other languages (cf. Palmer 2001: 138–142). It should be noted that the imperative is not zero-marked – since it makes use of a distinct paradigm of endings. Moreover, historically, there can be no identifiable source from which these have developed, since a similar imperative source construction is found in the earliest forms of the language.5 This seems to contradict the prediction in Bybee et al. (1994: 210) that “[n]on-zero expression of imperative must arise as overt material grammaticizes” – as well as the more general implication that all grammatical elements must have a “prior lexical history” (cf. Hopper & Traugott 1993: 128–129). In other words, as Joseph (this volume) puts it, there are cases of affixes which seem to remain stable over thousands of years and for which we must assume that historically “it’s been affixes all the way down”. In any case, the imperative cannot be used as evidence for (or against) the assumption of grammaticalization or unidirectionality, as far as its origin is concerned. However, there may still be a case for grammaticalization from speakeroriented to subordinating modality, if the imperatives in sentences like (2) can be analysed as a conditional construction of some sort: (2) Ela sto parti ke tha perasume kala Come-imp to-the party and fut pass-1pl well “If you come to the party, we’ll have a good time.” (Lit.: “Come to the party and we’ll have a good time.”) Thus, to the extent that (2) involves a subordinate structure, it might be considered an instance of subordination as a late stage in the development of a modal gram. It
Anastasios Tsangalidis
is possible to object to the analysis of (2) as a subordinate structure since it includes ke (and) and is thus an instance of coordination.6 However, ke is also used as a marker of subordination in a number of other cases (such as ksero/boro ke γrafo “I can write” (lit.: “I know/can and write”)). Moreover, examples like (2) need not involve ke at all, as in Ela sto parti, na δis ti tha pathis (“If you come to the party, you’ll see what may happen to you” (especially when used as a threat – lit.: Come to the party, that you see what will suffer-you).
. Indicative and subjunctive Formal facts in Greek have been claimed to support the assumption of a binary mood system beyond the inflectional imperative/non-imperative contrast. The indicative is expressed either on the verb form alone or in combination with the future marker tha, whereas the subjunctive is always formed with na or as, i.e. periphrastically from a historical standpoint and even synchronically to the extent that they do not admit of an affixal analysis for the contemporary language. The most salient formal feature that characterizes the subjunctive, as argued by Philippaki-Warburton (1994), is the different form of negation it takes (min, rather than indicative δen). As already mentioned, the most usual formal objection to an analysis of these forms in terms of a mood contrast reflects the hesitation to admit non-affixal expression of moods. However, it has also been argued that both tha and na (though not as) should be analysed as affixes rather than particles (e.g. Joseph 1990, 2001; Tsimpli 1990). The synchronic and diachronic status of the three particles is briefly discussed below.
.. The case of tha Tha has been discussed as one of the prime examples of the grammaticalization of a verb of desire into a marker of futurity and then into a marker of epistemic modality (cf. Meillet 1912; Jespersen 1924; Joseph 1990; Bybee et al. 1994; Tsangalidis 1999; Pappas & Joseph 2001; Joseph & Pappas 2002; Heine & Kuteva 2002). Currently, tha constitutes the main means of future-time reference (and all associated uses, such as the expression of intentions, promises, threats, as well as habits and eternal truths) and it is also a very clear marker of epistemic modality. Like all Greek modal particles, it may accompany any finite verb form, and therefore may function as a conditional or counterfactual, in particular when combined with [+past][–perfective] verb forms. Tha can only function either as a future marker or a marker of epistemic modality (in other words, there is no volition in its semantics). The epistemic sense is excluded when reference is to future time and it is forced when reference is (necessarily and deictically) to past time. Both epistemic and future are available in all other cases (cf. Tsangalidis 1999, 2001).
Unidirectionality in the grammaticalization of modality in Greek
Jespersen (1924) contrasts tha with will in English and notes that “the idea of volition seems to have been completely obliterated from the combinations with tha” and that therefore tha “has now become a pure temporal particle” (1924: 260– 261).7 The stages of development in the case of tha are summarized in (3). (3) thelo γrapsein >thelo na γrapso >(thelei na γrapso) want.1sg write.inf >want.1sg na write.1s >(want.3sg na write.1sg) >the na γrapso >tha γrapso >reduction of thelei >the + na > tha (cf. Joseph 1983, 1990, 2001; Tsangalidis 1999) Leaving aside the initial observation that tha is no longer a marker of volition, it should be stressed that it has also developed a clear use as a marker of epistemic modality for present and past situations. In other words, it is not just the bleaching of all ‘desire’ that makes it interesting but its development into a marker of epistemic modality which seems to conform to the overall predictions in Bybee et al. (1991) and (1994). The semantic ages for futures established in their 1991 corpus of 67 languages, are listed in (4) below, and as they report (1994: 279), “we found significant relationships between the four semantic ages and the measures of formal grammaticization”. (4) The semantic ages for futures (according to Bybee et al. 1991, 1994: 279): Futage 1: Futures with the agent-oriented uses of obligation, desire, and ability. Futage 2: Futures with the later agent-oriented uses of intention, root possibility, and the specific use of immediate future. Futage 3: Grams with simple future as their only use. Futage 4: Futures with epistemic, speaker-oriented, and subordinate uses. Therefore, tha is arguably a quite clear example of an age 4 future gram, in that it has particle (or even affixal) status and cannot signal desire (or ‘agent-oriented modality’). However, the stages in (4) should not be taken as precise pictures of any actual historical changes. As examples like (5) show, the source construction for tha had developed in the mid-17th century a quite clear epistemic use – which did not depend either on the reduction of thelo into tha or the loss of the infinitive.8 (5) koimastai thelei alethina, gia keino den probainei sleep-inf want-3sg really, for that neg walk-3sg “He will be sleeping really, and so he is not coming.” (Holton 1993: 124) In fact, then, the details of the historical development may not support the stages of the development presented above in a strict sense. However, from the point
Anastasios Tsangalidis
of view of grammaticalization theory, they need not, since “‘stage’ refers, for descriptive convenience, to a focal point and/or a clearly profiled characteristic of linguistic behaviour” (Kuteva 2001: 11 building on Heine 1993). More generally, as also argued there, “two important regularities should be noted: conceptual shift from lexical to grammatical content precedes all other shifts; and cliticization and erosion normally begin later than the other shifts”.
.. The case of na The historical development of na (cf. Joseph 1981, 1983, 1990; Horrocks 1997) has also been discussed in the literature on grammaticalization (e.g. Bybee et al. 1994: 212–230) as an example of a conjunction reduced to a particle with a primary subordinating function – which however has been generalized to a great variety of contexts and uses. In fact, as shown in (6) and (7), na is the particle with the most diverse uses and semantics in Greek, ranging from the expression of various modalities to the function of a semantically null – “merely relational” – element in cases like (6c) and (6d). (6) Subordinate uses of na a. Thelo (protimo, epithimo, prepi. . . ) na fiγo want.1sg (prefer.1sg, wish.1sg, must.3sg . . . ) NA leave.1sg “I want (prefer, desire, must. . . ) to go.” b. Pistevo na ta kataferi believe.1sg NA them manage.3sg “I believe he will make it.” (‘weak’ truth claim) c. Katafere (prospathise) na taxiδromisi to γrama managed.3sg (tried.3sg) NA post.3sg the letter “He managed (tried) to post the letter.” d. Ton vlepo na fevji. him see.1sg NA leave.3sg “I can see him go.” (Veloudis 2001: ‘live reports’) In (7) the main clause uses of na are presented. Non-factuality and ‘speakerorientation’ should be noted as the common basis of all its main clause uses – and this contrasts with the subordinate cases where speaker-orientation or even non-factuality (as in (6d)) may be missing. (7) Main clause uses of na (examples from Joseph & Philippaki-Warburton 1987: 180–181) a. na pas na δis ton patera su! NA go-2sg NA see-2sg the father-acc you-gen “(You should) go to see your father!”
Unidirectionality in the grammaticalization of modality in Greek
b. na zisete! NA live-2pl “May you live (a long life)!” (a conventional exclamation of wellwishing) c. ke na skeftis pos δen mu milise and NA think-2sg comp neg me-gen spoke-3sg “And to think that he did not speak to me!” d. ax, na se ixa konda mu oh NA you-acc had-1sg near me-gen “Oh, I wish I had you near me!” As far as the diversity of the subordinate uses of na in (6) is concerned, it is interesting that Bybee et al. (1994: 241) predict that “a single gram can develop multiple subordinate clause uses by separate paths. Thus all the uses of a subjunctive may not be contiguous on a single path, and may not be closely related semantically”. The history of na is quite complicated (cf. Joseph 1981, 1983, 1990). As briefly shown in (8), starting from a purposive conjunction, it has ended up as a marker of general subordination as well as (main-clause) marker of speakeroriented modality. (8) ´hina>´ina>i´na>na so that>so that>that purposive>generalized subordinator + main clause marker of speakeroriented modality As discussed above in relation to tha, the formal changes do not historically correlate with the semantic changes. As shown in Horrocks’ description of the situation in New Testament Greek in (9), again ‘late’ uses were found with the ‘source construction’: (9) it is clear that this ‘conjunction’, like its successor . . . na in Modern Greek, could already be used with a main verb to express permission/obligation, and that it had accordingly been downgraded in certain contexts to the status of a mood marker, thus becoming available in main as well as subordinate clauses. (Horrocks 1997: 95) In fact the history of na has even led Bybee et al. to admit some variation in the application of their “harmony principle”. Briefly, according to that principle, it is expected that a marker of imperative or hortative in main clauses may in time come to function as a complement of a verb like ‘order’ or ‘oblige’– and, similarly, one found in optative main clauses may in time function as a complement of verbs like ‘want’ (cf. Bybee et al. 1994: 212–225). On the basis of the history of na this
Anastasios Tsangalidis
principle is argued to operate along with a parallel system of development of main clause uses from subordinate clause uses (cf. Bybee et al. 1994: 224). Strictly speaking, the story of na contradicts the more general observation that “[s]ubjunctives from different sources all have in common the fact that the subordinating uses show up very late on the grammaticization paths” (Bybee et al. 1994: 213). Given that na had originated as a subordinating conjunction, and only later acquired its main clause uses as a marker of speaker-oriented modality, it runs counter to Bybee et al’s general prediction.9 The quotation in (10) refers to the second century BC and illustrates the point. (10) The damaging effects of sound change therefore led to a growing need to ‘mark’ subjunctives as such, and . . . ina began to develop languageinternally as an ‘empty’ mood marker, first in subordinate, but eventually also in main clauses that required a modal verb form (a process that was finally completed in the Middle Ages). (Horrocks 1997: 76, emphasis added)
.. The case of as This particle is usually described as a hortative/concessive particle (formally a member of the subjunctive paradigm) which etymologically derives from the imperative form of the verb afiemi “let, leave”. It should be noted that the uses of as constitute a subset of the uses of na: practically all the examples with as in (11) might include na instead. Crucially, the reverse does not hold, and as cannot substitute for na in subordinate contexts. In fact, unlike na, as can only appear in main clauses. (11) The uses of as a. as pame ki emis. AS go.1pl and we “Let us go, too.” (hortative) (Karantzola 1995) afu δe thelun. b. as min erthun AS neg come.3pl since neg want.3pl “Let them not come, since they don’t want to.” (permissive) (Karantzola 1995) c. ela sto parti mas, ki as feris ke ton come.imp to-the party ours and AS bring.2sg and the andra su! husband yours “Come to our party, even if you have to bring your husband.” (concessive) (Karantzola 1995)
Unidirectionality in the grammaticalization of modality in Greek
d. as kanis ta mathimata su ke tha se afiso na AS do.2sg the lessons your and fut you let.1sg subj vjis ekso go.2sg out “If you do your homework, I’ll let you go out.” (conditional; Nikiforidou 1990) e. as vreksi the mu! AS rain.3sg god my “If only it would rain, God!” (expressing wish; Nikiforidou 1990) f. as itane ikosi xrono peripu AS was-3sg twenty years-gen.pl about “He may have been approximately twenty years old” (‘probable inference’; Joseph & Philippaki-Warburton 1987) As far as the history of the development of as is concerned, there are numerous examples in the New Testament, where the source construction for as is found (Nikiforidou 2001). This involves the imperative of the verb afiemi, namely afes, in sentences like (12) (which apparently confused the King James translators):10 (12) hoi de loipoi elegon afes idomen ei erxetai Elias soson the prt rest said let.imp see.subj.1pl if comes Elias save.inf auton him “The rest said, Let be, let us see whether Elias will come to save him.” (King James Version) (Matthew 27.49) Nikiforidou (2001: 1066) locates the first use of the form as in the 6th or 7th century. The relevant grammaticalization may be briefly outlined in (13): (13) afes>as leave.imp>invariable particle glossing ‘let/lets’ imperative of Classical lexical verb afiemi “leave/let” > grammatical hortative. . . concessive particle Some of the problems with the analysis of the two particles as subjunctive markers should be obvious from this presentation. On one hand, as is never found in subordination, and this could be a problem for describing it as “subjunctive”, by definition. On the other hand, it is interesting that both particles select the same negation marker (and different from the indicative negator) and that their main clause uses seem to coincide. In fact, all of the main clause uses of both particles may be accurately described as instances of “speaker-oriented” modality in the sense of Bybee et al. 1994 (based on Bybee 1985: 166–169).11 The deontic source,
Anastasios Tsangalidis
in Lyons’ (1977: 843) terms is always the speaker – and this is a crucial difference from the situation with the modal verbs (Tsangalidis 2001). Moreover, it is interesting to note that the Greek imperative and subjunctive data conform to the problematic pattern acknowledged by Bybee et al. (1994: 241). Whereas their analysis “would predict that subjunctives would be more grammaticized and thus more bound than imperatives”, their own statistics “show that imperatives are more likely to be bound”.
. Modal auxiliaries Modal auxiliaries are hardly ever mentioned in Greek descriptive grammars, and even if they are, what is usually noted is that they are impersonal – rather than modal or auxiliary.12 Their auxiliary status is questioned as they differ from the standard have and be auxiliaries in important ways that will not be discussed here (but see Tsangalidis 2001). As shown in (14a) and (14b), the impersonal verb prepi may express both epistemic and non-epistemic necessity. Moreover, prepi – although impersonal – can still inflect for tense and thus a further impersonal past form is available (shown in (14c)), which, unsurprisingly, may not be used in the epistemic sense (cf. Palmer 1986; Joseph & Philippaki-Warburton 1987; Iakovou 2001; Tsangalidis 2001): Prepi na γrapsis must-3sg NA write-perf-pres-2sg “You must write.” (epistemic excluded) b. Prepi na γrafis must-3sg NA write-imperf-pres-2sg “You must be writing.” (both senses permitted) c. Eprepe na γrafis/γrapsis must-past-3sg NA write-(im)perf-pres-2sg “You had to write.” (epistemic excluded)
(14) a.
The impersonal and personal versions of boro may also express both epistemic and non-epistemic possibility. However, the epistemic sense is available only in the case of the invariable impersonal version (as in (15a)), whereas ability, permission etc. can only be expressed by the fully inflecting personal boro, as in (15b) and (15c): (15) a.
bori na γrafis/γrapsis may-3sg NA write-(im)perf-pres-2sg “You may write.” (epistemic only)
Unidirectionality in the grammaticalization of modality in Greek
b. boris na γrafis/γrapsis can-pres-2sg NA write-(im)perf-pres-2sg “You can/may write.” (epistemic excluded) c. boruses na γrafis/γrapsis can-past-2sg NA write-(im)perf-pres-2sg “You were able to write.” (epistemic excluded) It should be clear from this brief presentation of the two modals that their current status exemplifies (a version of) the form-meaning co-variation hypothesis of Bybee et al. (1994: 20), which assumes “a direct, even causal, link between semantic and phonetic reduction in the evolution of grammatical material”. Although there is no phonetic reduction involved, there is paradigmatic reduction in that only one form, namely the 3rd person singular, has survived in the impersonal case. Briefly, both verbs have started life as fully inflecting lexical verbs. Of the two verbs only boro still survives (cf. (14) above), in the agent-oriented sense of “being able to”.13 In view of the paths of development shown in Table 1, it is only to be expected that the semantics of the fully inflecting lexical boro will be closer to the initial stages of the development, whereas the invariable bori will be closer to the final stages. Prepi, being impersonal but not completely invariable may be seen to correspond to an intermediate stage, as shown in Table 2. Again, stages should not be taken as pictures of historical periods.14 However, one obvious advantage of a grammaticalization analysis of these facts concerns precisely the synchronic description of the modal auxiliaries. As has been argued (Tsangalidis 2001), the two versions of bori/boro are best associated at the descriptive level in the manner suggested in Table 2. In other words, there is no reason to treat the two versions of bori/boro as distinct items (as traditional descriptions of Greek seem to be doing) for two reasons. First, there are numerous cases of ambiguity between the invariable bori and the homophonous 3rd person singular bori of the personal verb. Secondly, the traditional distinction suggested misses the recurrent alternation between epistemic and agent-oriented modality in the semantics of modals of possibility and necessity. This typologically valid generalization (cf. van der Auwera & Plungian 1998; and similarly, Heine 1993; Bybee et al. 1994) can only apply to the Greek case if the personal boro can be associated with the impersonal bori. Table 2. Greek modal verbs: current stages of development (Tsangalidis 2001) boro
prepi
bori
Form fully inflecting impersonal but still past/non-past impersonal and invariable Function epistemic excluded agent-/speaker-oriented and agent-/speaker-oriented epistemic excluded
Anastasios Tsangalidis
Similarly, as was mentioned briefly above, the two verbs cannot be termed auxiliary in the traditional grammar sense. However, it is important that they are treated as somehow different from lexical verbs which express some modal notion, since they are clearly different both from verbs like order, promise, wish etc. and from expressions like I think, I believe, It is possible, etc.
. Conclusions and implications for the general theory In sum, the Greek facts briefly discussed above point to the following conclusions. First of all, the various subsystems of modality in Greek constitute clear evidence against the assumption of ‘functional need’ as an explanation in grammaticalization. Since the uses of as have been a subset of the uses of na for a number of centuries (see (7) and (11) above), the rise of grammatical meaning in the case of as cannot be explained by the need to fill some gap. Moreover, the persistence of the imperative is quite striking in this respect, since it has been there all along, despite pressures from all directions. As far as Palmer’s prediction is concerned (that one type of modal markers will in time replace the others), one is tempted to argue that this is precisely what has happened with the Classical Greek (inflectional) Subjunctive and Optative. It may seem that these have been replaced by na and as forms, respectively, but this replacement is restricted to main clause uses only. The lack of verbal features on all three particles constitutes the main property which distinguishes them from modal verbs and determines their modal value. This may well be accommodated within the predictions of Bybee et al. (1994), as the particles have moved away from agent-oriented modality and are used in the later senses, speaker-oriented modality being the next stage (cf. Table 1 above). In syntactic terms, the particles are simply not able (or no longer able) to assign any thematic role (since they lack all verbal features). Thus, in actual use, the deontic source in the case of the particles will always be identified with the speaker. More generally, the prediction in Bybee (1985) that reduced forms cannot express agentoriented modality is borne out in view of the behaviour of each class of modal items in Greek, summarized in Table 3. Table 3. Greek mood and modality: current stages of development (Tsangalidis 2001)
Modal verbs Modal particles Inflectional imperative
Agent-oriented
Speaker-oriented
Epistemic
Subordinating
+ – –
+ + +
+ + –
– + (+)
Unidirectionality in the grammaticalization of modality in Greek
As far as the various claims regarding the unidirectionality of grammaticalization are concerned the following points may be noted: That the process of grammaticalization involves movement from the lexical to the grammatical (or from the less to the more grammatical) follows ex definitio. Similarly, the term ‘grammaticalization’ implies unidirectionality tautologically, in the general case (cf. Joseph 2001 and Haspelmath this volume). When it comes to details, each particular case has to be evaluated in its own terms. It was seen that unidirectionality in the sense of movement from the lexical to the grammatical (or from the less grammatical to the more grammatical) can be argued to hold in the development of thelo into tha, of afes into as, of prepo into prepi, of boro into bori and even of hina into na. However, this sense of unidirectionality cannot be supported in view of the development of main clause uses of na from a subordinating conjunction or of the stability of the imperative. Movement towards greater grammatical status is not “monotonic” (in Joseph’s terms, this volume) and can be violated (as the factors affecting change can be numerous – cf. Newmeyer 1998; Janda 2001; for example). Moreover, the outcome of any given grammaticalization can not always be predicted as shown by the development of main clause uses in the case of ina/na. Thus, the nature of grammaticalization paths (and stages) has to be reconsidered: It seems that rather than depicting a predictable course of development, they can only describe some attested change in a given language – which may be comparable to some other change in another language. The development of modal verbs (in, say, Greek and English) is a case in point. Even reconstruction of previous stages may be quite impossible – thus challenging the view expressed in Bybee et al. (1994: 12) that “the unidirectionality hypothesis is fundamentally an assertion about the orderliness and tractability of semantic change”. However, it is important to realize that synchronically unrelated items (or distinct uses of a single item) can be related diachronically. Thus, future and epistemic tha (or epistemic and non-epistemic readings of modal verbs in, say Greek, English or German) may not be considered accidentally homophonous (cf. Heine 1993; Tsangalidis 2001). In this respect, the important argument for positing a principle of unidirectionality in grammatical change is not a claim to its explanatory power but rather to its generalizability at the descriptive level.
Notes * I am grateful to Olga Fischer, Brian Joseph, an anonymous reviewer and the audience at the International Conference on New Reflections on Grammaticalization 2 for comments and suggestions. All errors and inadequacies remain, of course, mine. . Throughout the paper “Greek” refers to “Modern Greek” unless specified otherwise.
Anastasios Tsangalidis . However, there are not many objections to more general definitions such as the one found in Palmer (1986: 21), that “[t]he distinction between mood and modality is then similar to that between tense and time, gender and sex”. . Given that this paper is concerned with modality in Modern Greek I will leave aside the oppositions found in (inflectional) mood in Classical Greek. These included the optative, the subjunctive and the imperative, as well as the indicative. . Joseph (1983, 1985) argues that Modern Greek imperatives are in fact not marked for person but rather only for number on the basis that they lack a person paradigm. Thus, the second person interpretation is pragmatically determined. . In fact, the relevant reconstructions go back to Proto-Indo-European. It should be noted, however, that in Classical Greek the imperative paradigm also included third person (singular and plural). . I owe this observation to an anonymous reviewer who also points out that similar uses of imperatives with co-ordinated clauses are found in English and other languages (cf. the literal gloss in (2)). . Jespersen’s statement was meant to juxtapose volition with temporal use, ignoring the modal and conditional uses of tha. . The development of today’s clearly epistemic uses of tha, i.e. tha + V [past] is rather complicated especially since a precursor can arguably be found in Medieval Greek constructions involving past forms of thelo + infinitive (see Pappas 2001). . The situation is further complicated by the appearance in Greek of a homophonous stressed version of na that will not be discussed here (but see Joseph 1981; Christidis 1990; Veloudis 2001; Tsangalidis 2002). . The afes idomen syntagm is arguably monoclausal, in that afes only accompanies the subjunctive form idomen to reinforce the intended hortative/optative meaning. A hortative interpretation of a bare subjunctive might have been unlikely by the time the text was written, given that the indicative/subjunctive distinction had been already neutralized. In other words, afes idomen should be rendered as ‘Let us see’ instead of the translation offered by the King James translators in example (12) in the text. . Strictly speaking not all main clause uses are speaker-oriented – to the extent that examples like (11f) are indeed instances of epistemic modality. However, the crucial point is that, as predicted by Bybee et al. (1994), they can only mark speaker-oriented modality as opposed to agent-oriented modality. . That modality and impersonalness often go together is shown by changes in the modal system of Middle English and other languages too. On the other hand, the description of Greek modals as impersonals in descriptive grammars generally misses the correspondence of the personal verb boro with its impersonal counterpart bori. . A personal form prepo (which used to be freely used in much earlier forms of the language, as in (i), meaning something like ‘fit’ or ‘resemble’) can still be found only in very restricted – poetic, fossilized or otherwise marginal – contexts (as in (ii)):
Unidirectionality in the grammaticalization of modality in Greek
(i)
prepeis thugateron morphen miai fit-2sg daughters-gen appearance-acc one-dat “You look like one of the daughters.”
(Euripides, Bacchae, 917)
(ii) tu prepun times ke δokses (marginal) him fit-3pl honours and glories “He is worthy of great honours.” Note that this form can appear only in the 3rd person and accompanied by a benefactive clitic pronoun. As such it will be excluded from the discussion. Moreover, even if this personal version were to be regarded as part of the language along with the impersonal prepi, given that the meanings are so different, one has to wonder if they really are to be considered as still related. . It is interesting to note in addition that the source construction of prepi is several centuries older than that of boro/bori – cf. Iakovou (2001).
References Bybee, Joan L. (1985). Morphology: A Study of the Relation Between Meaning and Form. Amsterdam & Philadelphia: Benjamins. Bybee, Joan L., William Pagliuca, & Revere Perkins (1991). “Back to the Future”. In Elizabeth Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization, Volume 2 (pp. 17– 58). Amsterdam & Philadelphia: Benjamins. Bybee, Joan L., Revere Perkins, & William Pagliuca (1994). The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago & London: The University of Chicago Press. Christidis, Anastasios-Phivos (1990). “On the Categorial Status of Particles: The Case for ‘Holophrasis”’. Lingua, 82, 53–82. Haspelmath, Martin (this volume). “On Directionality in Language Change with Particular Reference to Grammaticalization”. Heine, Bernd (1993). Auxiliaries: Cognitive Forces and Grammaticalization. Oxford: Oxford University Press. Heine, Bernd & Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Holton, David (1993). “The Formation of the Future in Modern Greek Literary Texts Up to the 17th Century”. Proceedings of Neograeca Medii Aevi, Vol. A, 118–128. Venice. Hopper, Paul & Elizabeth Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Horrocks, Geoffrey C. (1997). Greek: A History of the Language and its Speakers. London & New York: Longman. Iakovou, Maria (2001). “From Deontic to Epistemic Modality: Modals in Modern Greek”. In Christos Clairis (Ed.), Recherches en linguistique grecque, Vol. 1 (pp. 235–238). Paris: L’Harmattan. Jespersen, Otto (1924). The Philosophy of Grammar. London: George Allen & Unwin Ltd.
Anastasios Tsangalidis
Joseph, Brian D. (1981). “On the Synchrony and Diachrony of Modern Greek na”. Byzantine and Modern Greek Studies, 7, 139–154. Joseph, Brian D. (1983). The Synchrony and Diachrony of the Balkan Infinitive: A Study in Areal, General and Historical Linguistics. Cambridge: Cambridge University Press. Joseph, Brian D. (1985). “Complementizers, Particles, and Finiteness in Greek and the Balkans”. Folia Slavica, 7, 390–411. Joseph, Brian D. (1990). Morphology and Universals in Syntactic Change: Evidence from Medieval and Modern Greek. New York & London: Garland Publishing Inc. Joseph, Brian D. (2001). “Is There Such as Thing as ‘Grammaticalization’?” Language Sciences, 23, 163–186. Joseph, Brian D. (this volume). “Rescuing Traditional (Historical) Linguistics from Grammaticalization Theory”. Joseph, Brian D. & Panayiotis Pappas (2002). “On Some Recent Views Concerning the Development of the Greek Future System”. Byzantine and Modern Greek Studies, 26, 247–273. Joseph, Brian D. & Irene Philippaki-Warburton (1987). Modern Greek. London: Croom Helm. Karantzola, Eleni (1995). “Let’s Talk about Concession: The Case of the Modern Greek Particle as”. Journal of Pragmatics, 24, 55–75. Kuteva, Tania (2001). Auxiliation: An Enquiry into the Nature of Grammaticalization. Oxford: Oxford University Press. Lyons, John (1977). Semantics. Cambridge: Cambridge University Press. Meillet, Antoine (1912). “L’evolution des formes grammaticales”. Scientia, 12. Reprinted in A. Meillet (1948), Linguistique historique et linguistique générale, 1 (pp. 130–148). Paris. Newmeyer, Frederick J. (1998). Language Form and Language Function. Cambridge, MA: MIT Press. Nikiforidou, Kiki (1990). Conditional and Concessive Clauses in Modern Greek: A Syntactic and Semantic Description. Ph.D. Dissertation. University of California, Berkeley. Nikiforidou, Kiki (2001). “As: An Example of Syntactic Change”. In Anastasios-Phivos Christidis (Ed.), History of the Greek Language (pp. 1065–1068). Thessaloniki: Centre for the Greek Language & Institute for Modern Greek Studies. Palmer, Frank R. (1986). Mood and Modality. Cambridge: Cambridge University Press. Palmer, Frank R. (2001). Mood and Modality (2nd edition). Cambridge: Cambridge University Press. Pappas, Panayiotis A. (2001). “The Microcosm of a Morphological Change. Variation in thelo: + infinitive Futures and e:thela + infinitive Counterfactuals in Early Modern Greek”. Diachronica, 18, 59–92. Pappas, Panayiotis A. & Brian D. Joseph (2001). “The Development of the Greek Future System: Setting the Record Straight”. In Y. Agouraki et al. (Eds.), Proceedings of the 4th International Conference on Greek Linguistics, Nicosia, September 1999 (pp. 354–359). Thessaloniki: University Studio Press. Philippaki-Warburton, Irene (1994). “The Subjunctive Mood and the Syntactic Status of the Particle na in Modern Greek”. Folia Linguistica, 28, 297–328. Tsangalidis, Anastasios (1999). Will and Tha: A Comparative Study of the Category Future. Thessaloniki: University Studio Press.
Unidirectionality in the grammaticalization of modality in Greek
Tsangalidis, Anastasios (2001). “Types of Modality in English and Greek”. Paper presented at the Conference Modality in Contemporary English, University of Verona, September 6–8, 2001. Tsangalidis, Anastasios (2002). “Homonymy, Polysemy, Category Membership: The Case of Greek Modal Particles”. In Ad Foolen, Ton van der Wouden, & Piet van de Craen (Eds.), Particles, Special Issue of Belgian Journal of Linguistics, 16, 135–150. Tsimpli, Ianthi Maria (1990). “Clause Structure and Word Order in Modern Greek”. In J. Harris (Ed.), UCL Working Papers in Linguistics, 2 (pp. 226–255). van der Auwera, Johan & Vladimir A. Plungian (1998). “Modality’s Semantic Map”. Linguistic Typology, 2, 79–124. Veloudis, Ioannis (2001). “Na (Without Stress) and (Stressed) na”. In Y. Agouraki et al. (Eds.), Proceedings of the 4th International Conference on Greek Linguistics, Nicosia, September 1999 (pp. 243–250). Thessaloniki: University Studio Press.
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.1 (211)
How cognitive is grammaticalization? The history of the Catalan perfet perifràstic Ulrich Detges University of Munich
.
The ‘auxiliation constraint’
It is widely recognized that grammaticalization processes are polygenetic in nature. In many genetically unrelated languages in different parts of the world, grammatical elements of given functional categories derive from lexical source expressions with similar meanings. Thus, a prototypical diachronic source of future markers in many different languages are go (to)-constructions (see e.g. Bybee et al. 1994: 266). This empirical finding has been interpreted as the manifestation of an inherent cognitive affinity between the grammatical functions in question and the respective source concepts. A radical formulation of this view is the so-called “auxiliation constraint” (Heine 1993: 85, see also Kuteva 1995) which stipulates that “for each grammatical category, there is only a small pool of possible source concepts” (Heine 1997: 6). From this perspective, grammaticalization seems to be driven by universal forms of conceptualization (Heine et al. 1991: 27), which, in turn, impose (at least) two kinds of restriction on possible lexical source expressions: a.
According to Kuteva (1995), the source expressions eligible for grammaticalization are the most direct encodings of certain basic conceptual structures, the so called “kinesthetic image-schemas”, which are themselves limited in number. This would explain why verbs meaning “to go”, which are the immediate linguistic expressions of the cognitive schema source-path-goal, are good candidates for grammaticalization whereas verbs meaning “to walk” are not (Kuteva 1995: 379). b. The development of lexical items into grammatical elements reflects “the basic conceptualization principle that abstract notions (such as grammatical verb meanings) are conceptualized [. . . ] in terms of concrete ones” (Kuteva
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.2 (212)
Ulrich Detges
2001: 22). The matching of lexical source expressions to grammatical functions gives rise to further constraints: [I]t is always the goal-oriented movement verb structure that turns into auxiliary structure. Thus, whereas go to + Noun Phrase → go (to)-future and come to + Noun Phrase → come (to)-future are attested as auxiliation paths, no lexical structure go from + Noun Phrase or come from + Noun Phrase has developed into a future construction. (Kuteva 2001: 22)
In the present paper, I wish to show that this position is problematic. In particular, it will be argued that the forces which trigger grammaticalization processes are not primarily cognitive in nature, but discourse-pragmatic. In other words, it is not the source concepts as such which drive processes of this kind, but the fact that the concepts in question are useful for certain very basic discourse-pragmatic strategies.
. The perfet perifràstic of Catalan The empirical issue which will be discussed in this paper is the perfet perifràstic of Catalan, the Romance language spoken (and written) in the eastern part of the Iberian Peninsula. The perfet perifràstic (vaig cantar “I sang”, lit. “go:1sg:pres sing:inf ”), an analytical construction, is the default past tense marker of this language (see (1)). It marks perfective events which are in no special relation to the moment of speech. Thus, when telling a story, speakers of Catalan will use the perfet perifràstic for each of the events of the narrative sequence. (1) The Catalan perfet perifràstic El seu discurs va causar un gran impacte en The his talk go:3sg:pres produce:inf a great effect on l’auditori. the audience “His talk produced a great effect on the audience.” The Catalan perfet perifràstic fits most uncomfortably with the point of view outlined in Section 1 (see Pérez Saldanya 1998: 261), because diachronically it goes back to a goal-oriented go (to)-construction. Hence, proponents of the aforementioned position would have to assume that in this particular case, a go (to)construction was selected to conceptualize pastness. Such a view, however, would contradict the spirit of restriction (b) which leads one to expect that markers of the past should not derive from goal-oriented go (to) + infinitive but rather from source-oriented come from or go from, which is indeed often the case.1
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.3 (213)
How cognitive is grammaticalization?
As will be shown however, the evolution of a goal-oriented go (to)-construction into an aoristic past tense marker in Catalan was not merely an accident: each single step of this evolution is paralleled by similar developments in other languages. More specifically, the history of the perfet perifràstic refutes the idea that the grammaticalization of tense markers is motivated by a desire (or a need) to conceptualize time.
. Step 1: From movement in space to inchoative aspect The origins of the perfet perifràstic go back to the Middle Ages. Its immediate ancestor, a construction with aspectual meaning, was used throughout a dialect continuum which stretched from Anglo-Norman French in the North to the Languedoc in the South East and Catalonia in the South West (Gougenheim 1971 [1929]: 93– 95). Instances of this construction, which expressed an abrupt start, are documented in Old French (see ex. (2)); OFr. (il) va dire meant – as does Modern French il se met à dire – not simply “he starts to say”, but more specifically “all of a sudden, he says” (Gougenheim 1971 [1929]: 96). In the following example, the construction is used in the past tense. (2) Old French Et pour ce, les barons passerent devant Troies [. . . ] et se alerent logier
1
“And for this [reason], the men passed in-front-of Troyes [. . . ] and started [to]
en la praerie d’Isles.
2
set-up-camp in the prairie of Isles”.
(Joinville, cf. Gougenheim 1971 [1929]: 92) How does a go (to)-construction which originally refers to movement in space come to denote inchoative aspect? Consider (3): (3) Mary {moves + goes + jumps}1 (to the doorplace ) to let2 Peter in. If an agent moves to some place (event 1) with the intention of carrying out some action there (event 2), then she is visibly making a gesture which will take her to the beginning of this action. In this event sequence, movement and beginning are in contiguity (see Koch 1999), since movement (= event 1) and event 2 generally overlap at the beginning of the latter (see (4)). (4) Conceptual setting: From movement to beginning
(event 1)
(event 2)
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.4 (214)
Ulrich Detges
Note that the beginning of an action is not abstract (as might be claimed by proponents of the position outlined in Section 1) if the action itself is of a concrete nature. However, movement is an especially salient manifestation of beginning, since under normal conditions, the impulse of the agent who moves towards the action will be easier to perceive than the beginning itself (see Colon 1978b: 157 and Pichon 1933). When the shift from movement to beginning is completed, the go (to)-construction is no longer confined to telic verbs with agentive subjects. Typically, it starts to appear together with atelic verbs, such as ‘to sleep’, ‘to be’, ‘to feel’ etc. For OCat. go + inf. this is the case from the mid-14th century onwards (Colon 1978b: 170). Moreover, as a construction with purely aspectual meaning, it now admits impersonal verbal phrases (for Middle French, see examples in Gougenheim 1971 [1929]: 95). Abundant evidence for parallel developments exists in many languages, i.e. verbs denoting different types of movement which come to express inchoative meaning. (5) Parallel developments: a. Fr. se mettre à faire qc. “to start doing s.th.” < “to sit down to do s.th.” b. It. mettersi a fare qc. “to start doing s.th.” < “to sit down to do s.th.” c. Sp. ponerse a hacer ac. “to start doing s.th.” < “to put o.s. in a position to do s.th.” d. E. to start doing s.th. < ME. sterten “to jump, to leap” To these, one might add Lat. incipere and Germ. anfangen, both going back to lexemes denoting ‘to catch, to grasp’ (see (5e, f)): if an agent grasps some object, he will probably start doing something to this object. As in the cases described above, the gesture of the movement towards some object is a salient manifestation of beginning a manipulation of this object. (5) e. f.
Lat. incipere “to start” < “to grasp, to catch” Germ. anfangen “to start” < “to grasp, to catch”
Considering the different sources listed under (5), we find that we are not dealing with a “small pool of possible sources”. More specifically, concepts such as go, put, sit down, jump, grasp etc. are instances of very different types of source-pathgoal schemas. Consequently, the conceptual setting under (4), which is common to all the individual changes mentioned in (5), is not a “kinesthetic image-schema” in the sense outlined in Section 1. Rather, it characterizes a general conceptualization strategy, which leaves ample room for variation. Hence, at least in the cases under discussion, restriction a) of the ‘auxiliation constraint’ is found to be invalid. Moreover, the inchoative constructions listed under (5), which are the outcome of the cognitive operation sketched in (4), although aspectual in meaning, are
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.5 (215)
How cognitive is grammaticalization?
not grammatical elements. In particular, they do not have auxiliary status, which can be seen from the fact that they readily combine with nominal complements, as in E. Peter started a fight with Paul or Fr. Marie se met au travail “Mary sets herself to work” etc. Finally, unlike OFr. aller + inf. and Cat. anar + inf., the inchoative constructions in (5), i.e. Engl. to start, Germ. anfangen, Fr. se mettre à or Sp. ponerse a do not show the slightest symptoms of turning into past tense auxiliaries. The conceptualization strategy (4) plays no role in the eventual grammaticalization of Cat. anar + inf. and of OFr. aller + inf .
. Step 2: Inchoative constructions as discourse-structuring devices An important property of inchoative constructions becomes apparent if one compares the English sentences under (6). (6) a.
Aspectually unspecified event Bernie hits Paul. b. Inchoative construction denoting an incipient event (Suddenly) Bernie starts hitting Paul.
The event represented in (6a) is aspectually unspecified, whereas in (6b) the same state of affairs is expressed by an inchoative construction and thus represented as an incipient event. Clearly, the second, inchoative version is more dynamic and therefore more appropriate to catch the listener’s attention. From this, one might deduce a general cognitive principle (7): (7) All other things being equal, incipient events are perceived as more dynamic and more spectacular than aspectually unspecified ones. Principle (7) reflects a universal rule of perception which governs the allocation of attention to objects and events. As has been shown by abundant experimental evidence (see, among many others, Yantis 1993; Remington, Johnston, & Yantis 1992), sudden changes in situation are especially salient for human perception (and, in fact, for most living beings). Principle (7) explains a striking particularity of our go-construction in Old French and Old Catalan. The full verb most frequently accompanied by it during the Middle Ages in all the dialects mentioned is the lexeme ferir “to strike, to hit” (for OFr. Gougenheim 1971 [1929]: 93; for OCat. Pérez Saldanya 1998: 264). A typical context is given under (8), taken from the OFr. Song of Roland. (8) Inchoative go + ferir “to strike” in OFr.
(Song of Roland, 3567–3568)
En mi le camp amdui s’entr’encuntrerent
1
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.6 (216)
Ulrich Detges
“In [the] middle [of] the field the-two-of-them met Si se vunt ferir, granz colps s’entredunerent and they-start-to fight, hard blows [they] dealt-to-one-another”.
2
In this example, vunt ferir “they start to strike”, “they start to fight” describes a highly dramatic event. In the Song of Roland there are 31 such instances of aller ferir (see Gougenheim 1971 [1929]: 93). In combinations with ferir “to strike”, “to fight” the go-construction functions as a marker of suspense: it is a rhetorical device used to represent especially dramatic points of narrative sequences (see Colon 1978a: 125). The same observation applies to Old Catalan (see especially Colon 1978b: 144). (9) contains some of the most frequent verb combinations of OCat. anar + inf. as given in Colon (1978a). When used in narrative sequences, these verbs typically denote dramatic events. (9) OCat.: list of most frequent verbs (Colon 1978a: 130) abraçar “to embrace”, besar “to kiss”, brocar “to give a horse the spurs, to ride”, ferir “to strike”, gitar “to throw, to shoot (an arrow)”, trencar “to break, to smash”. Use of inchoative linguistic representations for non-inchoative events is a rhetorical technique for which parallel evidence exists from other languages. Latin had an inchoative verbal suffix -sc- which could freely be added to aspectually unspecified verbs and which modified their meaning to inchoative. Surprisingly, most of the Romance verbs which go back to an inchoative sc-form of the corresponding Latin verb have an aspectually unspecified meaning. A case in point is Fr. finir “to finish” with the 1st person plural nous finissons “we finish”, whose origin is not Lat. finimus “we finish” but inchoative *fini-sc-imus “we start to finish” (see (10)). An explanation for semantic change from inchoative to unspecified is that many of the inchoative sc-forms were regularly ‘over-used’ in spoken Latin to refer to events which in reality were not incipient and thus did not really justify the use of the inchoative form. A possible motivation for this ‘over-use’ is provided by the cognitive principle (7): if incipient events are perceived as more relevant than aspectually unspecified ones, then linguistic representations of incipient events, i.e. inchoative forms, will suggest communicative relevance. (10) Inchoative -sc- in Vulgar Latin (Rheinfelder 1967: 190) a. VLat. *fini-sc-imus “we start to finish” (from finimus “we finish”) > “we finish” (> OFr. fenissons “we finish”) b. VLat. *flore-sc-et “it starts to blossom” (from floret “it blossoms”) > “it blossoms” (> Sp. florece “it blossoms”) In the examples discussed under (10), aspectually marked verb forms (*fini-scimus, *flore-sc-et) are reinterpreted as aspectually unspecified. Parallel to this se-
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.7 (217)
How cognitive is grammaticalization?
mantic reinterpretation, the former inchoative suffix -sc-, now devoid of its original aspectual content, is morphologically reanalyzed as being part of the verbal stem. In a further step, some of the reinterpreted/reanalyzed sc-forms come to be treated as the normal forms of the respective verb, i.e. after coexisting for some time with the simple forms of the verb (e.g. finimus, floret), they finally replace them. Importantly, each of these changes is applied to individual verbs only. This type of change, which is lexical in nature, has nothing to do with grammaticalization. This means that the assumed rhetorical ‘dynamification’ of non-inchoative events is in itself not a sufficient motivation for grammaticalization – something else had to happen to turn the inchoative go-constructions of Old French and Old Catalan into grammatical markers. The factor which proved to be decisive for the further development of our go-constructions was that they were conventionalized as procedures to structure narrative texts. Within narrative sequences of discrete foreground events, go + inf. came to be used to foreground the ‘turning-point’ event, i.e. the most relevant single event or the first element of a sub-series of particularly surprising or noteworthy actions (see Colon 1978a: 126–127). This technique exploits the ‘suspense effect’ (7) described above. A translation for OCat. van trencar (les portes) in (11) which renders this effect would be “suddenly, they broke (the gates)”. (11) Old Catalan
(Desclot, IV, 13th c., cf. Colon 1978b: 135–136)
E aquells del rey de Franssa, que viren1 açò, cuydaren2 -sa que ·ll rey
1
“And when those [i.e. the men] of-the King of France saw1 this, they-thought2 the King
d’ Aragó fos lahins e no·u tengueren3 a festa, e giraren4 les testes als
2
of Aragon was in-there and they were3 not too happy about it, and they-turned4 their
cavalls e tornaren5 -se’n d’esperó vers la ost del rey de Franssa.
3
horses’ heads and returned5 quickly towards the army of-the king of France.
E al tornar, passaren6 per ·I· monestir de dones monges de la orda de
4
And on [their] return, they-stopped6 by a convent of lady nuns of the Cistercian Order,
Sistell, qui era fora la vila de Perpinià, e van trencar7 les portes del
5
which was outside the city of Perpignan, and [then suddenly] go:3pl:pres break7
monestir e barrajaren8 e robaren9 la sglésia e totes quantes coses hi hach,
6
the convent’s gates and plundered8 and looted9 the church and every thing there was,
que hanch no·y lexaren10 staca que tot no se’n ho
7
so-that they did not leave10 there [even one] post any-more which
aportassen ab ssi.
8
they-wouldn’t-have-taken with them.”
In Middle French, the construction is often found to specify verba dicendi, e.g. il va dire “suddenly, he said” (Gougenheim 1971 [1929]: 96–97). This usage of go + inf., which has been interpreted as evidence for a quotative or even evidential function
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.8 (218)
Ulrich Detges
(Steinkrüger 1999: 225), can be explained more adequately by the present account: unexpected speech acts are typical ‘turning-point’ events of narrative sequences. Note that the textual function of go + inf. as exemplified in (11) is not confined to written narratives. 16th-century grammarians in France, in the Languedoc and in Catalonia criticize these constructions as vulgar and uneducated (Colon 1978b: 146–147). This strongly suggests that their use in literary discourse reflects an oral technique of story-telling (see below, Section 6). As an ‘extra-heavy’ foregrounding device, the go-construction is used only sparingly, at least in elaborated literary discourse. As shown in (11), it normally appears no more than once per narrative episode (Colon 1978b: 169). Another detail which can be observed in (11) is its use in the present tense, even if the events preceding and following it are coded as past tense. In its present-tense realization, the construction combines two effects, namely (a) the already mentioned ‘dynamification effect’ (7), which is metonymic in nature, and (b) the historical present (Gougenheim 1971 [1929]: 97), i.e. a metaphorical extension of the actuality-effect which is normally associated with the present tense (see Weinrich 6 2001 [1964]: 38, 191). In the narrative genres of medieval Western Romance, the present tense is commonly used in fairly unsystematic alternation with past tenses, especially in earlier texts. In our go-construction, the present tense is originally optional, as can be seen in (12), where OCat. anar appears in its past tense form (anaren). (12) OCat. anar + inf. in past tense form (Desclot, II, cf. Colon 1978b: 159) Sobr·assò vengren-li bé ·X· cavalers justats e 1 “There-upon about ten knights in-armour came-up-with him and [then suddenly-
anaren-lo ferir, si que·l abateren a terra e aquí murí.
2
go:3pl:past hit him, and they-struck him down to the-ground and there he-died.”
The existence of such cases (for OFr. see also (2), many similar examples for Middle French are presented by Gougenheim 1971 [1929]) shows that the praesens historicum is dispensable for the construction’s foregrounding effect. However, the present-tense variant, i.e. the ‘stronger’ alternative, was used more frequently at all times; consequently, it was this variant which finally came to be conventionalized as the only possible form of the foreground-marker. This process, which is an instance of the gradual loss of the construction’s paradigmatic variability as described in Lehmann’s (1995 [1982]) parameters, is a direct consequence of its progressive routinization. Once the inchoative construction has become a conventional procedure with a clear textual function, we can say that it has turned into a grammatical item (see, e.g., Pérez Saldanya 1998: 266). Now, what precisely is cognitive about this process? It has, of course, a cognitive background, namely the principle (7), according to
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.9 (219)
How cognitive is grammaticalization?
which incipient events are perceived as more spectacular than aspectually unspecified events. However, as we have seen in this section, this principle is in itself not sufficient to explain the grammaticalization process under discussion. It gives rise to a grammatical procedure only because it is exploited for a discourse strategy of the following type (13): (13) Discourse strategy: When telling a story, represent the ‘turning-point’ event of an episode as incipient, because this will help to catch your listener’s attention. Discourse strategy (13) is not a linguistic rule. Rather, it is a technique of efficient speaking which, as such, is not specific to particular languages or to particular historical periods. The only linguistic prerequisite for (13) is the existence of a construction meaning ‘to start doing something’ in the respective language. Cornips (2000) reports that jokes told by immigrant children of Moroccan and Turkish origin in present-day Dutch contain an extremely high rate of gaan “go” + inf.2 In Middle English, inchoative gan “begin:past” + inf. served, among other things, to give prominence to dramatic narrative events (see Fischer 1992: 265–267). Having analyzed the parallel developments of inchoative Lat. -sc- (see (10)) and that of OFr./OCat. go + inf. (see (11)) in some detail, we are now a in position to account for the difference between lexical change and grammaticalization. As follows from our argument, this difference cannot be defined on purely cognitive grounds, since in both of the examples mentioned, the same cognitive principle (rule (7)) is implied. In the case of Lat. -sc-, (7) is applied to individual noninchoative concepts. However, with the go-constructions, principle (7) yields a technique of structuring narrative sequences (namely (13)), which in turn gives rise to grammaticalization. In principle, narrative texts consist of unspecific events – any kind of event can be part of a narrative sequence. In particular, the question of which event will be considered the ‘turning point’ of an episode depends largely on the story which the speaker wants to tell, not on the event itself. Hence, as a foregrounding technique, the go-construction is not confined to certain individual concepts – it can be applied to any kind of event.3 Moreover, as a foregrounding routine, go + inf. is used with significant frequency. These two qualities – high frequency and independence from particular lexical contexts – distinguish grammaticalization from lexicalization.
. Digression: Grammaticalization and areal convergence So far, I have drawn no explicit difference between the various linguistic realizations of go + inf. in Catalan, in Occitan and in French respectively. This is,
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.10 (220)
Ulrich Detges
of course, a simplification, since each of the varieties concerned used its specific go-construction. Thus, we find aller + inf. in French but anar + inf. in Catalan. What is spread (or borrowed) across the different neighbouring varieties of Romance between the Isle de France, the Languedoc and Catalonia is not a (languagespecific) grammatical construction, but principally the narrative strategy (13), i.e. a cultural technique which is essentially extra-linguistic in nature. This technique has, however, direct linguistic consequences, if all the varieties concerned use a similar type of linguistic expression for the implementation of (13), namely their own (language-specific) inchoative go-construction. When these originally free constructions turn into conventional grammatical markers, all of the varieties in question see the rise of highly similar grammatical procedures. This scenario describes the interplay sometimes observed between grammaticalization and language- or dialect contact (see Bisang 1998). In particular, it provides an explanation for areal convergence and Sprachbund-phenomena (Haspelmath 2001; van der Auwera 1998), i.e. partially parallel internal developments in adjacent languages and dialects.4 However, areal convergence must not be confused with the polygenetic change mentioned in Section 1, i.e. parallel, but independent change in languages of different historical periods and in different parts of the world. This feature, which is often found in grammaticalization processes, will be discussed in Section 7.
. Step 3: From ‘extra-heavy’ foregrounding to aoristic tense In Section 4, I explained how go-constructions came to be used in Middle French, Medieval Occitan and Old Catalan as devices to foreground ‘turning-point’ events in narrative sequences. Even though these constructions had a grammatical function, they were much less frequent in use than tense markers, since they only foregrounded the most important event in the sequence. They were marked grammatical constructions, used for a ‘special effect’ only (see Pérez Saldanya 1998: 271). In French and Occitan, the respective go-construction never developed further than this. In Occitan it became obsolete in the 16th century (Meyer-Lübke 1925: 106). In French it was abandoned during the first half of the 17th century (Gougenheim 1971 [1929]: 95–96). In Catalan however, anar + inf. supplanted the old synthetic past simple and became the default marker for the aoristic past tense. Arguably, this process was favoured by the absence of a written standard (Colon 1978b: 172; Steinkrüger 1999: 223). In the 16th century, Catalan was ousted as an official written language by Castilian Spanish, and when a new Catalan standard was promoted by the nationalist renaixença movement in the 19th century, the grammaticalization of anar + inf. was already completed.5
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.11 (221)
How cognitive is grammaticalization?
Recall that in Medieval Catalan, anar + inf. was, as a rule, only used once in the same sequence of events (see Section 4). This applied, above all, to narratives written in an elaborate literary style (for Old Provençal see Meyer 1865: LXVI). However, in Ramon Muntaner’s Crònica, a late-medieval document representative of the spoken popular Catalan of its time (Sobré 1978: 43), this restriction is violated (see (14)): (14) Inflationary ‘over-use’ of anar + inf. (Muntaner, ∼ 1330, cf. Colon 1978b: 169) E tantost que açò hac fet, el cavall se sentí1 ferit, e llevà’s2
1
“As soon as she had done this, the horse felt1 [it had been] beaten, and it-reared2 up for-
davant e detràs aixi que fóra caüt si no fos que era ab cadena fermat
2
wards and backwards, so that he-would-have fallen if he-hadn’t-been attached by a chain
en la sella. Què us diré ? Ella va metre3 la man a l’espaa, e venc4 a
3
to the saddle. What shall-I tell you? She suddenly-put 3 her hand to the sword, and came4 to
un portell altre, e anà ferir5 lo cavall per la testera; e el cavall estec6
4
another breach, and suddenly-she-hit 5 the horse over the headstall; and the horse
estabornit. Què us diré ? Lo cavall va pendre7 per la regna e cridà8 :
5
fainted6 . What shall-I-tell you? She-suddenly-took7 the horse by the bridle and shouted8 :
“Cavaller, mort sóts, si no us retrets”. E el cavaller tenc9 -se per mort,
6
‘Sir, you are dead if you don’t retreat.’ And the knight thought9 he-was [going to be] dead,
e va tenir10 lo bordó e va’l gitar11 en terra, e retré’s12 a ella;
7
and suddenly-he-took10 the spear and suddenly-he-dropped11 it and surrendered12 to her;
et ella pres13 lo bordó e trasc14 -li la llança de la cuixa, e així mès15 -lo
8
and she took13 the spear and detached14 the lance from his thigh, and in-this-state
dins Peralada.
9
she-sent15 him [back] into Peralada.”
This example (discussed in Colon 1978b: 169) documents some simple rhetorical procedures, typical of oral story-telling, e.g. qué us diré? “what shall I tell you?” (lines 3, 5). In accordance with this general attitude, it shows a heavy ‘overdose’ of go-constructions. Instead of only one instance we find five, two of them directly following one another (line 7). An obvious motivation for this is to make the narrative appear even more dramatic: each of the five events in question is presented as a ‘turning point’, even though four of them do not qualify for this status. The eventual grammaticalization of anar + inf. as an unmarked anterior was due to a massive extension of this rhetorically motivated over-use. Application of this technique to an ever-growing portion of the narrative event-chain will trigger a process of rhetorical devaluation (Dahl 1998). The more of the information conveyed is marked as surprising and noteworthy, the less relevance will be attributed to each individual piece of information. On the linguistic construction employed, over-use
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.12 (222)
Ulrich Detges
of this kind will have two effects: on the one hand, the construction’s frequency will be seen to rise dramatically; on the other hand, it will lose its ‘special-effect’ value. In doing so, it will turn into an unmarked default item. Steinkrüger (1999), who has analyzed written documents of private character (diaries, private letters etc.) dating from the 17th and 18th centuries, shows that the use of anar + inf. is mainly motivated by the ‘foregrounding’ function discussed above. Thus, in the diary of Miquel Parets, a chronicler of mid 17th century everyday life in Barcelona (see Amelang 1991), go-constructions are particularly frequent in the passages which relate the outbreak of the plague in 1651. Within these passages, anar + inf. is by far the most frequent in the parts where the author reports the death of his own family (Steinkrüger 1999: 229). This draws attention to another important point: speakers use the go-construction in order to highlight their involvement in the events related (Steinkrüger 1999: 230). Thus, when in the course of the 18th century, the construction’s relative frequency rises dramatically in the different parts of Catalonia, the available documents – which, in the absence of a written standard, are particularly faithful to oral usage – show that it is the first person singular which leads the development. This suggests that the ultimate grammaticalization of anar + inf. is triggered by speakers who give prominence to their own past actions.
. Digression: The ‘hot-news’ perfect in contemporary Peninsular Spanish The process I have hypothesized for Catalan has parallels in other languages. Just like the English present perfect, the modern standard Spanish perfecto compuesto is a marked construction, used to refer to past events with current relevance. Among other things, use of both tenses is excluded for coding narrative events (15a, b). By contrast, the analytical perfects of French and German can be used in exactly such contexts (15c, d). (15) Two types of analytical perfects a. E. *First, he has opened the door, then he has seen Mary, and then . . . b. S. *Primero, ha abierto la puerta, entonces ha visto a María, y después ... c. F. D’abord, il a ouvert la porte, ensuite, il a vu Marie, et alors . . . d. G. Zuerst hat er die Tür geöffnet, dann hat er Maria gesehen, und dann . . . For the French passé composé, the testimony of mid-16th century grammarians demonstrates that it did have current relevance, but that it lost this feature (Berschin, Felixberger, & Goebl 1978: 151–153; see also Harris 1982). A mecha-
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.13 (223)
How cognitive is grammaticalization?
nism which brings about change of this kind is the so-called ‘hot-news’ strategy (Schwenter 1994). In (16a), taken from a radio news broadcast in present day English, the present perfect makes the news appear especially fresh and hence more relevant to the listener. In English, the ‘hot-news’ perfect is a rhetorical technique, specific to journalistic style. Note that it is only the news headline which is coded in the present perfect – here, the restriction is respected according to which this tense must not be used to code the narrative event sequences themselves. As is shown in (16b), the same technique can be found in European Spanish (although less frequently than in English), and, at least in journalistic style, use of the perfecto compuesto displays the same restrictions as does the present perfect in (16a). (16) a. The ‘hot-news’ technique in written English Rock Musician Frank Zappa has died. A family spokesman reported that the entertainer passed away at his home Saturday after a long bout with colon cancer. (cf. Schwenter 1994: 1003) b. The ‘hot-news’ technique in Spanish journalistic style Ford ha1 fichado al doble campeón mundial Carlos Sainz para pilotar el 1 “Ford has1 engaged the double world champion Carlos Sainz for driving the
segundo Focus WRC en el campeonato del mundo de rallys las dos
2
second Focus WRC in the Formula 1 world championship [for] the next
próximas temporadas, informó2 la marca automovilística mediante un
3
two seasons, [as] the car brand informed2 in an
comunicado. announcement.”
4 (Agencia EFE, 4 nov. 1999)
In spoken Peninsular Spanish, however, speakers do not respect the restrictions observed in (16a, b). Example (17) documents the ‘hot-news’ perfect in spoken discourse: (17) Spoken Peninsular Spanish (Corpus oral 1992) Y entonces me dice1 . . . él: “Pues no sé”, la conferencia la tiene2 que dar 1 “And then he tells1 me: ‘Well, I-don’t know’, the lecture, he-has to give2 it
en. . . el centro de a las 7; hemos acabado3 de comer a las 5 2 in . . . the centre of at 7; we-have finished3 to eat at 5,
y. . . no ha comido4 casi nada, y entonces le he preguntado5 : “Oye, José Luis, 3 and . . . he-has eaten4 almost nothing and then I-*have asked5 him: ‘Listen, José Luis,
que no. . . no has comido”. Y entonces me ha dicho6 . . . me ha dicho7 : “No, 4 you-haven’t eaten’. And then he-*has told6 me . . . he-*has told7 me: ‘No,
no, si yo, comer, como muy poco;lo que hago es beber muchísimo”. . 5 no, yes I, eat, I-eat very little; what I-do is drink a lot”.
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.14 (224)
Ulrich Detges
The speaker is telling a funny story. In the event sequence preceding the story’s point, he strongly over-uses the perfecto compuesto. In the English translation, this is marked by the asterisks. From a normative point of view, rhetorical usage of this kind is clearly ‘false’, but in colloquial style it is accepted, because it has a conventional function. It is easy to see that spread of this technique could eventually make the perfecto compuesto lose its current relevance and turn it into an unmarked past tense marker with aoristic value. The comparison between English and Spanish in (16a, b) provides an indication of how polygenetic development may occur (see Section 1). The desire to make one’s information appear as relevant as possible is a strong pragmatic, i.e. extralinguistic, motivation (Grice 1975: 46). It leads to the elaboration of languageunspecific discourse strategies which can be expected to be found around the world. If, for the implementation of such a technique, two languages recruit linguistic items with similar values, then both items may well undergo a parallel evolution independently of each other. Let us compare the evolution of the Spanish perfecto compuesto to the case of the perfet perifràstic of Catalan. As source constructions for grammaticalization processes, both clearly have very different values: on the one hand, the perfet perifràstic of pre-18th-century Catalan is a foreground marker for narrative sequences, i.e. a construction which is already aoristic – on the other hand, the Spanish perfecto compuesto is a construction with current relevance and therefore banned from usage as an aoristic tense in narratives. From a strictly grammatical viewpoint, both constructions have little in common. But they share a feature which makes them eligible for the ‘hot-news’ strategy: they are marked constructions, designed for a ‘special effect’, and in both cases this effect is associated with communicative relevance. Thus, it is not the value of the source constructions as such which makes them eligible for grammaticalization, but the fact that these constructions prove to be useful for very basic communicative strategies. As we have seen in Sections 3, 4 and 6, the function of anar + inf. in presentday Catalan is the result of three discrete evolutionary steps, each of which was triggered by an independent motivation. However, none of these steps was motivated by a desire or a need to conceptualize time (or tense) – the grammatical function of the perfet perifràstic in modern Catalan is the unintended by-product of discourse techniques which aimed at rhetorical efficiency. This means that, at least in the case discussed in this article, the basic assumption underlying the idea of the ‘auxiliation constraint’ is unsustainable.
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.15 (225)
How cognitive is grammaticalization?
Notes . See Bybee et al. (1994: 56). A case in point is French venir de as in je viens d’arriver, “I just arrived”. . Note, however, that Cornips’s explanation of the facts differs substantially from mine. . However, certain literary genres favor certain concepts as ‘turning-point’ events. Thus, the verbs under (9) are the linguistic expressions of events which are ‘turning points’ typical of medieval epic poems and chronicles. Nevertheless, this does not exclude other verbs from ‘turning-point’ positions. . A Sprachbund will arise if the free discourse technique in question is spread across neighbouring dialects which are not genetically related. . Apart from the county of València.
References Amelang, James S. (Ed.). (1991). A journal of the plague year: the diary of the Barcelona tanner Miquel Parets, 1651. New York: Oxford University Press. Auwera, Johan van der (1998). “Phasal Adverbials in the Languages of Europe”. In Johan van der Auwera & Donall P. O Baoill (Eds.), Adverbial Constructions in the Languages of Europe (pp. 25–145). New York, NY: Mouton de Gruyter. Badía Margarit, Antonio (1951). Gramática Histórica Catalana. Barcelona: Noguer. Berchem, Theodor (1968). “Considérations sur le parfait périphrastique vado + infinitif en catalan et gallo-roman”. In Antonio Quilis (Ed.), Actas del XI Congreso de Lingüística y Filología Románicas (pp. 1159–1170). Madrid: R.F.E. Berschin, Helmut, Josef Felixberger, & Hans Goebl (1978). Französische Sprachgeschichte. München: Hueber. Bisang, Walter (1998). “Grammaticalization and Language Contact, Constructions and Positions”. In Anna Giacalone Ramat & Paul J. Hopper (Eds.), The Limits of Grammaticalization (pp. 13–58). Amsterdam & Philadelphia: Benjamins. Bybee, Joan L., Revere Perkins, & William Pagliuca (1994). The Evolution of Grammar. Chicago & London: University of Chicago Press. Colon, Germà (1978a). “El perfet perifràstic català ‘va + infinitiu”’. In Germà Colon (Ed.), La Llengua Catalana en els seus textos, Vol. II (pp. 119–130). Barcelona: Curial. Colon, Germà (1978b). “Sobre el perfet perifràstic ‘vado + infinitiu’, en català, en provençal i en francès”. In Germà Colon (Ed.), La Llengua Catalana en els seus textos, Vol. II (pp. 131–174). Barcelona: Curial. Cornips, Leonie (2000). “The Use of gaan + infinitive in Narratives of Older Bilingual Children of Moroccan and Turkish Descent”. In Helen de Hoop & Ton van der Wouden (Eds.), Linguistics in The Netherlands 2000 (pp. 57–67). Amsterdam & Philadelphia: Benjamins.
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.16 (226)
Ulrich Detges
Dahl, Östen (1998). “Grammaticalization and the Life-Cycle of Constructions”. Keynote lecture at the 17th Scandinavian Conference of Linguistics, August 20–22, 1998. (Paper presented at the Linguistischen Kolloquium of the Seminar für Sprachwissenschaft in Tübingen, Mai 3rd 1999). Fischer, Olga (1992). “Syntax”. In Richard M. Hogg (Ed.), The Cambridge History of the English Language, Vol. II (pp. 207–408). Cambridge: Cambridge University Press. Gougenheim, Georges (1971 [1929]). Étude sur les périphrases verbales de la langue française. Paris: Nizet. Grice, Herbert P. (1975). “Logic and Conversation”. In Peter Cole & Jerry L. Morgan (Eds.), Speech Acts (pp. 41–58). New York: Academic Press. Harris, Martin (1982). “The ‘Past Simple’ and the ‘Present Perfect’ in Romance”. In Nigel Vincent & Martin Harris (Eds.), Studies in the Romance Verb. Essays offered to Joe Cremona on the occasion of his 60th birthday (pp. 42–70). London & Canberra: Croom Helm. Haspelmath, Martin (2001). “The European Linguistic Area: Standard Average European”. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher, & Wolfgang Raible (Eds.), Language Typology and Language Universals, Vol. 2 (pp. 1492–1510). Berlin: de Gruyter. Heine, Bernd (1993). Auxiliaries. Cognitive Forces and Grammaticalization. New York & Oxford: Oxford University Press. Heine, Bernd (1997). Possession. Cognitive Sources, Forces and Grammaticalization. (= Cambridge studies in linguistics, 83.) Cambridge: Cambridge University Press. Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer (1991). Grammaticalization. A Conceptual Framework. Chicago & London: University of Chicago Press. Koch, Peter (1999). “Frame and Contiguity. On the Cognitive Bases of Metonymy and Certain Types of Word Formation”. In Klaus-Uwe Panther & Günter Radden (Eds.), Metonymy in Language and Thought (pp. 139–167). Amsterdam & Philadelphia: Benjamins. Kuteva, Tania A. (1995). “The Auxiliation Constraint and Reference”. In Richard Geiger (Ed.), Reference in Multidisciplinary Perspective: Philosophical Object, Cognitive Subject, Intersubjective Process (pp. 374–386). Hildesheim: Olms. Kuteva, Tania A. (2001). Auxiliation: an Enquiry into the Nature of Grammaticalization. Oxford: Oxford University Press. Lehmann, Christian (1995 [1982]). Thoughts on Grammaticalization. (= Lincom studies in theoretical linguistics, 1.) München: Lincom. Meyer, Paul (Ed.). (1895). Guillaume de la Barre. Roman d’aventures par Arnaut Vidal de Castelnaudari. Paris: Firmon Didot. Meyer-Lübke, Wilhelm (1925). Das Katalanische. Seine Stellung zum Spanischen und Provenzalischen. Heidelberg: Winter. Pérez Saldanya, Manuel (1998). Del Llatí al Català. Morfosintaxi verbal històrica. Valencia: Universitat de Valencia. Pichon, Jacques (1933). “De l’accession du verbe aller à l’auxiliarité en français”. Revue de Philologie Française, 45, 65–108. Remington, Roger W., James C. Johnston, & Steven Yantis (1992). “Involuntary Attentional Capture by Abrupt Onsets”. Perception & Psychophysics, 51, 279–290. Rheinfelder, Hans (1967). Altfranzösische Grammatik, Vol. 2. München: Hueber.
TSL[v.20020404] Prn:25/02/2004; 14:17
F: TSL5910.tex / p.17 (227)
How cognitive is grammaticalization?
Schwenter, Scott A. (1994). “Hot News and the Grammaticalization of Perfects”. Linguistics, 32, 995–1028. Sobré, Josep Miquel (1978). L’ èpica de la realitat: l’escriptura de Ramon Muntaner i Bernat Desclot. (= Biblioteca Torres Amat, 5.) Barcelona: Curial Ed. Catalanes. Steinkrüger, Patrick (1999). “Das katalanische perfet perifràstic – ein ehemaliger Evidential?” In Rolf Kailuweit & Hans-Ingo Radatz (Eds.), Katalanisch: Sprachwissenschaft und Sprachkultur (pp. 219–235). Frankfurt a. M.: Vervuert. Weinrich, Harald (6 2001 [1964]). Tempus: besprochene und erzählte Welt. München: Beck. Yantis, Steven (1993). “Stimulus-Driven Attentional Capture and Attentional Control Settings.” Journal of Experimental Psychology: Human Perception and Performance, 19, 676–681.
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.1 (229)
Perfect and resultative constructions in spoken and non-standard English Jim Miller University of Auckland, New Zealand
.
Introduction
This paper rests on the assumption that grammaticalisation processes reside in particular varieties (of a given ‘language’) and not in a ‘language’ tout court. This assumption is implicit in Brown and Miller (1982) and Miller (1993), which argue that the grammar of Scottish English is systematically different from that of standard English English, and it is explicit in, e.g., Harris (1984), Winford (1994), Filppula (1999) and Tagliamonte (2000). It is implicit in Elsness (1997), but Elsness compares two standard varieties of English, British and American, rather than standard and non-standard varieties.1 A particular grammaticalisation may take place in one variety but not in another – that is, different groups of speakers have different linguistic behaviours; alternatively, the results of a particular grammaticalisation may become firmly fixed in one variety but in another variety be affected by later changes; and the original construction that was grammaticalised may merely survive in one variety but survive and flourish in another. This paper examines these general processes as instantiated in the possessiveresultative construction – I have the letter written – which developed into the Perfect with a more general and abstract meaning – I have written the letter. The possessive-resultative construction now also permits have got, which emphasises the possession component of its meaning. It has been joined by other constructions based on the resultative participle; these constructions are not readily accepted by all users of standard English English but they occur in non-standard English throughout Britain and are prominent in the Scottish English data on which this paper is mainly (but not exclusively) based. The possessive-resultative construction has not just persisted but acquired a property which sets it apart from the Perfect,
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.2 (230)
Jim Miller
and the resultative participle has not just persisted but become the keystone of various resultative structures. The persistence of the possessive-resultative structure and the development of other resultative structures based on the resultative participle suggest that speakers require constructions that enable them to focus on the results of past actions. Resultative constructions are not peculiar to English, witness the vast amount of data from many language families in Nedjalkov (1988). Similarly, speakers appear to require constructions that enable them to present a single event as on-going, witness the ‘Progressive’ constructions throughout Germanic and Romance languages – see Ebert (2000) and Bertinetto (2000) – and in non-Indo-European languages such as Finnish and Turkish. Talk of speakers’ requirements is speculation, but there is scope for a cross-language investigation into grammaticalisation, what source structures persist and diversify, what source structures grammaticalise and are absorbed into the new structure and what grammaticalised structures are replaced by new structures. What of the English Perfect? The generally accepted view is that the possessiveresultative structure grammaticalised into a prototypical Perfect, at least in formal written English.2 The source of, e.g. I have cooked the meal is I have the meal cooked. The latter can be seen as conflation of two clauses, I have the meal and The meal is cooked, which correspond to two propositions HAVE (I, MEAL) and COOKED (MEAL). In I have the meal cooked the resultative participle cooked applies to meal. In contrast, I have cooked the meal is generally taken as a single clause corresponding to a single proposition COOK (I, MEAL) to which the Perfect operator applies. The classic analysis (as in, e.g., Dahl 1999) is of one structure with four major interpretations, exemplified in (1a–d). (1) a. b. c. d.
Have you ever drunk rakija? I have written up my thesis The Minister has (just) arrived I’ve been waiting for an hour
experiential/indefinite anterior resultative hot-news/recent past extended-now/persistent situation
The Perfect is described as excluding past time adverbs referring to specific times, as in (2). (2) *I have discussed this yesterday/last year/five minutes ago In formal written English the Perfect construction is solidly fixed, in frequent use and protected by grammars of standard English and by editorial practice. In nonstandard English and in spontaneous spoken English (standard and non-standard) the Perfect is not so central. Not only does it face competition from the Simple Past (see §4 below) and the resultative constructions alluded to above (see §§4.3 and 6 below) but there is strong reason to suppose that the classic Perfect construction
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.3 (231)
Perfect and resultative constructions in spoken and non-standard English
is splitting into different constructions distinguished by adverbs, as discussed in §5. If this interpretation of the data turns out to be correct, this process can be thought of as anti-grammaticalisation; the possessive-resultative construction with relatively specific meaning evolves into the Perfect with a more general and more abstract interpretation; the Perfect in turn splits into different constructions, each with a more specific meaning than the Perfect. The above reference to spontaneous spoken English is important. It is gradually being recognised that, although spontaneous spoken English overlaps with formal written English, each type has many grammatical structures and discourse organising devices peculiar to it. These differences are paralleled in other languages, see the references in Miller and Weinert (1998). Just as constructions can arise or be preserved in standard but not in non-standard varieties and vice-versa, so can they arise or be preserved in spontaneous spoken language but not in written language, and vice-versa.3 Some of the resultative constructions presented in §§4.3 and 6 occur regularly in spontaneous speech but irregularly or not at all in writing. In the preceding paragraph the classic Perfect was said to be splitting into different constructions distinguished by adverbs. When the distinction between formal written English and spontaneous spoken English is taken into account, another interpretation presents itself: that the classic adverb-less Perfect exists only in formal written English and that the different Perfect constructions distinguished by adverbs have always existed in spoken English.
. The data We begin with a comment on notation. Following Dahl (1985), labels for aspects and tenses in a particular language have an initial capital letter: e.g. the English Perfect, the English Simple Past. PERFECT is the label for the cross-language category. Following Lyons (1968), lexemes (lexical items) are written in capital letters and the various forms of a given lexeme are in italics: thus, HAVE is realised by the forms has, have, had. The analyses presented here rest mainly on transcripts of spoken Scottish English amounting to 71,600 words. The transcripts are as follows. A
6,000 words of spontaneous conversation between a field worker and two eighteen-year old female students at a state school in East Lothian. This chunk is taken from the 250,000 word corpus of Scottish English conversation (the Edinburgh Corpus of Spoken Scottish English or ECOSSE), collected, transcribed and put into digital form for Brown and Miller’s project on the syntax of Scottish English (1977–1980). For further details see Brown and Miller (1982).
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.4 (232)
Jim Miller
B
12,500 words of spontaneous conversation from the above corpus but between a field worker and two eighteen-year old male students at a private school in Edinburgh. C 8,400 words of conversation recorded in 1998 between a research student and four fourteen-year old female students at a state school in Edinburgh. D 10,000 words of spontaneous conversation between the same research student and the mother of one of the students recorded in the mother’s home in 1998. E 18,900 words of spontaneous conversation between the same research student and a teacher in the state school, recorded in 1998. F 15,800 words of spontaneous speech (business discussions) recorded on the premises of a Glasgow firm in 1996 for a group of researchers investigating the discourse structures of business discussions. To gain a historical dimension, albeit exiguous, the dialogues in the first half of John Galt’s novel The Entail (Volume One of the two-volume edition published in Edinburgh by John Grant in 1936) were examined for occurrences of the Perfect, what meaning each one had (Recent Past, Experiential, etc.) and whether they were modified by an adverb of time. Galt published The Entail in 1823 and the characters use the Scots and Scottish English heard and used by Galt as he was growing up in Ayrshire from 1779 to 1800. Dialogue in a novel is not the same as naturally occurring dialogue, but Galt is recognised as a master of Scots dialogue. The project in (A) above focused on the syntax of Scottish English but both Scottish English and English English examples were collected informally as they occurred in conversation, on radio, on television or in printed sources and this practice has been continued by the author. When used in this paper, such examples will be labelled as informal. Although informal, they are reliable, since the general construction types have been cross-checked with other investigators of Scottish English and Scots and with grammars of English English. It must be emphasised that the Scottish English syntax project was not primarily sociolinguistic. Its aim was to find out what syntactic constructions were used by speakers of Scottish English, to analyse their structure and their use in discourse, and to establish a preliminary set of grammatical differences between standard Scottish English, Scots and standard English English. Sociolinguistic judgments are however available to the author, since he and his family are users of standard Scottish English, lead their social lives among speakers of Scots and speakers of Scottish English, and daily either make or hear sociolinguistic judgments about spoken and written usage, including judgments of the sort made by Winford (1994: 176–182). Winford talks of the Trinidadian working classes being less familiar than the middle classes with Standard English norms and this contrast is analogous to the varying degrees of familiarity with standard written English in Scotland and indeed in all parts of the UK.
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.5 (233)
Perfect and resultative constructions in spoken and non-standard English
Although this paper draws heavily on Scottish English data, it is not claimed that any of the constructions are unique to Scottish English. (See, e.g., the discussion of (26a, b) in §4.3.) The data are, however, non-standard and spoken, and it will be useful to compare briefly the data and analysis offered here with recent accounts of the Perfect in other (non-standard) varieties of English. Tagliamonte (2000) examines data from the English-speaking enclave of Samana in the Dominican Republic, in which the Perfect function (to use T’s term) is encoded not just by HAVE + past participle but by a number of forms, such as lone past participles, pre-verbal been (she been married), pre-verbal done (I done been to Miami), verb stems (I never like the city) and others. Tagliamonte sees the Samana data as illuminating the range of variation available before the HAVE structure became specialised as the Perfect construction. Winford (1994: 155) treats done as a variant of have in examples such as She done know wha’ going on and He done livin’ here more than a year and I done tell him I is a poor man. Both Tagliamonte and Winford are concerned to establish a grammatical variable which can be presented as realised by one of a set of constructions labelled ‘Perfect’. The author’s approach here and in previous work is to treat formally distinct structures as equipollent and to place them on a continuum going from ‘Broad Scots’ to ‘standard written English’. There are several reasons why the concept of variable is not appropriate. There are different systems with different structures and different numbers of contrasts. Tagliamonte claims that the different structures in Samana English are in complementary distribution, the determining factors being, e.g., degree of remoteness in past time and lexical aspect. But degree of remoteness is reminiscent of languages with near and remote Pasts, and from Winford’s discussion it is perfectly possible that done in she done know is an inceptive. The problem is highlighted by the fact that Simple Past forms are used in Present Perfect contexts (to use Tagliamonte’s term) but Simple Past is not a variant of the Perfect. A further problem is that different structures tend to acquire different meanings (that is, speakers with different structures at their disposal tend to assign different meanings to the different structures). Tagliamonte talks of non-standard varieties of British English having the structures in use in Samana. It is not clear that any British variety had all the structures simultaneously, and the examples which T cites (2000: 337) of lone past participles look very like non-standard Simple Pasts: and I seen um both a-hanging in chains; I’m better than the best collect he ever done business with. The central point, however, is that the Scottish English data is not being presented as affording a view of an earlier stage in the development of the HAVE Perfect but as demonstrating that spontaneous spoken English and non-standard English follow their own paths of grammaticalisation and thereby shed light on the nature of the process. In this respect this paper has much more in common with Harris (1984) and Filppula
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.6 (234)
Jim Miller
(1999), who demonstrate that Hiberno-English has simply never developed a single Perfect construction with four major interpretations.
. The Perfect and definite past time adverbs Grammars of (standard) English maintain that the Perfect excludes definite past time adverbs, as in (2) above. The fact is that combinations of the Perfect and specific past time adverbs occur in written English as well as in spontaneous spoken English. The example in (3) was in an uncensored written text while (4) was in a newspaper text – with the intensive use of computers newspapers are no longer edited as rigorously as they once were. (3) The invoice has been sent off to Finance for payment before I went off on holiday [letter from staff in university finance department] (4) BEI’s success is all the more welcome as Britain has lost ground 10 years ago with the Saudis’ decision to opt for American frequencies of 60 hertz rather than the British 50 hertz – giving the American manufacturers a head start. [The Times] Examples from spontaneous speech are in (5) and (6). (5) I’ve talked to the player this morning and he isn’t leaving the club [TV interview] (6) Some of us have been to New York years ago to see how they do it [BBC News at Ten interview, January 2002 Simon Hughes, Liberal Democrat MP] Do (3)–(6) represent an innovation in spoken English which could be interpreted as a further stage of grammaticalisation, in that the more specialised Perfect is evolving into a general Past Tense? How ‘un-English’ are they? It is worth pointing out that they are not completely unusual. The constraint on the combination of Perfect and specific past time adverbs applies only in relation to actual moments of speech or writing. These can be thought of as primary deictic centres. Events narrated by means of the Perfect may relate, not to the moment of speech, but to some other time, as in the narrative in (7). These Perfects do combine with specific past time adverbs, such as the ones in bold in (7). (7) At 11 o’clock on the evening of Friday 1st September 1998 a man is trudging up Dundas Street in Edinburgh. He has had a row with his boss at 5 o’clock, has drunk too much between 6 o’clock and 8 o’clock and has lost his wallet when he fell down some steps at quarter past eight.
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.7 (235)
Perfect and resultative constructions in spoken and non-standard English
The deictic centre for (7) can be thought of as secondary, and secondary deixis can be invoked as the explanation for the combination of specific past time adverbs with Perfect infinitives, which by virtue of being non-finite are not anchored to a fixed moment of speech. Examples are in (8) and (9). (8) It’s annoying to arrive at the station at 10.30 only to find that the train has left at 10.15 and not 10.50 (9) To have arrived at 2pm on Tuesday was a miracle [is a miracle/will be a miracle] With respect to (3)–(6) as innovations, Elsness (1997: 250) cites similar examples from Shakespeare, Pepys and dialogue in Galsworthy, and Denison (1993: 352) asserts that in Middle English the Present Perfect and the Simple Past were used interchangeably with definite past time adverbs. Comrie (1985: 33) comments that ‘with considerable dialectal and idiolectal variation’ some speakers of English combine specific past time adverbs with the Perfect and points out that adverbs referring to recent points or periods of time – recently, this morning, etc. – regularly combine with the Perfect. Whether (3)–(6) represent a continuation of that usage or its re-emergence after a period of absence requires historical data that is not available to the author, but there is an important pragmatic factor that should be mentioned. When speakers refer to an event in the past they are under pragmatic pressure to enable listeners to locate the event accurately in past time. This is achieved by producing an appropriate adverb referring to a specific past time. There is no reason to suppose that this pragmatic pressure has ever ceased to apply and no reason to suppose that the construction is returning to English after a period of absence. Rather, users of English in Britain, as can be observed daily, follow other linguistic patterns away from the scrutiny of editors and prescribers.
. Perfect and Simple Past . Recent Past The Perfect is the required construction in formal written English for reference to recent past time (possibly in combination with just). Very common in spontaneous spoken standard English (and also in non-standard English) is the Simple Past (which is the norm in North America). Examples are in (10)–(12). (10) Er, as Charlie just pointed out, it is of great concern [Discussion at Trades Union Congress, recorded in the British National Corpus]
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.8 (236)
Jim Miller
(11) Sorry, Jane’s not in. She just went out (=has just gone out) [informally noted in conversation] (12) Hey! You just stepped on my foot! (?You’ve just stepped on my foot!) [informally noted in conversation] In the Scottish English data in (A)–(F) there were only two instances of the Perfect of Recent Past and none of the Simple Past plus just. However, the author regularly hears examples of the combination and when, around 1973 at about age four, his children began to talk explicitly about immediate past time, they used the Simple Past plus just, not the Perfect. Support for the author’s intuition that the Simple Past is used for reference to recent past time comes from Macaulay (1991), who provides several examples, such as (13). The author would probably say you’ve forgotten the boy and I haven’t forgotten the boy. (13) my father bought a round of drinks after the meal there wasnae one for me you see and one of the men happened to comment he says “Bob” he says “you forgot the boy” “No” he says “I didnae forget the boy” [Macaulay 1991: 197–198] Elsness (1997: 115) reports that in one of his databases just occurs thirty-eight times. Twenty nine of the occurrences are in a corpus of non-printed texts, business and social letters. Twenty of the occurrences were with the Present Perfect and only two with the Simple Past. Seven were with other verb forms (left undescribed by Elsness). In his corpus of printed British texts just occurs five times, all with the Present Perfect. In the corpus of printed American texts just occurs four times, none of them with the Present Perfect. Two were with the Simple Past and two with other verb forms. Elsness organised an elicitation test, which unfortunately elicited only acceptability judgments and did not require informants to fill in blanks. He found that more speakers of American English than of British English accepted the combination of Simple Past and just, and fewer of them accepted the Present Perfect and just. The results were statistically significant. Elsness’ findings are no surprise for American English but they are at odds with what occurs in non-standard varieties of British English and in spontaneous spoken British English. His results demonstrate the need to keep written and spontaneous spoken English apart and not to generalise from standard to non-standard. They also remind us that it is necessary to know who a given set of informants are and how long they have been exposed to formal education. The usual account of reference to recent past time in British English needs more, and more detailed, scrutiny, as does the notion of ‘British English’. We conclude this section with an alternative construction for reference to recent past time. It occurs regularly in Scottish English. (14a–b) are constructed
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.9 (237)
Perfect and resultative constructions in spoken and non-standard English
examples to demonstrate the properties of the construction but (15) is a naturally occurring example from the business meetings in (F). (14) a. b. c. d.
Somebody just kicked your car there Somebody was just kicking your car there I spoke to Susan on Friday there I was speaking to Susan there
(15) A
and we should be able to, you know, ‘I’m gonna do a job and finish the exercise’ right? OK I worked some kind of hours out there and I reckon the memo they sent I’m basing it on 250 sheets an hour but that’s you know that’s just a rough estimate
B A
The construction involves the Simple Past and the deictic adverb there. There is used to point at entities either close to the addressee or distant from both speaker and addressee. Typically there is used when the entity being pointed at is visible (actually or potentially) to both speaker and addressee. Speakers who utter examples such as (14a–d) point to events that are in the recent past and metaphorically visible, that is, they are examples of grammaticalisation involving the use of basically spatial terms for temporal reference. Note that there occurs with both Simple and Progressive aspect and the phrase Friday there in (14c) can only refer to the most recent Friday.
. Experiential The experiential meaning is regularly expressed by means of the Simple Past combined with ever, as in (16)–(18). (16) Did you ever hear the joke about the Glasgow man who goes into a bar with his pet crocodile? [informally recorded in conversation] (17) Did you ever try to give up smoking?
[current advert on TV]
(18) you said you enjoyed fishing were you ever interested in football [conversation in (A)] Such examples are probably an older usage continued rather than innovatory. In his detailed exploration of the Perfect and Preterite in digital databases of English, Elsness (1997: 247) observes that Old English had a Present Tense and a Preterite Tense (Simple Past). The Preterite had most of the functions later taken over by the Present Perfect. The assumption of these functions by the Present Perfect was neither smooth nor complete. Elsness (1997: 248) comments that texts
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.10 (238)
Jim Miller
from earlier periods of English contain numerous instances of the Present Perfect in constructions from which it is barred in current English, or not barred given (3)–(6) above. Equally, the Preterite/Simple Past was not instantly replaced by the Present Perfect in all the contexts where prescriptive grammars now require the Perfect. Elsness (1997: 340) observes that in his Old English and Middle English databases the distribution of Present Perfects and Simple Pasts develops clearly and with a consistent pattern. The development in the Modern English corpus is less clear and consistent; the rapid increase in the occurrence of Present Perfects comes to an end; the British English section of the Modern English corpus shows a slight increase but the American English section shows a marked decrease in the occurrence of Present Perfects between 1750 and the present day. Elsness’ discussion matches the author’s data with respect to the occurrence of Simple Past tokens with experiential meaning. Such tokens turn up in one of the corpuses which are towards the Scots end of the continuum mentioned in the penultimate paragraph of Section 2 above; that is, the conversations recorded for the Scottish English syntax project. Examples are (18) and (19) and (20a–b). (19) <ml68>
aye cos em i went away for an interview the other day and he telt me to dae homework and he was daein it on the board the next day and i’m stuck and wellhe asked the whole class if they got stuck at any bit – he was writing it down on the board – and I said aye and I showed him what bit and he just carried on with the next question <x m120> did he <x ml69> and he never showed me yet
(20) a.
uh huh did you ever run away or anything like that or wander go out wandering? b. who’s brenda did I meet her? [in context = ‘. . . have I met her?’
The other (very Scottish) corpus contains the business meetings, but there are only three Experiential Perfects in that corpus, and no Simple Pasts with experiential meaning. In the dialogues from Galt’s novel, however, there are eleven Simple Pasts (and ten Perfects) with experiential meaning. Three examples are in (21a–c). (21) a.
I ne’er did sic a thing in a’ my days; odd, I’m unco blate to try’t (p. 153) = ‘I never did such a thing in all my days; I tell you, I’m reluctant to try it’
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.11 (239)
Perfect and resultative constructions in spoken and non-standard English
b. Man, Johnny, this is the roughest brute that ever was created (p. 29) c. I’m no denying your parentage-I ne’er said a light word about it [in context = ‘. . . I’ve never said. . . ’ That is, the use of the Simple Past with adverbs such as ever to express experiential meaning is not a new feature of current Scottish English or Scots but a persisting, much older usage.
. Result Spontaneous spoken standard English has a number of constructions other than the Perfect which are used to convey resultative meaning. They all involve the past participle (passive), which was originally resultative (see Haspelmath 1993). One of the regularly used constructions is the one that is taken to be the source of the English Perfect, as exemplified in (22). (23) has a coordinate resultative structure and was produced by a surgeon talking about new technology in Accident and Emergency departments. (24) shows how the originally possessive construction has had its possessive nature reinforced by the introduction of have got. (22) I have the letter written (23) you have access to a vein gained and cardiac analysis done within one minute (24) I’ve got the letter written (25) is an existential-presentative construction with a resultative participle and is used to introduce a situation treated as new. (25) a. there’s something fallen down the sink b. there’s a cat trapped up the tree c. there’s one person injured in the explosion What is the structure of (26a–b)? They are instances of a copula construction: that is connected by the copula ‘s with noun phrases containing a resultative participle, the letters written and posted and him consulted. In this respect they are parallel in structure to reverse WH clefts such as That’s what you need to do. (Note that ‘reverse’ captures the fact that such clefts are the reverse of What you need to do is go on a long holiday, in which the WH phrase is in sentence-initial position. It is not to be interpreted as indicating that reverse WH clefts are reversible; What you need to do is that is peculiar.) Reverse WH clefts are used to reinforce a point and to bring a section of dialogue to a close – see the overview of previous work and the analyses of naturally occurring examples in Weinert and Miller (1996). Examples such as (26a, b) have the same discourse function as reverse WH clefts, and on the
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.12 (240)
Jim Miller
basis of the shared structural property (copular construction) and shared discourse function (26a, b) are here treated as reverse clefts, though not reverse WH clefts. (26) a. that’s the letters written and posted [Kennedy 1994] b. that’s him consulted [TV comedy show ‘Harry Enfield and Chums’] c. here’s the tyre repaired and good as new [conversation in garage] (26a–c) could in principle be replaced by the basic Perfect construction – I’ve written and posted the letters, I’ve consulted him and the tyre has been repaired and is as good as new. The replacements, however, introduce an explicit or implicit Agent in a non-cleft construction, thereby changing what is highlighted and reducing the salience of the current result. It is worthwhile stating explicitly that the construction is not unique to Scottish English. (26b) is from a television comedy programme from southern England and the construction is in regular use in Colchester in Essex.4 As an endnote to this section we should note that a detailed discussion of the resultative construction in (22)–(24) would have to give an account of the related causative-resultative construction with get and have got: to get him cairried to the pithead (Macaulay 1991 – the speaker is talking about a mining accident and the fact that he and his fellow-miners had to use their jackets as a stretcher to carry the injured man – to get him cairried – to the pithead), it might take you a week to get an alert concurred – in (F), we’ve got a training matrix established – in (F), as long as she gets her work done her be- her school work – in (D), a mini’s easy to manoeuvre i mean you can get it parked easier – in (A).
. Is the Perfect changing? (13)–(17) exemplify the current use of Simple Past Tense plus ever to signal the experiential meaning. Yet, ever and just are three adverbs that occur regularly in speech (and writing) in the English Perfect. They are usually treated as optional additions to the Perfect but their effect on acceptability is so great that the question arises as to whether they should be treated as giving rise to three separate constructions; that is, the Perfect in spoken English can be seen as splitting into three different constructions marked by yet, ever and just. The possibility of on-going change and the importance of the adverbs occurred to the author when reading Moens (1987). Moens revises Vendler’s taxonomy of lexical aspect by means of two features, with/without consequences and atomic/extended. hiccough and wink have no consequences because they produce no change in the participants in given situation; win the race and recognise do have consequences because someone has become winner of a given race
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.13 (241)
Perfect and resultative constructions in spoken and non-standard English
and someone has new knowledge about the participants in some situation. Similarly, build a house has consequences in that a new entity is created but play the piano has no such consequences. Moens (1987: 101) asserts that ’ease of occurrence’ with a perfect auxiliary isolates verbs having consequences. Moens observes, correctly, that the verbs knock, blink and hiccough do not occur easily with have, as in (27a–c). (27) a. ?He has knocked on the door b. ?She has blinked c. ?I have hiccoughed (27a–c) can be made acceptable if one devises special contexts but that is just the force of Moen’s word ‘easily’. If the verbs did combine straightforwardly with have it would not be necessary to think up special contexts. There are two interesting facts. One is that the examples are far less peculiar in interrogative and negative constructions. The other is that the addition of yet, ever and just makes the examples entirely acceptable, as shown by (28)–(30). (28) a. Has she knocked on the door (yet)? b. Has he blinked (yet)? c. Has the baby hiccoughed (yet)? (29) a.
Have you ever knocked on the door [cf. I have never knocked on the door] b. Has she ever blinked? [She has never blinked] c. Have you ever hiccoughed? [I have never hiccoughed]
(30) a. He has just knocked on the door b. She has just blinked. (Possible as a signal to someone.) c. The baby has just hiccoughed The above discussion of adverbs makes no reference to frequency of adverbs but only to acceptability. Is the role of the adverbs in acceptability reflected in the proportion of Perfects that occur with adverbs? The first step towards an answer was to code the verbs in the spoken Scottish English corpuses (A)–(F) and count the tokens of Simple Past with and without adverbs and the tokens of Perfect with and without adverbs. It turned out that for every Simple Past verb with an adverb there were nine without one, whereas for every Perfect with an adverb there were 2.2 without one. This result was not unwelcome, showing that there were proportionately many more Perfects with adverbs than Simple Pasts. That is, there did seem to be some connection between Perfects and adverbs. However, the one hundred and seventyfive Perfect verbs with no adverb ran counter to the hypothesis about the importance of adverbs – or seemed to run counter to it until the verbs were analysed
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.14 (242)
Jim Miller
in terms of their meaning. (The classic categories of Recent Past, Result, Persistent Situation and Experiential could not be easily applied to all the examples and two further categories were used, Experiential-Result and Experiential-Persistent Situation. For the purposes of this analysis Result and Experiential-Result were combined. Only examples of the sort I have finished the letter were counted as Result; not I have the letter finished.) A central fact is that ninety-five of the Perfects with no adverbs were Result, and a further fifty were Experiential-Result, giving a combined total of one hundred and forty five. Only eight Result and seventeen Experiential-Result verbs were modified by adverbs. For the other categories the ratio of verbs with adverbs and verbs without adverbs was this: Experiential 26/14; Persistent Situation 22/6; Recent Past 7/6. Result Perfects are clearly differentiated from the other Perfects; one reason may be that in context this meaning is clear and does not require an adverb. In context Recent Past is also clear, but the ratio for Recent Past Perfects, almost one-to-one, is very different from the ratio for Result Perfects. The ratio for Experientials is almost two-to-one, and for Persistent Situation Perfects it is over three-to-one. As shown in Sections 4.3 and 6, resultatives appear to be an important category to which not just one but several constructions are allotted; the above ratios drive home the difference between resultatives, including the Result Perfects, and the other Perfects. The above ratios are not quite matched by those for Galt’s dialogues: the ratios of verbs with adverbs to verbs without adverbs are: Experiential 5/5, Persistent Situation 6/6, Recent Past 1/0. The interesting ratio is the one for Result/Experiential Result: 4/69. Two hundred years ago, at least in Galt’s dialogues, Result Perfects already had a different pattern of usage from the other Perfects.
. Resultative constructions in Scottish English/Scots Scottish English has another resultative construction in addition to the constructions in (22)–(26). The resultative participles in those examples modify Patient nouns (which are not necessarily direct objects). In the resultative construction exemplified in (31)–(34) the participle modifies the Agent or subject noun. The subject noun may refer to an Agent, as in (31b, c) – and in (31a), depending on the analysis of see. The participle also modifies an Agent subject noun in (32), (33) and (34). (It is important to note that these examples are not analogous to the fallen leaves and undescended testicles examples discussed in Bresnan (1982: 29–30) which involve participles modifying Themes – to use Bresnan’s terminology.) (31) a.
but that’s me seen it (= I’ve seen it now)
[Macaulay 1991]
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.15 (243)
Perfect and resultative constructions in spoken and non-standard English
b. that’s you finished (= ‘You have finished (the task)) [Map Task Dialogues; see Weinert & Miller (1996)] c. that’s Ian arrived [informally recorded in conversation] (32) and -eh when I was waiting on the milk eh-the farmer came oot and he says ‘That you left the school noo Andrew?’ . . . says I ‘It is’ he says ‘You’ll be looking for a job’ [Macaulay 1991] (33) j50 m51 j51 m52 j52 m53 j53 m54 j54
once she goes to haddington it’ll no be so bad i mean she’s moving to haddington in june even haddington joyce is far enough you know i know martin but at least it’s not all wee windy roads to it [‘windy’ = winding] no you can get on the a1 you’re on the main road to haddington i think you can stay on the main road aye the musselburgh road and you’re there through macmerry and that way and then you’re at haddington aren’t you aye that’s right haddington’s the next one even so joyce it’s still a waste to go where does she go after that3 that’s her finished i think after haddington she goes to college again for a few weeks [conversation in (A)]
That’s her finished in (33) is a reverse cleft and the interpretation is not that she has been finished off but that she will have finished her period of training once she completes her stint at Haddington. In (32) That you left the school is a reduced interrogative reverse cleft. The full structure is Is that you left the school as shown by the reply It is. (34) now I’m trying to encourage Sally – you get on in life if you would just do that little bit more you know erm even though you’ve only got one more sentence to do – why not just do it? Even if it’s 20–34 minutes past – instead of 30 minutes – Sally is the kind of person that will say ‘that’s me done it’ and then stop and she’ll go off – but er to me you’ve got to just to get anywhere in life you just push that wee bit more – and do that wee bit more detail [conversation in (D)] (35) and (36) offer examples of the resultative construction in the past tense, the equivalent of Pluperfects. (35) the bus come at twenty-five minutes to six in the morning and he started at seven o’clock twenty-five to six now see if he didnae get that bus that was him
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.16 (244)
Jim Miller
he wasnae oot (=he hadn’t got out of the house) he needed to get the first one in the morning going doon the pit or that’s him slept in [Macaulay 1991] (36) he just lay doon on the settee and turned over and that was him gone (= he had gone) [Macaulay 1991]
. Conclusions We note first the properties of persistence and divergence. In contrast with the focus of, e.g., Hopper and Traugott (1993), it is constructions that persist and diverge, not lexical items which evolve into grammatical items. The resultative construction which was the source of the Perfect – e.g., She has the letter written – persists and is frequently used. It has also diverged from the Resultative-Perfect in that it allows both have and have got. That is, the interpretation of possessing an entity with a particular property as a result of a past action is emphasised by have got, which does not occur in the Perfect – *She has got written the letter. Secondly, the evolution of the Perfect construction did not lead to its adoption as the sole construction for the Experiential and Recent Past interpretations in all varieties of English. In spontaneous spoken (standard) English and in nonstandard English these interpretations are frequently conveyed by the Simple Past plus ever, just or there, as in (10)–(13) and (16)–(20). Contra Elsness (1997), this is typical of spontaneous spoken British English and non-standard British English. The resultative meaning is conveyed by a number of constructions, exemplified in (22)–(26) and in (31)–(34). The various resultative constructions all have a resultative participle at their core but differ in their discourse functions, as explained in Section 4.3. Finally, change does not stop. The Resultative-Possessive construction gave rise to what is analysed as a single Perfect construction with the major interpretations listed in (1a–d). There is reason to suppose, as discussed in Section 5, that the single Perfect is splitting into three separate constructions differentiated by the occurrence of particular adverbs or, for the Result Perfect, by the absence of adverbs. The process is in its earliest stages; various adverbs occur with the Experiential Perfect, not just ever. That is, specialisation has not taken place.
Notes . I am indebted to a generous but anonymous referee, to Keith Mitchell for indicating the importance of primary and secondary deixis and to Keith Brown for many discussions about aspect.
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.17 (245)
Perfect and resultative constructions in spoken and non-standard English . Denison (1993: 340–370) is content to regard late Old English, when the have Perfect became available for any lexical verb which did not conjugate with be, as the stage when perfect have had become an auxiliary verb, since it suggests that have was being used transparently, i.e. without argument structure or selectional restrictions of its own. But he points out that the literature contains different views on the relationship between the Possessive-Resultative construction and the Perfect, and on the relationship between the Possessive-Resultative construction in Old English and the Modern English structure I have the letter written. The Perfect is held to have developed first for highly transitive verbs denoting actions resulting in salient results such as physical changes in an entity and the creation or destruction of an entity. The construction took in intransitive verbs only in the last stages of its development. In addition to its resultative interpretation the Perfect then acquired the experiential, recent-past and extended-now interpretations exemplified in (1a–d). In a wider perspective the difficulties listed by Denison are counterbalanced by the links between Possessive-Resultative constructions and Perfect constructions in Romance, Germanic and Slavic. . With respect to spoken and written English, consider Chafe’s (2002: 409) assertion that the future be going to originated in the lexical verb go but that awareness of its literal origin has vanished. This might well be true for users who deal only or mostly with spoken English, especially as be going to has eroded to gonna. Is it true for users who make constant use of written English? In the written language be going to has not eroded and even new users of written English who have not connected the future gonna with go in speech might make (are very likely to make) new connections from their experience of writing. This conjecture is even more likely given the frequent movement metaphors on which reference to future time is based: taking things forward, moving on, rushing into some act, coming to the anniversary of some event, and so on. (And is it true for users who are aware that gonna in fast informal speech is equivalent to going to in careful formal speech?) Similar considerations apply to will. . I owe this information to Dave Britain, University of Essex.
References Bertinetto, Pier Marco (2000). “The Progressive in Romance, as Compared with English”. In Östen Dahl (Ed.), Tense and Aspect in the Languages of Europe (pp. 559–604). Berlin & New York: Mouton de Gruyter. Bresnan, Joan (1982). “The Passive in Lexical Theory”. In Joan Bresnan (Ed.), The Mental Representation of Grammatical Relations (pp. 3–86). Cambridge, MA: MIT Press. Brown, Keith & Miller, Jim (1982). “Aspects of Scottish English Syntax”. English World-Wide, 3(1), 1–17. Chafe, Wallace (2002). “Putting Grammaticalization in its Place”. In Ilse Wischer & Gabriele Diewald (Eds.), New Reflections on Grammaticalization (pp. 395–412). Amsterdam & Philadelphia: John Benjamins. Comrie, Bernard (1985). Tense. Cambridge: Cambridge University Press. Dahl, Östen (1985). Tense and Aspect Systems. Oxford: Blackwell.
TSL[v.20020404] Prn:24/02/2004; 16:44
F: TSL5911.tex / p.18 (246)
Jim Miller
Dahl, Östen (1999). “Perfect”. In Keith Brown & Jim Miller (Eds.), Concise Encyclopaedia of Grammatical Categories. Oxford: Elsevier. Denison, David (1993). English Historical Syntax (pp. 340–370). London: Longman. Ebert, Karen H. (2000). “Progressive Markers in Germanic Languages”. In Östen Dahl (Ed.), Tense and Aspect in the Languages of Europe (pp. 605–653). Berlin & New York: Mouton de Gruyter. Elsness, Johan (1997). The Perfect and the Preterite in Contemporary and Earlier English. Berlin & New York: Mouton de Gruyter. Filppula, Markku (1999). The Grammar of Irish English. Language in Hibernian Style. London & New York: Routledge. Harris, John (1984). “Syntactic Variation”. Journal of Linguistics, 20, 303–327. Haspelmath, Martin (1993). “Passive Participles Across Languages”. In Barbara Fox & Paul J. Hopper (Eds.), Voice. Form and Function (pp. 151–177). Amsterdam & Philadelphia: John Benjamins. Hopper, Paul J. & Traugott, Elizabeth Closs (1993). Grammaticalization. Cambridge: Cambridge University Press. Jackendoff, Ray (1976). “Toward an Explanatory Semantic Representation”. Linguistic Inquiry, 7, 89–150. Kennedy, A. L. (1994). Looking For the Possible Dance. London: Minerva. Lyons, John (1968). Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press. Macaulay, Ronald K. S. (1991). Locating Dialect in Discourse. Oxford: Oxford University Press. Michaelis, Laura A. (1994). “The Ambiguity of the English Present Perfect”. Journal of Linguistics, 30, 111–157. Miller, Jim (1993). “The Grammar of Scottish English”. In James Milroy & Lesley Milroy (Eds.), Real English (pp. 99–138). London & New York: Longman. Miller, Jim & Weinert, Regina (1998). Spontaneous Spoken Language. Syntax and Discourse. Oxford: Clarendon Press. Moens, Marc (1987). Tense, Aspect and Temporal Reference. Ph.D. dissertation, University of Edinburgh. Nedjalkov, Vladimir P. (Ed.). (1988). Typology of Resultative Constructions. Amsterdam & Philadelphia: John Benjamins Tagliamonte, Sali (2000). “The Grammaticalization of the Present Perfect in English. Tracks of Change and Continuity in a Linguistic Enclave”. In Olga Fischer, Anette Rosenbach & Dieter Stein (Eds.), Pathways of Change. Grammaticalization in English (pp. 329–354). Amsterdam & Philadelphia: John Benjamins. Weinert, Regina & Miller, Jim (1996). “Cleft Constructions in Spoken Language”. Journal of Pragmatics, 25, 173–202. Winford, Donald (1994). “Variability in the Use of Perfect have in Trinidadian English: A Problem of Categorial and Semantic Mismatch. Language Variation and Change, 5, 141–188.
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.1 (247)
Grammaticalization and standardization Lea Laitinen University of Helsinki
.
Introduction
Grammaticalization processes are generally studied as unidirectional languageinternal changes, deriving supposedly from broader cognitive tendencies. In this paper, I will discuss the analysis of these changes in relation to language-external factors, the creation and establishment of a standard language, its codification and consciously shaped norms. Grammaticalization phenomena were at first investigated in languages with large text corpora extending over centuries, but today spoken varieties and oral languages are also being studied within the same framework. The standardizing processes have, however, influenced different languages and language varieties in different ways. I argue that we should take more seriously into consideration the data we use when analyzing linguistic changes. How and to what extent do the data used affect the results of our analysis? This issue has not been widely discussed by grammaticalization theorists. In Finland we have been obliged to consider these questions, because Finnish is a language with a relatively short written history and a recently created standard form. As the documents of public discussions at times of standardization indicate, the language reformation often aimed at halting, regulating or redirecting the ongoing linguistic changes that can be identified as grammaticalization. In my paper, I will look at three different examples, concentrating on one particular case, the almost forgotten spontaneous process of grammaticalization of the negation word, which was stopped in standard Finnish during the 19th century.
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.2 (248)
Lea Laitinen
. Questions about the analysis of standardized languages Firstly, when studying languages that have gone through standardization we analyze languages that are at least partly artificial. Modern standard Finnish was not shaped until the late 19th century. Besides creating thousands of words, scholars had to select the common grammatical forms from different dialects. Both the educated, Swedish-speaking upper class – one or two per cent of the population – and the majority, the common people who spoke Finnish dialects, had to learn the new standard language. In a way, this meant a twofold change of language: at the beginning, standard Finnish was a foreign language to everybody, Finnish-speakers and Swedish-speakers alike. Secondly, the quality of the texts analyzed should be taken into consideration. Finnish texts from the 16th to the 19th century were mainly translations from Swedish religious and legal texts, and still in the 19th century, the majority of published texts were translations. Naturally, the language of translated texts differed in many respects from the language of texts that were originally written in Finnish. This is another reason why Finnish linguists have primarily based their research on the grammaticalization phenomena of spoken Finnish, especially on extensive dialect corpora, and nowadays also on conversational recordings. Of course, the grammaticalization phenomena in written languages may differ from those in spoken varieties in many other respects as well (see Herlin & Kotilainen, this volume). Thirdly, we could ask which ideological trends or whose linguistic values we are studying in analyzing standardized languages. There are times in the history of languages when such aspects are more or less explicitly documented. The Finnish language reform in the 19th century was a deliberate national project with ideological principles, a compromise made between several dialects and old literary Finnish. This project made people unusually well aware of the linguistic variation and ongoing changes in Finnish. Language was widely discussed in public. There were vivid debates on grammatical details, reflecting different linguistic ideologies, but also conflicts between various political groups. These ideologies influenced the way grammaticalization phenomena emerged in Finnish. Many reformers, especially the linguists specialized in Finnish, aimed at a pure, homogeneous and regular Finnish. They tried not only to avoid ‘foreign’ linguistic constructions but also to minimize the variation, asymmetry or polymorphism of the ‘original’ forms. Such national tendencies were widely common in Europe during the 19th century, and they also influenced languages that were standardized much earlier than Finnish. Even today, there are similar processes taking place in languages all over the world, where speakers are simultaneously creating their national identity, literacy and standard language. Thus, the questions posited here do not only concern
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.3 (249)
Grammaticalization and standardization
Finnish but are meant to provoke a more general discussion about the complicated role of standardization in grammaticalization phenomena as well. The fourth issue that makes this discussion relevant for grammaticalization studies has to do with the unidirectionality in grammaticalization. According to the principle of unidirectionality, the grammaticalizing forms tend to develop towards non-referential, more abstract functions, expressing subjective, interactional or textual meanings. Unfortunately, these functions are not necessarily present in the reference grammars that are used in typological studies of grammaticalization phenomena. It has been noticed in many different languages that – besides variation and oral features – prescriptive codification removes emotional or subjective meanings from standard languages (Stein 1997). In other words, it is often the most context-bound, purely indexical functions of linguistic forms (cf. Silverstein 1976) that tend to be avoided. This has undoubtedly had an influence on the intuitions of native speakers as well.
. Unexplicated changes and explicated norms . The standardization of the constructions of necessity My first example of the relation between grammaticalization phenomena and the standardization process in the 19th century is the modal construction formed by verbs of necessity (the NEC construction). The NEC construction had become grammaticalized in the spoken language hundreds of years earlier. It is morphosyntactically marked in many respects. Firstly, unlike most Finnish verbs, the modal verb (e.g. Finnish täytyä “must, have to” and pitää “ought to”, “should”) is always used with 3rd person singular verbal inflection, both in the standard language and in dialects.1 Secondly, the subjectlike argument in the NEC clause is often in the genitive case, contrary to both non-modal and other modal verbs such as the verbs of possibility, which have nominative subjects. In non-standard Finnish, the case marking of the first argument depends on its position in the NP hierarchy of agentivity, or indexical status as speech act participants (cf. Silverstein 1976, 1981). In the top of the hierarchy, the personal pronouns, including the logophoric 3rd person pronouns hän and he, always refer to speech act participants, and they are marked with genitive without exception. By contrast, inanimate or abstract NPs are in the nominative case. The NPs in the middle of the hierarchy have either the genitive or the nominative case, depending on their status as a potential speech act person in the context. Thus, even minimal pairs are possible, such as in (1a) and (1b) (for details see Laitinen 1995 and 1997).
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.4 (250)
Lea Laitinen
(1) a.
Morsiamen pitää olla kaunis bride.gen must be.inf beautiful “The bride must be beautiful” “It is necessary for the bride to be beautiful” (i.e., the bride herself is obligated to be beautiful) b. Morsian pitää olla kaunis bride-nom must be.inf beautiful “A bride must be beautiful” “It is necessary that a bride is beautiful” (i.e., there is a need that the bride be beautiful)
The two earliest grammars of Finnish syntax (Koskinen 1860 and Jahnsson 1871) presented non-standard cases such as (1b) as acceptable. In the 1870s, NEC constructions were a topic of public debate, especially because the vernacular case marking options (illustrated by (1a, b)), as well as the epistemic functions of the NEC verbs (cf. (1b)) were feared to be a result of a Swedish influence. After the description of syntax by Setälä (1880), however, the nominative variants were rejected in standard Finnish as ‘ungrammatical’. As the translation paraphrases in (1) illustrate, there is a link between the status of a speech act person and the modal function of necessity. In (1a), the subject (“bride”) is treated as an intentional being, a speaking subject to whom the deontic duty or norm can be directed, whereas in (1b) the necessity is practical, based on expediency; the intentionality and even the potential faculty of speech of the bride is irrelevant. Thus, the non-standard grammar of NEC constructions offers an opportunity to the speaker to construct the modalities and the participation framework of the described situation in alternative ways. Obviously, it was difficult to base the standard norm on such subjective distinctions. Thus, even if the case marking and verb inflection options of these constructions were systematically documented in studies on Finnish dialects in the last decades of the 19th century, the standardizers settled on another solution. In Standard Finnish, the use of the nominative case in NEC constructions is largely restricted to the so-called existential clauses introducing new referents, with a preverbal locative or possessive as in (2).2 In other words, the case marking is based on the propositional meaning of the clause. (2) Minulla pitää olla morsian I.ade must be.inf bride.nom “I must have a bride” Today, the standard norm is consistently followed in formal situations. The alternative however, i.e. the vernacular case marking, is still widely used in speech and in informal writing, even if people are not aware of it. Thus, the path of the
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.5 (251)
Grammaticalization and standardization
grammaticalization process can be reconstructed on the basis of dialectal data and other spoken or informal varieties. In previous articles I have provided an indepth analysis of he grammaticalization of certain lexical intransitive verbs with 3rd person nominative subjects into NEC constructions with infinitival complements and genitive arguments (see e.g. Laitinen 1995, 1997). My present concern has been to warn against an uncritical use of sources for the reconstruction of grammaticalization processes.
. From logophoric pronouns to discourse particles My second example concerns the status and development of the so-called logophoric pronoun in Finnish. Interestingly, some of the grammaticalization processes going on in the 19th century were totally ignored by the reformers and scholars of Finnish linguistics. Their negligence may even have fostered some changes to progress, at least in the written language. An example of this is the development of the third person singular pronoun hän, “s/he”, into an epistemic discourse particle -hAn (roughly) “I wonder; as you know; after all” (see Laitinen 2002).3 The source for the development of the clitic particle -hAn can be traced to Finnish dialects in which the pronoun hän is logophoric. Logophoric pronouns are used to indicate coreference with an individual whose speech or point of view is being reported, mostly in indirect discourse (see e.g. Hagège 1974; Hyman & Comrie 1981). The cognate logophoric (LOG) pronoun sun in the genetically related Saami languages has adopted similar discourse functions as a particle. Examples like (3), which is from Setälä (1883), are common in all Finnish dialects. (3) Se sano, että kyllä hän tiätää mitä se tekee s/he said that aff log knows what.par s/he does “S/hei said that surely s/hei knows what s/hej (another person) is doing” The parallel development of Finnish and Saami helps us to find contexts where the process could have started, i.e. interrogative clauses expressing the speaker’s ignorance of the situation of the LOG subject’s referent. The process is traceable in spoken varieties of both languages (for details see Laitinen 2002). Sometimes it is still impossible to decide whether the element is used as a pronoun or as a particle in speech. In (4) hän has been analyzed by the transcriber as an enclitic particle. If it had been written as a separate word (mikä hän), it would have had a pronominal identity. (4) Se ol Samppa, mikähäl lie ollus sukusis se he was [name] what+HÄN be.3sg. pot.pst kindred that viinankeittäjä distiller
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.6 (252)
Lea Laitinen
“His name was Samppa – I wonder where he came from, that distiller of spirit” However, it was only in Finnish that the particle spread to all other clause types as well, and this spreading happened mainly in the 19th century. This extension, as well as the extension of the LOG pronoun to inanimate referents and finally to loss of referentiality, can be followed in Finnish texts throughout the century. Example (5) from the end of the 18th century illustrates both processes. Here, the grammatical category of the word hän is again ambiguous. There are two arguments in favour of it being a particle: first, it obeys the vowel harmony conventions in Finnish, and second, there is a postverbal inanimate NP (juoma “drink”) that resembles dislocations and could have taken over the role of the ‘real’ subject of the clause. (Example (4) could illustrate the same development.) (5) Jopa han on juoma wähä waljahtanut already HÄN is drink a little stale “Why, the drink is already (indeed) a little stale” The scholars of the time did not comment on this gradual development, although some descriptions of Finnish dialects (Lönnbohm 1879: 65; Setälä 1883: 92) mentioned the hypothesis of the pronominal origin of -hAn. At that time, the researchers published comprehensive syntactic studies on the basis of the dialect data they had obtained on their field trips in different parts of Finland. The main motive of this work was to collect enough linguistic data for creating rules for standard Finnish. Nevertheless, even though the logophoricity of hän in dialects was thoroughly described in several studies, it was not mentioned in Finnish grammars.4 In addition, the logophoric function of hän was excluded from written texts, where its use was very inconsistent until the 20th century. In writing, hän was mainly used as the 3rd person pronoun for humans (“s/he”). But it was used for non-human referents as well, as an alternative for the pronoun se (“it”). This use emerged in translations from source languages with gender, especially from Swedish, and following the development of the Swedish pronouns han “he” and hon “she”, also hän was gradually limited in texts to animates, and finally to human referents alone (Laitinen, forthcoming). Thus, in written language, the relation of hän to the particle hAn was quite opaque, which may have been a precondition needed for hAn to extend quietly and inconspicuously into new functions. To summarize thus far, whereas the use of the NEC constructions was restricted to certain non-indexical meanings in the standard language, the grammaticalization of the LOG pronoun to a purely indexical discourse particle was not even noticed by the language reformers during the 19th century. In the former case, the grammaticalization only went on in dialects and informal speech and writing; in the latter one it happened in both vernacular and standard varieties. Our final
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.7 (253)
Grammaticalization and standardization
example deals with a third type. An incipient grammaticalization was debated for quite a long time by standardizers, and then it was finally excluded.
. From verb to particle – an interrupted change To illustrate genuine public conflicts between the processes of grammaticalization and standardization, let us look at the development of the non-agreeing negation word (NEG). As can be seen in (6), Finnish verbs normally agree in person and number with the nominative subject. The parentheses indicate that the presence of subjects in the first and second persons is optional. (6) olla “to be” in standard Finnish SG PL 1. (minä) olen “I am” (me) olemme “we are” 2. (sinä) olet “you are” (te) olette “you are” 3. hän on “s/he is” he ovat “they are” The negation word in Finnish is not a particle but an auxiliary verb. It lacks past tenses, non-indicative modes, and passive and infinitival forms. However, it always takes a connegative verb, i.e. a main verb in a special uninflected form, and it agrees in number and person with the nominative subject, as in (7). (7) NEG in standard Finnish SG 1. (minä) en ole “I am not” 2. (sinä) et ole “you are not” 3. hän ei ole “s/he is not”
PL (me) emme ole “we are not” (te) ette ole “you are not” he eivät ole “they are not”
There are still three separate dialect areas (among the south-western, western and eastern dialects) where the negation word can be used impersonally, having only the 3rd person singular form, as in (8). In these cases, the subject pronoun is obligatory.5 Like particles, the non-agreeing negation word occupies the initial position in the clause. (8) NEG in three local dialects SG 1. ei minä ole “I am not” 2. ei sinä ole “you are not” 3. ei hän ole “s/he is not”
PL ei me ole “we are not” ei te ole “you are not” ei he ole “they are not”
In written Finnish, the agreement of the negation word with the subject was gradually established, as is roughly described in examples (9) to (12).6 As shown in
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.8 (254)
Lea Laitinen
(9), there was much variation during the first century of written Finnish. In the first translation of the Bible into Finnish, published in 1642, the number distinction was disappearing. This is shown in (10), where the square brackets refer to infrequent options. Subsequently, the plural forms came back gradually, as outlined in (11) and (12). By the mid-20th century, the Finnish negation word agreed completely in person and number with the subject NP. (9) 16th century texts SG PL 1. en ∼ ei minä emme (me) ∼ ei me 2. et ∼ ei sinä ette (te) ∼ ei te 3. ei hen e(i)vät (he) ∼ ei he (10) 1642 Bible SG 1. en minä 2. et sinä 3. ei hän
PL en me [∼ emme (me)] et te [∼ ette (te)] ei he
(11) Texts of 1645–1810 PL 1. en me ⇒ emme (me) 2. et te ⇒ ette (te) (12) Texts of 1810–1970 PL 3. ei he ⇒ eivät he However, some other processes were also affecting the negation word in the 19th century. It was grammaticalizing, taking on interactional and textual functions, developing in certain contexts into a non-agreeing particle in the 3rd person singular form. The whole history of this process had been totally forgotten when I accidentally came across it whilst studying 19th century standardization debates in Finland. In the 1870s there were intense polemics about the form of the negation word.7 Finally, the proponents of agreement won, and now both written and spoken Finnish (except in the few regional dialects mentioned above) comply with the norm. The debates concerned the possible particlehood of the word ei in two particular contexts: firstly, negative answers to questions, and secondly, co-ordinating complements and adjuncts in a predicate. I will introduce examples of both context types.
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.9 (255)
Grammaticalization and standardization
. Answers to questions As shown in (13), the verb is typically repeated in answers to yes-no questions, the subject being indicated by the person affix of the verb. In negative answers, the connegative verb is optional. (For answers in Finnish see further Hakulinen 2001.) (13) Tuletko (sinä) mukaan? come.2.sg.q you.sg along “Are you coming along?” – Tulen. / En (tule). come.1.sg. neg.1sg. come “– Yes, I am” / – “No, I’m not” The early grammars of Finnish mentioned an exception to this rule. The exception was first noticed in 1824 by Reinhold von Becker, a Finnish grammarian (see example (14)). Example (15) is from Jahnsson, 1871, who comments: “If the negation in an answer relates to one of the dependents of the predicate in the question, the third person singular is used.” (14) Hevoistako haet? horse.par.q search.2sg.prs “Is it the horse you are looking for?” – Ei, härkiä neg.3sg oxen.par.q “– No, the oxen.” (15) Kalevalaako luet? Kalevala.par.q read.2sg “Is it the Kalevala you are reading?” – Ei Kalevalaa, vaan Kanteletarta; neg.3sg Kalevala.par but Kanteletar.par “– No, not the Kalevala but the Kanteletar” The next description of Finnish syntax however (Setälä 1880), held an opposing view of the marking of negation: “The negation word in Finnish is a verb and agrees in person and number.” This is illustrated in example (16), where the negation word is in the 1st person singular.8 (16) Etkö luekaan Kalevalaa? neg.2sg read.cl Kalevala.par.q “Aren’t you reading the Kalevala?” – En Kalevalaa, vaan Kanteletarta. neg.1sg Kalevala.par but Kanteletar.par “– No, not the Kalevala but the Kanteletar”
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.10 (256)
Lea Laitinen
The negation particle ei is also found in spoken dialogues in fictional texts, and Finnish conversation analysts have lately found it in answers to questions in their data (17). (17) (A:) (B:)
Oot sä käyny näissä huutokaupoissa kiertämässä? have.2sg you gone these.ine auction.pl.ine strolling.ine Ei, minä oon nyt vaan toistaiseksi vasta katsellu neg.3sg I have.1sg now only so.far only look.ptc “Have you been going around to these auctions?” “– No, I have just been looking around so far”
Historically, the verb-like usage of ei as an auxiliary that agrees in person and number with the subject belongs to the Uralic features of Finnish. Its use as an uninflected particle in conversational language is thus a later phenomenon, developed in the process of grammaticalizing and assuming interactional functions. As we saw, these were excluded from the standard language. Next, I will turn to the textual functions of ei, the use of eikä in co-ordinations.
. Compound co-ordinating conjunction The second example of non-agreeing negation is the compound co-ordinative conjunction eikä “neither”, which was first discussed by G. A. Avellan in 1853 (example (18)). In modern Standard Finnish (example (19) from www.kolumbus.fi/tommi. tuura/nytuho.html), eikä is not allowed; the first person form (enkä) is required. (18) En tunne asiaa eikä ihmisiä neg.1sg know thing.par neg.3sg-cl people.par “I know neither the thing nor the people” (19) En ymmärrä enkä edes halua ymmärtää neg.1sg understand neg.1sg-cl even want understand.inf “I don’t understand nor do I even want to understand” Jahnsson accepted this use of eikä in his grammar in 1871. He writes: “The clitic particle -ka, -kä, united with the third person singular negation verb, is used as a connecting word between two dependents of the predicate. If this word, however, is related to the subject or to the whole predicate, then the negation verb is inflected.” Setälä, on the other hand, ignored this use of eikä entirely in his grammars. And yet, there were numerous examples of it in his study on dialects; the following example is from a fairy tale (Setälä 1883: 108):
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.11 (257)
Grammaticalization and standardization
(20) Et saa tulla vaatteissa eikä vaatteetoinna, neg.2sg may come.inf cloth.pl.ine neg.3sg-cl without-cloth.ess eikä hevosella eikä hevosetonna, eikä neg.3sg-cl horse.ade neg.3sg-cl without-horse.ess neg.3sg-cl tiätä eikä tiänsivua road.par neg.3sg-cl roadside.par “You are not allowed to come either with clothes on or without clothes, either riding a horse or without riding a horse, either on the road or on the roadside” The non-agreeing eikä was also used in texts written by educated people in the first decades of the 19th century. The number of occurrences was, however, quite small: in a corpus of 4,000 examples, it represented less than one percent. Example (21) was written by Elias Lönnrot, professor of Finnish and the creator of Kalevala, the Finnish national epic: (21) Muualla emme löydä turvaa eikä onnea, kun Elsewhere neg.1pl find shelter.par neg.3sg-cl happiness.par as uskossa - faith.ine “We find neither safety nor happiness elsewhere, but in faith - -” Where we do find frequent use of eikä is in the texts of self-educated writers towards the turn of the century. Example (20) was written by Frans Fredrik Björni, born in 1850 (Makkonen, ed. 2002: 187): (22) Tamperella en lainkaan poikenut mennessä, eikä Tampere.ade neg.1sg at-all stop-off.ptc go.inf.ine neg.3sg-cl palatessa return.inf.ine “I did not stop off at Tampere on the way there nor back” In the spoken language of today, the phenomenon has not yet been investigated. It can be found however in dialogues in contemporary fiction. Elsewhere in modern texts, eikä is used sometimes, but then it is usually co-ordinated with 3rd person plural forms: (23) Kaikki eivät välitä äänittää eikä kuunnella all neg.3pl bother record.inf neg.3sg-cl listen.inf kasetteja cassette.pl.par “All people do not bother to record or listen to cassettes”
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.12 (258)
Lea Laitinen
Thus, the non-agreeing eikä in co-ordinated structures has more limited usage in present-day Finnish than the simple ei in answers to questions. At any rate, it is only the extensive dialectal and conversational data that will show us how the grammaticalized (i.e. non-agreeing) forms of the negation word in interactional and textual functions coexist with its older verb-like variants in spoken Finnish.
. The ideology of agreement In the ideologically loaded national debates on the Finnish language in the 19th century, all participants aimed to find the most ‘national’ forms and categories for the new standard language under construction. Thus, modern standard Finnish and Finnish linguistics were created simultaneously and can hardly be treated separately. One of the ‘genuine Finnish’ features discussed in connection with the category of the negation word was the agreement of verbs with their subjects. However, the rule of SV agreement was not violated by the negation particle because the apparent non-agreement of ei was based on its closer relationship with arguments other than subjects. Moreover, both parties involved in the debate believed that ei would undergo change into a particle in the near future. What they disagreed on was how it should be categorized in the standard language. One of the parties wanted the norm to be based on the view of the majority of the speakers of Finnish. They were prepared to wait for the language community’s view to be established with respect to the agreement of the negation word. In addition, they accepted that negation could belong to two grammatical categories at the same time: both to verbs and to particles. Most of these liberal-minded standardizers were originally Swedish-speaking – that is, they had switched to the use of Finnish as adults9 – and they were not professionals in the study of the Finnish language. The other party, by contrast, considered it important that the negation word would belong to one grammatical category only. They justified their view on historical grounds and tried to select the norm from the ‘purest’ dialects – that is, the dialects they thought to be the least influenced by other languages. These reformers were also experts on the Finnish language. Ultimately, the entire problem was solved by the highest authority in the field, the professor of Finnish E. N. Setälä, whose intuition more often than not served as the basis for the Finnish standard language. In his inaugural speech as professor (1894), he explicitly refused to propose norms for the Finnish language. The only advice he wanted to give was that one should stick to the language of the best authors. However, Setälä created implicit norms in his grammars for many features of Finnish by simply ignoring colloquial forms and constructions found in dialectal data, including grammati-
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.13 (259)
Grammaticalization and standardization
calization phenomena. Thus, leaving something unsaid turned out to be a way of creating self-evident norms for the standard language.
. Summary In this paper, I have discussed three cases in which the grammaticalization process and the reform of the Finnish standard language conflicted with one another in different ways. My first example, the necessity construction, is much older than the creation of the Finnish standard language. In the process of grammaticalization, the case marking options (nominative vs. genitive) of the subject in these constructions were based on their indexical status as speech act participants. In the project of the standardization of Finnish in the 19th century, it proved difficult to describe this system with a single rule, and the use of the nominative case was restricted to one specific referential function only. In the end, the standard norm was adopted only in the written language. The second case, the development of Finnish and Saami logophoric pronouns into enclitic discourse particles was a process that started off in certain interrogative contexts at the turn of the 19th century, and in Finnish the enclitic particles spread to all clause types during the 20th century. According to my hypothesis, this expansion was triggered by several factors. First, the logophoric function of the pronoun hän was excluded from the written language, and secondly, the use of hän was heavily influenced by the use of pronouns in texts from other (Indo-european) languages which were translated into Finnish. Because of these confusing factors, the standardizers did not pay attention to the development in which hän adopted new, non-referential functions. The negative auxiliary, the third case analyzed in this paper, has many particlelike features, and certain modern dialects, as well as the five hundred years’ history of the standard language, show a considerable proportion of non-agreeing forms. In 19th century texts, its non-agreeing use as a particle was already limited to a restricted number of contexts, and after the public standardization debate, the negation word was placed in the category of verbs in all its syntactic and pragmatic functions. Knowing its history in the written language helps us to study the recent rise of non-agreeing forms in modern Finnish. Having remained hidden for over 150 years, this usage will be mentioned in a new descriptive grammar of Finnish where also spoken language is widely discussed (Hakulinen et al. forthcoming) as a separate variant of the auxiliary construction.
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.14 (260)
Lea Laitinen
Abbreviations ACC ADE CL ELA ESS GEN INE INF LOG NEC
accusative adessive unspecified clitic elative essive genitive inessive infinitive logophoric necessity
NEG NOM Q PAR PST PX PL PTC SAP SG
negation word nominative question clitic partitive past tense possessive suffix plural (past) participle speech act person singular
Notes . The NEC verbs have been grammaticalized from specific lexical verbs that were mainly used with 3rd person non-personal subjects. Only in limited areas with close proximity to Swedish dialects do the NEC verbs agree with nominative subjects in person and number. This is a recent phenomenon which cannot be used as evidence in reconstructing the development of the NEC construction (Laitinen 1992; contrary to Campbell 1998: 244–246). . For the term existential sentence, see Jespersen (1992 [1924]: 154–156); for its use in Finnish e.g. Vilkuna (1989: 155–160), Helasvuo (2001: 61–63, 97–103). . According to the vowel harmony rule, the enclitic particle -hAn (where A is an archiphoneme) is realized either as hän (see example (4)) or as han (example (5)) depending on the vowel quality in the root of the word. In addition, the final -n of both the pronoun and the particle can be assimilated with the initial lateral or nasal consonant of the next word (see example (4): mikähäl lie). . It is not mentioned today, either, but it will be in Hakulinen et al. forthcoming. Thus, although they use it in their speech, people are not aware of the logophoricity of hän. An illustrative consequence of this can be found in the classic article on logophoricity by Hagège (1974), where Finnish is misleadingly presented as a peculiar language with a third person pronoun (hän) that does not mark any differences between the persons referred to. . Also in Standard Estonian, the negation word (ei) is always used in the 3rd person singular form. . In order to simplify the description, I exclude the connegative verbs, and all the 3rd person pronouns except singular hän and plural he from the picture. According to the detailed account of this process in Savijärvi (1977: 259), the agreement of 3rd person plural forms (see (10)) was gradually adopted in texts as follows: in 1810–1849 50.5%, in 1850–1889 78.6%, in 1890–1929 90.1%, and in 1930–1969 98.5%. . The main opponents in this debate were August Ahlqvist, professor of Finnish, and A. W. Ingman, professor of exegetics and chairman of the first board of linguistics and language use in Finland. Both were distinguished translators and language reformers.
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.15 (261)
Grammaticalization and standardization . Setälä did however present several examples of its use as particle in his syntactic description of dialects (and so did other researchers). Paradoxically, in his grammar of spoken dialects, Setälä recommended not using the uninflected NEG word ei in writing: “One has to ask, what is the most correct and logical form, [i.e.] the form that agrees with the nature of the language and is based on the use of most speakers of Finnish and cognate languages” (Setälä 1883: 111). . Of course, all language standardizers were educated in Swedish, even those who had been born to Finnish-speaking families.
References Avellan, G. A. (1853). “Om Sättet att i Finska Periodbyggnaden Uttrycka Negativa Begrepp”. Suomi. Tidskrift i Fosterländska Ämnen, 1–48. Helsinki: SKS. Becker, Reinhold von (1824). Finsk Grammatik. Helsinki. Campbell, Lyle (1998). Historical Linguistics. An Introduction. Cambridge, MA: MIT Press. Hagège, C. (1974). “Les Pronoms Logophoriques”. Bulletin de la Société de Linguistique de Paris, LXIV, 287–310. Hakulinen, Auli (2001). “Minimal and Non-Minimal Answers to Yes-No Questions”. Pragmatics, 11, 1–15. Hakulinen, Auli et al. (forthcoming). Iso Suomen Kielioppi. Helsinki: SKS. Hakulinen, Auli & Eeva-Leena Seppänen (1992). “Finnish Kato: from Verb to Particle.” Journal of Pragmatics, 18, 527–549. Helasvuo, Marja-Liisa (2001). Syntax in the Making. The Emergence of Syntactic Units in Finnish Conversation. Amsterdam & Philadelphia: John Benjamins. Hyman, Larry M. & Bernard Comrie (1981). “Logophoric Reference in Gokana”. Journal of African Languages and Linguistics, 3, 19–37. Jahnsson, A. W. (1871). Finska Språkets Satslära. För Läroverkets Behof. Helsinki: SKS. Jespersen, Otto (1992 [1924]). The Philosophy of Grammar. With a New Introduction and Index by James D. McCawley. Chicago & London: The University of Chicago Press. Kannisto, Artturi (1901). “Lauseopillisia havaintoja läntisen Etelä-Hämeen kielimurteesta”. Suomi, III, 20. Helsinki: SKS. Koskinen, Yrjö (1860). Finska Språkets Satslära. Åbo: J. W. Lillja. Laitinen, Lea (1992). Välttämättömyys ja persoona. Helsinki: SKS. Laitinen, Lea (1995). “Metonymy and the Grammaticalization of Necessity in Finnish”. SKY 1995, Yearbook of the Linguistic Association of Finland. Helsinki. Laitinen, Lea (1997). “Norms Made Easy: Case Marking with Modal Verbs in Finnish”. In Jenny Cheshire & Dieter Stein (Eds.), Taming the Vernacular. From Dialect to Written Standard Language (pp. 110–124). London & New York: Longman. Laitinen, Lea (2002). “From Logophoric Pronoun to Discourse Particle: A Case Study of Finnish and Saami”. In Ilse Wischer & Gabriele Diewald (Eds.), New Reflections on Grammaticalization (pp. 327–344). Amsterdam & Philadelphia: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 16:07
F: TSL5912.tex / p.16 (262)
Lea Laitinen
Laitinen, Lea (forthcoming). “Hän, the third speech act pronoun”. In Ritva Laury (Ed.), Minimal Reference in Finnic: The Use and Interpretation of Pronouns and Zero in Finnish and Estonian Discourse. Studia Fennica. Helsinki: SKS. Lönnbohm, O. A. F. (1879). “Jääsken, Kirvun ja osittain Rautjärven ja Ruokolahden pitäjien kielimurteesta”. Suomi, II, 13. Helsinki: SKS. Makkonen, Anna (Ed.). (2002). Karheita kertomuksia. Itseoppineiden omaelämäkertoja 1800luvun Suomessa. Helsinki: SKS. Savijärvi, Ilkka (1977). Itämerensuomalaisten kielten kieltoverbi I. Suomi. Helsinki: SKS. Setälä, E. N. (1880). Suomen kielen lause-oppi. Oppikirjan koe. Helsinki: K. E. Holm. Setälä, E. N. (1883). “Lauseopillinen tutkimus Koillis-Satakunnan kansankielestä”. Suomi, II, 8. Helsinki: SKS. Setälä, E. N. (1894). “Oikeakielisyydestä suomen kielen käytäntöön katsoen” I. II. Valvoja, 81–99, 190–217. Helsinki: SKS. Silverstein, Michael (1976). “Shifters, Linguistic Categories, and Cultural Description”. In Keith H. Basso & Henry A. Selby (Eds.), Meaning in Anthropology (pp. 11–55). Albuquerque: University of New Mexico Press. Silverstein, Michael (1981). “Case Marking and the Nature of Language”. Australian Journal of Linguistics, 1, 227–247. Stein, Dieter (1997). “Syntax and Varieties”. In Jenny Cheshire & Dieter Stein (Eds.), Taming the Vernacular. From Dialect to Written Standard Language (pp. 35–50). London & New York: Longman. Vilkuna, Maria (1989). Free Word Order in Finnish. Its Syntax and Discourse Functions. Helsinki: SKS.
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.1 (263)
External factors behind cross-linguistic similarities* Ilona Herlin and Lari Kotilainen University of Helsinki
.
Introduction
The idea that grammaticalisation proceeds from lexicon to grammar but not vice versa, the so-called ‘principle of unidirectionality’, has received different kinds of criticism. The main line of criticism has focused on counter-examples (see e.g. Ramat 1998: 115; Campbell 1991; Joseph & Janda 1988). Furthermore, the (possible) reversibility of developments has been discussed theoretically (see Haspelmath 1999; Campbell 2001). Ramat (1998) has also suggested that the idea of directionality needs to be elaborated on by distinguishing between different subcategories of unexpected developments. In our article we concentrate on the latter issue. We develop the idea of directionality by drawing attention to the complex factors underlying (seemingly) expected developments. In addition, we comment on the methodological weakness of drawing diachronic conclusions from artificial, synchronic data. The principle of unidirectionality is at least partially based on the assumption of universal semantic developments, i.e. the cross-linguistic similarities in the development of grammatical meaning. Sometimes these universal paths seem to be so self-evident that whenever one sees a polysemous element, one automatically assumes that the more abstract meaning has developed from the less abstract. An example of this kind of reasoning can be found in Heine and Kuteva (2002: 11): “. . . if in a given grammar the author states that the adverb ‘behind’ is “homophonous” with or “resembles” the noun ‘back’, then the assumption made here on the basis of a larger corpus of cross-linguistic data is that we are dealing with an instance of the grammaticalization of a body part noun to a locative adverb.” Another typical example of this kind are elements having temporal and causal uses: their causal meaning is assumed to have developed from the temporal one
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.2 (264)
Ilona Herlin and Lari Kotilainen
by a conventionalisation of causal implicatures. Traugott and König (1991: 197) describe this development by citing the Finnish conjunction kun as one of their examples. Their assumption, repeated in Heine and Kuteva (2002: 291), seems to rely solely on the fact that kun has both temporal and causal uses in modern Finnish. We question the simplicity of this reasoning by discussing two Finnish constructions: the conjunction kun “as, when, while” (example (1)) and the converb essa (example (2)). (1) Tapasimme heidät, kun matkasimme Helsinkiin. meet-1pl.past they-acc when travel-1pl.past Helsinki-ill “We met them when we travelled / were travelling to Helsinki” (2) Tapasimme heidät matkatessamme Helsinkiin. meet-1pl.past they-acc travel-essa-ps Helsinki-ill “We met them when we travelled / were travelling to Helsinki” These two constructions have very different syntactic characteristics: kun is a subordinating conjunction and essa is a ‘converb’ (Haspelmath 1995: 3) – a non-finite verb form whose function is to mark adverbial subordination.1 In literary Finnish, however, kun and essa have similar meanings. Both usually mark a temporal overlap of two events, and both have causal uses. So these constructions may be seen as potential examples of temporal to causal development. We argue, however, that the claim that ‘meanings tend to develop from temporal to causal’ is too strong. We adopt the viewpoint that changes in grammatical meaning cannot be automatically ascribed to universal cognitive processes and idealised pragmatic principles. Instead, we point to other, external factors, such as language standardisation, language contact, etc. The picture of the development of grammatical constructions may change dramatically if the analysis includes a simultaneous effect of both language-internal and language-external factors. In practice, this means that before conclusions could be drawn about the direction of and reasons for grammatical change, the uses of polysemous constructions need to be analyzed in detail. This analysis should also be related to the cultural and socio-historical environment in which the changes have taken place. Furthermore, the different uses of constructions in the different varieties of language should also be taken into account. In the case of the converb essa and the conjunction kun, for example, it should not be ignored that in addition to the temporal and causal uses, they both also have conditional and contrastive uses, and that their uses have different distributions in written and spoken Finnish. The method used in this paper contrasts with the usual methods used in typologically oriented studies, which often rely on the language of reference grammars (see, e.g. Bybee et al. 1994: 27–50). Using only reference grammars as data is problematic because the reference grammars tend to describe the literary language only.
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.3 (265)
External factors behind cross-linguistic similarities
Modern literary Finnish is a result of a extensive standardisation process which took place mainly in the 19th century (see Laitinen, this volume). At that time, standard Finnish, a non-Indo-European language, was created mainly on the basis of dialects, but also under the strong influence of the Indo-European languages. For instance, the second official language of Finland, Swedish, has affected Finnish profoundly. Sweden is the western neighbour of Finland, and Swedish has originally also been the language of the Finnish educated classes. In fact, the creators of modern, standard Finnish were mainly native speakers of Swedish.2 In addition, Latin, German and English have had an impact on Finnish. These languages have been the source languages of literature translated into Finnish – an essential part of the new literary culture of the 19th century. Thus, Finland being a part of “Europe as a cultural unit” (Kortmann 1997: 46), we must take into account the constructions of Indo-European languages when explaining language change in Finnish.
. The conjunction kun As a starting point, we adopt a critical attitude towards the simplified view of kun as a temporal-causal conjunction that has acquired its causal meanings through the conventionalisation of causal implicatures. First, a distinction needs to be drawn between the kun in written Finnish and kun in spoken Finnish. In both written and spoken Finnish, kun is used temporally and causally. Furthermore, kun has contrastive uses in the written variety, and comparative and similaritive uses in the spoken variety (Figure 1). After first providing an overall picture of the use of kun in spoken Finnish (2.1), we proceed by concentrating on the written variety and the changes that have led to the difference between the two varieties.
M F
Contrastive
Temporal Causal
F
Comparative Similaritive
Figure 1. The use of kun in written and spoken Finnish
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.4 (266)
Ilona Herlin and Lari Kotilainen
. Kun in the spoken variety of Finnish In modern spoken Finnish, kun is used to denote various subordinating relations: temporal ((3)), causal ((4)), comparative ((5)) and similaritive ((6)). In spoken Finnish, kun mostly takes the form ku. (3) Anna soitti ku mä söin. Anna call-3sg.past while I eat-3sg.past “Anna called while I was eating” (4) Ei Anna tuu ku se on kipee. neg-3sg.pres Anna come because she be-3sg.pres ill-part “Anna won’t come because she is ill” (5) Tää on parempi ku toi. this be-3sg.pres good-comp than that “This one is better than that one” (6) Tää on yhtä hyvä ku toi. this be-3sg.pres as good as that “This one is as good as that one” The list below shows the translation equivalents of kun in English. Temporality: Causality: Comparison: Similarity:
when, while, as as, because than (parempi ku(n) “better than”) as (yhtä hyvä ku(n) “as good as”)
While kun is predominantly used as a conjunction, it is also occasionally used as an emphatic particle as a second position clitic: (7) Minä kun en kalaa syö! I emph neg-1sg.pres fish-part eat-inf “I certainly do not eat fish!” Although this use of kun is clearly not subordinating, it is semantically connected with the causal uses of kun (see Herlin 1998: 141–145). In sum, instead of being a mere temporal conjunction with causal meanings, kun in spoken Finnish is an ambiguous marker of various subordinating relations and an emphatic particle.
. Kun in the written variety of Finnish We will now explore the use of kun in literary Finnish. From the 16th to the 19th centuries, the form cuin covered most usages in written Finnish that kun had in spoken Finnish. During the first centuries of the history of written Finnish, the
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.5 (267)
External factors behind cross-linguistic similarities
written and the spoken varieties were therefore similar, with their main difference lying in the form of the conjunction: kun in the spoken, cuin in the written variety. Thus, in the examples below, which are excerpts from the first Finnish Bible published in 1642, the form cuin functions both as a comparative (8) and as a temporal (9) conjunction. (8) Parembi on cuulla wijsan curitusta cuin good-comp be-3sg.pres hear wise-gen chastisement-part than (1642 Bible) cuulla hulluin lauluja. hear-inf fool-pl.gen song-pl.part “It is better to hear moans of the wise than it is to hear songs of fools” (9) Ja cuin hän nämät sanonut oli otettin hän ylös And when he these say-pcp be-past take-pass.past he up heidän nähdens ja pilwi wei hänen ylös heidän they-gen see-inf-ps and cloud take-3sg.past he-acc up they-gen (1642 Bible) silmäins edestä. eye-pl.gen front-elat “And when he had spoken these things, he was taken up while they beheld, and a cloud took him out of their sight” This system remained stable for centuries, until two major changes took place in the written variety and caused a split between the spoken and the written language during the 19th century. First, kun lost its comparative and similaritive uses. Second, kun acquired a contrastive use. Two questions arise: first, where did the contrastive use come from?, and secondly, how did the comparative and similaritive uses disappear? Concerning the first question, as mentioned previously, the form cuin had practically the same usage as kun had in spoken Finnish. But in the late 18th century, some writers began to use the form kun without i (Häkkinen 1994: 384). One reason for this may have been that non-scholastic writers were unfamiliar with the form cuin because it did not occur in their speech. The variation (kun/kuin) was subsequently eliminated during the standardisation of Finnish. A formerly ambiguous conjunction was split in two: kun as such and kuin with the extra i. Kun was to be used as a temporal or causal (or contrastive) conjunction. Kuin, on the other hand, was to be used in comparative and similaritive functions. This artificial norm was first recommended at the beginning of the 19th century in a reference grammar by von Becker (1824). However, during the 19th century, kun and kuin were mostly described as alternative forms for both temporal and comparative (and other) uses. Among educated writers, there were proponents of both kun and kuin. One of the most outstanding contributors to standard Finnish, Elias Lönnrot (e.g. 1839), used and recommended the form
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.6 (268)
Ilona Herlin and Lari Kotilainen
kun for all subordinate functions. His student G. E. Eurén on the other hand, the author of several influential grammar books (e.g. 1852), followed the tradition of old written Finnish and used the form kuin exclusively. Finally, the authoritative reference grammar by E. N. Setälä (1898) succeeded in setting the norm, and since then, kun and kuin have had separate functions in written Finnish. There is, however, a ‘grey area’ between temporality and comparison, i.e. temporal comparison, which causes variation that shows that the artificial distinction has not completely been adopted. Example (10) shows this variation. According to the standard, one should write kuin, but the temporality of the complex conjunction niin kauan kuin (“as long as”) leads one often to write kun: (10) Syö niin kauan kuin [/kun] ruokaa riittää. Eat-2sg.imp as long as food-part be enough-3sg.pres “Eat as long as there is food” The late 19th century grammar books played an important role in standardising the Finnish language (see Laitinen, this volume). The authors of the grammar books all knew Swedish, as well as other Indo-European languages, and the split between kun and kuin can indeed be seen as an indirect result of language contact. For example, Swedish has a temporal conjunction när “when”, a comparative än “than” and som “as” to mark similarity relations. These Swedish constructions therefore provided a model for distinguishing between temporal conjunctions on the one hand and comparative conjunctions on the other. The standardisation thus brought kun closer to the conjunctions of (familiar) Indo-European languages. To give a more complete picture of the status of kun in modern written Finnish, the rise of its contrastive use (as in (11)) also needs to be explained. (11) Kun ruotsalaisnaapuri toivoi pari oksaa While the Swedish neighbour hope-3sg.past a couple branch-part sahattaviksi, suomalainen kaatoi koko puun. to sew-pcp the Finn cut-3sg.past whole tree-acc “While the Swedish neighbour hoped for a couple of branches to be sawed, the Finn cut down the whole tree” This example contains two clauses that express simultaneous situations. In addition to the simultaneity, there is clearly a certain contrastivity between the processes. However, contrastivity has been conventionalised to the extent that there no longer needs to be temporal overlap between the two situations (12): (12) Kun 60-luvulla vielä oli tavallista matkustaa while 1960-ade still be-3sg.past common travel-inf kansipaikalla, on nyt useimmilla oma hytti. deck-ade be-3sg-pres now most (people)-ade own cabin.
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.7 (269)
External factors behind cross-linguistic similarities
“While in the ’60s it was still common to have deck accomodation, most travelers now have a cabin of their own” The change illustrated by the above modern Finnish examples did not occur until the 19th century. This is an example of an extension of meaning, which could be explained by conventionalisation of implicatures, i.e. conventionalisation of contrastive implicatures. The contrastive meaning of kun, peculiar to the written variety of Finnish, has arisen under the influence of other languages. Several Indo-European languages – the source languages of translations into Finnish – have temporal subordinators that also have contrastive uses (English while is an example, Swedish medan “while, whereas” is another). Both changes described in this section have brought kun closer to the conjunctions of the familiar Indo-European languages. Moreover, they have both been affected by language-external factors: language contact and standardisation. These changes have also led to a situation in which using modern reference grammars only may lead to false diachronic assumptions, because these grammars do not mention that kun originally had comparative and similaritive uses, too. Whereas the (early) history of kun is beyond the scope of this article, it may be noted in passing, however, that the use of kun as an emphatic particle is a possible source for its causal uses (for a detailed discussion and for alternative explanations, see Herlin 1998: 183–195). It should also be observed that crosslinguistically, modal meanings such as comparativity and similaritivity are sources of both temporal and causal meanings (see Kortmann 1997: 188–204).
. The converb essa Among the European languages, Finnish has an exceptionally rich converb system and this is one of the typological characteristics that distinguishes Finnish from the languages of Western Europe (see I. Nedjalkov 1998). In the following, we will focus on one of the Finnish converbs, i.e. the essa form, which consists of an infinitival sign -e- and a locative case ending -ssa.3 When analysed in terms of referentiality, three types of converbs can be distinguished: same-subject converbs (the subject of the converb is always coreferential with the subject of the superordinate verb), different subject converbs (the subjects of two verbs are never coreferential) and varying-subject converbs (the converbs that may appear in both type of constructions) (e.g. V. Nedjalkov 1995: 110–111). The essa converb is of the varying-subject type. The same-subjectness is most often marked with the possessive suffix on the converb as in (13) and differentsubjectness is marked with an NP in the genitive case as in (14):
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.8 (270)
Ilona Herlin and Lari Kotilainen
(13) Hän sulki oven lähtiessään. She/he close-past door-acc leave-essa-ps “She/he closed the door when she/he left” (14) Hän lähti auringon laskiessa. She/he leave-past sun-gen go down-essa. “She/he was gone at sunset” In modern literary Finnish, the converbial affix -essa can be attached to all transitive and intransitive verbs. Nevertheless, the range of the essa uses has not been stable during the time of literary Finnish. Next, we will concentrate on the changes essa has undergone.
. Changes in the use of essa in literary Finnish The main function of the converb essa is to mark the temporal overlap of two events. In fact, this is its only use in spoken Finnish and it was also clearly the dominant use in old written Finnish. However, in modern literary Finnish, the essa infinitive has a wider range of usage and marks causal, conditional and contrastive relations. Figure 2 presents an overview of the changes which the essa infinitive has undergone in literary Finnish during approximately the last 500 years – the time of the written standard for Finnish. The starting point here, referred to as ‘Stage 1’, illustrates the current state in spoken Finnish and the dominant state of affairs in early written Finnish (ca. 1500 – ca. 1800). At Stage 1, the verb in the essa form denotes a process that has either been mentioned previously in the text or is otherwise well known. The latter type includes, for instance, the sunset in (14) above. However, the earliest Finnish texts indicate that writers began to use the temporal essa infinitive in a wider context. In other words, through this change, the essa infinitive was not only used to specify an already active process but essa began to denote the presence of new information. This is referred to as ‘Stage 2’ in Figure 2 and is exemplified in (15), with the English equivalent in (15’): (15) Ja hen lowui heiste / ia wlgosmeni caupungista Bethanian / ia oli sielle. Mutta homenelda palaitesans [to return-essa-ps] caupungihin / isosi hen. (Agricola 1548, Mt-21: 17–18)4 (15’) And he left them, and went out of the city into Bethany; and he lodged there. Now in the morning as he was returning into the city, he was hungry. (King James Bible 1769, Mt-21: 17–18) Converbs have been classified semantically into specialised and contextual converbs (for examples, see V. Nedjalkov 1995). The function of contextual converbs
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.9 (271)
External factors behind cross-linguistic similarities STAGE 1: A temporal adverbial
STAGE 2: A temporal adverbial with broadened uses
STAGE 3: Temporal uses
Causal uses
Conditional uses
Contrastive uses
Figure 2. Changes in the use of the essa converb
varies depending on the context. A good example is the English -ing, which has varying interpretations depending on its context. A specialised converb, on the other hand, has one or two functions regardless of the context in which it appears. At Stages 1 and 2, essa was a typical specialised converb: it had only one function, namely the temporal one, in all possible contexts. However, after Stage 2, essa gained a number of new functions. In modern written Finnish, the essa infinitive has causal uses, but these uses did not arise independently. At the same time, the use of essa has diversified: causal (16), conditional (17) and contrastive (18) uses emerged as well. Thus, the essa converb changed from a specialised converb into a contextual one. (16) Puuttuessaan kiistoihin sotilaallisesti YK:n intervene-essa-ps strife-pl.ill in a military manner UN-gen uskottavuus heikkenee. credibility undermine-3sg “The credibility of the UN is undermined by its military intervention” (17) Testin onnistuessa parannuksia käytetään seuraavassa Test-gen succeed-essa improvement-pl.part use-pass next-ine
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.10 (272)
Ilona Herlin and Lari Kotilainen
osakilpailussa. Grand Prix-ine. “If the test succeeds, the improvements will be used in the next Grand Prix” (18) Tannerin käydessä rajua vaalikamppailua Tanner (name)-gen do/run-essa fierce-part election campaign-part kehotti Skp boikotoimaan vaaleja. urge-past Skp (political party) boycott-inf election-part. “While Tanner had a fierce election campaign, the Communist party told people to boycott the election” These are the major changes in the use of essa during the history of literary Finnish. We will now discuss the factors underlying these changes. First, it is important to bear in mind that essa in spoken Finnish is still used as a temporal adverbial only. Thus, it was only literary essa that changed. We suggest that this is because the literary variant of Finnish has been influenced by other languages. In the following, we will show how particular changes in the use of essa can be attributed to certain contact situations that are specific to written Finnish only.
. Translation contacts Harris and Campbell (1995: 120) point out that syntactic borrowing is ‘perhaps the most neglected and abused area of syntactic change’. A basic claim about this type of borrowing is that if it is possible at all, the languages in contact should be at least typologically close. Harris and Campbell adopt an agnostic view on this ‘structural-compatibility requirement’. The evidence from the Finnish essa infinitive shows that when it comes to the evolution of a literary language, typologically very different languages can have a profound influence. In the case of essa, the change began as a simple borrowing of foreign structures, but the effects have been far-reaching for the whole system of essa in literary Finnish. Literary Finnish was created in the 16th century and was modelled heavily on the familiar written standards from Indo-European languages. Latin and Greek were particularly important source languages since the earliest written Finnish consisted mainly of translated religious texts. However, the infinitive verb systems of Latin and Greek differ substantially from the Finnish one. It has been argued by the Finnish linguist Lindén (1966), that the change from Stage 1 to Stage 2 (cf. Figure 2) can be linked to some specific constructions of Latin (namely, ablativus absolutus and participium conjunctum). Examples (19) and (20) are Latin sentences and their Finnish translation in the first Finnish New Testament (1548) by Michael Agricola.
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.11 (273)
External factors behind cross-linguistic similarities
(19) Et egredientibus illis ab Jericho, secuta est com turba multa. (Vulgate, ablativus absolutus) Ja heiden wlgoskieudesens Jerichost, and they-gen depart-essa-ps Jericho-ela (Agricola) seurasi hende palio Canssa. follow-past he-part much crowd. “And as they departed from Jericho, a great crowd followed him” (20) Mane autem revertens in civitatem, esuriit. (Vulgate, participium conjunctum) Mutta homenelda palaitesans caupungihin, isosi hen. But tomorrow return-essa-ps town-ill feel hunger-past he (Agricola) “But in the morning as he was returning into the city, he was hungry” A similar effect has been attributed to some Greek constructions. Examples (19’) and (20’) show the previous sentences in classical Greek. (19’) Kai ekporeuomen¯on aut¯on apo Ierich¯o, ¯ekolouth¯esen aut¯o okhlos polys. (genetivus absolutus) (20’) Pr¯oias de epanag¯on eis t¯en polin, epeinase.
(participium conjunctum)
Itkonen-Kaila (1991) has shown that the influence of Greek was direct, i.e. not only via Latin. For instance, she cites the Greek structure en t¯o + infinitive (roughly “in” + an article + infinitive, for example en t¯o einai “in the being”), which has been translated with the essa infinitive in the first Finnish Bible. As seen in (21), Vulgate uses temporal clause in the same Bible passage. (21) En de t¯o katheudein tous anthr¯opous, ¯elthen autou ho ekhthros – –. (Greek, en t¯o + infinitive) – – cum autem dormirent homines venit inimicus eius – –. (Vulgate, a temporal clause) Mutta inhimisten maates, tuli henen wihoilissens, ia but people lie-essa come-past his enemy-ps and (Agricola 1548) kylui ohdakeet nisuijn sekan – –. sow-past thistle-pl.acc wheat-gen among “But while people were sleeping, his enemy came and sowed thistle among the wheat” From examples (19) to (21) it becomes clear how the broadening of the possible contexts of temporal essa could occur: (1) certain constructions are translated with the essa form, (2) the essa converb adopts the discourse functions of these source language constructions and (3) the new use of essa emerges.
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.12 (274)
Ilona Herlin and Lari Kotilainen
Later, Finnish has also been in intense contact with Swedish and English. The effect of Swedish has been the most influential factor in the process where essa has acquired new uses in the 19th century. This will be discussed in detail in the next section, but first we will take a look at the possible effect of English, which seems to be very similar to that of Latin and Greek. The further change of the essa infinitive from a specialised to a contextual converb (i.e. from Stage 2 to Stage 3) may have been influenced by the English converb -ing, a typical contextual converb. This form can be translated into Finnish using a converb in the essa form, as in (22) found in an English textbook: (22) Nähdessään, että kaikki oli tehty, Harry lähti see-essa-ps that everything be-past do-pcp Harry go-past kotiin. home-ill “Seeing that everything had been done, Harry went home” Ebert (1978: 16f.) has argued that loan syntax can not only be qualitative, but also quantitative: the construction of a foreign language may cause an increase in the frequency of use of a native construction. In this case, the statistical effect of English has been documented: Eskola (2002: 139) shows that in modern Finnish translations from English, the essa infinitive is twice as frequent as in the original Finnish texts. She also maintains that essa is most often used as a translation equivalent of -ing (p. 77). Therefore it seems likely that the vagueness of English -ing has affected Finnish essa. And, indeed, this fits into the picture of the development of essa. The rise of the use of non-temporal essa converbs can be attributed to quite recent times – the latter half of the 20th century. At the same period, the importance and influence of English has grown. Moreover, there seems to be a correlation between the frequency of non-temporal essa converbs and the text types and social environments in which the influence of English is strong, e.g. on the Internet. (See Herlin & Kotilainen forthcoming for a more detailed discussion on English influence on essa.)
. The influence of Swedish It is not only converbs of other languages that have affected the changes of essa. Swedish is clearly not a converb language. Nonetheless, Swedish has had a clear effect on the change from Stage 2 to Stage 3 (Figure 2). The change is also intertwined with the interplay between the conjunction kun and the converb essa. While Swedish has no structural counterparts to Finnish converbs, the essa converb is used as the translation equivalent of conjunctions such as då “when”, när “when, because” and medan “while, whereas” . The confusing factor here is
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.13 (275)
External factors behind cross-linguistic similarities
that the conjunction kun is also the equivalent for these Swedish conjunctions. In other words, both Finnish kun and the Swedish conjunctions mentioned above are used to denote not only temporal but also causal and other relations. Example (23) shows Swedish då, which is used causally, and its possible translations: (23) Swedish:
Då han hörde att fadern ogillade planen övergav han den. Finnish with essa: Kuullessaan, että isä ei pitänyt suunnitelmasta, hän lopetti. Finnish with kun: Kun hän kuuli, että isä ei pitänyt suunnitelmasta, hän lopetti. English: When he heard that his father did not like the plan, he abandoned it.
In modern written Finnish, the original differences between the use of kun and the use of the essa converb (cf. also Section 2) have diminished. This can clearly be attributed to translation effects as well as to reference grammars. For example, in 1871, Jahnsson wrote a grammar in Swedish which uses the Swedish conjunction då to describe the meaning of both essa and kun. In his authoritative reference grammar, E. N. Setälä (1891), states the similarity between kun and essa even more clearly when he writes that essa is used as an equivalent to temporal clauses that use kun. In part, this has led to the present state of affairs where kun and essa are semantically close to each other in written Finnish yet clearly different in spoken Finnish.
. Summary and theoretical implications The following figure summarises the changes in the range of use of our two Finnish constructions in the written language.
the essa infinitive the conjunction kun
the the essa infinitive conjunction kun
the essa infinitive
kuin
kun
Figure 3. The initial range of use of essa and kun in early Finnish, change in use and current use in modern written Finnish
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.14 (276)
Ilona Herlin and Lari Kotilainen
Before the changes, the kun conjunction had a significantly wider range of use than the essa infinitive. Throughout the history of literary Finnish this infinitive developed a wider range of use (broadened temporality, causal and conditional uses)5 while the range of uses of the conjunction kun diminished (i.e. split into kun and kuin). The current situation in modern written Finnish is that the differences between the essa infinitive and the conjunction kun have been reduced considerably. While showing the main lines of the development of two seemingly synonymous constructions, we have paid particular attention to the developments in written Finnish and to factors underlying them. As we have pointed out, this change has mostly been caused by external factors such as language contact and standardisation.6 Thus, as kun was made a more specific conjunction, essa developed towards a more typical converb (when compared to familiar, i.e. Indo-European, languages) and kun and essa were made synonyms in grammatical descriptions, these two constructions became almost equivalent in meaning. To summarise, we have shown that the fact that both kun and essa happen to denote temporal as well as causal meanings is not a manifestation of a universal cognitive-pragmatic principle whereby meanings tend to develop from less abstract to more abstract, e.g. from temporal to causal. The polysemy of a construction does not allow one to conclude that an underlying cognitive principle has caused it. In the case of essa, the rise of causality may be seen as a part of a wider change caused by both language-internal and language-external factors. In the case of kun, the causal meaning need not have anything to do with temporality. What is common to these two constructions is that the changes they have undergone are a symptom of a wider typological change which has brought Finnish closer to the Indo-European languages. Several authors have noted (see, e.g. Tauli 1966) that various Uralic languages (including Finnish) have had the tendency to replace verb-nominal constructions with subordinate clauses of the dominant Indo-European type. The tendency we have described here is different, but related and shows that the influence of Indo-European languages is more complicated: the infinitival/converbal (‘verb-nominal’) constructions and subordinate clauses have (themselves) become closer in use to their ‘equivalents’ in Indo-European languages.
Abbreviations ACC ADE CND COMP
accusative adessive conditional mood comparative
INE INF neg NEGCL
inessive infinitive negation negative clitic
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.15 (277)
External factors behind cross-linguistic similarities
CONS ELAT EMPH ESS essa GEN ILL IMP
concessive elative emphatic particle essive essa infinitive genitive illative imperative
PART PASS PAST PCP PL PRES PS SG
partitive passive past participle plural present possessive suffix singular
Notes * We would like to thank Lea Laitinen and Heli Pekkarinen for valuable comments on an earlier version of this paper. . The concept ‘converb’ is chosen here instead of ‘infinitive’ to help the comparison between our example constructions. Typologically, converbs and conjunctions are often seen as semantically analogous. For example, V. Nedjalkov (1995:100f.) has suggested that ‘we can group together conjunctions and converbs semantically, the main difference being their formal coding.’ However, the comparison between infinitives and conjunctions is far less obvious. For instance, not all Finnish infinitives are used to mark adverbial subordination. . The importance of Swedish in Finland dates back to the 12th century. Sweden ruled Finland until 1809, and Swedish has remained an official language of Finland. Finnish, on the other hand, did not acquire the status of a fully official language of Finland before the year 1883 (for a more detailed discussion on the role of Swedish in Finland, see Embleton 2001). . The Finnish infinitive verb system is divided into three main categories on morphological grounds: the A infinitive, the E infinitive and the MA infinitive (also called first, second and third infinitive of Finnish). These three are infinitive suffixes that can be added to any verbal stem. These three infinitives are further divided into subcategories, according to the case endings they can take: the A infinitive appears in two different cases, the MA infinitive in six cases and the E infinitive in two cases. The different case variants of the same main category (A, MA, or E) are quite specialized, i.e. they may differ syntactically and/or semantically. For example, the two forms of the E infinitive – the instructive and the inessive case variant (essa) – are both used as adverbials, but the instructive case variant expresses mainly manner and the inessive case variant expresses mainly temporality. It should be noted, also, that not all Finnish infinitives can be counted as converbs: the lative case variant of the A infinitive does not express ‘adverbial subordination’, but is used in subject and object positions. . Agricola mainly used the form esa, not essa. In the 16th century, a standard orthography had not yet been set. Incidentally, Agricola is usually referred as ‘the father of written Finnish’. . It is interesting to note that the cognate converb -es in Estonian, a closely related language, has even more uses: in addition to temporality, causality and conditionality, it may express purpose and concession (EKG). This is most likely due to the influence of its neighbouring language, Russian, a typical contextual converb language.
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.16 (278)
Ilona Herlin and Lari Kotilainen . It is evident that in the history of Finnish language contact and standardisation are closely interrelated. An example of their interplay is that other languages are not only sources of borrowing, but are also models for standardisation.
References Becker, Reinhold von (1824). Finsk grammatik [The Grammar of Finnish]. Åbo. Bybee, Joan, Revere Perkins, & William Pagliuca (1994). The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: The University of Chicago Press. Campbell, Lyle (1991). “Some Grammaticalization Changes in Estonian and their Implications”. In Elizabeth Closs Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization, Vol. I (pp. 285–299). Amsterdam: Benjamins. Campbell, Lyle (2001). “What’s Wrong with Grammaticalization?” Language Sciences, 23, 113–161. Ebert, Robert Peter (1978). Historische Syntax des Deutschen. Stuttgart: Metzler. EKG = Eesti keele grammatika [The Grammar of Estonian] (1993). Tallinn: Eesti Teaduste Akadeemia Keele ja Kirjanduse Instituut. Embleton, Sheila (2001). “Bilingualism in Contemporary Finland: whither Swedish?” Journal of Finnish Studies, 5:2, 7–30. Eskola, Sari (2002). Syntetisoivat rakenteet käännössuomessa. Suomennetun kaunokirjallisuuden ominaispiirteiden tarkastelua korpusmenetelmillä [Synthesising Structures in Translated Finnish. A Corpus-based Analysis of the Special Features of Finnish Literary Translations]. Joensuu: University of Joensuu. Eurén, G. E. (1852). Suomalainen kielioppi suomalaisille [A Finnish Grammar for the Finnish People]. Åbo. Harris, Alice C. & Lyle Campbell (1995). Historical Syntax in Cross-linguistic Perspective. (= Cambridge Studies in Linguistics 74). Cambridge: Cambridge University Press. Haspelmath, Martin (1995). “The Converb as a Cross-linguistically Valid Category”. In Haspelmath & König (Eds.), 1–56. Haspelmath, Martin (1999). “Why is Grammaticalisation Irreversible?” Linguistics, 37, 1043–1068. Haspelmath, Martin & Ekkehard König (Eds.). (1995). Converbs in Cross-Linguistic Perspective. Structure and Meaning of Adverbial Verb Forms – Adverbial Participles, Gerunds. Berlin: Mouton de Gruyter. Heine, Bernd & Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Herlin, Ilona (1998). Suomen kun [Finnish kun]. Helsinki: Finnish Literature Society. Herlin, Ilona & Lari Kotilainen (forthcoming). “Itsenäistyvä infinitiivi. essa-infinitiivin kehitys kirjakielen aikana” [The Development of the essa Infinitive in Literary Finnish]. In Ilona Herlin & Laura Visapää (Eds.), Suomen kielen infiniittisten verbirakenteiden historia [The History of the Infinitival Verb Constructions in Finnish].
TSL[v.20020404] Prn:25/02/2004; 9:10
F: TSL5913.tex / p.17 (279)
External factors behind cross-linguistic similarities
Häkkinen, Kaisa (1994). Agricolasta nykykieleen [From Agricola to Contemporary Finnish]. Helsinki: WSOY. Itkonen-Kaila, Marja (1991). “Agricolan Uuden testamentin temporaalirakenteet: suomi klassisten ja germaanisten kielten taustaa vasten” [The Temporal Construction in the Agricola’s New Testament: Finnish against the Background of Classic and Germanic languages]. Virittäjä, 95, 255–280. Jahnsson, A. W. (1871). Finska språkets satslära [The Syntax of Finnish]. Helsinki. Joseph, Brian D. & Richard Janda (1988). “The how and why of Diachronic Morphologization and Demorphologization”. In Michael Hammond & Michael Noonan (Eds.), Theoretical morphology: Approaches in modern linguistics (pp. 193–210). New York: Academic Press. Kortmann, Bernd (1997). Adverbial Subordination. A Typology and History of Adverbial Subordinators Based on European Languages. Berlin: Mouton de Gruyter. Lindén, Eeva (1966). “Latinan vaikutus temporaalirakenteeseen suomen kirjakielessä” [The Effect of Latin to the Temporal Construction in Literary Finnish]. Virittäjä, 70, 1–8, 339–347. Lönnrot, Elias (1839). “Graece sunt; non leguntur”. Mehiläinen 10/1839. Nedjalkov, Igor V. (1998). “Converbs in the Languages of Europe”. In Johan van der Auwera (Ed.), Adverbial Constructions in the Languages of Europe (pp. 421–456). Berlin: Mouton de Gruyter. Nedljakov, Vladimir P. (1995). “Some Typological Parameters of Converbs”. In Haspelmath & König (Eds.), 97–136. Ramat, Anna Giacalone (1998). “Testing the Boundaries of Grammaticalization”. In Anna Giacalone Ramat & Paul J. Hopper (Eds.), The Limits of Grammaticalization (pp. 107– 127). Amsterdam & Philadelphia: John Benjamins. Setälä, E. N. (1891). Suomen kielen lauseoppi [The Syntax of Finnish]. Oppikouluja varten (3 edition). Helsinki: K. E. Holm. Setälä, E. N. (1898). Suomen kielioppi [The Finnish Grammar]. Äänne- ja sanaoppi oppikouluja ja omin päin opiskelua varten. Helsinki: Otava. Tauli, Valter (1966). Structural Tendencies in Uralic languages. The Hague: Mouton. Traugott, Elizabeth C. & Ekkehard König (1991). “The Semantics-pragmatics of Grammaticalization Revisited”. In Elizabeth Closs Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization I (pp. 189–218). Amsterdam & Philadelphia: John Benjamins. Virittäjä = The journal Virittäjä. Published by the Society for the Study of Finnish (Kotikielen Seura).
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.1 (281)
What constitutes a case of grammaticalization? Evidence from the development of copulas from demonstratives in Passamaquoddy Eve Ng University of Pittsburgh
.
Introduction
Grammaticalization has been characterized as a phenomenon that may involve semantic, syntactic, morphological, and phonetic changes (e.g. see Heine & Reh 1984: 67; Bybee & Pagliuca 1985; Hopper & Traugott 1993: 2–4). In many of the well-known examples, all of these changes can be seen. For example, the complementizer that in English developed from a demonstrative that. Complementizer that is no longer a referring term; it is restricted syntactically to a position at the beginning of a subordinate clause; only the singular distal form can be used as a complementizer; and it is commonly pronounced with a reduced vowel, /ð6t/. Since grammaticalization occurs over time, however, it is typical that at the earlier points of the process, only some linguistic changes will have occurred. Thus, semantic and syntactic development may be evidenced without accompanying changes in either phonetic form or inflectional behavior. Alternatively, there may be changes in semantic, syntactic, and inflectional characteristics without reduction in phonetic form. In well-documented cases of grammaticalization, such as the development of the Late Latin verb habere ‘have’ into future suffixes in Romance (see Harris 1978; Fleischman 1982; Vincent 1982; Pinkster 1987), it is relatively unproblematic to propose that when the forms are showing only semantic and syntactic changes, this is an incipient period of grammaticalization. However, when there are insufficient written records, it is less obvious how to treat some item X which does not show phonetic or inflectional differences compared to the suspected source item Y, since in such cases, we cannot be im-
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.2 (282)
Eve Ng
mediately confident that item X is not simply evidencing part of the functional range of the supposed source item. For example, in many languages, the forms with the functions generally associated with demonstratives are the same as the forms used as regular third person pronouns, with no difference in inflectional behavior (e.g. see Greenberg 1978). Although there are languages where the historical record shows that third-person pronoun forms subsequently diverged phonetically from the demonstratives (such as the development of third-person pronouns in the Romance languages from Latin demonstratives), in other languages there is no sign of such a development, and it seems clear that the latter situation can be stable for some time. In this paper I address the issue of assessing whether grammaticalization has taken place when we are confronted with this sort of data. I will do so by examining the occurrence of a number of demonstrative-forms in clauses with nominal predicates in Passamaquoddy, an Algonquian language of the United States and Canada. I use the label ‘demonstrative-form’ to refer to a morpheme which shares its phonological form with an item that in other contexts is used for deictic reference; however, I argue that the demonstrative-forms I will discuss no longer necessarily have the functions commonly associated with demonstratives. Rather, these demonstrative-forms can be analyzed as having developed the function and syntactic characteristics of copulas, with variation in their inflectional behavior depending on the type of verbless construction in which they occur. In one construction, the demonstrative-form still inflects fully for number and animacy; in other constructions, the demonstrative-form inflects for number and animacy in the plural but not in the singular; and in still other constructions, the demonstrative-form does not inflect for number and animacy in either the singular or the plural. I will suggest that it is crucial to look at how the range of clauses with nominal predicates behaves in the language in order to determine the analysis for the verbless clause type where the demonstrative-form still has the inflectional and phonetic characteristics associated with demonstratives more generally. I will conclude that, while early stages of grammaticalization may pose certain problems for analysis, a careful consideration of the syntactic properties of the construction in question along with the characteristics of related constructions proves valuable in determining whether grammaticalization has occurred.
. An introduction to demonstrative-forms in Passamaquoddy Like other Algonquian languages, Passamaquoddy has a large paradigm of demonstratives that are available for use to refer to entities such as people, animals, objects, and places. These demonstratives are morphologically differentiated for
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.3 (283)
What constitutes a case of grammaticalization?
Table 1. Passamaquoddy non-absentative demonstrative-forms
Animate
sg pl
Inanimate
sg pl
proximate obviative proximate obviative
Proximal
Distal
Remote
wot yuhtol yuktok, yukt, yukk yuhuht yut yuhtol
not nihtol niktok, nikt, nikk nihiht nit nihtol
yat yehtol yektok, yekt, yekk yeheht yet yehtol
three deictic distances (proximal, distal, and remote), number, animacy, obviation (obviative value is associated with reference to a less topical entity) for animate forms, and absentativity (absentative value is associated with reference to an entity formerly present but currently absent, or formerly but not currently owned). Table 1 gives only the non-absentative demonstrative-forms of Passamaquoddy.1 When demonstratives are used to refer deictically or anaphorically to entities, any of the forms in Table 1 can be used, provided they agree with the referent for the relevant inflectional categories. In Sections 3–4, I will discuss verbless constructions in Passamaquoddy which contain demonstrative-forms that can be analyzed as copulas, where, following authors such as Hengeveld (1992), a copula is defined as a morpheme that creates a predication when the predicate is non-verbal.2 The demonstrative-forms which I argue have the identity of copulas I call demonstrative-copulas. Demonstrativecopulas make use of only a subset of the demonstrative paradigm – they are all morphologically non-absentative, non-obviative, distal forms. I will also use the label term to refer to the nominal expressions in the verbless constructions, whether they are pronouns or lexical NPs, and whether they are serving as arguments or predicates. In the interests of space, I will only give data for affirmative sentences.3
. Demonstrative-forms in a verbless construction with two lexical NP terms In this section, I first present data for a verbless construction which contains two lexical NPs and a demonstrative-form, and then consider three different analyses for the demonstrative form: a demonstrative pronoun, a demonstrative-copula, or a demonstrative-form belonging to the same grammatical class as demonstrative pronouns but functioning as a copula. In sentences of this type, there is an argument term and a predicate term. A demonstrative-form agreeing in animacy and number with the argument term occurs between the two terms.4 When the predicate term is non-referential, the
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.4 (284)
Eve Ng
sentence has a classificatory meaning, e.g. Nomehs not polam “A salmon is a fish” or Tepit not taktal “David is a doctor.” (Constructions where the predicate term refers to a kind have sometimes been called ‘predicate nominal’ constructions, where the semantically more general term, e.g. nomehs “fish” or taktal “doctor” in the examples here, is the predicate nominal.) In other clauses, the predicate term is referential, such as the nominal amsqahsewey meaning “the first one” in Piley nit amsqahsewey “The new one is the first one.” In the examples below, I have underlined the demonstrative-form. Generally speaking, when both terms are non-referential, the terms can occur in either order. When the argument term is referential, there is a strong preference for the order of the sentence to be [argument term]-[demonstrative-form]-[predicate term]. In the interests of space, not all possible word orders are given for all examples below. (1)–(4) are examples with singular terms. An inanimate singular demonstrativeform nit occurs in (1)–(2) where the argument term tomhikon “axe” is inanimate singular, while (3)–(4) are examples with an animate singular argument term polam “salmon”, so the demonstrative-form is the animate singular form not. (1) Tomh-ikon nit wehke-w-akon. beat.ai-nmlz.inan inan.sg.dist use.ti-der-nmlz.inan “An axe is a tool.” or (2) Wehke-w-akon nit tomh-ikon. use.ti-der-nmlz.inan inan.sg.dist beat.ai-nmlz.inan “An axe is a tool.” (3) Polam not nomehs. salmon.an an.sg.dist fish.an “A salmon is a fish.” or (4) Nomehs not polam. fish.an an.sg.dist salmon.an “A salmon is a fish.” (5)–(6) are examples with plural terms. In (5), the demonstrative-form is the inanimate plural form nihtol, which agrees with the inanimate plural argument term kompiyutawol “computers.” In (6), the demonstrative-form is the animate plural form niktok, agreeing with the plural argument term Mali naka Tepit “Mary and David.”
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.5 (285)
What constitutes a case of grammaticalization?
(5) Kompiyuhtaw-ol nihtol pili mosin-ol. computer.inan-pl inan.pl.dist new machine.inan-pl “Computers are new machines.” (6) Mali naka Tepit nikt taktal-ok. Mary and David an.pl.dist doctor.an-pl “Mary and David are doctors.” Although the two terms usually match in animacy in these constructions, occasionally they do not. For example, in (7)–(8), the argument maltuhsis “hammer” is animate while the predicate wehkewakon “tool” is inanimate. We see that the demonstrative-form is the animate form not, agreeing with the argument, not the predicate; if the demonstrative-form were the inanimate demword form nit, the results would be ungrammatical.5 (7) Maltuhs-is not wehke-w-akon. hammer.an-dim an.sg.dist use.ti-der-nmlz.inan “A hammer is a tool.” or (8) Wehke-w-akon not maltuhs-is. use.ti-der-nmlz.inan an.sg.dist hammer.an-dim “A hammer is a tool.” In some ways, the demonstrative-form in this construction looks like some sort of anaphoric or cataphoric pronoun forming a clause with the predicate nominal. For example, if not in (7) were analyzed as an anaphoric pronoun referring back to maltuhsis “hammer”, the more literal interpretation would be something like “Hammer, that [is] a tool.” Similarly, if not is analyzed as a cataphoric pronoun in (8), then the more literal interpretation is something like “That [is] a tool, hammer.” The structures for such interpretations of (7) and (8) are given in (9) and (10), in which the demonstrative-form not forms a clause with predicate term wehkewakon “tool”, while maltuhsis “hammer” is some sort of adjunct occurring before or after this clause. (9) [Maltuhs-is] [not wehke-w-akon]CLAUSE . hammer.an-dim an.sg.dist use.ti-der-nmlz.inan “A hammer is a tool.” (more literally, “Hammer, that [is] tool.” or “Hammer, it [is] tool.”) (10) [Wehke-w-akon not ]CLAUSE [maltuhs-is]. use.ti-der-nmlz.inan an.sg.dist hammer.an-dim “A hammer is a tool.” (more literally, “That [is] tool, hammer.” or “It [is] tool, hammer.”)
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.6 (286)
Eve Ng
Analyzing the demonstrative-form as a pronoun means that two clause structures are proposed: [argument-predicate] as in (9) and [predicate-argument] as in (10). This in itself is not a particular problem; although the [predicateargument] order is usual for clauses with a lexical NP predicate and a pronominal argument, the opposite order can also occur. However, if the demonstrative-form is a pronoun, forming a clause with the lexical NP predicate, then there is no obvious reason why example (11), which under this analysis would consist of a [predicateargument] clause wehkewakon not “that [is] tool” preceded by maltuhsis “hammer”, is unacceptable. Example (12), which under an analysis of the demonstrativeform as being a pronoun would consist of the [argument-predicate] clause not wehkewakon “that [is] tool” followed by maltuhsis “hammer”, is also unacceptable. (11) *Maltuhs-is wehke-w-akon not. hammer.an-dim use.ti-der-nmlz.inan an.sg.dist (12) *Not wehke-w-akon maltuhs-is. an.sg.dist use.ti-der-nmlz.inan hammer.an-dim Also, if maltuhsis “hammer” is actually outside of the clause as shown in the structures (9) and (10), then it would not be surprising to find some kind of phonological cue that maltuhsis is outside of the posited clause. However, there is no evidence that this is the case; (7) and (8) do not have different intonational patterns, with, say, a pause after maltuhsis in (7) and a pause before maltuhsis in (8). In summary, the demonstrative-form in this type of construction shows certain differences from demonstratives that are used to refer pronominally to entities. First, it is syntactically restricted to the position between the two terms. Second, such a demonstrative-form must always be a distal form, while entityreferring demonstratives make use of the entire demonstrative paradigm. Thus, the demonstrative-forms discussed here show a decrease in syntactic variability as well as a reduction in the number of items in the paradigm, both characteristics associated with grammaticalization. Demonstratives have been documented to grammaticalize into copulas in other languages, including Mandarin, Hebrew, and Panare, through either topiccomment to subject-predicate reanalysis or comment-afterthought to predicatesubject reanalysis. Thus, Li and Thompson (1977) propose that a sentence that originally had the structure [argumenti]TOPIC – [demonstrativei – predicate nominal]COMMENT was reanalyzed as [argument]SUBJECT – [copula – predicate nominal]PREDICATE in Mandarin and Hebrew. Diessel (1999: 143– 148) argues that the facts of Hebrew support a slightly different analysis, where the demonstrative originally agreed with the predicate nominal, not the topic term: [argumenti]TOPIC – [demonstrativei – predicate nominali]COMMENT → [argument]SUBJECT – [copula – predicate nominal]PREDICATE . Finally, Gildea
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.7 (287)
What constitutes a case of grammaticalization?
(1993) observes that in a language like Panare, which is predicate-initial, a mirror-image analysis to Li and Thompson’s (1977) account fits the data better: [predicate nominal – demonstrativei] – [argumenti]AFTERTHOUGHT → [predicate nominal – copula] – [argument]SUBJECT . In the Passamaquoddy data, we saw that the demonstrative word agrees with the argument term, not the predicate term, thus ruling out Diessel’s analysis. However, both Li and Thompson’s (1977) and Gildea’s (1993) accounts are available for Passamaquoddy, depending on the word order of the clause, if the Passamaquoddy demonstrative-form is indeed a copula. An alternative analysis would be that such demonstrative-forms belong to the same grammatical class as demonstratives used to refer to entities, but that demonstrative-forms like nit in (1) and (2) or not in (3) and (4) are occurring in a context commonly associated with copula function. Such an analysis might be argued to be analogous to treating English wh-words occurring as interrogative elements, as in (13)–(16), as belonging to the same class as wh-words that mark a relative clause, as in (17)–(20). Interrogative wh-words do not have identical functions or distributions to relative wh-words, but since all wh-words can occur in both syntactic contexts, one plausible analysis is that the functional range of a single class of wh-words encompasses both uses. (13) (14) (15) (16)
Who did you see? Where did they go? What happened today? Why are we here?
(17) (18) (19) (20)
They didn’t believe who I had seen. I think I know where they went. We didn’t see what happened. You will be told why you are here.
Under this sort of analysis, then, ‘demonstratives’ in Passamaquoddy have a larger functional range than is usually associated with demonstratives; the restriction on the syntactic position of the demonstrative in the verbless clauses discussed here would be argued to be dictated by the construction, and the limitation of the demonstrative to the distal forms is explained by positing that in more uncommon linguistic contexts (verbless clauses are relatively rare in verb-centric languages like Passamaquoddy and other Algonquian languages), it is the most unmarked forms of a paradigm that will be used, which are the distal forms in a three-distance demonstrative system. In order to further consider the merits of the different analyses given here, in the following section I present some additional Passamaquoddy data of demonstrative-forms occurring in other types of verbless clauses. I will show that in those constructions, the demonstrative forms no longer agree consistently in number and animacy with the terms, and thus show further differentiation from the behavior of entity-referring demonstratives.
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.8 (288)
Eve Ng
. Demonstrative-forms in other verbless clauses This section demonstrates that there are other verbless constructions where we find demonstrative-forms specific to those constructions. In 4.1, I present constructions where the demonstrative-form agrees for number and animacy with the terms in the clause only when they are plural, and in 4.2, I show a construction where the demonstrative-form fails to agree with the terms in the clause for number or animacy.
. Clauses where the demonstrative-form agrees in the plural but not in the singular There are three types of constructions where the demonstrative-form agrees for animacy and number in the plural, but not in the singular.
.. Clauses with one demonstrative term In these clauses, there is one focused demonstrative term followed by a demonstrative-form, and the resulting clause is translated as “X is the one” or “It is X”, where X is expressed by the demonstrative term. The demonstrative-form is always distal; the demonstrative term, on the other hand, can be any of the forms in the demonstrative paradigm. The demonstrative-form in sentences with singular terms is always the inanimate singular form nit; thus, this demonstrative-form fails to vary with the animacy of the term. Examples of such sentences are given in (21)–(22), with the demonstrative-form underlined and the focused term bolded. (21) Yut nit. inan.sg.prox inan.sg.dist “It’s this (one) [inan].” (or “This [inan] is the one.”) nit. (22) Wot an.sg.prox inan.sg.dist “It’s this (one) [an].” (or “This [an] is the one.”) In contrast, in affirmative sentences with plural terms, the demonstrative-form agrees in both animacy and number with the term. Examples are given in (23)– (24), again with the demonstrative-form underlined and the focused term bolded. (23) Yuhtol nihtol. inan.pl.prox inan.pl.dist
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.9 (289)
What constitutes a case of grammaticalization?
“It’s these (ones) [inan].” (or “These [inan] are the ones.”) (24) Yuktok niktok. an.pl.prox an.pl.dist “It’s these (ones) [an].” (or “These [an] are the ones.”)
..
Clauses with one personal pronoun term as focus and one lexical NP term as topic In these clauses, there are two terms: a personal pronoun as the focus term and a lexical NP as the topic term. The personal pronoun term occurs first and the nominal term last, and a demonstrative-form occurs between the two terms; the reverse order is ungrammatical.6 Personal pronouns are inherently animate, and the NP term in these sentences will also be animate. When the terms are singular, there is some variation amongst different speakers and occasionally from the same speaker as to whether the demonstrative-form agrees in animacy with the terms; hence, in the data below, I give both the inanimate nit and the animate not demonstrative forms as options. The demonstrativeform is underlined and the focused term bolded. In (25), the context is one where the speaker is seeking to identify the person who is a doctor amongst a group of people; (26) would be the corresponding response from another speaker who is the sought-after individual. Thus, taktal “doctor” is the topic term, and the personal pronoun is the focused term. (25) Kil nit / not taktal? 2sg inan.sg.dist / an.sg.dist doctor.an “Are you the doctor?” or “Are you a doctor?” (26) Nil nit / not taktal. 1sg inan.sg.dist / an.sg.dist doctor.an “I’m the doctor.” or “I’m a doctor.” When the terms are plural, the demonstrative-form agrees in both animacy and number with the terms. An example is given in (27). (27) Nilun nikt taktal-ok. 1plex an.pl.dist doctor.an-pl “We’re the doctors.” or “We’re doctors.”
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.10 (290)
Eve Ng
..
Clauses with one demonstrative term as focus and one lexical NP term as topic In these clauses, there are two terms: a demonstrative as the focus term and a lexical NP as the topic term. Comparably to the construction in 4.1.2, the demonstrative term occurs first and the lexical NP term last; the reverse order is ungrammatical.7 A demonstrative-form usually occurs between the two terms.8 In sentences with singular terms, the demonstrative-form is the inanimate singular form nit, regardless of whether the terms are inanimate (e.g. mitsut “fork”) or animate (e.g. emqansis “spoon”), as illustrated by (28)–(29). The demonstrativeform is underlined and the focused term is bolded in the English translation. (28) Yut nit mitsut. inan.sg.prox inan.sg.dist fork.inan “This is the [only] fork.”/“This is a fork [not one of the spoons].” or “This is the fork [amongst many, that I was talking about].” nit emqansis. (29) Wot an.sg.prox inan.sg.dist spoon.an “This is the [only] spoon.”/“This is a spoon [not one of the forks].” or “This is the spoon [amongst many, that I was talking about].” With plural terms, the demonstrative-form generally agrees in both animacy and number with the terms, as shown in (30)–(31). In (30), the terms are animate and plural, so the demonstrative-form is the animate plural form nikt. In (31), the terms are inanimate and plural, so the demonstrative-form is the inanimate plural form nihtol. nikt emqansis-ok. (30) Yuktok an.pl.prox an.pl.dist spoon.an-pl “These are the spoons.”/“These are spoons [not the forks].” or “These are the spoons [amongst many, that I was talking about].” (31) Yuhtol nihtol mitsuti-yil. inan.pl.prox inan.pl.dist fork.inan-pl “These are the forks.”/“These are forks [not the spoons].” or “These are the forks [amongst many, that I was talking about].” It should be mentioned that plural sentences where the demonstrative-form does not agree with the terms occasionally occur in elicitation, as shown in (32), which was provided in the context of identifying the doctors as some people standing some distance away. The demonstrative-form is the inanimate singular form nit, which fails to agree with the animate plural terms, yektok “those (yonder)” and taktalok “doctors”.
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.11 (291)
What constitutes a case of grammaticalization?
(32) Yektok nit taktal-ok. an.pl.remt inan.sg.dist doctor.an-pl “Those (yonder) are the doctors.”
. Clauses where the demonstrative-form does not agree with the terms Recall that in 4.1.1, we saw data for a construction with one focused demonstrative term, which is followed by a demonstrative-form that agrees in number and animacy with the term only if the term is plural. This section shows a structurally similar construction, but where the one focused term is a personal pronoun. In this construction, the demonstrative-form which follows the term does not agree in animacy and number with either singular or plural terms; an inanimate singular demonstrative-form nit occurs regardless of the animacy and number of the term. The resulting clause is translated as “It is X” or “X is the one”, where X is expressed by the personal pronoun. Examples are given in (33)–(36), with the demonstrativeform underlined and the focused term bolded. In (33)–(34), the term is singular; in (35)–(36), the term is plural. (33) Nil nit. 1sg inan.sg.dist “It’s me.” (or “I’m the one.”) (34) Kil nit. 2sg inan.sg.dist “It’s you (sg).” (or “You’re the one.”) (35) Nilun nit. 1plex inan.sg.dist “It’s us.” (or “We’re the ones.”) (36) Kiluwaw nit. 2pl inan.sg.dist “It’s you (pl).” (or “You’re the ones.”)
. The relevance of information structure For certain of the clauses with two terms, it is interesting to contrast constructions that do not have a demonstrative-form with comparable constructions that do. Compare (37) with (38), and (39) with (40), where the focus term is in bold in the English translation:
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.12 (292)
Eve Ng
(37) focus topic Taktal nil. doctor.an 1sg “I’m a doctor.” (38) focus topic Nil nit / not taktal. 1sg inan.sg.dist / an.sg.dist doctor.an “I’m the doctor.”/“I’m a doctor.” (39) focus topic Emqansis wot. spoon.an an.sg.prox “This is a spoon.” (40) focus topic Wot nit emqansis. an.sg.prox inan.sg.dist spoon.an “This is the spoon.”/“This is a spoon.” We can see the (37) and (39) types of sentences as having the unmarked informational status situation. A lexical NP will inherently bear more semantic content than a demonstrative or personal pronoun; thus, the focus term is more likely to be a lexical NP, because focused information is not presupposed and hence is more likely to be encoded with expressions having high lexical semantic content. Conversely, the topic term is more likely to be a pronoun, since topic information is presupposed and hence more likely to be encoded with expressions having low lexical semantic content. In contrast, the (38) and (40) types of sentences have the opposite information status situation, where it is the pronoun that is the focus, and the lexical NP that is the topic. It is in the latter constructions, which have the more marked information status, that a demonstrative-form is required between the two NP terms. Note also that for the one-term constructions which have demonstrative-forms, discussed in 4.1.1 and 4.2, there is also a situation where a pronoun term is the focused one, which fits the profile of clauses with a pronoun as the focus term as containing a demonstrative-form.9
. Grammatical status of the demonstrative-forms in Section 4 The demonstrative-forms of Section 4 show properties comparable to the demonstrative-forms in constructions with two lexical NPs, as discussed in Section 3. First, they are restricted in syntactic position: for constructions with two terms, as in 4.1.2 and 4.1.3, the demonstrative-form occurs between the terms, while for constructions with one term, as in 4.1.1 and 4.2, the demonstrative-form occurs
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.13 (293)
What constitutes a case of grammaticalization?
after the term. Second, the demonstrative-forms are all distal forms. Thus, like the demonstrative-forms of Section 3, these demonstrative-forms show two characteristics associated with grammaticalization: compared to the entity-referring demonstratives, they have a reduction in syntactic variability and a reduction in the number of items in the paradigm. Furthermore, unlike the demonstrative-forms in the Section 3 construction, none of the demonstrative-forms in any of the constructions of Section 4 agree consistently with the terms in the clause. Also, in these constructions, the terms are not just lexical NPs, but also involve pronouns, including first-person and secondperson pronouns, for which analyses where the demonstrative-forms are anaphoric or cataphoric pronouns, along the lines of (9) and (10), seem unlikely. Thus, it is even harder to argue that they belong to the same word class as demonstratives used to refer to entities, which make use of the entire demonstrative paradigm and do not show the same types of distributional restrictions.
. Discussion The construction discussed in Section 3, with two lexical NPs, differs in a couple of respects from those discussed in Section 4. One is that for the Section 3 construction, the information status of the terms is not relevant with respect to whether a demonstrative-form occurs; the rule is simply that a verbless clause with two lexical NPs will have a demonstrative-form between them. Additionally, the demonstrative-forms of Section 3 agree consistently in number and animacy with the term(s) in the clause, unlike the demonstrative-forms in the constructions described in Section 4. First, regarding the fact that the information status of the terms does not influence whether or not there is a demonstrative-form in the Section 3 construction, this can be explained by the fact that there is no pronoun term in that construction, so whichever of the two possible topic-focus situations obtains, neither one is more intrinsically marked than the other. Second, it is true that the demonstrative-forms in Section 3 still look like entity-referring demonstratives inflectionally, if we set aside the reduction in the size of the paradigm that these demonstrative-form have with respect to the distance category (recall that all of the demonstrative-form must always be distal forms). If demonstrative-forms in all verbless constructions consistently showed the same inflectional properties and paradigmatic range as entity-referring demonstratives, the case would be stronger that they all belong to a single grammatical class. This, however, is not the case; even if the demonstrative-forms of Section 3 were to be analyzed as belonging to the same grammatical class as entity-referring
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.14 (294)
Eve Ng
demonstratives, we would still need to account for the demonstrative-forms of Section 4. Furthermore, let us return for a moment to the phenomenon mentioned earlier of demonstrative forms in some languages serving as demonstrative pronouns as well as third person pronouns. This is a functional range that can easily be argued makes sense, since whether the form is being used to make reference to unexpected, contrastive, or emphatic entities (e.g. see Himmelmann 1996 for a discussion of functions associated with demonstratives) or is simply anaphoric, the central function of reference is a characteristic. In contrast, classing together entity-referring demonstratives with demonstrative-forms functioning as copulas would place items used to refer together with items used to allow non-verbal expressions to serve as predicates; this is a grouping which lacks the naturalness of the first. What the demonstrative-forms in all of the constructions presented in Sections 3 and 4 share is their occurrence in verbless clauses; their syntactic position, which can be described as occurring after the first term in the clause and before the second term when there is one; and the morphological limitation to distal demonstrative forms. Thus, one analysis which accounts for these facts is that all of the demonstrative-forms are copulas which have developed from demonstratives, although they are at different stages with respect to the loss of inflectional behavior. There are a number of possibilities how Passamaquoddy demonstrativecopulas could have developed. They may have arisen from pronominal demonstratives independently in each of the constructions in which they are now found. Alternatively, a demonstrative-copula may have first developed in a smaller set of constructions, or in a single construction, and then spread in use to other verbless constructions; for example, demonstrative-copulas might have first developed in constructions with two terms and then come to be used in constructions with one term, or vice versa. Craig (1991) discussed ‘polygrammaticalization’ as a phenomenon where a single item grammaticalizes into more than one other kind of item. In the Passamaquoddy data being considered here, all of the demonstrativeforms show some evidence of development into copulas, but it appears that a number of related processes may now be occurring concurrently in the different types of verbless constructions such that the rate of loss of inflectional behavior has not been completely uniform. This is not surprising given our knowledge of how grammaticalization can occur; for example, the various developments of Romance pronouns and determiners from Latin demonstratives did not occur exactly simultaneously (e.g. see Harris 1978: 66–87, 97–111; Carrasquel 1996). It should also be mentioned that previously discussed pathways for the grammaticalization of demonstratives into copulas have focused on constructions which originally consisted of a clause containing the demonstrative, along with a term that is outside that clause – the topic term in Li and Thompson (1977) or the af-
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.15 (295)
What constitutes a case of grammaticalization?
terthought in Gildea (1993) – which then developed into two-term sentences where the original demonstrative is now a copula. However, any complete analysis for Passamaquoddy also needs to account for the development of the demonstrativecopulas in one-term constructions, presented in 4.1.1 and 4.2. For a more detailed discussion of the grammaticalization of demonstrative-copulas, see Ng (2002: 339– 365).
. Summary In this paper, I considered the question of determining whether grammaticalization has occurred by examining a range of Passamaquoddy verbless constructions containing demonstrative-forms that do not function to refer to entities. The demonstrative-forms in one of these constructions agree consistently in animacy and number with the terms in the clause, but demonstrative-forms in the other constructions do not. I argued that these inflectional differences are not unexpected if several processes of demonstrative-to-copula grammaticalization have been taking place, and that the demonstrative-forms in each of the constructions presented share other similarities of paradigmatic range and syntactic distribution that support treating them all as distinct from entity-referring demonstratives. Analyzing all of these demonstrative-forms as copulas is the most attractive account, since demonstrative-forms with the grammatical behaviors described here occur only in clauses with non-verbal predicates, a environment where copulas are commonly found. In addition, demonstratives have been documented to grammaticalize into copulas in various other languages, so it is not surprising that it has occurred, or is occurring, in Passamaquoddy. The fact that inflectional changes have not occurred to the same extent in all constructions means that in one construction, the copula has a greater resemblance to the entity-referring demonstratives from which it developed, but its other grammatical behavior supports placing it with the demonstrative-derived copulas that are found in other Passamaquoddy verbless constructions. Thus, although there is not ample written documentation detailing the history of these developments, an examination of the various verbless constructions where the demonstrative-forms occur provides valuable evidence in favor of a grammaticalization analysis for the less analytically obvious construction.
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.16 (296)
Eve Ng
List of abbreviations 1 1plex 2 ai an cop der dim
first person first person plural exclusive second person animate intransitive verb animate copula derivational suffix diminutive
dist inan nmlz pl prox remt sg ti
distal inanimate nominalizer plural proximal remote singular transitive inanimate verb
Notes . Passamaquoddy has twelve consonants – /p/, /t/, /v/, /k/, /kw /, /s/, /h/, /m/, /n/, /l/, /w/, and /j/ – and five vowels – /i/, /e/, /a/, /o/, and /6/. I use an orthography that was developed for Passamaquoddy in the 1970s by Philip LeSourd and others; it differs slightly from the IPA symbols in that /v/ is written as c; /kw / as q; /h/ as an apostrophe before a word-initial consonant; /j/ as y; /o/ as u; and /6/ as i before y, u before w, and o in other environments. Passamaquoddy also has contrastive accent, which is not reflected in this orthography. . Another common definition of copula is that it is a morpheme which links an argument and a non-verbal predicate (e.g. see Trask 1993: 64), but this is a less satisfactory definition, since not all copulas occur in clauses with two such expressions. Consider the following Mandarin Chinese data: (I)
a.
b.
(II) a.
b.
Ni shi shei? 2sg cop who “Who are you?” Wo shi jingcha. 1sg cop police “I’m a police officer.” Shi shei? cop who “Who is it?” Shi jingcha. cop police “It’s the police.”
In both (Ia–b) and (IIa–b), shi (historically a demonstrative) functions as a copula in creating a predication with the following predicate nominal, shei “who” or jingcha “police (officer)”. In (I), shi could also be analyzed as linking the two nominal expressions, the subject pronoun and the predicate nominal. However, this is not possible in (IIa–b), since there is a predicate nominal but (unlike a language like English) no overt subject. . There is some variation amongst the constructions as to whether negative sentences contain demonstrative-forms when the corresponding affirmative sentences do. This is not
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.17 (297)
What constitutes a case of grammaticalization?
surprising if a grammaticalization process is relatively new, in which case the relevant item might not be obligatory for all instances of some construction, and be found only in the linguistic structures that occur most frequently. In this instance, affirmative clauses are more frequent than negative clauses in Passamaquoddy (and languages generally). . My language consultants generally rejected sentences without such a demonstrativeform, but there are exceptions in older texts, where there are clauses that have a simple juxtaposition between two nominal terms. . Although copula agreement with the subject is the familiar pattern for English and a number of other European languages, agreement with the predicate nominal also exists, as Diessel (1999: 143–148) argues for Hebrew. . If the lexical NP is the focus while the personal pronoun is the topic, then the lexical NP occurs first and the personal pronoun last, and there is no demonstrative-copula: Taktal nil. doctor.an 1sg “I’m a doctor.” I will return to this point in 4.3. . Also comparably to the construction in 4.1.2, if the converse topic-focus situation holds, such that the lexical NP is the focus while the demonstrative term is the topic, then the lexical NP occurs first and the demonstrative term last, and there is no demonstrative-copula: Emqansis wot. spoon.an an.sg.prox “This is a spoon.” . My language consultants were fairly consistent in rejecting sentences without such a demonstrative-form, but exceptions can be found in older texts, where there are sentences involving a simple juxtaposition between the argument and predicate expressions. Also, in cases where the nominal term is a participle derived from a verb, it is more common to find clauses that do not have a demonstrative-form between the terms. In this section, however, I will discuss only clauses where a demonstrative-form is present. . What one might imagine to be the equivalent one-term construction with a lexical NP term does not exist in Passamaquoddy distinct from a construction that has two terms. That is, there is no construction meaning “It’s a girl” distinct from the construction translating as “That’s/she’s a girl”; both would be as given as below, with one lexical NP as the focus term and one demonstrative as the topic term. This is the same type of construction as illustrated by (28) in this section. Pilsqehs-is not. girl.an-dim an.sg.dist That’s/She’s a girl. (or It’s a girl.)
TSL[v.20020404] Prn:1/03/2004; 12:21
F: TSL5914.tex / p.18 (298)
Eve Ng
References Bybee, Joan & William Pagliuca (1985). “Cross-Linguistic Comparison and the Development of Grammatical Meaning.” In Jacek Fisiak (Ed.), Historical Semantics and Historical Word Formation (pp. 59–83). Berlin: Mouton de Gruyter. Carrasquel, Jose Ramon (1996). The Evolution of Demonstrative ille from Latin to Spanish: A Grammaticalization Analysis. Ph.D. dissertation, University of Washington, Seattle. Craig, Colette (1991). “Ways to Go in Rama: A Case Study in Polygrammaticalization.” In Elizabeth Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization, Vol. II (pp. 455–492). Amsterdam and Philadelphia: John Benjamins. Diessel, Holger (1999). Demonstratives: Form, Function, and Grammaticalization. Amsterdam and Philadelphia: John Benjamins. Fleischman, Suzanne (1982). The Future in Thought and Language: Diachronic Evidence from Romance. Cambridge: Cambridge University Press. Gildea, Spike (1993). “The Development of Tense Markers from Demonstrative Pronouns in Panare (Cariban).” Studies in Language, 17, 53–73. Greenberg, Joseph (1978). “How Does a Language Acquire Gender Markers?” In Charles Ferguson & Edith Moravcsik (Eds.), Universals of Human Language, Vol. 3: Word Structure (pp. 47–82). Stanford, CA: Stanford University Press. Harris, Martin (1978). The Evolution of French Syntax: A Comparative Approach. London: Longman. Heine, Bernd & Mechthild Reh (1984). Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Hengeveld, Kees (1992). Non-Verbal Predication: Theory, Typology, Diachrony. Berlin: Mouton de Gruyter. Himmelmann, Nicholas (1996). “Demonstratives in Narrative Discourse: a Taxonomy of Universal Uses.” In Barbara Fox (Ed.), Studies in Anaphora (pp. 205–254). Amsterdam and Philadelphia: John Benjamins. Hopper, Paul & Elizabeth Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Li, Charles & Sandra Thompson (1977). A Mechanism for the Development of Copula Morphemes. In Charles Li (Ed.), Mechanisms of Syntactic Change (pp. 414–444). Austin, TX: University of Texas Press. Ng, Eve (2002). Demonstrative Words in the Algonquian Language Passamaquoddy: a Descriptive and Grammaticalization Analysis. Ph.D. dissertation, State University of New York at Buffalo. Pinkster, Harm (1987). “The Strategy and Chronology of the Development of Future and Perfect Tense Auxiliaries in Latin.” In Martin Harris & Paolo Ramat (Eds.), The Historical Development of Auxiliaries (pp. 193–233). Berlin: Mouton de Gruyter. Vincent, Nigel (1982). “The Development of the Auxiliaries HABERE and ESSE in Romance.” In Nigel Vincent & Martin Harris (Eds.), Studies in the Romance Verb (pp. 71–96). London: Croom Helm.
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.1 (299)
Multi-categorial items as underspecified lexical entries The case of Kambera wàngu* Marian Klamer University of Leiden
.
Introduction
Grammaticalization processes are often described as a particular item going through distinct diachronic stages. At the same time, it is generally agreed upon that grammaticalisation is a gradual process, 1 i.e., such descriptions are abstractions from what happens in reality: grammaticalisation takes place along a continuum, not as a sequence of discrete stages. We also know that the items involved in such grammaticalisation continua are often semantically vague and structurally ambiguous, and often undergo changes in ‘word class (affiliation)’ or ‘(lexical) category’. Most authors assume that a change of category at least involves a reanalysis of the underlying structure of the grammatical context in which an item is used.2 Since the reanalysis causes a change in the lexical properties of an item, the item appears in the sentence with a different word class: it is now part of a different constituent type, and it has a different interpretation than it originally had. Note that such a characterisation of the different stages of category change is not a description of the actual process of change: it remains unclear exactly which transitions took place in the lexicon, and why. The reason is that a ‘category (label)’ is not a cognitive unit, and a ‘category (label) change’ is not a lexical process. The former is a theoretical construct to characterise the combined distributional properties of words, and the latter is a coarse description of the outcome of a sequence of many (often small) changes in the distributional properties of an item. In order to formally characterise processes of category change more precisely, it is necessary to start by analysing the characteristics of the original lexical item involved and the items derived from it: exactly which semantic and/or grammatical
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.2 (300)
Marian Klamer
features are involved when an item changes the way it combines with other linguistic elements? And how are these features manipulated during the change: are semantic and/or grammatical features added, lost, or both? Which contexts are favorable to changes in the combinatorial possibilities of an item, and which contexts simply allow for it? Only by studying the lexical item and its contexts in very much detail will we be able to characterise the changes that took place in the original lexical item to arrive at its different word class affiliation. Since category change does not take place overnight nor in clearly distinct stages, but is a slow process along a continuum, we expect to find cases of category change ‘on the go’. Cases to look for are those where in one synchronic stage a single item has various functions, interpretations and combinatorial properties, and occurs in distinct but co-existing structures, as well as in a sizeable number of ambiguous contexts. This paper studies such a instance of category change ‘on the go’ in Kambera, an Austronesian language spoken on the island of Sumba in Eastern Indonesia (Klamer 1998). In this language, there is a word wàngu, with various synchronic functions and combinatorial properties. It appears in a continuum which includes the following six distinct syntactic contexts: (1) Points in the (synchronic) continuum of contexts where wàngu occurs3 Corpus: 12 hours of spoken and transcribed texts, N = 1384 a. As an independent, main instrumental verb: 5,1% (7/138) b. As the second member of verbal compound: 50,7% (70/138) c. Ambiguous between verb in a compound, or a preposition: 11,6% (16/138) d. As the head of a PP: 5,1% (7/138) e. As an ambiguous clause linker (P, C, V?) 12,3% (17/138) f. As a matrix (‘raising’) verb in biclausal constructions: 15,2% (21/138) Observe that the ratio of the distinct occurrences of wàngu shows a large variation. For example, as an independent instrumental verb it only occurs in about 5% of the cases, while its use as the second member in a verbal compound accounts for more than 50% of its occurrences. The sentences (2)–(4) illustrate three major points in the continuum: from a main instrumental verb translatable as “use”, in (2), via a prepositional function translatable as “with”, as in (3), wàngu becomes an (untranslatable) matrix verb heading an embedded clause, as in (4): (2) Nda ku-wàngu-a huru ba ku-ngangu neg 1sn-wàngu-mod spoon cnj 1sn-eat “I don’t use a spoon when I eat” (elic.)
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.3 (301)
Multi-categorial items as underspecified lexical entries
(3) Ku-palu-ha da ahu-mu nyumu wàngu ài 1sn-hit-3pa artdog-2sg you wàngu wood “I hit your dogs with a stick” (elic.) (4) Talanga la anda-ka nyungga hi na-wàngu pa-urang while loc road-prf I cnj 3sn-wàngu sr-rain “While I was on the road it began to rain” Significantly, however, in stage (c) and (e) in (1) the context of wàngu is ambiguous and its categorical status unclear. This ambiguity is not to be ignored as ‘noise’ in the data: the ambiguous contexts make up about 25% of the contexts where wàngu occurs, and are about as frequent as the unambiguous stages (i.e., (d) and (f) together). In other words, wàngu belongs to two word classes synchronically: Verb – with the subclasses of main (1a, b) and matrix (1f) verb – and Preposition (1d), while we must also account for the fact that in about 24% of its occurrences, it cannot be assigned to any particular class at all, as in (1c, e).5 The main issues addressed in this paper is therefore how a multifunctional (or multicategorial), ambiguous element should be represented in the lexicon. Analogous to the way phonologists study the feature make-up of particular phonological segments by analysing processes of (synchronic and diachronic) change in which those segments are involved, it is also possible to study the lexical-conceptual features of an item by looking at its synchronic and diachronic variation. Since there are no written records from earlier stages of Kambera, this paper only analyses the synchronic patterns of variation of wàngu. However, since synchronic variation is the source of diachronic change, there should be no principled distinction between the formal representation of an item that is synchronically multicategorial, and one that changes its category over time. This paper is organised as follows. In Section 1 I present some background information on Kambera morpho-syntax. In Section 2, the various stages in the grammaticalization chain of wàngu are discussed in more detail, by linking the distinct functions of wàngu to distinct syntactic configurations. In Section 3 I present a scenario for how wàngu may have become the item it is now. In the remainder of this introduction I address issues relating to the lexical representation of multifunctional items such as wàngu that appear to occur in a so-called ‘grammaticalisation chain’. Standard theories of lexical representation generally assume that, in principle, a lexical element should belong to one word class only. In case we find the same form in different contexts and with different functions, we assume either that we are dealing with homophonous words (whose similarity in form is accidental), or with polysemous words (where the similarity in form goes back to one lexical unit). In the former case, there is no problem to assign the words to different categories, since there is no semantic connection between them. In the case of polysemy, the situation is rather more complex. An
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.4 (302)
Marian Klamer
example of a polysemous word in present-day English is the adjective6 like. This adjective (stripped from its adjectival inflections, Maling 1983: 277) allows a preposition (P), complementiser (C), or adverb (Adv) interpretation, as in the following illustrations (from Klamer 2000: 96): (5) Cry like a baby He ran like crazy It looks like he will win I wouldn’t mind, it’s just like I prefer not to It goes like “bang”
like = A or P? like = A, P or C? like = C like = C like = C, P, or Adv?
The distinction between A like, C like, P like and Adv like, if it exists, is clearly gradual and entirely determined by the different syntactic contexts of like. It is therefore not immediately clear if, and how, like could be categorized in a classic lexical model: if we assign the different likes to distinct categories, we lose the generalisation that they are synchronically related; and if we lump all its functions and treat it as one item, we fail to account for all its distributional properties. Examples like this can be multiplied for English, and indeed occur in every language. For Kambera wàngu the classic model provides roughly two options. First, we may assume that there are three (homophonous) items wàngu: one an instrumental verb, one a preposition, and one an untranslatable matrix (‘raising’) verb. This option has two drawbacks: it denies the obvious semantic connections between the various occurrences of this item (they are more often than not polysemous), and it does not account for the fact that in about 25% of its occurrences it is fundamentally ambiguous and cannot be assigned to any particular category. The second option is to assume that the instrumental verb, the preposition, and the matrix verb wàngu are polysemous items that go back to one ‘basic’ form, in this case presumably the main instrumental verb. The basic lexical representation of wàngu would then be as a main instrumental verb. The problem with this option is that it does not take into account how wàngu is actually used: the ratios given in (1a) above show that wàngu is used as an independent instrumental verb in only 5% of the cases, while at the same time, it is not used as a verb in about 43% of its occurrences (i.e., contexts (1c–f)). If in almost half of its actual occurrences it is not used as a verb, it seems odd to characterise its distribution as basically verbal in the lexicon. In sum, a lexicon which employs discrete categories does not seem to allow for synchronically multifunctional items such as Kambera wàngu. Heine (1992, 1993: 79, 112–116) has explicitly pointed out this problem in relation to multifunctional items in ‘grammaticalization chains’, and argues that we need a new type of categorization to account for items that occur in such chains. He proposes to analyse a ‘grammaticalization chain’ as a distinct type of category, based on the taxonomic principles of ‘family resemblance’ logic. A family resem-
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.5 (303)
Multi-categorial items as underspecified lexical entries
blance category is a set of items that share a form, while every member shares at least one attribute with one or more of the other members. A family resemblance category is somewhat similar to a ‘prototype category’ (cf. Rosch 1978, and later work), since not every member is equally representative of its category, and the category has fuzzy boundaries. However, a family resemblance category is crucially distinct from a prototype category, because the latter has one prototype member – an item that combines all attributes that define the membership of the category –, whereas a family resemblance category does not have such a prototype. Also, in a prototype category, all members of the category share at least one attribute, whereas there is no such shared attribute in a family resemblance category (Heine 1993: 114). In short, Heine suggests that items in a grammaticalization chain are a special type of category, based on the taxonomic principles of family resemblance, and that this type of category should be part of the lexicon alongside the classic discrete lexical categories. His proposal therefore introduces a new set of taxonomic principles into the lexicon, in addition to the old set. If we want to keep our theoretical model of grammar as simple as possible, this is a less attractive move. In this paper I argue that it is also unnecessary, since the formal categorization of items in ‘grammaticalization chains’ can be incorporated into a lexical model with discrete categories, if we allow for a properly articulated theory of lexical representation that can handle the variable, gradient properties of individual lexical items. Also addressing the issue of category change, Haspelmath (1998) argues that we can express gradient word class membership and word class changes by using graded notations. (For instance: V1.0 for ordinary verbs, V.7/ P.3 for preposition-like verbs, V.2/ P.8 for verb-like prepositions, etc., see Haspelmath 1998: 330.) The proposal presented here is similar to Haspelmath’s in that it assumes that gradient word class membership can and must be formally expressed, but it expresses the gradience in a different way. Though the categories used in my proposal are discrete lexical categories, I do allow categories to be more or less specified. The variable, gradient properties of individual lexical items are then expressed by manipulating (i.e., adding, deleting, or changing) the lexical-conceptual features of that item. The features referred to here are very general notions that in some way or other recur in all theories about argument and lexical conceptual structure: the number of arguments of an item, their hierarchical organisation (internal/external), their syntactic category, and their semantic role (cf. Jackendoff 1990; Rappaport Hovav & Levin 1998; Bresnan 2001). The following hypotheses concerning lexical representation are specifically explored in this paper: 1. Lexical items are formally expressed as predicate-attribute combinations. Attributes are variable in type and number. For example, the lexical argument
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.6 (304)
Marian Klamer
structure of a verb may vary in the number of its arguments (1, 2, 3) and their type (Agent, Patient, Beneficiary, etc). 2. The semantic bleaching of a verb may imply: – A change in attribute number (e.g., loss or addition of one or more argument(s), see Vincent 1999 and references cited there, and Klamer 2000), and/or – A change in attribute type (e.g., loss of a specific semantic role). The result of semantic bleaching is a lexical item with a less complex/specified argument structure: an “underspecified” lexical item. 3. Underspecified lexical items exist side-by-side with fully specified lexical items. 4. An underspecified lexical item can get a specific interpretation through its syntactic context, and various contexts may invoke distinct interpretations of one single underspecified item. With respect to the analysis of the specific Kambera item wàngu, it is proposed that originally wàngu must have been an instrumental verb, which, as a result of the well-known ‘semantic bleaching’ has been stripped of part of its argument structure and semantic roles: it lost its external argument slot, the Agent role of that argument, and the Instrument role of the internal argument. As a result, the original verb is now an underspecified lexical item represented as X <x>. The different interpretations of this item when it functions as either an instrumental verb, a preposition or a matrix verb are caused by the different constructions in which it occurs. In other words, we witness that a full verb has grammaticalised into a semantically bleached, category neutral lexical item X <x>, which allows different contextually induced interpretations. In this way, we have characterised about 95% of the actual occurrences of wàngu. The remaining 5% are occurrences of wàngu in its original function as independent instrumental verb, but we will see that even as independent instrumental verb its lexical and morpho-syntactic properties are reduced.
. Preliminaries: Kambera argument marking Before going into the further details of multifunctional wàngu, I outline some aspects of Kambera morpho-syntax that are relevant to interpret the data. Kambera is a head-marking language; verbal arguments are commonly marked on the verb by pronominal clitics. The Agent argument of a simple declarative sentence and the single argument of an intransitive predicate are canonically marked with a nominative proclitic, while a definite Patient is canonically marked with an accusative, as illustrated in (6). The coreferent NPs are optional adjuncts.
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.7 (305)
Multi-categorial items as underspecified lexical entries
(6) Na tau wútu na-palu-ka nyungga art person be.fat 3sn-hit-1sa I “The big man hit me” Definite Beneficiaries/Recipients are canonically marked with a dative clitic, as the contrast between (7a–b) shows: in (7a), the object clitic marks the Patient and is accusative, in (7b) it marks the Recipient and is dative. Grammatical definiteness is signalled by the presence of one of the definite articles: na (singular), da (plural) or i (proper name). Only grammatically definite objects are marked on the verb; indefinite Patients or Recipients are never crossreferenced, compare (7a–b) with (7c). (7) a.
Da-ngàndi-ya na uhu 3pn-take-3sa art rice “They take the rice” b. Da-ngàndi-nya na uhu i Ama 3pn-take-3sd art rice art father “They bring father the rice” c. Da-ngàndi uhu 3sn-take rice “They take (some) rice”
The dative clitic paradigm is a hybrid category, since it marks both Recipients and Patients: any transitive verb that ends in -ng or -ngu must mark its object with a dative clitic; even if it is semantically a Patient. An example is the object of píngu “know” in (8). Observe that -ngu and the dative clitic are in complementary distribution. (8) Nda ku- pí -nya na laku-mu neg 1sn-know -3sd art go-2sg “I didn’t know that you’d gone” Because the item under study in this paper, wàngu, also ends in -ngu, its object clitic must always be dative, even if it expresses a Patient. Indeed, we will see below that the semantic role of the dative object of wàngu may vary from Instrument, Recipient/Beneficiary, to Patient, while it may also be semantically ambiguous. A dative object whose role is ambiguous is illustrated in (9), where -nya refers to ‘the hymn’, which may be the Patient of “sing” or the Instrument of wà- “use”: (9) Ku-rongu-kau ba u-ludu wà-nya na ludu hali 1sn-hear-2sa cnj 2sn-sing wàngu-3sd art song be.holy “I heard you sing the hymn” (lit. “I heard you and you sang (using) the hymn”)
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.8 (306)
Marian Klamer
This sentence also illustrate the use of conjunction ba “and, as, while, because” in Kambera. Kambera only has coordinating conjunctions such as ba in (9), other examples are hi “and, so”, ka “so that”, and jàka “if, when”. The language does allow clauses to be embedded, but employs particular non-finite morphemes to mark those (for example, there are special markers for relative clauses and for controlled clauses); while nominalised clauses are marked with a genitive subject. In short, Kambera has no subordinating complementisers, a fact we will return to in the analysis of the clause-combining function of wàngu in Section 2.5 below.7
. The distinct functions of wàngu As was mentioned above, the occurrences of wàngu in a 12-hour corpus of transcribed spoken texts indicate that wàngu occurs in a continuum of grammatical contexts, in which we can distinguish at least six distinct stages. They are given in (1) and repeated here briefly: (a) as independent instrumental verb (5,1%), (b) as second member of verbal compound (50,7%), (c) ambiguous between verb in a compound, or preposition (11,6%), (d) as the head of a PP (5,1%), (e) as an ambiguous clause linker (12,3%), (f) as a matrix verb in biclausal constructions (15,2%). The following subsections present analyses of these six contexts.
. Wàngu as an independent instrumental verb Sentences (10) and (11) illustrate that wàngu can be used as an independent, instrumental verb. Note, however, that neither of them expresses the two arguments of wàngu at the same time: in (10), the subject of wàngu is marked, but the object is not; in (11), the subject is unmarked, while the object is. (10) Nda ku-wàngu-a huru ba ku-ngangu neg 1sn-wàngu-mod spoon cnj 1sn-eat “I don’t use a spoon when I eat” (11) Njadi-mbu jàka wà-nya yena? be.possible-also if wàngu-3sd this.one “Is (it) also possible with this one?”/”Is (it) also possible if (I/we) use this one?” There are indications that synchronically, wàngu is losing its function as independent instrumental verb. First, it functions as a main instrumental verb in only 5.1% of its occurrences – Kambera speakers also use the loan word paki (from Indonesian pakai “use”). There are communicative reasons for this: wàngu has (by now) developed a rather generic semantics (including “use”, “with”, “because of ”,
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.9 (307)
Multi-categorial items as underspecified lexical entries
“while”, etc., see the subsections below), while paki is more specific. In addition, wàngu is homophonous with yet another generic verb: the quotative verb wà(ng) “report/say”, which is used extremely frequently in Kambera discourse.8 It is also significant to note that the corpus does not contain any examples of wàngu used as an independent instrumental verb with both its arguments expressed overtly (i.e., it occurs with a subject clitic or with an object clitic, but never with both). I take this to indicate that the morpho-syntactic properties of the verb in its function as a main, instrumental verb are breaking down as a result of the fact that elements of its argument structure have also been lost. (This will be further discussed in the subsections below.) But if wàngu is hardly functioning as an independent instrumental verb anymore, why then claim that this is its ‘basic’ or ‘original’ function? The reason is that more than 50% of its occurrences are (still) typically verbal, and instrumental, when wàngu appears as the second member of a compound verb. This is discussed in the next subsection.
. Wàngu as the second verb in a verbal compound In the majority of cases, instrumental wàngu is the second verb in a verbal compound. Compound verbs are a productive morphological category in Kambera. They are derived by combining (any) two verbs.9 Illustrations are tila wàrung “kick & dispose of ” > “kick (s.o./s.t.) away” and palai nyara “run & chase” > “run after (s.o./s.t.)” In a similar vein, the instrumental verb wàngu combines with any other verb to derive a compound verb, whereby wàngu is always the second verb in the compound. Some examples are: (12) taku riki hayidi tanda pabanjar
“scoop X” “laugh” “play games” “know X” “talk”
taku wàngu riki wàngu hayidi wàngu tanda wàngu pabanjar wàngu
nggidik
“shake”
nggidik wàngu
meti
“die”
meti wàngu
“scoop X using Y” “laugh about/because of Y” “play games on Y” “know X because of Y” “talk about Y”; “talk using language Y” “shake because of Y”, “be worried because of Y” “die because of Y”
In all these compounds, the object Y is an instrument: a thing to scoop with, or the instrument by which one can have a laugh, play a game, know something, talk or say something, shake, become worried or die. The Kambera compound verb is lexically and syntactically a unit (Klamer 1998: Chapter 7). This is an important distinction between wàngu in compound verbs and wàngu in other positions in
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.10 (308)
Marian Klamer
the sentence. Syntactically, the two verbs that constitute a compound are one verbal unit; more particularly, in this configuration, wàngu never projects a separate syntactic constituent. The compound verb as a whole is the head of the Predicate Phrase (PredP). A Kambera PredP may also contain adverbs (which are separate words), and is the domain of attachment for the clitic cluster.10 In (13) the PredP of the second clause contains the compound verb panuang wàngu “continue doing/with something” and the reduplicated adverb ju-juang “only”. Note that the clitics -ma-ki-a-da-ka attach to the PredP as a whole, including the adverb ju-juang, and not to an individual verb. (13) . . . kei-ma-danya-ka uda hawiang, [[panuang wangu]V buy-emp-3pcont-prf emp.3p some continue wàngu ju-juang Adv ]PredP -ma-ki-a-da-ka rdp-only -emp-mod-mod-3pg-prf “. . . they are buying some, and they simply continue (with it). . . ”11 In (13) the shared subject of panuang wàngu “continue with s.t.” is expressed with a genitive enclitic -da, and there is no overt object. In (14), however, the compound verb meti wàngu “die because of something” has both an overt subject (ku-) and an overt object (-nya).12 Since the first member of the compound is an intransitive verb (meti “die”), the object of the compound verb must be the complement that originally belonged to wàngu. (14) Ai ndia, puli-bia-ngga bùdi, jàka ku-meti exc no release-mod-1sd in.fact if 1sn-die wà-ma-nya-i una, nda nggàra ehi-a wàngu-emp-3sd-iter emp-3s neg what content-mod “O no, just let me go, if I die as a result of it – I don”t care” Notice, however, that neither the compound verb, nor wàngu alone, forms a syntactic constituent with this object: the clitic -nya is separated from the PredP by an emphatic clitic -ma. In fact, more modal clitics could be added here: in the clitic cluster pronominal clitics are preceded by modal and emphatic clitics, and the compound verb can never form a syntactic constituent with its complement clitic to the exclusion of the other clitics. Lexically, the two verbs are also a unit, since they have a single, merged, argument structure. If we assume that an instrumental verb has two arguments, X and Y, X would be commonly considered the external argument, and semantically be an Agent, while the Y would be the internal argument, and canonically an Instrument. When instrumental compound verbs are derived, we expect that the argument structure of the verb wàngu merges with the argument structure of
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.11 (309)
Multi-categorial items as underspecified lexical entries
the verb it combines with, and this is sketched in (15a) for intransitive base verbs, and (15b) for transitive base verbs: (15) a.
Intransitive base verb + wàngu => Transitive compound verb Example: riki <x> “X laugh” + wàngu <x > “X use Z” => riki wàngu <x > “X laugh about Z” b. Transitive base verb + wàngu => Ditransitive compound verb Example: taku <x > “ X scoop Y” + wàngu <x > “X use Z” => taku wàngu <x > “X scoop Y using Z”
The argument merger results in a structure where the Agent of both verbs is shared, and the compound verb always has at least one internal argument. (If the base verb is transitive, it has two). The sentence in (16) consists of two clauses, both of which contain an instrumental compound with a shared Agent and a shared Patient: (16) . . . ka ku- langu li wà -ki-nya na aya-nggu, cnj 1sn- word word wàngu -mod-3sd art older.sibling-1sG “. . . so I can say (s.t.) about my brother, ka ku- paní wà -nya na aya-nggu cnj 1sn- tell wàngu -3sd art older.sibling-1sg so I can tell (s.t.) about my brother.” In case the compound verb has two shared internal arguments, only one of those is usually morpho-syntactically expressed (the other one being implied). There is no structural preference for one argument in particular: either one of the shared arguments may be expressed, pragmatic and/or discourse considerations usually decide which one is. For example, in (17), the object clitic refers to the Patient argument originally belonging to the verb “take” (wife), while the Instrument (dowry) is implied. In (18), the object clitic must refer to the argument of wàngu since there is no interpretation available where the statue could be the Patient of “tell”.13 (17) . . . hi mài pa-piti wà-nyaj [na kuru uma-nggu]j hu dita la cnj come sr-take wàngu-3sd art wife-1sg dir up loc Jawa Java “. . . and (I) came to get a wife with (it, i.e. dowry) up there in Java” (18) . . . hi kiri wà-ma nyuma na paní wà-njaj [da katoda cnj begin wàngu-1pg we art tell wà-3pd art pole kawindu]j yard “. . . and we began to tell (s.t. to) the katoda kawindu”
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.12 (310)
Marian Klamer
In many cases, the referent of the object clitic is ambiguous, and may be interpreted as the argument of either of the two verbs. For example, the object of the compound verb in (19) allows two interpretations: as Patient of “discuss” or Instrument of wàngu. (19) Mu- hili lua-ki ná haromu, apu, wà-na-nya 2sn- again go-mod dei tomorrow granny report-3sg-3sd “ “You go again tomorrow, gran,” he told her, muda-a una nuna nú, u- báta wà pàku be.easy-mod emp-3s that.one dei 2sn- discuss wàngu in.fact -nya mod-3sd “it’s easy, you should just discuss it (with words)” ” To conclude, the argument structure of instrumental wàngu in a compound verb merges with that of the other verb, and together they also constitute one syntactic unit. In the latter respect, the compound verb configuration is distinct from configurations where wàngu heads an independent constituent in the clause. Such configurations are discussed in the next section.
. Wàngu as either a verb or a preposition In this section we will see that wàngu can also occur outside the Predicate Phrase and form an independent, separate syntactic constituent, while it is semantically still part of the compound verb. The contrast is illustrated in (20a–b). In (20a) wàngu occurs in the ‘original’ configuration: as part of a verbal compound, heading the PredP. In (20b), the same (semantic) compound verb is used, but wàngu is now moved to a position in the clausal periphery (following the adjunct NP “your dogs”). This position is normally occupied by temporal or locative adjuncts, the latter of which are often PPs, as illustrated in (21). In other words, the constituent with wàngu in (20b) has the positional properties of a preposition. (20) a.
Ku-palu wà-nja ài da ahu-mu nyumu 1sn-hit wàngu-3pd wood art dog-2sposs you “I hit your dogs with a stick” b. Kupalu-ha da ahu-mu nyumu wà-nja ài 1ssubj- hit-3pa art dog-2sg you wàngu-3pd wood “I hit your dogs using a stick (on them)”
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.13 (311)
Multi-categorial items as underspecified lexical entries
(21) . . . hi ku-pu-puha-bia-ya una [na karobu tunu]NP [la cnj 1sn-rdp-drop-mod-3sa emp3s art pumpkin roast loc mbombang]PP . . . ground14 “. . . and I just dropped the roasted pumpkin on the ground” Observe, however, that while the object is cliticised once in (20a), it occurs twice in (20b): once on “hit”, and once on wàngu. The two major distinctions between verbs and prepositions in Kambera are that (i) object clitics attach to PredP’s only, and (ii) that predicates may consist of a single verb, but never of a single preposition. In other words, since the extraposed wàngu in (20b) has an object clitic, and object clitics never attach to prepositions, wàngu cannot be a preposition in that context. Morpho-syntactically, then, it is a verb, but note that this verb occurs in the peripheral position that is typically reserved for PPs. While wàngu in (20b) can still be called a verb for morpho-syntactic reasons, and the interpretation of the sentence suggests that this verb is still a semantic unit (a compound) with the other verb in the clause, there are also contexts where wàngu is truly ambiguous between a verb and a preposition. In such contexts, wàngu appears without an object clitic and is directly adjacent to both the verb (with which it would form a compound) and the object noun (with which it would form a PP). In such contexts, wàngu can be analysed as in (22a), or as a preposition, as in (22b). Illustrations of such configurations are given in (23) and (24). (22) a. [V wàngu] V [N] b. [V] [wàngu N]PP (23) Ngangu kokuru jua-a, ngangu wàngu tolung eat coconut only-mod eat wàngu meat “Eat coconut only, with meat” (24) Na kaweda-na na tau nuna lundu na-bei wàngu art be.old-3sg art person that.one until 3sn-crawl wàngu kamata kaba top husk “Because of the age of that person s/he crawls using coconut husks”15 Note that the context where this ambiguity arises is quite restricted. Wàngu must be uninflected, it must have an indefinite object, the object noun must be bare, and wàngu and its object must be linearly adjacent. In other words, wàngu in (25) is not ambiguous: the object na iyang “fish” is adjacent to wàngu but not a bare noun (since it has a definite article), and kawàdak “money” is indefinite and a bare noun, but is not linearly adjacent to wàngu:
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.14 (312)
Marian Klamer
(25) Ee, kawàdak ta-kei wàngu na iyang wà-nggu excl sacrifice 1pn-buy wàngu art fish report-1sg “Oh my, money to buy the fish with, I said” Despite these restrictions, we still find that in 11.6% of the cases examined in the database, wàngu occurs in such ambiguos contexts, which thereby outnumber the percentage of contexts (5.1%) where an unambiguous Preposition-interpretation is available (see the next subsection). Ambiguity is clearly one of major characteristics of wàngu.
. Wàngu as the head of a prepositional phrase In 5.1% of the contexts examined, wàngu is interpreted as an unambiguous preposition since it appears in the clausal periphery along with a bare complement noun, just like a canonical preposition would. (Semantically, however, it is still a unit with the verb in the Predicate Phrase, as in the compound verb constructions discussed in 2.2.) Sentences (26a–b) illustrate the preposional function of wàngu. They contrast with (20a–b), where wàngu is interpreted as an (inflected) verb. (26) a.
Ku- palu-ha da ahu-mu nyumu wàngu ài 1sn-hit-3pa art dog-2sg you wàngu wood “I hit your dogs with a stick” b. Wàngu ài ba ku-palu -ha da ahu -mu nyumu wàngu wood cnj 1sn-hit-3pa art dog-2sg you “With a stick I hit your dogs”
Additional examples of the prepositional use of wàngu are (27) and (28), where we know that wàngu occurs outside the PredP because it follows the object clitic that is attached to the edge of the Predicate Phrase. As a preposition, it governs a complex NP: (27) . . . ka u-dunda-nya [wàngu [ngahu-mu dangu pulu-mu] cnj 2sn-call.together-3sd wàngu spirit-2sg and word-2sg “. . . call him with your spirit and your word” In (28), wàngu projects a prepositional phrase which follows the peripheral, temporal adjunct ni kawài “just now”: (28) [[. . . pabanjar-di-manya-i] duma] [ni kawai] [wàngu hilu speak-emp-1pcont-iter emp.1p dei just.now wàngu language Humba] Sumba “. . . we have also been speaking Sumbanese just now. . . ”
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.15 (313)
Multi-categorial items as underspecified lexical entries
In sum, if wàngu occurs outside the PredP, with a nominal complement but without an object clitic, it has the structural properties of a canonical Kambera preposition.
. Wàngu as ambiguous item with a clausal complement Although complements of prepositions are canonically referential nominals, prepositions also have a proposition – a clause – as their complement. Kambera has one general locative preposition la, and this preposition also governs subordinate clauses. In biclausal constructions, la introduces the second, subordinate, clause as the explicit purpose or goal of the matrix clause. Examples are (29) and (30). Observe that the subordinate clauses are marked with the subordinating (SR) morpheme pa-, and that their subject is controlled by the matrix subject. (Arguments for this analysis are presented in Klamer 1998: Chapter 8.) La is preceded by an intonational break. (29) Hi ku-njadi la pa- piti-ya cnj 1sn-be.able loc sr- take-3sobj “So I’ll be able to take her/him/it” (30) Da-puru-ka uda nú, ngàndi-danya bi kabela 3pn-descend-prf emp.3p dei bring-3pcont der machete bi nímbu nú der spear dei “They came down, bringing machetes and spears la pa-pa-meti wà-nya da ular dàngu da wuya loc sr-cau-die wàngu-3sd art snake and art crocodile (in order to) to kill the snakes and the crocodiles with (it)” In a similar vein, the form wàngu may have a clausal complement, and link a subordinate clause to a matrix clause, as illustrated in (31), (32) and (33). Like la, wàngu is preceded by an intonational break. (31) Mandapung-danya ndalihu wàngu pa-pa-ndalihung sit-3pcont segment wàngu sr-cau-segment-ng “The segments (joints of the corn stalk) appear and get their shape” (lit. “the segments settle while becoming segments”) (32) . . . patiang ana mandài-ndài wàngu pa-buta ana rumba wait dim rdp-be.long.time wàngu sr-pluck dim grass “. . . (we) wait for some time, weeding some grass in the meantime”
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.16 (314)
Marian Klamer
(33) Mbaca mangilu wàngu pa-ngangu read first wàngu sr-eat “(We) read before eating” Wàngu occurs in biclausal configurations such as these in 12,3% of the cases studied. Semantically, it has scope over the second clause, and in (32) and (33) it indicates that the events expressed in the subordinate clause happen simultaneously with, or in immediate sequence to, the event of the matrix clause. In their clause-linking function, then, the canonical preposition la and the item wàngu pattern alike: (34)
S VP come down read first
XP X la wàngu
IP to kill to eat
“Come down to kill” / “Read before eating”
Yet, there are significant syntactic differences between subordinate clauses introduced by la and those introduced by wàngu. First, clauses with la cannot be preceded by a conjunction, while clauses with wàngu can. This applies to Kambera conjunctions in general and is illustrated with the conjunction hi in (35)–(37): (35) Hi ku-njadi *hi la pa- piti-ya cnj 1sn-be.able cnj loc sr- take-3sobj “So I’ll be able to take her/him/it” (36) Mbaca mangilu hi wàngu pa-ngangu read first cnj wàngu sr-eat “We read and then we eat” (37) Talanga la anda-ka nyungga hi wàngu pa-urang while loc road-prf I cnj wàngu sr-rain “While I was on the road it began to rain” The fact that a coordinating conjunction such as hi precedes wàngu and a subordinated clause is remarkable because clause coordinators by definition mark paratactic clause relations. Moreover, under a prepositional reading of wàngu, the coordinator would subcategorise for a PP, which would result in the syntactically illformed configuration in (38) (compare (34)):16
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.17 (315)
Multi-categorial items as underspecified lexical entries
S
(38) VP read first
C and
PP P wàngu
IP to eat
“(We) read first (and then we) eat”
In Section 2.6 we will see that in such contexts wàngu is often interpreted as a verb, since it gets a subject proclitic. In other words, the configuration where a conjunction is followed by an uninflected, categorically ambiguous wàngu governing a subordinated clause is not very ‘stable’, because the syntactic context excludes any prepositional interpretation of wàngu, while forcing its verbal reading. The second syntactic contrast between clause-linking prepostion la and ambiguous wàngu is that subordinate clauses with la must appear with a matrix clause, while clauses with wàngu may occur independently. Sentence (39) illustrates this. Structurally, the wàngu clause in this sentence is a subordinate clause (since it is marked with the subordinating morpheme pa-), but it appears without a matrix clause: it is linearly preceded by an interjection/exclamation (“good heavens”), and, before that, by a clause with a nominal predicate that is not its matrix clause. In other words, grammatically wàngu would link two clauses, but one of these is absent. Such ‘reduced’ clausal sequences are impossible with the preposition la. (39) Yena ama, jàka wài huhu mini katoba-ya, this.one father if water milk male be.crazy-3sa “Here, dad, if this is ‘mini katoba”s milk, ka nggiki-na wà-mu, ba wàngu pa-nda padàdu-mbu-nya cnj how-3sg report-2sg cnj wàngu sr-neg endure-also-3sd good heavens (lit. How do you say it!), it is hard to stand na wau-na na wài huhu nuna.17 art odour-3sg art water milk that.one the stench of that milk. . . ” Thirdly, the semantic relation between clauses linked by wàngu is less specified than between clauses that are linked by la: while la always expresses a purpose or goal, the semantic function of wàngu is much less clear, as illustrated in (39). In many cases, it expresses a temporal notion (in particular, “happen/do simultaneously, or subsequently” as in (31), (32), (33), (40), see below). In sum, compared to clauses introduced by the preposition la, subordinate clauses with wàngu enjoy a relative syntactic freedom: they may be preceded by an optional coordinating conjunction, as well as an optional matrix clause. Moreover,
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.18 (316)
Marian Klamer
the semantic relation between a wàngu clause and the matrix clause (if there is one) is vague and open to various interpretations, depending on the context. In general, an increase of syntactic and semantic freedom of an item is formally expressed as a decrease in its lexical specifications and/or restrictions. For the lexical representation of wàngu this suggests that it must have fewer (syntactic) subcategorization restrictions and fewer (semantic) attributes (arguments, semantic roles) than, for example, a canonical verb, or a preposition such as la. In particular, it seems that wàngu has no external argument of its own (anymore), and that the semantic content of its internal argument (Instrument) is lost.
. Wàngu as a matrix verb When wàngu occurs in biclausal contexts, and especially when it is preceded by a conjunction, it is interpreted as a (matrix) verb. Its verbal function is overtly marked by the presence of a subject proclitic.18 This matrix verb subcategorises for a subordinate clause (a clause marked with pa-), as illustrated in (40) and (41): (40) Talanga la anda-ka nyungga hi na-wàngu pa-urang while loc road-prf I cnj 3sn-wàngu sr-rain “While I was on the road it began to rain” (41) . . . jàka ningu-ka banda pa-ngàndi-nggu nú haromu, if be-prf cattle rel-bring-1sg dei tomorrow “. . . if I bring dowry tomorrow ka u-wàngu [pa-puru toma-ka] cnj 2sn-wàngu sr-descend meet-1sa you will come down to meet me (because of it/in order to get it)” Wàngu functions as such a matrix verb with an overt subject marker in 15,2% of its total occurrences. What is the meaning of this matrix verb? Its semantics are extremely vague: it may express a temporal relation “happen/do simultaneously, subsequently”, as in (40) and (42), it may express a general instrumental or causal relation, as in (41), and it may mean “use”, as in (43). (42) Ndedi ana luhu nú, not.yet dim exit dei “He hadn’t left (the woods) yet, nda na-wàngu pa-pàdang mapini ngga-nggàra, ndia-ma neg 3sn-wàngu sr-sense hear rdp-what no-emp (and) he didn’t hear a thing, nothing at all”
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.19 (317)
Multi-categorial items as underspecified lexical entries
(43) Jàka ta-wàngu paní hau, ‘nyuna na kapala hakola’, if 1pn-wàngu word one.clf he art head school wà-da-nya report-3pg-3sd “In (lit. If we use) other words, “he’s the head of the school”, they tell him” In addition, it links clauses expressing two simultaneous states of affairs: in (44) the state of affairs of not yet having pulled out the knife is simultaneous to the hitting, in (45) the state of affairs of having finished the cutting is simultaneous to the trying to lift it up. (44) Ee, ndedi-ma na-butuh-ya na kirih-na; excl not.yet-emp 3sn-pull.out-3sa art k.o. knife-3sg “He hadn’t yet pulled out his knife. . . na-wàngu pa-pangàmbah tú-tú -ma-a-na-nja nú 3sn-wàngu sr-hit.with.fists rdp-put -emp-mod-3sg-3sd dei he just hit them like this (gesture of hitting with fists)” (45) Hàla hi na-wàngu pa-katiri-ya na bi ai, be.finished cnj 3sn-wàngu sr-cut-3sa art der wood “After he had cut the wood hi na- kama pa-pajàjak-nya cnj 3sn- try sr-lift.up-3sd he tried to lift it up. . . ” The matrix-verb-interpretation of wàngu is clearly related to the context discussed in the previous subsection, where it connects two clauses while being preceded by a conjunction. We saw that this resulted in illformed configurations like the one in (38). However, this illformed construction can easily be turned into a wellformed one by analysing wàngu in (38) as a verb rather than a preposition. The resulting configuration is represented in (46): wàngu is now a verb, and projects its own clause which is coordinated with the previous one (I bring dowry) by the conjunction (and). At the same time, wàngu is also the matrix verb of an embedded, non-finite clause (come down to meet me), whose subject is controlled by the subject of wàngu.
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.20 (318)
Marian Klamer
(46)
S VP I bring dowry
PP
C and I
VP
V you-wàngu
IP come down to meet me
“... I bring dowry and you will come down to meet me”
Observe that, though the subject of wàngu controls the embedded subject, its referent is the Agent of the embedded verb. (For example, the subject of wàngu in (41) and (46) refers to the Agent of ‘come down to meet’). In other words, it appears that wàngu has a syntactic subject, but no Agent argument. This is what we expect to be the case if wàngu is indeed a reanalysed ambiguous lexical item (category neutral between P and V) in the contexts discussed here (cf. Sections 2.4, 2.5), since prepositions never have an Agent argument. At the same time, the embedded verb has an Agent. But since subordinate clauses marked with pa- cannot have overt subjects, this Agent can only be expressed as the subject of the matrix verb. In sum, lacking its own Agent, wàngu expresses the Agent of the embedded verb as its subject. In conclusion, pressure to avoid an ill-formed configuration forces the ambiguous item wàngu to function as a matrix verb in those biclausal contexts where it is preceded by a coordinator. Since wàngu no longer has an external argument of its own, its syntactic subject expresses the external argument of the embedded verb.19
. The wàngu “grammaticalization chain” is in fact only one lexical item On the basis of the evidence presented in Section 2, I analyse wàngu as an underspecified lexical item: X <x>. Its function as instrumental verb in compound constructions, its prepositional function, its use as a semantically vague matrix verb, and its numerous ambiguous occurrences are all expressions of one, generic lexical item. The internal argument of wàngu, <x>, is always expressed as a syntactic complement, but in different configurations. It may be expressed as: (i) (one of) the object(s) of a compound verb, (ii) the nominal complement of a preposition, or (iii) a clausal complement of an ambiguous item or of a matrix verb. I propose the following scenario for the development of wàngu from a main instrumental V into the underspecified lexical item X:
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.21 (319)
Multi-categorial items as underspecified lexical entries
1. Independent instrumental verb, with two semantically specified arguments: V ‘wàngu’ < Agent < Instrument >> 2. For various functional reasons, including homophony with the quotative verb wà(ng) and the availability of Indonesian loan paki, wàngu falls into disuse as independent instrumental verb, and loses some of its morpho-syntactic properties. 3. As an instrumental verb, it is still frequently used in compound verbs. Since Kambera compound verbs have a merged argument structure, wàngu and the other verb share their Agents. As a result, the Agent of wàngu is never expressed as a separate morpho-syntactic entity. Since the context of compound verbs do not allow it ever to be visible at the surface, it is lost from the lexical argument structure of wàngu altogether: V ‘wàngu’ < Instrument > 4. The internal argument of wàngu is also part of the merged argument structure of compound verbs, and as a result (some of) its semantic features are also lost, so that the interpretation of the complement of wàngu becomes more generic, ‘object’-like: V ‘wàngu’ < Instrument/Goal/Beneficiary/Patient/. . . > 5. Without semantic role for its only argument, wàngu is no longer a typical verb (since prepositions and complementisers also have complements). Wàngu has de facto become an underspecified lexical item with a single, unspecified, internal argument: X ‘wàngu’ < x > 6. Given the appropriate syntactic context, it can now be interpreted as either a main verb, a preposition, a matrix/raising verb, or remain ambiguous between any of these interpretations.
. Conclusions By analysing the grammaticalization chain of wàngu as different surface manifestations of one single, impoverished lexical item X <x>, we account for almost all of its synchronic occurrences and functions. Though I assume that wàngu used to be a fully fledged instrumental verb this is not crucial to the analysis. The central idea of the analysis presented here is that synchronically, wàngu has an extremely simple lexical argument structure (no external argument, and one semantically generic internal argument), while its various functions are entirely determined by the different syntactic configurations in which it occurs. As contextually derived functions, they do not need to be lexically represented.
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.22 (320)
Marian Klamer
In this view, the lexicon contains underspecified items as well as items with a fully specified predicate and argument structure. The lexicon is a dynamic entity, where items may lose and gain argument structure and semantic content as a result of both functional constraints and syntactic pressure. I hope to have shown that in a slightly more articulated theory of lexical representation, combined with an analysis of the interaction between different linguistic modules (here, lexicon and morpho-syntax), there is no need to treat items in grammaticalization chains as a ‘special species’ in the lexicon. Multicategorial items, or items in grammaticalization chains can be categorized using the same discrete categories and taxonomic principles we use for other lexical items. The only thing in which they differ from canonical lexical items is that they have fewer lexical features, so that their function and interpretation is largely dependent on the syntactic context in which they appear.
Notes * I would like to thank Elly van Gelderen for discussion about the syntactic analysis of wàngu, and the anonymous reviewer for insightful and helpful comments on the prefinal draft of the paper. Abbreviations: 1, 2, 3 = person, A = Accusative, ART = Article, CAU = Causative prefix, CLF = Classifier, CNJ = Conjunction, D = Dative, DEI = Deictic, DER = Derogatory marker, DIM = Diminutive, DIR = Directional particle, EMP = Emphasis, EXCL= Exclamation, G = Genitive, ITER = Iterative aspect clitic, LOC = Locative preposition, MOD = Modal clitic, N = Nominative, NEG = Negation, p = plural, PRF = Perfective aspect clitic, RDP = Reduplication, s = singular, SR = Clitic marking subordinated clauses. . See for example Lightfoot (1991: 162–163) and Haspelmath (1998: 329). . See the discussion of reanalysis in Harris and Campbell (1995: Ch. 4). Some classic examples of processes where V changes into P can be found in Lord (1973), Heine and Reh (1982), Hopper and Traugott (1993), and Harris and Campbell (1995). A recent study of category change (P changing into C in English) within the generative (Minimalist) framework is Van Gelderen (1998). . There is another, homophonous verb wà(ng) ‘report (to); do’, which is used very frequently in Kambera discourse to report direct/indirect speech, visions, and sounds. The (near) homophony of this verb and instrumental wàngu suggests a shared ancestor. In addition, because both verbs have very generic semantics, it is tempting to try to analyse them as one and the same verb from a synchronic point of view too. This is what I set out to do when I started the research for this paper, and what an anonymous reviewer suggested after reading the previous draft of the paper. However, homophony and semantic vagueness are not sufficient evidence to assume that synchronically quotative wà(ng) and instrumental wàngu are the same verb, especially since there are a number morpho-syntactic distinctions between them: (i) quotative wà is an intransitive root verb (the form wà-ng is its applicative
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.23 (321)
Multi-categorial items as underspecified lexical entries
derivation), while instrumental wàngu is a transitive root verb; (ii) wà only takes a genitive subject, wàngu a nominative or a genitive subject; (iii) applicative wà-ng takes subject and object pronominal clitics simultaneously, instrumental wàngu does not allow simultaneous subject and object clitics; (iv) quotes always precede the quotative verb, instrumental objects always follows the instrumental verb; (v) ‘derived’ functions of the quotative verb include being used in naming strategies (“to call someone X”) and as a discourse particle (“you know”), while the ‘derived’ functions of instrumental wàngu are that it can be used as a preposition or as a matrix verb in biclausal constructions. To me, it does not seem appropriate to derive all these functions from one generic lexical item; for one, because it would be unclear what the argument structure of that item would be: quotative wà is fundamentally intransitive (and becomes transitive by applicative derivation), while wàngu is transitive in all its occurrences. For a full exposition of the characteristics of quotative wà I refer to Klamer (2000), where the syntactic and lexical properties of this verb are compared with quotative verbs in two genetically related languages (Buru and Tukang Besi), and to Klamer (2002), which discusses the various (derived) functions and grammaticalization properties of wa(ng). These articles extend the analysis of quotative wà in Klamer (1998: Section 8.2.4). The present paper supplements and extends the discussion of instrumental wàngu in Klamer (1998: Section 7.2.1). . Most examples in this paper were taken from the same corpus of conversations and narratives that was used to calculate the frequencies. None of the examples is elicited through an intermediate language, but some examples were brought up as isolated sentences during discussions with native speakers about the use of wàngu. All examples have been checked by native speakers. . Among other things, this implies that synchronically speakers may have two, three or more interpretations of the same form available, varying per context (contra Haspelmath 1998: 341). It does not imply, however, that speakers are necessarily aware of the (historical) relation between the competing interpretations, since a historical scenario is a linguist’s construct and does not necessarily reflect a speaker’s perspective (cf. Joseph this volume). . Despite the homophony, verbal like (OE lician, cf. Allen 1995) and adjectival like (OE gelic, Maling 1983) are not cognates. . For more information on Kambera complementation, see Klamer (1998: Chapter 8), and Klamer (2000, 2002). . See Note 3 for a motivation why I assume that the semantically generic verbs wà and wàngu are not the same verb (synchronically). . Note that Kambera compound verbs are not serial verbs. For arguments, see Klamer (1998: Chapter 8). . The Kambera clitic cluster that attaches to the PredP contains up to 9 clitics: modal, pronominal and aspectual. The PredP and the clitic cluster cannot be separated from each other, i.e. adjuncts always follow the clitic cluster or precede the Pred P. . Context: Person talking about how easy it is for the people who live in town and have a constant supply of fish. . The object clitic -nya refers to the proposition in the preceding clause (‘you should just let me go’).
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.24 (322)
Marian Klamer . Since a katoda kawindu is a religious statue in the yard, used to address prayers to, in the context of (20) the argument of wàngu is interpreted as a Goal rather than an Instrument. . Mbombang is a specific term for the ground under a house on stilts, the place where domestic animals are kept. . Kamata is the top part of a fruit, the location where a stalk connects it to the tree. Kaba is the husk of a coconut. Disabled Sumbanese people crawl using the top part of a coconut husk to protect their hands. . An alternative would be to analyse wàngu as it appears in (40) as a conjunction introducing a (subordinated) CP. However, in Section 2.6 we will see that Kambera speakers tend to re-analyse wàngu in contexts like (40) as a (main, matrix) verb, which goes against the subordinating conjunction analysis. Moreover, it is unclear whether the latter analysis would be available to Kambera speakers in the first place, since all of the Kambera conjunctions are coordinating (Section 1). Lacking a class of subordinating conjunctions, the CP analysis of a wàngu clause would presuppose the introduction of a new functional category in Kambera. . Context: Main character of folk narrative is assigned with finding a bottle of the (stinking) milk of a mythological human-like person (mini katoba). . Note that if prepositions and conjunctions took pronominal marking (which they do not in Kambera), they would be expected to have an object, not a subject. . This use of wàngu may be compared with English raising verbs such as seem in He seems to lie (< It seems that he lies).
References Allen, Cynthia L. (1995). Case Marking and Reanalysis. Oxford: Clarendon Press. Bresnan, Joan (2001). Lexical-functional Syntax. Oxford: Blackwell Publishers. Gelderen, Elly van (1998). “The Future of for to.” American Journal of Germanic Linguistics & Literatures, 10, 45–71. Harris, A. C. & L. Campbell (1995). Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press. Haspelmath, Martin (1998). “Does Grammaticalization Need Reanalysis?” Studies in Language, 22, 315–351. Heine, Bernd (1992). “Grammaticalization Chains.” Studies in Language, 16, 335–368. Heine, Bernd (1993). Auxiliaries: Cognitive Forces and Grammaticalization. Oxford: Oxford University Press. Heine, Bernd & Mechtild Reh (1982). Patterns of Grammaticalization in African languages. Arbeiten des Kölner Unviersalienprojekts, 47. University of Cologne: Institute of Linguistics. Hopper, Paul & Elizabeth Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Jackendoff, Ray (1990). Semantic Structures. Cambridge, MA: The MIT Press. Klamer, Marian (1998). A grammar of Kambera. Berlin: Mouton de Gruyter.
TSL[v.20020404] Prn:13/04/2004; 16:12
F: TSL5915.tex / p.25 (323)
Multi-categorial items as underspecified lexical entries
Klamer, Marian (2000). “How Report Verbs Become Quote Markers and Complementisers.” Lingua, 110, 69–98. Klamer, Marian (2002). “Report Constructions in Kambera (Austronesian).” In Tom Güldemann & Manfred von Rucador (Eds.), Reported discourse (pp. 323–340). Amsterdam: John Benjamins. Lightfoot, David (1991). How to Set Parameters: Arguments from Language Change. Cambridge, MA: MIT Press. Lord, Carol (1973). Serial Verbs in Transition. Studies in African Linguistics, 4, 269–296. Maling, Joan (1983). “Transitive Adjectives: a Case of Categorial Reanalysis.” In Frank Heny & Barry Richards (Eds.), Linguistic Categories: Auxiliaries and Related Puzzles (pp. 253– 289). Dordrecht: D. Reidel Publishing Company. Rappaport Hovav, Malka & Beth Levin (1998). “Morphology and Lexical Semantics.” In Arnold Zwicky & Andrew Spencer (Eds.), Handbook of Morphology (pp. 248–271). Oxford: Blackwell Publishers. Rosch, Eleanor (1978). “Principles of Categorization.” In E. Rosch & B. B. Lloyd (Eds.), Cognition and Categorization (pp. 27–48). Hillsdale, NJ: Lawrence Erlbaum Assoc. Vincent, Nigel (1999). “The Evolution of C-Structure: Prepositions and PPs from IndoEuropean to Romance.” Linguistics, 37, 1111–1153.
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.1 (325)
The acquisition of polysemous forms The case of bei2 (“give”) in Cantonese Kwok-shing Wong University of Hong Kong
.
Introduction
In the last two decades, parallels have been drawn between diachronic changes and ontogenetic processes, for example, in the acquisition of the noun classifier system by Mandarin speaking children (Erbaugh 1986) and the present perfect by English speaking children (Slobin 1994). The discussion seemed to have reached its height when it was found that Mandarin speaking children apparently relexify the object-marker bˇa to mean “hold” which is ungrammatical in the adult language (Ziegeler 1997). The term ‘ontogenetic grammaticalization’ was thus coined to account for the phenomenon whereby lexical items acquire grammatical meanings or grammatical items become more grammaticalized in the course of first language acquisition. Despite such optimism displayed in extending the theory of grammaticalization to first language acquisition research, caution has been expressed that ontogeny may not recapitulate diachrony (Slobin 2000; Ziegeler 1997, 2000). Slobin (2000) argued that what is salient to the cognition of a two-yearold is different from the processes triggering grammaticalization in adult language. Nevertheless, a more serious concern, as I believe, may be the lack of analyses of ontogenetic data, since comparing ontogeny with diachrony, in theory, requires expertise in both language acquisition and grammaticalization. As a contribution to the discussion, this paper provides detailed developmental data on the use of the polysemous form bei2, meaning “give”, among eight monolingual Cantonese-speaking children. To briefly outline this paper, it first examines the contexts where this polysemous form occurs (Section 2): transfer, dative, permissive and passive. Second, the links between these constructions will be explored (Section 3). After stating my predictions for the developmental sequence in Section 4 and describing the nature of the corpus used in Section 5, a detailed discussion
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.2 (326)
Kwok-shing Wong
on the acquisition of the form bei2 is given in the results and discussion section in 6. There are three aspects to the discussion: the frequency of constructions, age of emergence and verbs used in the constructions. Explanations were also sought to account for the specific developmental sequence. The basic findings of this corpus study are that there exist close interrelationships among the constructional types studied: transfer, dative, permissive and passive. The first three share the same semantics (the core meaning ‘transfer’) whereas the passive construction is linked to the rest via the permissive construction because of their shared syntax. Developmentally, the sequence of acquisition is found to be transfer > permissive > dative > passive, which corresponds closely to the syntactic complexity of the constructions examined and their semantic distance to the meaning ‘transfer’. It is argued that the permissive and the dative constructions must come after the transfer construction because they are syntactically more complex and their functions seem to elaborate the transfer activities concerned: ‘purpose of the transfer’ in the permissive and the ‘particular transfer act’ in the dative. Passive comes last because of its indirect link to the transfer use via the permissive construction and its non-canonical order, namely, patient preceding agent. Detailed longitudinal data on language development such as provided here is felt to be essential because in order to draw valid comparisons between ontogeny and diachrony, sufficient data on both children’s language and historical change is required. It appears that in such comparisons important aspects of acquisition are sometimes overlooked. Two aspects of language acquisition are particularly essential: developmental order and explanations for a given order. In Ziegeler (1997), for example, it has not been made explicit whether the Mandarin children were found to use the object marker bˇa to mean ‘hold’ before the emergence of the grammatical use. That is, the developmental sequence was unclear, and it is not certain from the paper at what age the lexical use started to appear. Also, the proportion of correct uses (bˇa as an object marker) versus bˇa as a ‘verb’ is unknown. A frequency count would have solved the problem by allowing us to know the extent to which the error surfaces in the production data. Finally, without data on the longitudinal development of these children, the anomalous sentence (1), from Ziegeler (1997), may well be an eclipse of a possible sentence such as (2). That is, the theory of grammaticalization may not have been justifiably invoked. (1) Child: Wˇo bˇa yˇızi I take chair “I’m taking/holding the chair.” (2) Possible match: Wˇo bˇa yˇızi [mài-le] I Obj-chair sell-ASP “I have sold the chair.”
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.3 (327)
The acquisition of polysemous forms
Given the number of issues raised in the above paragraph, it is clear that a careful observation on a sufficient quantity of data is needed. In fact, the same object marker bˇa in Mandarin once led to the hypothesis that there is an on-going word order change in Mandarin from SVO to SOV. However, a careful analysis of classical Chinese texts has shown that the basic word order in Mandarin has been stable in the past 2000 years (Sun 1996). Although at this stage there does not yet exist any definitive work on the phenomenon of grammaticalization in Cantonese, a number of researchers have started to look at this language (Wong, in progress; Matthews, Xu, & Yip, in progress), and at languages that are related (Ansaldo 1999; Ansaldo & Matthews 1999) or that are spoken in the Southeast Asian subcontinent (Yap & Iwasaki 1998; Enfield 2000, 2001). Hopefully, this work might serve to help remove a big obstacle for anyone who wishes to compare ontogeny with diachrony.
. bei2: Range of functions Before examining the form bei2, some background information is needed. Firstly, Cantonese is a tonal language (the number that appears in bei2, for example, indicates a high rising tone) spoken in the Guangdong province of China and is the mother tongue of the majority of inhabitants living in Hong Kong. Secondly, in this language, there exists a number of polysemous forms (bei2 “give”, lei4 “come”, dou3 “arrive”, etc.) and each form may appear in a number of different, though related, constructions (Wong, in progress). Returning to the form bei2, it has multiple functions: transfer, dative, permissive and passive, which are distinguished by the specific construction in which the form bei2 occurs and the particular assignment of thematic roles in the construction concerned. The following is an illustration of the four different types of constructions. Transfer (3) giver – bei2 – theme – recipient ngo5 bei2 cin2 nei5 I give money you “I’ll give you money.” Dative (4) (agent) – Verb – (theme) – bei2 – recipient1 ngo5 lo2 di1je5 bei2 nei5 I get something give you “I’ll bring you something.”
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.4 (328)
Kwok-shing Wong
Permissive (5) (causer) – bei2 – causee – Verb – (Object) ngo5 bei2 nei5 tai2 din6si6 2 I let you watch TV “I’ll let you watch TV.” Passive (6) (patient) – bei2 – agent – Verb ngo5 bei2 man1 zam1 I give mosquito bite “I was bitten by a mosquito.” It can be readily seen that the transfer function requires one verb, namely bei2 and allows maximally three noun phrase arguments which are the giver, the theme (mostly a concrete object that is transferred) and the recipient. The reason why ‘maximally’ is used is because, unlike in English or other Germanic languages, noun constituents can be dropped or omitted provided that they can easily be traced in the context or understood in the discourse (Matthews & Yip 1994: 83). Another interesting observation is that unlike in English, the Cantonese transfer construction forces the theme to precede the recipient, unless the theme is exceptionally heavy (Matthews & Yip 1994: 137). For the dative construction, in addition to the three noun arguments as mentioned in the transfer use: the giver, theme and recipient, there is another verb which serves to specify the transfer action involved. In (4), the action involved is ‘to get’. Notice that it might be argued that there are two clauses in (4), one headed by the verb lo2 and the other by bei2. However, syntactically, this analysis is unacceptable, because bei2 cannot carry an aspect marker, as in (7). Only the verb preceding bei2 can carry an aspect marker (8). (7) *ngo5 lo2 di1je5 bei2-zo2 nei5 I get something give-ASP you (8) ngo5 lo2-zo2 di1je5 bei2 nei5 I get-ASP something give you “I’ve brought you something.” In addition, it must be stressed that the bei2 in the dative construction does not function as a benefactive marker. To convey the meaning ‘I’ll watch TV for you’, the verb bong1 (“help”) must be used. Compare (9) and (10): (9) *ngo5 tai2 din6si6 bei2 nei5 I watch TV give you.
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.5 (329)
The acquisition of polysemous forms
(10) ngo5 bong1 nei5 tai2 din6si6 I help you watch TV “I’ll watch TV for you.” The permissive allows three arguments. However, the object that appears at the end of the construction should not be understood as a theme transferred but rather as an argument of the verb that appears after the causee, since there are situations where the object is not transferred at all. In (5), din6si6 (TV) is not transferred, but it is an argument of the verb tai2 “watch”. Secondly, it might be suggested that the same example can be analyzed as an instance of the transfer construction, with the verb phrase [tai2 din6si6] being the theme argument. That is, the permissive construction could perhaps be derived from transfer by the inclusion of a verb clause as its ‘theme’ argument (‘I give you [watch TV]’ > ‘I’ll let you watch TV’). However, such an analysis is problematic because unlike in English, the theme argument in Cantonese canonically precedes the recipient in the transfer construction. It is more likely that the extension from transfer to permissive involves the bleaching of the meaning ‘give’ and the juxtaposition of a verb indicating the permitted act to the main clause. Another observation concerning example 5 is that despite the strong cross-linguistic links between permissive and causative semantics, this example does not mean ‘I’ll make you watch TV’ – the causation is much ‘weaker’ in this case. Finally, unlike in the dative construction, bei2 in the permissive construction can carry an aspect marker, as illustrated in (11). (11) ngo5 bei2-zo2 nei5 tai2 din6si6 I let-ASP you watch TV “I have let you watch TV.” Lastly, in the case of the passive construction, unlike in the afore-mentioned constructions, the semantic role that occupies the first slot, namely the topic position, is a patient, rather than an agent (‘giver’ in the transfer and dative constructions, ‘causer’ in the permissive construction). Nevertheless, despite such a semantic difference, syntactically the passive has the same structure as the permissive: if we ignore the thematic role assignment, they have the same skeletal structure which is Noun – bei2 – Noun – Verb. Figure 1 illustrates the syntactic proximity between the passive and the permissive constructions. Permissive causer – bei2 – causee – Verb | Noun – bei2 – Noun – Verb
=
Passive patient – bei2 – agent – Verb | Noun – bei2 – Noun – Verb
Figure 1. The proximity of the passive construction and the permissive construction
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.6 (330)
Kwok-shing Wong
This proximity presents a challenging situation for Cantonese speakers (let alone children): How are they going to disambiguate between the permissive and passive constructions if the key to disambiguation does not rest on sentence structure, given the syntactic closeness between the two constructions? A misanalysis of causer as patient or vice versa might change the meaning of the sentence completely because, theoretically, a causer is almost the opposite of a patient in terms of volitional control, and the two simply rank differently in the animacy hierarchy (Comrie 1989), with causer ranking higher than patient. Consider again (6), repeated here: (6) Passive (patient) – bei2 – agent – Verb ngo5 bei2 man1 zam1 I give mosquito bite “I was bitten by a mosquito.” A permissive interpretation would be “I allow the mosquito to bite (an unnamed entity).” whereas a passive reading would be “I was bitten by a mosquito” as stated. An attempt will be made in the following section to show how the ambiguity is resolved.
. Links between the functions In this section, I will make explicit the links between the constructions discussed in Section 2. I will argue that the links are instantiated in the areas of semantics and syntax. First, in terms of semantics, the transfer, dative and permissive constructions share a core meaning, namely ‘give’. That is, this common meaning of ‘give’ serves to anchor these constructions directly. As for syntax, the passive is linked to this ‘core’ sense of transfer via the permissive construction because as argued, the two share the same syntactic structure. In other words, the linkage between passive and the other constructions is indirect. Secondly, the syntax of transfer is sometimes retained in other constructions. The following three sub-sections illustrate the shared semantics among the transfer, dative and permissive uses, the linkage of passive to the ‘core’ sense of transfer and the issue of syntactic retention respectively.
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.7 (331)
The acquisition of polysemous forms
. Shared semantics: Transfer, dative and permissive The meaning of ‘give’ is obvious in the case of the transfer construction and therefore, more attention will be paid to the dative and permissive constructions.
.. The dative construction Recall (4), the dative construction has the structure of: agent – Verb – theme – bei2 – recipient Notice that in addition to bei2, there is a verb appearing after the agent. This verb specifies the particular type of transfer activity involved (give by sending, bringing, etc.). As for the form bei2, functionally, it seems to have become specialized in marking the recipient, as omission of the recipient role is unacceptable in the dative construction (see (12)), unlike in the transfer construction, in which noun omission is permissible so long as the context is clear. (12) *ngo5 gei3 di1je3 bei2 I mail something give The following example taken from the file CCC20907.cha in the Cantonese child language database CANCORP further illustrates the underlying meaning of the dative construction and the functional shift of the form bei2. (13) Investigator: zaau2-zo2 bei2 bin1go3 aa3 change-ASP give who SFP “For who have you changed the money?” Child: bei2 maa1maa1 lo1 give mummy SFP “For mummy.” In response to the investigator’s question whereby the main verb is zaau2 “change money” and which seeks information concerning the identity of the recipient, the child replied by saying literally “give mummy” in Cantonese. The meaning of transfer is transparent in that if you are changing money for someone, the money will end up in the hand of somebody or be given to that party.
.. The permissive construction Similar to the dative construction, there is a verb in addition to the form bei2. Nevertheless, this verb serves to specify what the recipient can do with the theme on receiving it if it is concrete, or what the recipient is allowed to do if what is given is the right to act. Example (14) from CANCORP demonstrates the situation whereby an investigator requests for a theme from a child so that she can act on the theme (by eating). The child, however, does not grant her the permission by
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.8 (332)
Kwok-shing Wong
not giving the theme to her. So it is clear that the concept of permission is a natural extension of the concept of ‘giving’: one gives (or does not give) something to a party in order to allow (or disallow) something to happen. (14) Investigator: bei2 ngo5 sik6 aa1 give me eat SFP “Let me eat your food.” Child: m4 bei2 not give “I won’t let you.”
. The linkage of the passive construction to the ‘core’3 Recall that the crucial difference between the passive and the permissive constructions rests on the assignment of thematic role to the topic position: patient in passive while causer in permissive. Nevertheless, there exists a situation whereby this thematic distinction between the two is neutralized. This happens because, when children or adults produce a passive in daily conversation, the subject position (the patient role) is usually empty (Wong, in progress). Sometimes the noun with the patient role may even appear directly after the verb at the end of the construction, with the result that the surface structures in relation to the assignment of thematic roles between the passive and the permissive constructions have become almost identical (Wong, in progress): Permissive: Passive:
bei2 – causee – Verb – theme bei2 – agent – Verb – patient
The resulting structures are extremely similar in terms of syntax and thematic role assignment: a causee is a kind of agent, and a theme argument is similar to a patient argument. Example (15) below (from the file CGK20818.cha) illustrates this interesting situation: (15) Child:
bei2 di1 cung4 aa3 ngaau5 dou2 keoi5 give CL bug Prt bite VerP he “He was bitten by a mosquito.” Investigator: bei2 di1 cung4 ngaau5 dou2 keoi5 me1 give CL bug bite VerP he SFP “Was he really bitten by a mosquito?”
Notice that both the child’s and the adult’s utterances have a passive reading. Nevertheless, on reflection, giving the example a permissive reading, that is, ‘an unnamed
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.9 (333)
The acquisition of polysemous forms
agent allows the bug to bite him’, may not be too bad because the unnamed agent is most likely interpreted as coreferential with the theme/patient keoi5.
. Syntactic retention Cantonese is a language which permits topicalization (Matthews & Yip 1994), a mechanism which promotes elements from non-prominent positions to the topic position. Omitting the agent, a topicalized transfer would thus look like: theme – bei2 – recipient (Compare it with an ordinary transfer without an agent: bei2 – theme – recipient.) It is interesting that this topicalized transfer has ‘coincidentally’ the same structure as part of the dative construction. The part in brackets is identical to a topicalized case of transfer: agent – Verb – [theme – bei2 – recipient] (16) and (17) present cases of topicalized transfer and dative respectively. (16) Topicalized transfer paau2ce1 bei2 nei5 race car give you “(I) give you the race-car.”
(HHC: 2;06.10)
(17) Dative ngo5 maai5 [zi1 faa1 lei4 bei2 nei5] I buy CL flower come give you “I’ll buy you a flower.”
(HHC: 2;11.08)
In the case of the permissive, (18) below shows that the permissive construction can be analyzed as a combination of the transfer construction and a verb to denote the special caused activity. (18) Permissive sin1saang1 bei2 ngo5 sik6 ping4gwo2 teacher give I eat apple “The teacher gave me an apple to eat.” This example can be decomposed into: Transfer: Activity:
sin1saang1 bei2 (ping4gwo2) ngo5 teacher give (apple) I ngo5 sik6 ping4gwo2 I eat apple
(HHC: 2;10.13)
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.10 (334)
Kwok-shing Wong
. Predictions on developmental sequence Summing up what has been presented so far, the transfer, dative and permissive constructions all share the sense of transfer and this sense can be traced most readily in the transfer construction. Secondly, the dative and permissive constructions differ from the transfer construction in that both require an extra verb to denote the specific transfer activity (in the case of dative) and the particular action allowed in the permissive. Finally, passive is seen to be most distant from the transfer function because the most directly observable link is its syntactic structure which strongly resembles the permissive. Based on the above observations, I predict the following developmental patterns in Cantonese children: The transfer sense is acquired first because of its basic nature: the learning of this sense and the associated transfer construction is a pre-condition for learning the other related constructions: dative, permissive and passive. To falsify this claim, cases where children acquire, for example, the dative construction, without first going through the stage of transfer have to be found. Secondly, the dative and the permissive constructions should come after the transfer construction because structurally, both require an extra verb to give extra information: a main verb denoting the special transfer activity in the dative and a secondary verb specifying the allowed action in the permissive. That is, the dative and permissive constructions are structurally more complex and semantically more specific than transfer. Finally, passive should appear last because of its distant resemblance to the transfer construction: it is linked to the transfer use via the permissive construction. Moreover, the order of constituents in the passive is non-canonical in that, unlike in the non-topicalized forms of transfer, the dative and permissive constructions, the subject position is occupied by a patient. In order to validate the above predictions on language development, I have resorted to CANCORP which is currently the only large-scale corpus available for child Cantonese. The following section describes the nature of this corpus.
. Corpus and utterance extraction There are two versions of CANCORP: the original one, which can be found at the CHILDES link on the internet http://childes.psy.cmu.edu, and a modified version (Fletcher, Leung, Stokes, & Weizman 2000), developed at the Division of Speech and Hearing Sciences at the University of Hong Kong. The modified version consists of 128 transcripts of audio-recordings from a longitudinal study of
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.11 (335)
The acquisition of polysemous forms
Table 1. Gender, age range, MLU range of the children Children Younger CCC CGK CKT MHZ Older HHC LLY LTF WBH
Gender
Age range
Range of MLU
male female male male
1;11.21–2;09.07 1;11.22–2;08.18 1;10.30–2;07.02 2;00.03–2;08.06
1.792–3.208 2.020–3.790 1.660–2.890 1.610–3.240
male female female female
2;04.08–3;04.14 2;08.10–3;04.22 2;02.10–3;02.18 2;04.15–3;02.20
1.650–3.529 2.090–4.447 1.760–4.029 1.800–3.327
eight Cantonese-speaking children, aged between 22 and 42 months. This version is adopted because some of the tagging errors that appeared in the original version have been corrected. In addition, to facilitate comparison, some early files (where the children were too young and were not producing many words) were removed from the original, resulting in a reduction of sample size to 128 files, that is, a uniform 16 files for each child. Table 1 describes the gender, the age range of the children and the range of mean length of utterance (MLU). The eight children are divided into an older group and a younger group because some of the children had their first recordings made considerably earlier than the others (Wong, in progress). The utterances containing bei2 are extracted and then categorized into four different constructions, namely, transfer, dative, permissive and passive.4 My approach to classification is conservative in that to count an utterance, for example, as transfer, the syntax has to closely resemble a transfer construction. Also, the five utterances preceding and following the extracted utterances with the form bei2 are examined to ensure that the categorization is semantically sound: this approach avoids the problems of rich interpretation and of rejecting utterances where the order of constituents as used by children differs from that in adult language. For example, children are known to produce transfer utterances with the recipient preceding the theme sometimes. (19) CGK:
jat1zan6 bei2 suk1suk1 cin2 aa3 later give uncle money SFP “I’ll give uncle the money later.”
(2;03.11)
To reject this example as transfer because of its non-canonical order between the recipient and theme will underestimate the total number of transfer utterances.
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.12 (336)
Kwok-shing Wong
. Results and discussion This section is divided into four parts. The first handles the issue of frequency. The second one deals with the age of emergence for the various constructions, while the last subsection is concerned with the collocation patterns observed.
. Frequency Running the frequency command freq +t*CHI +t%mor –t* +o +k *.cha generates a frequency word list of descending order. It is found that bei2 is the 44th most frequently used word by children of Cantonese. This finding is consistent with Newman’s (1996) observation that cross-linguistically the verb ‘give’ is a high frequency form that surfaces early on in child language. Tables 2 and 3 show respectively the average number of occurrences of the various bei2 constructions produced by the two groups of children and the corresponding examples. The frequency data on individual children is provided in Table 4. The results show that frequency of constructions is sensitive to age. Over two-thirds of the bei2 utterances produced by the younger group of children (69%) fall into the category Table 2. Number of occurrences of different bei2 constructions Construction
Children
Dative Passive Permissive Transfer Total
Younger
Older
9.75 (11%) 1.25 (1%) 10.75 (12%) 62.25 (69%) 90
19 (19%) 1.75 (2%) 31.5 (31%) 41.25 (39%) 102.25
Table 3. Examples for different bei2 constructions Children Younger Dative Passive Permissive Transfer Older Dative Passive Permissive Transfer
Examples
Meaning
lo2 je5 bei2 ngo5 aa3. bei2 ngo5 zip3 zyu6 zo2. m4 bei2 maa1mi4 tai2. bei2 cin2 ngo5.
“Get something for me.” “It was picked up by me.” “Won’t let mommy see.” “Give money to me.”
maa1mi4 maai5 bei2 ngo5 gAa3. bei2 man1man1 ngaau5 aa3. m4 bei2 nei5 tai2. nei1 go3 bei2 nei5.
“Mommy bought (this) for me.” “Been bitten by a mosquito.” “Won’t let you see.” “Give you this one.”
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.13 (337)
The acquisition of polysemous forms
Table 4. Frequencies of different constructions in individual children Children Younger CCC CGK CKT MHZ Older HHC LLY LTF WBH Total
Dative
Permissive
Passive
Transfer
9 22 5 3
3 21 11 8
0 5 0 0
31 56 80 82
28 22 23 3 115
22 41 53 10 169
2 1 1 3 12
59 33 53 20 414
of the transfer construction. This is considerably higher than the proportion for the older group of children (39%). In the case of the structurally more complex constructions, the permissive construction is found to be more frequent than the dative construction, especially in the older group of children. This suggests that the lexical function “to let” may have an advantage over the grammatical function of recipient-marking. Finally, passive is rare in either group of children. This supports the cross-linguistic observation that passive is a late-acquired construction.
. Age of emergence After rejecting all the unclear and imitative utterances from the sample, the first time that a particular construction occurred for each child was noted down. Table 5 below presents the age of emergence for each constructional type in each child. It clearly shows variation among the children (which is normal): they exhibit different rates in the acquisition of constructions. By referring to the average row for the younger group of children, the age of emergence of bei2 constructions follows the order of transfer > permissive > dative > passive, which is identical to the pattern for frequency of constructions. That is, the lexical functions of bei2 are more frequent and are acquired earlier than its grammatical functions. The average age of acquisition for the older children is not particularly informative because by the time these children received their first recording, most of them would have already acquired the transfer, dative and permissive constructions, which are, on average, acquired below the age of 2;04. Finally, passive is late-acquired and does not emerge until age 3, on average.
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.14 (338)
Kwok-shing Wong
Table 5. Order of acquisition Children Younger CCC CGK CKT MHZ Average Older HHC LLY LTF WBH Average
Transfer
Dative
Permissive
2;04.10 1;11.22 1;10.30 2;00.03 2;00.24
2;05.23 2;00.08 2;03.17 2;04.07 2;03.14
2;04.10 1;11.29 2;03.17 2;03.28 2;02.29
2;04.08 2;11.01 2;03.02 2;04.15 2;05.22
2;09.30 2;09.14 2;03.02 2;07.14 2;07.15
2;07.21 2;11.08 2;04.27 2;07.14 2;07.25
Passive
2;02.07
Not applicable 3;03.11 3;04.22 2;06.01 2;11.06 3;00.11
. Collocation patterns The following tables show the verbs found in the dative (6), permissive (7) and passive constructions (8) produced by the younger and older children. The verbs are first sorted in descending order by the younger children. It is observed that both groups of children have preference for certain verbs in different constructions, suggesting that the use of constructions may be centered on a few verbs in the early stage of acquisition – a notion that is comparable to the verb-island hypothesis (or the item-based learning hypothesis) of Tomasello (1992, 1995, 1999, 2000, 2001). In the case of the dative construction, the verbs that appear in the pattern serve to further specify a transfer activity by providing details on how the giving is performed. The four most frequent verbs in the younger group of children are: maai5 “buy”, lo2 “carry”, daa2 “dial” and sung3 “give as a gift”, while the top three verbs in the older group follow the pattern of sung3 “send as a gift”, maai5 “buy” and lo2 “carry”. In the case of the permissive construction, the verbs that appear are activity verbs which serve to specify the action permitted. The top four frequently used verbs in the younger group of children are: sik6 “eat”, jam2 “drink”, tai2 “see” and waan2 “play”. A similar pattern is observed for the older group. In the case of the passive construction, the only verb that is shared by both groups of children is ngaau5 “bite”. (20) is an instance of the passive construction produced by a child in the younger group. (20) bei2 di1 cung4 ngaau5-dou3 keoi5 give CL bug bite-part he “He was bitten by a bug.”
(CGK: 2;08.18)
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.15 (339)
The acquisition of polysemous forms
Table 6. Verbs used in the dative construction Verbs maai5 lo2 daa2 sung3 zing2 daai3 ling1 Others Total
Meaning buy carry dial send make bring carry
Children Older
Younger
20 (26%) 4 (5%) 0 (0%) 32 (42%) 4 (5%) 3 (4%) 2 (3%) 11 (15%) 76
12 (31%) 7 (18%) 5 (13%) 5 (13%) 2 (5%) 0 (0%) 0 (0%) 8 (20%) 39
Table 7. Verbs used in the permissive construction Verbs sik6 jam2 tai2 waan2 hoi1 ke4 zaa1 ceot1 co5 jiu3 Others Total
Meaning eat drink see play open ride drive go out sit have
Children Older
Younger
23 (18%) 9 (7%) 48 (38%) 9 (7%) 1 (1%) 1 (1%) 1 (1%) 5 (4%) 0 (0%) 4 (3%) 25 (20%) 126
11 (26%) 5 (12%) 5 (12%) 4 (9%) 2 (5%) 2 (5%) 2 (5%) 1 (2%) 1 (2%) 0 (0%) 10 (48%) 43
Due to the lack of examples, it is difficult to ascertain whether the typological tendency for passive to carry adverse meanings also holds true for language acquisition. To sum up, the developmental data on bei2 does support the hypothesis that developmental sequence is a function of syntactic complexity and semantic resemblance. Constructions that are relatively less complex and resemble the transfer construction more tend to be acquired earlier (and be more frequent) than other constructions. It is also observed from the corpus data that the permissive construction is acquired slightly earlier than the dative construction. This suggests that the lexical function of bei2 (“to let”) might have an advantage over its grammatical function (recipient-marking). Finally, children’s preference for certain verbs in the
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.16 (340)
Kwok-shing Wong
Table 8. Verbs used in the passive construction Verbs
Meaning
laau5 lo2 ngaau5 tai2 zip3 lem2 co5 laai1 ze1 Others Total
scold get bite look catch lick sit catch cover
Children Older
Younger
0 (0%) 0 (0%) 1 (14%) 0 (0%) 0 (0%) 1 (14%) 1 (14%) 1 (14%) 1 (14%) 2 (0%) 7
1 (20%) 1 (20%) 1 (20%) 1 (20%) 1 (20%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 5
constructions studied (permissive, dative and passive) shows that during the initial stage of acquisition, the use of constructions follows item-based learning patterns.
. Conclusion This paper discusses first the contexts (constructions) where the form bei2 occurs, followed by an examination of the inter-relationships among the transfer, dative, permissive and passive constructions: the first three are argued to be intimately linked because of their shared semantics whereas the passive construction is argued to be closely related to the permissive construction because of their syntax. Three specific predictions were then made. First, the transfer use should be acquired first because of its basic nature. Secondly, as children’s language abilities improve over age, more complex constructions such as the dative and the permissive constructions should appear next. Finally, passive should come last because of its distant resemblance to the core meaning ‘give’ and its non-canonical thematic order. These predictions are supported by the CANCORP database. The developmental sequence does appear to be a function of syntactic complexity and semantic resemblance. The order of acquisition and frequencies of constructions follow the order of: transfer > permissive > dative > passive A construction that is the least complex and accommodates the meaning “give” most readily (i.e. the transfer construction) is acquired first (and is more frequent)
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.17 (341)
The acquisition of polysemous forms
whereas a structure that is the most complex and is distant to the meaning ‘give’ (i.e. passive) is acquired last (and is the least frequent). Returning to the issue of the relationship between ontogeny and diachrony, this paper serves to provide concrete data on the developmental change seen in the use of bei2 and may thus function as a yardstick for future comparisons.
Key ASP SFP VerP CL Prt Obj
= = = = = =
Aspect marker Sentence-final particle Verbal particle Classifier Particle Object marker
Notes . Items in brackets can be omitted. . The gloss of bei2 is “let” instead of “give” because in this environment, bei2 is functionally equivalent to the verb “let” or “allow” in English. . Schematically, passive could perhaps be interpreted as a kind of transfer because in a passive situation, the patient receives an impact from the agent. Nevertheless, unless the subject position is empty, the order of thematic roles in a passive is still different from that of a non-topicalized transfer because the topic position is occupied by a patient rather than an agent. . Utterances which contain non-comprehensive units are excluded.
Corpus CANCORP (monolingual Cantonese) child language corpus by T. Lee et al.: http://childes. psy.cmu.edu (folder: canton.zip under ‘Other language corpora’) Fletcher, P., Leung, S., Stokes, S., & Weizman, Z. (2000). Cantonese Pre-school Language Development: A Guide. Hong Kong: Standing Committee on Language Education and Research (SCOLAR).
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.18 (342)
Kwok-shing Wong
References Ansaldo, Umberto (1999). Comparative Constructions in Sinitic: Areal Typology and Patterns of Grammaticalization. Stockholm: Allduplo Tryckeri AB. Ansaldo, Umberto & Matthews, Stephen (1999). “The Min Substrate and Creolization in Baba Malay”. Journal of Chinese Linguistics, 27, 38–68. Enfield, Nicholas (2000). Linguistic Epidemiology: on the Polyfunctionality of ‘Acquire’ in Mainland Southeast Asia. Ph.D. dissertation, University of Melbourne. Enfield, Nicholas (2001). “On Postverbal Modal Use of the Southeast Asian Verb of ‘Acquisition’: a Typological Poise Explanation”. Paper presented at the XIth meeting of the Southeast Asian Linguistics Society, Bangkok. Erbaugh, Mary S. (1986). “Taking Stock: the Development of Chinese Noun Classifiers Historically and in Young Children”. In C. Craig (Ed.), Noun Classes and Categorization (pp. 399–436). Amsterdam and Philadelphia: John Benjamins. Matthews, Stephen, Xu, Huiling, & Yip, Virginia (in progress). “On Chaozhou ‘Intransitive Passives’: Passive and Unaccusative in Jieyang”. Matthews, Stephen & Yip, Virginia (1994). Cantonese: A Comprehensive Grammar. London: Routledge. Newman, John (1996). Give: a Cognitive Linguistic Study. Berlin; New York: Mouton de Gruyter. Slobin, Dan I. (1994). “Talking Perfectly. Discourse Origins of the Present Perfect”. In W. Pagliuca (Ed.), Perspectives on Grammaticalization (pp. 119–133). Amsterdam: John Benjamins. Slobin, Dan I. (2000). “Form-function Relations: How do Children Find Out What They Are?” In M. Bowerman & S. Levinson (Eds.), Language Acquisition and Conceptual Development (pp. 406–447). Cambridge: Cambridge University Press. Sun, Chaofen (1996). Word-order Change and Grammaticalization in the History of Chinese. Stanford, California: Stanford University Press. Tomasello, Michael (1992). First Verbs: A Case Study of Early Grammatical Development. Cambridge: Cambridge University Press. Tomasello, Michael (1995). “Pragmatic Contexts for Early Verb Learning”. In M. Tomasello & W. Merriman (Eds.), Beyond Names for Things: Young Children’s Acquisition of Verbs (pp. 115–146). Hillsdale, NJ: L. Erlbaum. Tomasello, Michael (1999). The Cultural Origins of Human Communication. New York: Cambridge University Press. Tomasello, Michael (2000). “Do Young Children Have Adult Syntactic Competence?” Cognition, 74, 209–253. Tomasello, Michael (2001). “The Item-based Nature of Children’s Early Syntactic Development”. In M. Tomasello & E. Bates (Eds.), Language Development: The Essential Readings (pp. 169–186). Malden, MA: Blackwell Publishers. Wong, Kwok-Shing (in progress). The Emergence of bei2 (“give”) Constructions in Cantonese-speaking Children. Ph.D. dissertation, University of Hong Kong.
TSL[v.20020404] Prn:13/04/2004; 16:13
F: TSL5916.tex / p.19 (343)
The acquisition of polysemous forms
Yap Foong-Ha & Shoichi Iwasaki (1998). “A Typological and Semantic Analysis of the Relationship between Causatives and Passives in the Periphrastic ‘Give’ Constructions of some East and Southeast Asian languages”. Paper presented at the Conference on Southeast Asian Linguistics (SEALS), Kuala Lumpur. Ziegeler, Debra P. (1997). “Retention in Ontogenetic and Diachronic Grammaticalization”. Cognitive Linguistics, 8, 207–241. Ziegeler, Debra P. (2000). Hypothetical Modality: Grammaticalization in an L2 Dialect. Amsterdam & Philadelphia: John Benjamins.
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.1 (345)
Phonetic absence as syntactic prominence Grammaticalization in isolating tonal languages* Umberto Ansaldo and Lisa Lim University of Amsterdam
.
Grammaticalization in isolating tonal languages
One of the salient aspects of the grammaticalization process is the physical reduction of morphemes that evolve from a lexical to a grammatical item. This reduction process typically functions as an indicator of the fact that grammaticalization has occurred. In this paper we address the following question: how does this apply to languages of the isolating, tonal type? The nature of reduction and erosion in isolating tonal languages (ITLs) is an important question for the theoretical status of grammaticalization studies severely criticized in recent works (cf. Campbell & Janda 2001; Newmeyer 1998, 2001). It is obvious that, if we cannot account for the emergence of grammar in isolating languages within the framework of grammaticalization, the universal validity of such a framework will be undermined. Strongly isolating languages typically do not allow yesterday’s syntax to become today’s morphology (cf. Ansaldo 1999). In Sinitic languages, where syllable boundaries are discrete and phonotactic constraints rule out reduced syllables of the kind observed elsewhere, the material available for reduction is not easily found at the morphological level. Does this mean that grammaticalization cannot occur in these languages? We argue that grammaticalization does occur in ITLs, but can more often be found in suprasegmental features. Additionally, however, tones have lexical significance in ITLs, and can occur in different registers and show various contour behaviors, depending on the language. The suprasegmental features available to reduction/erosion are therefore not only more than one, but are also complex, and need to be thoroughly investigated. To illustrate this, our paper explores the reduction and erosion observable in functional words of two Sinitic varieties: Cantonese and Hokkien as spoken in Singapore. In these languages, morphologically strongly isolating and tonally rather complex, we can observe a clear
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.2 (346)
Umberto Ansaldo and Lisa Lim
tendency for grammaticalized morphemes to exhibit phonetic erosion in terms of vowel/syllable reduction compared to their lexical counterparts. This, we suggest, may be related to the prosodic prominence of the constructions in which the grammaticalized morphemes occur. Indeed, we claim, in ITLs of the type investigated here, prosodic erosion may be the most salient diagnostic of grammaticalization effects and it is therefore of extreme relevance to a universally acceptable framework of grammaticalization that such differences be thoroughly investigated.
. Phonetic erosion in Sinitic In ITLs like those found in the Chinese family, suprasegmental features can show traces of erosion in grammaticalized constructions: Mandarin,1 for example, exhibits (a) tonal reduction, in the sense that the grammaticalized morpheme guo51 is assigned a neutral tone (e.g. (2)), or (b) weakened stress, as on the morpheme zai51 (e.g. (4)): (1) ta55 guo51 ma214 lu51 s/he pass road “s/he crosses the road” (2) wo214 chi55 guo fan51 I eat exp rice “I have eaten” (3) wo214 zai51 jia55 I be.at home “I am at home” (4) wo214 zai51 jia55 kan51 shu55 I loc home read book “I am reading at home” Within Sinitic varieties, however, Mandarin is typologically not always representative of the whole family (cf. Ansaldo 1999). With regard to tonal categories, for example, Mandarin exhibits a simpler system than southern varieties of Sinitic. Two aspects of the Mandarin system are of particular relevance for grammaticalization studies: it has only one register, and it does allow the neutral tone option. Below we look at the significance of these two features.
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.3 (347)
Phonetic absence as syntactic prominence
. Tonal registers in Sinitic The absence of tonal contrast in different registers – i.e. where the relevant feature of word-identifying pitch behavior is the relative height of the syllabic pitches concerned within the speaker’s pitch-span – in the tonal system of Mandarin creates a relatively simple environment in which to observe phonetic erosion. A tone perceived to be lower than average, for example, might already be interpreted as erosion; weakened stress, as shown above, may be realized as a lower tone and interpreted as such.2 Sinitic varieties such as Cantonese and Hokkien, however, all present us with three different registers: low, mid and high. A low tone is just as marked (or unmarked) as a higher one and does not constitute an indication of erosion. It is this challenging aspect which restricts the range of possible interpretations for variation that lies at the basis of our decision to investigate grammatical markers in Cantonese and Hokkien.
. The neutral tone option and tone sandhi Mandarin has a neutral tone category, neutralization meaning lack of tonal contrast. Neutral tones have variable realizations, i.e. they can coarticulate with adjacent tones as much as they want. The realization is dependent on context since different neutralized morphemes tend to be realized similarly in the same context. This could be the direct consequence of tonal erosion (Zhang p.c.). Indeed, the neutral tone is often associated with functional words in Mandarin, a clear indication that tonal erosion has happened. The neutral tone option is not available in Cantonese and to the best of our knowledge has not been argued for in Hokkien either. As pointed out in Bao (1999) the default tone varies from language to language and can be postulated only on language-specific tone sandhi (i.e. tonal changes that may depend on phonological context and apply across word boundaries, see e.g. Pike 1947; Yip 2002) facts. In Hokkien, sandhi patterns have been described along the lines of the ‘tonal clock’ (Bodman’s tone cycle, cf. Bodman 1987). This means that each tone changes in a cycle between a citation and a context tone, except for the tone of the final tone-bearing unit (TBU).3 In Cantonese, sandhi is related to lexicalization and not grammaticalization. In this light it seems to make sense to look firstly for phonetic (as well as phonological) changes accompanying grammaticalization.
. The grammaticalized constructions examined In this section we briefly illustrate the nature of the constructions under investigation. The choice of the following constructions is motivated by several factors: it
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.4 (348)
Umberto Ansaldo and Lisa Lim
needs to be kept in mind that the study of grammaticalization in Sinitic varieties is not as advanced as, say, for Indo-European languages. In that sense, especially in the case of non-Mandarin varieties, we find several markers that would not be accepted as grammaticalized without controversy, as well as a whole series of constructions still to be properly described in terms of grammaticalization. We have therefore focused on choosing structures that have been either reasonably described in the literature or that may be regarded not too controversial candidates for grammaticalized morphemes. In order to be able to appreciate the differences between lexical and grammaticalized morphemes, we obviously had to choose items that, though grammaticalized, still retain some vestiges of lexical usage. The rather high degree of polyfunctionality found in isolating languages of East and Southeast Asia (cf. Ansaldo 1999; Enfield 2000) facilitates this task though one is still faced with complications (cf. Section 4 below).
. Cantonese4 The comparative marker gwo33 found in Cantonese has been described in terms of functions and grammaticalization pathway in Ansaldo (1999): gwo33 : “to pass/cross” > V/ RV (V2) > surpass marker As we can see below, the full lexical function is still preserved in very limited contexts: (5) ngo23 gwo33 ma23 lou22 I cross road “I cross the road” Also, we can still observe some semi-verbal uses in constructions such as (6). This is not uncommon as polyfunctionality tends to be more frequent in isolating languages of this type: (6) keoi23 haang21 gwo33 di55 saan53 s/he walk pass pl mountain “s/he crossed the mountains” Most frequently however, the morpheme gwo33 can be found suffixed to a verb as a comparative marker, e.g.: (7) ngo23 daai21 gwo33 lei23 I big sur you “I’m taller than you”
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.5 (349)
Phonetic absence as syntactic prominence
The second morpheme studied for Cantonese is the resultative verb dou33 “to arrive” (cf. Yip & Matthews 2001) as in: (8) geng55 dou33 sei35 scare rv die “scared to death” Note that this item can be used lexically as in: (9) keoi23 dou33 zo35 mei22 aa33 s/he arrive asp neg ptcl “has s/he arrived yet?”
(Matthews p.c.)
. Hokkien5 As in the case of Cantonese, for Hokkien two structures are chosen, one of which has been previously described in literature on grammaticalization, the other one not. The former concerns the polyfunctional marker ho33>21 la]24 : literally “to give people”, which can function as a verb (10), as a causative marker (11), as a permissive marker, (12) and as a passive (13) (cf. Bodman 1987; Yap & Iwasaki 1998): (10) i55>33 ho33>21 la]24>33 te33 s/he give people bag “s/he gave them a bag” (11) i55>33 ho33>21 la]24>33 sien33 s/he caus be.bored “s/he made them bored” (12) i55>33 ho33>21 la]24>33 tsia‘5 s/he perm eat “s/he let them eat” (13) i55>33 ho33>21 la]24>33 me33 s/he pass scold “s/he was scolded” The suggested path for the grammaticalization of the passive marker is as follows: lexical give → permissive → passive (most grammaticalized, cf. Yap & Iwasaki 1998; Matthews et al. 2001). Finally we consider the aspectual (perfective) marker khi21 : “to go” (cf. Bodman 1987) as in: (14) i55>33 si53>55 khi21>53 liao53 s/he die asp asp/ptcl “s/he died”
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.6 (350)
Umberto Ansaldo and Lisa Lim
and in its lexical occurrence: (15) su(53>55) khi21>53 liao53 Sue go asp/ptcl “Sue left already”
. Materials and data collection and analysis . Controlling the data A number of variables can potentially confound the measurements of duration, fundamental frequency (F0) and intensity, and thus the phonological environment in which the morphemes under investigation are found was controlled as far as possible: i.
to avoid phrase-final prominence, particularly lengthening, morphemes were placed in non-phrase-final position; ii. as the duration of an item decreases with an increasing number of syllables in the word or utterance, the number of syllables in the utterances was kept constant, or at least was kept comparable across sets A and B (the lexical vs. the grammaticalized sets of morphemes); iii. in the Hokkien material, to control the potential effect of lowering of pitch caused by the lower tone of the following syllable, the item following the morpheme was of 55 or 33 tone; iv. in the Hokkien material, tone sandhi was taken into consideration; v. to avoid differences due to fundamental frequency declination across an utterance, test items were located in identical or at least approximately similar positions in comparable utterances across sets A and B. The following examples illustrate the kinds of control exerted over the material.
.. ho33>21 “to give” (Hokkien) Because ho33>21 la]24>33 “to give people” gives a passive reading by default, all utterances used a {i55>33 ho33>21 la]24>33 X} “s/he gives them (people) X” construction, where, as an example of a set, X is kiu55 , which can give the meaning of either “s/he gave them ginger”, “s/he made them shrink”, “s/he let them shrink”, or “s/he was shrunk”, with a context provided to elicit each of the readings, e.g.: Context: “They needed various ingredients to cook that particular dish, and everyone contributed something. He gave them ginger.”
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.7 (351)
Phonetic absence as syntactic prominence
(16) i55>33 ho33>21 la]24>33 kiu55 s/he give people ginger Context: “The magician wasn’t careful with his spells and didn’t keep an eye on the children. So what happened? He made them shrink (by accident).” (17) i55>33 ho33>21 la]24>33 kiu55 s/he perm shrink Context: “The scientist was experimenting with some chemicals on some people. What did he manage to do? He made them shrink (on purpose).” (18) i55>33 ho33>21 la]24>33 kiu55 s/he caus shrink Context: “He made the wizard furious with him. So what happened to him? He was shrunk.” (19) i55>33 ho33>21 la]24>33 kiu55 s/he pass shrink
.. khi21 (Hokkien) Pairs of utterances were created where the phonological environment was virtually identical, with the difference lying in whether the item being tested was the lexical or grammaticalized morpheme, e.g.: (20) su(55>33) khi21>53 liao53 Sue go asp/ptcl “Sue has gone” (21) su55 khi21>53 liao53 lose asp already “lost already” (22) a bwe(33>21) khi21>53 pasat bwe33>21 kue53 Ah Buay go market (Malay ‘pasar’) sell cake “Ah Buay goes to the market to sell cakes” (23) i55>33 bwe33>21 khi21>53 pasat bwe33>21 e24>33 kue53 s/he sell asp market sell poss cake “s/he sold the market’s cakes”
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.8 (352)
Umberto Ansaldo and Lisa Lim
.. gwo33 (Cantonese) This morpheme has a very restricted verbal usage, viz. one can gwo malou/tiu kiu/tiu ho “cross road/ bridge/ river” but not much else, which made for difficulties in creating comparable sentences. Nonetheless, utterances were constructed keeping the number of syllables and phonological environment as comparable as possible, e.g.: (24) a ceong21 gwo33 ma23 lou22 Ah Cheong cross road “Ah Cheong crosses the road” (25) ho21 ceong21 gwo33 ma23 lou22 river long sur road “the river is longer than the road”
.. dou33 (Cantonese) Again, pairs of utterances were created where the number of syllables and the phonological environment were virtually identical, with the difference lying in whether the item being tested was the lexical or grammaticalized morpheme, e.g.: (26) a tim21 dou33 liu23 Ah Tim arrive already “Ah Tim has arrived” (27) keoi23 tim21 dou33 sei35 s/he sweet asp die “s/he is really sweet”
. The subjects Subjects were 6 male university undergraduates, 3 of whom had acquired Hokkien and 3 Cantonese as a first language in childhood, alongside English, and who use the language on a daily basis with members of the family. The utterances were recorded in the phonetics laboratory, and were analyzed using the Kay Multispeech speech analysis software, with the following measurements being made: i. ii. iii. iv.
the first and second formant frequencies (F1 and F2) of the vowels; syllable and vowel duration; fundamental frequency (F0) at the start, middle and end of the vowel; and intensity peak.
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.9 (353)
Phonetic absence as syntactic prominence
Paired 2-tailed t-tests were subsequently conducted on the means of the measurements (i) to (iv) for all the morphemes, to test if the difference between the means of the two sets of data, viz. the lexical and the grammaticalized items, is significant.
. Discussion . Vowel quality The lexical items in general have a quality in the more extreme reaches of vowel space. The vowel in lexical gwo is significantly lower (p < 0.05) and more back (p < 0.001) than its grammaticalized counterpart. The vowel in lexical dou starts lower and more back and moves to become higher and more round (p < 0.005), showing more diphthongal movement than its grammaticalized counterpart. The vowel in lexical khi, when pronounced [kh G] by one speaker, is significantly higher (p < 0.005) and more back (p < 0.01). Lexical ho has a vowel quality which is more back; the difference from the permissive morpheme is significant (p < 0.05), while the difference from both the causative and passive morphemes is somewhat less so (p < 0.1). Lexical ho also tends to be more diphthongal.
. Duration As Figures 1 and 2 illustrate, with all four items, the grammaticalized morpheme is significantly shorter than its lexical counterpart. Grammaticalized gwo with a mean syllable duration of 0.141s is significantly shorter than lexical gwo with 0.224s (p < 0.001). Similarly, grammaticalized dou with a mean syllable duration of 0.148s is significantly shorter than lexical dou with 0.219s (p < 0.005).
0.224
gwo
0.141 0.219
dou
0.148
lexical gram
0.233
khi
0.179 0
0.05
0.1
0.15
0.2 0.25 duration/s
Figure 1. Duration of lexical and grammaticalized gwo33 , dou33 and khi21
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.10 (354)
Umberto Ansaldo and Lisa Lim
0.199
0.162
lexical permissive
0.109
causative
0.105
0.166
passive
0.114
0.158
0
0.05
0.179
0.1
0.15
0.2
ho lang
0.25
0.3
0.35 0.4 duration/s
Figure 2. Duration of lexical and grammaticalized ho33>21 la]24>33
Grammaticalized khi is also significantly of a shorter duration at 0.179s compared to the lexical counterpart at 0.233s (p < 0.05). Grammaticalized ho and la] are both of a shorter duration than their lexical counterparts, with the causative and passive morphemes showing a more significant difference (ho p < 0.01 for both; la] p < 0.005 for the causative and p < 0.01 for the passive) than the permissive morpheme (ho p < 0.1; la] n.s.).
. Fundamental frequency (F0) As seen in Figures 3 and 4 respectively, with the Cantonese morphemes gwo and dou, the grammaticalized item is lower in pitch than its lexical counterpart, on av160 140 120
max
100 F0/Hz
gwo-lexical gwo-gram min
80 60 40 20 0
F0-1
F0-2
F0-3
max
135.77
135.77
135.77
gwo-lexical gwo-gram min
118.42
114.9
113.85
116.1 99.91
112.6 99.91
111.45 99.91
Figure 3. F0 (at 3 points) of lexical and grammaticalized gwo33
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.11 (355)
Phonetic absence as syntactic prominence 160 140 120 max dou-lexical dou-gram min
100 F0/Hz
80 60 40 20 0
F0-1
F0-2
F0-3
max
142.08
142.08
142.08
dou-lexical dou-gram min
119.22
114.37
113.84
113.93 104.11
111.46 104.11
114.93 104.11
Figure 4. F0 (at 3 points) of lexical and grammaticalized dou33 160 140 120
max khi-lexical khi-gram min
100 F0/Hz
80 60 40 20 0
F0-1
F0-2
F0-3
max
141.02
141.02
141.02
khi-lexical
118.7 110.52 90.48
111.12 106.73 90.48
108.34 104.63 90.48
khi-gram min
Figure 5. F0 (at 3 points) of lexical and grammaticalized khi21>53
erage by about 2–3Hz, but the differences are not statistically significant. The pitch of grammaticalized khi, lower on average by 5–8Hz compared to its lexical counterpart, as illustrated in Figure 5, does make for a difference which is significant (p < 0.05). The differences in pitch, seen in Figure 6, for the three grammaticalized
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.12 (356)
Umberto Ansaldo and Lisa Lim 160 140 120 100 F0/Hz
max lexical permissive causative passive min
80 60 40 20 0
max lexical permissive causative passive min
F0-1 143.79 124.73 120.15 117.45 120.42 97.77
F0-2 143.79 119.57 123.7 120.17 121.33 97.77
F0-3 143.79 112.55 116.73 114.76 116.22 97.77
F0-4 143.79 109.02 116.56 111.15 108 97.77
F0-5 143.79 106.09 105.95 103.86 104.25 97.77
F0-6 143.79 106.79 106.23 104.43 103.18 96.98
Figure 6. F0 (at 3 points) of lexical and grammaticalized ho33>21 la]24>33
morphemes for ho and la] compared to their lexical counterparts are however all not significant.
. Intensity Figure 7 illustrates the intensity of all the lexical and grammaticalized morphemes. Grammaticalized gwo is lower in intensity at 71.66dB than its lexical counterpart at 72.74dB (p < 0.01). Similarly grammaticalized dou has a lower intensity at 72.28dB compared to lexical dou at 75.49dB (p < 0.1). Grammaticalized la] is significantly lower in intensity than lexical la], once again with the causative and passive morphemes having a more significant difference (p < 0.05) than the permissive (p < 0.01). No significant differences in intensity are found however for ho, nor for khi.
. Summary Table 1 provides a quick summary to the findings discussed above and indicates the features found for the grammaticalized morphemes as compared to those of their lexical counterparts. Given the findings discussed above, it is beyond doubt that phonetic erosion can be consistently observed in the grammaticalized morphemes while it does not
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.13 (357)
74.4 74.08
72
72.28
73
72.74
74
71.66
intensity/dB
75
75.27
76
74.01 73.38 73.08
75.49
77
76.32 75.89 74.82 75.27
Phonetic absence as syntactic prominence
lexical gram 1 gram 2 gram 3
71 70 69
gwo
dou
khi
ho
lang
Figure 7. Intensity of lexical and grammaticalized gwo33 , dou33 , khi21 and ho33>21 la]24>33 (where for the latter morpheme, gram 1: permissive, gram 2: causative, gram 3: passive)
occur, or occurs to a lesser degree, in the lexically used occurrences. More specifically, this consistent erosion occurs to a significant degree in syllable duration, as well as in vowel quality, the effects of the latter being seen either as an undershoot in vowel quality or a reduction in diphthongal quality. The results for fundamental frequency call for special comment: while one might normally expect to see a reduction in pitch with grammaticalized items, as one tends to do in languages not of the isolating tonal type, the findings indicate Table 1. Features of grammaticalized morphemes Morpheme
Vowel quality
Duration
Intensity
Pitch
gwo33
shorter**** shorter *** shorter**/ns shorter***/**/ns shorter**
lower** lower ns lower ns lower*/ns lower ns
lower ns lower ns variable ns variable ns lower*/ns
dou33 ho33>21 la]24>33 khi21>53 Notes: i. ii. iii. iv.
**** p < 0.001; *** p < 0.005; ** p < 0.01; * p < 0.05; ns p <0.1, not significant. [gwf] > [gf] in all cases of the grammaticalized item. [ho] > [huo] in 1 speaker in the lexical (verbal) item. [kh i] > [kh G] in 1 speaker in both lexical (verbal) and grammaticalized items.
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.14 (358)
Umberto Ansaldo and Lisa Lim
this not to be the case, not to any significant degree, with our data. This can actually be seen to make perfect sense in the domain of tonal languages: if tonal contrast needs to be maintained for linguistic contrast, then any erosion that accompanies grammaticalization has to occur elsewhere. This is particularly so in tone languages such as Cantonese and Hokkien which have different registers for tone, where a mid level tone, for example, reduced in pitch height, may be misinterpreted as low level. With tone languages which do not make a distinction between pitch registers such as Mandarin, such a reduction would not entail as severe an implication.
. Significance Our findings present several direct implications for grammaticalization theory: Implication I:
They are prima facie evidence that phonetic erosion does occur in ITLs.
This can be seen primarily in (a) vowel quality, where the grammaticalized items occupy more centralized vowel space or whose diphthongs are more monophthongized; and (b) duration, where the grammaticalized morphemes are almost half in length compared to the lexical counterparts; and, to a less significant degree, (c) pitch and loudness, where the grammaticalized items have relatively lower fundamental frequency and intensity. Implication II:
This reduction of phonetic form shows that grammaticalization effects occur even in languages of the strongly isolating, tonal type.
This is even clearer when we consider the various degrees of semantic bleaching as well as syntactic obligatorification on the structural level that accompany the reduction of phonetic material (see Bisang 1991, Matisoff 1991, Bisang 1996, Ansaldo 1999). Though in these languages the morphological structure does often not undergo the obvious reductions observable in languages of the non-isolating type, we can positively claim that grammaticalization processes take place, thus adding weight to universal claims behind grammaticalization. Implication III: The phonetic reduction is related to syntactic prominence. It is important to note that the reduction observed can be due to the fact that when grammaticalized, the morpheme tends to occur as an element of a syntactically and semantically closer/tighter phrasal unit, usually in a prosodically weak position, i.e. between two content morphemes, e.g. (29) and (31), where khi and gwo directly
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.15 (359)
Phonetic absence as syntactic prominence
follow a content word with which they form a tight unit, while (28) and (30) by contrast show khi and gwo as the main predicates: (28) [su][khi (liao)] s – w “Sue has gone” (Hokkien)
(29) [[si khi] liao] s – w “died already” (30) [ngo][gwo malou] s – s–s “I cross the road” (31) [keoi][[daai gwo][lei]] s – w – s “s/he is bigger than you”
(Cantonese)
To formalize, the items/constructions studied in this paper could be represented as follows, where (a) gives the structure in which the morpheme occurs as a lexical item, and (b) gives the grammaticalized construction. The stress pattern (SP) is given to the right: I.
Cantonese gwo:
(a) verbal: (b) comparative: II. Cantonese dou: (a) verbal: (b) resultative III. Hokkien ho la]: (a) verb-object (b) passive IV. Hokkien khi: (a) verbal (b) perfective
[NP-gwo-NP]; SP: s-s-s [VP-gwo-NP]; SP: s-w-s [NP-dou]; SP: s-s [V-dou-V]; SP: s-w-s [ho la]-NP]; SP: s-s-s [NP-ho la]-V]; SP: s-w-w-s6 [NP-khi]; SP: s-s [V-khi]; SP: s-w
It appears that there is a tendency to reduce duration, pitch and intensity to maintain or emphasize the trochaic (strong-weak) foot (also found in compounds). In other words, the observed reduction may be due to secondary prosodic modifications, possibly all related to duration. In particular, the grammaticalized morpheme typically occurs in the proximity of metrically stressed items, assuming that lexical items are more stressed than functional ones (cf. Duanmu 1997).7 This may be the very reason, or at least one of the reasons, why morphemes becoming grammaticalized tend to undergo phonetic reduction as they find themselves next to, or sandwiched between, content morphemes, which are naturally stressed or accented.
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.16 (360)
Umberto Ansaldo and Lisa Lim
. Conclusions We hope to have provided clear, if preliminary, evidence for the fact that grammaticalized items as the ones discussed above show phonetic erosion when compared to their lexical counterparts in Sinitic languages. The erosion is primarily in terms of duration and vowel quality. This is a significant result as there has not been any previous agreement about the nature of phonetic differences between grammaticalized and non-grammaticalized lexical items. It is obvious that this is only the first step in an area that still needs to be more thoroughly researched and understood. Nonetheless we hope to have provided a framework for investigating phonetic reductions in tonal languages that future researchers will be able to develop further. The consistency of the initial results however shows a clear difference in the phonetic realization of grammaticalized markers in Cantonese and Hokkien that needs to be taken seriously. We have also claimed that it is of extreme relevance to a universally acceptable framework of grammaticalization that such differences be investigated for ITLs as they may be the most salient diagnostics of the grammaticalization process. As observed in Section 1, typologically different languages will show grammaticalization effects in different ways and therefore we should not expect the old adage of yesterday’s syntax becoming today’s morphology as universally valid. In ITLs, as we have shown above, we may have to look at other aspects of structure to find grammaticalization effects. Finally, we have suggested that the mechanism by which the reductions occur may be related to metrical structure. When grammaticalized morphemes typically occur next to syntactically salient, stressed units, they are assigned weak stress, the first step in the phonetic erosion process. To conclude, we have argued for the following three, related points: I.
Grammaticalized morphemes in languages of the isolating tonal type undergo erosion of phonetic features when compared to their lexical counterparts. II. This erosion is a significant diagnostic of grammaticalization available in these languages as the manifestations of grammaticalization are relative to the morphological typology of the language being investigated. III. The mechanisms by which erosion occurs may be related to metrical patterns of the syntactic construction in which an item grammaticalizes.
List of abbreviations ASP aspect CAUS causative EXP experiential
LOC NEG PASS
locative negative passive
s SP SUR
strong stress pattern surpass comparative
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.17 (361)
Phonetic absence as syntactic prominence
F0 F1 F2 ITLs
fundamental frequency formant 1 formant 2 isolating tonal languages
PERM PL POSS PTCL
permissive RV resultative verb plural TBU tone-bearing unit possessive w weak particle
Notes * We want to express our gratitude to Stephen Matthews at Hong Kong University and Zhang Jie at Harvard University for sharing their invaluable expertise with us. We also want to thank Alan Goh and Alvin Leong at NUS for their intuitions and judgments on data. Finally, many thanks to Debra Ziegeler, Richard Wong and other participants at NRG2 for comments. . For Mandarin we use Hanyu Pinyin, with tones given in pitch level numbers. . In this respect, Greenberg and Zee (1979) show that speakers cannot hear contour tones below 90ms. . Wright (1983) suggests that this is not the case: context tones should be treated as basic and citation ones as derivative; this has significant implications for the theory but not for the purpose of this paper. We will therefore follow the traditional sandhi predictions in this paper. . For Cantonese we follow the Linguistic Society of Hong Kong romanization, with tones in pitch level numbers. . For the Hokkien data we use IPA throughout, with tones in pitch level numbers. The tones are indicated as ‘predicted’ by sandhi rules. . In some related varieties the contraction: ho la] > ho] occurs. This would yield a stress pattern of ‘NP-ho]-V’; SP: [s-w-s]. . It is far from established whether metrical prominence can be related to morphosyntactic patterns. While it is true that all four morphemes under investigation could occur in what appears to be a destressed position, within phonological theory the necessity of postulating an additional tier of accent for Sinitic languages (cf. e.g. Chen 2000) has been criticized. Most recently, for example, Bao (2002, 2003) undermines the overall necessity of metrical structure in Chinese phonology, presenting a view in which tone systems as the ones found in Sinitic varieties oscillate between tone and accent, rather than combine them, with no system being clearly metrical.
References Ansaldo, Umberto (1999). Comparative Constructions in Sinitic. Areal Typology and Grammaticalization Patterns. Ph.D. dissertation, Stockholm University. Bao, Zhiming (1999). The Structure of Tone. New York & Oxford: Oxford University Press.
TSL[v.20020404] Prn:13/04/2004; 16:22
F: TSL5917.tex / p.18 (362)
Umberto Ansaldo and Lisa Lim
Bao, Zhiming (2002). “Accentualism in Chinese”. Paper presented at the International Symposium on Chinese Language and Linguistics-8, Academia Sinica, Taiwan, 8–10 November 2002. Bao, Zhiming (2003). “Tone, accent and stress in Chinese”. Review article of Matthew Y. Chen, Tone Sandhi: Patterns across Chinese Dialects. Cambridge University Press, 2000. Journal of Linguistics, 39, 147–166. Bisang, Walter (1991). “Verb serialization, grammaticalization and attractor positions in Chinese, Hmong, Vietnamese, Thai and Khmer”. In H. Seiler & W. Premper (Eds.), Partizipation: Das sprachliche Erfassen von Sachverhalten (pp. 509–562). Tübingen: Gunter Narr Verlag. Bisang, Walter (1996). “Areal typology and grammaticalization: Processes of grammaticalization based on nouns and verbs in South-east Asian languages”. Studies in Language, 20(3), 519–597. Bodman, Nicholas C. (1987). Spoken Amoy Hokkien. New York: Spoken Languages Services. Campbell, Lyle & Richard Janda (2001). “Introduction: Conceptions of Grammaticalization and their Problems”. Language Sciences, 23, 93–112. Chen, Matthew (2000). Tone Sandhi. Cambridge: Cambridge University Press. Duanmu, San (1997). “Recursive Constraint Evaluation in Optimality Theory: Evidence from cyclic compounds in Shanghai”. Natural Language and Linguistic Theory, 15, 465– 507. Enfield, Nick (2000). “On the Polyfunctionality of ‘Acquire’ in Mainland Southeast Asia. A Case Study in Linguistic Epidemiology”. Ph.D. dissertation, University of Melbourne. Greenberg, Stephen & Eric Zee (1979). “On the Perception of Contour Tones”. UCLA Working Papers in Phonetics, 45, 150–164. Matisoff, James M. (1991). “Areal and Universal Dimensions of Grammaticalization in Lahu”. In Elizabeth C. Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization, Vol. II (pp. 383–453). Amsterdam & Philadelphia: John Benjamins. Matthews, Stephen J., H. Xu, & Virginia Yip (2001). “On Chaozhou ‘Intransitive Passives’: Passive and Unaccusative in Jieyang”. Ms., University of Hong Kong. Newmeyer, Frederick J. (1998). Language Form and Language Function. MIT Press. Newmeyer, Frederick J. (2001). “Deconstructing Grammaticalization”. Language Sciences, 23, 187–230. Pike, Kenneth L. (1947). “Grammatical Prerequisites to Phonemic Analysis”. Word, 3, 155– 172. Wright, Martha (1983). A Metrical Approach to TS in Chinese Dialects. Ph.D. dissertation, University of MA: Amherst. Yap, Foong-Ha & Shoichi Iwasaki (1998). “A Typological and Semantic Analysis of the Relationship between Causatives and Passives in the Periphrastic ‘Give’ Constructions of some East and Southeast Asian Languages”. Paper presented at the 8th Annual Meeting of the Southeast Asian Linguistics Society, Kuala Lumpur, Malaysia, 20–22 July 1998. Yip, Moira (2002). Tone. Cambridge: Cambridge University Press. Yip, Virginia & Stephen J. Matthews (2001). Intermediate Cantonese. A Grammar and Workbook. London & New York: Routledge.
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.1 (363)
Grammaticalization of word order Evidence from Lithuanian* Sergey Say ILI RAN, St. Petersburg
.
Introductory remarks
The role of word order phenomena in the study of grammaticalisation is somewhat unclear.1 On the one hand, word order processes do not seem to fit into the narrow understanding of grammaticalisation as a development of once lexical units into grammatical ones. On the other hand, ever since Meillet’s path-breaking article on grammaticalisation (1912), it has been acknowledged that “the grammatical fixing of word order (. . . ) is a phenomenon “of the same order” as the grammaticalization of individual words” (Hopper & Traugott 1993: 23 and further reference to Meillet). Moreover, regardless of theoretical persuasions of individual researchers, ‘grammaticalisation of word order’ is a collocation that is repeatedly used in diachronic syntax referring to the process of emergence of a rigid ‘grammatical’ word order instead of loose ‘pragmatic’ one. And yet, apparent reversibility of word order changes as well as differences in the very character of developing linguistic forms led to viewing the fixing of word order as at best a rather peripheral case of grammaticalisation (see Hopper & Traugott 1993: 50–56 for an in-depth discussion).2 In what follows I present a case study of a diachronic word order process in Lithuanian and try to build it into the theory of grammaticalisation, which is primarily based on the study of the more usual lexicon-to-grammar scenario of grammaticalisation. Lithuanian is a so-called ‘free word order language’ (Ambrazas 1982: 101; Ambrazas 1985; DLKG 1996); according to the now commonly accepted views, this means that word order, at least at the clausal level, reflects here not only the structure of grammatical relations (subject, object etc.), but also depends on a number of communicative and pragmatic factors, such as for instance the theme vs. rheme distinction (Girdenien˙e 1971; Valeika 1974; Ambrazas 1985: 671–682),
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.2 (364)
Sergey Say
emphasis, relative communicative weight and the like (DLKG 1996: 645).3 As expected, there are, however, some structures in Lithuanian grammar, in which word order is fixed, or grammaticalised, that is, pragmatic factors are ruled out.4 One of these is the genitive construction, in which the GEN-N pattern is clearly unmarked in contemporary Lithuanian: (1) t˙ev-o švarkas father-gen jacket “father’s jacket” (2) auks-o žiedas gold-gen ring “golden ring” (3) kelion-˙es tikslas journey-gen purpose “the purpose of the journey” For these and further examples see (Ambrazas 1985: 686, 1986: 96–97; Ambrazas (Ed.) 1997: 702; DLKG 1996: 655–656). Postposition of the Genitive is often claimed to be an emphatic or poetical device (example quoted from Ambrazas (Ed.) 1997: 702, see also Ambrazas 1985: 687; Schmalstieg 1988: 318): (4) Kaip puik-¯us sl˙eniai sraun-ios Dubys-os! How magnificent-pl vales swift-gen Dubysa-gen “How magnificent are the vales of the swift Dubysa!” What is more relevant for the issues raised here is that the genitive is consistently postposed if the head noun conveys quantity (5), and inconsistently if the genitive has a partitive function (6): (5) litras pien-o liter milk-gen “(one) litre of milk” (6) dal-yje tarm-iu˛ or: tarm-iu˛ dal-yje part-loc dialects-gen dialects-gen part-loc “in (one) part of the dialects” In one subtype of the genitive construction the relative order of elements has distinctive value, namely, where the head noun functions as a ‘container’: (7) a.
stiklin˙e arbat-os glass tea-gen “glass of tea”
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.3 (365)
Grammaticalization of word order
b. arbat-os stiklin˙e tea-gen glass “tea-glass” (8) a.
lentyna knyg-u˛ shelf books-gen “shelf of books” b. knyg-u˛ lentyna books-gen shelf “book-shelf ”
If preposed, the genitive in such constructions does not refer to any particular physical entity in the world and merely denotes a general characteristic intrinsic to the entity referred to by head of the NP. Thus, arbat-os stiklin˙e may refer to a glass that is actually empty; it simply refers to a type of glass that is used for tea. This kind of preposed genitival modifier can scarcely have dependents of their own; typically they describe a characteristic of the head noun that holds a high position on the time-stability scale (that is, arbat-os stiklin˙e “tea glass” can not become alaus stiklin˙e “beer glass”).5 In other words, they show some signs of the loss of nominal properties, typical of non-referential genitives (Koptjevskaja-Tamm 2002: 158). Postposed genitives in the constructions headed by a noun with the meaning of container are quite different. Actually, the role of stiklin˙e “glass” in stiklin˙e arbatos “a glass of tea” is very similar to that of a quantifier whose prototypical function cross-linguistically is to make mass nouns referential. The question whether or not arbatos is itself referential here can hardly be unambiguously answered; however, there are some clear indications that these genitives hold a position nearer to the referential pole of the referentiality continuum (see Givón 1984: 423ff.), than those discussed in the preceding paragraph. For instance, these genitives may have dependents of their own and in some of these cases the whole dependent NP is clearly referential: (9) puodukas šit-os arbat-os cup this-fem.gen tea-gen “a cup of this tea” (10) puodukas mano mam-os išplikyt-os arbat-os cup my mother-gen made-fem.gen tea-gen “a cup of the tea prepared by my mother” Of course, in many cases the dependent genitive in the constructions at issue does not have dependents of its own and refers to substance rather than to a subset of it, and, thus, is not semantically referential. Even in this cases, the genitive has a
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.4 (366)
Sergey Say
property of what can be called potential referentiality, which can manifest itself in further discourse, cf.: (11) Aš išg˙eriau puoduk-a˛ arbat-osi ir jii man pasirod˙e labai I drank.1sg cup-acc tea-gen and she me.dat seemed very skani. tasty “I have drunk a cup of tea and it seemed very tasty to me.” Here ji “it (feminine)” is coreferential with arbatos “tea” and refers to the portion of tea that has been drunk by the speaker; thus, referentiality of the noun denoting substance is created by its use in the genitive construction at issue.6 In other words, arbatos “tea-GEN” appears to be referential on the discourse level, although it may be not on the propositional level. It is curious that in other constructions, unlike those discussed above, the genitive is found in preposition regardless of its referentiality or its non-referentiality, which gives rise to semantic ambiguity: (12) vyr-o darbas man-gen job a. “the job of the man (referential)” b. “a job that is appropriate for a man, a job of a real man (nonreferential)” The (relative) fixedness of word order in the genitive construction is a comparatively innovative feature of Lithuanian. In the following analysis of its emergence I intend to implement a scholarly precept by Olga Yokoyama proposed with respect to the study of word order in Russian: “(. . . ) if there exists a kind of continuum between grammar and pragmatics (. . . ), then there may be considerable theoretical interest in re-examining the interaction of word order and grammatical relations (. . . ) in terms of grammaticalisation of anthropological attitudinal factors” (Yokoyama 1986: 171). In other words, the ordering patterns of a particular language could be found to be motivated by the speaker’s construal of the nonlinguistic reality, while the processes of word order change may be possibly viewed as conventionalisation of these pragmatically motivated patterns.
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.5 (367)
Grammaticalization of word order
. Genitive construction in Old Lithuanian . General facts and previous research In the course of the rather short (beginning in the 16th century) written history of the Lithuanian language,7 its syntax, and word order in particular, have undergone few significant changes (Zinkeviˇcius 1981: 208). However, in Old Lithuanian genitives are found in postposition as often (and in some texts even more often) as in preposition. The position of adnominal genitives in the Old Lithuanian texts has been thoroughly studied by Vasiliauskien˙e (1994, 1997, 1998). She argues that the position of the genitive depends on a number of factors, and particularly, on whether any component of the NP is emphasised. In the latter case the genitive may be found in the marked position, that is, after the head noun (Vasiliauskiene˙ 1997: 108; cf. similar observations for the agreeing modifiers in Girdenien˙e 1981). No semantic differences between the two ordering patterns are reported in the literature (to my knowledge). There has been much speculation as to what the original word order pattern was in this domain; since no conclusive evidence has been put forward, the question will not be discussed in any detail below and the picture characteristic of the Old Lithuanian writings (16–17th centuries) will be taken as a starting point for further analysis.
. Anchoring vs. non-anchoring relations in the genitive construction In order to make further discussion profitable, an important distinction between anchoring and non-anchoring genitives must be drawn. This distinction is most clearly put by Maria Koptjevskaja-Tamm: [I]n many cases GAs (genitive adnominals – S. S.) function as anchors (. . . ) or reference point entities (. . . ) for identification of head’s referents. In other words, in many instances we can identify the referent of a nominal via its relation to the referent of the GA. Thus, knowing who Peter is we can identify Peter’s bag, arm, brother; knowing what the table refers to we can identify the edge of the table etc. (Koptjevskaja-Tamm 2000: 126) Obviously, an important precondition for GAs to function as reference point entities, or anchors, is that they themselves involve referential noun phrases. Particular good anchors are also those whose reference is firmly established in the preceding discourse and/or situation (. . . ). (ibid.: 128)
Non-anchoring genitives are different in that (. . . ) their primary function consists in qualifying or classifying the head nominals, rather than identifying their referents. (. . . ) Non-anchoring genitives
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.6 (368)
Sergey Say
provide a powerful tool for qualifying entities by focusing on various aspects – material they are made up of, age, size, purpose, temporal and locational characteristics and so on. (ibid.: 141)
Contemporary Lithuanian is among those languages that use basically the same means to express anchoring and non-anchoring relations. Thus for instance in (12) the genitive vyro may refer to a definite ‘man’ and function as anchor for the darbas “job”; however, it may also be used as a mere classifier of the head noun darbas, indicating that the job at issue refers to a particular kind of job. In the majority of languages, however, these two types of relations are coded differently. Most often the two patterns are kept distinct with the help of the determiner system, as in the following examples from Italian (as well as in their English translations): (13) a.
la casa di un professore art.def house of art.indef teacher “the house of a teacher” b. la casa di pietro art.def house of stone “the house of stone” (Koptjevskaja-Tamm 2002: 156)
This type of opposition is of course not attested in those languages which lack a determiner system, as is the case in Lithuanian. There are, however, other strategies that make it possible to distinguish between the two types of constructions and which do not involve the use of articles. Russian denominal adjectives as opposed to true genitives (14) or German compounds as opposed to true genitives (15) are typical examples of such strategies: (14) a.
krest’janskaja loshad’ peasant.adj horse “(a) horse for work in the field” b. loshad’ krest’janin-a horse peasant-gen “(the) peasant’s horse, horse of the peasant”
(15) a.
Kinder-krankheit children-illness “(a) disease typical of children, (a) childhood disease” b. Krankheit der Kinder illness def.gen children.gen “(the) illness of the (particular) children”
A preliminary typology of the means used to make the anchoring/non-anchoring distinction is proposed in (Koptjevskaja-Tamm 2002).8
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.7 (369)
Grammaticalization of word order
. Referentiality and word order in the Old Lithuanian genitive construction The main hypothesis proposed here with respect to the ordering patterns in the Old Lithuanian genitive construction is that genitives used referentially and having possession-like meaning tended to be postposed, while preposition was favoured by those genitives that did not have a particular referent and denoted an abstract quality of the object. In other words, in the Old Lithuanian texts (as well as in some dialects, where word order preserves many archaic features) relative order of GEN and N statistically reflected the anchoring vs. non-anchoring distinction. This may be exemplified by a typical NP from the Mažvydas’ “Katekizmas”: (16) žadis da˛ng-aus karalist-as word heaven-gen kingdom-gen’ “the word of the kingdom of heaven” Here the relation between žadis and postposed karalistas is possession-like: karalistas indicates the source of the words. In contrast, preposed da˛ngaus may be interpreted as an epithet of the kingdom. This general hypothesis accounts well for the fact that the distribution of preand postposition is not the same for various types of genitival meanings. Thus, for instance, the meaning of material or abstract quality is usually expressed by the genitive in preposition (examples (17)–(24) are taken from Baranauskas’ dialectal texts quoted from Vasiliauskien˙e 1998: 223–224): (17) sidobr-o blizguczai silver-gen spangles “silver spangles” (18) ape sanow-as gadyni about old.times-gen days “about (the) days of yore” Highly referential genitives in the constructions with the meaning of inalienable possession (19)–(21), as well as objective (22) and subjective (23)–(24) genitives are usually found in postposition: (19) isz akiu Aniut-es from eyes Aniute-gen “from Aniute’s eyes” (20) unt bałsa nabasznik-a Wyskup-a on voice dead.person-gen bishop-gen “in the voice of the late bishop”
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.8 (370)
Sergey Say
(21) su runku numirel-a with hand dead.man-gen “with the hand of the dead (man)” (22) pasamdimas kut-ias renting sty-gen “(the) renting of the sty” (23) isz pasakojimo paukszt-es from saying bird-gen “from the words of the bird” (24) unt kłausima karalos on question king-gen “to the question of the king” Appropriate statistical analysis of the data is complicated by the fact that in many cases exact semantic interpretation is difficult to perform. This is, for instance, the case with the word diewo god-GEN, which is one the most frequent attributes in the Old Lithuanian writings. It is difficult to determine whether it must be interpreted as “god’s, of god” or “divine, godly” in each particular case. Position of this modifier shows a good deal of variation. However, Vasiliauskien˙e notices that the only NP in Sirvydas’ “Punktai sakymu˛” and Bretk¯unas’ “Postil˙es” in which diewo favours preposition is diewo ´zodis “god’s (or possibly divine) word” (Vasiliauskien˙e 1997: 107). This observation fits well into the hypothesis proposed above: preference for preposition is not observed in cases where the genitive has possessive-like meaning and cannot be understood as qualifier, such as Diewo akim / akim Diewo “to the god’s eye’, Diewo baymes / baymes Diewo “of the fear of god”. In some cases word order in the Gen-N structure seems to reflect the referential structure of the text, cf. the following passage from “Punktai sakymu˛”: (25) Vnt gało pakrutina ir inartina marias At end stirs and attracts seas-acc menuo, kuris kad auga ir wunduo MARIU moon, which when grows and water seas-gen kiełasi ir kayp butu auga, kad ma´zin yet ir gayszta, ir raises and as.if were growing, when smaller goes and wanes, and mariu wunduo nupuoła ´ziemiaus seas-gen water goes.down lower “Finally the seas are stirred and attracted by the moon: when it grows the water of the seas (lit.: “water – seas-gen”) is also raising and, as it were, grows; when it (= the moon) becomes smaller and wanes, the sea water
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.9 (371)
Grammaticalization of word order
(lit.: “seas-gen water”) goes down, too.” (quoted from Vasiliauskien˙e 1997: 108–109) In the first case mariu “seas-GEN” (these seas has already been mentioned at the beginning of the sentence!) has the function of anchor for the water and thus is found in postposition. In the second case, there is no need to identify the nominal once again and the genitive is preposed. The hypothesis at issue is also in agreement with Vasiliauskien˙e’s (1997: 111) observation that the genitive is particularly often postposed at the beginning of a new paragraph or thought.
. The problem of foreign influence It is sometimes claimed that postposition of the genitive attribute was basically alien to Old Lithuanian (as it is alien to contemporary Lithuanian) and that it was due to the influence of other languages (Zinkeviˇcius 1981: 216, 1998: 181). Indeed, the proportion of pre- and postposed genitives in Old Lithuanian texts differs drastically from one text to another even when they date from the same period. This fact is difficult to account for by the factors intrinsic to the Lithuanian language. However, Vasiliauskien˙e (1997: 121) argues that “(. . . ) influence of other languages cannot be the only possible explanation for the so frequent postposition of Genn in Old Lithuanian. For such alternation of placement of Genn to be possible at all, the language must be, in a sense, prepared for it”. This way of thinking relies on a well-known claim proposed by Jakobson: “[l]a langue n’accepte des éléments de structure étrangers que quand ils correspondent à ses tendences de développement” (Jakobson 1938: 54; cf. also Weinreich 1953: 25). The two opposite views on the role of foreign influence quoted above could in fact appear not to be as contradictory as they seem at the first glance. These two approaches are reconciled if one recognizes the idea of multiple causation, according to which the presence of system-internal factors which facilitate a particular pattern of change does not automatically preclude the possibility of foreign influence as a triggering force of that change, and vice versa (cf. for instance Thomason & Kaufman 1988: 57ff.). The hypothesis on the distribution of Old Lithuanian genitival positions proposed above allows us to reinterpret the commonly accepted view on the role of contact phenomena in the development of Lithuanian ordering patterns. Indeed, it is well-known that the range of genitival uses in Lithuanian covers a very wide spectrum, which corresponds to both genitives and relational adjectives in a number of neighbouring languages, for instance in Polish, which was the basic source for the Lithuanian literary tradition. To put it somewhat crudely, those Lithuanian genitives that are used non-referentially and have the meaning of some abstract quality correspond to the Polish relational adjectives, rather than genitives. One may further note that, while Polish genitives are almost exclusively
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.10 (372)
Sergey Say
found in postposition, the adjectives may appear on either side of the noun. In general, preposition of adjectives is slightly preferred, although not for relational ones. In other words, there is both semantic and structural contrast between genitival constructions and constructions with relational adjectives respectively, and this contrast is partially manifested in the ordering patterns. In this respect Polish influence could have indeed initiated (or at least strengthened) the tendency to express referential/non-referential distinction by the position of the genitive in Old Lithuanian. This distinction could be to some extent an epiphenomenon of the tendency to preserve original (post-nominal) position of genitives in the translated texts. What is more important, however, is that the Polish system of nominal modifiers could have triggered or strengthened the differentiation of the two types of genitives in Lithuanian. In other words, the source and target systems have been assimilated, even though the morphosyntactic patterns of the two languages were different.
. The Lithuanian genitive construction from a diachronic point of view . Emergence of rigid word order as a process of grammaticalisation There is no doubt that the ordering of elements within the Old Lithuanian genitive construction was a tricky matter, which reflected communicative, stylistic (as proposed in the relevant literature) and/or referential (as suggested here) factors. However, this pragmatically/referentially oriented system developed into a system that fixes word order regardless of any pragmatic or communicative nuances in the meaning of particular genitive phrases. A few contemporary cases of word order ‘doubling’9 may be interpreted as fossilised remnants of the older system: in contemporary genitive constructions whose head has the meaning of container, position of the genitive reflects its function in a rather regular fashion; as was typical of Old Lithuanian, attributive genitives are in preposition, while those genitives that are used in pseudo-partitive constructions and thus hold a higher position on the referentiality hierarchy are in postposition. Although not in line with the most popular trend to view grammaticalisation as exclusively the process of “the increase of the range of a morpheme advancing from a lexical to a grammatical or from a less grammatical to a more grammatical status” (Kuryłowicz 1965: 52; cf. also Heine et al. 1991: 2, and many others), this kind of development does present a case of grammaticalisation if the latter is “treated in a wider perspective of the life cycles of grammatical constructions” (Dahl 2001) and is understood as “the gradual drift in all parts of the grammar toward tighter structures, toward less freedom in the use of linguistic expressions at all levels” (Haspelmath 1998: 318; cf. also Haspelmath 1999).
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.11 (373)
Grammaticalization of word order
If one assumes that it is not necessarily lexical units, but rather grammatical constructions that grammaticalise, the development of the Lithuanian genitive construction has shown such constitutional features of grammaticalisation as loss in semantic complexity, pragmatic significance and syntactic freedom (cf. Heine & Reh 1984 for an elaboration of these parameters). If we treat word order as a kind of linguistic sign, we should acknowledge that this sign has acquired a new, more abstract signifié, its new function being primarily marking the very structure of grammatical relations, instead of indicating (though somewhat indirectly) the referential status of particular NPs. This kind of ascension from more concrete to more abstract meaning is typical of any grammaticalisation. In the case of container construction word order has not acquired a new, more abstract signifié, but syntactic freedom has also decreased in that a statistically preferred encoding strategy which could serve as a merely inferential cue for the listener grew to be an obligatory rule which has clearly semantically distinctive value. One specific aspect of the process at issue, which puts it somewhat apart from the more common grammaticalisation pathways, is that it did not lead to the fixing of constraints on interpretation; rather, correct understanding of the majority of contemporary Lithuanian genitive constructions requires more inferential effort on the part of the hearer in determining the implicatures that are necessary for successful communication. In particular, more implicit inferences are needed to choose between the referential and non-referential reading of contemporary genitive constructions – a development that goes against the general pattern found in grammaticalisation processes, which, as argued by LaPolla, usually create new built-in constraints on interpretation, thus decreasing the inferential aspect of understanding (LaPolla 1997).10 Thus, instead of grammaticalising the anchoring/non-anchoring distinction Lithuanian collapsed the two subtypes of the genitive construction altogether; unlike many other patterns of grammaticalisation, the development discussed above was not based on the conventionalisation of a once pragmatically motivated contrast.
. Direction of extension of ordering patterns The very fact that word order in the Lithuanian genitive phrase was grammaticalised is not surprising: it is well known that word order in the genitive phrase shows a low degree of intra-language variation (Rijkhoff 1998). The mechanism that has led to the fixation of word order was the extension of the pattern typical of non-anchoring genitives onto anchoring genitives.11 The question, however, arises why did grammaticalisation of word order happen the way it did? In other words, why did Lithuanian grammaticalise preposition of the genitives, and not postposi-
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.12 (374)
Sergey Say
tion? Although questions of this kind can never be answered conclusively, there are some speculations that may be germane. First, it may be noticed that while expressing possessive relations is a primary function of adnominal genitives in various languages, it is the richness of the various non-anchoring genitival uses that results in the notable frequency of genitives in Lithuanian and Baltic generally (Pritzwald 1936; Fraenkel 1950: 102; Zinkeviˇcius 1981: 210 passim; Berg-Olsen 1999). Thus the proportion of non-referential genitives in Lithuanian is much higher than in languages with more typical genitival functions. This fact may have resulted in the shift of the quantitative, if not functional, core of the genitive construction from the domain of possession into the domain of more abstract qualification of the head noun; the latter was typically expressed by the preposed genitive in Old Lithuanian (as argued above). Second, the functions of relational adjectives and genitives overlap in Lithuanian even more than in other languages (see also below). Translational practice in the Old Lithuanian period (see 2.4) corroborates the view that Lithuanian genitives were easily identified with the adjectives of other languages. Thus, the preposition of agreeing modifiers could have played an analogical role in the levelling of the position of other adnominal dependents with respect to the noun; that is, the versatility of the Lithuanian genitives could have led to their expansion into the adjectival domain triggering subsequent alignment of word order according to the model provided by attributive phrases. Thus, the uniform pre-head position of nominal modifiers in Lithuanian was achieved by harmony-by-extension development in terms of Harris and Campbell (1995: 212). In any case the Lithuanian scenario of development could be relevant for the on-going debate on the nature of the tendency towards harmonic (that is, either consistently head-initial or consistently head-final) word order patterns in the languages of the world (see for instance Hawkins 1983). On the one hand, it may be true that the reasons for this tendency may be historical, as is argued by many scholars (Aristar 1991; Bybee 1988). On the other hand, it is not likely that the well-attested correlations between ordering patterns in the constituents of various types is only due to the preservation of the older word order in the process of reanalysis of a particular construction (harmony-through-reanalysis), as is argued for instance by DeLancey (1993), who rules out any possibility of synchronic forces leading to the harmony of word order. In fact, the process of analogical elimination of word order doubling, such as the one attested in Lithuanian, may reflect such a synchronic force, which is in this case simply a tendency towards uniformity based on the comparison of structurally different but functionally similar constructions.
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.13 (375)
Grammaticalization of word order
. On functional convergence between genitives and adjectives The word ‘genitive’ primarily refers to a particular member of the Indo-European case system attested in older IE languages and preserved in some contemporary IE branches, Baltic among them. However, in the typological literature ‘genitive’ (Givón 1990; Croft 1995: 99) or more often ‘possessive’ (Nichols 1992) constructions are identified on syntactic (or semantic) grounds as typical constructions used in a certain language to link two nominals thus modifying one of them. Preferably these constructions must have a possessive core. That is why phrases like English ‘(the) daughter of my friend’, German Frauenbild “portrait of a woman” or Russian Petina knizhka “Petya’s book” are often counted as genitive constructions in cross-linguistic studies, regardless of the fact that from a morphological point of view their ‘possessors’ are expressed by prepositional phrases, adjectives or even parts of compound nouns. This is a typical example of a somewhat opportunistic solution to a wellknown problem of non-universality of syntactic constructions. Given that the morphosyntactic systems of individual languages are sometimes wildly deviant, there is no other way to identify constructions cross-linguistically than by basing them primarily on their functions. Thus, for instance English of -phrases are often (for instance, in typological studies on word order) compared to other languages’ true genitives precisely because they have similar functions. Lithuanian is a kind of opposite extreme in this respect: morphological genitives are somewhat abundant and have many functions that are typical of adjectives, rather than genitives in other languages. Thus, there are some logical reasons to identify at least some of Lithuanian morphological genitives with adjectives in cross-linguistic studies. It seems that such an approach is not only logically plausible, but would also capture some facts intrinsic to the grammar of Lithuanian, such as the ability of the Lithuanian genitive to be coordinated with adjectives: (26) rank-u˛ ir protinis darbas hands-gen and intellectual work “hand (physical) and intellectual work” (27) vidur-io ir lyviškoji tarm˙es middle-gen and Livonian dialects “central (-Latvian) and Livonian dialects” An oblique argument for the functional convergence between adjectival and genitival modifiers in Lithuanian comes from Latvian, where the system of modifiers is generally similar to the Lithuanian one. However, there is a large group of the socalled ˙genit¯ıvini, i.e., indeclinable adjectives that have the morphological form of ã genitives of non-existing (or extinct) nouns: bezb¯ernu ˙gimene “childless (Gen.Plur.)
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.14 (376)
Sergey Say
family”, pretgripas l¯ıdzeklis “anti-flu (Gen.Sg.) medicin”, tr¯ısistabu dz¯ıvoklis “threeroom (Gen.Plur.) flat”. The rise of ˙genit¯ıvini is nothing less than the process of lexã source of adjectives in the languages icalisation of genitives, which is a well attested of the world. The possibility for two phrases to be co-ordinated is often viewed as a manifestation of the fact that they belong to the same syntactic category. Thus, judging from the combinatory abilities of Lithuanian morphological adjectives and genitives, three major classes must be distinguished: (1) anchoring genitives, (2) non-anchoring genitives and relational adjectives, (3) qualitative adjectives.
. The Lithuanian NP as a counterexample to typological word order tendencies There is a strong universal tendency to place genitives in the periphery of the NP. Lithuanian is the only language in Rijkhoff ’s (1998) European languages’ sample in which the Adj-Gen-Noun pattern is nevertheless unmarked, although Gen-AdjNoun pattern is sometimes possible, too (actually Latvian, which is very near to Lithuanian in this respect, was not included in the sample, see also below). However, an explanation for this unexpected finding may be offered in the light of the above suggestions. Attributes are often classified on the basis of the degree to which their meanings are integrated into the meaning of the head noun; it is assumed that word order in the NP must iconically reflect these layers of intimacy (Rijkhoff 1992). Relational adjectives, especially terminological ones typically hold a very high position on this hierarchy (cf. the tendency to place relational adjectives nearer to the noun than qualitative ones, which is attested in various languages). Contrariwise, possessors and other meanings typically expressed by the genitives are usually in the most peripheral layer of adnominal dependents, which is reflected in the cross-linguistic word order pattern discussed by Rijkhoff. In the light of the interpretation proposed in Section 4 Lithuanian genitives are split: some of them (non-referential ones) must be adjacent to the head noun in the cognitive structure of the NP, and others (referential ones, ‘true’ possessive genitives) must hold a peripheral position. It is curious that Latvian, which is otherwise very close to Lithuanian, conforms to the essence of Rijkhoff ’s principles in that the split between referential and non-referential genitives (specifier and descriptive genitives in Christen’s terms) is reflected in its word order patterns. Thus Latvian speakers prefer the structure in (28a) to the one in (28b) in case of specifier genitives and under some further conditions (Christen 2001: 505):
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.15 (377)
Grammaticalization of word order
(28) a.
maz-¯as meiten-es jaunais little-gen girl-gen new b. jaunais maz-¯as meiten-es new little-gen girl-gen “the little girl’s new chair”
kr¯esls chair kr¯esls chair
Thus the Adj-Gen-Noun pattern is persistent for the descriptive non-referential genitives only (for example, a non-referential meaning “the new little girl’s chair” is likelier to be conveyed by the word order pattern in (28b)). Contrariwise, in Lithuanian the two types of genitives are conflated and do not differ in their word order behaviour. Both types of genitives tend to behave as Latvian descriptive genitives, cf. preference for (29a) over (29b) (ibid.: 505): (29) a.
nauja maž-os mergait-˙es new little-gen girl-gen b. maž-os mergait-˙es nauja little-gen girl-gen new “the little girl’s new chair”
k˙ed˙e chair k˙ed˙e chair
This observation corroborates the above hypothesis that in the history of Lithuanian the ordering discrepancies between the two types of genitives have been analogically levelled and that the pattern typical of descriptive (non-referential) genitives extended onto other genitives.12
. Concluding remarks Linguistic notions such as grammaticalisation are supposed to capture some regularities in the real functioning of language rather than to be artificial concepts imposed upon it. In this respect one must agree with Östen Dahl in that the traditional understanding of grammaticalisation à-la Meillet may seem too narrow. The emergence of fixed word order, for instance, would only be subsumable under grammaticalization when the position of morphemes which are on their way to becoming grammaticalized is concerned. Yet, we would want to see such processes as a unitary phenomenon. A more general definition of grammaticalization would generalize it to all processes by which grammatical phenomena develop. (Dahl 2000: 8)
Grammaticalisation of word order as the shift from so-called free word order to fixed word order in a particular syntactic domain is basically a development of a new syntactic means or, rather, an up-grading of such a means on a well-known grammaticalisation scale proposed by Givón (1979: 83):
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.16 (378)
Sergey Say
discourse > syntax > morphology > morphophonemics > zero It is a typical grammaticalisation process in that it produces more regular linguistic signs with more abstract semantics out of less regular and more concrete signs. Supposedly, extension is the basic mechanism of this type of development. If a distribution between the two competing ordering patterns is reshaped in such a way that the complicated pragmatics- or discourse-oriented rule yields to a more transparent syntactic rule, it means that one of the patterns has been extended over time. As was shown above, the very possibility as well as the direction of the extension is conditioned by the identification of syntactic constructions and/or categories. The major deviation from the more usual patterns of grammaticalisation is that grammaticalisation of word order is clearly a reversible process. Syntactically conditioned (rigid) and pragmatically conditioned (loose) word order patterns interchange diachronically in the history of various languages. Moreover, the two tendencies may run simultaneously in different structures of one language. Thus, in contemporary Lithuanian there is a tendency to place pronominal (especially topical) objects before the verb, while nominal objects are usually found in postverbal position. If the development of this split is innovative, and there are some grounds to think so, then it clearly undermines the more or less consistent SVO pattern of Lithuanian, thus making it less grammaticalised. Do these facts mean that we should abolish altogether the unidirectionality principle with respect to the grammaticalisation of word order? Yes and no. If the patterns of word order change are taken in isolation, there is no ground for viewing these processes as unidirectional. However, word order change is often involved in longer grammaticalisation chains, or channels, such as for instance a pronoun > clitic > affix channel, cf. (Lehmann 1995). Word order adjustment of a certain kind is a necessary, though not sufficient, condition for many of these developments. And it is only a particular type of word order change that can provide grist for the grammaticalisation mill. Thus, for instance, Lithuanian would have to grammaticalise the emerging ordering contrast between nominal and pronominal objects of the verb in order to develop pronominal clitics, as has happened in Romance and some Slavic languages. Pragmaticisation of word order never leads to further processes of grammaticalisation, while syntactisation of word order may do so; moreover, emergence of contrastive ordering patterns with respect to a once single syntactic construction is a good starting point for the splitting of a syntactic category, which opens up the way for further grammaticalisation chains. It is in that sense that grammaticalisation of word order is unidirectional: once a novel syntactic unit goes one step further in the grammaticalisation channel, the word order re-patterning becomes irreversible.
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.17 (379)
Grammaticalization of word order
Notes * The research reported here was partially supported by the president of the Russian Federation’s programme for the support of major academic schools (grant No. NSh-2325.2003.6). ILI RAN stands for Institut lingvisticheskix issledovanij Rossijskoj akademii nauk – Institute of linguistic research of the Russian Academy of Sciences. . This study owes its existence to a number of persons. Vanda Kazanskien˙e was my main informant, and a co-operative linguist at the same time. Many thanks go to Emma Geniušien˙e, an anonymous referee and especially Olga Fischer for their valuable comments on earlier versions of the paper. Finally, I would like to express my deep gratitude to Maria Koptjevskaja-Tamm, for both general inspiration that I found in her research and very specific help in recommending some relevant references and providing access to the then unpublished material. None of these kind people bears any responsibility for the possible mistakes and shortcomings in the present article. . The situation is further complicated by the fact that ‘word order change’ usually refers to the shift in basic word order patterns (such as, for instance, SOV to SVO or Adj-Noun to Noun-Adj development). This kind of language change, too, has some features in common with the prototypical cases of grammaticalisation, but it is of course usually viewed as a distinct process (for the comparison of the two types of phenomena see McMahon 1994: 139– 170). Evidently, fixing of word order is much nearer to the core of grammaticalisation than the shift of basic word order. . As argued by Bhat (1991) it means that semantic and pragmatic relations within the sentence, especially between the major constituents, are encoded by basically the same means and are not kept apart as is typical of languages with fixed word order. . It has been repeatedly pointed out in the literature that languages with highly flexible clause structures may have inflexible NPs; among the European languages this combination of properties is found (besides Lithuanian, see below) in e.g. Udmurt and Armenian (Bakker 1998: 392). The opposite combination is possible as well; however, the relative order of V and O is cross-linguistically much more liable to flexibility than the order of the elements within the NP (for some figures see Bakker 1998: 388). . It has been noticed in the literature that the position of a modifier relative to its head noun may depend on whether or not this modifier refers to an inbuilt characteristic of the entity or rather to a characteristic that is asserted by the use of the NP. A lengthy discussion of such phenomena in Old English and some other languages can be found in Fischer (2001). It is convincingly shown in this article that pre-existing, thematic properties of the noun tend to be iconically expressed by pre-head modifiers (adjectives in those cases), while possibly temporary and rhematic properties are expressed by postposed modifiers. . The picture is further blurred by the fact that it is not quite clear which noun is a semantic head in puodukas arbatos ‘a cup of tea’. Along with the contexts similar to (11) where one speaks about tea rather than about the glass (cf. the use of the verb ‘to drink’) this NP can be also found in examples like:
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.18 (380)
Sergey Say
(11) a.
Ant stalo stov˙ejo puodukas arbatos. On table stood cup tea-gen “There was a cup of tea on the table.”
Here the whole NP refers to the glass itself rather then to a particular portion of tea. Finally, somewhat embarrassingly, some contexts are registered in which the whole NP has a kind of split reference: (11) b.
Aš išg˙eriau puoduka˛i arbatos ir išmeˇciau ji˛i pro langa˛. I drank.1sg cup-acc tea-gen and threw.1sg it-acc through window “I have drunk a cup of tea and threw it out through the window.”
Here puoduka˛ arbatos seems to refer to a portion of tea in the first clause, although in the second clause, ji˛ that is syntactically coreferential to it, clearly refers to the glass itself. . The first written Lithuanian texts appeared in the 16th century. As in many European countries literacy was brought by Christianity. Lithuanian literary tradition developed more or less simultaneously in two countries, the Grand Duchy of Lithuania and the Duchy of Prussia. The study of Old Lithuanian syntax is complicated by the fact that Old Lithuanian writings were for the most part compilations of almost word-by-word (eventually even morpheme-by-morpheme) translations of religious texts from Polish or Latin. One of the most important sources is the first Lithuanian book, Martynas Mažvidas’ catechism (“Katekizmas”) published in 1547 in Königsberg, Prussia. The other literary sources that are directly referred to in this article include a book of sermons (“Postil˙e”) published by Jonas Bretk¯unas in 1591 and Konstantinas Sirvydas’ “Gospel Points” (“Punktai Sakymu˛”), a collection of original sermons published in 1629. Due to historical circumstances, the language of various Old Lithuanian writings shows a good deal of dialectal variation and orthographic inconsistency, although some efforts have been made to normalize the language during the 16–17th centuries and the first grammars and dictionaries were published in this period. For a detailed description of the early written history of the Lithuanian language, see (Zinkeviˇcius 1998: 227–255; Zinkeviˇcius 1988). . It may be noticed that in many languages the two types of constructions used for the expression of anchoring and non-anchoring relations respectively differ with respect to word order patterns found in them (see for instance examples from German and Russian above). However, word order alone does not signal the distinction in any of the European languages from Koptjevskaja-Tamm’s sample (2002). . To use Rijkhoff ’s term, who applies it to those cases, when “some or all modifiers of a certain category (. . . ) may appear on either side of the noun” (Rijkhoff 1998: 343). . The argument of LaPolla is based on the widely attested mechanism of grammaticalisation, namely, conventionalisation of a once expressive and more explicit linguistic coding. Thus, for instance, grammaticalisation of you guys into a 2Pl. pronoun as opposed to you as a 2Sg. pronoun in some varieties of English is indeed a case of such a development. The use of you guys is more explicit with respect to the number of addressees and thus constrains the hearer’s interpretation of the utterance more heavily than Standard English you does. . Anette Rosenbach (this volume) argues that the scenario of development of the English NP-structures was due to the analogy between referential/non-referential types of adnom-
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.19 (381)
Grammaticalization of word order
inal dependents, cf.: “There is both structural variation (absence of presence of POSS ’s) as well as semantic variation ([± referential] possessor), which leads to a mismatch between form and meaning. (. . . ) I assume that such form/meaning mismatches have given rise to some kind of analogy, with hearers having two possible interpretations ([± referential] possessor/first noun) for one surface construction (. . . ). Likewise, speakers have two constructions (with or without POSS ’s) at their disposal to convey the same meaning (either a referential or a non-referential possessor/first noun). For the scenario I am suggesting it is (. . . ) crucial (. . . ) that (. . . ) the distinction between the constructions becomes blurred, which makes it possible for the properties of one construction to ‘jump’ to another one. Given such an analogy, it is plausible that the constructions have mutually affected each other in the rise of phrasal first elements in the s-genitive and N+N constructions” (Rosenbach, this volume). . It may be added, that the unusual Gen-Adj-Noun pattern in Lithuanian can be iconically motivated, but this motivation reflects the properties of the adjective, rather than of the genitive, cf.: “The adjective attribute goes directly before the modified word if it makes a closer or terminological unit with the modified word” (Schmalstieg 1988: 319) and an example from the same source: (11) a.
motin-os vestuvin˙e suknel˙e mother-gen wedding.adj dress “mother’s wedding dress”
References Ambrazas, Vytautas (1982). “Žodžiu˛ tvarka ir baltu˛ kalbu˛ sakinio tipo rekonstrukcija” [“Word order and reconstruction of Baltic sentence type”]. Baltistica, 18, 100–118. Ambrazas, Vytautas (1985). Grammatika litovskogo jazyka [The grammar of the Lithuanian language]. Vilnius: Mokslas. Ambrazas, Vytautas (1986). “Dabartin˙es lietuviu˛ kalbos žodžiu˛ tvarkos modeliai” [“Word order patterns in the contemporary Lithuanian language”]. Lietuvos TSR mokslu˛ akademijos darbai, A ser.: 3, 92–102. Ambrazas, Vytautas (Ed.). (1997). Lithuanian Grammar. Vilnius: Baltos Lankos. Aristar, Anthony R. (1991). “On Diachronic Sources and Synchronic Pattern: An Investigation into the Origin of Linguistic Universals”. Language, 67, 1–33. Bakker, Dik (1998). “Flexibility and Consistency in Word Order Patterns in the Languages of Europe”. In Siewierska (Ed.), 383–420. Berg-Olsen, Sturla (1999). A Syntactic Change in Progress: The Decline in the Use of the Non-Prepositional Genitive in Latvian, with a Comparative View on Lithuanian. MA paper. University of Oslo. Bhat, D. N. S. (1991). Grammatical Relations. The Evidence against their Necessity and Universality. London & New York: Routledge. Bybee, Joan L. (1988). “The Diachronic Dimension in Explanation”. In John A. Hawkins (Ed.), Explaining Language Universals (pp. 350–379). Oxford: Basil Blackwell.
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.20 (382)
Sergey Say
Christen, Simon (2001). “Genitive Positions in Baltic and Finnic languages”. In Östen Dahl & Maria Koptjevskaja-Tamm (Eds.), Circum-Baltic Languages: Typology and Contact, Vol. 2 (pp. 499–520). Amsterdam & Philadelphia: John Benjamins. Croft, William (1995). “Modern Syntactic Typology”. In Masayoshi Shibatani & Theodora Bynon (Eds.), Approaches to Language Typology (pp. 85–143). Oxford: Oxford University Press. Dahl, Östen (2000). “The Tense-Aspect Systems of European Languages in a Typological Perspective”. In Östen Dahl (Ed.), Tense and Aspect in the Languages of Europe (= Eurotyp, 20–26) (pp. 3–25). Berlin & New York: Mouton de Gruyter. Dahl, Östen (2001). “Grammaticalization and the Life Cycles of Constructions”. RASK – Internationalt tidsskrift for sprog og kommunikation, 14, 91–134. DeLancey, Scott (1993). “Grammaticalization and Linguistic Theory.” In Jule Gomez de García & David Rood (Eds.), Proceedings of the 1993 Mid-America Linguistics Conference (pp. 1–22). Boulder: Department of Linguistics, University of Colorado. DLKG (1996). Dabartin˙e lietuviu˛ kalbos gramatika [The grammar of the contemporary Lithuanian language]. Vilnius: Mokslas. Fischer, Olga (2001). “The position of the adjective in (Old) English from an iconic perspective”. In Olga Fischer & Max Nänny (Eds.), The Motivated Sign. Iconicity in Language and Literature 2 (pp. 249–276). Amsterdam & Philadelphia: John Benjamins. Fraenkel, Ernst (1950). Die baltischen Sprachen. Heidelberg. Girdenien˙e, D. (1971). “Aktualioji skaida ir jos vaidmuo bei reiškimo b¯udai lietuviu˛ kalboje” [“Functional sentence perspective, its role and means of expression in Lithuanian”]. Kalbotyra, XXII(1), 35–47. Girdenien˙e, D. (1981). “Derinamojo pažyminio vieta tarmin˙eje ir bendrin˙eje lietuviu˛ kalboje” [“The placement of the agreeing modifier in Lithuanian dialects and Standard Lithuanian”]. Baltistica, 17, 156–169. Givón, Talmy (1979). “From Discourse to Syntax: Grammar as a Processing Strategy”. In Talmy Givón (Ed.), Discourse and Syntax (Syntax and Semantics, 12) (pp. 81–112). New York, San Francisco & London: Academic Press. Givón, Talmy (1984/1990). Syntax. A Functional-Typological Introduction. Vols. 1–2. Amsterdam & Philadelphia: John Benjamins. Greenberg, Joseph H. (1963). “Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements”. In Joseph H. Greenberg (Ed.), Universals of Language (pp. 73–113). Cambridge, MA: Massachussets Institute of Technology. Harris, Alice C. & Lyle Campbell (1995). Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Haspelmath, Martin (1998). “Does Grammaticalisation Need Reanalysis?” Studies in Language, 22, 315–352. Haspelmath, Martin (1999). “The Change of Behavioral and Coding Properties in Constructional Grammaticalization”. Paper presented at the conference New reflections on grammaticalization, Potsdam, June 1999. Hawkins, John A. (1983). Word Order Universals. New York: Academic Press. Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer (1991). Grammaticalization. A Conceptual Framework. Chicago & London: The University of Chicago Press.
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.21 (383)
Grammaticalization of word order
Heine, Bernd & Mechthild Reh (1984). Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske Verlag. Hopper, Paul J. & Elizabeth C. Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Jakobson, Roman (1938). “Sur la théorie des affinités phonologiques des langues”. Actes du quatrième Congrès International de Linguistes, 48–59. Copenhagen: Einar Munksgaard. Koptjevskaja-Tamm, Maria (2000). “Romani Genitives in Cross-Linguistic Perspective”. In Viktor Elšík & Yaron Matras (Eds.), Grammatical Relations in Romani. The Noun Phrase. (Current Issues in Linguistic Theory 211) (pp. 123–149). Amsterdam & Philadelphia: John Benjamins. Koptjevskaja-Tamm, Maria (2002). “Adnominal Possession in the European Languages: Form and Function”. Sprachtyplogie und Universalienforschung, 55, 141–172. Kuryłowicz, Jerzy (1965). “The Evolution of Grammatical Categories”. Diogenes, 51, 55–71. Reprint: Kuryłowicz, Jerzy (1975). Esquisses linguistiques II (= International Library of General Linguistics, 37) (pp. 38–54). München: W. Fink. Lehmann, Winfred P. (1974). Proto-Indo-European Syntax. Austin & London: University of Texas Press. LaPolla, Randy (1997). “Grammaticalization as the Fossilization of Constraints on Interpretation. Towards a Single Theory of Communication, Cognition and the Development of Language”. Paper presented at City University of Hong Kong Seminar in Linguistics, November 6, 1997. Lehmann, Christian (1995). Thoughts on Grammaticalisation. München: Lincom Europa. Li, Charles N. & Sandra A.Thompson (1975). “The Semantic Function of Word Order: A Case-Study in Mandarin”. In Charles N. Li (Ed.), Word Order and Word Order Change (pp. 163–196). Austin: University of Texas Press. McMahon, April (1994). Understanding Language Change. Cambridge: Cambridge University Press. Meillet, A. (1912). “L’évolution de formes grammaticales”. Scienta (Rivista di Scienza), 12(26), 6. Nichols, Johanna (1992). Language Diversity in Space and Time. Chicago: The University of Chicago Press. Pritzwald, St. von (1936). Das Attribut im Altlitauischen. Heidelberg. Rijkhoff, Jan (1992). The Noun Phrase: The Typological Study of its Form and Structure. Dissertation, University of Amsterdam. Rijkhoff, Jan (1998). “Order in the Noun Phrase in the Languages of Europe”. In Siewierska (Ed.), 321–382. Schmalstieg, William R. (1988). A Lithuanian Historical Syntax. Columbus, OH: Slavica Publishers. Siewierska, Anna (Ed.). (1998). Constituent Order in the Languages of Europe. Berlin & New York: Mouton de Gruyter. Thomason, Sarah Grey & Terrence Kaufman (1988). Language Contact, Creolization and Genetic Linguistics. Berkeley: University of California. Valeika, L. (1974). Word Order in Lithuanian and English in Functional Sentence Perspective. Vilnius.
TSL[v.20020404] Prn:25/02/2004; 14:39
F: TSL5918.tex / p.22 (384)
Sergey Say
Vasiliauskien˙e, V. (1994). “On the Placement of the Genitive Attribute in Old Lithuanian”. Dept. of General Linguistics, University of Umeå. Report 35. 99–115. Vasiliauskien˙e, V. (1997). “Daiktavardžio ir ˛ivardžio genityvo vieta senojoje lietuviu˛ kalboje” [“The placement of the nominal and pronominal genitive in Old Lithuanian”]. Lietuviu˛ kalbotyros klausimai, 38, 104–121. Vasiliauskien˙e, V. (1998). “Atributiniu˛ junginiu˛ modeliai Mikalojaus Daukšos “Postil˙eje” ir Antano Baranausko tarmiu˛ tekstuose” [“Patterns of the attribute phrase in Daukša’s “Postil˙e” and dialectal texts compiled by Baranauskas”]. Baltistica, 33, 215–232. Venneman, Theo (1975). “An Explanation of Drift”. In Charles N. Li (Ed.), Word Order and Word Order Change (pp. 269–305). Austin: University of Texas Press. Weinreich, Uriel (1953). Languages in Contact: Findings and Problems. (Publications of the Linguistic Circle of New York, 1.) New York. Yokoyama, Olga T. (1986). Discourse and Word Order. Amsterdam: John Benjamins. Zinkeviˇcius, Zigmas (1981). Lietuviu˛ kalbos istorin˙e grammatika [A historical grammar of the Lithuanian language]. Vol. II. Vilnius: Mokslas. Zinkeviˇcius, Zigmas (1988). Lietuviu˛ kalbos istorija [The history of the Lithuanian language]. Vol. III. Senu˛ju˛ raštu˛ kalba. Vilnius: Mokslas. Zinkeviˇcius, Zigmas (1998). The History of the Lithuanian Language. Vilnius: Mokslo ir enciklopediju˛ leidybos institutas.
TSL[v.20020404] Prn:10/05/2004; 11:30
F: TSL59LI.tex / p.1 (385)
Language index
Cantonese dou 349, 352, 353, 354–357, 359 ge 148, 150, 151 gwo 348, 352, 353, 354, 356–359 Catalan anar 212, 215–218, 220–222, 224 Chinese (Mandarin) ba 127, 325–327 bei 11, 325–333, 335–341 de 6, 13, 137–139, 148–151, 160–163, 165 dei, dé 5, 122–128, 130 guo 346 s¯an tóng 30 shi 296 sì huà 30 zai 346 Dutch gaan 219 hij, zij 30 moeten 132 mogen 132 tig 38 English a(n) 10, 84 anti 31 burger 3, 31 but 3, 30, 117, 120 dare 4, 5, 34, 97–112 down 117 dwarf, dwarves 63, 67
‘em 56, 57 ex 31 ever 237–241, 244 -gate 118 going to, gonna 106, 245 have to 14, 106, 108, 112, 117 his 75, 76, 89, 90, 92 if 3, 30, 117, 120, 121 ism 30, 118 just 235–237, 240, 241, 244 like 302, 321 Mc 56, 59, 66 must 30, 106–108, 112, 123, 124 need 4, 97–112 -ness 67 pro 31 off 3, 31 suppose, supposing 6, 7, 169–190 teens 3, 31 that 281 there 237, 244 to 29, 57, 60, 61, 67, 88, 131, 189 ultimate 56, 59, 66 up 3, 31, 117 while 269 yet 240, 241 Finnish ei 253–256, 261 eikä 256–258 essa 9, 264, 269–277 hän 249, 251, 252, 259, 260 kun, kuin 9, 264–269, 274–276 päällys 31 pitää 249, 250
TSL[v.20020404] Prn:10/05/2004; 11:30
F: TSL59LI.tex / p.2 (386)
Language index
French aller 215–218, 220 tutoyer 29 venir de 225 German anfangen 214, 215 angenommen daß 187 mögen 38, 121, 132 wegen 28 Hindi ne 22, 37 Hokkien ho (lang) 349–351, 353, 354, 356–357, 359, 361 khi 349–350, 353–359 Irish muid
3, 29, 118
Italian ammesso che 187 anta 31 supponendo (che) 169, 174, 176, 178, 179, 186 Japanese ga 29 no 5, 6, 137–139, 141–148, 158–165 Kambera la 313–316 -nya 305, 308, 321 pa- 313, 315, 316 paki 306, 307, 319 wà(ng) 319–321 wàngu 10, 11, 300–322 Latin ad 24 habere 281
habet ibi 61 incipere 214 nec 29, 30 negare 29, 30 quilibet 52, 65 quivis 52, 65 re 29 -sc- 8, 216, 217, 219 subpono 187 Malay, Indonesian -nya 151, 165 punya, empunya 6, 137–139, 151–163, 165 yang 151, 165 Modern Greek as 200, 201, 204, 205 boro, bori 202, 203, 205, 207 ksana 29 na 7, 198–200, 204–206 prepi, prepo 202, 203, 205–207 tha 7, 51, 53, 196–198, 205, 206 tos 51 Spanish deber (de) 117, 131 dentrar 31 hay 61 -mente 54, 55, 66 sobrar 31 Swedish än 268 dummer 32 då 274, 275 han 252 medan 269, 274 må 38 när 268, 274 som 268
TSL[v.20020404] Prn:7/05/2004; 15:27
F: TSL59NI.tex / p.1 (387)
Name index
A Agricola, Michael 270, 272, 273, 277 Ahlqvist, August 260 Ahrens, Hans 38 Allen, Andrew S. 33 Allen, Cynthia L. 4, 75, 76, 78, 79, 89, 90, 131, 321 Altenberg, Bengt 80 Ambrazas, Vytautas 363, 364 Amelang, James S. 222 Andersen, Henning 4, 52, 86 Ansaldo, Umberto 11, 12, 327, 345, 346, 348, 358 Anttila, Raimo 65 Aoki, Haruo 141, 144 Aristar, Anthony B. 374 Aronoff, Mark 45 Auwera, Johan van der viii, 4, 17, 24, 30, 32, 36, 38, 105, 108, 112, 115, 117, 118, 121, 122, 125, 140, 184, 187, 203, 220 Avellan, G. A. 256
Bertinetto, Pier Marco 230 Besten, Hans den 90 Beths, Frank 4, 34, 98, 115 Bhat, D. N. S. 379 Biber, Douglas 91, 106, 111 Bisang, Walter 220, 358 Bloomfield, Leonard 37 Bodman, Nicholas C. 347, 349 Bopp, Franz 49 Börjans, Kersti 29 Bourdin, Philippe 131 Bowden, John 45, 64 Bresnan, Joan 242, 303 Britain, Dave 229, 234, 235, 245 Brown, Keith 112, 229, 231, 244 Brugmann, Karl 38 Brunner, Karl 89 Burridge, Kate 39 Butt, Miriam 22, 37 Bybee, Joan L. 7, 19, 21, 29, 33, 65, 111, 118, 121, 181, 193–206, 211, 225, 264, 281, 374
B Bakker, Dik 379 Bao, Zhiming 124, 131, 347, 361 Barentsen, Adrie viii Battye, Adrian 35 Bauer, Laurie 74, 91 Beames, John 37 Becker, Reinhold von 31, 255, 267 Bentley, Delia 131 Benveniste, Émile 30, 118 Berg-Olsen, Sturla 374 Berschin, Helmut 222
C Campbell, Lyle 1, 8, 17, 23, 25–27, 29, 31, 33, 36, 37, 64–67, 76, 88, 130, 260, 263, 272, 320, 345, 374 Carrasquel, Jose Ramon 294 Carstairs, Andrew 74 Chafe, Wallace 181, 245 Chang, Lili 124, 131 Chao, Yuen Ren 123, 124 Chappell, Hilary 131 Chen, Matthew 361 Chomsky, Noam 38
TSL[v.20020404] Prn:7/05/2004; 15:27
F: TSL59NI.tex / p.2 (388)
Name index
Christen, Simon 376 Christidis, Anastasios-Phivos 206 Clancy, Patricia 142 Claudi, Ulrike 45, 115–117, 126 Colon, Germà 214, 216–218, 220, 221 Comrie, Bernard 37, 235, 251, 330 Condillac, Étienne Bonnot de 45 Cook, Haruko Minegishi 141, 144 Cornips, Leonie 219, 225 Cowie, Claire 88 Craig, Colette 294 Croft, William 24, 121, 375 Crowley, Terry 31 D Dahl, Östen 23, 221, 230, 231, 372, 377 Dasher, Richard B. 6, 104, 105, 108, 123, 132, 170, 187, 188 Dawson, Hope 64 DeLancey, Scott 374 Demske, Ulrike 78, 85, 91 Denison, David 175, 189, 235, 245 Detges, Ulrich 7, 8, 211 Diessel, Holger 8, 286, 287, 297 Diewald, Gabriëlle 1, 7, 45, 122 Duanmu, San 359 DuBois, John W. 92 Dürr, Michael 51 E Ebert, Karen H. 230, 274 Ebert, Robert Peter 230, 274 Echols, John M. 165 Elsness, Johan 229, 235–238, 244 Embleton, Sheila 277 Emonds, Joseph E. 189 Enfield, Nick J. 123, 240, 327, 348 Erbaugh, M. S. 325 Erdal, Marcel viii Eskola, Sari 274 Eurén, G. E. 268 Evans, B. 139, 152
Evans, Nicholas 139, 152 Eythórsson, Thórhallur 76 F Felixberger, Josef 222 Ferguson, Charles A. 18, 20 Filppula, Markku 229, 233 Fischer, Olga C. M. 1, 8, 12, 14, 29, 60, 64, 79, 80, 84, 88, 89, 92, 117, 131, 175, 187–189, 205, 219, 379 Fitzmaurice, Susan 29 Fleischman, Suzanne 281 Fortescue, Michael 59, 182 Fraenkel, Ernst 374 Frajzyngier, Zygmunt 17 Fraurud, Kari 84, 85, 92 Fujimura, Taiji 142 G Gabelentz, Georg von der 120 Galt, John 232, 238, 242 Gelderen, Elly van 320 Giacalone Ramat, Anna 1, 45, 46, 64, 119, 263 Geurts, Bart 17 Gil, David 157–159 Gildea, Spike 286, 287, 295 Girdenien˙e, D. 363, 367 Givón, Talmy 365, 375, 377 Goebl, Hans 222 Goossens, Louis 123 Gougenheim, Georges 213–218, 220 Greenberg, Joseph 19, 66, 128, 282, 361 Grice, Herbert P. 224 H Hagège, Claude 30, 31, 118, 251, 260 Häkkinen, Kaisa 267 Hakulinen, Auli 255, 259, 260 Harris, Alice C. 23, 27, 31, 33, 36, 37, 272, 320, 374 Harris, John 229, 233 Harris, Martin 222, 281, 294
TSL[v.20020404] Prn:7/05/2004; 15:27
F: TSL59NI.tex / p.3 (389)
Name index
Haspelmath, Martin 1–4, 17, 18, 21, 24, 28, 37–39, 73, 84, 89, 92, 104, 115, 118–120, 140, 162–164, 187, 205, 220, 239, 263, 264, 303, 320, 321, 372 Hawkins, John A. 374 Heine, Bernd viii, 1, 7, 9, 14, 17, 26, 27, 36, 45–48, 50, 51, 55, 58, 64–66, 90, 108, 109, 115–117, 122, 126–128, 131, 138–140, 196, 198, 203, 205, 211, 263, 264, 281, 302, 303, 320, 372, 373 Helasvuo, Marja-Liisa 260 Hengeveld, Kees viii, 283 Herlin, Ilona 9, 248, 263, 266, 269, 274 Himmelmann, Nikolaus 39, 294 Hock, Hans Henrich 53, 67 Hoenigswald, Henry M. 25 Holton, David 197 Hopper, Paul J. 1, 2, 4, 10, 26, 31–33, 36, 45, 46, 53, 58, 59, 65–67, 73, 75, 97, 98, 106, 115–118, 120, 122, 131, 139, 195, 244, 281, 320, 363 Horie, Kaoru 5, 137, 141, 143, 144, 146, 161–163 Horne Tooke, John 45 Horrocks, Geoffrey C. 198–200 Huang, Chu Ren 149 Huddleston, Rodney 60, 67, 83, 90, 91 Hundt, Marianne 109 Hünnemeyer, F. 45, 115–117, 126 Hyman, Larry M. 251 I Iakovou, Maria 202, 207 Ingman, A. W. 260 Itkonen-Kaila, Marja 273 Iwasaki, Schoichi 145, 327, 349 J Jackendoff, Ray 303 Jahnsson, A. W. 250, 255, 256, 275
Jakobson, Roman 371 Janda, Richard D. 1, 23–26, 28, 29, 36, 37, 53, 59, 62–64, 67, 73, 75, 76, 89–92, 115, 119, 140, 205, 263, 345 Jespersen, Otto 60, 75, 80, 89, 90, 196, 197, 206, 260 Jiang, Lihong 149 Johnston, James C. 215 Joseph, Brian D. viii, 1, 3, 4, 36, 45, 51, 53, 57, 61, 63, 65–67, 77, 195–199, 201, 202, 205, 206, 263, 321
K Karantzola, Eleni 200 Katz, Aya 46 Kaufman, Terrence 371 Kemenade, Ans van viii, 21 Kennedy, A. L. 240 Kiefer, Ferenc 171, 182 Kim, Hyeree 17 Kitagawa, Chisato 148 Klamer, Marian viii, 10, 11, 299, 300, 302, 304, 307, 313, 321 Klausenburger, Jurgen 115 Klemola, Juhani 81 Koch, Peter 213 König, Ekkehard 9, 175, 181, 189, 264 Konishi, Jin’ichi 143 Koptjevskaja-Tamm, Maria 92, 365, 367, 368, 379, 380 Kortmann, Bernd 176, 181, 187, 189, 265, 269 Koskinen, Yrjö 250 Kotilainen, Lari 9, 248, 263, 274 Koziol, Herbert 80 Kroch, Anthony 34 Krug, Manfred 97, 99, 104, 106, 109, 111 Kuhn, Adalbert 49 Kuryłowicz, Jerzy 45
TSL[v.20020404] Prn:7/05/2004; 15:27
F: TSL59NI.tex / p.4 (390)
Name index
Kuteva, Tania A. 1, 6, 7, 9, 14, 27, 45, 46, 48, 50, 51, 58, 64, 196, 198, 211, 212, 263, 264 L Labov, William 53 Lai, Violet 131, 132 Laitinen, Lea 9, 247, 249, 251, 252, 260, 265, 268, 277 Lakoff, George 127 LaPolla, Randy 373, 380 Lass, Roger viii, 2, 25, 36, 37, 77, 87, 89, 90 Lazzeroni, Romano 31 Le, Van Cu 143, 175 Leacock, Stephen 67 Lehmann, Christian 1, 2, 21, 22, 26, 28, 38, 45, 65, 76, 140, 218 Lehmann, Winfred P. 378 Lehrer, Adrienne 118 Lessau, Donald A. 45, 51 Levin 303 Levinson, Stephen 116 Li, Charles N. 132, 149, 150, 164, 286, 287, 294 Lichtenberk, Frantisek 110 Lightfoot, David 35, 39, 320 Lim, Lisa 11, 12, 345 Lindén, Eeva 272 Lindstedt, Jouko viii Löbner, Sebastian 84, 92 Lommatzsch, E. 175 Lönnbohm, O. A. F. 252 Lönnrot, Elias 257, 267 Lord, Carol 67, 172, 184, 320 Los, Bettelou viii Luraghi, Silvia 32 Lü, Shuxiang 123–125 Lyons, John 6, 146, 161, 163, 169, 202, 231 M Macaulay, Ronald K. S. 236, 240, 242–244
Maddieson, Ian 20 Makkonen, Anna 257 Maling, Joan 302, 321 Matisoff, James M. 66, 358 Matsumoto, Y. 29 Matthews, Stephen J. 5, 131, 137, 151, 327, 328, 333, 349, 361 McMahon, April 379 Meillet, Antoine 25, 45, 196, 363, 377 Méndez Dosuna, J. 29 Mervis, Caroline 127 Meyer Lübke, Wilhelm 220 Meyer, Paul 221 Millar, Robert McColl 79 Miller, Jim 8, 9, 229, 231, 239, 243 Mitchell, Bruce 79, 84, 244 Moens, Marc 240, 241 Mowrey, Richard 21 Murtoff, Jennifer 54 Mustanoja, Tauno 89 Mutt, O. 81
N Nedjalkov, Igor V. 269 Nedjalkov, Vladimir P. 230, 269, 270, 277 Nevalainen, Terttu 90 Nevis, J. 29 Newman, J. 180, 336 Newmeyer, Frederick J. 1, 2, 17, 23, 25–27, 29–32, 34, 35, 37, 39, 47, 64, 65, 73, 88, 115, 118, 120, 140, 205, 345 Ng, Eve 10, 281, 295 Nichols, Johanna 375 Nikiforidou, Kiki 201 Noël, Dirk 190 Norde, Muriel 1, 4, 8, 14, 23, 28–32, 36, 37, 73, 77–79, 85–92, 115, 117, 120, 140, 187, 188 Nordlinger, Rachel 126
TSL[v.20020404] Prn:7/05/2004; 15:27
F: TSL59NI.tex / p.5 (391)
Name index
O Osthoff, Hermann 38 P Pagliuca, William 21, 45, 111, 118, 121, 181, 281 Palmer, Frank R. 193–195, 202, 204, 206 Pansegrau, Petra 39 Pappas, Panayiotis A. 65, 196, 206 Payne, John 85 Pérez Saldanya, Manuel 212, 215, 218, 220 Perkins, Revere D. 111, 118 Philippaki-Warburton, Irene 196, 198, 201, 202 Pichon, Jacques 214 Pike, Kenneth L. 347 Pinkster, Harm 281 Pintzuk, Susan 21 Plank, Frans 14, 17, 77, 87, 88, 90, 92, 120 Plungian, Vladimir A. 4, 24, 30, 38, 105, 108, 112, 121, 122, 125, 203 Pritzwald, St. von 374 Pullum, Geoffrey K. 60, 61, 67, 83, 90, 91 Q Quirk, Randolph 111, 112
83, 90, 100, 101,
R Ramat, Paolo 30–33, 36, 37, 115, 118 Rappaport, Hovav 303 Raumolin-Brunberg, Helena 8 Reh, Mechtild 26, 55, 122, 281, 320, 373 Remington, Roger W. 215 Rey, A. 175 Rhee, Seongha viii Rheinfelder, Hans 216 Rijkhoff, Jan 373, 376, 380
Rissanen, Matti 175 Roberts, Ian 35 Roma, Elisa 29 Rosén, Hannah 52 Rosch, Eleanor 127, 303 Rosenbach, Anette 4, 73, 80, 84, 85, 89, 90, 92, 117, 187, 380, 381 Ross, Claudia 148 S Sadock, Jerrold 90 Saussure, Ferdinand de 37, 47, 50 Savijärvi, Ilkka 260 Say, Sergey 12, 363 Scatton, Ernest 65 Schmalstieg, William R. 364, 381 Schultze-Berndt, Eva viii Schwenter, Scott A. 223 Seppänen, Aimo 79 Seppänen, Eeva-Leena 79 Setälä, E. N. 250–252, 255, 256, 258, 261, 268, 275 Shadily, Hassan 165 Shi, Yuzhi 124, 131, 149, 296 Silverstein, Michael 249 Simpson, Andrew 148, 164, 165 Slobin, Dan I. 325 Smith, Nicholas 105–107, 111 Sobré, Josep Miquel 221 Soga, Matsuo 142 Stein, Dieter 249 Steinkrüger, Patrick 218, 220, 222 Summers, Della 109 Sun, Chao/Fen 122, 123, 125, 132, 327 Sweetser, Eve 6, 181 T Tabor, Whitney 76, 117, 189 Taeymans, Martine 4, 5, 97 Tagliamonte, Sali 8, 229, 233 Tauli, Valter 276 Taylor, John 83 Thomas, Russel 84, 91, 173
TSL[v.20020404] Prn:7/05/2004; 15:27
F: TSL59NI.tex / p.6 (392)
Name index
Thomason, Sarah 371 Thompson, Sandra 132, 150, 286, 287, 294 Tiersma, Peter 63 Tobler, A. 175 Tomasello, M. 338 Trask, Larry 296 Traugott, Elizabeth Closs viii, 1, 2, 4, 6, 9, 17, 26, 31–34, 36, 45, 46, 53, 66, 73, 75, 76, 83, 84, 87, 92, 97, 98, 104, 105, 108, 109, 115–121, 123, 125, 132, 139, 140, 169, 170, 181, 182, 187–189, 195, 244, 264, 281, 320, 363 Tsangalidis, Anastasios 7, 193, 196, 197, 202–206 Tsimpli, Ianthi Maria 196 V Valeika, L. 363 Vasiliauskien˙e, V. 367, 369–371 Veloudis, Johan 198, 206 Vendler, Zeno 240 Verhagen, Arie 175 Veselinova, Ljuba 38 Vezzosi, Letizia 84, 89, 90 Vilkuna, Maria 260 Vincent, Nigel 21, 281, 304 Visconti, Jacqueline 6, 7, 169–171, 174, 177, 180, 187–189 Visser, Frederikus Th. 98, 175, 189 W Warner, Anthony R. 97–99 Watkins, Calvert 50 Weerman, Fred 85 Weinert, Regina 231, 239, 243 Weinreich, Uriel 371
Weinrich, Harald 218 Westney, Paul 108 Wichmann, Soren 32 Wiedenhof, Jeroen viii Wilkins, David 139, 152 Winford, Donald 8, 229, 232, 233 Winter, Werner 18 Wischer, Ilse 1, 7, 45 Wit, Petra de 85 Wittgenstein, Ludwig 127 Wong, Kwok-shing (Richard) 11, 131, 325, 327, 332, 335, 361 Wright, Martha 361 Wu, Xiu-Zhi Zoe 148, 164, 165 Wurzel, Wolfgang Ullrich 18 Wyld, Henry Cecil 75 X Xu, H. 124, 131, 327 Xu, Zhi Ming 124, 131, 327 Y Yantis, Steven 215 Yap, Foong-Ha 5, 6, 131, 137, 165, 188, 327, 349 Yip, Virginia 151, 327, 328, 333, 347, 349 Yokoyama, Olga T. 366 Z Zee, Eric 361 Zhang, Bojiang 150, 347, 361 Ziegeler, Debra P. viii, 5, 7, 45, 115, 119, 127, 140, 161, 163, 181, 182, 184, 185, 187, 190, 325, 326, 361 Zinkeviˇcius, Zigmas 367, 371, 374, 380 Zwicky, Arnold M. 76
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.1 (393)
Subject index
A ablativus absolutus see case absentativity 10, 283 accent(ed) 296, 359, 361 acceptability judgments 236 accusative see case adjectives 12, 92, 117, 368, 371, 372, 374–376, 379 adjective inflection, strong/weak 78, 84, 91 qualitative adjectives 376 denominal adjectives 368 relational adjectives 371, 372, 374, 376 adjuncts 6, 170, 176–178, 186, 189, 254, 304, 310, 321 adverbs 6, 8, 31, 173, 174, 230, 231, 234, 235, 239–242, 244, 308 past time adverbs 230, 234, 235 affix, phrasal 76 age of emergence 326, 336, 337 agent see thematic roles agreement 253–259, 261, 284, 285, 287–291, 293, 295, 297 agreeing modifiers 367, 374 loss of NP-internal agreement 4, 79, 80, 87, 90 non-agreeing negation 253–259, 261 Albanian 57 ambiguity 107, 130, 138, 158, 184, 203, 301, 311, 312, 330, 366 ambiguous contexts 300, 301 ambiguous item 313, 318 American English see English
analogy 2, 3, 25, 28, 47, 51, 61, 63, 82, 83, 119, 188, 380, 381 anchor(ing) 84, 85, 92, 330, 368, 371 Anglo-Norman 213 animacy 10, 282, 283, 285, 287–291, 293, 295, 330 antigrammaticalization 2, 3, 17, 27–31, 33, 34, 36, 37, 121, 231 aorist(ic) see tense areal convergence 219, 220 argument (structure) 245, 281–287, 303, 304, 306–310, 313, 316, 319–321, 329 external argument 304, 308, 316, 318, 319 internal argument 304, 308, 309, 316, 318, 319 types see thematic roles Armenian 57, 379 aspect 7, 8, 12, 49, 182, 213, 214, 233, 237, 240, 244, 320, 328, 329, 341, 347, 349, 360, 373 inchoative aspect 7, 213–217 lexical aspect 233, 240 perfective 12, 195, 212, 320, 349, 359 progressive 80, 127, 164, 218, 230, 237 associative 56, 142, 146, 157, 164, 165 auxiliary see verb auxiliation constraint 211, 214, 224 auxiliation paths 212 Avestan 49, 57
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.2 (394)
Subject index
B back-formation 31, 38 Bari 55 beneficiary see thematic roles biclausal constructions 300, 306, 313, 314, 316, 318, 321 bidirectionality 115–117, 131, 162 bleaching (semantic) 2, 10, 12, 13, 51, 58, 197, 304, 329, 358 blend constructions 99, 100, 102, 110 borrowing 13, 54, 57, 61, 272, 278 bridging context 80, 82, 90, 139, 143, 158 British English see English Broad Scots 233 Bulgarian 50 Buru 321 C calquing 61, 157, 165 CANCORP 11, 331, 334, 340 Cantonese 11, 12, 148, 150, 157, 325, 327–331, 333, 334, 336, 345, 347–349, 352, 354, 358–361 case case marking 249, 250, 259 ablativus absolutus 272, 273 accusative 49, 164, 260, 276, 304, 305, 320 dative 11, 24, 28, 305, 320, 325–331, 333–340 ergative 22 genitive 12, 78–82, 85, 90, 91, 165, 365, 367–369, 371–377 classifying genitives 81, 82, 91 genitivus absolutus 273 anchoring genitives 373, 376 descriptive genitives 85, 91, 376, 377 group genitives 79–81, 90 his-genitive 75, 76, 90
genitive marker/suffix 6, 29, 79, 92, 117, 141 measure genitives 91 s-less genitive 81, 82, 87, 91 specifying genitives 81, 82, 91 split genitives 90 instrumental 22 partitive 260, 277, 364 Catalan 7, 8, 211–213, 215–217, 219–222, 224 categories 28, 32, 49, 53, 110, 127, 128, 152, 162, 163, 195, 211, 242, 258, 277, 283, 301–303, 320, 346, 378 basic level categories 128 categorial status 60 conceptual category 127 discrete categories 302, 303, 320 lexical category 127, 152, 303 syntactic category 303, 376, 378 radial categories 127 categorization 90, 302, 303, 335 categorization of grammaticalization chain 5, 33, 301–303, 318, 319 causal(ity) 9, 61, 90, 203, 263–267, 269–271, 275–277, 316 causative 12, 32, 320, 329, 349, 353, 354, 356, 357, 360 causative-resultative construction 240 chains 5, 7, 127, 128, 148, 233, 302, 303, 320, 378 change, language change 3, 7, 13, 17–19, 21–24, 27, 35–37, 46–48, 50, 52, 54, 62–64, 97, 99, 104, 115, 170, 186, 265, 379 category change 5, 13, 299, 300, 303, 320 change schema 4, 86–88 change types 36, 74, 87
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.3 (395)
Subject index
conceptual change 5, 7, 87, 115–118, 198 counterdirectional (change) 5, 87 language-external factors of language change 9, 10, 247, 264, 269, 276 lexical change 219 semantic change 5–7, 13, 20, 26, 27, 38, 56, 61, 105, 116, 117, 131, 140, 163, 169, 170, 179, 186, 188, 205, 216 sound change 18, 20–22, 26, 27, 47, 51–54, 61–63, 67, 200 syntactic change 21, 117, 272 CHILDES 334 Chinese 5, 6, 13, 30, 115, 122, 123, 125–127, 130, 137–140, 148, 150, 157, 158, 161, 163–165, 286, 296, 325–327, 346, 347, 358, 361 Early Mandarin 123 Middle Chinese 123 classifier terms 127 clause matrix clause 172, 176, 177, 179, 186, 189, 313–316 relative clause 52, 141–145, 147–149, 157, 158, 161, 162, 165, 287 headed 144, 148, 161, 162 head-internal 143 headless 142, 144, 145, 147, 149, 158, 162, 165 cleft 150, 151, 158, 159, 161, 162, 165, 243 reverse cleft 243 reverse WH clefts 239, 240 cline (of grammaticalization) 2, 4, 7–9, 14, 73, 75, 77, 78, 86, 87, 118, 119, 161, 188 morphosyntactic cline 75, 87 clipping 31, 55 clitic 2–4, 6, 22, 28–31, 37, 74–77, 86, 89, 90, 117–119, 165,
207, 251, 256, 260, 266, 276, 305, 307–313, 320, 321, 378 object clitic 305, 307, 309–313, 321 subject clitic 34 subject proclitic 315, 316 codification 9, 247, 249 cognition 13, 325 cognitive (principle(s)) 215, 216, 219, 276 collocation 61, 336, 338, 363 Colonial Quiché 51 comparative 9, 12, 113, 265–269, 348, 359 Comparative Method 48–50, 65 complementizer 6, 60, 141, 143–148, 161, 162, 164, 165, 181, 188, 281 complement clausal 122, 313, 318 nominal 103, 313, 318 prepositional 313 conceptual networks 127–130, 140, 162, 211 conceptualization 211, 214, 215 conditional 6, 9, 98, 113, 132, 169, 170, 187, 188, 195, 196, 201, 206, 264, 270, 271, 276 conjunction 6, 7, 9, 66, 170, 176, 177, 180, 181, 186, 198–200, 205, 256, 264–268, 274–276, 306, 314–317, 320, 322 coordinating 314, 315 compound coordinating 256 contiguity 130, 138, 139, 145, 162, 163, 199, 213 Continental Scandinavian 3 continuum 90, 213, 233, 238, 299, 300, 306, 365, 366 control 6, 18, 46, 120, 176, 177, 179, 186, 187, 189, 330, 350 conventionalization 139, 184 of conversational implicatures 121
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.4 (396)
Subject index
converb, contextual/specialized 9, 264, 269–271, 273–277 conversion 30, 31 coordinator see conjunction copula see verb counterdirectionality 5, 87, 120, 140 cross-linguistic tendencies 3 D dative see case definiteness 84, 85, 91, 92, 305 degrammaticalization 2, 4, 14, 17, 22, 23, 27–32, 36, 39, 73–78, 85–89, 98, 99, 115–121, 130, 131, 140, 161, 163 deixis 235, 244 deictic elements 6 demonstrative(s) 6, 10, 25, 149, 150, 164, 281–283, 286–297 distal 10, 283, 294 proximal 283 remote 283 indexical function 9, 249, 252, 259 secondary deixis 235, 244 delexicalization 52 demodalization 115, 121, 122, 125, 126, 130 demonstrative(s) see deixis demorphologization 36 demotivation 32 deontic see modality derivation 32, 49, 51, 165, 321 desemanticization 117 determiner 4, 76, 78, 81, 83–87, 91, 92, 368 definite 4, 78, 83–87, 91, 92 determiner function of the s-genitive 85, 92 indefinite 10, 13, 84 possessive 75, 76, 92 developmental sequence 325, 326, 334, 339, 340 diachronic correspondence 52, 56
diachronic phonology 20 diachronic process/mechanism 56 direct lineal descent 54–56 directionality 2, 3, 17–20, 35, 57, 65, 88, 138, 140, 148, 150, 160, 161, 163, 193, 263 directionality constraints 3, 17–20, 35 discourse 6, 8, 9, 28, 32, 97, 99, 104, 110–112, 141, 152, 156, 175, 177, 178, 218, 219, 223–225, 231, 232, 239, 240, 244, 251, 252, 259, 273, 307, 309, 320, 321, 328, 366, 367, 378 discourse markers 6 discourse particles 9, 251, 259 discourse-pragmatic 7, 14, 212 distribution 57, 75, 83, 98, 100, 105, 121, 127, 233, 238, 295, 302, 305, 369, 371, 378 divergence 10, 98, 125, 244 duration 12, 126, 127, 350, 352–354, 357–360 Dutch 30, 38, 85, 90, 132, 219 Dyirbal 127 E East and Southeast Asia 348 ellipsis 55, 124, 149 emergent grammar 46, 67 emphasis 9, 49, 61, 138, 139, 200, 320, 364 English passim American English 5, 55, 56, 111, 112, 229, 236, 238 British English 4, 5, 111, 187, 233, 236, 238, 244 spoken 8, 230, 231, 233, 234, 236, 240, 244, 245 written 97, 103, 111, 230, 231, 235 Early Modern English 34, 81, 84, 87, 90, 91, 97–99, 109, 110, 112, 170–174, 176–179, 181, 184
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.5 (397)
Subject index
English English 229, 232 Hiberno-English 234 Middle English 34, 74–76, 79–81, 84, 85, 91, 92, 97, 112, 170–173, 176, 179, 188, 206, 219, 235, 238 Modern English 4, 34, 54, 56, 60, 61, 63, 66, 74, 75, 81, 84, 87, 89–91, 97, 109, 110, 112, 170–174, 176–179, 181, 184, 188, 238, 245 non-standard English 8, 229, 231, 233, 235, 244 Old English 4, 10, 32, 34, 54, 56, 57, 60, 67, 74, 75, 77, 79, 83–85, 87, 89, 91, 92, 97, 112, 123, 188, 237, 238, 245, 321, 379 present-day English 5, 97, 99, 103–105, 109–112, 184, 302 Samana English 233 Scottish English 229, 231–233, 236, 238–242 spoken 231, 241 standard English 229, 232 spoken 235, 239, 244 written 223, 230–235, 245 epiphenomenon 26, 38, 61, 372 epistemic see modality erosion see reduction Eskimo 59 Estonian 29, 260, 277 etymology 22, 32, 56–58, 60, 165 evidential see modality exaptation 87, 92 experiential/indefinite anterior see tense F F0 (fundamental frequency) 350, 352, 354–356, 361 family resemblance 127, 128, 144–146, 148, 302, 303 Finnish 9, 18, 31, 230, 247–261, 264–278
written Finnish 9, 253, 254, 265, 266, 268, 270–272, 275–277 fixation of word order see word order folk etymology 32 foregrounding 139, 217, 218–220, 222, 224 formant frequency 352 French 7, 24, 29, 30, 170, 174, 175, 178, 182, 183, 189, 213–215, 217–220, 222, 225 Middle French 182, 214, 217, 218, 220 Old French 174, 213, 215–219 frequency 5, 12, 19, 53, 89, 97, 99, 100, 105, 106, 110–112, 120, 128, 159, 172, 173, 175, 219, 222, 234, 241, 274, 321, 326, 336, 337, 340, 374 frozen constructions 109–111 function words 12, 29, 30, 38, 53 future see tense fuzzy boundaries 303 G generic lexical item 318, 321 generic semantics 306, 320 genitive see case German 28, 30–34, 38, 39, 78, 85, 89, 91, 118, 121, 122, 127, 170, 181, 187, 193, 205, 222, 265, 368, 375, 380 Old High German 33, 122 Germanic 49, 56, 67, 85, 92, 132, 170, 187, 230, 245, 328 Gothic 67, 132 grammaticalization passim see also anti-, de-, poly-, post- and regrammaticalization chain 5, 33, 301–303, 318, 319 definition 25, 26, 38, 88, 377 diagnostic 346, 360 path 8, 24, 39, 148, 159, 194 reversibility of 28
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.6 (398)
Subject index
theory 1, 3, 22–24, 45–48, 58, 61, 62, 64, 65, 198, 358 Greek 7, 19, 29, 51, 63, 193, 199, 204–206, 273 Ancient or Classical Greek 19, 49, 63, 193, 204, 206, 273 Medieval Greek 63, 206 Modern Greek 7, 29, 51, 199, 205, 206 H Hanyu Pinyin 361 Head-Driven Phrase Structure Grammar (HPSG) 46 head-marking 304 Hebrew 286, 297 Hellenistic Koine 63 Hittite 49 Hokkien 12, 148, 157, 345, 347, 349–352, 358–361 homonymy 10 homophony 10, 55, 75, 92, 203, 205, 206, 263, 301, 302, 307, 319–321 ‘hot-news’ strategies 223, 224 hypercorrection 61, 75, 76 hypoanalysis 121 I iconicity 46 ideologies, linguistic 248 imperative see mood implicational map 160, 162 inchoative see aspect indefinite article see determiner indefiniteness 84, 92 indexical function see deixis indicative see mood Indo-European 9, 32, 67, 130, 195, 259, 265, 268, 269, 272, 276, 348, 375 inference 51, 56, 57, 61, 108, 112, 117, 182–185, 201 infinitive 7, 8, 14, 29, 60, 63, 97, 98, 101, 102, 104, 106, 109, 110,
117–119, 131, 164, 197, 206, 212, 251, 253, 260, 269–274, 276, 277 inflationary processes 23 inflection(al) 2, 4, 21, 32, 37, 38, 49, 74–79, 80, 84, 85, 90–92, 102, 118, 120, 193, 194, 196, 204, 206, 249, 250, 281–283, 293–295 information structure 291 innovation 18, 52, 67, 76, 104, 234 instrument see thematic roles instrumental see case intensity 12, 350, 352, 356–359 interactional and textual functions 254, 258 interrogative 6, 25, 98, 102, 105, 106, 109, 149, 150, 173, 175, 177, 178, 186, 195, 241, 243, 251, 259, 287 intonation 138, 153, 154 IPA (International Phonetic Alphabet) 296, 361 Irish 3, 29, 56, 118 Old Irish 56 isolating tonal languages (ITLs) 11, 12, 345, 346, 357, 358, 360, 361 isomorphism 82 Italian 7, 29, 31–33, 38, 169, 170, 174, 176, 178, 181, 186–190, 368 item-based learning 338, 340 J Japanese 5, 6, 29, 127, 137–146, 148, 150, 158–161, 163–165 K Kambera 10, 299–302, 304, 306–308, 311, 313, 314, 319–322 kinesthetic image-schema 214 Kono 51 L language change see change language contact 10, 38, 264, 268, 269, 276, 278 language history 54, 58, 61
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.7 (399)
Subject index
Languedoc 213, 218, 220 Latin 18, 24, 29, 30, 32, 33, 49, 52, 54, 55, 57, 59, 61, 62, 170, 187, 193, 216, 265, 272–274, 281, 282, 294, 380 Latvian 375–377 layering 33, 105, 110 legal diction 50 levelling 63, 374 lexical entry 299 features 320 item (category neutral) 10, 304 representation 301–303, 316, 320 specifications and/or restrictions 316 split 10 lexical-conceptual features 301, 303 structure 303 Lexical Memory Traces hypothesis 119 lexicalization 117, 118, 120, 121, 128, 376 phrasal 78, 80, 87, 91 relexicalization 126, 129, 130 lexification 5 Linguistic Society of Hong Kong 361 Lithuanian 12, 363, 364, 366–381 Old Lithuanian 12, 367, 369–372, 374, 380 logophoricity 252, 260 M Macedonian 59 main verb see verb Malay 5, 6, 137–140, 151, 153, 157–161, 163–166, 351 Mandarin see Chinese matrix (‘raising’) verb see verb metanalysis 52 metaphor 24, 56 metaphoric(al) extension 51, 61, 218
metonymy 115–117, 128, 131, 218 metrical patterns 360 prominence 361 structure 360, 361 Minimalist Syntax 46 modality agent-oriented 194, 197, 203, 204, 206 deontic 107, 112, 123, 125, 129, 130, 132, 170, 181, 182, 184, 186, 187, 201, 204, 250 external necessity 4, 105, 106, 108, 112, 190 internal necessity 105, 108, 112 epistemic 6, 7, 112, 123–125, 129–131, 170, 173, 174, 178, 180–182, 184–187, 196, 197, 202–206, 250, 251 evidential 137, 141, 144, 150, 181, 182, 186, 187, 190, 217 permissive 5, 11, 12, 200, 325–340, 349, 353, 354, 356, 357, 361 speaker-oriented 200, 201, 204, 206 subordinating 195 modal (verbs) see verb monotonic(ity) 63, 67, 205 mood imperative 6, 7, 172, 173, 175, 188, 189, 195, 196, 199–202, 204–206, 277 indicative 46, 48, 65, 99, 102, 104, 154, 175, 196, 201, 206 optative 7, 57, 178, 199, 204, 206 subjunctive 57, 178, 189, 193, 195, 196, 199–202, 204, 206 subordinating 194
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.8 (400)
Subject index
morphology isolating 4, 18, 21, 28, 32, 33, 36, 45, 91, 100, 116, 345, 360, 378 morphological processes 49 morphological structure 358 morphological typology 360 movement 45, 58, 59, 62, 63, 104, 111, 117, 121, 205, 212–214, 220, 245, 353 multifunctional (or multicategorial), ambiguous element 11, 301, 320 multilateral comparison 66 N N(oun) + N(oun) constructions 4, 78, 80–83, 87, 91, 381 narrative 6–8, 212, 216–224, 234, 321, 322 necessity see modality negation 9, 101, 102, 105, 106, 113, 196, 201, 247, 253–256, 258–260, 276, 320 Neogrammarian(s) 3, 48, 53, 62 neutralization 105, 206, 332, 347 nominalizer 137, 141, 143–145, 148, 150, 151, 158, 160, 164, 165, 296 nominalizing element 6 nominals 313, 367, 375 non-assertive contexts 98, 100, 104–106, 110 non-referential see referential non-standard (language) see standard language non-phonetic conditioning 52, 53 Norse 32, 56 numeral 10, 124 O object clitic see clitic object 121, 215, 242, 282, 305, 321, 378 obviation 10, 283
Occitan 219, 220 onomatope 54 ontogenetic 119, 325 ontological 6, 89, 138, 146, 148, 160–163 optative see mood Optimality Theory 46, 92 P Panare 286, 287 panchrony 126, 127 parameter settings 35 participium conjunctum 272, 273 particle, sentence final 150, 151, 164 partitive see case Passamaquoddy 10, 281–283, 287, 294–297 passé composé see tense passive 11, 12, 100, 182, 183, 186, 187, 239, 253, 277, 325–330, 332, 334–341, 349, 350, 353, 354, 356, 357, 359, 360 past participle (passive) 233, 239, 260 past tense marker see tense pathway 7, 9, 24, 54, 55, 61, 66, 73, 85, 120, 123, 127, 128, 130–132, 137, 140, 141, 148, 150, 160, 163, 182, 348 patient see thematic roles perfect see tense perfective see aspect perfecto compuesto see tense perfet perifràstic see tense peripheral position 311, 376 periphrastic expressions 5, 106, 110, 112 permissive see modality Persian 49, 57 phonogenesis 31, 32, 58, 66 phonology 18–21, 36, 48, 49, 57, 65, 361 phonological context 347 phonological environment 350–352
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.9 (401)
Subject index
phrasal lexemes 80–82, 90, 91 phrasal unit 358 phrase marker 76, 78, 79, 83, 85–88, 90, 92 pitch 347, 350, 354, 355, 357–359, 361 pluperfects see tense poetic formula 50 Polish 12, 371, 372, 380 politeness 144 polyfunctional(ity) 57, 348, 349 polygenetic 211, 220, 224 polygrammaticalization 294 polysemy 9, 10, 39, 51, 117, 122, 127, 130, 170, 181–183, 276, 301 Portuguese 119 possession 123, 154–156, 159, 229, 369, 374 possessive 4, 73–76, 78, 83–85, 89–92, 148, 152–157, 165, 239, 250, 260, 269, 277, 361, 374–376 compound 84 construction 165, 239 marker 4, 148, 153, 156 phrasal 80, 85, 90, 91 possessive-resultative construction 229–231, 245 post-grammaticalization 117, 128, 131 post-modal 121, 125, 126, 130, 131 postposition 12, 28, 29, 51, 364, 367, 369, 371, 372 pragmatic 1, 5, 6, 9, 61, 108, 112, 117, 121, 131, 137, 140, 141, 163, 169, 186, 187, 224, 235, 259, 264, 309, 363, 364, 372, 373, 379 pragmatic markers 5, 140, 141 preposition(al) 11, 12, 28, 60, 61, 300–302, 304, 306, 310–321, 366, 367, 369, 370, 372–375 present perfect see tense preterite see tense preterite-present see verb
principle of least effort 120 (pro)nominalizers 5, 137, 138, 140, 145, 160, 161, 165 progressive see aspect prominence 345, 350, 358, 361 pronoun indefinite 34 logophoric 9, 251, 259 possessive 75, 89, 90, 92 pronominal marking 322 weak nominative pronoun 51 Proto-Celtic 56 Proto-Eskimo 59 Proto-Indo-European (PIE) 49, 50, 56, 57, 59, 62, 81, 206 prototype category 303
R reanalysis 2, 3, 10, 26, 28, 31, 47, 52, 55, 59–61, 76, 92, 117, 119, 149, 170, 176–178, 181, 186, 187, 189, 286, 299, 320, 374 recent past see tense recipient see thematic roles reconstruction 1, 88, 205, 251 internal reconstruction 2, 49–52, 65 reduction, phonetic 2, 3, 7, 11–13, 47, 51, 53, 56–58, 66, 120, 154, 197, 198, 203, 281, 286, 293, 335, 345–347, 356–360 tonal erosion/reduction 346, 347 reference referent 81, 141, 142, 157, 251, 283, 310, 318, 367, 369 referential(ity) 81, 82, 84, 85, 92, 137, 141, 143, 252, 259, 269, 284, 313, 365–367, 369–373, 376, 380, 381 referential anchor 84, 85, 92
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.10 (402)
Subject index
non-referential(ity) 81, 82, 91, 249, 259, 283, 284, 365, 366, 371–374, 376, 377, 380, 381 register 103, 346 regrammaticalization 116 regularity of sound change 47, 53, 54 reinforcement 118, 140 relative 6, 37, 52, 89, 141–145, 147–150, 157, 158, 161, 162, 165, 222, 287, 306, 315, 347, 360, 364, 366, 369, 379 relexicalization see lexicalization remotivation 32 resultative 8, 12, 229–231, 239, 240, 242–245, 349, 359, 361 retention 121, 126, 127, 330, 333 retraction 3, 33, 34, 104 rhetorical(ly) 122, 216, 217, 221, 223, 224 Romance languages 7, 8, 18, 32, 33, 54, 169, 170, 173, 174, 185, 187, 193, 212, 216, 218, 220, 230, 245, 281, 282, 294, 378 Russian 50, 121, 277, 366, 368, 375, 379, 380 S Saami 29, 251, 259 Samana see English sandhi 62, 347, 350, 361 Sanskrit 22, 37, 49, 57 Scandinavian 3, 29, 117, 118 scope increase/reduction 76, 177, 186 Scots 178, 180, 232, 233, 238, 239, 242 semantic content 292, 316, 320 depletion 52 head 379 map 24, 108, 112 roles see thematic roles vagueness 320 variation 82, 381
similaritive 9, 265–267, 269 simple past see tense Sinitic 12, 345–348, 360, 361 Slavic 50, 59, 245, 378 sociolinguistic judgments 232 sound change see change source concept 7, 14, 211, 212 Spanish 31, 32, 54, 55, 61, 66, 117, 177, 220, 222–224 Medieval Spanish 54 specifier 376 spoken (language) 106, 124, 231, 249, 257, 259 Sprachbund 225 stance marker 6, 47, 53, 73, 77, 137, 138, 141, 144–148, 150, 151, 157–165 standard language 8, 247–249, 252, 256, 258, 259 standardization 8–10, 13, 247–249, 253, 254, 259, 264, 265, 267–269, 276, 278 stress stress pattern 359–361 weakened stress 346, 347 structural variation 82, 381 subcategorization 316 subjectification 6, 7, 169, 170, 172, 176, 179, 181, 186 subject embedded 318 narrative 6 syntactic 6, 318 subjunctive see mood subordinate (clause) 199, 200, 281, 313–316 subordination 7, 9, 57, 194–196, 199–201, 204, 205, 264, 266, 277, 306, 313, 315, 322 subordinator 7, 29, 60, 67, 199 suprasegmental features 12, 345, 346 surpass marker 348 survival of the frequent 18, 19, 21 Swahili 65, 139, 164
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.11 (403)
Subject index
Swedish 9, 32, 38, 85, 90, 91, 120, 248, 250, 252, 260, 261, 265, 268, 269, 274, 275, 277 switch contexts 138, 139, 152, 153, 156 synchronic analysis 60, 67, 77 syntactic borrowing 272 context 304, 315, 319, 320 obligatorification 358 pressure 320 prominence 345, 358 Systemstörung 14, 87 T target 7, 14, 20, 128, 139, 372 taxonomic principles 302, 303, 320 temporal (meaning) 269 temporal-causal 265 temporality 266, 268, 276, 277 tense aorist(ic) 49, 50, 213, 220, 224 experiential/indefinite anterior 230 future 7, 22, 51, 53, 55, 63, 66, 121, 182, 183, 196, 197, 205, 211, 212, 245, 281 passé composé 222 past tense marker 7, 212, 213, 224 recent past 230, 232, 235–237, 242, 244 perfect 8, 9, 63, 178, 222, 223, 229–241, 244, 245, 325, 358 have perfect 233, 245 ‘hot-news’ perfect 222, 223 present perfect 222, 223, 233, 235–238, 325 pluperfect 243 perfecto compuesto 222–224 perfet perifràstic 7, 211–213, 224 present tense 218 historical present 218
preterite/simple past 105, 112, 230, 231, 233, 235–241, 244 near and remote pasts 233 textual domain 6 thematic roles agent 108, 169, 213, 214, 240, 242, 304, 308, 309, 318, 319, 326–333, 341 beneficiary 24, 304, 305, 319 goal 214, 319 instrument 304, 305, 307–310, 316, 319, 322 patient 242, 304, 305, 309, 310, 319, 326, 328–330, 332–334, 341 recipient 11, 24, 305, 327–329, 331, 333, 335 themes 2, 50, 242 token reversal 28 tokens 28, 74, 86, 110, 166, 238, 241 tonal languages see isolating tonal languages tonal registers 347 tone 98, 327, 346, 347, 350, 358, 361 context tone 347 neutral tone 346, 347 tone sandhi 347, 350 tone-bearing unit (TBU) 347, 361 topicalization 333 traditional historical linguistics 45, 47, 48, 54, 58, 59, 61, 62, 64 translation 57, 175, 178, 206, 217, 224, 250, 254, 266, 272, 274, 275, 290, 291 translation contact 272 trochaic 359 Tukang Besi 321 Turkish 219, 230 type/token distinction 4, 87 typology 193, 360, 368
U Udmurt 379
TSL[v.20020404] Prn:11/05/2004; 14:06
F: TSL59SI.tex / p.12 (404)
Subject index
unidirectionality 1–7, 13, 17, 20–23, 25–28, 32–39, 58, 59, 62, 64–67, 73, 74, 77, 78, 87, 88, 98, 104, 108, 111, 112, 115–118, 120, 127, 129–131, 137, 138, 140, 160–163, 170, 181, 186, 187, 193–195, 205, 249, 263, 378 semantic unidirectionality 7, 104, 115, 117, 130, 131, 162 uniformitarianism 25 universals of language change 3, 17–19, 27 Uralic languages 276 V variable, grammatical 233 variation, semantic/structural 82, 381 Vedic Sanskrit 49 verb auxiliary 5, 9, 22, 30, 34, 60, 99–102, 112, 117, 122, 125, 126, 130, 131, 193, 202, 204, 212, 215, 241, 245, 253, 256, 259 negative auxiliary 9, 259 compound verb 11, 307–310, 312, 318 copula (construction) 239, 281–295 embedded verb 318 instrumental verb 300, 302, 304, 306–308, 318, 319, 321 main verb 4, 5, 11, 34, 63, 97, 98, 104, 109, 112, 121, 124, 125, 199, 253, 319, 331, 334 matrix (‘raising’) verb 6, 11, 182, 300, 302, 304, 306, 316–318, 321, 322
modal verbs 4–6, 97, 100, 102, 110–112, 121, 127, 132, 178, 182, 188, 189, 194, 202–206, 249 modal auxiliary verbs 100, 101, 193, 202, 203 modal necessity verbs 9 movement verbs 212 preterite-present 4, 97 resulative verbs (RV) 348, 349, 361 semi-modal 4, 97, 99, 103–105, 109–112, 117 verb (form) 196, 200, 264 verb-island hypothesis 338 verb-second phenomena 21 verbal (compound) 300, 306, 307, 310 verbless constructions 10, 283, 288, 293–295 vernacular 250, 252 vernacular case marking 250 vowel quality 12, 260, 353, 357, 358, 360
W wh-words, English 287 why-questions 18, 24 word class 10, 293, 299–301, 303 word-formation, delocutive 3, 29–31 word order 12, 21, 38, 287, 327, 363, 364, 366, 367, 369, 370, 372–380 writing 111, 202, 231, 234, 238, 240, 245, 250, 252, 261 written language 8, 9, 220, 231, 245, 251, 252, 259, 267, 275
In the series Typological Studies in Language the following titles have been published thus far or are scheduled for publication: 1 2
HOPPER, Paul J. (ed.): Tense-Aspect. Between semantics & pragmatics. 1982. x, 350 pp. HAIMAN, John and Pamela MUNRO (eds.): Switch Reference and Universal Grammar. Proceedings of a symposium on switch reference and universal grammar, Winnipeg, May 1981. 1983. xv, 337 pp. 3 GIVÓN, T.: Topic Continuity in Discourse. A quantitative cross-language study. 1983. vi, 492 pp. 4 CHISHOLM, William, Louis T. MILIC and John A.C. GREPPIN (eds.): Interrogativity. A colloquium on the grammar, typology and pragmatics of questions in seven diverse languages, Cleveland, Ohio, October 5th 1981-May 3rd 1982. 1984. v, 302 pp. 5 RUTHERFORD, William E. (ed.): Language Universals and Second Language Acquisition. 1984. ix, 264 pp. 6 HAIMAN, John (ed.): Iconicity in Syntax. Proceedings of a symposium on iconicity in syntax, Stanford, June 24–26, 1983. 1985. vi, 402 pp. 7 CRAIG, Colette G. (ed.): Noun Classes and Categorization. Proceedings of a symposium on categorization and noun classification, Eugene, Oregon, October 1983. 1986. vii, 481 pp. 8 SLOBIN, Dan I. and Karl ZIMMER (eds.): Studies in Turkish Linguistics. 1986. vi, 294 pp. 9 BYBEE, Joan: Morphology. A Study of the Relation between Meaning and Form. 1985. xii, 235 pp. 10 RANSOM, Evelyn N.: Complementation: its Meaning and Forms. 1986. xii, 226 pp. 11 TOMLIN, Russell S.: Coherence and Grounding in Discourse. Outcome of a Symposium, Eugene, Oregon, June 1984. 1987. viii, 512 pp. 12 NEDJALKOV, Vladimir P. (ed.): Typology of Resultative Constructions. Translated from the original Russian edition (1983). Translation edited by Bernard Comrie. 1988. xx, 573 pp. 14 HINDS, John, Shoichi IWASAKI and Senko K. MAYNARD (eds.): Perspectives on Topicalization. The case of Japanese WA. 1987. xi, 307 pp. 15 AUSTIN, Peter (ed.): Complex Sentence Constructions in Australian Languages. 1988. vii, 289 pp. 16 SHIBATANI, Masayoshi (ed.): Passive and Voice. 1988. x, 706 pp. 17 HAMMOND, Michael, Edith A. MORAVCSIK and Jessica WIRTH (eds.): Studies in Syntactic Typology. 1988. xiv, 380 pp. 18 HAIMAN, John and Sandra A. THOMPSON (eds.): Clause Combining in Grammar and Discourse. 1988. xiii, 428 pp. 19:1 TRAUGOTT, Elizabeth Closs and Bernd HEINE (eds.): Approaches to Grammaticalization. Volume 1: Theoretical and methodological issues. 1991. xii, 360 pp. 19:2 TRAUGOTT, Elizabeth Closs and Bernd HEINE (eds.): Approaches to Grammaticalization. Volume 2: Types of grammatical markers. 1991. xii, 558 pp. 20 CROFT, William, Suzanne KEMMER and Keith DENNING (eds.): Studies in Typology and Diachrony. Papers presented to Joseph H. Greenberg on his 75th birthday. 1990. xxxiv, 243 pp. 21 DOWNING, Pamela A., Susan D. LIMA and Michael NOONAN (eds.): The Linguistics of Literacy. 1992. xx, 334 pp. 22 PAYNE, Doris L. (ed.): Pragmatics of Word Order Flexibility. 1992. viii, 320 pp. 23 KEMMER, Suzanne: The Middle Voice. 1993. xii, 300 pp. 24 PERKINS, Revere D.: Deixis, Grammar, and Culture. 1992. x, 245 pp. 25 SVOROU, Soteria: The Grammar of Space. 1994. xiv, 290 pp. 26 LORD, Carol: Historical Change in Serial Verb Constructions. 1993. x, 273 pp. 27 FOX, Barbara A. and Paul J. HOPPER (eds.): Voice: Form and Function. 1994. xiii, 377 pp. 28 GIVÓN, T. (ed.): Voice and Inversion. 1994. viii, 402 pp. 29 KAHREL, Peter and René van den BERG (eds.): Typological Studies in Negation. 1994. x, 385 pp. 30 DOWNING, Pamela A. and Michael NOONAN (eds.): Word Order in Discourse. 1995. x, 595 pp. 31 GERNSBACHER, Morton Ann and T. GIVÓN (eds.): Coherence in Spontaneous Text. 1995. x, 267 pp.
32 BYBEE, Joan and Suzanne FLEISCHMAN (eds.): Modality in Grammar and Discourse. 1995. viii, 575 pp. 33 FOX, Barbara A. (ed.): Studies in Anaphora. 1996. xii, 518 pp. 34 GIVÓN, T. (ed.): Conversation. Cognitive, communicative and social perspectives. 1997. viii, 302 pp. 35 GIVÓN, T. (ed.): Grammatical Relations. A functionalist perspective. 1997. viii, 350 pp. 36 NEWMAN, John (ed.): The Linguistics of Giving. 1998. xv, 373 pp. 37 GIACALONE-RAMAT, Anna and Paul J. HOPPER (eds.): The Limits of Grammaticalization. 1998. vi, 307 pp. 38 SIEWIERSKA, Anna and Jae Jung SONG (eds.): Case, Typology and Grammar. In honor of Barry J. Blake. 1998. 395 pp. 39 PAYNE, Doris L. and Immanuel BARSHI (eds.): External Possession. 1999. ix, 573 pp. 40 FRAJZYNGIER, Zygmunt and Traci S. CURL (eds.): Reflexives. Forms and functions. Volume 1. 2000. xiv, 286 pp. 41 FRAJZYNGIER, Zygmunt and Traci S. CURL (eds.): Reciprocals. Forms and functions. Volume 2. 2000. xii, 201 pp. 42 DIESSEL, Holger: Demonstratives. Form, function and grammaticalization. 1999. xii, 205 pp. 43 GILDEA, Spike (ed.): Reconstructing Grammar. Comparative Linguistics and Grammaticalization. 2000. xiv, 269 pp. 44 VOELTZ, F. K. Erhard and Christa KILIAN-HATZ (eds.): Ideophones. 2001. x, 436 pp. 45 BYBEE, Joan and Paul J. HOPPER (eds.): Frequency and the Emergence of Linguistic Structure. 2001. vii, 492 pp. 46 AIKHENVALD, Alexandra Y., R.M.W. DIXON and Masayuki ONISHI (eds.): Non-canonical Marking of Subjects and Objects. 2001. xii, 364 pp. 47 BARON, Irène, Michael HERSLUND and Finn SØRENSEN (eds.): Dimensions of Possession. 2001. vi, 337 pp. 48 SHIBATANI, Masayoshi (ed.): The Grammar of Causation and Interpersonal Manipulation. 2002. xviii, 551 pp. 49 WISCHER, Ilse and Gabriele DIEWALD (eds.): New Reflections on Grammaticalization. 2002. xiv, 437 pp. 50 FEIGENBAUM, Susanne and Dennis KURZON (eds.): Prepositions in their Syntactic, Semantic and Pragmatic Context. 2002. vi, 304 pp. 51 NEWMAN, John (ed.): The Linguistics of Sitting, Standing and Lying. 2002. xii, 409 pp. 52 GÜLDEMANN, Tom and Manfred von RONCADOR (eds.): Reported Discourse. A meeting ground for different linguistic domains. 2002. xii, 425 pp. 53 GIVÓN, T. and Bertram F. MALLE (eds.): The Evolution of Language out of Pre-language. 2002. x, 394 pp. 54 AIKHENVALD, Alexandra Y. and R.M.W. DIXON (eds.): Studies in Evidentiality. 2003. xiv, 349 pp. 55 FRAJZYNGIER, Zygmunt and Erin SHAY: Explaining Language Structure through Systems Interaction. 2003. xviii, 309 pp. 56 SHAY, Erin and Uwe SEIBERT (eds.): Motion, Direction and Location in Languages. In honor of Zygmunt Frajzyngier. 2003. xvi, 305 pp. 57 MATTISSEN, Johanna: Dependent-Head Synthesis in Nivkh. A contribution to a typology of polysynthesis. 2003. x, 350 pp. 58 HASPELMATH, Martin (ed.): Coordinating Constructions. x, 553 pp. + index. Expected Spring 2004 59 FISCHER, Olga, Muriel NORDE and Harry PERRIDON (eds.): Up and down the Cline - The Nature of Grammaticalization. 2004. viii, 383 pp. + index. 60 BHASKARARAO, Peri and Karumuri Venkata SUBBARAO (eds.): Nonnominative Subjects. Volume 1. vi, 324 pp. + index. Expected Fall 2004 61 BHASKARARAO, Peri and Karumuri Venkata SUBBARAO (eds.): Nonnominative Subjects. Volume 2. vi, 311 pp. + index. Expected Fall 2004
