The Paradox of Grammatical Change: Perspectives from Romance (Amsterdam Studies in the Theory and History of Linguistic Science, Series IV: Current Issues in Linguistic Theory, CILT 293)

The Paradox of Grammatical Change

AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F.K. KOERNER (Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin) Series IV – CURRENT ISSUES IN LINGUISTIC THEORY Advisory Editorial Board Lyle Campbell (Salt Lake City); Sheila Embleton (Toronto) Brian D. Joseph (Columbus, Ohio); John E. Joseph (Edinburgh) Manfred Krifka (Berlin); E. Wyn Roberts (Vancouver, B.C.) Joseph C. Salmons (Madison, Wis.); Hans-Jürgen Sasse (Köln)

Volume 293

Ulrich Detges and Richard Waltereit (eds.) The Paradox of Grammatical Change Perspectives from Romance

The Paradox of Grammatical Change Perspectives from Romance

Edited by

Ulrich Detges

Ludwig-Maximilians-Universität München

Richard Waltereit Newcastle University

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA

4-

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 The paradox of grammatical change : perspectives from romance / edited by Ulrich Detges, Richard Waltereit. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 293) Includes bibliographical references and index. 1. Linguistic change. 2. Grammar, Comparative and general. 3. Romance languages--Grammar, Historical. I. Detges, Ulrich. II. Waltereit, Richard. P142.P365    2008 440'.045--dc22 2007044470 ISBN 978 90 272 4808 4 (Hb; alk. paper) © 2008 – 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

CONTENTS Introduction Ulrich Detges and Richard Waltereit

1

Syntactic change from within and from without syntax: A usage-based analysis Richard Waltereit and Ulrich Detges

13

On explaining the rise of c’est-clefts in French Andreas Dufter

31

The role of the plural system in Romance Elisabeth Stark

57

Morphological developments affecting syntactic change Maria Goldbach

85

Grammaticalisation within the IP-domain Susann Fischer

107

Imperfect systems and diachronic change Giampaolo Salvi

127

From temporal to modal: Divergent fates of the Latin synthetic pluperfect in Spanish and Portuguese Martin Becker

147

Non-lexical core-arguments in Basque, Romance and German Hans-Ingo Radatz

181

Towards a comprehensive view of language change: Three recent evolutionary approaches Esme Winter-Froemel

215

Subject Index

251

ACKNOWLEDGMENTS We would like to thank all people who were involved in the preparation of this volume. The German Research Foundation (Deutsche Forschungsgemeinschaft) supported the project by sponsoring Giampaolo Salvi as invited speaker at a workshop of the Deutscher Romanistentag in September 2005 at the University of Saarbrücken, where the papers were initially presented. All our authors gracefully endured our numerous requests for major and minor changes. We would like to thank a number of anonymous scholars who reviewed the individual papers and the entire project. Our special thanks go to the series editor who supported and steered the project from the very beginning and to Anke de Looper at John Benjamins for her generous help with all editorial and technical matters. Munich and Newcastle October 2007

Ulrich Detges Richard Waltereit

INTRODUCTION

ULRICH DETGES University of Munich RICHARD WALTEREIT Newcastle University

Recent years have seen intense debates between formal (generative) and functional linguists, particularly with respect to the relation between grammar and usage (Haspelmath 1999; Newmeyer 2003, 2005; Bybee 2005, to name just a few). This debate is directly relevant to diachronic linguistics, where one and the same phenomenon of change can be, and has been, explained from various theoretical perspectives. In this context, a look at the evolution of Romance languages is particularly useful as their history is richly documented. The evolution of various languages from one common ancestor with their converging, and possibly diverging, histories provides an excellent testing ground for competing hypotheses on grammatical change. What is at stake in controversies over the nature of language change is not the reconstruction of the historical facts ‘as such’. Rather, diverging views on language change reflect controversial convictions concerning the nature of language itself, and especially that of synchronic grammatical systems. This observation can be shown to hold for both the functionalist as well as the generativist camp. The basic problem for any approach to language change is what Coseriu (1978 [1957]) has termed the paradox of change: if synchronically, languages can be viewed as perfectly running systems, then there is no reason why they should change in the first place. And yet, as everyone knows, languages are changing constantly. Within the generative tradition, there are at least four different types of explanations that have been advanced to overcome this problem. First, it has been argued that change is not inherent to language (i.e., syntax) as such, but that it occurs in language acquisition. New generations of speakers do not have direct access to their parents’ grammar, but only to the latter’s speech production, which serves them as evidence for the reconstruction of the underlying grammatical system. On this view, language change is brought

2

ULRICH DETGES & RICHARD WALTEREIT

about by generations of children who ‘misacquire’ their parents’ grammar. Imperfect transmission of this type is, still according to the position outlined here, due to changes in the evidence available to the language learners (Lightfoot 1979, 1991, 1997, 1999). This position often implies that language change is non-directional, that is, “a random ‘walk’ through the space of possible combinations or parameter resettings” (Battye & Roberts 1995: 11). In this perspective, language change is epiphenomenal — it is an accidental transition between synchronically stable grammars, and is limited only by the restrictions imposed by Universal Grammar. A radical formulation of this view is Lightfoot’s (1979, 1999) repeated remark that there can be no particular generative theory of language change – all there can be is a theory of differences between (synchronic) grammars (see also Roberts & Roussou 2003: 11). A second type of explanation targets language contact as a possible source for changes in the trigger evidence. Meisel (2001) shows that bilingual children produce more deviant word-order patterns than monolingual ones do. If those children acquire only one of the two languages to a certain degree of perfection, then they may become a source of language change when speaking the second language to their own children. However, change triggered by contact between different languages is not the normal case. As is argued by Kroch (1989, 1994, 2006), the existence of competing grammatical systems (within a society as a whole, but also within the individual) is, under normal circumstances, not a precondition, but a consequence of innovative change. Thus, unless change is viewed as arising ‘out of nowhere’, as indeed claimed by the first position sketched here, other types of explanations are needed. A seemingly attractive alternative has been proposed by proponents of a third position. Thus, according to Roberts (1993), Roberts & Roussou (2003) and van Gelderen (2004), syntactic change is motivated by structure simplification in acquisition. Structural simplicity, in turn, is measured in terms of (theoryinternally defined) markedness. However, it is unclear how this approach avoids the logical pitfall of all markedness theories: if diachronic change consists in replacing marked structures by relatively unmarked ones, then there is no reason why, in the long run, languages should exhibit marked structures at all. A fourth model of explanation is provided by the so-called ‘Inertia Theory’ (Longobardi 2001, 2003). Building on the observation sometimes made by traditional historical syntacticians that syntax is a domain especially resistant to short-term change, this position stipulates that syntax is diachronically inert. To fully understand this claim, one has to keep in mind that this theory, too, is an implicit answer to the paradox of change: if the grammar of a given language is compatible with Universal Grammar (as must be the case, since otherwise this language would not exist), then this grammar should not undergo diachronic change. Like the aforementioned three generative positions, Inertia theory is anxious to confine diachronic change to peripheral domains of language, but here, the ‘periphery’ is defined in different terms. If

INTRODUCTION

3

syntax is the core module of grammar, then a possible answer to the paradox of language change could be that diachronic change always originates in other, less central modules of grammar, e.g. in morphology, phonology, or in the lexicon. This means that syntax is always only secondarily affected by change arising from other, more peripheral modules of language, but that change never originates in syntax itself. More than the generative tradition, the functionalist camp is a patchwork of sometimes conflicting but also overlapping positions. These approaches are characterized by, first, the conviction that language structure is motivated by language use (the functionalist answer to the paradox of change), and, second, their generally anti-generative attitude. Most prominent among these positions is grammaticalization theory (Lehmann 1995 [1982]) which describes the life cycle of constructions as a history of progressive routinization that can be measured in terms of structural parameters. This current is considered as ‘functionalist’ mainly for certain anti-generative implications (Haspelmath 1998), but, crucially, grammaticalization theory as such does not focus on language usage as an explanation for language change. While it seems to be uncontroversial that it makes descriptive predictions about a certain type of diachronic change, its explanatory power is far from clear. According to some (mainly generative) critics (Roberts 1993, Newmeyer 1998, 2003, 2005), grammaticalization as described by grammaticalization theory is not in itself an explanation, but rather is a phenomenon that still goes largely unexplained. Moreover, recent years have seen generative attempts providing structural explanations for grammaticalization cycles (see Roberts & Roussou 2003 and van Gelderen 2004). Another prominent functionalist position stipulates that frequency of usage shapes the structure of grammatical systems, as already suggested by Hugo Schuchardt in 1885 (cf. Greenberg 1966, Bybee 1985, 2001, 2005). The effects of high frequency on language systems, it is argued, are of diachronic as well as of synchronic nature: in diachrony, high frequency favors various kinds of language change. Frequent tokens, for example, are stored in the human memory by rote learning and are therefore reproduced en bloc rather than being rule-generated. Therefore, in such high-frequency contexts, originally rule-generated forms or structures typically become reanalyzed as underived units. At the same time, this very effect directly manifests itself at the synchronic level: irregularity (i.e. the absence of rules) and autonomy (i.e. the status of not being derived from other items) are normally found among the most frequently used forms and structures of the system. Thus, the explanations proposed by the frequency-based approach bring together diachrony and synchrony in a maximally close fashion. A radical formulation of this is Hopper’s (1987) conception that grammar is always emergent in usage. While, as we have seen, important proponents of the generative camp pleaded for an almost complete disregard of diachrony (see above), the emergent-grammar model seems to imply that grammatical systems

4

ULRICH DETGES & RICHARD WALTEREIT

have no synchrony. A third approach, which at first sight appears only to deal with language change, i.e. with diachrony, but which bears important implications as to the synchronic level of language, is the invisible-hand model of language change (Keller 21994 [1990]). According to this position language change is the unintended by-product of speaker activities which are aimed at assuring communicative advantages to the latter. Thus, according to Keller, the prestige decline of the Old High German high-status word vrouwe “lady”, which became the neutral Frau “woman” in Modern German, was triggered by a courtesy strategy whereby speakers tended to extend the usage of vrouwe “lady” also to address women of lower social status, thereby assuring themselves obvious advantages in the communication with such addressees. Now, even though this change is usage-based — insofar as it is motivated by communicative considerations, its synchronic outcome — a rhetorically devaluated address-form — is maximally independent of the original motivation. This view reconciles language use as motivation for diachronic change with the possibility of arbitrariness in the synchronic system. The various ways of coming to terms with the paradox of change are the backdrop of the papers collected in this volume. Even though only a small number of them take an explicit stand in the generative vs. functionalist debate, all of them do refer, in one form or another, to at least one of the following questions: a) What is the locus of change? Does the change originate in the language system itself, in language acquisition or in language usage? b) What is the relationship between the various modules of grammar? Is syntactic change triggered by syntactic factors, or is it motivated by foregoing changes outside syntax? c) Is the change in question a ‘random walk’ or does it have an inherent directionality? And if the latter, how can this directionality be accounted for? In “Syntactic Change from within and from without Syntax: A usage-based analysis”, we present a critical discussion of Inertia Theory. According to the authors, the best solution to the paradox of change is the notion that change of language systems originates in language usage. This includes the possibility that genuinely syntactic factors play a role in syntactic change. In order to substantiate this claim, two syntactic changes are studied in more detail, namely the rise of French est-ce que as an interrogative particle and the reanalysis of presentational hay + noun “there is + noun” in Spanish. The authors show that in the first case (rise of est-ce que), the change is pragmatically motivated: the originally free construction est-ce que “is it (the case) that …” structurally separates focal from presupposed information and is therefore used in speech acts with a particularly strong illocutionary force. Crucially, the reanalysis of this construction as a marker of ‘strong’ interrogation does not occur in language acquisition but through frequent usage. The second change discussed in this paper, i.e. the reanalysis of impersonal presentative hay + noun in certain varieties of Spanish, is shown to

INTRODUCTION

5

be triggered by conflicting coding strategies for certain types of arguments, to which presentational constructions are exposed cross-linguistically. We argue that therefore, this process must be considered an instance of syntactically motivated syntactic change. Andreas Dufter’s study “On Explaining the Rise of c’est-clefts” in French calls into question some standard assumptions about cleft constructions and the reasons for their enormous frequency increase from Old to Modern French. Since the most important function of c’est-clefts is to mark focus on the clefted constituent, it has generally been assumed that their strong frequency increase can reasonably be explained by the progressive loss of alternative focusing techniques, in particular by the loss of word stress which, by the 16th century, had been replaced by phrasal stress. Another long-term change which, according to the prevailing view, created a need for cleft sentences as focusing devices was the progressive fixation of word order, since the rise of rigid SVO was increasingly narrowing down the possibilities of focusing by noncanonical word order patterns within the normal sentence. Based on solid corpus evidence, Dufter argues that both claims are highly problematic, since the frequency of c’est-clefts can be shown to have still risen long after the shift from word-based to phrase-based stress and the fixation of SVO were achieved. The real reason for the spread of c’est-clefts, he argues, was clefting ‘beyond necessity’, i.e., the use of cleft sentences for purposes other than focusing the clefted constituent. A major reason for this was the introduction and eventual expansion, from the 16th century onwards, of the so-called informative-presupposition cleft sentence, i.e., a construction-type in which the clefted constituent expressed given or inferable rather than focal information. According to Dufter, this innovation, which is pragmatic in nature, provides a natural explanation not only for the heavy frequency increase of French c’estclefts well after the 16th century, but also for some further, seemingly unrelated syntactic changes. Elisabeth Stark, in “The Role of the Plural System in Romance” studies the interaction of the rise of indefinite and partitive determiners in Romance, and the concomitant loss of nominal morphology being sensitive to mass/count distinctions. While one might be tempted to explain these changes by their apparent functional complementarity, Stark argues that such an account would be too coarse-grained for the intricate grammatical correlations she observes. Specifically, Romance languages that allow argument bare nouns also have residues of the Latin neuter (i.e., Italian and Spanish, as opposed to French). Furthermore, Romance languages with partitive articles also have some morphological ambiguity in their plural marking (namely French and Italian, as opposed to Spanish). Her minimalist analysis suggests the rise of the functional projection Pl(*) as the decisive step. Pl(*) can take two values, Pl* or Pl, which yield an interpretation as mass or plural on the one hand, or as singular on the other hand, respectively. The loss of inflectional endings in Latin

6

ULRICH DETGES & RICHARD WALTEREIT

favoured not only the reanalysis of the numeral as indefinite determiner, but also and crucially the introduction of Pl(*) as the locus of semantic pluralization. Noun phrases in Romance have had henceforth two functional projections: Num (Number) and Pl(*). The interplay of these projections allows her to account elegantly for the descriptive correlations mentioned. Maria Goldbach’s paper on “Morphological Developments Affecting Syntactic Change” deals with the evolution of infinitive constructions in Latin and Old French in a minimalist perspective, with occasional glances at Old Spanish and Old Italian. According to her, the syntactic capacities of the infinitive constructions in the respective languages were heavily affected by morphological changes. In Latin, the formal richness in the infinitive paradigm permitted the expression of the functional features voice, tense, and modality in infinitive complement sentences (i.e., in ACI constructions). The qualities of this paradigm licensed a strong functional category Infl(ection) in the syntactic structure of the ACI. In Old Romance (Old French, Old Italian and Old Spanish) however, the infinitive paradigms covered neither a complete tense nor a complete voice axis, nor could they express modality – consequently, in contrast to Latin, the Old Romance infinitive constructions were characterized by a weak Infl-category. As Goldbach shows particularly for Old French, this fundamental property correlated with further syntactic restrictions on infinitive constructions, especially with the impossibility of using object clitics and with the complete absence of negation in infinitives. That is, Latin infinitive constructions exhibit characteristics that in Old French are shared by finite constructions only. Goldbach conceives of this difference as a difference between infinitive constructions which express propositions (i.e., the Latin ACI) and infinitive constructions which do not (i.e., the Old French infinitive constructions). In the second part of the paper, it is shown that during the evolution from Old to Middle French, the formerly weak Infl-category was gradually strengthened. According to Goldbach, this change was once again brought about by important modifications of the verb morphology, notably by the expansion of the object clitics. Whereas in Old French, only unbounded object pronouns can be placed in front of an infinitive, in Middle French, object clitics are regularly sequenced in this position. As Goldbach argues, object clitics are verbal inflectional affixes, indicative of the propositional capacities of the construction in which they occur. Therefore, their rise in frequency in Middle French infinitive constructions reflects the establishing of a strong Infl-category. Susann Fischer’s contribution “Grammaticalisation in the IP-Domain” is strongly committed to the minimalist program. It deals with a case of wordorder change from Old to Modern Romance (Catalan, Spanish, French). Her explanation is an elaborated generativist response to the functionalist criticism formulated in our contribution (see this volume). Whereas Old Romance had postverbal clitics in finite sentences and stylistic fronting, both features have

INTRODUCTION

7

been lost in Modern Romance. Fischer’s account challenges two long-standing opinions, namely, first, that Old Romance was verb-second (and consequently that postverbal clitics disappeared when this was no longer the case), and, second, that stylistic fronting was needed to guarantee a verb-second structure and/or to fill the subject gap within this structure. Importantly for her argument, Fischer shows that in Old Romance postverbal clitic placement and stylistic fronting not only excluded each other, but furthermore that both were in complementary distribution with negation. Her conclusion is that postverbal clitic placement, stylistic fronting, and negation targeted the same position within the sentence structure, namely a functional category labeled ΣP, which models the speech-act level. Within ΣP, several operators can be realized, among other things ‘emphatic’ affirmation, ‘neutral’ affirmation, and negation. Fischer argues that both stylistic fronting and postverbal clitics were attracted by ‘emphatic’ negation, whereas ‘neutral’ affirmation triggered the structure clitic-verb. How, then, can it be explained that stylistic fronting and postverbal clitics are lost in Modern Romance? In this respect, Fischer subscribes to the ‘classical’ generativist position, according to which language acquisition drives language change by favoring simpler, unmarked structures over structurally more complex representations. Moved items are more marked than basegenerated/merged items, and items merged are more marked than items which are phonologically not realized. Stylistic fronting and postverbal cilitic placement both are instances of movement, and therefore represent the most marked case. Hence, according to Fischer, it is not surprising that they became obsolete. Fischer’s argument makes ample reference to the paradox of change. Her answer to this apory lies in language variation, more specifically in the coexistence of competing grammars (Kroch 1989, 1994, 2006). Giampaolo Salvi, in “Imperfect Systems and Diachronic Change”, provides strong arguments in favor of the ‘random-walk’ view outlined above. The paper addresses the relation of long-term diachronic tendencies to small, local changes. Salvi chooses Italian si-constructions as an example. In Old Italian, a si-construction, like si deve vendere “it should be sold” was a passive construction, just like in Modern Italian. However, in Old Italian si did not occur in impersonal constructions such as Modern Italian si è contenti “one is happy”. Salvi provides a detailed analysis of how si gradually extended to impersonal constructions. At the beginning of the 17th century, si begins to show up in some copula constructions with the auxiliary essere “to be”. This is the first sign of an impersonal si. At the end of the 18th century, more impersonal characteristics become available to the si-construction, namely the full range of copula constructions and compatibility with some direct object clitics. By the middle of the 19th century, the modern system is well established, at least with some authors. Salvi stresses, however, that this is only apparently a systematic long-term transformation. According to him, any assumption of systematic long-term transformations runs a risk of resorting to

8

ULRICH DETGES & RICHARD WALTEREIT

teleological explanations. Rather, he argues, impersonal constructions with si arose in small and in principle unrelated diachronic steps, each of which may have had a completely different motivation. He cites analogy and language contact as the reasons for the local changes. In this sense, syntactic change is much closer to morphological change than commonly thought, since, as Salvi argues, both types of change are regulated by local, often analogical, adjustments, with imperfections remaining. Martin Becker’s corpus-based study “From Temporal to Modal: Divergent Fates of the Latin Synthetic Pluperfect in Spanish and Portuguese” provides an alternative explanation to changes with a long-term format. The paper traces the evolution of the morpheme -ara, a descendant of Latin -averam. In Latin, the functions of -averam were mainly temporal, but already included usages with a modal flavor. Curiously, in Spanish and Portuguese -ara evolved in opposite directions: whereas it is an exclusively temporal category in Modern Portuguese, it acquired counterfactual and irrealis functions in Spanish, and was eventually transformed into a past subjunctive. According to Becker’s account, which crucially relies on formal logical and conceptual analyses, this shift was brought about by a long-term sequence of smaller changes. It started out in conditional sentences and their cognate structures. At later stages, -ara was increasingly used as an irrealis marker in independent clauses. As Becker shows, the individual steps of this long-term change were far from random – their overall direction was determined by the cognitive schemata underlying the respective grammatical constructions. The main mechanism of change involved was, according to Becker, reanalysis: at different stages of the development, the modal extensions of the original pluperfect meaning were taken to be the form’s proper meaning. As a consequence of this, the speakers finally reversed the inherited relationship between meaning and meaning extension: -ara, formerly a morpheme with temporal meaning which allowed modal extensions, was reinterpreted as a modal form with temporal extensions. This reversal was accompanied by a shift in quantity of the respective functions: in Spanish, the frequency of temporal uses dramatically declined until, by the end of the 15th century, they dropped out of usage. In “Non-lexical Core-arguments in Basque, Romance and German: How (and why) Romance Syntax is Shifting towards Sentential Head-Marking and Morphological Cross-Reference”, Hans-Ingo Radatz deals with the problem of long-term typological shifts. In particular, he addresses the typological characterization of Romance person/number inflection and clitic pronouns, a long-standing issue in Romance diachronic linguistics. Basing himself on Johanna Nichols’ (1986) distinction of head-marking vs. dependent-marking languages, he contrasts Ibero-Romance and French mainly with Basque but also with other languages. In head-marking languages like Basque, NPs are not realizations of the verb’s arguments; rather, they are facultative appositions, the arguments themselves being realized on the verb as the sentence’s head itself.

INTRODUCTION

9

Accordingly, Basque does not have agreement but cross-reference between NP and verb. Taking this as a backdrop for the analysis of Romance, he argues that Ibero-Romance has a system with the typical Indo-European full head-marking for subjects, but is increasingly developing head-marking also for objects. A consequence of this claim is that argument-NPs in Ibero-Romance are in fact not phrasal arguments but rather free appositions. Romance languages are therefore, according to Radatz, diachronically shifting towards headmarking on the sentence level. The volume is rounded off by a paper which addresses the explanation of grammatical change from a broader perspective. In “Towards a Comprehensive View of Language Change: Three recent approaches”, Esme Winter-Froemel discusses evolutionary approaches to language change. Her survey of these approaches provides fine-grained conceptual distinctions which, according to her, can be made fruitful for the theory of language change in general. In particular, the author distinguishes three types of evolutionary approaches: firstly, metaphorical ones, which choose biological evolution as a mere analogy to language change, secondly, biologistic views, for which language change is part of the evolution of the species, and finally generalized views, which take language change and biological evolution as both instantiations of a more general process of variation-and-selection. Cross-cutting these, there is a further distinction between adaptive and two-level views. Adaptive views try to account for long-term diachronic changes, whereas two-level views focus on the interplay of low-level speaker-hearer interaction and the drift of the language. Winter-Froemel argues that the conceptual distinctions provided by her survey shed light on the problems faced by any account of language change. She proposes an explanation scheme for language change which carefully separates levels of linguistic analysis, stages and factors involved in language change. We hope that the present volume is of interest to a broad public within linguistics. Syntacticians, typologists, historical linguists and romanists alike may find it a helpful contribution to their respective field of study.

REFERENCES Battye, Adrian & Ian G. Roberts. 1995. “Introduction”. Clause Structure and Language Change, ed. by Adrian Battye & Ian G. Roberts, 3-28. Oxford: University Press. Bybee, Joan. 1985. Morphology: a study of the relation between meaning and form. Amsterdam & Philadelphia: John Benjamins. ----------. 2001. Phonology and Language Use. Cambridge: University Press.

10

ULRICH DETGES & RICHARD WALTEREIT

----------. 2005. “The Impact of Use on Representation: Grammar is usage and usage is grammar”. LSA Presidential Address, LSA Annual Meeting 2005 [http://www.unm.edu/~jbybee/Bybee%20plenary.pdf]. Coseriu, Eugenio. 1978 [1957]. Sincronia, diacronia e historia: El problema del cambio lingüístico. Madrid: Gredos. Greenberg, Joseph. 1966. Language Universals. With special reference to feature hierarchies. The Hague: Mouton. Haspelmath, Martin. 1998. “Does Grammaticalization Need Reanalysis?”. Studies in Language 22. 315-51. ----------. 1999. “Optimality and Diachronic Adaptation”. Zeitschrift für Sprachwissenschaft 18. 180-205. Hopper, Paul. 1987. “Emergent Grammar”. Berkeley Linguistics Society: Proceedings of the 13th annual meeting of the Berkeley Linguistics Society. Berkeley, Calif., 139-157. Keller, Rudi. 1994 [1990]. Sprachwandel. Von der unsichtbaren Hand in der Sprache. 2nd ed. Tübingen & Basel: Francke. Kroch, Anthony. 1989. “Reflexes on Grammar in Patterns of Language change”. Journal of Language Variation and Change 1. 199-244. ----------. 1994. “Morphosyntactic Variation”. Proceedings of the Thirtieth Annual Meeting of the Chicago Linguistics Society. Vol. II, ed. by K. Beals, 180-201. Chicago: Linguistics Society. ----------. 2006. “Modeling Language Change and Language Acquisition”. [http://www.ling.upenn.edu/~kroch/courses/lx556/lsa-forum.pdf]. Lehmann, Christian. 1995 [1982]. Thoughts on Grammaticalization. Munich: Lincom. Lightfoot, David. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press. ----------. 1991. How to Set Parameters: Arguments from language change. Cambridge, Mass.: MIT Press. ----------. 1997. “Shifting Triggers and Diachronic Reanalyses”. Parameters of Morphosyntactic Change, ed. by Ans van Kemenade & Nigel Vincent, 253-272. Cambridge: Cambridge University Press. ----------. 1999. The Development of Language: Acquisition, Change, and Evolution. Malden, Mass. & Oxford: Blackwell. Longobardi, Giuseppe. 2001. “Formal Syntax, Diachronic Minimalism, and Etymology: The history of French chez”. Linguistic Inquiry 32. 275-302. ----------. 2003. “On Parameters and Parameter Theory”. Syntaxtheorien. Modelle, Methoden, Motive, ed. by Elisabeth Stark & Ulrich Wandruszka, 273-290. Tübingen: Gunter Narr. Meisel, Jürgen. 2001. From Bilingual Language Acquisition to Theories of Diachronic Change. (= Arbeiten zur Mehrsprachigkeit - Working Papers in Multilingualism, Universität Hamburg: Sonderforschungsbereich 538 “Mehrsprachigkeit”.)

INTRODUCTION

11

Newmeyer, Frederick J. 1998. Language Form and Language Function. Cambridge, Mass.: MIT Press. ----------. 2003. “Grammar is Grammar and Usage is Usage”. Language 79. 682-707. ----------. 2005. “A Reply to the Critiques of ‘Grammar is Grammar and Usage is Usage’”. Language 81. 229-236. Nichols, Johanna. 1986. “Head-marking and dependent-marking grammar”. Language 62. 56-119. Roberts, Ian. 1993. “A Formal Account of Grammaticalization in the History of Romance Futures”. Folia Linguistica Historica 13. 219-258. ---------- & Anna Roussou. 2003. Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: University Press. Schuchardt, Hugo. 1885. Über die Lautgesetze: Gegen die Junggrammatiker. Berlin: Robert Oppenheim. (Reprint in The Lautgesetz-Controversy: A documentation, with an introduction by Terence H. Wilbur, Amsterdam: John Benjamins, 1977.) van Gelderen, Elly. 2004. Grammaticalization as Economy. Amsterdam & Philadelphia: John Benjamins.

SYNTACTIC CHANGE FROM WITHIN AND FROM WITHOUT SYNTAX A USAGE-BASED ANALYSIS

RICHARD WALTEREIT Newcastle University ULRICH DETGES University of Munich

This article presents a critical discussion of Inertia Theory (Longobardi 2001; 2003), according to which syntactic change never originates in syntax itself. We argue that syntactic change originates in language usage, which includes the possibility that it may be triggered by genuinely syntactic factors. In order to substantiate this claim, two syntactic changes are studied in detail, namely the rise of French est-ce que as an interrogative particle and the reanalysis of presentational hay + noun “there is + noun” in Spanish. We show that the reanalysis of est-ce que as a marker of ‘strong’ interrogation is pragmatically motivated and brought about by frequent usage. By contrast, the reanalysis of impersonal presentative hay + noun in certain varieties of Spanish is shown to be triggered by conflicting linking strategies, and must therefore be considered an instance of syntactically motivated syntactic change.

1. The problem: The locus of syntactic change According to a recent hypothesis, syntax is diachronically inert (Longobardi 2001). This claim is based on arguments of two different kinds: First, it seems to be an elaboration of the observation made by traditional historical linguists that syntax is a domain especially resistant to short-term change. Second, it represents a state of discussion in current generative theory. As pointed out in Lightfoot 1999 and Longobardi (2001:277, 2003:283-4), diachronic change is a challenge to basic assumptions of generative theorybuilding. From a synchronic perspective, languages compatible with universal grammar are perfectly running systems. Hence there can be no synchronic mechanism, for instance a single syntactic rule or a syntactic parameter, which carries in itself the germ of change. Change of a perfectly running system cannot be caused by properties of the system itself. For a theory of the language system, change is beyond the scope of description. Put more simply: from this point of view, diachronic change should not exist (Lightfoot 1999, Longobardi 2001:277), which is obviously contrary to fact. Generative theory

14

RICHARD WALTEREIT & ULRICH DETGES

has proposed various ways to resolve this paradox, all of which confine change to peripheral domains of language. Many authors invoke, in various ways, language acquisition as a ‘weak point’ in the transmission of grammar, that is, as the locus of change. For Roberts (1993) and Lightfoot (1999), all modules of grammar, including syntax, are subject to possible change in acquisition. From these approaches, the so-called Inertia Theory distinguishes itself in that the ‘periphery’, i.e. the locus of change, is defined in different terms. If syntax is the core of the grammar, it might be reasonable to assume that change only originates in other, less central, modules of grammar. In other words, syntax can be affected by language change, but such change never originates in syntax itself. On this view, syntactic change can only arise as a consequence of semantic, phonological, etc. change. From what has been said so far, a logical problem of Inertia Theory becomes apparent. It clearly is not an elaboration of the ‘pre-theoretical’ observation made by traditional historical linguists that syntax is relatively resistant to change (see above). In particular, it makes no prediction about the scarcity of ‘pre-theoretically’ defined syntactic change. Recall that it only excludes change originating in syntax, not change concerning syntax. This means that Inertia Theory, even though apparently maximally restrictive, makes no empirically testable predictions about syntactic change. Certain claims inherent in Inertia Theory, however, are interesting. In particular, it distinguishes itself from other frameworks, for instance from classical grammaticalization theory (cf. Lehmann 1985), or from earlier generative approaches (cf. Lightfoot’s 1979 ‘transparency principle’) in that it excludes change induced by the language system itself. We would subscribe to this view, but for different reasons. Language systems are diachronically inert because they are conventional objects. Social conventions do not change by themselves. In order to alter them, special things have to happen. What is needed, then, is a theory that provides for both the relative stability of language systems as well as for their openness to change. The above-mentioned paradox of language change has already been discussed in the structuralism of the 1950s. The solution proposed by Coseriu (1978 [1957]:68-9) is that language systems have to be constantly enacted in order to continue their existence. The necessity of continuous enactment and re-enactment opens the door for change. Put more simply, language change arises in language use. The notion that language change arises in language use involves no restriction as to the domains where it can originate. This means that syntactic change may very well arise in the usage of the respective syntactic structure itself. As all change, syntactic change is often caused by semantic and pragmatic reasons, but sometimes it is also motivated by genuinely syntactic factors, although these do not have the status of synchronic rules. Spontaneous innovations triggered by language use can directly affect syntax without

SYNTACTIC CHANGE FROM WITHIN AND FROM WITHOUT SYNTAX

15

necessarily passing through other components of grammar (e.g. pragmatics, semantics, phonology etc.). This claim will be substantiated in section 3. Usage-based change is not unconstrained. As we are going to show, there is one very general but at the same time extremely powerful principle, the principle of reference, which underlies any type of functional reanalysis (see section 2). However, this principle is not a rule with clear-cut domains of application and predictable outcomes. Furthermore, we will discuss certain tacit assumptions on language change which prevail especially in the generative literature. For reasons outlined above, the generative tradition invokes imperfect language acquisition as the locus of change. According to this view, a necessary condition for change is a shift in the evidence available to the acquirer. Generative approaches to language change differ as to the type of evidence for which they allow. For Roberts (1993), syntactic change passes through structure simplification in acquisition, which in turn may be motivated by a previous shift in usage at a stylistic level. These processes can still be measured by the categories of traditional historical description. Lightfoot (1999), in the spirit of chaos theory, conceives of change as triggered by unpredictable probabilistic cues, which are inaccessible to large-scale historical description. That is, for Roberts (1993) and Lightfoot (1999), the shift in the available evidence can, in principle, originate in syntax itself. For Longobardi (2001; 2003), in contrast, the loss of a lexical element (that is, a change which by definition must occur outside syntax, see above) leads to change in the available evidence, thereby triggering syntactic reanalysis.1 Our position differs from this viewpoint in that change based on linguistic evidence is considered only one among many mechanisms of change, and not even a particularly important one (see section 3). In order to substantiate our claims, we will discuss two cases of syntactic change, one of which is motivated by pragmatic factors, the other one by syntactic factors. Moreover, both examples represent two different types of reanalysis.

1 According to Longobardi (2001), the Latin full noun casa ‘house’ developed into the French preposition chez ‘at’ via what he calls a “construct state” in which casa was used as a full noun in a prepositional phrase ‘in the house of’. What caused the reanalysis of the “construct state” as a preposition was the loss of the full noun casa (which is replaced by mansione). However, there seems to be a logical inconsistency in this argument: If the full noun casa disappeared because it was replaced by mansione, why then was mansione not available for the “construct state” and for what reason did casa survive in this construction? This state of things seems to indicate that the Romance “construct state” was itself already the product of a foregoing reanalysis and / or an entrenchment process. The construct state, then, would not be a possibility of Universal Grammar, but an idiosyncratic construction in the sense of Construction Grammar.

16

RICHARD WALTEREIT & ULRICH DETGES

2. The rise of French est-ce que as an interrogative particle: a pragmatically motivated syntactic change In Modern French, est-ce que is a sentence-initial particle in yes/no interrogatives (1) as well as in wh-questions (2): (1) Est-ce que Pierre viendra demain? Pierre come.FUT.3S tomorrow INT “Will Pierre come tomorrow?” (2) Pourquoi est-ce que Pierre viendra why INT Pierre come.FUT.3S “Why will Pierre come tomorrow?”

demain? tomorrow

In Modern French, est-ce que is a not a free combination, even thought all of its elements, est “is”, ce “it, this”, and que “that”, exist as independent items. Rather, est-ce que is a particle with a non-compositional meaning. First, it does not mean “is it (true) that” as it should if it were a free combination of autonomous words. Second, its elements are not exchangeable with otherwise synonymous expressions (3a). In particular, est-ce que is not sensitive to tense (3b). Its non-compositional character is the result of a language change which we will discuss in this section. (3) a.*Est-cela que Pierre be.3S-that that Pierre “Is that that Pierre will come?”

viendra? come.FUT.3S

(3) b.*Sera-ce que Pierre be.FUT.3S-this that Pierre “Will it be that Pierre will come?”

viendra? come.FUT.3S

Sentence-initial est-ce que guarantees a relatively simple syntactic structure, because the rest of the sentence has exactly the same form as the corresponding declarative (4a). In particular, it is syntactically simpler than the interrogative inversion, another interrogative sentence type (4b), but less simple than the question marked only by a rising intonation contour (4c): (4) a. Pierre viendra demain. Pierre come.FUT.3S tomorrow “Pierre will come tomorrow.” (4) b. Pierre viendra-t-il demain? Pierre come.FUT.3S-he tomorrow “Will Pierre come tomorrow?”

SYNTACTIC CHANGE FROM WITHIN AND FROM WITHOUT SYNTAX

17

(4) c. Pierre viendra demain? Pierre come.FUT.3S tomorrow “Will Pierre come tomorrow?”

Compared to their competitors (4b) and (4c), est-ce que interrogatives are syntactically ‘optimal’: First, est-ce que indicates the sentence type at their beginning, which makes them perceptually easy to process. Second, the rest of the clause is identical with a corresponding declarative, thus being easy to produce. In spite of this, est-ce que is relatively rare. It only covers about 5-14% of yes/no-questions in spoken French (Söll 1971:497, Seelbach 1983:277, Hansen 2001:479) and even less in written French. This has to do with its restricted function. In discourse, est-ce que questions are marked interrogative devices. According to the literature, they appear at prominent places in discourse. For example, they figure text-initially, or in the conclusion, or mark a transition between discourse topics (Söll 1971:498, Seelbach 1983:286, Weinrich 1982:743, Hansen 2001:510). They are ‘strong’ interrogatives in that, unlike questions formed with other constructions, est-ce que questions always require an answer. Rhetorical questions formed with est-ce que are exclamative speech acts with a particularly strong illocutionary force: (5) Qu’ est-ce qu’ il a travaillé! what INT he has worked “Damn, how hard did he work!”

At the same time, however, est-ce que interrogatives are topical in that they aim at information which normally is not new to the speaker. Hansen (2001) shows that they are often used to ask for something which is already accessible in the universe of discourse. In the following example, the point of the question is not to elicit new propositional information. Rather, it aims at the viewpoint of the addressee on this problem. (6) RL: alors à / à propos / à propos des listes e, e les élections “so with / with respect / with respect to the tickets er, er the European elections européennes ont lieu, à la représentation proportionnelle c’:est en will be proportional elections effet l’occasion pour chaque parti d’être représenté pour ce qui it’s indeed an occasion for each party to be represented nous concerne, e c’est / c’est de cette manière,que nous le / que as far as we’re concerned, er it’s / it’s in this way, that we, that nous concevons l’élection e, d’une façon toute simple we see the elections er quite simply Ch: est-ce que pour e pour la majorité pour la gauche, c’est mieux for er for the majority for the left, is it better

18

RICHARD WALTEREIT & ULRICH DETGES

d’avoir une liste ou deux listes de l’opposition? to have one or two opposition tickets? RL: ben je viens de vous […] (VS2, 16–17, cf. Hansen 2001:509) well just now I […]”

We would like to assume that both the ‘strong’ function of Modern French est-ce que as well as its topical character are conventionalizations of implicatures attached to its diachronic predecessor. In Old French, as free constructions, est-ce que interrogatives were complex sentences, est-ce being the predicate of an autonomous matrix clause, while que introduced a subordinate clause (“is it that…?”). According to the grammar of Old French, the interrogative character of the main clause was marked by inversion of subject and predicate (estV-ceSUBJ “is it?”). Est-ce que is attested earlier in wh-questions than in yes/no questions, where it is found only from the mid-16th century onwards (Foulet 1921:264-5). Wh-interrogatives with est-ce que are the counterpart of cleft-constructions of the type c’est X qui/que… “it is X who/which …”. In other words, the primary function of these interrogatives is to focalize the argument represented by the wh-phrase: (7) a. Focal est-ce que wh-question… Que c'est ce qu' ele a tant chacié? (LANCELCB) what this is this which she has so.much chased “What is it that she has been looking for so hard?” (7) b. …and the corresponding cleft-construction C’est cela qu’ ele a tant chacié. This is that which she has so.much chased “It’s that what she has been looking for so hard.”

The grammatical properties of free interrogative est-ce que combine three aspects. First, it strongly highlights the participant represented by the whelement. Second, the content of the embedded que-clause is treated as presupposed, i.e., as information taken for granted by the speaker at the time of utterance. Third, the highlighted and the presupposed elements are neatly kept apart by assigning them to the different clauses of the complex sentence. These three features are what we will call the grammatical meaning of free est-ce que. By virtue of these structural properties, the construction qualified for at least two different textual functions. In the following passage taken from the Cent Nouvelles Nouvelles, a crook promises three gentlemen to show them a church where they can see God. Arrived at that church, the three gentlemen open up the following dialogue:

SYNTACTIC CHANGE FROM WITHIN AND FROM WITHOUT SYNTAX

(8) Où est ce que nous verrons where is this that we see.FUT.1P “Where is it that we shall see God?”

19

Dieu? (CNNCB 63) God

Je le vous monstreray I him you.DAT show.FUT.1S “I’m going to show Him to you.”

The point of the free est-ce que construction here is not only to highlight the wh-element où “where”, but also to treat the state of affairs expressed in the subordinate clause (“we shall see God”) as topical information, that is, information already accepted as valid by both speaker and hearer. The pragmatic function of this is to remind the crook of his previous promise (“you will see God”), made a couple of lines before the quoted passage. The main function of free est-ce que in Old French, however, exploits the focal character of the wh-element. (9) Qui est ce, Diex, qui me descuevre? who is this, God, that me uncover.3S “God, who is it that is uncovering me?” (Trist. I, 232, cf. Schulze 1888:94-5)

Interrogative sentences can be used to convey speech acts other than questions: (10) Qu’est ce, deable? Quex sunt ti dit ?/ Qu’ est ce que tu as ici dit? what is this, devil? what are your sayings? what is this that you have here said? “What is it, devil? What are you saying? What is it that you have said?” (MEUGL2:11496)

In (10), the est-ce que interrogative encodes not a question, but an exclamative (the speaker knows actually very well what the other person said and merely expresses her surprise). Still, we can observe all the effects already discussed with respect to questions. The est-ce que sentence conveys topical information (it is the third in a series of roughly synonymous exclamations), and it focuses on the wh-argument. But why should a non-question carry focus on a wh-argument? The reason is that this makes the entire utterance a speech act with a strong illocutionary force. This holds also for the use of the interrogative as a question, as in (9), as can be seen from the exclamative particle Diex! “God!”. In fact, marking the illocutionary force of a speech act as particularly strong seems to have been the main motivation of free est-ce que in medieval French: Partout au XIIe et au XIIIe siècle qui est ce que, qu’est ce que expriment surprise, indignation, colère, dégoût. C’est tout au plus si les locutions s’affaiblissent à ne marquer qu’un sentiment de vive curiosité. (Foulet 1921:253)

20

RICHARD WALTEREIT & ULRICH DETGES

“Everywhere in the 12th and 13th century, qui est ce que, qu’est ce que express surprise, indignation, anger, disgust. Only exceptionally these constructions weaken so as to mark a feeling of strong curiosity.”

Hence, originally, the strength of the illocutionary force is a side-effect of the focal character of est-ce que as a free construction. To sum up, free est-ce que is a construction which grammatically separates focal from presupposed information. This structural property can be exploited to make strong speech acts, above all strong questions, i.e. questions requiring an answer. This means that a distinction is needed between the grammatical meaning of free est-ce que as such and the function it can fulfil in discourse. The relationship between both is of course not random, as the function of marking strong speech acts can be derived from the grammatical meaning by pragmatic inferencing. But as long as est-ce que is a free, non-conventional construction, this function must be compositionally computed on the basis of the various constituents, every time a speaker builds a new structure with est, ce, and que. Diachronic reanalysis changed the originally free est-ce que into the particle we know from Modern French. In the outcome of this process, the discourse function mentioned, namely ‘strong’ question, was turned into the conventional meaning of the construction as a whole. This process is motivated by the frequent use of the construction with that particular function, in the course of which the function becomes progressively associated with the construction. In the literature, this process is referred to as entrenchment (Langacker 1987:59-60, Bybee 2001:10-12, 112-117). Progressive entrenchment means that the function of the construction is no longer computed on the basis of its elements by pragmatic inferencing, but that it is directly assigned to the construction as a whole. The mechanism underlying this change is the most important principle of diachronic reanalysis: the “principle of reference” Detges & Waltereit 2002:156), which stipulates: (11) Match the sound string you hear with what seems to be its function in the situation.

(11) is designed to simplify the relation between meaning and function. It is a strategy of interpretation and understanding (for more details, see Detges & Waltereit 2002). In this sense, it is a hearer’s strategy. As a consequence of this reanalysis, est-ce que is no longer a focal construction. In particular, it ceased to be the interrogative counterpart of the cleft-construction c’est X qui (7b). (For the resulting difference in interpretation of wh-questions, see Foulet 1921:253). Hence the new unitary meaning of reanalyzed est-ce que is (one of) its former function(s). As a consequence of this semantic-pragmatic step, the formerly complex syntactic structure is lost and the construction is reanalyzed as an interrogative particle.

SYNTACTIC CHANGE FROM WITHIN AND FROM WITHOUT SYNTAX

21

For a critical discussion of Inertia Theory and its implications, a brief look is needed at certain historical circumstances of the reanalysis sketched in the previous paragraph. Before syntactic reanalysis, est-ce que constructions exhibit full variability of their elements. In particular, they are marked for tenses other than the present tense (12a), for persons other than the 3S (12b), and they contain pronouns other than the demonstrative neuter ce (12c) (Kaiser 1980) or even no pronoun at all (Schulze 1888:91-94) (12d): (12) a. A quel jeu sera que nous jourons encependant? play.FUT.1P meanwhile at what play be.it.FUT.3S that we “What game will it be that we shall play in the meantime?” (Farces, cf. Kaiser 1980:120) (12) b. Mais quelz gens sont-ce qui y sont? But which people are-this who there are “But who are the people (who are) there?” (Farces, cf. Kaiser 1980:120) (12) c. Qui est celuy qui nous espie? who is the.one who us spy.3S “Who is the one (over there) who spies on us?” (Farces, cf. Kaiser 1980:120) (12) d. Dex, fait il, ki est ki la gist? God, says he, who is.it who there lie.3S “God, he said, who is it that is lying there?” (Perceval 21026, Schulze 1888:92)

The question as to when exactly the reanalysis of est-ce que occurred is a controversial issue in French philology (see Kaiser 1980:112-125 for discussion). There are, however, certain indications for its progressive entrenchment from the 14th century onwards: In particular, some texts exhibit spellings as esse instead of est-ce, which is suggestive of a non-compositional analysis of the sound string est-ce que.2 Hence, it seems that phonological change from est-ce que to esse que is a (possible) consequence of syntactic reanalysis (which is itself triggered by semantic-pragmatic factors). (13) Qu’ esse qu' il a? (PATHELCB, 15th century) what INT he has “What does he have?”

Importantly, the entrenchment of the construction as a conventional marker of interrogation and its concomitant reanalysis as a particle do not exclude a 2 Note, however that these spellings concern est-ce not only as part of est-ce que, but also in other syntactic contexts.

22

RICHARD WALTEREIT & ULRICH DETGES

continuing use of the original free construction in contexts other than ‘strong’ questions (in fact, this use continues till this day).3 But we can be sure that estce que has been reanalyzed as a particle once it is found to introduce a second, free, est-ce que construction meaning “is it (true) that” as in (14). Here it is obvious that est ce que is simply used as a device to mark the interrogative character of free (ce) sera que “it will be (the case) that…”, since otherwise the construction would be redundant. (14) Quant esse que1 sera que2 vous donrrez à when INT1 be.it.FUT.3S that2 you give.FUT.2P to “When will it be that you will give a banquet?” (Farces, cf. Kaiser 1980:125)

banqueter? banquet

As for the relationship between grammar modules, our analysis has shown a primacy of function over form. Reanalysis based on the principle of reference changes the meaning of an originally free construction. Concomitantly, this construction turns into a syntactically frozen formula. The erasure of syntactic boundaries may then be followed by the erasure of phonological word boundaries (est-ce > esse). We have thus the following ordering of grammar modules: (15) Pragmatics/semantics > syntax > phonology

This change seems to confirm the assumption, basic to Inertia Theory, that syntactic change is caused by factors external to syntax. In particular, according to our analysis, this change cannot be explained by the rise of SVO at the expense of V2 in French. As is often argued in the literature (Foulet 1921), est-ce que interrogative sentences conform to the SVO pattern of declaratives (see above, (4a-b)), and therefore could have been favored by a general tendency to replace the original Old French V2 pattern by SVO from the 13th century onwards. However, this view is clearly contradicted by the fact that the ‘breakthrough’ of SVO in declarative sentences was not paralleled by an analogous advance of est-ce que interrogatives. The relative frequency of est-ce que questions in Modern French is basically the same as it was in the 15th century. A major limit to a further rise in frequency was that est-ce que interrogative sentences are marked constructions with a very specific function (i.e. ‘strong’ interrogation). Contrary to central assumptions of Inertia Theory and generative theory in general however, the reanalysis of est-ce que is not acquisition-based. The key to the reanalysis lies in frequency. Repeated use of est-ce que in the function of a ‘strong’ question leads to entrenchment of the form-function-matching and finally gives rise of a new linguistic element. Frequency, however, is not a 3 This is evidenced by examples such as Qui était-ce qu'on enterrait ? “Who was it that they buried there?” (Jules Romains, 1911, FRANTEXT).

SYNTACTIC CHANGE FROM WITHIN AND FROM WITHOUT SYNTAX

23

parameter of the language system or of language acquisition, but of language use. Another critical remark concerns the role of linguistic evidence in reanalysis: According to generativist authors, reanalyses are usually triggered by a change in the comparative evidence available to children. However, what could be evidence of such type in the case of est-ce que? The ‘evidence’ responsible for the reanalysis of est-ce que is the fact that already the free construction has frequently been used in the very function that is to become its conventional meaning. But this clearly is not comparative evidence. Still worse: the construction has been reanalyzed in spite of the existence of counter-evidence. In (14) we find est-ce que as an interrogative particle along with a free (ce) sera que “will it be that…”, which is evidence for the free state of the construction. Put in more general terms, the tacit assumption underlying the generativist ‘reanalysis scenario’ poses an empirical problem: it predicts that the free construction, i.e. the ‘input’ to reanalysis and its ‘output’, i.e. the new construction, should not exist side by side in the same grammar. Yet, empirically, this constellation is rather frequent: In French, pas as a negation particle coexists with pas as a lexeme “step” (to which it goes back diachronically). In the same fashion, English has going to as a marker of the near future along with the full verb to go, which is its historical ancestor. Spanish possesses a complex preposition frente a “in front of” as well as the latter’s lexical source, the full noun frente “front”. Analogous examples in many languages spring to mind quickly.4 We do not claim that comparative linguistic evidence never plays a role in language change. However, change based on such comparative evidence is only one type among others, and it requires certain additional conditions. This claim will be substantiated in the next section. 3. Shift of grammatical function in Spanish presentational constructions: a syntactic change caused by syntactic factors We would now like to discuss a case of syntactic change which is caused by syntactic factors while at the same time being usage-based. Like many languages, Spanish has a special presentational construction:

4 This is also an argument against Longobardi’s (2001) argumentation. According to him, the construct state chez ‘at X’s place, in X’s house’ (< Lat. casa ‘house’) in Old French could only be reanalyzed as a preposition because casa ‘house’ as a full noun was replaced by mansione ‘dwelling’. Our argument, however, implies that chez acquired the prepositional meaning independent of the loss of the full noun use (cf. note 1). Similarly, German mithilfe von (or mit Hilfe von, according to the new German spelling) ‘by means of’ is analyzed as a complex preposition in spite of the existence of Hilfe ‘help’ as a full noun.

24

RICHARD WALTEREIT & ULRICH DETGES

(16) Hay coches en el patio. there.is car.PL in the courtyard “There are cars in the courtyard.”

Coches, the only complement in this construction, is a direct object, as can be seen from the lack of agreement (cf. (16)) and the possibility of anaphoric substitution (see (16’), (16’’)): (16’) El coche, lo hay en the car 3SG.ACC there.is in “The car, it is in the courtyard” (16’’)*El coche, el hay en the car 3SG.NOM there.is in “The car, it is in the courtyard”

el the

patio. courtyard

el patio. the courtyard

In (16’), lo is a direct object in the accusative case; in (16’’), however, nominative el is incompatible with el coche because of the latter’s direct object status. However, in certain dialects of Spanish, especially of American Spanish (Bentivoglio & Sedano 1989:76-77, Fontanella de Weinberg 1992:152-154), this construction displays agreement between haber and the complement: (17) Habían soldados en el patio. there.were soldier.PL in the courtyard “There were soldiers in the courtyard.”

Here, the complement (soldados) has changed to subject status, as can be seen from the agreement behavior of the verb. Curiously, the choice of this variant seems to be favored by past tense usage. In the present tense, the same varieties favor the construction type (16), with coches as a direct object. The standard explanation for this asymmetry is that the presentational present tense form hay is not the regular 3SG present of the verb haber, but the idiosyncratic diachronic outcome of a fusion of ha “have-3SG” and y “there” in the 13th century (Penny 1991:162). Thus, in the present tense, presentational haber/hay + NP does not have a regular plural counterpart. However, in some dialects presentational plural forms built on hay, such as hayn or haen, do exist, albeit sporadically (Montes Giraldo 1982:384). This shows that morphological considerations alone cannot explain the scarcity of subject agreement in present tense usages. The contradictory syntactic patterning of haber + noun in Spanish is symptomatic of presentational constructions cross-linguistically. Presentational constructions are specialized in introducing new participants into discourse. By virtue of their being new, these participants are focal (Lazard 1994:4). Normally, they have the status of single arguments of the construction. This follows from the function of presentatives, which do not predicate something

SYNTACTIC CHANGE FROM WITHIN AND FROM WITHOUT SYNTAX

25

about a referent but simply introduce the new referent and locate it in discourse (Lambrecht 1987:227). Cross-linguistically, the single arguments of these constructions are sometimes encoded as subjects, sometimes as objects and at times as something which is neither subject nor object (Lazard 1994, Givón 2001:191-192). Thus, English has the presentational construction there is: (18) There are apples in the garden.

Here, the coding properties of there are (agreement) point to a subject status of the only argument apples. Other properties however, in particular behavioral properties of this constituent, seem to deny apples a subject status: (19) a. b. c.

There are apples in the garden. → I expect there to be apples in the garden. → *I expect apples to be there in the garden.

(20) a. b. →

He comes. I expect him to come.

In the control construction, there in (19b) occupies the same syntactic slot as the subject constituent of a standard sentence (20b), while this position is excluded for apples (19c). This in turn means that apples in (19a) does not have the full range of properties of a normal subject (cf. Lazard 1994:12-14). Another example is the following presentational construction of Standard French: (21) Il arrive des gens. 3SG arrive.3SG PARTITIVE.PL people “There are people arriving”

Coding properties (agreement) and behavioral properties (post-verbal position in a strict SVO language) point to an analysis of des gens as a direct object, the formal subject being the expletive element il. The problem with this interpretation, however, is that arriver is not a transitive verb and therefore should not govern a direct object. In non-presentational standard sentences, les gens has the coding and behavioral properties of a regular subject: (22) Les gens arrivent. The.PL people arrive.3PL “People arrive”

What is the reason for this bizarre patterning? As already pointed out, presentational constructions are specialized for introducing new participants into discourse. New participants represent by definition focal information. Focal participants in turn are most naturally coded as non-subjects, typically as direct objects (Givón 1979:52). At the same time however, presentational

26

RICHARD WALTEREIT & ULRICH DETGES

constructions normally have (for reasons pointed out above) only one argument-participant (Lazard 1994:3).5 The unmarked syntactic role of only arguments is the subject status. Thus, presentational constructions are subject to two conflicting coding constraints observable in ‘normal’ sentences elsewhere in the language: (23) 1. Focal information is coded as non-subject 2. Single arguments are coded as subject

Individual presentational constructions can only satisfy one of these constraints and must necessarily violate the other one (Koch 2003:157). Returning to our construction in (16) hay coches en el patio vs. (17) habían soldados en el patio we can observe that (16) conforms to (23)-1 and violates (23)-2, whereas (17) satisfies (23)-2 but is in contradiction with (23)-1. In the change from (16) to (17), the criterion for the status of the only argument is shifted from (23)-1 to (23)-2. This change is likely to occur in constructions with singular NPs like (24) which are open to both interpretations:6 (24) Había un soldado en There.was INDEF.SG soldier.SG in “There was a soldier in the courtyard”

el the

patio. courtyard

The reanalysis leaves the surface of the construction unaltered. Hence it goes unnoticed as long as the argument in question is realized as a singular NP. The change consists in analyzing an opaque sound string by means of a constraint which can be observed to underlie transparent sound strings of the same language (in this case constraint (23)-2 rather than (23)-1). (25) a. Before reanalysis: Había un soldadoDO en el patio [← (23)-1] b. After reanalysis: Había un soldadoSUBJ en el patio [← (23)-2]

This type of reanalysis is different from the one described in section 2. It follows from a principle which we have termed the “principle of transparency” (Detges & Waltereit 2002:159):

5

Lambrecht (1987:227-8) discusses, under the label of presentational constructions, French allnew sentences of the type ya un camarade d’usine qui m’a ramené en voiture “there’s a pal from work who took me home in his car”. Lambrecht himself acknowledges that sentences of this type consist of two constructions, namely a) a presentational clause ya X ‘there is X’ with only one argument, and b) an embedded relative clause containing a further predication about the newly introduced participant. 6 Note that this does not mean that the construction is ambiguous. Before reanalysis, its single argument is coded as direct object. This interpretation, and only this one, is sanctioned by the traditions of the speech community until the reanalysis takes place. By “openness” we only mean that the non-subjecthood of the only argument is not overtly marked.

SYNTACTIC CHANGE FROM WITHIN AND FROM WITHOUT SYNTAX

27

(26) Match the sound string you hear with other sound strings of the same language.

Note that reanalyses of this type tacitly comply with the principle of reference pointed out in section 2, since the function of the construction, namely the introduction of the new discourse participant un soldado, is the same in (25a) as in (25b). Even though the grammatical structure of the two sound strings is different, their communicative function is identical. This means that the principle of reference (11) constrains the principle of transparency (26). In fact, we are not aware of any reanalysis based on the principle of transparency which violates the principle of reference. Hence, of the two principles, the principle of reference (11) is the most important one. It can be expected to underlie any kind of reanalysis in grammar (Detges & Waltereit 2002). Whereas reanalyses that rely exclusively on the principle of reference are based on high token frequency, reanalyses of the type sketched in (25a/b) are favored by low token frequency. They are typically brought about by speakers who are unsure as to the conventionally established grammatical structure of the construction. This is the case with constructions which are either new to the language or which have a relatively low frequency. 7 Now we can also explain the otherwise mysterious fact that the reanalysis of presentational haber + NP is strongest in non-present tense uses (Bentivoglio & Sedano 1989:72). The reason is not the defective morphology of haber / hay. In spoken language, presentational constructions are most frequently used in the present tense. Hence, of all the forms of haber + NP, the irregular present tense hay + NP is the most solidly entrenched one. Reanalyses based on low frequency will therefore more likely occur in non-present tenses. Bentivoglio & Sedano (1989:72) show that the tense most affected by this reanalysis is indeed the imperfecto, which is the least frequent of the two Spanish past tenses (Berschin, Fernández-Sevilla & Felixberger 1987:214). Recall that according to Longobardi (2001), all syntactic change is caused, i.e., driven by other components of grammar. Now, the change under discussion here is clearly determined by the conflicting constraints described in (23). But what is syntactic about this change? Crucially, the constraints (23)-1 and (23)-2 have a double status. On the one hand, the coding of argument configurations is idiosyncratic to individual verbs and, as such, to lexical items. For example, I think that and it seems to me that are projections of largely identical semantic arguments to different syntactic constructions. The choice of the construction is specific to the individual verb (seem and think, respectively). Hence, it could be argued that argument projections of this type 7 Strictly speaking, newness is a special case of low frequency, given that a construction which is new necessarily starts out from zero frequency. This is an argument against Koch’s (2002:94, note 77) critique of the link between low token frequency and reanalyses based on the principle of transparency exposed in Detges (2001:419-20).

28

RICHARD WALTEREIT & ULRICH DETGES

must be specified in the lexical entry of the verb. If this line of reasoning is correct, then the constraints in (23) can be observed in the organization of the lexicon and are therefore lexical, not syntactic, in nature. On the other hand, constraints such as (23)-1 and (23)-2 describe a property of a language’s syntax. Basically, they belong to the inventory of a language’s linking rules. More specifically, they describe the unmarked case, leaving aside special syntactic constructions such as passive, special topicalization and focus constructions, which have to be dealt with in the grammar of a language. This may be the reason why in the Chomskyan paradigm rules such as (23)-1 and (23)-2 are normally considered objects of the syntax, not of the lexicon. 4. Conclusion We hope to have shown that a usage-based approach to syntactic change can provide a convincing solution to the apparent paradoxes of grammatical change. Reanalyses are not triggered by properties of the language system. Rather, they are conditioned by frequency of usage. High token frequency of a construction will lead to its entrenchment and to a reanalysis according to the principle of reference. Low frequency can entail reanalysis according to the principle of transparency. Moreover, there seem to be certain types of constructions, e.g. presentational constructions, which seem to favor syntactic reanalysis because they are subject to conflicting coding constraints. Syntax is not inert, not even in the weak sense proposed by Inertia Theory. Syntax is exposed to possible change to the same degree as other modules of grammar are. Note that the mechanisms of syntactic change are not specific to syntax. The principle of reference and the principle of transparency apply to any kind of functional change in language, that is to syntax as well as to the lexicon. Neither is historical stability anything specific to syntax. A language’s syntax is relatively stable because linguistic conventions, be they syntactic or other, tend to self-perpetuate. Furthermore, as we have shown, syntactic change can be triggered from within syntax itself as well as from other domains of grammar. In principle, this claim is not incompatible with the generative viewpoint underlying Longobardi’s (2001) theory that syntax is the ‘most central’, the ‘core’ domain of grammar in the synchronic perspective. But neither does it offer any new evidence in favor of that viewpoint.

REFERENCES Secondary Literature Bentivoglio, Paola & Mercedes Sedano. 1989. “Haber: ¿Un verbo impersonal? Un estudio sobre el español de Caracas”. Estudios sobre español de América y lingüística afroamericana. Ponencias presentadas en el 45

SYNTACTIC CHANGE FROM WITHIN AND FROM WITHOUT SYNTAX

29

congreso internacional de americanistas (Bogotá, julio de 1985), 58-81. Bogotá: Instituto Caro y Cuervo. Berschin, Helmut, Julio Fernández-Sevilla & Josef Felixberger. 1987. Die spanische Sprache. Verbreitung, Geschichte, Struktur. München: Hueber. Bybee, Joan. 2001. Phonology and Language Use. Cambridge, UK: Cambridge University Press. Coseriu, Eugenio. 1978 [1957]. Sincronia, diacronia e história. El problema del cambio lingüístico. Madrid: Gredos. Detges, Ulrich. 2001. Grammatikalisierung. Eine kognitiv-pragmatische Theorie. Habilitation thesis University of Tübingen. ---------- & Richard Waltereit. 2002. “Grammaticalization vs. Reanalysis: A Semantic-Pragmatic Account of Functional Change in Grammar”. Zeitschrift für Sprachwissenschaft 21. 151-195. Fontanella de Weinberg, Maria Beatriz. 1992. El español de América. Madrid: MAPFRE. Foulet, Lucien. 1921. “Comment ont évolué les formes de l’interrogation”. Romania 47. 243-348. Givón, Talmy. 1979. On Understanding Grammar. New York: Academic Press. ----------. 2001. Syntax: an introduction. Vol. 1. Amsterdam & Philadelphia: John Benjamins. Hansen, Maj-Britt Mosegaard. 2001. “Syntax in Interaction. Form and Function of yes/no Interrogatives in Spoken Standard French”. Studies in Language 25. 463-520. Kaiser, Egbert. 1980. Strukturen der Frage im Französischen. Tübingen: Narr. Koch, Peter. 2002. “‘Il ne me faut plus nule rien’ – changement sémantique, métataxe et réanalyse”. Syntaxe et sémantique 4. 67-108. ----------. 2003. “From Subject to Object and from Object to Subject: (De)personalization, Floating and Reanalysis in Presentative Verbs”. Romance Objects, ed. by Giuliana Fiorentino, 153-185. Berlin: Mouton de Gruyter. Lambrecht, Knud. 1987. “On the Status of SVO Sentences in French Discourse”. Coherence and Grounding in Discourse, ed. by Russel S. Tomlin, 217-261. Amsterdam and Philadelphia: John Benjamins. Langacker, Ronald W. 1987. Foundations of Cognitive Grammar. Vol.1: Theoretical prerequisites. Stanford: University Press. Lazard, Gilbert. 1994. “L’actant H. Sujet ou objet?”. Bulletin de la Société de linguistique de Paris 89/1. 1-28. Lehmann, Christian (1985). “Grammaticalization: Synchronic Variation and Diachronic Change”. Lingua e stile 20. 303-318. Lightfoot, David. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press. ----------. 1999. The Development of Language: Acquisition, Change, and Evolution. Malden, Mass.: Blackwell.

30

RICHARD WALTEREIT & ULRICH DETGES

Longobardi, Giuseppe. 2001. “Formal Syntax, Diachronic Minimalism, and Etymology: The History of French chez”. Linguistic Inquiry 32. 275-302. ----------. 2003. “On Parameters and Parameter Theory”. Syntaxtheorien. Modelle, Methoden, Motive, ed. by Elisabeth Stark & Ulrich Wandruszka, 273-90. Tübingen: Narr. Montes Giraldo, José. 1982. “Sobre el sintagma haber + substantivo”. Thesaurus 37. 383-5. Penny, Ralph. 1991. A History of the Spanish Language. Cambridge: Cambridge University Press. Roberts, Ian. 1993. Verbs and Diachronic Syntax: A Comparative History of English and French. Dordrecht: Kluwer. Schulze, Alfred. 1888. Der altfranzösische direkte Fragesatz. Ein Beitrag zur Syntax des Französischen. Leipzig: Hirzel. Seelbach, Dieter. 1983. “Fokussierung mit der est-ce que-Frage”. Grammatik, Konversation, Interaktion. Beiträge zum Romanistentag 1983, ed. by Elisabeth Gülich & Thomas Kotschi, 277-312. Tübingen: Niemeyer. Söll, Ludwig. 1971. “Der neufranzösische direkte Fragesatz in einem Corpus der Kindersprache”. Sprache und Geschichte. Festschrift für Harri Meier zum 65. Geburtstag, ed. by Eugenio Coseriu & Wolf-Dieter Stempel, 493506. München: Fink. Weinrich, Harald. 1982. Textgrammatik der französischen Sprache. Stuttgart: Klett. Old French Texts LANCELCB = Chretien de Troyes. Le Chevalier de la Charrette (1177-81). Base de Français Médiéval. CNNCB = Anonymous. Les Cent Nouvelles Nouvelles (1456-1467). Base de Français Médiéval. MEUGL2 = Jean de Meun. Le Roman de la Rose (fin) (1269-78). Base de Français Médiéval. PATHELCB = Anonymous. La Farce de Maistre Pierre Pathelin (1456-1469). Base de Français Médiéval. Base de Français Médiéval, IXe au XVIe siècle. Rassemblée par l’équipe ‘Linguistique et Informatique’ sous la direction de Christiane MarchelloNizia et Benoît Habert. CD gravé le 21/01/2002.

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH* ANDREAS DUFTER University of Munich

In Contemporary French, c’est-clefts are claimed to occur with significantly higher frequency than their counterparts in other Romance languages and in older stages of French. Starting out from the assumption that c’est-clefts exist in order to mark focus on the clefted constituent, historical linguists commonly seek to explain the observed increase from Old to Modern French as resulting from the decline of alternative focusing strategies. In particular, the loss of flexible focus accents and the severe restrictions on non-canonical constituent orders are generally held responsible for the rise of clefting. This contribution puts standard explanations to the test of corpora and argues that they fail to account for a number of observations. Finally, it proposes that a more comprehensive account of the evolution of c’est-clefts needs to take into account not only phonological and syntactic change, but also an independent pragmatic innovation, viz. the emergence of the informative-presupposition cleft type as defined by Prince (1978).

1. Introduction The term cleft sentence was coined by Jespersen (1937) to refer to English biclausal structures such as (1a), which intuitively qualify as variants of monoclausal sentences, at least with respect to truth conditions. Subsequently, the notion of clefting has been extended to other kinds of complex sentences which appear to be the result of ‘splitting up’ a simple clause in two. In particular, sentences as in (1b) and (1c), called wh- or pseudo-clefts, have been added to the set of systematically available syntactic variants and are often analyzed on a par with so-called it-clefts as in (1a). * Most parts of the reasoning put forward in this contribution have been presented at talks in Berlin (Free University and ZAS), Munich, Paris, Regensburg, and Saarbrücken. From all of these audiences, I received insightful comments and suggestions. In particular, Christophe Benzitoun, Knud Lambrecht, and Morgane Sénéchal provided invaluable advice. Probably none of them, however, would subscribe to all of the views expressed here. In addition, I am indebted to an anonymous reviewer for commenting on an earlier version of this article. I regret that I was unable to consult Sophie Scappini’s Aix-en-Provence dissertation on c’estclefts in Contemporary spoken French at the time of writing. All remaining errors are, of course, mea culpa.

32

(1) a. b. c. d.

ANDREAS DUFTER

It was a weekend in Paris (that) he won What he won was a weekend in Paris A weekend in Paris was what he won He won a weekend in Paris

While earlier transformational grammar explored various kinds of derivational relationships between the sentences in (1), more recent work has generally acknowledged that each of them is subject to specific distributional restrictions, due to different syntactic and information structural constraints. In addition, some researchers propose to analyze cleft sentences as inherently non-compositional constructions in the sense of Construction Grammar (Kay & Fillmore 1999). Therefore, clefts may be subject to language-specific conditions on use and interpretation (Davidse 2000, Lambrecht 2001, Pavey 2004; but see Clech-Darbon, Rebuschi & Rialland 1999 for a defense of compositional analyses with regard to French clefts). Since Jespersen’s seminal works, typological studies have identified numerous counterparts to the English cleft sentences in (1a,b,c) crosslinguistically (Harris-Delisle 1978, Di Tullio 2006:487). Nonetheless, it-clefts have also been characterized as a syntactic strategy which is most typical of European languages (Löfstedt 1966:276-277). Within Europe, the availability of it-clefts seems to decrease from west to east (Miller 1996). In particular, these clefts are reported to occur extensively in Modern Irish, English, French and Portuguese (Wehr 2005). Therefore, they have been interpreted as evidence for an Atlantic sprachbund within Western Europe, possibly ascribable to shared Celtic substrate (Sornicola 1991). More specifically, the commonness of it-clefts has repeatedly been considered one of the most salient structural characteristics of Modern French, and was already highlighted by foreign language instructors in the early nineteenth century (see references given in Metzeltin 1989:195 and Léard 1992:25). Nonetheless, historical French scholarship is unanimous in stating that clefts do not constitute a prominent feature of older stages of the language. While clefting is attested since the Old French period, and indeed since pre-classical Latin, it is claimed to occur less frequently and to be less grammaticalized at those times than in the contemporary usage. In order to explain the rise of itclefts in French, diachronic studies generally take it for granted that clefts are employed in order to mark focus on the clefted constituent (hence on a weekend in Paris in the case of (1a,b,c)), and argue that this syntactic focusing device increased in frequency because alternative prosodic and syntactic focus markings became more restricted over time. First, the fundamental rhythmic change reconstructable for Middle and Renaissance French implies a drastic reduction of in situ constituent foci signaled by focal accents. Second, clauseinternal constituent order variation, which could be exploited to mark focus on a constituent by positioning it ex situ, has become more and more heavily

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

33

constrained since later Old French. By the Renaissance French period, the vast majority of lexical arguments appear in canonical SVO order. Surprisingly enough, however, the alleged interdependence between these changes and the growing frequency of clefts has never been put to the test by systematic diachronic corpus research. This article will investigate variation and change in French it-cleft homologues headed by ce “this”, henceforth referred to as c’est-clefts for short.1 In particular, it intends to provide a more solid empirical basis for assuming a diachronic increase of c’est-clefts and a division of labor between clause-internal foci and clefting. In the next section, the defining characteristics of c’est-clefts will be described in some more detail. Section 3 surveys the distribution of c’est-clefts in Contemporary French, comparing written and spoken usages. In the light of these synchronic findings, some claims about the explanatory potential of prosodic change for the evolution of French clefts will be re-assessed. Diachronic corpora are examined in section 4 in order to evaluate standard assumptions about the impact of rigid constituent order on clefting. By way of conclusion, I will argue in section 5 that the emergence of cleft sentences containing new information in the embedded clause, designated informative-presupposition clefts by Prince (1978), is a major factor contributing to the increase of c’est-clefts since the Renaissance French period. 2. The grammar of c’est-clefts in French The different subparts of cleft sentences are referred to by various names in the literature. In (2), I give the terminologies used by Hedberg (2000) and Lambrecht (2001), and introduce my own shorthand labels. These are intended to be maximally theory-neutral and will be used throughout in the remainder of the text. (2) Syntactic terminology for it-cleft sentences

Lambrecht (2001) Hedberg (2000) This paper

It C’ matrix subj. cleft pron. CE

was était copula copula EST

a weekend in Paris un weekend à Paris focus phrase clefted constituent X

(that) he won qu’ il a gagné relative clause cleft clause QU IPQU CPQU

According to Harris-Delisle (1978), cross-linguistic variation in the syntax of cleft sentences typically reflects syntactic differences found in the grammar 1 Note that in addition to it-cleft homologues headed by ce, Contemporary French permits elements such as voici “here is”, voilà “there is”, il y a “there is” or j’ai “I have” to introduce sentences that qualify as clefts syntactically and semantically (see the definition cited in section 2). For a more comprehensive treatment of clefting in Modern French, especially in its spoken varieties, the reader is referred to Lambrecht (1994).

ANDREAS DUFTER

34

of identificational clauses, i.e. sentences such as It was a weekend in Paris, and in possible strategies of relativization. Not surprisingly, therefore, the same element CE that functions as an obligatory subject in French identificational clauses is also required in cleft sentences. In an opinion sanctified by tradition and still taken for granted in recent handbooks (Di Tullio 2006:484), CE-type function words in clefts are merely expletive subjects, devoid of any referential properties and therefore systematically absent in null-subject languages. While this may be true for Italian and Spanish, Hedberg (2000) points out that there are other pro-drop languages such as Russian where a ‘spurious’ CE element is significantly more likely to occur in clefts than as an argument of raising or weather predicates. Likewise, Old French authors regularly omit noncontrastive subject pronouns, but use CE in clefts almost without exception (Jochimsen 1907:22). For this reason, the term c’est-cleft seems a reasonable choice to designate sentences corresponding to English it-clefts for all periods of French. The second compulsory element in French clefts is the copula EST. Starting in Middle French, agreement in person between X and EST was gradually lost, and third person verb forms were generalized. This change, completed by the sixteenth century (Hatcher 1948), sets Modern French apart from Old French, and from other Romance languages, as can be seen in the examples given in (3). (3) a.

b.

c.

Ce estes vous que je doy remercier it be.2PL you.PL that I must.1SG thank [Antoine de la Sale (1456); BFM1] “It is you (PL / SG.POLITE) whom I have to thank.” C’ (*êtes / est) vous que je dois remercier it (*be.2PL / be.3SG) you.PL that I must.1SG thank “It is you (PL / SG.POLITE) whom I have to thank.” (Sois / *Es) vosotros a los to who.PL (be.2PL / *be.3SG) you.PL que tengo que agradecer that have.1SG that thank “It is you (PL) whom I have to thank.”

(Middle French)

(Modern French)

(Peninsular Spanish)

Another characteristic of Modern French, and in particular contemporary spoken varieties, is the ongoing spread of singular forms of EST with plural clefted constituents X (cf. (4a,b)). By contrast, number agreement between EST and X remains obligatory in other Romance languages, as can be seen from (4c).

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

(4) a.

b. c.

C’ est eux qui choisissent it be.3SG they who choose.3PL [COR fpubmn02] Ce sont eux qui choisissent it be.3PL they who choose.3PL (Son / *Es) ellos los que eligen (be.3PL / *be.3SG) they who.PL that choose.3PL “It is them who choose.”

35

(Spoken French)

(Spanish)

This overall trend toward invariable EST in French c’est-clefts is further enhanced by restrictions on tense and mood. In many cases, the present indicative is employed as a default option for EST in Contemporary French, irrespective of the tense of IPQU, and periphrastic tenses such as the passé composé in (5b) are marginal at best. (5) a.

b.

C’ est elle qui a choisi la couleur it is she who has chosen the color [FRANTEXT 1993 Dormann] ??Ça a été elle qui a choisi la couleur it has been she who has chosen the color “It’s been her who’s chosen the color.”

Notice also that the embedded clause CPQU in c’est-clefts is obligatorily introduced by a relative pronoun QU. At this point, one could argue that French cleft sentences are more advanced in their degree of grammaticalization than their counterparts in languages such as English and Portuguese. In both of these languages, there is variation between clefts containing a QU element and “that-less clefts” (Ambar 2005:97; cf. also example (1a) above). With respect to the choices available for QU, it has sometimes been assumed (cf. Krötsch & Sabban 1990:94, Léard 1992:30) that Contemporary French c’est-clefts only permit two relative pronouns qui “who, what” and que “that”, whereas all other relativizers (dont “of whom, of what”, où “where” and the inflected lequel “which”) cannot be employed in this particular syntactic context. Of course, such a restriction (and the consequent generalization of invariable que to all cleft types except subject clefts) would constitute a further piece of evidence for an overarching trend towards structural invariance of c’est-clefts in French. More recently, however, Blanche-Benveniste (2000) and Muller (2003) have contested these restrictions on empirical grounds. Therefore, the present study will in principle consider all instantiations of QU to be possible. In Contemporary French, the order of the obligatory subparts in c’est-clefts is quite rigid. CE and EST invariably precede QU, which in turn invariably heads the CP. Only CE and EST may be inverted, and EST CE has become fossilized in the interrogative particle est-ce que (literally “is it [the case] that”; see Foulet 1921 and Rouquier 2003 for detailed diachronic description). In Modern French declarative sentences, however, EST CE orderings are only

ANDREAS DUFTER

36

licensed by a small number of sentence-initial adverbs such as peut-être “perhaps”, and have become quite rare. In our corpus study, we basically adopt the definition of clefts formulated by Lambrecht, without thereby committing ourselves to any theoretical viewpoint about the nature of constructions in grammar: A CLEFT CONSTRUCTION (CC) is a complex sentence structure consisting of a matrix clause headed by a copula and a relative or relative-like clause whose relativized argument is coindexed with the predicative argument of the copula. Taken together, the matrix and the relative express a logically simple proposition, which can also be expressed in the form of a single clause without a change in truth conditions. (Lambrecht 2001:467)

Notice that this definition relies on syntactic and truth-conditional semantic aspects of sentence structure, but makes no reference to either prosody or pragmatics. In this approach, which amounts to analyzing clefts as structurally complex variants of monoclausal sentences, there are at least two subtypes of syntactic structures which in our opinion could be categorized as c’est-clefts, while Lambrecht either does not take them into consideration or explicitly excludes them from the class of cleft sentences. A first subcase which is traditionally included into the class of c’est-clefts, but strictly speaking not covered by the above definition, is clefting of copula complements. As can be seen in (6a), in this structure, the post-QU subpart l’abus du tems obligatorily surfaces without a verb form and thus does not constitute a “relative-like clause”. Nonetheless, it would not seem unreasonable to assume an elliptical copula in CP, and thus underlyingly a biclausal syntactic format. In any event, (6a) has to be considered a marked syntactic variant of (6b). (6) a.

b.

C’ est un grand mal que l’ abus du tems (*est) it’ s a big evil that the abuse of.the time (*is) [FRANTEXT 1750 Rousseau] L’ abus du tems est un grand mal the abuse of.the time is a big evil “Wasting time is a big evil.”

A second clarification is in order for the class of sentences exemplified in (7). Following much recent work (see Di Tullio 2006:488-489 and references cited there; see Lambrecht 2001:503-504 for a different view) we interpret (7a) below as a borderline case of a c’est-cleft, with empty X. In the literature, the structures illustrated in (7) have come to be known as “inferential sentences” (see Delahunty 1995; for inferentials in Romance, see Pusch 2006). Notice that, like their English counterparts, Modern French inferential clefts permit negative and some other sentential adverbs between EST and QU, as in (7b,c).

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

(7) a. b. c.

37

Inferential c’est-sentences C’est qu’il était devenu vieux, subitement “It’s that he had become old, suddenly.” [FRANTEXT 1993 Dormann] Euh c’est pas que j’aime pas “Oh it’s not that I don’t like [it].” [COR ffamdl01] C’est surtout que j’avais pas fait de gaffe “It’s mostly because I didn’t pay attention.” [COR ftelpv06]

However, not all CE EST CPQU sequences qualify as inferentials. In cases such as those exemplified by (8), CE is clearly anaphoric, and relates back to a left-dislocated subject (8a) or a free relative within a pseudo-cleft structure (8b). In an inferential c’est-cleft, by contrast, no anaphoric relationship is established with the preceding context. (8) a. b.

Non-inferential c’est que La preuve, c’est que je suis mort lit.: “The proof, it’s that I’m dead.” [FRANTEXT 1993 d’Ormesson] Car ce qui était important, c’est qu’il allait vite “Because what was important was that he went fast.” [FRANTEXT 1993 Roubaud]

In line with traditional French descriptive grammar (cf. Grevisse 1993:697), I decided to take into consideration both inferential sentences and sentences involving fronted copula complements in my corpus searches for c’est-clefts. The examples given in (9) illustrate a subtype of CE EST X CPQU sequences which is generally excluded from consideration in descriptive accounts of Modern French c’est-clefts. In all the sentences in (9), the pronoun ce “this / that” occurs at the right edge of X. Lacking a full vowel, ce can never receive prosodic prominence in Modern French. In fact, in the most natural pronunciation, its shwa is regularly dropped and shorter sentences such as (9a) are most likely to exhibit no internal division into phonological phrases. In a natural pronunciation of (9b), the first phonological phrase will end after voit or even after souvent. Such a prosodic structure is untypical for c’est-clefts, since the right edge of X tends to coincide with the end of a phonological phrase (often somewhat misguidingly referred to as “mot phonétique” [phonetic word] in the French tradition). Recall, however, that Lambrecht’s definition of cleft sentences does not make reference to prosodic aspects of sentence structure, so that we cannot exclude the sentences in (9) from consideration on phonological grounds alone. A second argument against classifying CE EST ce CPQU sentences as clefts is that, strictly speaking, (9a,b) do not conform to the definition of clefts as complex variants of monoclausal sentences, since the non-clefted counterparts are ungrammatical with verbs other than être “to be”. However, a minor “lexical adjustment” (Lambrecht 2001:499), viz. the substitution of non-clitic

ANDREAS DUFTER

38

cela (or ceci) for clitic ce, permits to convert (9a,b) into the perfectly grammatical sentences (10a,b). It is important to note that similar alternations between clitic and non-clitic allomorphs are also necessary to uncleft a sentence like (11a), cited as an example of a c’est-cleft by an authoritative French grammar. (9) a. b. c.

(10) a. b. c. (11) a.

b.

C’est-cleft candidates with X = ce in Contemporary French En plus c’est ce qui complique les choses “Moreover, that’s what makes things complicated.” [COR ftelpv26] C’est ce qu’on voit le plus souvent quand même maintenant “Still, that’s what you see most often nowadays.” [COR fpubdl04] C’est ce qui est écrit dans les livres “That’s what is written in the books.” [FRANTEXT Claudel] Non-clefted variants of (9a,b,c) En plus (*ce / cela) complique les choses On voit (*ce / cela) le plus souvent quand même maintenant C’est écrit dans les livres Lexical adjustment in c’est-clefts C’est (moi / *je) qui le ferai it’s (me / *I) who it do.FUT.1SG “I will do it.” (Grevisse 1993:696) (*Moi / Je) le ferai (*me / I) it do.FUT.1SG “I will do it.”

It may well be true that in the majority of cases, CE EST ce CPQU sentences are identificational, with anaphoric sentence-initial CE. However, I am inclined to be sceptical about excluding sentences as in (9) from the outset, since at least some occurrences in our corpus data do appear to constitute marked variants of simple predicational clauses. Moreover, as mentioned by Lerch (1934:227), there are clear cases of c’est-clefts with X = ce in older stages of the language examined. In view of this situation, I decided to include cases which in my interpretation involve clefted ce, but to count these cases separately in the corpus searches.2 The findings will be presented in the following two sections.

2 Limitations of space preclude further discussion of the difficulties in distinguishing between ‘genuine’ c’est-clefts and “cleft look-alikes” (Katz 2000; see Dufter in prep. for more on this issue). Those who have worked on historical syntax will know that grammatical concepts which rely on a normative definition of ‘canonical’ sentence structures cannot always straightforwardly be extended to diachronic description. Highly instructive in this respect is Pagani-Naudet’s (2005) discussion of identifying dislocation structures in older stages of French.

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

39

3. On the alleged trade-off between clefting and focus in situ 3.1 Clefts in speech and in writing In variationist studies of French, and in Romance linguistics in general, not much systematic attention has been paid to the relative importance of clefts in speech and in writing. Simplifying somewhat, three mutually incompatible assumptions can be found in the literature. According to the first group of scholars, clefting in French does not display variation between spoken and written varieties (Koch & Oesterreicher 1990:96), since c’est-clefts are frequent in both of them (Katz 2000:253-254, Schöpp 2002:48). The second view is maintained by linguists such as Lerch (1934:229), who notes that c’estclefts are characteristic of spoken French. Third, several more recent studies on Romance, among them Metzeltin (1989), Sedano (1990), or Gil (2003), have endorsed just the opposite claim. According to these authors, it is the spoken varieties of Romance where clefting is less common. Similar hypotheses appear to have gained widespread acceptance beyond the field of Romance studies. With respect to English, Declerck (1984:270) states that clefts are “much more frequent in written language than in speech” (see also Collins 1991:ch. 7 and Doherty 2001:460). Typically, the proponents of the third view seek to explain the alleged written language bias of clefts by pointing out that — unless using special typographic provisions — accentual focus marking can be much less reliably encoded on paper than acoustically. Following this line of argumentation, the recourse to clefts is motivated in writing only as a “compensatory device” (Doherty 2001:457, Di Tullio 2006:487), and avoided whenever alternative focus marking strategies are available. The reasoning behind this appears straightforward: By virtue of their biclausal structure, clefts are more ‘costly’ than all indications of focus which preserve canonical word orders, relying exclusively on prosody. In a cross-linguistic perspective, it has even been suggested that prosody is the most basic and potentially a universal exponent of focus structure (see Roberts 1998:147).3 In any event, clefts are predicted to appear in writing at least as often as in speech. In order to evaluate these claims, a solid empirical basis, ideally from a representative sample of languages, would be needed. Unfortunately, no such cross-linguistic set of data is available at present. For English at least, authors such as Collins (1991) provide information about distributions and token frequencies of clefts in written and spoken sources. The figures that are crucial for the present context are cited in Table 1.

3

Notice, however, that in her typological survey of contrastive focus marking, Harris-Delisle (1978:421) reaches just the opposite conclusion. All languages in her sample possess syntactic means to express contrast, while not all languages turn out to have clear phonological cues.

40

ANDREAS DUFTER

# It-clefts / 10,000 words # Wh-clefts / 10,000 words

Writing Speech 5.7 4.3 4.1 14.3

Table 1: Frequencies of clefts in English [LOB and LL corpora] (Collins 1991:179-180)

For it-clefts (cf. (1a) above), the predicted higher incidence in writing is thus borne out, whereas for wh-clefts (cf. (1b,c)), just the opposite trend is observed. Comparing different genres of written English texts, Collins (1991:214-215) observes that it-clefts prevail in “opinionative” text types and in formal registers. While it may not be straightforward to account for the observed affinity of clefting and text type, one might attempt to relate the register correlation to the higher syntactic complexity of clefts. However, whclefts, especially of the ‘reverse’ type, with the embedded clause following the copula (cf. (1c)), occur predominantly in informal registers of spoken English. On the whole, then, the relationship between clefting and medium or register is less clear than one might have expected. In a follow-up study, Nelson (1997) investigates clefts in the International Corpus of English (ICE). His results fail to confirm the predicted written language preference even for it-clefts. For European Spanish, Gómez-González & Gonzálvez-García (2005) were unable to detect significant differences in clefting frequencies between written and spoken varieties. Within written Spanish, clefts are more likely to be employed in narratives than in academic or journalistic texts — just the reverse trend as the one stated by Collins for it-clefts in English. In general, it is not obvious how the postulated correlations with text and discourse type could be explained by reference to focus marking alone. Furthermore, the findings presented so far suggest that statistical results obtained for it-cleft sentences in one language may not carry over to other languages. The present investigation of c’est-clefts in Contemporary French is based on two corpora: For spoken French, I selected the French part of the C-ORALROM corpus (COR for short, cf. Cresti & Moneglia (eds.) 2005). The 206 French transcripts in COR add up to almost 270,000 words and cover more than 26 hours of speech. Moreover, the corpus represents a wide range of formal as well as informal situation types. For written French, I consulted the enormous version non-catégorisée of FRANTEXT and randomly selected one year, 1993, thereby obtaining a 540,000 word subcorpus.4 All texts in this sample are literary in nature, as are about 80 % of the entire database. It is clear that the absence of non-literary text types severely limits the representativeness of the corpus for written Contemporary French in general. However, the subcorpus does establish a reasonably good basis for the diachronic com4

The corpus is made up of the following six literary works: Geneviève Dormann, La petite main, Annie Ernaux, Journal du dehors, Jean d’Ormesson, La douane de mer, Eric Orsenna, Grand amour, Jacqueline de Romilly, Les Œufs de Pâques, and Philippe Sollers, Le secret.

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

41

parisons presented in the next section. The strategy employed for searching Contemporary French declarative c’est-clefts is based on the description of section 2 and results in the query given in (12). Notice that in order to construct a regular search expression, some upper bound had to be imposed on the word length of clefted constituents X under consideration. (12)

Search pattern for Contemporary French c’est-clefts Search all sequences CE (N’) EST X QU where CE: ce or c’, (N’): n’, ne or the empty string, EST: any synthetic 3rd person singular form of être, X: any (possibly empty) string containing up to 6 words, QU: any relative pronoun.

Of course, queries for the regular expression defined in (12) will not yield all and only the wanted c’est-clefts in the corpora. On the one hand, the majority of the search results in COR and FRANTEXT 1993 are sentence tokens which present CE, EST, and QU elements in the same linear order as genuine c’est-clefts, but lack a semantically equivalent monoclausal variant. Therefore, I had to check all query results individually, taking into account their context of occurrence, to sort out the cleft look-alikes, i.e. identificational sentences as those in (13).5 (13) a. b.

Sentences formally identical to c’est-clefts, but lacking monoclausal variants C’est une des premières choses que les garçons remarquent “This is one of the first things that the boys notice.” [COR fmedts05] Voilà ça c’est quelque chose… qui … qui a été pérennisé “Look that, that is something … which … which has become permanent.” [COR ffammn02]

On the other hand, the formula in (12) will not detect declarative c’estclefts with inversions of EST and CE, and, more importantly, clefts with X exceeding six words in length. For a rough estimation of the proportion of longer clefted constituents, I conducted an additional search, retrieving all sequences CE EST in FRANTEXT 1993 and collecting the first one hundred tokens of c’est-clefts. Of these, only eight turned out to exceed the upper bound 5

Many studies postulate systematic prosodic differences between French c’est-clefts and cleft look-alikes like those in (13). In particular, c’est-clefts are claimed to have different intonation contours (Krötsch & Sabban 1990:90), and the pitch accent on the clefted constituent X would be realized “usually stronger than the one that is typically found at the end of a rhythmic group” (Katz 2000:254). Interestingly, however, this standard assumption has been challenged by more recent experimental work (Clech-Darbon, Rebuschi & Rialland 1999). These authors failed to detect any robust prosodic correlate specific to c’est-clefts. However, since the present contribution exclusively relies on transcripts and on written texts, I could not have used prosodic cues for sorting out cleft look-alikes anyway.

42

ANDREAS DUFTER

chosen in (12). While the frequencies of c’est-clefts in our subsets will thus not reflect overall frequencies totally faithfully,6 it seems plausible to assume that an exhaustive search for c’est-clefts would not yield significantly different results. On the whole, then, the subsets of c’est-clefts obtained by the query in (12) should furnish at least an approximate comparison. Their frequencies are presented in Table 2.

Total # of non-inferential c’est-clefts, X ≠ [… ce] Total # of inferential c’est-clefts Total # of c’est-clefts, X = [… ce] Total # of words Non-inferential c’est-clefts, X ≠ [… ce] / 10,000 words ALL c’est-clefts / 10,000 words

Writing Speech 300 376 36 29 49 86 541,474 269,959 5.5 13.9 7.1 18.2

Table 2: Frequencies of declarative c’est-clefts with |X| ≤ 6 in Contemporary French [FRANTEXT 1993; COR]

The figures in Table 2 constitute solid evidence against interpreting c’estclefts in written French as a substitute for focus accents. Even if focus accents were completely absent from Contemporary spoken French, one should expect token frequencies to be roughly the same for both mediums. The fact that the c’est-clefts occur about 2.5 times more often in speech than in writing remains inexplicable under any hypothesis which holds that clefting is nothing more than a substitute for in situ accentual focus marking. 3.2 Implications for diachrony Evidence for prosodic change from Middle to Modern French comes from a variety of sources. Phenomena from morphosyntax, especially new restrictions on pronoun placement, metalinguistic comments found in pronunciation handbooks for foreigners, as well as developments in versification clearly indicate that the prosodic autonomy of the individual word forms was gradually declining. Conversely, phrasal phonological units above the word-level, demarcated by domain-final lengthening, became more and more important, not only as domains of sandhi phenomena, but also as the basic units of rhythmic structure (cf. Dufter 2003:137-146). While word stress had played a central role in the evolution from Vulgar Latin to Gallo-Romance, and is considered a salient prosodic feature of Old and Middle French, its phonetic exponence has subsequently weakened and arguably been lost in all 6 To the extent that written texts are syntactically more complex than spoken utterances, we may expect average constituent lengths in writing to exceed those found in speech. Therefore, it is unlikely that the proportion of X constituents with more than six words in C-ORAL-ROM outnumbers that in FRANTEXT. I am grateful to Christophe Benzitoun for pointing out this issue to me.

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

43

positions except at the end of phonological phrases. The sixteenth century, in particular, has been judged crucial for this transition from word-based to phrase-based rhythm (Kuryłowicz 1945). Traditional historical grammars (Lerch 1934:229; Gamillscheg 1957:563) as well as more recent investigations (Schöpp 2003:50; Wehr 2005:368-370) consider this prosodic change primordial for the increase of clefting in French. The following citation is representative of the general line of argumentation: The frequency of the cleft construction in French is doubtlessly connected with the loss of the single word accent. […] The decisive factor is, in my opinion, the change from an autonomous single word to the mot phonétique, in which the word stress has been lost. […] whereas in Proto-Old French the word accent was probably the strongest among the Proto-Romance languages, as some diphthongisation processes show, it is virtually non-existent in Modern French, which is without parallel in the Romance languages (with the exception of Portuguese […]). (Wehr 2005:369)

Of course, the theoretical premises outlined in the previous subsection can be taken to bear on diachrony as well. More specifically, any theory which takes it that in situ prosodic focusing is ceteris paribus preferred over more ‘costly’ foci ex situ will predict that an increase of the latter marking type is not a natural development. From a derivational perspective, ex situ foci imply movement; in a constraint-based framework, non-canonical (re-)orderings or biclausal (re-)arrangements clearly do not constitute a case of ‘emergence of the unmarked’. All other things being equal, clefts should increase in frequency only if more economic focus markings, and in particular in situ focal accents, become restricted by changes which need to be explained on independent grounds. However, to the extent that the frequencies given in Table 2 are indicative of a general spoken language bias of c’est-clefts in Contemporary French, they provide clear counter-evidence to interpreting clefts as a compensatory device. Turning from synchronic variation to diachrony, the findings reported so far also shed doubt on attempts to explain the diachronic increase of c’est-clefts by referring to prosodic change restricting in situ accentuations and possible focus-accent alignments. In fact, c’est-clefts appear to be employed ‘beyond necessity’ in a significant number of cases, at least in Contemporary spoken French. In the next section, historical corpus data will provide additional evidence against explaining the rise of c’est-clefts as being due exclusively to prosodic and syntactic changes. 4. Syntactic motivations for the rise of c’est-clefts in French 4.1 Discourse-governed constituent order from Latin to Modern French The second major factor which is generally held responsible for the rise of clefts in French is the massive limitation of possible major constituent orders. In the evolution from Old to Modern French, lexical arguments in declarative sentences have come to occur in SVO order with overwhelming frequency.

44

ANDREAS DUFTER

Conversely, preverbal objects (OV) and postverbal subjects (VS) have gradually been restricted to very specific contexts (Marchello-Nizia 1995:69113). Following Lambrecht (1994, 2001), I take predicate focus to be the unmarked case, and both argument and sentence focus to be configurations that are likely to be differentiated, albeit minimally. Since both of the ‘noncanonical’ linearizations of major constituents could be exploited to mark ‘noncanonical’ argument foci (on the fronted object or the inverted subject, respectively), the fixation of SVO may be expected to correlate with an increase in alternative argument focus markings. As noted in the last section, flexible in situ focus accentuation had become heavily constrained by the sixteenth century, so the only generally available option was clefting. In a synchronic perspective, a division of labor between discourse-governed constituent order and clefting is already assumed by Jespersen (1937:85) and explicitly adopted by Lambrecht as a cross-linguistic generalization: PRINCIPLE 1: The occurrence of cleft constructions in a language correlates with the degree of positional freedom of prosodic accents and syntactic constituents in that language. (Lambrecht 2001:488)

In a similar vein, French linguists and grammarians typically seek to explain the lower incidence of c’est-clefts in Old French by pointing out the gamut of linear constituent orders (Jochimsen 1907:110-112, Sornicola 1991:48-49, Grevisse 1993:695, among others). This agreement of scholarly opinion seems convincing enough — indeed almost too convincing, for we are left to wonder why clefts occur at all in Latin or Old French, given that less costly focus markings are always available. In fact, Löfstedt (1966) adduces an impressive range of clefts from a wide variety of Latin sources, and Jochimsen (1907) provides a treasure of examples of Old and Middle French c’est-clefts. To obtain a more precise view of the alleged diachronic increase of this syntactic structure, some quantitative findings will be presented in the next subsection. 4.2 Corpus evidence In order to track back the evolution of c’est-clefts in French, I consulted all 73 texts in the Base de Français Médiéval 1 (BFM1) corpus before 1500 (i.e., all available texts with the sole exception of the Tiers Livre by Rabelais). For the subsequent periods, I arbitrarily selected the years 1550, 1650 and 1750 from the version non-catégorisée of FRANTEXT. For all subcorpora, the search strategy adopted was the same as the one for the Contemporary French investigation described in the previous section. However, due to the wider variety of forms and spellings for each of the canonical subparts CE, EST, and QU, and because of the absence of lemma-based query options in the BFM1, the search expressions were considerably more complicated. In particular, I

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

45

attempted to account for all synthetic finite forms and graphic variants of EST, including first and second person forms. (Recall that the restriction to third person EST in c’est-clefts only became compulsory during the sixteenth century, cf. (3a) above.) In addition, the range of morphological and graphic variants for CE and QU found in medieval texts is also somewhat greater than in Contemporary French. In (14), the alternative instantiations of the criterial subparts of c’est-clefts are presented in some more detail. (14)

Search pattern for Old and Middle French c’est-clefts in BFM1 Search all sequences CE (N’) EST X QU where CE: c’, ce, ço, ceo …, (N’): n’, ne or the empty string, EST: 37 graphic forms of finite estre “be”, X: any (possibly empty) string containing up to 6 words, QU: any form beginning in qu- or k-, ou, dont / dunt / don, cui / coi, all forms of lequel.

As in the corpus study presented in 3.1, this strategy will fail to detect c’est-clefts with clefted constituents longer than six words and inversion of EST and CE. Due to the higher overall flexibility of constituent orders in older stages of French, the proportion of c’est-clefts with inverted EST CE may be expected to be higher than in the contemporary language. Moreover, since Old and Middle French permit non-overt subjects, we have to expect (N’) EST X QU clefts lacking a cleft pronoun. However, as already mentioned in section 2, Jochimsen (1907:22) reports that the great majority of clefts is introduced by CE from early Old French onwards. As with the query results from Contemporary French corpora, each of the almost 3,500 search results had to be checked individually in its context of occurrence. After sorting out the numerous cleft look-alikes, the absolute and relative frequencies of clefts of our BFM1 and FRANTEXT samples were calculated separately for each century.7 The figures are presented in Table 3.8

7

For simplicity, we included c’est-inferentials and cases of clefted ce. Note, however, that we would have observed exactly the same distributional trends if we had excluded one or both of these subtypes from our counts. 8 There are very few clear cases of c’est-clefts in texts before 1100. See Jochimsen (1907:39, 45, and 47) for two examples from Saint Lethgier (second half of the tenth century) and two from the Vie de Saint Alexis (about 1050).

46

ANDREAS DUFTER

Century 12th 13th 14th 15th 16th 17th 18th 20th # C’est-clefts 85 244 24 106 133 68 98 385 # Words 728,570 1,051,984 424,662 585,417 305,047 70,585 81,950 541,474 C’est-clefts / 10,000 words 1.2 2.3 0.6 1.8 4.4 9.6 12.0 7.1 Table 3: Frequencies of c’est-clefts in French literary texts [BFM1 1100-1500; FRANTEXT 1550, 1650, 1750, 1993]

The relative frequencies clearly present evidence for an overall increase of c’est-clefts in the history of French. However, this development is not monotonous, especially in the BFM1 corpus. While most of the texts included in this database are literary in nature, we find both verse and prose works pertaining to a variety of different genres. Given this heterogeneity, one might suspect the unexpected sharp decrease in frequency between the thirteenth and fourteenth centuries to follow from different compositions of the relevant BFM1 subcorpora. In fact, verse predominates in the thirteenth century subcorpus, while the great majority of fourteenth century texts in BFM1 happen to be in prose. If c’est-clefts in Old and Middle French occurred preferably in versified texts, the local deviation from the overall trend towards more clefting might reflect the different proportions of the two literary forms rather than linguistic evolution. The figures in Table 4, however, fail to confirm this hypothesis.

# C’est-clefts # Words # C’est-clefts / 10,000 words

Verse Prose 225 234 1,269,993 1,497,202 1.8 1.6

Table 4: C’est-clefts in Old and Middle French according to literary form

Another factor that should be taken into account is literary genre. I am aware of the difficulties inherent in establishing genres, in particular for medieval texts. Nonetheless, the classification adopted by the creators of the BFM1 seems to me at least an insightful preliminary typology. The three genres that make up for more than 60 % of the entire BFM1 database are romance, chronicle and theater. Calculating the frequencies for c’est-clefts separately for each of these subcorpora yields the results given in Table 5. Romance Chronicle Theater # C’est-clefts 274 22 40 # Words 1,085,528 423,326 249,125 # C’est-clefts / 10,000 words 2.5 0.5 1.6 Table 5: C’est-clefts in Old and Middle French according to genre

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

47

The differences in clefting frequencies are obvious. Moreover, they help us to understand the surprising decrease of c’est-clefts from the thirteenth to the fourteenth century, since the prevailing genre in the fourteenth century subcorpus is chronicle (Froissart, Morée), while in the thirteenth century subcorpus of BFM1 romances have much more weight. Of course, the statistical trends observed in Table 5 are themselves in need of a principled explanation. Moreover, larger subcorpora of FRANTEXT would have to be examined in order to obtain a more fine-grained picture of the evolution of c’est-cleft structures from the Renaissance to the contemporary language. In her descriptive grammar of seventeenth century French, Fournier (1998:135136) maintains that the distribution and frequency of c’est-clefts in the siècle classique do not differ at all from the situation in the contemporary language.9 On the whole, the data in Table 3 lend empirical support to the common assumption of an increase of c’est-clefts from Old to Contemporary French. Upon closer scrutiny, however, attempts to explain this rise of clefting by reference to the decrease of competing focusing strategies lose much of their initial appeal. To begin with, the figures in Table 3 suggest that c’est-clefts continue to increase in frequency even after the prosodic change described in the last subsection. Similarly, the rise of this biclausal structure appears to extend well beyond the period during which rigid SVO was almost categorically established in French. While the generalization of SVO is a complex evolution spanning several centuries, it is clear that by the midsixteenth century, clause-initial objects are only licensed in highly limited discourse contexts (Combettes 2003:175-182, Marchello-Nizia 2005:413-416). At the same time, the proportion of postverbal subjects in declarative sentences has dropped significantly (Prévost 2001). Moreover, an overall increase of clefting has been reported for Spanish as well (Helfrich 2003), a language in which major constituent orderings have remained much more flexible to date. A second observation casting doubt on the clefts-as-compensation view pertains to the solid attestation of c’est-clefts as early as Old French that might seem ‘unwarranted’, since focus marking by clefting could easily have been avoided. For instance, Wehr (2005:362) adduces the sentence in (15a) from the Song of Roland, where it is clear from the context that in the clause introduced by et “and”, there is a contrastive focus on the subject pronoun jo “I”. Since Old French is a null subject language, the presence of an overt pronoun, which in all likelihood carries a focus accent in accordance with the metrical template, suffices to license contrastive inferences. Nonetheless, several examples of clefted pronominal subjects are found in the BFM1 as early as the twelfth century. One of them is cited in (15b). 9

In this respect, clefting differs fundamentally from dislocation, which was subject to massive criticism by prescriptive authorities of the classical period and subsequently came to be avoided in written French (Pagani-Naudet 2005).

ANDREAS DUFTER

48 (15) a.

b.

Old French Dunez m’en, sire, le bastun e le guant, e jo irai al Sarazin espan “Give, Sire, to me the glove, also the wand, I will seek out the Spanish Sarazand.” [Roland 268-269; Wehr 2005:362; trans. Scott Moncrief] Et ce est il qui me requiert “And it is him who wants me.” [BFM1 Chrétien, Yvain 2131]

Third, a further reason to question the assumption of c’est-clefts as a substitute of simpler focusing strategies can be gained from investigating the syntactic function of the clefted constituent X with regard to the embedded clause CPQU. In Table 6, the percentages of the dominating types subject, (direct or indirect) object, adverbial, and inferential (see section 2 above) are compared.

% Subjects % Objects % Adverbials % Inferentials % Others

Old French Middle French 20th c. (12th and 13th c.s) (14th and 15th c.s) Written Spoken 49.8 43.8 42.1 38.5 30.7 23.1 19.0 22.6 16.4 21.5 28.1 32.8 0 0 9.4 5.9 3.0 11.5 1.6 0.2

Table 6: Syntactic functions of clefted constituents X in c’est-clefts

Given that the restrictions on flexible, discourse-governed orderings outlined above crucially affect arguments, one would expect the proportions of clefted subjects and objects to increase over time. The figures in Table 6 show that this prediction is not borne out. Instead, two other developments are observed. First, the ratio of adverbial X in c’est-cleft has augmented over time. Second, inferential constructions are completely absent from Old and Middle French (cf. also Jochimsen 1907:104), whereas they are solidly attested in the written and the spoken corpora of Contemporary French. In the concluding section, a unified account for both of these observations will be offered. 5. On explaining clefting ‘beyond necessity’ For a more adequate description of the evolution of c’est-clefts, we need to look beyond their capacity to put a focus on the clefted constituent X. In particular, it has been pointed out several times (cf. Delin & Oberlander 1995, submitted, among others) that cleft sentences may differ from monoclausal variants by carrying additional presuppositions. While in (16b), contrastiveness on the pronominal subject can be successfully signaled by an appropriate prosodic profile on the strong pronoun, the presupposition that somebody else will inevitably have to pay the bill is only conveyed by the clefted variant (16a). More generally, c’est-clefts may help speakers to avoid focus ambiguities, cf. (17a) and the non-clefted variants in (17b,c). In some of the

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

49

transcripts of spontaneous spoken French, additional motivations for resorting to ‘unnecessary’ c’est-clefts appear to be the ordering “topic before comment”, as in (18), or the optimization of argument structure, by avoiding more than one new lexical argument per clause (cf. (19); see Du Bois 1987, 2003 for the theory of Preferred Argument Structure). (16) a. b. (17) a. b. c. (18)

(19)

C’est pas moi qui vais payer “It’s not me who is going to pay.” [COR ftelpv26] Moi, je vais pas payer “I am not going to pay.” C’est peut-être cela que j’ai guetté “Perhaps it is that what I observed.” [FRANTEXT 1993 Orsenna] Peut-être que j’ai guetté cela Lit.: “Perhaps that I observed that.” J’ai peut-être guetté cela “Perhaps I observed that.” Sophie tu la connais depuis longtemps, enfin c’est elle que tu connaissais d’abord “Sophiei, you have been knowing heri for a long time, indeed it was heri that you knew first.” [COR ffamdl16] C’est le riche qui demande au pauvre maintenant “It’s the rich one who asks the poor one now.” [COR fnatpr03]

With respect to it-clefts in Contemporary English, it has been argued that at least two pragmatic types have to be distinguished (see Prince 1978, Declerck 1984, Geluykens 1988; see Lambrecht 2001:483-485 for a dissenting view). In cleft sentences of the first type (called stressed-focus clefts in Prince 1978, type 1 clefts in Collins 1991), X bears focus (and heavy stress), while CP presents discourse-old or inferable material. The second type, named informative-presupposition cleft in Prince (1978) (i-p cleft for short), introduces new information in CP, so that (parts of) CP can arguably be analyzed as constituting (part of) the focus domain(s). Although subsequent work elaborated on Prince’s taxonomy of the discourse functions of clefts (cf. Collins 1991, Delin & Oberlander 1995, submitted, Hedberg 1990, 2000, Katz 2000, Pavey 2004, among others), the distinction between stressed-focus and ip clefts continues to be fundamental for theoretical and empirical studies. Two well-known examples of English i-p it-clefts, the first text-initial, the other occurring in the middle of a longer passage, are cited in (20). (20) a. b.

It was just about 50 years ago that Henry Ford gave us the weekend (Prince 1978:898) The federal government is dealing with AIDS as if the virus was a problem that didn’t travel along interstate highways and was none of its business. It’s this lethal national inertia in the face of the most devastating epidemic of the late 20th century that finally prompted one congressman to strike out on his own (Hedberg 1990, ch. 6:27)

Within the class of i-p clefts, we can distinguish between two important subtypes, according to the status of X as given (or, at least, inferable) or new

ANDREAS DUFTER

50

(cf. the distinction between type 2 and type 3 clefts in Collins 1991). Clefts of the second type, that simultaneously serve to introduce a new discourse referent and predicate something about it, have also come to be known as allnew, thetic or presentational clefts. Besides, inferential sentences have been subsumed under the class of i-p clefts, since they share with them the introduction of new information in the embedded clause. The French corpora I examined contain only few clear instances of clefts that encode new information both in X and in CP. Conversely, there is an increasing number of c’est-clefts with given or inferable X and new information in CP. Many of these given X constituents refer back to preceding discourse referents or propositions (cf. the class of cohesive c’est … qui / que in Krötsch & Sabban 1990). Upon closer scrutiny, it turns out that these usages are unavailable in the BFM1 corpus of Old and Middle French, while they are clearly attested in the 1550 FRANTEXT corpus. In (21), examples for the various subtypes of i-p c’est-clefts from Renaissance French are provided: cohesive clefts in (21a,b), a presentational cleft sentence in (21c) (note that the Muse has not been mentioned before in the Ode), and an inferential sentence in (21d). (21) a. b.

c.

d.

Renaissance French C’est tout cela que je puis dire en somme “It’s all I can actually say.” [FRANTEXT 1550 Bèze] C’est à ce jour / que nostre muse oisive / veut rompre pour toi son sejour “It is on this day that our otiose muse wants to interrupt her stay for you.” [FRANTEXT 1550 Ronsard] C’est la Muse qui engarde / les bons de ne mourir pas “It’s the Muse who is working to prevent the good ones from dying.” [FRANTEXT 1550 Ronsard] C’est qu’elle estoit contante de parler à luy “It’s that she was happy to talk to him.” [FRANTEXT 1550 Marguerite de Navarre]

Of course, the emergence of i-p clefts in Renaissance French is itself something to be explained.10 Tentatively at least, one could relate these developments to an increasing need for clarity of expression and textual organization felt among Humanist authors (Jochimsen 1907:112-113). In addition, language-internal factors such as the loss of verb second and clauseinitial cohesive devices such as Old French si “and” should also be taken into consideration (see Ball 1994 for an in-depth study of the rise of i-p clefts in Late Middle English). However, further research is needed to evaluate these hypotheses. At least, the recognition of the i-p class, defined by the availability 10

An anonymous reviewer raised the question of whether the emergence of i-p clefts should be interpreted as the delayed “actualization” of the full pragmatic potential of c’est-clefts, or as the emergence of a new c’est-cleft construction. I tend towards the former view since no systematic prosodic, syntactic or semantic differences have been observed for the different pragmatic types of clefts in Modern French.

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

51

of CP as a potential focus domain, appears to constitute an insightful generalization, which accounts for the more or less simultaneous emergence of cohesive, presentational, and inferential c’est-clefts in sixteenth century French. Moreover, the proportional increase of clefted adverbials finds a natural explanation, since i-p clefts are found most frequently with adverbial X in the corpora examined (see Hasselgård 2004 for similar observations with respect to English). Contrary to what some functionalist approaches to clefting would lead us to expect, not all pragmatic functions of it-cleft sentences are available in all languages or language stages (see Huber 2006 for similar remarks with respect to modern Germanic languages). Conversely, the diachrony of clefting is certainly not a “random walk” (Detges & Waltereit 2007) through the space of conceivable information-structural arrangements. Finally, I would like to suggest that the very same innovation of cleft structures with focus on the embedded clause also applies to interrogative sentences containing est-ce que (lit. “is it that”; cf. section 2), which at least originally may be interpreted as c’est-cleft with interrogative inversion. While there are several attestations of est-ce que in wh-questions as early as in Old French, it is only in the sixteenth century that it gets regularly employed in yesno interrogatives (Foulet 1921:264-268, Rouquier 2003:339-340). In the light of the preceding discussion, the delayed extension to yes-no questions no longer comes as a surprise. In wh-questions introduced by a wh-constituent, this clause-initial constituent arguably bears the focus and the rest of the clause is backgrounded, so that inserting est-ce que immediately after the wh-subpart will yield a sentence with the same focus-before-background order as in stressed-focus c’est-clefts. By contrast, yes-no questions need not have their focus subpart in clause-initial position. Therefore, it is only when c’est-clefts with focus on (a subpart of) the embedded CP become available that the use of est-ce que may be extended to yes-no interrogatives. In conclusion, this contribution has argued against a standard view according to which clefts function as a ‘costly’ substitute for more economic focus markings, especially in situ focus accents in speech. First, it is difficult to see how this view could account for the fact that in Contemporary French, c’est-clefts are found to occur considerably more often in speech than in writing. Second, the diachronic corpus data presented suggest that the rise of c’est-clefts in French cannot be satisfactorily explained as a reaction to the loss of flexible, discourse-governed sentence prosody and constituent orders. C’estclefts are solidly attested before the emergence of phrase-final rhythm and rigid SVO, and continue to augment in the texts examined well after these developments. Moreover, it seems unclear how the observed increase in the proportion of clefted adverbials and the emergence of c’est que-clefts with zero clefted constituents in the sixteenth century could be insightfully characterized in grammaticalization-theoretic terms of semantic and discursive attrition (Gil 2003:206) or gradual “demarking” (Sornicola 2006:408). By contrast, both of

52

ANDREAS DUFTER

these changes can be shown to follow naturally from pragmatic change, viz. the emergence of informative-presupposition (i-p) c’est-clefts, and the concomitant loss of a one-to-one correspondence of focus-background structure and the bipartite syntax of cleft sentences.

REFERENCES Corpora BFM1 = Base de Français Médiéval 1, version avril 2005. [http://bfm.enslsh.fr/; accessed september 2005] COR = Cresti, Emanuela & Massimo Moneglia, eds. 2005. C-ORAL-ROM. Integrated reference corpora for spoken Romance languages. (= Studies in Corpus Linguistics, 15.) Amsterdam & Philadelphia: John Benjamins. FRANTEXT = Base textuelle FRANTEXT, version non-catégorisée. [http:// www.frantext.fr/; accessed september 2005] Further References Ambar, Manuela. 2005. “Clefts and Tense Asymmetries”. UG and External Systems. Language, brain and computation, ed. by Anna Maria Di Sciullo, 95-127. Amsterdam & Philadelphia: John Benjamins. Ball, Catherine N. 1994. “The Origins of the Informative-Presupposition ItCleft”. Journal of Pragmatics 22. 603-628. Blanche-Benveniste, Claire. 2000. “Présence et absence de prépositions dans les clivées et les pseudo-clivées”. Actes du XXIIe Congrès International de Linguistique et de Philologie Romanes, Bruxelles, 23-29 juillet 1998, vol. 6: De la grammaire des formes à la grammaire du sens, ed. by Annick Englebert, 55-65. Tübingen: Niemeyer. Clech-Darbon, Anne, Georges Rebuschi & Annie Rialland. 1999. “Are there Cleft Sentences in French?”. The Grammar of Focus, ed. by Georges Rebuschi & Laurice Tuller, 83-118. Amsterdam & Philadelphia: John Benjamins. Collins, Peter C. 1991. Cleft and Pseudo-Cleft Constructions in English. London: Routledge. Combettes, Bernard. 2003. “Variation dans la structure du syntagme verbal”. Évolution et variation en français préclassique: Études de syntaxe, ed. by Bernard Combettes, 171-210. (= Linguistique historique, 1.) Paris: Champion. Davidse, Kristin. 2000. “A Constructional Approach to Clefts”. Linguistics 38. 1101-1131. Declerck, Renaat. 1984. “The Pragmatics of It-Clefts and WH-Clefts”. Lingua 64. 433-471.

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

53

Delahunty, Gerald P. 1995. “The Inferential Construction”. Pragmatics 5. 341364. Delin, Judy L. & Jon R. Oberlander. 1995. “Syntactic Constraints on Discourse Structure: the case of it-clefts”. Linguistics 33. 465-500. ----------. Submitted. “Cleft Constructions in Context. Some suggestions for research methodology”. Detges, Ulrich & Richard Waltereit. 2007. “Introduction”. This volume. Di Tullio, Ángela. 2006. “Clefting in Spoken Discourse”. Encyclopedia of Language and Linguistics (second edition), ed. by Keith Brown, vol. 2, 483-491. Amsterdam: Elsevier. Doherty, Monika. 2001. “Discourse Functions and Language-Specific Conditions for the Use of Cleft(-like) Sentences: a prelude”. Linguistics 39. 457462. Du Bois, John W. 1987. “The Discourse Basis of Ergativity”. Language 64. 805-855. ----------. 2003. “Argument Structure: grammar in use”. Preferred Argument Structure. Grammar as architecture for function, ed. by John W. Du Bois, Lorraine E. Kumpf & William J. Ashby, 11-60. Amsterdam & Philadelphia: John Benjamins. Dufter, Andreas. 2003. Typen sprachrhythmischer Konturbildung. Tübingen: Niemeyer. ----------. In prep. Zur Diachronie der Fokusmarkierungen im Französischen und Spanischen. Foulet, Lucien. 1921. “Comment ont évolué les formes de l’interrogation”. Romania 47. 243-348. Fournier, Nathalie. 1998. Grammaire du français classique. Paris: Belin. Gamillscheg, Ernst. 1957. Historische französische Syntax. Tübingen: Niemeyer. Geluykens, Ronald. 1988. “Five Types of Clefting in English Discourse”. Linguistics 26. 823-841. Gil, Alberto. 2003. “Zur Geschichte des Spaltsatzes und seiner strukturellen Varianten im Romanischen”. Aufgaben und Perspektiven der romanischen Sprachgeschichte im dritten Jahrtausend, ed. by Alberto Gil & Christian Schmitt, 195-217. Bonn: Romanistischer Verlag. Gómez-González, María de los Angeles & Francisco Gonzálvez-García. 2005. “On Clefting in English and Spanish”. The Dynamics of Language Use: functional and contrastive perspectives, ed. by Christopher S. Butler, María de los Angeles Gómez-González & Susana M. Doval-Suárez, 155-196. Amsterdam & Philadelphia: John Benjamins. Grevisse, Maurice. 1993. Le bon usage: grammaire française (13th edition). Gembloux: Duculot. Harris-Delisle, Helga. 1978. “Contrastive Emphasis and Cleft Sentences”. Universals of Human Language, ed. by Joseph H. Greenberg, vol. 4, 419-486. Stanford, Calif.: Stanford University Press.

54

ANDREAS DUFTER

Hasselgård, Hilde. 2004. “Adverbials in It-Cleft Constructions”. Advances in Corpus Linguistics: papers from the 23rd International Conference on English Language Research on Computerized Corpora (ICAME 23), Göteborg 22-26 May 2002, ed. by Karin Aijmer & Bengt Altenberg, 195211. Amsterdam: Rodopi. Hatcher, Anna G. 1948. “From Ce suis je to C’est moi (the ego as Subject and as Predicative in Old French)”. PMLA 63. 1053-1100. Hedberg, Nancy. 1990. The Discourse Function of Cleft Sentences in English. Ph.D. dissertation, University of Minnesota. ----------. 2000. “The Referential Status of Clefts”. Language 76. 891-920. Helfrich, Uta. 2003. “Hendidas y seudo-hendidas: un análisis empírico-diacrónico”. Actas del XXIII Congreso Internacional de Lingüística y Filología Románica, Salamanca 2001, ed. by Francisco Sánchez Miret, vol. 2, 439-451. Tübingen: Niemeyer. Huber, Stefan. 2006. “The Complex Functions of It-Clefts”. The Architecture of Focus, ed. by Valéria Molnár & Susanne Winkler, 549-578. Berlin & New York: Mouton de Gruyter. Jespersen, Otto. 1937. Analytic Syntax. London: Allen & Unwin. Jochimsen, Paul. 1907. Beiträge zur Geschichte der deiktischen Hervorhebung eines einzelnen Satzteiles, bezw. eines Satzes mittelst c’est (…) que (qui). Doctoral dissertation, University of Kiel. Katz, Stacey L. 2000. “Categories of C’est-Cleft Constructions”. Canadian Journal of Linguistics 45. 253-273. Kay, Paul & Charles J. Fillmore. 1999. “Grammatical Constructions and Linguistic Generalizations: the What’s X doing Y? construction”. Language 75. 1-33. Koch, Peter & Wulf Oesterreicher. 1990. Gesprochene Sprache in der Romania: Französisch, Italienisch, Spanisch. Tübingen: Niemeyer. Krötsch, Monique & Annette Sabban. 1990. “Bleu, je veux — Remarques sur la focalisation en français”. Zeitschrift für romanische Philologie 96. 80-97. Kuryłowicz, Jerzy. 1945. “Le changement accentuel dans la langue française du XVIe siècle”. Bulletin Linguistique [Bucuresti] 13. 39-45. Lambrecht, Knud. 1994. Information Structure and Sentence Form: topic, focus and the mental representations of discourse referents. Cambridge: Cambridge University Press. ----------. 2001. “A Framework for the Analysis of Cleft Constructions”. Linguistics 39. 463-516. Léard, Jean-Marcel. 1992. Les gallicismes: étude syntaxique et sémantique. Paris & Louvain-la-Neuve: Duculot. Lerch, Eugen. 1934. Historische französische Syntax, vol. 3: Modalität. Leipzig: Reisland. Löfstedt, Bengt. 1966. “Die Konstruktion c’est lui qui l’a fait im Lateinischen”. Indogermanische Forschungen 71. 253-277.

ON EXPLAINING THE RISE OF C’EST-CLEFTS IN FRENCH

55

Marchello-Nizia, Christiane. 1995. L’évolution du français: ordre des mots, démonstratifs, accent tonique. Paris: Colin. ----------. 2005. La langue française aux XIVe et XVe siècles. Paris: Colin. Metzeltin, Michael. 1989. “Zur Typologie der romanischen Spaltsätze”. Variatio linguarum. Beiträge zu Sprachvergleich und Sprachentwicklung. Festschrift zum 60. Geburtstag von Gustav Ineichen, ed. by Ursula Klenk, KarlHermann Körner & Wolf Thümmel, 191-203. Stuttgart: Steiner. Miller, Jim. 1996. “Clefts, Particles and Word Order in Languages of Europe”. Language Sciences 18. 111-125. Muller, Claude. 2003. “Naissance et évolution des constructions clivées en c’est…que…: de la focalisation sur l’objet concret à la focalisation fonctionnelle”. La cognition dans le temps: Études cognitives dans le champ historique des langues et des textes, ed. by Peter Blumenthal & JeanEmmanuel Tyvaert, 101-120. Tübingen: Niemeyer. Nelson, Gerald. 1997. “Cleft Constructions in Spoken and Written English”. Journal of English Linguistics 25. 340-348. Pagani-Naudet, Cendrine. 2005. Histoire d’un procédé de style: la dislocation (XIIe-XVIIe siècles). Paris: Champion. Pavey, Emma Louise. 2004. The English It-Cleft Construction: a Role and Reference Grammar analysis, Ph.D. dissertation, University of Sussex. Prévost, Sophie. 2001. La postposition du sujet en français aux XVe et XVIe siècles. Analyse sémantico-pragmatique. Paris: CNRS. Prince, Ellen F. 1978. “A Comparison of Wh-Clefts and It-Clefts in Discourse”. Language 54. 883-906. Pusch, Claus D. 2006. “Marqueurs discursifs et subordination syntaxique: La construction inférentielle en français et dans d’autres langues romanes”. Les marqueurs discursifs dans les langues romanes. Approches théoriques et méthodologiques, ed. by Martina Drescher & Barbara Frank-Job, 173188. Frankfurt a.M.: Peter Lang. Roberts, Craige. 1998. “Focus, the Flow of Information, and Universal Grammar”. The Limits of Syntax, ed. by Peter W. Culicover & Louise McNally, 109-160. San Diego, Calif.: Academic Press. Rouquier, Magali. 2003. “La séquence est-ce dans les interrogatives en qui / que en ancien et en moyen français”. Journal of French Language Studies 13. 339-362. Schöpp, Frank. 2002. “Funktionen der Cleft-Konstruktion im Französischen (Teil I)”. Beiträge zur Fremdsprachenvermittlung 40. 47-66. ----------. 2003. “Funktionen der Cleft-Konstruktion im Französischen (Teil II)”. Beiträge zur Fremdsprachenvermittlung 41. 37-53. Sedano, Mercedes. 1990. Hendidas y otras construcciones con ser en el habla de Caracas. Caracas: Universidad Central de Venezuela. Sornicola, Rosanna. 1991. “Origine e diffusione delle frasi scisse nelle lingue romanze”. Actes du XVIIIe Congrès International de Linguistique et de Phi-

56

ANDREAS DUFTER

lologie Romanes (Université de Trèves (Trier) 1986), ed. by Dieter Kremer, vol. 3, 41-54. Tübingen: Niemeyer. ----------. 2006. “Interaction of Syntactic and Pragmatic Factors on Basic Word Order in the Languages of Europe”. Pragmatic Organization of Discourse in the Languages of Europe, ed. by Giuliano Bernini & Marcia L. Schwartz, 357-544. Berlin & New York: Mouton de Gruyter. Wehr, Barbara. 2005. “Focusing Strategies in Old French and Old Irish”. Opening Windows on Texts and Discourses of the Past, ed. by Janne Skaffari, Matti Peikola, Ruth Carroll, Risto Hiltunen & Brita Wårvick, 353-379. Amsterdam & Philadelphia: John Benjamins.

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE1 ELISABETH STARK Freie Universität Berlin

This paper presents a diachronic formal morphosyntactic analysis of the role of the functional projection Pl/Pl* in Romance indefinite nominals, responsible for number and the countability distinction. Reinterpreting the complex system of indefinite nominal determination in two central Romance languages, viz. French and Italian, which both feature an indefinite article and a ‘partitive article’ as a device of ‘nominal classification’ in a broad sense in contrast to Romance languages without such an element, viz. Spanish, it argues that this ‘classification system’ arose when nominal declension in Latin was partially or completely lost. The application of the latest minimalist assumptions on agreement processes in the syntax both to modern Romance languages and to (Late) Latin allows us to describe and explain the obvious differences between French, Italian and Spanish and to relate them to the interaction of gender and number marking in Romance indefinite nominals.

Introduction: Different systems of indefinite nominal determiners in Romance nominals as results of language change As I have repeatedly shown,2 there is considerable cross-linguistic variation in indefinite nominal determination in Romance, despite some well-known generalizations assuming homogeneous semantic and syntactic systems of nominal determination for all Romance languages (e.g. Chierchia 1998, Longobardi 2001a). See the data in (1a-f):

1.

1

This article was presented in two versions, the first at the Workshop on Grammatical Language Change and its Explanations in Saarbrücken in September 2005 and the second at the International Workshop on the Typological Comparison of Italian and German at the Johann-Wolfgang-Goethe Universität, Frankfurt am Main, in February 2006. Special thanks for very helpful comments go to Guido Mensching and Natascha Pomino from Freie Universität Berlin as well as to Ulrich Detges, Richard Waltereit, Michele Loporcaro, Ulrich Wandruszka, Günter Grewendorf, Alessandra Tomaselli, Cecilia Poletto, Carme Picallo and one anonymous reviewer. All errors and shortcomings are, of course, mine. 2 Cf. some earlier work on the cross-linguistic divergence in Romance indefinite nominals and its syntactic implications, cf. Stark (2006 and forthcoming).

58

ELISABETH STARK

(1) a.

Has visto *(un) águila? As-tu vu *(un) aigle? Hai visto *(un) aquila? “Did you see an eagle?”

(Spanish) (French) (Italian)

b.

Compro pan. J’achète *(du) pain. Compro (del) pane. “I buy (some) bread”

(Spanish) (French) (Italian)

c.

Me falta agua. Il me faut *(de l’)eau. Mi occorre (dell’)acqua. “I need (some) water”

(Spanish) (French) (Italian)

d.

Veo (a unos) estudiantes en el edificio. Je vois *(des) étudiants dans le bâtiment. Vedo (degli) studenti nell’edificio. “I see (some) students in the building”

(Spanish) (French) (Italian)

In fact, there are two main differences to be observed: Firstly, bare noun phrases in argument position occur in Spanish and Italian (and other Romance languages) under very restricted grammatical conditions: bare plurals surface mostly postverbally with subject and object function regardless of the lexical entry of the noun (normally with non-specific interpretation of the nominal, cf. 1d). Bare singulars are also possible with ‘mass-denoting nouns’ (not with ‘entity-denoting nouns’ in so-called singular countable NPs, see 1a) postverbally with subject and object function (see 1b and 1c), again with non-specific interpretation of the nominal. Conversely, the only Romance language which almost never permits bare noun phrases in argument position is French. Secondly, every Romance language possesses a so-called indefinite article derived from the Latin numeral unus, “one”, which accompanies singular count noun phrases. However, only two of the three Romance languages discussed here, i.e. French and Italian, have a further indefinite determiner, the so-called ‘partitive article’ derived from the composition of Latin de and the definite article. It marks indefinite non-countable singular noun phrases, usually with ‘mass-denoting nouns’, in pre- and postverbal subjects and objects. It is obligatory in French and (still) optional in Italian, where its use is increasing constantly, however (cf. Korzen 1996). Functionally, its morphological plural (Fr. des, It. dei, degli) is to be considered as the plural of the indefinite article (see examples 1b-d). If we compare this picture to Latin, the main differences or changes within nominals become immediately obvious: for indefinite nominals as well as definite ones, obligatory determination has become a property of Romance languages in contrast to Latin. See example (2):

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

59

(2) …ubi se tamen montes illi, inter quos ibamus, aperiebant et faciebant vallem infinitam ingens, planissima et valde pulchram, et trans vallem apparebat mons sanctus Dei Syna. “...where in the meantime the mountains, between which we were going, opened out and formed an endless valley – huge, very flat and very beautiful – and across the valley there appeared Sinai, the holy mountain of God.” (Vincent 1997:156, from the Peregrinatio Aegeriae, late 4th century AD)

The first mention of the valley (et faciebant vallem... ‘and they formed a valley’) in a postverbal object does not show any indefinite marking (via aliquis, quidam ‘anyone’, ‘someone’ or else), and even when this discourse referent is referred to anaphorically (et trans vallem apparebat... ‘and across the valley there appeared…’) in a prepositional adverbial, it is left without a definite determiner and not referred to via (clitic) pronouns.3 In short: Latin did not need (obligatory) overt marking of nominal determination in argumental noun phrases and did not have third person pronouns (cf. Vincent 1997 and others). The Romance languages now do have overt nominal determination, which is obligatorily overt in certain syntactic contexts, and they also have (clitic) third person pronouns. This change is of considerable typological impact and can be summed up under diverse headings such as ‘from synthetic to analytical coding’, from ‘dependent-marking’ to ‘head-marking’ or, in a minimalist framework, as changes in the lexical feature structure (see below) of universal functional heads such as C, T, v and D. Alternatively, one could think, as Vincent 1997 did for definite nominals, of the rise of new functional categories in Romance via reanalysis of ambiguous syntactic input or, less radically, of a different parameterized projection of functional categories in different languages. Whatever name we give to the change, it has to be described and explained in a way that satisfies both the aspect of the initial innovation, which is most probably linked to universal cognitive principles, and the later individual mutations of the different Romance linguistic systems. These differ to a considerable extent, both in their history and in their present morphosyntactic make-up, and their linguistic evolution can therefore not be subsumed under some vague general functional coding or decoding strategies, as will be shown below. In the following I shall in particular argue in favour of a formal account of the morphosyntactic changes leading to the new feature structure of a functional head ‘Pl’ in addition to the already (now widely accepted) projection NumP inside nominals 4 , and their language-specific features triggering different overt realizations of nominal determination in different Romance languages. 3

Prepositional phrases are typical syntactic contexts in which nominal determination does not need to be marked overtly, even in languages with nominal determination, cf. modern Romanian. However, modern French, Italian or Spanish would require a definite determiner in the case in question. 4 Cf. especially Ritter (1993, 1995), Picallo (1991, 2005), Munn & Schmitt (2005) and Zamparelli (2004, 2005) for the discussion of NumP.

60

ELISABETH STARK

The remainder of this article is structured as follows: section 2 gives an overview of the nominal morphology and indefinite determiners in French, Italian and Spanish compared with the Latin declensional system in order to reveal two fundamental correlations. Section 3 introduces the formal model, following Chomsky’s most recent proposals for modelling agreement processes and for deriving the correct morphological and semantic quantification of nominals (section 3.1), before applying it to indefinite nominals in modern French, Italian and Spanish. Section 4 tries to retrace the morphosyntactic evolution from Latin to modern Romance languages by applying the ‘probe model’ to questions of diachronic syntax, especially to the functional projection Pl(*), after which the conclusions given in section 5 will summarize the findings and relate them to some parallel approaches in the recent generative literature. 2. A functional ‘explanation’: Romance indefinite determiners as ‘classification devices’ I have repeatedly pointed to the fact that Romance indefinite determiners can be understood as ‘classification devices’ in a broad sense, coding what Seiler in his functional universal dimensions calls “apprehension”:5 APPREHENSION is the universal operational dimension with corresponding subdimensions which explicate the grasping and representation of concepts corresponding to objects or things by means of language. (Seiler 1986:145)

In Latin, the basic notional distinction between an individualized object and some uncontoured substance could be coded by using gender and number, especially by using the semantic and morphological residue of the IndoEuropean opposition between neuter and masculine/feminine. See examples (3a-c): (3) a.

caseus “one single (piece of) cheese”, olea “olive” / “olive tree”

b.

caseum “cheese as a substance”, oleum “oil”

c.

acinus/acinum “berry”, acina “grape”, frumentum “wheat”, frumenta “corn”6

(3a) shows lexical roots with masculine and feminine gender, resulting in ‘entity-denoting nouns’, whereas the nouns from the same root in (3b) with 5

Cf. Seiler (1986) and Stark (2005, 2006, and forthcoming). Cf. in detail Hofmann & Szantyr ((1997) [1965]:7-10), Meisterfeld (1998:56ff.) and for Late Latin analogical neuter plurals following the same pattern cf. Morani (2000:228). 6

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

61

neuter gender are ‘mass-denoting nouns’. 7 In addition, (3c) shows the wellknown ‘collective’ semantics of the Latin neuter plural ending in -a (cf. Schön 1971, Windisch 1973). Although theses oppositions are not systematic, the Latin neuter and especially the Latin neuter plural in -a – both unambiguously marked in spoken and written varieties – can be re-interpreted as a partly generalised ‘classification system’ mainly denoting the opposition between ‘single, contoured object’ (e.g. one piece of cheese, one olive, one berry) and ‘noncontoured substance’ (e.g. cheese, oil) or ‘collective’ (grapes). Furthermore, this important semantic opposition is as much related to gender as to number: the Latin plural is neither automatically interpreted as additive (cf. Link 1991) nor restricted to ‘entity-denoting nouns’. See examples (4a-b): (4) a. b.

frigora caloresque “an intense heat and cold”: plural indicating intensification acquae “waters”, cerae “wax tablets”: different appearances of a substance8

Although the Latin plural can have a sortal reading, bare plurals of abstract or ‘mass denoting-nouns’ are not automatically re-categorized as they for instance are in modern Romance languages (compare Fr. huile “oil”, des huiles “different sorts of oil”). Virtually any Latin noun can be pluralized, and frequent occurrences of plurals of ‘mass-denoting nouns’ or abstract nouns, as in (4) above, are in fact attested. 9 This indicates that Latin had no grammaticalized ‘countability distinction’ at the level of noun phrases. In contrast to Latin, modern Standard French, Italian and Spanish have almost completely lost the possibility of ‘classification’ via gender/number alternations (and declension class). The main morphological changes in the nominal system from Latin to Romance are well-known and comprise the 7

In describing the facts in these terms, following Löbel (1993:192ff.) and with reference to Gil’s (1987) typology, I assume a fundamental difference between the lexical categories ‘massdenoting’, ‘entity-denoting’ and ‘abstract noun’ (N), which derive from characteristics of the potential (extra-linguistic) referents and are based on denotational properties of the head noun, and the countability or non-countability of entire noun phrases. This last opposition is a grammatical category or a feature which depends on the internal syntactic structure of the noun phrase and is mainly characterized by compatibility with certain indefinite determiners (French/Italian: un/uno ‘one’, ‘a’, INDEF.ART vs. du/del ‘of.the’, PART = partitive article). This assumption is justified by the fact that virtually any noun in Romance languages (like in any language with a grammaticalized countability distinction in this sense) can in principle appear in any kind of noun phrase, cf. German Nach dem Unfall lag jede Menge Auto herum, “After the accident, there was much car lying around”. 8 Cf. Kühner & Stegmann (31955:69, 73), Hofmann & Szantyr ((1997) [1965]:18, 21). 9 Cf. Iturrioz Leza (1986:295f.): “This individualization strategy [ = pluralization of abstract nouns, E.S.] is more widespread in the classical languages (Greek, Latin) than in modern German or any other European language; it is thus often difficult to translate an abstract [plural, E.S.] NP without changing its number: [...] Asperitates viarum et angustiae [...] ‘The roughness(es) and narrowness(es) of the ways’”.

62

ELISABETH STARK

complete loss of morphological case, a reduction of gender (especially loss of the neuter), contrasted with a solid formal preservation of number, but now with only the ‘additive plural meaning’ left. In addition, modern Standard French shows the complete loss of the declension classes (already lost in Old French, cf. Delfitto & Schroten 1991:180f.). Gender and number are usually marked (in the phonetic code) by prenominal determiners only. Modern Standard Italian is different from French in that it has preserved 3 main declension classes, 2 overtly marked genders and overt number marking. However, the declensional endings -a and -e are far from being unambiguous markers of singular or plural, as they can either indicate feminine singular, (rarely) masculine singular or feminine plural (-a) or masculine singular or feminine plural (-e). The morpheme -o unambiguously indicates singular, but both masculine and (rarely) feminine gender. Just like Italian, modern (European) Standard Spanish has 3 main declension classes, 2 overtly marked genders and overt number marking. It is unambiguous with respect to number marking (“plurality is manifested consistently with the suffix /-s/”, Harris 1992:67). See (5a-c): (5) a.

b.

un ami (m) / une amie (f.) – des ami(e)s [œ)nami / ynami – dezami] “a male friend”/“a female friend”, “male/female friends”

(French)

Sg.: -o / -a / -e Pl.: -i / -a / -e

(Italian)

libr-o – libr-i (m.) “book” – “books” cas-a – cas-e (f.) “house” – “houses” can-e – can-i (m.) “dog” – “dogs” c.

mano – mani (f.) “hand” – “hands” poet-a – poet-i (m.) “poet” – “poets”

Sg.: -o / -a / -e; Pl.: -s: pas-o – pas-os (m.) “step” – “steps” pas-a – pas-as (f.) “raisin” – “raisins” jef-e – jef-es (m.) “chief” – “chiefs”

bracci-o – bracci-a (m. – f.) “arm” – “arms”

(Spanish) man-o – man-os (f.) “hand” – “hands” map-a – map-as (m.) “map” – “maps” nub-e – nub-es (f.) “cloud” – “clouds”:

Please note in (5b) the residue of the original Latin classification potential of the neuter plural in -a as opposed to a regular plural form in -i (originating in Late Latin, cf. Hofmann & Szantyr (1997 [1965]:21) and reanalysed as feminine (plural), always indicating a collective or at least ‘pair’ reading). Some nouns ending in -o (masculine singular), usually denoting concrete objects like body parts (It.: ginocchio “knee”, orecchio “ear” and so on, also

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

63

muro “wall” etc.), have a plural form in -a when denoting a plurality, body parts or a ‘collective reading’. However, they form a plural in -i when used metaphorically to denote something similar in form but without a collective denotation (e.g. It.: le braccia denotes both arms of an animate being, whereas i bracci denotes the arms of a river, It. le mura denotes the city wall, whereas i muri denotes the single walls of a building). Unlike Italian, Spanish does not seem to preserve a certain ‘classification potential’ in nominal (declensional) endings; the only slight ‘classification potential’ left in Spanish is the erroneously named ‘neuter’ (deriving from the Latin neuter singular) in the pronominal system which refers to CPs, i.e. quotations, statements of fact etc. (cf. Picallo 2002).10 If we try to relate the findings concerning the paradigm of indefinite determiners with the nominal morphology shown by French, Italian and Spanish up to now, we can observe two important correlations: Firstly, the possibility of having bare arguments, especially bare plurals in argument position (Italian, Spanish vs. French), correlates with the formal preservation of the Latin ‘neuter’ in nouns or pronouns, not only with the possibility of overt plural marking.11 Secondly, the presence of a ‘partitive article’ explicitly coding non-countability (Italian, French vs. Spanish) correlates with ambiguous plural marking.12 If we try to explain the general language change from Latin towards Romance in relation to the re-structuring of nominal morphology and the grammaticalization of nominal determination for indefinite argument nominals too, we could now easily take up Givon’s observations about the widespread grammaticalization of the numeral “one” into an indefinite article (cf. Givón 1981) and thus an indicator of countability via implicature (‘if something can be quantified as being one element out of a set in contrast to two, three and so on, it is a member of a set of countable discrete entities’). This element could thus have compensated for the loss of the complex Latin declension and assumed the function of ‘classification’ in a broad sense. Even if this ‘explanation’ seems convincing at first sight, it leaves us with a lot of open questions and does not provide a satisfying structural descriptive analysis of the individual phenomena occurring in the morphosyntactic structure of the different Romance languages. In particular, the new syntactic property of obligatory coding of countability in Romance indefinite nominals in contrast to 10

Spanish personal pronouns and demonstratives show a threefold morphological opposition, with forms ending in -e for masculine singular, -a for feminine singular, and -o for the socalled ‘neuter’ (e.g. span. no quiero pensar más en él/ella/ello “I do not want to think about him/her/it any longer”). 11 Contrary to what is postulated in Schroten (2001). 12 Cf. Stark (in prep.), with reference also to older stages of the Romance languages and the simultaneous development of the ‘partitive article’ and the loss of overt or unambiguous plural marking in French and Italian.

64

ELISABETH STARK

Latin remains unexplained, and the considerable differences between the single Romance languages which we showed above in section 1 are completely neglected – as are the specific properties of many other languages which have grammaticalized “one” into an indefinite article and are nevertheless very different from Romance in many respects (English, German, Hebrew and so on). In addition, we have to bear in mind that “[u]seful or needed things are not sufficiently explained by their usefulness or the need for them” (Haspelmath 1999:188). 3. A formal account of Romance indefinite nominals 3.1 ‘Manufacturing plurality’ Following the formal modelling of agreement in the latest versions of the minimalist program,13 lexical items and more generally heads “interact through agreement to influence the shape of syntactic structure and the process of semantic interpretation” (Pesetsky & Torrego 2004:1). Lexical items can be described as having certain bundles of grammatical features, so-called φfeatures, some valued, others unvalued (e.g. the number feature Nb on nouns is normally valued and can have the value singular, SING, or plural, PLUR, whereas it does not come already valued from the lexicon for determiners or adjectives, which accordingly have to agree with the noun in nominals, cf. Pesetsky & Torrego 2004:1f.). Features can thus be described as attribute–value pairs. Moreover, features are considered to be either interpretable or uninterpretable: “This distinction is concerned with a different question: whether or not a feature of a particular lexical item makes a semantic contribution to the interpretation of that item” (Pesetsky & Torrego 2004:2). The number feature is generally considered to be interpretable on nouns, whereas gender in modern Indo-European languages does not seem to make a semantic contribution to the respective lexical items (but see Radford 2004:288, Picallo 2005 and the discussion below). There is a direct relation between the notion of (un)valued features and (un)interpretable features: if a feature (e.g. number) is interpretable under a given category, it enters the derivation with a specified value (e.g. plural). In contrast, if the feature is uninterpretable, it is not specified; it will be valued via agree. That is, uninterpretable features must be deleted before Spell-Out, and in order to be deleted, they must be valued by agree, so that agreement can be understood as directly related to the interpretability of syntactic structures by the semantic component. 14 Agreement processes are conceived of as ‘probing’ processes: Unvalued features on a given category item act as a probe and search in their c13

Cf. Chomsky (2000, 2001, 2005), Radford (2004, chapter 8), Heycock & Zamparelli (2003), Zamparelli (2004, 2005), Pesetsky & Torrego (2004), Mensching (2005), Mensching & Remberger (2006). 14 For details see Radford (2004:289f.).

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

65

commanded domain for a suitable goal, i.e. an item with matching valued features. Both, the probe and its goal must be ‘active’ for agree to apply, i.e. they must contain at least one uninterpretable/unvalued feature.15 After match, the operation value takes place: the valued features of the goal are copied onto the probe, which can, but does not have to, instantiate an unvalued feature of its goal. After this agreement process the goal is mobile for syntactic movement. In order to apply this model, originally conceived for agreement processes at the sentence level (e.g. person and number agreement and nominative case assignment between T and the subject DP), to the internal structure of nominals (gender and number agreement and its interaction with indefinite determiners) and in order to derive their correct quantificational interpretation, we will use a modified version of Heycock & Zamparelli’s (2003) proposal for English indefinite nominals. Following Heycock & Zamparelli (2003), we assume first that nouns do not have a specific interpretation with regard to ‘count’ or ‘mass’ or anything else but that their denotation is merely a set of singleton elements irrespective of their morphological number (cf. Heycock & Zamparelli 2003:13 for a detailed argumentation). The interpretation of a nominal as being ‘semantically pluralized’, i.e. as denoting “a set composed in all the ways in which these [elements, E.S.] can be grouped together into pluralities” (as “having a join semilattice structure”, with the notation [+LATT]) is created by merge of a noun with a functional head, ‘Pl*’ which operates semantic pluralization (i.e. creates a ‘semilattice interpretation’, cf. Link 1983). The semantic feature [±LATT] is introduced by Heycock & Zamparelli (2003) in order to describe the contribution of the functional head Pl* to the interpretation of the whole nominal. Pl* creates sets of sets of different sizes out of the set of singleton elements denoted by the NP: For example, with 4 individual tall boys a, b, c, d in the domain, [NP tall boys] will denote {{a}, {b}, {c}, {d}}, and PlP will be: || [PlP tall boys] || = {{a, b, c, d}, {a, b, c}, {b, c, d}{a, c, d}, {a, b, d}, {a, b}, {a, c}, {a, d}, {b, c}, {b, d}, {c, d}, {a}, {b }, {c}, {d}}. At the next level up, NumP behaves as a filter. It regulates the cardinality of the PlP denotation, eliminating from it all the pluralities with the wrong number of atoms […]. (Heycock & Zamparelli 2003: 13)

This denotation of plural noun phrases is parallel to the one of mass terms, which also denotes sets of sets of portions of the same substance (cf. Link 1983). Later we will mention some structural parallels between plural noun phrases and non-countable nominals, independent of morphological number and demonstrating the relevance of the [LATT]-feature in the nominal derivation. The functional head Pl(*) is also responsible for the morphological number agreement inside the nominal. Heycock & Zamparelli (2003) work 15

Cf. Radford (2004:289ff.).

66

ELISABETH STARK

within a framework which still assumes feature checking, as in earlier versions of minimalism, and they apply the checking mechanism both for morphological/grammatical and semantic features (essentially number and [±LATT]), whereas we will apply the probe-model and assume agree processes only between grammatical attribute-value pairs. Thus, in our framework the number feature of Pl(*) is unvalued ([Nb: ]) and acts as a probe. This feature will be valued by agree with the lexical noun (cf. 6a) and gets deleted after valuation (cf. 6b), i.e. it gets invisible to the semantic, but remains visible to the phonological component. The interpretation of the nominal depends ultimately on the different properties of the heads in the derivation and on the interaction of different agreement processes in the structure: (6)

a.

probe and match

PlP

probe probe

Pl* N [Nb: ] [Nb: PLUR] [+LATT] oils probe & match

b.

value and delete

PlP Pl* N [Nb: PLUR][Nb: PLUR] [+LATT] oils value & delete

Note that, in this derivation, N does not have any uninterpretable features and is thus theoretically not ‘active’ in the sense described above. On the other hand, the probe in Pl(*) has to be instantiated for morphological number – and there is only one matching element in its c-commanded domain, N, with a valued Nb feature. For English nominals, therefore, we have to assume16 that agree takes place in this way, given the fact that the probing operation cannot take place erroneously between a probe and an inadequate goal. In the next step, PlP can optionally be merged with Num, the head that hosts cardinals, weak quantifiers etc.:17 (7)

merge with Num NumP Num n = [5] (five)

PlP Pl* N [Nb: PLUR][Nb: PLUR] [+LATT] oils

Since NumP acts as a filter regulating the cardinality of the PlP denotation (see above), only sets containing five elements will be left in the denotation of 16 17

Just as Pesetsky & Torrego (2004:1) seem to assume tacitly for Latin nominals as well. Cf. Heycock & Zamparelli (2003:11ff.).

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

67

NumP. Numerals are lexically specified for their semantic [±LATT] feature and have to possess the same [±LATT] feature as Pl(*) in order to derive a grammatical structure; the numeral “one” has [-LATT], numerals above “one”, [+LATT]. Heycock & Zamparelli 2003 (17f.) assume here a traditional feature checking mechanism between Num and Pl(*), which avoids, e.g., the ungrammatical structure *one oils. We will not go into any further detail here, given that the problem of explicit quantification by numerals is not central to our discussion. Also, NumP controls number agreement outside the nominal, e.g. in cases where the respective nominal is the sentential subject and has to agree in number with the finite verb. There are several arguments in favour of assuming that two (rather than only one 18 ) functional projections are involved in the plural marking of a nominal. One important argument comes from languages like Hungarian, which combines singular N with cardinals higher than “one”: (8)

a.

probe and match

b.

PlP

probe probe

Pl* N [Nb: ] [Nb: SING] [+LATT] hajót (‘ship’) probe & match

value and delete NumP Num PlP n = [5] Pl* N öt (‘five’) [Nb: SING] [Nb: SING] [+LATT] hajót (‘ship’) value & delete

If morphological number agreement and ‘semantic pluralization’ plus indication or filtering of the correct cardinality were situated in the same functional projection, say Num, the Hungarian structure could not be accounted for, as Num would host two contradictory features, i.e. the cardinality “five”, thus selecting a set of five elements out of a set of sets and regulating morphological plural agreement outside the nominal, and a probe being valued with morphological singular number inside the nominal. Most importantly for our argumentation, by separating morphological and semantic number (SING vs. PLUR and [±LATT]), Heycock & Zamparelli (2003) can account firstly for the difference between ‘count’ and ‘mass’ readings of singular nouns in languages like modern English, German or Spanish and secondly for the striking parallels in the syntactic behaviour of singular mass and plural nominals in many languages: In these languages, it is not the noun itself that carries [±LATT] information, nor is it the morphological number of a nominal, as virtually any noun can appear in both a countable or 18

Like Ritter (1993 and subsequent work) with NumP only.

68

ELISABETH STARK

non-countable nominal. It is merge of N with Pl* irrespective of the morphological number of N which creates a [+LATT] denotation.19 Now, what happens if N is not merged with Pl* and is therefore not semantically pluralized? In the case of a singular N, a default element, the indefinite article, has to enter the enumeration in order to derive the [-LATT] denotation: (9)

a.

probe and match

PlP

probe probe

Pl N [Nb: ] [Nb: SING] [-LATT] oil boy probe & match

b.

value and delete NumP Num n = [1]

PlP Pl N [Nb: SING] [Nb: SING] [-LATT] oil *(an) boy *(a) value & delete

According to Heycock & Zamparelli (2003), the obligatory presence of a(n) in a structure of this type is due to the fact that Pl would not contribute anything semantically to the denotation of the whole nominal. They further assume that a functional head which is semantically inactive and phonologically empty should not exist in the structure. 20 Alternatively, we could say that the [±LATT]-feature in Pl(*) needs an overt expression in a language for one or the other value, the opposite value being implicated in the absence of overt coding. Furthermore, Pl(*) is syntactically necessary for phrase internal morphological number agreement, which does not automatically coincide with external morphological number agreement (cf. Fr. la plupart des gens fument, Engl. this/*these committee are to meet tomorrow). This is enough reason to assume its existence inside (indefinite) nominals (see below). Be it as it be, in the case of the absence of Pl*, the Pl position in English is filled with an element providing the [-LATT] interpretation for the whole nominal, thus disambiguating overtly the semantic interpretation of the whole structure. Contrary to the Romance languages and their indefinite article deriving from the Latin numeral unus (see below), English a is merged in Pl on the basis of different distributional properties (cf. the impossible literal translation into Romance of the English many a time, too tall a boy etc.).21 Functionally, the obligatory indefinite article in Romance countable nominals assumes the parallel coding of [-LATT], among others functions (see below). 19

Cf. Heycock & Zamparelli (2003:16f.). Cf. Heycock & Zamparelli (2003:17). This is also in complete accordance with Economy Principles assumed already in Chomsky 1995. 21 Cf. Heycock & Zamparelli (2003:17f.). 20

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

69

3.2 Romance languages: gender, number and ‘classification’ How can we now relate this model to our initial problems, i.e. the striking differences between French, Italian and Spanish with regard to their systems of indefinite determiners, the possibility of having bare nominals in argument position and the respective language change from Latin to Romance? First of all, Romance languages and Latin have overt gender as an additional φ feature besides number, a feature which is usually considered to be uninterpretable. As gender is expressed overtly at least on some cardinals, weak quantifiers and the indefinite article, and as it has to be valued there in order to be deleted,22 there has to be a probe for gender in Num,23 and there will be two subsequent agree operations to carry out, the first one for morphological number between Pl* and N (but see below) and the second one for morphological gender between Num and N (Pl* in Spanish is not a matching goal for the probe for gender). Num controls gender agreement inside the nominal. A structure like the following one must accordingly be assumed for Spanish: (10) a.

agree between Pl* and N

PlP

probe probe

Pl* N [Nb: ] [Nb: PLUR] [+LATT] [Gd: FEM] aguas probe & match

(10) b.

agree between Num and N

NumP

probe probe

Num PlP [Gd: ] Pl* N n > [1] (muchas) [Nb: PLUR][Nb: PLUR] [+LATT] [Gd: FEM] aguas probe & match

22

Cf. Heycock & Zamparelli (2003:19f.). This, then, is an explanation for the necessary location of gender in Num: due to the mechanism of agree and feature deletion immediately after merge (cf. the ‘Earliness Principle’, which requires operations, especially agree, to apply as early as possible in a derivation, cf. Radford 2004:282), gender cannot be a probe together with number in Pl(*), as it would get valued and deleted. 23

70 (10) c.

ELISABETH STARK resulting configuration NumP Num PlP [Gd: FEM] Pl* N n > [1] (muchas) [Nb: PLUR][Nb: PLUR] [+LATT] [Gd: FEM] aguas

In the case of agua, we would get un(a) for Pl [-LATT] just like in English. Yet, in Spanish and in Romance in general the indefinite article is located in Num, because it also realizes the gender feature of this category. In Spanish, after full instantiation of the probes, N can additionally move from its position at least up to Num and further to D (as only DPs are arguments in Romance, cf. Chierchia 1998 and Longobardi 2001b) and can thus appear as a bare plural in argument position (cf. also Picallo 2005:110, though assuming a slightly different denomination and function for Pl(*), see below). The gender feature of N remains undeleted in this and the following derivations, because it is considered to be interpretable, roughly following Picallo 2005, who assumes that “gender declension is the morphological exponent of an abstract interpretable feature in the functional layer of nominal structures” (Picallo 2005:107ff.), an ‘abstract classification feature’ (cf. also Harley & Ritter 1999, 2002a,b for a feature-geometric approach to pronouns, where gender is the specification under a ‘classification node’). Note that, theoretically, a probe for gender could also be situated in Pl(*) together with the number probe (as proposed in Picallo 2005 for Spanish and Catalan). However, the below discussion of Italian and especially French will show that the assumption of two different probes in two different heads, one for gender in Num and one for number in Pl(*), makes clearer predictions for the syntactic behaviour of indefinite nominals in the respective languages. We will try to show in the following that the gender feature is ultimately indirectly related to the semantic [±LATT] feature introduced by Heycock & Zamparelli 2003. The problem is that the location of gender varies in the Romance languages, and the complete syntactic and semantic structure of Romance nominals can only be obtained by assuming a complex interplay between the functional heads Num and Pl(*) and their respective number and gender features. As has been pointed out in section 2, French nouns do not have a morphological marker for gender and number. They could be analysed as having lost any grammatical features and as not contributing to an agreement phenomenon of any kind. The only information they provide is their descriptive semantic content. Accordingly, morphological number must be specified and interpreted elsewhere, for example in NumP, by obligatorily inserted lexical items which bear a valued number but an unvalued gender feature. A probe for gender is

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

71

situated in Num, and agree will therefore take place not between Num and/or Pl(*) with N as in English or Spanish, but between Num and Pl(*) only. We should remember that, in Romance, NumP is the place for gender agreement inside and outside the nominal (cf. Heycock & Zamparelli 2003:19f.). Additionally, we assume that, for French at least, number is valued (and interpretable) in NumP and has to be assigned to Pl(*) after the agree operation (remember that fully valued probes can after value assign unvalued features to their goals, like the person-number-gender probe in T assigns nominative case to its ‘goal DP’, see above, the beginning of section 3.1). Evidence for this comes from the morphological difference between plural and singular determiners in French – as on N in Spanish, the inflectional endings of the French determiners indicate unambiguously whether a nominal is plural (obligatorily ending in [-e]) or not. Pl(*) is visible/active to the probe in Num because its number feature is again not valued, but it is not itself a probe. For this unvalued feature we will use the notation ([Nb: A]).24 After valuation of the gender probe of Num, the number value can be assigned to Pl(*) (and deleted afterwards). The derivation of French nominals is as follows: (11) a.

agree between Num and Pl* NumP

probe probe

Num PlP [Gd: ] N [Nb: PLUR] Pl* [Gd: FEM] [o] n [Nb: A] (eau[x]) [+LATT] probe & match

(11) b.

resulting configuration

NumP Num PlP [Gd: FEM] N [Nb: PLUR] Pl* [Gd: FEM] [o] n *(des) [Nb: PLUR] (eau[x]) [+LATT]

24

Note that this unvalued feature does not act as a probe. It is comparable to the unvalued Case feature of N.

72

ELISABETH STARK

During this agree operation the unvalued feature for number of Pl*, which doesn’t act as a probe, is instantiated. That is, the agree operation also serves for assigning number to Pl*. Thus, as French nouns do not participate in agreement processes at all, they remain totally disabled for movement to D – for this reason, bare noun phrases cannot appear in argument position in French. In contrast to Spanish, and because of the lack of number information in N and Pl(*), French indefinite singular noun phrases exhibit a morphological element for singular nominals with a [+LATT] interpretation, the ‘partitive article’, indicating morphological number and [+LATT]: (12)

a. agree between Num and Pl*

b. resulting configuration

NumP

probe probe

Num PlP [Gd: ] [Nb: SING] Pl* [Gd: FEM] n [Nb: A] [+LATT]

NumP Num PlP [Gd: FEM] [Nb: SING] Pl* [Gd: FEM] n *(de l‘) [Nb: SING] [+LATT]

N eau ([o])

probe & match

N eau

Please note that, in the derivations (11) and (12), the number feature remains undeleted in Num, just like the gender feature in Pl*. We will come back to this later. French countable nominals are to be derived in the following way: (13) a.

agree between Num and Pl: NumP

probe probe

Num PlP [Gd: ] Pl [Nb: SING] [Gd: FEM] n [Nb: A] [-LATT] probe & match

N eau ([o])

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

(13) b.

73

resulting configuration: NumP Num PlP [Gd: FEM] Pl [Nb: SING] [Gd: FEM] n *(une) [Nb: SING] [-LATT]

N eau ([o])

Just as in English, the indefinite article un(e) has to appear obligatorily with French singular countable noun phrases; however, it has to do so in order to indicate the [-LATT] interpretation (it also appears higher in the structure than English a, cf. Heycock & Zamparelli 2003:17f.) and to indicate singular number. Italian now has overt expression of number, but this is quite ambiguous, whereas gender, at least for the overwhelming majority of nouns ending in -o and -a, is marked on the noun and therefore assumed to be a valued feature on N here: (14) a.

agreement: NumP

probe 2 probe 2 probe 1 probe 1

Num PlP [Gd: ] N [Nb: SING] Pl* [Gd: ] [Gd: FEM] n [Nb: A] acqua [+LATT]  probe & match

(14) b.


probe & match

resulting configuration: NumP Num PlP [Gd: FEM] N [Nb: SING] Pl* [Gd: FEM] [Gd: FEM] n (dell’) [Nb: SING] acqua [+LATT]

This makes two subsequent agree operations necessary: first, one between the gender probe of Pl* and N, and second, one between the gender probe of

74

ELISABETH STARK

Num and Pl*.25 During this second agree operation the unvalued feature for number of Pl*, which doesn’t act as a probe, is instantiated. That is, the second agree operation also serves for assigning number to Pl*, like in French. Just as in Spanish, there are two agree operations, but this time two for gender. The main difference to Spanish lies in the fact that morphological number agreement takes place between Num and Pl* and not between Pl* and N. Here we can find a structural explanation for the empirical correlation of unambiguous number marking on the one hand and the absence of a ‘partitive article’ on the other. Thus, if agree has to take place between Num and Pl(*), the former possesses an element marking singular or plural, especially for nominals with a [+LATT] interpretation. Furthermore, only if morphological number cannot be obtained at all via the complex interplay between gender and number features in the structure, as it still can be in Italian (the ending -e e.g. is disambiguated as to its number feature, when the feminine or masculine gender is clear), this ‘partitive’ element will be obligatory, as in French. We can also already understand the second correlation, the one between bare nominals (plurals) in argument position (Italian, Spanish vs. French) and the expression of gender as an overtly coded feature on N. In this case, N takes part in an agree operation between N and Pl(*), and even in one between N and Num or Num and Pl(*) – and after these operations have taken place, it can move upwards in the structure into a functional projection compatible with argumenthood in Romance languages like D. As a further illustration of this, let us have a look at the derivation of Italian indefinite plural nominals: (15) a.

agreement: NumP

probe 2 probe 2 probe 1 probe 1

Num PlP [Gd: ] N [Nb: PLUR] Pl* [Gd: ] [Gd: FEM] n [Nb: A] uova [+LATT]  probe & match

25


probe & match

Please note that, according to the “Feature Visibility Convention” in Radford (2004:289), a deleted feature like the gender feature in Pl(*) after merge and agree with N remains visible to the syntax and PF even if it becomes invisible for the semantic component after deletion, so that the gender feature there can represent a matching feature for the gender probe in Num after merge with PlP.

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

(15) b.

75

resulting configuration: NumP Num PlP [Gd: FEM] N [Nb: PLUR] Pl* [Gd: FEM] [Gd: FEM] n (delle) [Nb: PLUR] uova [+LATT]

Finally, with a countable nominal: (16) a.

agreement: NumP

probe 2 probe 2 probe 1 probe 1

Num PlP [Gd: ] Pl N [Nb: SING] [Gd: ] [Gd: FEM] n [Nb: A] acqua [-LATT]  probe & match

(16) b.


probe & match

resulting configuration: NumP Num PlP [Gd: FEM] Pl N [Nb: SING] [Gd: FEM] [Gd: FEM] n un’ [Nb: SING] acqua [-LATT]

Again, there is an obligatory insertion of the indefinite article in order to derive the [-LATT] property of the whole nominal in Num, but not in Pl because of the gender feature. 4. From Latin to Romance: what happened in the structure? As Latin nouns are inherently specified (and marked) for gender and number and as, at least for concrete denotations, the combination of gender (especially neuter) and number gives relatively clear-cut information on the ‘classification’ of the intended referent (cf. section 2), we assume that the [±LATT] feature is a lexical feature on nouns in Latin, the functional projection

76

ELISABETH STARK

Pl having only a probe for number. Indefinite nominals in Latin thus have the following structure: (17) a.

neuter nouns: NumP Num PlP [Gd: NEUTR] Pl N n unum [Nb: SING] [Nb: SING] [Gd: NEUTR] [+LATT] oleum

(17) b.

feminine nouns: NumP Num PlP [Gd: FEM] Pl N n una [Nb: SING] [Nb: SING] [Gd: FEM] [-LATT] olea

If we abstract away from the Latin case marking system, there is a gender probe in Num in this derivation, as Latin indefinite determiners and quantifiers, including numerals, can be marked for gender, just like it has been assumed for modern Romance languages. However, Pl does not have a [±LATT] feature responsible for ‘semantic pluralization’ – the interpretation of the denotational characteristics of the whole nominal is a feature of N. This is reflected by the fact that there is no direct association in Latin between morphological plural and [+LATT], as has been shown above in section 2. Thus, the number probe and the [±LATT] feature have to be separated in languages without a grammaticalized ‘countability distinction’. Now, imagine a system which loses to a considerable extent its gender- and number marking devices through the (partial) loss of inflectional endings – like Late Latin: not only case, but also number and especially gender, carrying the most important conceptual distinction (‘apprehension’, ‘classification’) besides (in)definiteness for a nominal to be an argument. Of course, the speakers of this system can ‘re-analyse’ its numeral(s), especially “one”, as a ‘countability marker’ and as a singular marker and thus compensate for the inflectional classification system. However, this appealing functional explanation is too simplistic when it comes to the details of syntactic computation:

77

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

(18)

a.

b.

NumP Num [Gd: ] n

NumP PlP

Pl [Nb: ]

N [Nb: A] [Gd: B] [-LATT] ole-

Num PlP [Gd: ] Pl N n [Nb: A] [-LATT] [Nb: ] [Gd: B] unole-

?

Reanalysis defined in its original acceptation as a “change in the structure of an expression or class of expressions that does not involve any immediate or intrinsic modification of its surface manifestation” (Langacker 1977:58) could be seen here in the rearrangement in the feature structure of Num. But there are considerable problems for the modelling of gender and number agreement (unvalued gender feature in Num, no value for the number feature etc.) inside indefinite Late Latin and future Romance nominals, so that the right-hand derivation is simply not acceptable. What exactly happened during the syntactic evolution from Latin to Romance, then? How can we adequately describe and explain the actual syntactic make-up of Romance together with its differences from Latin, which are also topologically relevant? Please have a look at the following structure, which evolved from the left-hand one in the above Latin derivation: (19)

a.

Italian: NumP Num PlP [Gd: FEM] Pl N [Nb: SING] [Gd: FEM] [Gd: FEM] n un’ [Nb: SING] acqua [-LATT]

b.

French: NumP Num PlP [Gd: FEM] Pl [Nb: SING] [Gd: FEM] n une [Nb: SING] [-LATT]

N eau ([o])

The reanalysis of “one” as a countability marker and the whole ‘countability grammaticalization’ that took place after the Latin declension system broke down led to the dichotomic coding of [±LATT] in Pl(*) as the new location for ‘semantic pluralization’, but also as a key projection for the internal gender and number agreement inside Romance nominals.26 In central 26

Cf. the definition of grammaticalization in Roberts & Roussou 2003: “[…] grammaticalization is reanalysis ‘upwards’ along the functional structure. Since movement is always local and upward, categorial reanalysis is also local and upward” (Roberts & Roussou 2003:71).

78

ELISABETH STARK

Romance languages like French and Italian, where number marking is problematic on N (gender marking as well in French), the number feature appears in Pl(*), as has been postulated for English or Hungarian on the basis of independent facts by Heycock & Zamparelli (2003) (see section 3.1 above). This number feature is unvalued and has to receive a value by agree, but cannot be a probe, and for this reason, an additional agree operation has to take place between Num and Pl(*) without directly or exclusively involving N (as in Latin). There is now a valued number feature in Num, together with the original gender probe, while a valued gender feature for French is situated in Pl(*). One, for example a functionalist, could now ask why the assumption of a functional projection, Pl(*), with a [±LATT] feature, would be necessary in order to account for the morphosyntactic facts in modern Romance. The location of a valued number feature in Num seems to be enough, together with the re-interpretation (rather than the re-analysis) of “one” as a countability marker. Now, besides the fact that the assumption of creating (morphological and semantic) plurality in Num and filtering it at the same time has been rejected on independent grounds by Heycock & Zamparelli (see section 3.1 above), it also does not solve our agreement problems, and, more importantly still, it does not allow us to understand the interaction of noun morphology, nominal classification and nominal determination discussed in section 2 above. It would not give evidence for the obligatory insertion of the numeral “one” in Num in indefinite singular countable nominals. Gender seems to play a key role in this story, and, both historically and functionally, it makes perfect sense that it should do so in relation to a head with a [±LATT] feature: gender classes are assumed to have been directly related to extra-linguistic referent classes in Proto-Indo European (cf. Schön 1971, Windisch 1973). Also, in our model the gender feature is situated in Pl(*), either as a probe (Italian) or as a valued feature (French) – the functional projection now responsible for ‘semantic pluralization’ and, more than anything else, for the correct derivation of a ‘mass’ or ‘count’ reading of the (singular) nominal. All of this is only indirectly related to the filtering of the correct quantity that is done in Num (see section 3.1 above). Also, if we compare the structures of (12b) and (14b) above with the one for Spanish (and other Romance languages with unambiguous plural marking) in (20), we can derive the possibility or the need for a ‘partitive article’ in Italian and French:

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

(20)

79

Ibero Romance

NumP Num PlP [Gd: FEM] Pl* N n > [1] (mucha) [Nb: SING] [Nb: SING] [+LATT] [Gd: FEM] agua

This structure, derived from the defective one in Late Latin, now with a [±LATT]-feature in Pl(*) like in English, is fundamentally different from the central Romance one: just as in Italian, two agree operations have to take place, and the result is the possibility of moving N to a higher projection in order to make it available for argument position even if it is bare. In contrast to Italian and French, however, there is no association of gender with the [±LATT] feature in Pl(*), and there is therefore no ‘classification’ system and no explicit marking for the [+LATT] value in the singular case – the ‘partitive article’. This could not be explained if we assumed that gender and number were both situated in Num and functioned as a probe (Spanish), as a partial probe (Italian) or as valued features (French). 5. Conclusion In this paper, I have tried to describe and explain the grammatical change in the evolution of some devices of ‘nominal classification’ in a broad sense from Latin inflectional nominal endings (overt gender and number marking) to Romance indefinite determiners. Comparing the noun morphology and the need for indefinite determination in argument position in French, Italian and Spanish allowed us to observe two main correlations: firstly, the possibility of bare arguments (Italian and Spanish vs. French) correlates with the preservation of a quite clear-cut overt gender marking morphology on the noun (in contrast to Delfitto & Schroten 1991, who see an interdependency with number marking); secondly, unambiguous number marking correlates with the absence of a classifier for ‘mass readings’ (the ‘partitive article’: Spanish vs. French and Italian). If the loss of nominal declensional endings in Latin is analysed as the loss of the marking of the fundamental ‘mass-count distinction’ via noun morphology, the rise of the numeral “one” as an indefinite article indicating ‘countability’, a functional and typologically consistent analysis, is easily motivated. However, only fine-grained formal analyses of the different agreement processes inside indefinite nominals in different Romance languages based on Chomsky’s ‘probe model’ have revealed the syntactic change in Romance nominals which, just as it did for definite nominals with overt D-marking, led to the assumption of new features in Pl(*) (and Num), as a head responsible

80

ELISABETH STARK

now for the semantic [±LATT] feature. This assumption is consistent with other syntactic accounts of the change from Latin to Romance, the great majority of which show a tendency towards the explicit marking of former inherent (semantic or pragmatic) features.27 Pl(*) and its feature make-up seems to play a central role in determining the morphological structure and syntactic behaviour of Romance indefinite nominals: if there is a gender feature, whether unvalued or valued (i.e. interpretable, see also section 3.2 above) in Pl(*) (besides Num), explicit markers of [±LATT] for the singular exist (the indefinite and the ‘partitive’ article) and are merged in Num, inducing an agree operation between Num and Pl(*) for number, and, in Italian, one for gender between Pl(*) and N (thereby valuing the former uninterpretable gender feature; Pl(*) is therefore always the syntactic location with an interpretable gender instance). The question of the interpretability of features would thus be in line with the conception proposed in Pesetsky & Torrego 2004: “an uninterpretable feature must enter an Agree relation with an interpretable counterpart” and the requirement that “every feature must have at least one interpretable instance” (Pesetsky & Torrego 2004:8). This analysis is also quite parallel to the one presented in Picallo 2005, who assumes a functional projection c (‘class’) directly above N in which gender and number features are situated together; Picallo further assumes that “gender declension is the morphological exponent of an abstract interpretable feature in the functional layer of nominal structures” (Picallo 2005:107), this feature being non-overt in Romance (cf. Picallo 2005:109). Now, while Heycock & Zamparelli 2003 concentrate on morphological number in relation to ‘nominal classification’, Picallo 2005 considers gender to be central here, assuming that number features always depend on the expression of gender (which is true for Spanish and Catalan but much less obvious for Italian or French). By separating gender and number agreement processes in indefinite nominals in Romance on the one hand, as I propose here in contrast to Ritter 1993, and observing their interaction on the other, as also proposed here, the derivation of both intranominal morphological agreement and indefinite determination becomes obvious. If Pl(*), which is responsible for morphological number agreement inside the nominal, hosts the gender information in order to cause the uninterpretable gender feature in Num to be valued and deleted, there will be an overt element for this for either case (the indefinite article or the so-called ‘partitive article’). This is done via the number probe or the number feature valued in Num, thus triggering the agree process. If Pl(*) only hosts a number feature, with gender and number features being present on the N as in Spanish, there will be no overt element for the semantic feature [+LATT] in Pl* for the singular. 27

Cf. e.g. Longobardi 2001a for definiteness and D.

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

81

Finally, if N is not involved in an agree operation of any kind, it will not be enabled for syntactic movement to a higher projection, and consequently bare nominals will not be able to appear in argument position, which is true for French but not for Italian and Spanish. The formal account presented only as a sketch in this paper and based on independent assumptions has allowed us to derive and explain the syntactic behaviour of Romance indefinite nominals in opposition to the morphosyntactic regularities of Latin, and this in a much more detailed way than any (diachronic) functionalist description of the facts would have been capable of. Functional considerations are important for uncovering the general cognitive motivations for innovations in languages and changing systems; they are related to the interaction of the interfaces with the syntax, but in principle they are too general and extra-linguistically driven to be able to explain morphosyntactic regularities in language change related to the ‘narrow syntax’.

REFERENCES Chierchia, Gennaro. 1998. “Reference to kinds across languages”. Natural Language Semantics 6:4. 339-405. Chomsky, Noam. 1995. The minimalist program. Cambridge, Mass.: The MIT Press. ----------. 2000. “Minimalist Inquiries: The framework”. Step by Step. Essays on Minimalist Syntax in Honor of Howard Lasnik, ed. by Roger Martin, David Michaels & Juan Uriagereka, 89-151. Cambridge, Mass: The MIT Press. ----------. 2001. “Derivation by phase”. Ken Hale: A Life in Language, ed. by Michael Kenstowicz, 1-52. Cambridge, Mass: MIT Press. ----------. 2005. On phases. Ms. Delfitto, Denis & Jan Schroten. 1991. “Bare plurals and the number affix in DP”. Probus 3:2. 155-185. Gil, David. 1987. “Definiteness, noun phrase configurationality, and the countmass distinction”. The representation of (in)definiteness, ed. by Eric J. Reuland & Alice G.B. ter Meulen, 254-269. Cambridge, Mass.: MIT Press. Givón, Talmy. 1981. “On the development of one as an indefinite marker”. Folia Linguistica Historica II. 35-53. Harley, Heidi & Elizabeth Ritter. 1999. Meaning in Morphology: Motivating a feature-geometric analysis of person and number. MS, University of Calgary & University of Pennsylvania. ----------. 2002a. “Person and number in pronouns: a feature-geometric analysis”. Language 78. 482-526. ----------. 2002b. “Structuring the bundle: a universal morphosyntactic feature geometry.” Pronouns – Grammar and Representation, ed. by Horst J.

82

ELISABETH STARK

Simon & Heike Wiese, 23-39. Amsterdam & Philadelphia: John Benjamins. Harris, James W. 1992. “The form classes of Spanish substantives”. Yearbook of Morphology 1991, ed. by Geert Booij & Jaap van Marle, 65-88. Dordrecht: Kluwer. Haspelmath, Martin. 1999. “Optimality and diachronic adaptation”. Zeitschrift für Sprachwissenschaft 18:2. 180-205. Heycock, Caroline & Roberto Zamparelli. 2003. Friends and colleagues: Plurality, coordination, and the structure of DP. Ms, University of Edinburgh & Università di Bergamo. Hofmann, Johann Baptist. 1997 (2nd reprint of the first edition of 1965/1972). Lateinische Syntax und Stilistik: mit dem allgemeinen Teil der lateinischen Grammatik. Neubearbeitet von Anton Szantyr. München: Beck. Iturrioz Leza, José Luis. 1986. “Individuation and determination III: The concept of verbal plurality and the pluralization of abstractives”. Función 1:2. 201-308. Korzen, Iørn. 1996. L’articolo italiano fra concetto ed entità. Uno studio semantico-sintattico sugli articoli e sui sintagmi nominali italiani con e senza determinante – con un’indagine particolare sulla distribuzione del cosidetto ‘articolo partitivo’. 2 vols. Kopenhagen: Museum Tusculanum Press. Kühner, Raphael & Carl Stegmann. 31955. Ausführliche Grammatik der lateinischen Sprache. Satzlehre. Erster Teil. Leverkusen: Gottschalksche Verlagsbuchhandlung. Langacker, Ronald W. 1977. “Syntactic Reanalysis”. Mechanisms of Syntactic Change (Austin), ed. by Charles N. Li. 57-139. Santa Barbara: University of California. Lehmann, Christian. 1991. “The Latin nominal group in a typological perspective”. New studies in Latin linguistics. Selected papers from the 4th international colloquium on Latin linguistics, Cambridge, April 1987, ed. by Robert Coleman, 203-232. Amsterdam & Philadelphia: John Benjamins. Link, Godehard. 1983. “The logical analysis of plurals and mass terms”. Meaning, use and interpretation of language, ed. by Rainer Bäuerle, Christoph Schwarze & Arnim von Stechow, 302-323. Berlin: de Gruyter. ----------. 1991. “Plural”. Semantik. Ein internationales Handbuch der zeitgenössischen Forschung, ed. by Arnim von Stechow & Dieter Wunderlich, 418-440. Berlin & New York: Mouton de Gruyter. Löbel, Elisabeth. 1993. “On the parametrization of lexical properties”. The parametrization of universal grammar, ed. by Gisbert Fanselow, 183-199. Amsterdam & Philadelphia: John Benjamins. Longobardi, Giuseppe. 2001a. “The structure of DPs: some principles, parameters and problems”. The handbook of contemporary syntactic theory,

THE ROLE OF THE PLURAL SYSTEM IN ROMANCE

83

ed. by Mark Baltin & Chris Collins, 562-603. Oxford & Cambridge, Mass.: Blackwell. ----------. 2001b: “How comparative is semantics? A unified theory of bare nouns and proper names”. Natural Language Semantics 9:4. 335-369. Meisterfeld, Reinhard. 1998. Numerus und Nominalaspekt. Eine Studie zur romanischen Apprehension. Tübingen: Niemeyer. Mensching, Guido. 2005. “Sonden und Phasen in romanischen Sprachen”. Deutsche Romanistik – generativ, ed. by Georg Kaiser, 123-143. Tübingen: Narr. ---------- & Eva Remberger. 2006. “Probes: lack of agreement in romance”. Studies on Agreement, ed. by João Costa & Maria Cristina Figueiredo Silva, 173-201. Amsterdam & Philadelphia: Benjamins. Morani, Moreno. 2000. Introduzione alla linguistica latina. München: Lincom. Munn, Alan & Cristina Schmitt. 2005. “Number and indefinites”. Lingua 115. 821-855. Pesetsky, David & Esther Torrego. 2004. The Syntax of Valuation and the Interpretability of Features. Unpublished MS. Boston: MIT. Picallo, Carme. 1991. “Nominals and nominalizations in Catalan”. Probus 3. 279-316. ----------. 2002. . “Abstract agreement and clausal arguments”. Syntax 5:2. 116147. ----------. 2005. “Some notes on grammatical gender and l-pronouns”. Specificity and the evolution/emergence of nominal determination systems in Romance. Selected papers from the international workshop NEREUS II, October 2004 in Berlin (= Konstanzer Arbeitspapiere zur Sprachwissenschaft, 119), ed. by Klaus von Heusinger, Georg Kaiser & Elisabeth Stark, 107-121. Konstanz: Universität Konstanz. Radford, Andrew. 2004. Minimalist Syntax. Exploring the structure of English. Cambridge: CUP. Ritter, Elizabeth. 1993. “Where’s Gender?”. Linguistic Inquiry 82. 146-150. ----------. 1995. “On the syntactic category of pronouns and agreement”. Natural Language and Linguistic Theory 13. 405-443. Roberts, Ian & Anna Roussou. 2003. Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Schön, Ilse. 1971. Neutrum und Kollektivum. Das Morphem -a im Lateinischen und Romanischen. Innsbruck: Institut für Vergleichende Sprachwissenschaft der Universität Innsbruck. Schroten, Jan. 2001. “L’absence de déterminant en espagnol”. Typologie des groupes nominaux, ed. by Georges Kleiber, Brenda Laca & Liliane Tasmowski, 189-203. Rennes: Presses Universitaires de Rennes. Seiler, Hansjakob. 1986. Apprehension. Language, object, and order. Part III: The universal dimension of apprehension. Tübingen: Narr.

84

ELISABETH STARK

Stark, Elisabeth. 2002. “Indefiniteness and specificity in old Italian texts”. Journal of Semantics 19. 315-332. ----------. 2005. “Explaining article grammaticalization in Old Italian”. Romanistische Korpuslinguistik II. Korpora und diachrone Sprachwissenschaft, ed. by Claus D. Pusch, Johannes Kabatek & Wolfgang Raible, 455-468. Tübingen: Narr. ----------. 2006. Indefinitheit und Textkohärenz. Entstehung und semantische Strukturierung indefiniter Nominaldetermination im Altitalienischen (= Beihefte zur Zeitschrift für romanische Philologie, 336), Tübingen: Niemeyer. ----------. 2007. “Gender, number, and indefinite articles – about the ‘typological inconsistency’ of Italian”. Evolution and function of nominal determination (= Studies in Language and Linguistic Typology), ed. by Elisabeth Stark, Werner Abraham & Elisabeth Leiss, 49-71. Amsterdam & Philadelphia: John Benjamins. ----------. Forthcoming. “Typological correlations in nominal determination in Romance”. Proceedings of the Copenhagen Determination Symposium (August 2004), ed. by Alex Klinge et al., Amsterdam & Philadelphia: Benjamins. Van Kemenade, Ans & Nigel Vincent. 1997. Parameters of morphosyntactic change. Cambridge: Cambridge University Press. Vincent, Nigel. 1997. “The emergence of the D-system in Romance”. Van Kemenade & Vincent 1997. 149-169. Windisch, Rudolf. 1973. Genusprobleme im Romanischen. Das Neutrum im Rumänischen. Tübingen: Narr. Zamparelli, Roberto. 2004. “Manufacturing Plurality”. Talk given at the Copenhagen Determination Symposium, 26.-28. August 2004. ----------. 2005. “The structure of (in)definiteness”. Lingua 115. 915-936.

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE*

MARIA GOLDBACH University of Oxford This article is concerned with the role of the morphological paradigm in syntactic change. The discussion focuses on syntactic change in infinitival complements and the distribution of these complements in the context of infinitive embedding verb phrases. I argue that the syntactic structure of the infinitival complements is closely related to the co-occurrence of inflectional exponents on the infinite verb forms. The presence of inflectional exponents on infinitive verb forms determines the quality of the functional syntactic category Infl in the infinitival complements. Syntactic changes in the structure and/or in the distribution of infinitival complements originate from paradigm alterations concerning the inflectional exponents of the infinite verb forms.

1. Introduction In the present article I compare infinitive constructions in Latin with those in Old French, Old Spanish, and Old Italian. This study does not cover participle and gerund constructions but leaves them for future research. The texts on which I have based my analysis are the following: Cicero’s letters to his close friend Atticus is the textual basis for Classical Latin. The letters originate from the 1st century B.C. The Late Latin period I analyse on the basis of the historical work of Gregory of Tours, written in the second half of the 6th century A.D.1 Old French is examined on the basis of the chronicle of Villehardouin (dating from 1210), and Old Italian on the basis of the chronicle of Dino Compagni (composed 1310-1312).2 Old Spanish is analysed *

For very valuable and supportive comments I would like to thank an anonymous reviewer and Paul O’Neil. 1 Concerning the closeness of Cicero’s speech in his letters to spoken Latin, compare Jäger (1986). As regards Late Latin, Bonnet (1890:29) argues that Gregory’s mother tongue was Latin: “Aucun mot, d’ailleurs, dans tous ses écrits, ne ferait deviner qu’il eût jamais parlé une autre langue ou qu’il eût appris le latin comme une langue étrangère. Il répète souvent qu’il a peu étudié la grammaire, mais il ne dit jamais qu’il ait mal appris le latin.” Compare also Lot (1931), who assumes that Gregory’s spoken Latin did not differ from his written Latin. Altogether, it seems feasible to compare the Classical Latin of Cicero’s letters to Atticus with the historical writings of Gregory of Tours. 2 Compagni’s chronicle is preserved in a manuscript dating from the second half of the 15th century. For that reason it cannot be classified as a typical representative of Old Italian. However, my primary concern is to keep constant the type of text, i.e. chronicles. There are

86

MARIA GOLDBACH

on the basis of the Primera crónica general de España, composed between 1272-1290 in redaction of the King Don Alfonso X. (el sabio) and subsequently of his son Sancho IV. With regards to grammatical change and its explanation, I argue that at least some syntactic developments can be traced back to morphological changes and that these may be explained by the quality of an abstract, morphologically motivated, syntactic category. In this respect I follow the fourth generative model described by Detges & Waltereit (this volume) in the introduction: the Inertia Theory of Longobardi (2001). Regarding the infinitive constructions, I propose that changes in the infinitive paradigm (in Latin) or in the pronoun paradigm (in Old French) are related to the development of the syntactic distribution of these constructions. The development of the morphological paradigm gives rise to the weakening, in the development of Latin, or strengthening, in the development of French, of the abstract syntactic category Infl(ection), which in turn affects the syntactic distribution of the respective infinitive construction.3 Thus, I assume that the syntax of Infl depends on variations in the morphological paradigm, i.e. the correlation between the morphological paradigm and the syntactic distribution of the infinitive structures is close insomuch that one can predict a change in this distribution whenever a paradigm alteration occurs. More specifically, I interpret the predictability of a grammatical change as a piece of the explanation towards the change itself. Admittedly, this cannot be the complete explanation. For example, at present, I have no answer to the question of why the paradigm alterations occur, and an answer to this question would certainly contribute to a more far-reaching explanation. The benefit of predictability and of (relatively) strong correlations is that they narrow down the search area for the explanation of an observed phenomenon. Moreover, they facilitate the formulation of abstract, that is to say, grammar theoretical generalisations, which are not only valuable for historical or diachronic linguistics but also for synchronic syntactic theory. In the first section of this article I present the development of the Latin Accusativus-cum-Infinitivo construction (henceforth ACI) and I propose an two other Old Italian chronicles, both composed in the early 14th century. Both texts are much shorter than Compagni’s chronicle and, probably as a consequence of this, some constructions fail to appear. Moreover, one of them, the Cronica fiorentina (from an anonymous author) is preserved in fragmentary manuscripts, cf. Hartwig (1875). Since the principle aim of the present paper is not to describe the Old Italian infinitives but rather to compare infinitive structures cross-linguistically and, thus, to discover correlations between various phenomena, I consider the use of Compagni’s chronicle linguistically justifiable. 3 I fully concur with the view of Anderson (1982), Carstairs (1987, 1988), Stump (2001), etc. that the paradigm is an important object of morphological description and explanation. In this article, I emphasize that this is not only true for the morphological component but that paradigm alterations also have repercussions on the syntactic development, and I consider paradigm alterations as morphological changes.

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

87

explanation of its development. The second section compares the structure and the development of medieval French infinitives with those of Old Italian and Old Spanish infinitives and, finally, I contrast the development of the Latin ACI with the development of the medieval French infinitives. I conclude the paper with a discussion of the problems involved in seeking an explanation for grammatical change. 2. The development of the Latin ACI Classical Latin infinitive forms were morphologically specified for tense and voice as exemplified in table 1. active present amar-e “to love” perfect amaviss-e “to have loved” future amaturum esse “to will love”

passive amar-i “to be loved” amatum esse “to have been loved” amatum iri “to will be loved” 4

Table 1: The morphological paradigm of the Latin infinitive amare “to love”

Additionally, the infinitive can express modality, which is instantiated by an analytic form; that is, the participle future active plus the infinitive perfect active of the verb esse (“to be”): amaturus fuisse “will love” + “to have been” comes to mean “would have loved”, in this way the irrealis modality of the event is expressed (c.f. Priem 1889). In Cicero’s letters to Atticus all of the forms in table 1 appear in the ACI. The ACI is a non-finite complement sentence without complementizer, whose subject bears accusative case and whose verb appears as an infinitive, cf. (1ag). The Latin ACI corresponds to finite complement sentences with the complementizer che, que “that” in the Romance languages. (1) a. b.

4

Nam illi ita negant vulgo, ut mihi se debere dicant. (pres. act., Att. 1.1.1) “For in general those reject [it] by saying that they are indebted to me.” Nostris rationibus maxime conducere videtur Thermum fieri cum Caesare. (pres. pass., Att. 1.1.2)

In Classical Latin, the forms of the infinitive for the perfect passive and for future tense were analytic. Nevertheless, they form a morphological paradigm with the three synthetic forms. I suggest that the relevant morphological property licencing the existence and guiding the distribution of the ACI is the paradigmaticity of the infinitive forms, irrespective of their analytic or synthetic character. In Classical Greek the corresponding infinitive paradigm is completely synthetic, cf. Kieckers (1926), and there are also ACI complements. It seems to me that the paradigm coverage (in the sense of Carstairs 1988) concerning the morphosyntactic features of voice and tense is more significant for the distribution of the ACI than the question of analyticity or syntheticity of the infinitival paradigm. In fact, my aim is to derrive the strength or the weakness of the syntactic functional category Infl in the ACI from the degree of paradigm coverage in the infinitive paradigm, cf. Goldbach (2003).

88

MARIA GOLDBACH

c.

d.

e. f.

g.

“Concerning our [personal] matters [it] seems to be most conducive that Thermus is elected [consul] together with Caesar.” Quodsi voles in me esse durior, ambitionem putabis mihi obstitisse. (perf. act., Att. 1.1.4) “If [you] want to deal more severely with me, [you] will believe that [my] ambition has hindered me.” Abs te peto, ut mihi hoc ignoscas et me existimes humanitate esse prohibitum (perf. pass., Att. 1.1.4) “[I] beseech you that [you] forgive me for this and that [you] believe that I was hindered by appositeness.” Eius rei consolationem ad te L. Saufeium missurum esse arbitror. (fut. act., Att. 1.3.1) “[I] suppose that L.S. will send console to you.” Sed ductus odio properavit rem deducere in iudicium, cum illum plumbeo gladio iugulatum iri tamen diceret. (fut. pass., Att. 1.16.2) “But driven by odium [he] hurried to inform the court of this case when he said this one will be killed even with a leaden sword.” Demonstravi haec Caecilio simul et illud ostendi, si ipse unus cum illo uno contenderet, me ei satis facturum fuisse. (irrealis, Att. 1.1.4) “[I] demonstrated this to C. and at the same time [I] made obvious that I would have satisfied him if only this one had struggled with that one.”

That is, from these examples it can be observed how the richness of forms in the infinitive paradigm permits the expression of the functional features for voice, tense, and modality in the Latin infinitive complement sentence; i.e., in the ACI. In all of the ten books of Gregory of Tours’ Histories, two of the Classical Latin forms fail to appear: there are no constructions with the future passive infinitive (1f) or with the irrealis (1g). Both forms start to diminish in the texts of the post classic period, cf. Perrochat (1932). This would imply that the functional categories voice and modality are weakened or even disappear in the ACI-constructions of Late Latin. The question to be addressed now, then, is whether the weakening of the infinitival paradigm has bearing on the distribution of the ACI. In order to go about answering this question I analysed the occurrences of the ACI in the 1st book of letters to Atticus (about 683 sentences) and in the 2nd book of Gregory’s Histories (about 931 sentences). Table 2 summarizes the distribution of the ACI. The two texts differ regarding both the token and the type frequency. In order to explain these differences, first, it ought to be noted that the morphological paradigm of the infinitive in Classical Latin is complete with regard to tense and voice and that modality can also be expressed. To put it into the terms of generative grammar or especially into the terms of the Minimalist Program (cf. Chomsky 1995), I propose that in Classical Latin the quality of the morphological paradigm of the infinitive licences a strong functional category Infl(ection) (consisting of the checking capacity for tense, voice, and modality features) in the syntactic structure of the ACI.

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

Cicero’s letters to Atticus

ACI5 present active (amare) present passive (amari) perfect active (amavisse) perfect passive (amatum esse) future active (amaturum esse) future passive (amatum iri) irrealis (amaturum fuisse) total (token frequency) number of the ACI-embedding verb phrases (type frequency)

89

Gregory of Tours’ Histories

95 22 42 27 19 2 1

48 27 22 12 7 0 0

208

116

66

55

Table 2: The distribution of the ACIs

Furthermore, I suggest that the ample distribution of the ACI may be attributed to the strength of the Infl-category. From the fact that the morphological paradigm of the infinitive has become fragmentary in Gregory’s Histories I want to deduce that the strength of the syntactic Infl-category in the ACI is in decline. From this in turn follows that the distribution of the ACI recedes with respect to token and type frequency. The intended explanation consists of correlating a morphological property, the richness of the infinitive paradigm, with the syntactic distribution of the infinitive construction by means of the quality of an abstract syntactic category; i.e., its strength or, respectively, its weakness. In the proposed explanation I have neglected some factors. It is true that in Classical Latin the ACI was the standard form of a complement sentence; however, there also existed finite complement sentences. The finite complements were mostly in the subjunctive mood: in the 1st book of Cicero’s letters to Atticus there are 12 finite complements in the indicative mood vs. 108 in the subjunctive mood. It has been assumed that the Latin ACI is superseded by finite complement sentences with the complementizers quod, quia, quoniam, cf. the discussion and synopsis in Cuzzolin (1994), Mayen (1889), Perrochat (1932). To investigate this assumption, I compared the occurrences of the finite complement sentences in both texts. Table 3 reports the results of this investigation. If we examine the development of the distribution of the finite complement sentences we cannot detect that the number of these constructions increases in Gregory’s Histories to the disadvantage of the ACI.6 5

I included to the active constructions also those that contain a deponent verb because deponents syntactically behave like active verb forms (in that they can take a direct object) even though they turn up in passive morphology. 6 In the whole text of Gregory’s Histories the ACI-complements outnumber the finite complement sentences introduced by quod and quia, compare also Bonnet (1890:666).

90

MARIA GOLDBACH

finite complement sentences indirect interrogative ut-complementizer ne-complementizer null complementizer quin-complementizer utrum-complementizer quod-complementizer quia-complementizer total

Cicero’s letters to Atticus

Gregory’s Histories

36 42 13 21 3 0 5 0

23 48 13 3 0 1 14 8

120

110

Table 3: Occurrences of finite complement sentences

There are, in fact, some complement sentences embedded by verbs of saying and thinking with quod- and quia-comlementizers, which do not correspond to the classical usage. In Classical Latin quia introduces adjunct sentences (in the sense of “because”) but not complement sentences, and verbs of saying and thinking take an ACI as their complement. (2) a.

b.

[…] cum ait, quod Constantinus […] Constantium filium ad se de Hispaniis venire iussisset, ita disseruit: (Greg. Hist. 2,9, 19-21) “[…] when he says that C. gives the order that his son C. should come to him from Spain, he tells [it] in the following way” Vere enim cognovi, quia Deus non inridetur (Greg. Hist. 2,3, 17-18) “I have really recognized that God should not be derided”

Nevertheless, it seems to me, so far, that the development of the ACI can be correlated more naturally with the development of the infinitive paradigm than with the development of the finite complement sentences. In the next section I compare infinitive complements in medieval French, Spanish, and Italian. I seek to illustrate that the development of the infinitive in medieval French is contrasted with the development of the Latin infinitive, insomuch as it seems to develop in the opposite direction, in that a weak Inflcategory in Old French becomes strong in Middle French. 3. The evolution of French infinitives In order to expound the character of the medieval French infinitive, I compare the infinitive complements in Villehardouin’s chronicle with those of Dino Compagni’s chronicle and with those of the Old Spanish chronicle.7 ACIconstructions, like in Latin, cannot be found in either chronicle. In Compagni’s chronicle there are six sentences that resemble the ACI insofar as the 7

The Old French and the Old Italian chronicles have been studied in their entirety, in contrast to the Spanish chronicle, in which the first 1022 sentences of the first volume and 1009 sentences of the second volume of the edition of Menéndez Pidal (1977) have been analysed.

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

91

embedded infinitive has its own subject, cf. (3a). Similar constructions are found in the Old Spanish chronicle, cf. (3b). But in contrast to the Latin ACI, the subjects in Old Italian and in Old Spanish do not bear the accusative case. (3) a.

b.

e essendo l'uso della terra a simili raunate i cittadini sedere basso in su stuoie di giunchi (DC 1.20.101) (Old Italian) “and [it] being the custom of the city at similar meetings [that] the citizens sit down on their straw mats.” Grand cosa fue, Eneas, duna muger poder se deffender contra tantos enemigos (PCG 42a41-42) (Old Spanish) “It was a considerable affair, Eneas, that a woman could defend herself against so many enemies”

It is to be noted that in Villehardouin’s chronicle there has been no trace found of such a type of sentence with a subject in the infinitive complement. The infinitive paradigm of Old French, Old Spanish, and Old Italian, compared to the Latin paradigm, is fairly reduced. Apart from the present active infinitive there are only infinitive forms for perfect active (cf. (4)) and for present passive (cf. (5)). (4) a.

b.

c.

(5) a.

b. c.

il perché confessò avere ricevuta una testimonanza falsa per messer Niccola Acciaiuoli (DC 1.19.92) (Old Italian) “therefore, he confessed to have received a false testimonial by M. Niccola A.” Et dirent que onques mes n'avoit esté si hardi home qui osast avoir desfié l'empereeur de costentinoble en sa chambre. (Villeh. 215) (Old French) “And [they] said that there has never been such a bold man who dared to have defied the emperor of Constantinople in his chamber.” E ¿quando cuedas tu auer fecha cibdat que semeie a Carthago ...? (PcgE40a, 31-33) (Old Spanish) “And when do you believe to have founded a town that resembles Carthago …?” ... spronò il cavallo contro a messer Corso credendosi esser seguìto da' Cerchi (DC 1.20.104) (Old Italian) “and he spurred on the horse against M. Corso, opining to be followed by the Cerchi” conment porroit estre prise tele vile par force (Villeh. 77) (Old French) “how this city could be conquered by force” … assi que apenas puede seer sabudo el comienço de los que la poblaron (PcgE 4a, 19-20) (Old Spanish) “… so that the beginning of those who colonised it can hardly be known”

These examples contrast with those of the Latin paradigm expounded in table 1 in that the infinitive paradigms of the three Old Romance languages under consideration cover neither a complete tense nor a complete voice axis. Moreover, the infinitive constructions cannot express modality; i.e., there is no Old Romance correlative to sentences like (1g) in the infinitive. Next, we examine the distribution of infinitive complements, cf. table 4.

92

MARIA GOLDBACH

Villeh. Number of infinitive complements (token frequency) Number of verb phrases embedding infinitive complements (type frequency) Number of analysed sentences

DC

PcgE 1st part

PcgE 2nd part

573

605

328

267

26

88

42

39

2924

2383

1022

1009

Table 4: Distribution of infinitive complements in the chronicles of Villehardouin (Villeh.) and Dino Compagni (DC) and in la Primera crónica general de España (PcgE)

What is noticeable here are the differences in type frequency: Compagni’s chronicle contains more than three times as many infinitive embedding verb phrases than the French chronicle, and even the two sections of the Spanish chronicle, despite the fact that each part is nearly one third as large as the French chronicle, outnumber the Old French infinitive embedding verb phrases. As we have seen, this can obviously not be attributed to a difference concerning the infinitival tense and voice morphology since the three Old Romance languages are alike in this respect, cf. (4) and (5). I now turn my attention to other inflection-like elements co-occurring with infinitives, in particular object clitics. Table 5 illustrates the occurrences of object clitics in the infinitive. We observe that in Villehardouin’s chronicle there are no object clitics like me, te, le, la, li, se, les, leur, y, en, “me, you, him, her, to him/ to her, himself/herself/themselves, them, to them, there, of it”, in infinitive constructions, neither pre- nor post-verbally. This is all the more true for complement infinitives (Kok 1985), as well as for adjunct infinitives (cf. Goldbach 2006). The infinitives co-occur solely with unbounded (strong) object pronouns like nous, vous, lui, “us, you, to him” etc8. 8

An anonymous reviewer of my study disbelieves my assumption that Old French infinitives are incompatible with object clitics. He suggests that the unbounded object pronoun forms cooccurring with infinitives are object clitics that, due to the special stress system of Old French, in certain positions where the stress falls on the clitic, have to assume a form different from the form they take in an unstressed position. He argues that the position after prepositions like pour “in order to” is stressed and for that reason the object clitics me, te, se, and le appear in their stressed forms moi, toi, soi, and lui. As a consequence, we find pur lui sálúer (QLR 98,9) “in order to salute him” instead of *pur le sálúer (saluer governs a direct object as can be seen in the following sentence: Cume Cusaï, li amí David, vint devant Absalon, saluád le en ceste maníere “when Huschai, the friend of David, came to Absalom [he] saluted him this way ...” QLR 89,16). This assumption leaves several questions open: Why do we find in the Quatre Livre des Reis seven constructions of the type 1 pur lui sálúer and only two of the type 2 pur

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

Number of analysed sentences Object clitics in the infinitive

93

OF QLR* (1175)

OF Villeh. (1210)

MF Joinv. (1305)

OS PcgE1 (1270-90)

OS PcgE2 (1270-90)

OI DC (1310)

5836

2924

4113

1022

1009

2383

4

0

2

38

46

125

*QLR. = Li Quatre Livre des Reis; Villeh. = Villehardouin’s chronicle; Joinv. = Joinville’s chronicle, PcgE1 = Primera crónica, 1st part; PcgE2 = Primera crónica, 2nd part; DC = Dino Compagni’s chronicle. Table 5: Occurrences of object clitics in the infinitive9 sálúer le (QLR 95,15; 58,15)? This being the preferred structure in Old Italian and Old Spanish, cf. FN 6. Furthermore, why is this type 2 completely missing in the chronicles of Villehardouin and Joinville? Even if we want to assume that the phonological rule that diphthongises me, te, and se to yield moi, toi, and soi (cf. Rheinfelder 1953:69f.; in the QLR: mei, tei, sei) also derives le to lui in a stressed position (which I am not convinced of since one would have to explain why the forms *loi / *lei do not appear), it is not feasible to trace back the usage of els “them” in the example é pur els querre (QLR 17,3) “and in order to look for them” to the same phonological rule since the respective object clitic is les. In the finite sentences of the Quatre Livre des Reis, the postverbal pronoun for 1st person singular appears very often as mei (moi). As a consequence, the postverbal position in finite sentences should count as a stressed position where the clitic me must appear as mei (moi). But this replacement is not systematic; there are also examples with postverbal me (QLR 159,37; 165,35; 159,40). The same is true for the postverbal reflexiv pronoun; we also find finite sentences with postverbal se (QLR 34,39) instead of sei (soi). It seems to me, thus, that the replacement of me, te, and se by mei, tei, and sei (moi, toi, soi) in the finite context is not systematic and that it is not a model for the infinite context. In addition, in finite sentences, postverbal le, la (3rd sg dir. obj.) is never replaced by the unbounded object pronoun form lui (exept for coordination, QLR 16,18; 47,26; 66,20). Postverbal lui in finite sentences replaces 3rd sg ind. obj. mas. li. The distribution of lui in infinitive constructions seems to be governed by its own conditions. I surmise that in the Quatre Livre des Reis the pronoun forms mei, tei, sei, lui, etc. occurring in infinitive constructions do not result from the productive application of the aforementioned phonological rule to the object clitics me, te, se, and le, but they are instead the unbounded object pronouns by that time; i.e., that the unbounded object pronouns constitute a stable paradigm. Additionally, the example that the anonymous reviewer cites ne mei si encreper (QLR 34,29) (cf. FN8 below) is evidence that mei cannot be an object clitic, since if it were an object clitic it would not be separated from the infinite verb encreper by the adverb si. 9 In contrast to the Old Italian and to the Old and Middle French texts, the Old Spanish chronicle exhibits the phenomenon of interpolation; cf. Ramsden (1963), Rivero (1992) etc. (i) … en tal guisa ques non entendien unos a otros (PgcE 4b, 44-46) “… in such a way that [they] did not understand each other” In interpolation contexts, the object clitic (in the example above the reflexive s(e)) is separated from its verb (above: entendien) by the negation (non), or by subject pronouns like el “he”, tu “you”, etc., or by adverbs like y “there” and agora “now”. This phenomenon predominantly occurs in embedded contexts and rarely in matrix sentences (in the parts of the Old Spanish chronicle that I analysed there is no example of interpolation in matrix sentences). Interpolation never applies post-verbally; it can only be found in preverbal position. As Rivero 1992 states,

94

MARIA GOLDBACH

By contrast, Compagni’s chronicle features 125 infinitives with object clitics, and the sections of the Spanish chronicle contain 38 and 46 infinitives with object clitics. In the majority of cases the object clitics are in postinfinitival position in Old Italian and Old Spanish.10 One could surmise that it is a peculiarity of Villehardouin’s text or mere chance that there are no object clitics in infinitives. Therefore, I have studied another Old French text: Li Quatre Livre des Reis (dating from 1175) that contains four infinitive constructions with object clitics (pre-infinitival QLR 58,15; post infinitival QLR 95,15; QLR 108,16; QLR 72,23-73,23). In the early Middle French chronicle of Joinville (dating from 1305) there are two infinitives with object clitics (both in preverbal position, Join. 606, 359). It seems, therefore, that Old French infinitives have a tendency to avoid object clitics, whereas Old Italian infinitives prefer them. Further examples from other Old Italian texts are given by Cardinaletti (2005) and Egerland (2005). The question can be asked if more differences between Old French infinitives and the infinitive constructions of the other two Old Romance languages exist. This, I shall show, is the case with the possibility of negating the infinitive construction. Whereas Old Italian and Old Spanish infinitive constructions can be negated independently from the matrix sentence, this is not possible for Old French infinitives, cf. Moignet (1965). (6) a.

b.

c.

d.

per poterli levare da oste ... e non li dare la terra (DC 3.17.104) (Old Italian) “in order to prevent them from fighting […] and not to give them the city” (literally: “territory”) proposimi observarli, e non lasciare fare l'uficio contro alle leggi (DC 2.10.44) (Old Italian) “[I] resolved to observe them and not to let officiate against the laws” [...] que acquello auien a fazer e non se tirar dello por ninguna manera (PcgE 37b50-51) (Old Spanish) “[…] that [they] had to do this and not to avoid it by any means” […] por non auer guerra con los moros […] (PcgE343b) (Old Spanish) “in order to be not at war with the Moors”

this phenomenon disappears after the middle of the fifteenth century. In the analysed parts of the Primera crónica general de España, I found eight instances of interpolation in the first part of the chronicle, but there are also 19 sentences which satisfy the context of interpolation but in which interpolation does not take place. In the second part of the chronicle there are 14 examples of interpolation vs. 11 contexts of interpolation without interpolation. It seems to me that interpolation is not so productive in the Primera crónica general de España. 10 In Compagni’s chronicle there are two constructions with preverbal object clitics, cf. DC 3.17.104, 2.34.159, in the Old Spanish chronicle there are seven examples with preverbal object clitics in the first part, cf. PcgE 9b, 16a, 18b, 20a, 21b, 22a, 23a; and in the second part there are four examples of this type, cf. PcgE 323b, 325b, 332b, 333a. Compare also Schmidely (1998).

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

95

In order to confirm that the negation of the infinitive is not an accidental eccentricity of Compagni’s text I provide the examples in (7) from the Old Italian chronicle of Paolino Pieri (dating from 1305).11 (7) a. b.

dappoi che voi sete fermi di non servircene (Paolino Pieri 4, 27-28) (Old Italian) “since you are bent on not fulfilling our request” Et allora ordinaro, et fecerne decreto per istatuto di non lasciarla mai rifare (Paolino Pieri 5, 18-20) (Old Italian) “and then they gave order and they enact an ordinance by constitution to never let it reconstruct”

In the French area instances of negated infinitives as in (6, 7) start to emerge in texts composed by the end of the 13th century (cf. Moignet 1965:54). However, in Joinville’s chronicle I could not find an equivalent example.12 Table 6 summarizes the differences between the Old French infinitive complements on the one hand and the Old Italian and the Old Spanish constructions on the other hand. Villehardouin Subjects in the infinitive Object clitics in the infinitive Negation in the infinitive

no no no

Compagni

Primera crónica

yes yes yes

yes yes yes

Table 6: The differences between the chronicles with regard to the infinitive complements

That is, the Old Italian and Old Spanish infinitive constructions exhibit characteristics that in Old French are shared by finite constructions only. How can we grasp the observed differences on a more abstract level? Semantically, the nature of negation is the negation of a proposition (cf. Kneale & Kneale 11

Compare also Brunetto Latini, La Rettorica (dating from 1260) 9,18; 53,2; 60,9; 62,14, etc. The anonymous reviewer points out that in Quatre Livre des Reis infinitives can be negated and cites the following example: (i) N’i ad parole dunt te estuce curecher ne mei si encreper (QLR 34,29) “There is no word that requires [you] to enrage you nor to reprimand me so much” This type of negation, ne, is compatible with nominals: (ii) N’ai beü ne vin ne él par unt l’um se poïsse enívrer (QLR 5,15) “I did not drink no vine nor anything by which one can get drunk” Therefore, the co-occurrence of ne with the infinitive encreper in (i) does not prove that the Old French infinitive is negated in the same way as the Old Italian and Old Spanish infinitives in (6). In the text of the Quatre Livre des Reis, there are eight sentences with a negated infinitival form, cf. (iii) (also: QLR 38,4; 39,24; 45,15; 64,22; 82;24; 82,32; 201,8): (iii) N’esguarder pas á sa chiere ne á sun corsage! (QLR 31,7) “Don’t look at his face nor at his body!” All these instances are imperatives, but not infinitival sentences. In the Quatre Livre des Reis, there are no sentences comparable to the Old Italian and Old Spanish examples in (6) above. 12

96

MARIA GOLDBACH

1962 passim), and a proposition is a set of predications. That is, if the infinitive construction can be independently negated, it is a proposition. Consequently, it may involve a predication over its own subject. I propose that the distinction between the Old Italian and Old Spanish infinitive constructions on the one hand and the Old French ones on the other, is that the former are propositions while the latter are not. According to their propositional character, Old Italian and Old Spanish infinitives co-occur with subjects independent of the matrix sentence. How, then, can we integrate the distribution of object clitics in infinitive complements into this picture? Following Kaiser (1992), I consider object clitics as verbal inflectional elements; i.e., on a par with tense and voice inflectional elements.13 I suggest that in their capacity as verbal inflectional elements they are confined to propositional constructions. Since the infinitive complements in Villehardouin’s chronicle cannot be independently negated, we may assume that they are no propositions and, consistently, they do not cooccur with verbal inflectional elements. I assume that the non-propositional character of Old French infinitives correlates with the fact that they eschew the vicinity of object clitics. If, in the chronicle of Joinville, which is about 100 years younger than the chronicle of Villehardouin, there are two object clitics in the infinitive construction, is this evidence for the propositional character of the infinitives? After all, the infinitives in Joinville’s chronicle are not negated. Turning to this question, I present the type frequency of infinitive embedding verbs in Joinville’s chronicle. (8) Number of verbs embedding infinitive complements in the chronicle of Joinville: 56

That is, compared with Villehardouin’s chronicle, there are more than twice as many verbs which embed infinitive complements, however not as many as found in Compagni’s chronicle. No doubt the infinitival complementation is more productive in Joinville’s chronicle than in Villehardouin’s text. To throw light on the further development in Middle French, I counted the infinitive embedding verbs in Martineau’s (1990) corpus of the Cent Nouvelles Nouvelles (stemming from the 2nd half of the 15th century). I found there were 120 infinitive embedding verbs. Moreover, Martineau’s corpus reveals that the infinitive productively co-occurs with object clitics, specifically in environments where object clitics avoid appearing in Old French; i.e., in preinfinitival position. (9) a.

13

Quand il fut ou lieu ou l’on vouloit le avoir (CNNA 495/912, cf. Martineau 1990:385)

(Middle French)

For my purposes, I abstract from the question if object clitics behave like agreement affixes. It is conceivable that object clitics are inflectional elements without an agreement function. For a different view see Meisenburg (2000).

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

b.

“When he was in the place where one wanted to have him” car il va se coucher (CNNA 213/205, ibid.) “for he goes to sleep”

97

(Middle French)

In Old French, the object clitic le in (9a) would be positioned in front of the finite verb vouloit, just as well in (9b) se would precede va. Concomitantly, Martineau’s corpus contains negated infinitives, as in (10a, b), and subjects in infinitive constructions (10c), cf. Martineau (1990, 1994). (10) a.

b.

c.

faindant de ne s’en apparcevoir (CNNV 40/24, cf. Martineau 1994:60) (Middle French) “alleging not to notice it” Il nous promettera de jamais ne nous accuser (CNNV 57/134, cf. Martineau 1994:63) (Middle French) “he will promise us to never accuse us” Pareillement disoit avoir son pere beaucop de bons heritages (CNNV 25/51, cf. Martineau 1990:448) (Middle French) “Likewise, [he] said [that] his father to have a lot of good heritages”

I conclude that the co-occurrence of object clitics with the infinitive constructions is an indication of the propositional capacity of the construction in which they appear. Concerning the development of the infinitive construction from Old to Middle French, this means that infinitives acquire a more propositional character. The latter development means that Middle French infinitives can be independently negated and that they can have their own subject. In order to describe the development of the medieval French infinitive in a uniform manner with the development of the Latin infinitive (cf. section 1), a corresponding syntactic functional category the changing quality of which is the core of the development should be specified. Following the proposal of Pearce (1990), I assume that Old French infinitives lack a syntactic category Infl and that they obtain Infl in the course of Middle French. To this emergence of Infl, in turn, I suggest to attribute the larger productivity of the infinitive construction in Middle French. Recall that in section 1 the change in the syntactic distribution of the Latin ACI has been correlated with the change of a morphological characteristic, namely with the changing quality of the Latin infinitive paradigm. Can we, now, discover in medieval French a morphological characteristic whose change possibly affects the change of the distribution of infinitive complements? In my opinion we can. But in contrast to the Latin development, it is not the change in the infinitival paradigm. In medieval French the change occurs in the paradigm of the object clitics. To illustrate the French development, I compare the paradigm of the object clitics with the paradigm of the unbounded object pronouns, cf. table 7.

98

MARIA GOLDBACH

Object Clitics direct sg 1 2 3 pl 1 2 3

indirect

le, f. la

me te li

les

leur

Unbounded Object Pronouns reflexive

non-reflexive

reflexive moi toi

se lui nous/nos vous se euls/aus, f. eles

soi

soi

Table 7: The forms of Old French object pronouns (in Villehardouin’s chronicle)14

Table 7 shows that the paradigm of the object clitics shares two forms with the paradigm of the unbounded object pronouns; i.e., nous and vous. In comparison with Old French, there is no such syncretism in the Old Italian and Old Spanish object pronoun paradigms. In these medieval variants, no object pronoun form can be used both as an object clitic and as an unbounded object pronoun. Object Clitics direct sg

pl

1 2 3

lo, il f. la

1 2 3

li/gli, f. le

indirect mi ti li/gli, f. le ci vi li/gli

Unbounded Object Pronouns reflexive

si

si

non-reflexive

reflexive

(a, di, con) me (a, di, con) te (a, di, con) (a, di, con) se lui, lei (a, di, con) noi (a, di, con) voi loro (a, di, con) se

Table 8: The forms of Old Italian object pronouns Object Clitics direct sg

pl

1 2 3 1 2 3

indirect

Unbounded Object Pronouns reflexive

lo, f. la

me te le

se

los, f. las

nos uos les

se

non-reflexive

reflexive

prepos. + mi prepos. + ti prepos. + el/lo, prepos. + si f. ella, n. ello prepos. + nos prepos. + uos prepos. + prepos. + si ellos/los, f. ellas

Table 9: The forms of Old Spanish object pronouns

14

nos and aus each appear just once; for the rest there can be found only nous and euls.

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

99

In the French texts from the end of the 13th century an alteration can be seen concerning the distribution of li/lui (cf. Kok 1985:23). As a consequence of this alteration, there is in Joinville’s chronicle, in addition to nous and vous, a third pronoun form that figures both as object clitic and as unbounded object pronoun: li.15 Additionally, the unbounded object pronouns in Villehardouin’s and Joinville’s chronicles are sequenced in pre-infinitival position, and they can only follow the infinitive if they are coordinated with other nominal phrases. (11) Unbounded object pronouns in infinitive constructions in Villehardouin’s and Joinville’s chronicles

a. b.

c.

15

in pre-infinitival position: Et l’andemain ala parler a l’empereeur et lui veoir (Villeh. 275) (Old French) “and the next day [he] went to talk to the emperor and to see him” ne ne commanderoit ja a ses serjans que il constreinsissent les escommeniés a eulz fere absoudre (Join. 671) (Early Middle French) “and that [he] will not advise his servants that they induce the excommunicated [persons] to let themselves being absolved” in post-infinitival position: … pour requeillir le roy et le conte d’Anjou son frere, et … et moi (Join. 378) (Early Middle French) “… in order to retrieve the king and the count of Anjou his brother and ... and me”

In Villehardouin’s chronicle there are three finite sentences where lui is used as a 3rd sg. ind. obj. clitic instead of li, cf. Chascuns prist ostel tel conme lui plot “Every one took quarters where he pleased” Villeh. 251 (also Villeh. 60, 182). In all three examples lui occurs with a form of the verb plaire “to please”, which seems to be a formulaic expression by that time. This usage cannot be observed in Joinville’s chronicle; i.e., in this early Middle French text composed about 100 years later than the chronicle of Villehardouin, there are no sentences where lui replaces li as a clitic. Conversely, the edition of Villehardouin’s chronicle, which I analysed (based on the oldest manuscript B, Fond français 2137 in the BnF), contains only one example where li occurs in the function of the unbounded object pronoun lui, cf.: Li baron dirent qu'il em parleroient ensamble sanz li. (Villeh. 196) “The barons said that they would consult together without him”. Faral’s (1972) edition of Villehardouin’s chronicle reproduces this sentence as sanz lui, without any remark that the manuscript B contains li. Since I have not yet seen the original manuscript B in the Bibliothèque nationale de France, I cannot decide whether the object pronoun in this sentence is lui or li. In contrast, in Joinville’s chronicle there is only one instance of lui: … ou par lui ou par autrui (Join. 104) “either by him or by another”. Apart from this example, we can only find li in this text, both as an object clitic and as an unbounded object pronoun: et li avoit dit que il vouloit parler a li (Join. 46) “and [he] had said to him that he wanted to talk to him”. On the basis of Villehardouin’s and of Joinville’s texts, I conclude that in the Champenois dialect li has replaced lui as an unbounded object pronoun in the course of the 13th century, in contrast to what refers Kok (1985:23f.) that li was replaced by lui as an object clitic. In the texts analysed by Martineau (1990) (Cent Nouvelles Nouvelles, 2nd half of the 15th century), lui/luy replaces li in all contexts and there are no instances of li. At the present moment I have no idea why and when the replacement of li by lui/luy was finally established.

100

MARIA GOLDBACH

d.

… et il li requist que il me venist secourre moy et mes chevaliers (Join. 227) (Early Middle French) “… and he requested him that he came to rescue me and my chevaliers”

Villehardouin’s chronicle pre-infinitival unbounded object pronouns

7

post-infinitival unbounded object pronouns

0

Joinville’s chronicle 52 2* * cf. (11c, d)

Table 10: Sequencing of unbounded object pronouns in infinitive constructions

By contrast, the unbounded object pronouns in Old Italian and Old Spanish are not confined to the position in front of the non finite verb forms. (12) a. b.

scusando sé e accusando altri (DC 3.39.213) “exculpating himself and inculpating others” … et quel podrie por uentura atraer a si (PcgE321b, 33-34) “… and that [he] by chance could turn him to his side”

(Old Italian) (Old Spanish)

I argue that the increased number of homonymous forms in the medieval French pronoun paradigm, together with the strict order of unbounded object pronouns in infinitive constructions, is accountable for the fact that in the further development of Middle French object clitics recurrently co-occur with infinitives. In the texts stemming from the Middle French period object clitics almost exclusively occur in pre-infinitival position. Martineau (1990:92) reports that in the texts she analysed (written in the second half of the 15th century and in the beginning of the 16th century) there is no instance of an object clitic following the infinitive. That is, in Middle French object clitics are sequenced in front of the infinitive where in Old French we find the unbounded object pronoun. From the fact that in Joinville’s chronicle the form homonymy extends over three pronouns, and the observed increase of infinitive embedding verbs in contrast to Villhardouin’s chronicle, I conclude that the following sentences (13) of Joinville’s chronicle indicate that the Infl-category is established in infinitive constructions. (13) a. b.

et me pria que j'alasse vers li pour la reconforter (Join. 605) “and [she] asked me that I go to her in order to console her” Et le roy leur devoit jurer aussi a leur faire gré de deux cens mille livres (Join. 359) “and additionally the king had to swear them to compensate them with 200.000 pounds”

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

101

Nonetheless, pending further research I must leave the question open as to why negated infinitives still fail to appear in Joinville’s chronicle. At this point I compare the Latin and the French development. Concerning the Latin case, I have proposed that the decrease of ACI embedding verbs correlates with the weakening of the infinitive paradigm. On a more abstract level, I have described this correlation as the weakening of the functional syntactic category Infl(ection). Referring to the French development I argue that the extension of the form homonymy in the pronominal paradigm is conducive to the fact that in Middle French the object clitics start to co-occur with the infinitive. On par with my suggestion for the Latin development, I propose that the latter phenomenon correlates with the increase of infinitive embedding verbs. I regard object clitics as verbal inflectional affixes, indicating the propositional capacity of the construction in which they occur. That is to say, in abstract terms, infinitives are endowed with an Infl-node from when they co-occur with object clitics. In both cases I assume that the morphological change – in the paradigm of the infinitive or in the paradigm of the object pronouns – triggers the syntactic development. Thus, the nature of my explanation consists of ascribing the syntactic change to a morphological alteration by means of an abstract syntactic functional category, the strength or weakness of which corresponds to the morphological alteration. Admittedly, my explanation does not cover the questions of how and why the morphological alterations take place. 4. Explaining grammatical change In this section I turn to the core questions of this volume: how can we approach the paradox of grammatical change and what should be considered as a suitable explanation for a grammatical change? Surely, this question has an epistemological aspect (cf. Lass 1980), but I shall not enter upon this science theoretical discussion. I think that an explanation of grammatical change ought to be related to the following questions: a) what does it mean to know a human language?, and b) how does this knowledge evolve in the individual, cf. Chomsky (1980)? Consequently, I believe that an investigation of grammatical change should have results which not only satisfy historical or diachronic linguistics but also inspire synchronic theoretical linguistics and even psycholinguistics (language acquisition and perception research). In order to achieve this it seems to be a reasonable starting point to seek relatively strong correlations between a range of phenomena and to characterise the core of these correlations on an abstract level. Since, if we find relatively strong correlations by examining grammatical change, these correlations should also play a role in synchronic theoretical linguistics or in psycholinguistics. When studying grammatical change, it is often the case that the grammatical systems of both the source and the target languages are documented (e.g., Old French – Middle French – Modern French). If only the

102

MARIA GOLDBACH

source grammatical system is documented, the question must be posed if one can predict what kind of change will take place in a delimited grammatical area. Is it possible to say that contemporaneous infinitival complements have specific grammatical features that will develop in a specific manner under well-defined circumstances? I think an answer to this question goes beyond the current state of research. Up to now it is not possible to predict the development of grammatical change. Likewise, contemporary grammatical change is nearly opaque to empirical observation because most types of grammatical change cannot be discerned until a long period of time has passed. I believe that if we detect a substantial number of strong correlations between grammatical phenomena it is feasible to predict grammatical change. Although at the current state of research such predictions are not possible, this does not mean that in diachronic linguistics one has to renounce any predictability from the conception of explanation. A different aspect of predictability in historical syntax is taken up by Harris & Campbell (1995). The authors apply the historical comparative method to syntax in order to reconstruct an ancestor grammatical system from the comparison of its daughter grammatical systems. To translate their method into my field of investigation, I could try to reconstruct the Latin ACI on the basis of the Romance infinitive complements by comparing the infinitival sentence patterns. (14) a.

b.

c.

d.

il perché confessò avere ricevuta una testimonanza falsa per messer Niccola Acciaiuoli (DC 1.19.92) (Old Italian) “therefore, he confessed to have received a false testimonial by M. Niccola A.” Et dirent que onques mes n'avoit esté si hardi home qui osast avoir desfié l'empereeur de costentinoble en sa chambre. (Villeh.) (Old French) “And [they] said that there has never been such a bold man who dared to have defied the emperor of Constantinople in his chamber.” E ¿quando cuedas tu auer fecha cibdat que semeie a Carthago ...? (PcgE40a, 31-33) (Old Spanish) “And when do you believe to have founded a town that resembles Carthago …?” Quodsi voles in me esse durior, ambitionem putabis mihi obstitisse. (Att. 1.1.4) (Classical Latin) “If [you] want to deal more severely with me, [you] will believe that [my] ambition has hindered me.”

It is far from obvious how to reconstruct the grammatical structure of the Latin ACI in (14d) by comparing the Old Italian (14a) with the Old French (14b) and/or the Old Spanish (14c) infinitive structures. Therefore, I think that the only possibility to reconstruct grammar is to discover correlations between grammatical phenomena which are evident and to reduce them to an abstract common denominator. To put it another way, I consider parameter searching as proposed by Longobardi (2003) not only as a promising method for grammatical reconstruction but also for discovering predictive explanations in

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

103

diachronic syntax in general. Within the scope of this article I have suggested a correlation between a morphological and a syntactic phenomenon. I assume that the morphological phenomenon is accountable for the syntactic development and that the morphological quality determines the strength or weakness of the abstract functional syntactic category Infl. At the current point of my studies I am not able to formulate a grammatical parameter. Nonetheless, the purpose of my research is to find a cluster of abstract grammatical features underlying the structure of the various infinitive embeddings. If this project succeeds I would regard it as a contribution to the explanation of the grammatical changes exposed in the previous sections and as a promising, albeit partial, demystification of the paradox of grammatical change.

REFERENCES Latin texts Att. = Cicero. Letters to Atticus, ed. by D. R. Shackleton Bailey 1965-1968. Cambridge: Cambridge University Press. Gregor von Tours. Zehn Bücher Geschichten, ed. by Rudolf Buchner 1986. Darmstadt: Wissenschaftliche Buchgesellschaft. French texts CNNA = Les Cent Nouvelles Nouvelles Anonymes. Anonymous author, ed. by F. P. Sweetser 1966. Geneva: Droz. CNNV = Vigneulles, Philippe de. Les Cent Nouvelles Nouvelles, ed. by Charles H. Livingston 1972. Geneva: Droz. Join. = Joinville, Jehan de. La vie de Saint Louis, ed. by Noel L. Corbett 1977. Québec: Naaman. QLR = Li Quatre Livre des Reis. Anonymous author, ed. by Ernst Robert Curtius 1911. Dresden: Gesellschaft für Romanische Literatur. Villeh. = Villehardouin, Geoffroy de. La conquête de Constantinople, ed. by O. Derniame et al. 1978. La conqueste de Costentinoble d’après le manuscrit n° 2137 de la B.N. Nancy: Centre de Recherches et d’Applications Linguistiques. Spanish text PcgE = Primera crónica general de España. Redaction of Alfons X., ed. by Ramon Menéndez Pidal 19773. Madrid: Gredos. Italian texts DC = Compagni, Dino. Cronica delle cose occorrenti ne’ tempi suoi, ed. by Davide Cappi 2000. Dino Compagni, Cronica. Rome: Istituto storico italiano.

104

MARIA GOLDBACH

Cronica fiorentina anonima. Magl. XXV.565 Biblioteca Nazionale di Firenze (also II.IV 323), anonymous author, ed. by Alfredo Schiaffini 1954. Testi fiorentini del Dugento e dei primi del Trecento. Florence: Sansoni. Latini, Brunetto. La Rettorica, ed. by F. Maggini 1968. Florence: Le Monnier. Pieri, Paolino. Cronica, ed. by Anton Filippo Adami 1755. Rome: Monaldini. Secondary literature Anderson, Stephen R. 1982. “Where’s Morphology”. Linguistic Inquiry 13. 571-612. Bonnet, Max. 1890. Le latin de Grégoire de Tours. Paris: Hachette. Cardinaletti, Anna. 2005. “Il pronome personale obliquo”. Grammatica dell’italiano antico, ed. by Lorenzo Renzi & Giampaolo Salvi. Bologna: Il Mulino. Carstairs, Andrew. 1987. Allomorphy and Inflection. Beckenham: Croom Helm. ----------1988. “Some Implications of Phonologically Conditioned Suppletion”. Yearbook of Morphology, ed. by Geert Booij & Jaap Van Marle, 67-94. Dordrecht: Foris. Chomsky, Noam. 1980. Rules and Representations. Columbia: Columbia University Press. ----------. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Cuzzolin, Pierluigi. 1994. Sull’ origine della costruzione dicere quod: aspetti sintattici e semantici. Florence: La Nuova Italia. Egerland, Verner. 2005. “Proposizioni all’infinitivo”. Grammatica dell’italiano antico, ed. by Lorenzo Renzi & Giampaolo Salvi. Bologna: Il Mulino (available: http://geocities.com/gpsalvi/konyv/index.html). Faral, Edmond. 1938/39. Villehardouin, la conquête de Constantinople. Paris: Les Belles Lettres. Goldbach, Maria. 2003. “Repercussions of Morphology on Syntactic Checking: the Latin Accusativus Cum Infinitivo”. Linguistische Berichte 193. 57-90. ----------. 2006. “Kontrastiver Vergleich der syntaktischen Verteilungen der starken und schwachen Objektpronomen im Alt- und Mittelfranzösischen und im Altitalienischen”. Zeitschrift für romanische Philologie 122. 392428. Harris, Alice C. & Lyle Campbell. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press. Hartwig, Otto. 1875. Quellen und Forschungen zur ältesten Geschichte von Florenz. Marburg: Elwert. Jäger, Wolfgang. 1986. Briefanalysen. Zum Zusammenhang von Realitätserfahrung und Sprache in Briefen Ciceros. Frankfurt M.: Lang. Kaiser, Georg. 1992. Die klitischen Personalpronomina im Französischen und Portugiesischen. Eine synchronische und diachronische Analyse. Frankfurt M.: Vervuert.

MORPHOLOGICAL DEVELOPMENTS AFFECTING SYNTACTIC CHANGE

105

Kieckers, Ernst. 1926. Historische griechische Grammatik, IV: Syntax (2nd part). Berlin: de Gruyter. Kneale William & Martha Kneale. 1962. The Development of Logic. Oxford: Clarendon Press. Kok, Ans de. 1985. La place du pronom personnel régime conjoint en français: une étude diachronique. Amsterdam: Rodopi. Lass, Roger. 1980. On Explaining Language Change. Cambridge: Cambridge University Press. Longobardi, Giuseppe. 2001. “Formal syntax, diachronic minimalism, and etymology: The history of French chez”. Linguistic Inquiry 32. 275-302. ----------. 2003. “On Parameters and Parameter Theory”. Syntaxtheorien. Modelle, Methoden, Motive, ed. by Elisabeth Stark & Ulrich Wandruszka , 273-290. Tübingen: Narr. Lot, Ferdinand. 1931. “A quelle époque a-t-on cessé de parler latin?”. Archivum latinitatis medii aevi (Bulletin du Cange) 6. 97-159. Martineau, France. 1990. La montée du clitique en moyen français: une étude de la syntaxe des constructions infinitives. Ph.D. dissertation, University of Ottawa. ----------. 1994. “Movement of Negative Adverbs in French Infinitival Clauses”. French Language Studies 4. 55-73. Mayen, Georg. 1888/9. De particulis quod, quia, quoniam, quomodo ut pro acc. cum infinitivo post verba sentiendi et declarandi positis. Ph.D. dissertation, University of Kiel. Meisenburg, Trudel. 2000. “Vom Wort zum Flexiv? Zu den französischen Pronominalklitika”. Zeitschrift für französische Sprache und Literatur 110. 223-237. Moignet, Gérard. 1965. “L’opposition NON/NE en ancien français”. Travaux de linguistique et de littérature 3:1. 41-65. Pearce, Elizabeth. 1990. Parameters in Old French Syntax: Infinitival Complements. Dordrecht: Kluwer. Perrochat, Paul. 1932. Recherches sur la valeur et l’emploi de l’infinitif subordonné en latin. Paris: Les Belles-Lettres. Priem, Josef. 1889. Die irrealen Bedingungssätze bei Cicero und Caesar. Göttingen: Dieterich. Ramsden, Herbert. 1963. Weak-Pronoun Position in the early Romance Languages. Manchester: University Press. Rheinfelder, Hans. 1953. Altfranzösische Grammatik, Teil 1: Lautlehre. München: Hueber. Rivero, María-Luisa. 1992. “Clitic and NP Climbing in Old Spanish”. Current Studies in Spanish Linguistics, ed. by Héctor Campos & Fernando Martínez-Gil, 241-282. Washington: Georgetown University Press. Schmidely, Jack. 1998. “Los pronombres clíticos y las formas verbales no personales en la Primera Crónica general”. AnMal 21:1. 115-125.

106

MARIA GOLDBACH

Stump, Gregory T. 2001. Inflectional Morphology. A Theory of Paradigm Structure. Cambridge: Cambridge University Press.

GRAMMATICALISATION WITHIN THE IP-DOMAIN∗

SUSANN FISCHER University of Stuttgart

This paper will argue that grammaticalisation within the IP domain is responsible for the change in word order from Old to Modern Romance. Stylistic fronting, postverbal clitics, and other verb-third orders do not seem to be related at first sight. However, on closer investigation it seems as if all these phenomena can be brought together by the availability of an additional functional category in the IP domain that was involved in information structure. This paper will argue that stylistically fronted elements and the verb in postverbal clitic constructions target this additional functional category and thus change the meaning of the sentence. In the course of time, grammaticalisation (in the view of Roberts & Roussou 2003) reduced the availability of this functional category and therefore stylistic fronting and postverbal clitics are not an option in Modern Romance anymore.

1. Introduction The Old Romance languages show some peculiarities with respect to word order – postverbal clitics in finite sentences, stylistic fronting and other verbthird effects – that are ungrammatical in most Modern Romance languages. I will show that the change with respect to these phenomena can readily be accounted for by assuming grammaticalisation within the IP domain. In a generative framework, grammaticalisation is defined as the creation of new functional material through the reanalysis of lexical or functional material (Roberts 1993, Roberts & Roussou 2003). Under this assumption language change can be seen as a change in whether or how functional heads are phonetically realised (Fischer 2002, Roberts & Roussou 2003). Change occurs when the trigger experience for a parameter setting has become obscure or ∗

This paper is a revised version of a talk given at the II Simposi Internacional ‘Vers una sintaxi històrica del Català’, November 2004 and at the XXIX Romanistentag ‘Grammatischer Wandel und seine Erklärbarkeit’ in Saarbrücken, September 2005. I would like to thank the audiences at these two meetings for relevant observations and insightful discussions. I also wish to thank Artemis Alexiadou, Maria Goldbach, Tom McFadden, Claus Pusch, Manuel Pérez Saldanya and two anonymous reviewers for helpful comments and suggestions on an earlier version of this article. Thanks are also due to Sandhya Sundaresan for checking the language. Any remaining errors are of course mine.

108

SUSANN FISCHER

ambiguous for the learner, i.e. whether functional heads are phonetically realised by merge (lexical insertion) or move (material from elsewhere is moved to the functional head) or not at all (phonologically empty). Thus, language change in a generative perspective is connected to language acquisition and furthermore to the old idea of markedness (among many others Meillet 1912) where the learner shows a preference for less marked structures. With respect to Old Romance clause structure this means that an additional functional category below CP and above IP (involved in information structure) was phonetically realised via move and, that nowadays, this category is not phonetically realised anymore. In chapter two, I will give a brief overview of the attested data on stylistic fronting and clitic placement in Old Romance. Section three will connect the two phenomena and account for them by assuming an additional functional category. Chapter four will elaborate on the question of how language change is instantiated and how grammaticalisation, i.e. pathways of grammaticalisation can be explained within a minimalist framework. And, finally in chapter five, I will show how everything falls into place with respect to the change from Old to Modern Romance clause structure if we apply the proposed theory of language change and grammaticalisation. 2. The Data 2.1 Stylistic Fronting Stylistic fronting (SF) was argued to consist of a rule that moves a category to a position in front of the finite verb in those sentences where the position in front of the verb (SpecIP) is not occupied by an overt subject NP. The examples in (1) below illustrate instances of SF in Icelandic. (1a) represents the canonical word order, while (1b) illustrates the order after SF has applied. (1) a.

b.

Hver heldur þú að hafi who think you that has “Who do you think has stolen the bike?” Hver heldur þú að stolið hafi __ hjólinu

stolið stolen

hjólinu the.bike (Holmberg 2000)

Indisputably SF is only still operative in Modern Icelandic and Faroese (Jónsson 1991), but it has been shown that it was operative in all Old Germanic languages (cf. Falk 1993, Fischer 2004). SF in Icelandic is considered to show characteristics that are in opposite distribution to topicalisation (Barnes 1986, Maling 1990, Jónsson 1991, Holmberg 2000, Poole 1997): SF applies to heads whereas topicalisation applies to phrases, SF is clause-bound, topicalisation applies over clause-boundaries, SF observes minimality, topicalisation does not, and SF, but not topicalisation, requires a subject gap. On the basis of these characteristics, SF has been analysed as obligatory movement in order to guarantee a verb-second structure (cf. Maling 1990 among others), or in order to check off an EPP feature (Holmberg 2000). Topicalisation, on the other

GRAMMATICALISATION WITHIN THE IP-DOMAIN

109

hand, applies freely, or more precisely it is involved in information structure (Maling 1990, Holmberg 2000). With respect to the Romance languages, I have argued elsewhere (Fischer 2004) that what we perceive in Old Catalan looks identical in all respects but one, namely the subject gap. Note in the following, the assembled data for all Old Romance languages. Examples for SF that display head movement for Old Catalan1 are found under (2), for Old French under (3), and for Old Spanish under (4). (2) a.

b.

c.

Que feita aviets la corona del that made had the crown of.the “That you had made the crown of the emperor.” noble cavaler. que molt es that much be.3SG noble man. “that he is a very noble man.” qui demanar li vengés , who to-ask him came.3SG , “who came to ask him ..., ”

Emperi. emperor.

participle (Desclot/309) adverb (Desclot/275) infinitive (Desclot/154)

(3)

Et si ne sait que faire puisse __ infinitive and so not knows what to.do can __ “And so he doesn’t know what he can do ...” (Cardinaletti & Roberts 1991: 25)

(4) a.

Dexado ha __ heredades e casas e palaçios left has.3SG __ properties and houses and palaces “He has abandoned the properties, houses and palaces” bastir quiero ___ dos archas to-build want.1SG ___ two chests “I want to build two chests”

b.

participle (Fontana 1993:75) infinitive (Fontana 1993: 75)

As mentioned before SF is taken to be clause-bounded in contrast to topicalisation which can apply across clause boundaries. In the texts of Old Catalan, no example of SF which crossed a clause boundary was attested.2 1 The Catalan data presented in this paper is taken from the corpus assembled by Fischer (2002). The corpus covers the last eight centuries, i.e. 13th - 20th century. To make sure that a certain pattern of syntactic structure was not merely part of an author’s idiolect, two representative texts by different authors were chosen from each century. Each century is represented by at least 800 sentences taken from continuous text. The excerpts from medieval texts (13th to 14th century) presented here are: Llull, Ramon. Date of composition around 1288. Llibre de Meravelles. Desclot, Bernard. Date of composition between 1283 and 1288. Crònica de Bernat Desclot. Eiximenis, Francesc. Date of composition between 1373 and 1386. Contes i Faules. Metge, Bernard. Date of composition 1399. Lo Somni. The references used follow the pattern of my database. To create the corpus, the edition Barcino Els Nostres Clàssics was used. To make sure that the edition Barcino is reliable, some of the texts published by Barcino were compared with the original manuscripts. 2 In this paper, I only make claims with respect to the frequency of data and data that is not attested (NA) concerning Old Catalan since the corpus is representative and can be crosschecked in the database ‘Access’. Of course, it cannot be excluded that maybe in spoken language or in some other texts a certain construction might have been possible or a different

SUSANN FISCHER

110

(5) a.

b.

E dix que anat se n’era __ ja REF. there’was.3SG __ already and said.3SG that gone “and he said that the night had already gone ...” NA E anat dix que se n’era __ and gone said.3SG that REF there’was.3SG _

la nuit ... the night (Desclot/284) ja la nuit already the night

SF is argued to be subject to locality. The locality condition observed can be accounted for in terms of Relativized Minimality (Rizzi 1990) or the Minimal Link Condition (as in Holmberg 1997, 2000)3. Out of several elements which could, in principle, undergo SF, it is only the structurally most superior element that undergoes SF. This condition also holds for the stylistically fronted elements in Old Catalan, as well, as shown in (6). (6) a.

Lo scuder fo molt meravellat de la demanda que el the squire was very surprised of the question that the caveller li hac feta. cavalier him had made “The squire was very surprised by the question that the cavalier had asked him” (Llull/60) b. molt fo meravellat lo rey de les peraules del pagès, very was __ surprised the King of the words of.the page, “The king was very surprised by the words of the page,” (Llull/190) c. NA meravellat fo molt __ lo rey de les peraules del pagès

The subject gap condition is the one characteristic where we see that Old Romance, i.e. Old Catalan is different. As in Icelandic, we find SF in Old Catalan in all types of finite clauses with a subject gap, recalling also the examples of Old French (3) and Old Spanish (4). (7)

subject relative qui demanar li who to-ask him “who came to ask him,”

vengés __ , came __ , (Desclot/154)

distribution attested but, with respect to my corpus, the claims hold. With respect to the other Old Romance languages, it is not possible to say anything about the frequency of certain structures and whether a certain structure is not attested at all. 3 The Minimal Link effect (Chomsky 1995) is a more selective reinterpretation of the Relativized Minimality effect (Rizzi 1990). Relativized Minimality spells out as: in the configuration (i) …. X ….Z…..Y Y cannot be related to X if Z intervenes and Z has certain characteristics in common with X. So in order to be related to X, Y must be in a minimal configuration with X, where minimality is relativized to the nature of the structural relation to be established.

GRAMMATICALISATION WITHIN THE IP-DOMAIN

(8)

111

impersonal construction E el senyor del hostal dix que anat se n'era __ REF there’was.3SG __ and the landord said that gone ja la nuit, already the night, “And the landlord said that the night had already gone,” (Desclot/294)

(9) a.

b.

(10)

complement clause que feta li havie , , that made him had “that she had made him” , que fet m’hajats that made me’have , “that they had made me,” matrix clause més Dit has said have more “You have said more”

(Llull/25)

(Eixemenis/197)

(Metge/179)

However, the Old Romance languages are pro-drop languages where we find referential pro next to expletive pro in main and in embedded contexts. Therefore, we do not expect to find a difference between topicalisation (11) and SF (12) when a subject gap is taken as the distinctive feature. a aquestes peraules lo rey no lur to those words the king not them “The king did not answer them to those words.” b. és veritat que en algun lloc s’ha that in some place REF’has.3SG be.3SG truth “it is true that it is more ardently in certain places”

(11) a.

. respòs answered ___ (Desclot/150) pus ardentment __ more ardently __ (Metge/60)

And this is indeed the case. We do find SF together with a full subject. Compare (12a) to (12b). del gran honrament que feyt nos avets __ , of.the great honour that made us had.2PL __ “of the great honour that they have given us,” (Desclot/316) b. e adonchs con amà Deu e serví Déu de ço and so with love God and serve God of this que Déus donat li havia __, that God given him has __ “and thus to love God and serve God for that which God has given him,” (Llull/36)

(12) a.

The attested sentences where we find SF together with a full subject clearly show that SF cannot have applied to fill the subject gap, not even to guarantee

112

SUSANN FISCHER

a verb-second structure, since the finite verb in sentence (12b) turns up in fourth position. Sentences like the above can be taken as clear evidence against the traditional approaches.4 Let us now turn to the clitic distribution in Old Romance. 2.2 Clitic Distribution The lack of absolute initial clitics in Old Romance and the lack of postverbal clitics in subordinate contexts has often been explained by the clitics’ inherent enclitic character or/and by a description of them as being second position elements in a verb-second phrase structure (Tobler 1875; Mussafia 1888; Wanner 1987; Cardinaletti & Roberts 1991; Benincà 1995, Halpern 1995; Fontana 1993, among many others). However, a look at the clitic distribution in Old Romance reveals that we see a lot of attested data that challenge the generally held view that these object pronouns are prosodically enclitic elements or second position elements.5 (13) shows an Old French example where the clitic is in absolute initial position, (14) shows that there is only a prosodic break to the Old Catalan clitic’s left, to which it could attach and (15) represents a clear example that the clitic is in fourth and the verb is in third position. So, Old Romance is clearly not a verb-second language (Fischer 2002, Kaiser 2002). (13)

(14)

Me fetes vos droit de doner a la reine... me give you right to give to the queen “You give me the right to give to the queen ....”

(Benincà 1995:330)6

l’alberque e la viny que possehia, the’refuge and the vineyard that possessed.3SG, l’embargave algunes vegades ... him’hindered sometimes “the refuge and the vineyard he possessed, hindered him sometimes…” (Llull/342)

(15)

4

lo bon hom hac totes les vestedures pobres e the clothing poor and the good man had.3SG all mesquines que la dita infanta portà-li shabby that the said Infant carried.3SG-him “the good man had all the poor and shabby clothing that the Infant had given to him” (Desclot/34,19)

In the meantime, Gunnar Hrafnbjargarson has shown in his dissertation that, with respect to Icelandic SF as well, the traditional approaches do not hold (cf. Hrafnbjargarson 2004). 5 For an overview and examples of the attested clitic distribution in all Old Romance languages, please refer to Fischer (2002). 6 Benincà doesn’t provide a free translation. The free translation is mine (S.F.) and unfortunately without any knowledge of the context of the sentence.

GRAMMATICALISATION WITHIN THE IP-DOMAIN

113

On the basis of sentences as in (15), I argued elsewhere (Fischer 2002) that a further category between CP and TP is needed. On the basis of examples as in (16), it was argued that the reason for the postposition of the clitic behind the verb cannot be explained by any clause-structural reason but needs to be explained by information structure (Fischer 2002, Fischer 2003). (16) a.

.I. jorn sdevench-se que .I. juheu vench a one day became.to.pass-REF that .I. jew came to aquell sant hom, that sacred man, “One day it did happen that this Jew came to that sacred man,” b. .I. jorn se sdevench que lo ermità splugave REF became.to.pass that the hermit put.on one day son cilici, his penitential robe, “One day it happened that the hermit put on his penitential robe,”

(Llull/78)

(Llull/132)

Sentences (16a) and (16b) not only display the same structure (Subject Verb) but also identical words: the words preceding the sequence clitic-verb / verb-clitic I. jorn (one day), the verb sdevench (happen), and the clitic se (ref) are exactly the same. So there is no structural difference which could motivate the verb to precede the clitic on one occasion and to follow it on another. Only when we look at the contexts in which the sentences appear in the texts, does the difference become obvious. (16a) describes the daily routine of a hermit. Sentence (16b) states that the hermit (referred to as “the holy man”) is visited by someone. The verb-clitic sequence clearly marks the fact that the visit stood out in the life of the hermit, something that needs to be emphasized since a hermit does not usually receive any visitors. 3. Connecting and explaining the phenomena In the previous chapter, it was shown that the traditional explanations seem to hold neither for SF and nor for clitic placement if we consult a more extensive and/or more representative corpus. Look at the change in clitic distribution over the centuries. In Table 1, it is shown that Old Romance clitics can appear in pre- and postverbal position;7 however, they never appear in a postverbal position in negated sentences. 7 The table provides an overview of attested clitic distribution in all “major” Romance languages (Fischer 2002). Line 7 (C°/CP-V-Cl) only shows in which century postverbal clitics appear in Old Catalan. However, postverbal clitics in embedded sentences have also been attested for Old French (Kok 1985), Old Italian (Henz 1908, Cecilia Poletto p.c) and for Old Spanish (Granberg 1988, Gessner 1893), even though it is not clear in which century they could appear. See Hinzelin (2005: 68-75) for an extensive discussion on the phenomena of postverbal clitics in embedded sentences.

SUSANN FISCHER

postverbal

preverbal

114

1 2 3 4 5 6 7 8

Cl - V XP - Cl - V C°/CP - Cl - V Neg - Cl - V V - Cl XP - V - Cl C°/CP - V - Cl Neg - V - Cl

13th 14th 15th 16th 17th (a) (a) (a) (a) a a a a a a a a a a a a a a a a a a a a a a a a a a Ø Ø Ø a a Ø Ø Ø Ø Ø

a = attested, (a) = attested mainly in co-ordinated sentences or after a prosodic break, Ø = not attested.

Table 1: Clitic distribution with respect to finite verbs

Whenever there is a negation in the sentence, the clitic precedes the verb. Even though example (17) is again taken from Old Catalan, this claim seems to hold without exception with respect to all Old Romance languages (cf. Wanner 1987, Martins 1995). (17) a.

a la regina feya tots los plahers que podia, to the sovereign made.3SG all the wishes that could, no la poch alegrar, not her could please, “He fulfilled all the sovereign’s wishes, but he could not please her.” b. NAa la regina feya tots los plahers que podia, no poch-la alegrar,

(Llull p.38/l.3)

Looking at stylistic fronting, we find exactly the same constraint. As soon as there is a negation in the sentence, SF does not apply. The Old Catalan (18b) is not attested (for Old Spanish see Fontana 1997). (18) a.

b.

que feta li havie ___, that made.FEM him had ___, “That she had made him...,” NA que no feta li havie ____,

(Llull p.35/l.20)

Additionally, SF and postverbal clitics exclude each other. In the whole of the Old Catalan corpus, no postverbal clitic appears together with a stylistically fronted element. (19) is not attested. (19) a. b.

NA NA

que no feta havie-li __ , que feta havie-li __ ,

If we summarize the findings with respect to stylistic fronting, clitic distribution and negation we get the following pattern.

GRAMMATICALISATION WITHIN THE IP-DOMAIN

(20) A. B. C. D.

*NEG-VERB-CLITIC *NEG-SF-VERB *NEG- SF-VERB-CLITIC *NEG-SF-CLITIC-VERB

E. F. G. H.

115

* SF-VERB-CLITIC SF-CLITIC-VERB SF-VERB. VERB-CLITIC

Postverbal clitics are not possible whenever a negation is contained in the sentence. The same obviously holds for stylistic fronting. Surprisingly enough, no postverbal clitic together with a stylistically fronted element is attested. So, it seems to be the correct assumption that the verb in postverbal clitic constructions and the SFed element target the same position in the phrase structure. Since negation is part of this constraint on clitic and SF distribution, I propose building on Chomsky (1957) and Laka (1990) that the additional position is a functional category (ΣP) where different operators are realised: negation, ‘emphatic’ affirmation, ‘neutral’ affirmation. Within a minimalist framework, Σ is – like T or Agr - a functional category with features that need to be checked. Concerning Old Catalan (Old Romance) sentence structure, I propose that the different instantiations of Σ, depend on which operator is realised. (21) a. Σ° [−v] neutral b. Σ° [+v] emphasis c. Σ° [no] negation

The overall picture is the following: if a verbal element (the finite verb or an SF-element) moves to Σ to check off the strong v-feature, the structure verbclitic, or SF-verb appears. Σ with a weak v-feature results in the structure cliticverb, or the structure without a stylistically fronted element. Negative sentences are different from emphatically affirmative clauses in that, in negative clauses the sigma head (Neg°) is lexicalised8. Therefore, we never find the sequence verb-clitic or SF in negated sentences.

8 As has been shown in Laka (1990), replies to yes/no questions crucially involve the Σ projection. We know that Latin did not have a word for “yes” (cf. Pinkster 1988 and all references in there). Instead expressions like ita est, sane, vero etc. are used next to the bare inflected verb. The minimal affirmative answer in Latin involves the bare inflected verb and the negative answer involves the affix non together with the bare inflected verb (ii). (ii) Legistine librum? a. legi b. non legi read.2SG book read.1SG not read.1SG “Have you read the book?” “Yes” “No” The same effect can still be observed as an archaic use in European Portuguese (Martins 1995). (iii) Comeste o pãozinho que te trouxe? a. comi b. não the sandwich that you brought.1SG eat.1SG no eat.2SG “Did you eat the sandwich I brought you?” “Yes” “No”

116

SUSANN FISCHER

4. Explaining Syntactic Change 4.1 Language change within the generative tradition Most of the studies within the generative school concerning diachronic instability represent one or the other of two basic approaches to language change: the structuralist (cf. Lightfoot 1991) and the variationist (cf. Kroch 2001) approach. Both approaches assume a highly structured Universal Grammar consisting of principles and parameters that are set according to triggers in the language learner’s linguistic environment. They also share the view that language change and language acquisition are closely connected. According to this assumption, the language learner is the locus of change and variation. A problem for both approaches might be the apparent incompatibility between language change and a grammatical theory that believes in language as a perfect system (the strong minimalist thesis, cf. Chomsky 1995, 2000, 2001). Now, perfect systems should not vary over time. However, looking at a certain language at different points in its history, we notice that, obviously, language needs to be seen as a dynamic system in constant change. Furthermore, the synchronic variation among the world’s languages clearly shows that we find different types, each of which is perfect in its own way. A way out of the dilemma - of perfect systems that should not vary over time and the obvious ongoing change we perceive in the world’s languages - is to connect diachronic change to synchronic variation and propose that every speaker is able to refer to more then one I(nternal)-language. In other words, language learners in historical periods characterized by a high degree of linguistic variation are able to acquire two grammars or competing analyses for a given construction (cf. Kroch 2001, among others). The alternative analyses can remain side by side as part of the overall linguistic competence of speakers over extended periods of time. Ongoing changes are thus gradual and it is more likely that one generation differs from its predecessors in the frequency with which its speakers use certain forms than in whether those forms are possible at all. If this assumption is correct with respect to Old Romance, we should find some kind of optionality with respect to the variation in structural representation. And, in fact, this is the case. Looking at Catalan clitic distribution over the centuries (Table 2), it becomes clear that the change in clitic placement needs to be seen as a gradual change that is characterized by competing structures over an extended period of time. Two questions immediately arise when adopting an approach that claims that language change is an effect of competing structures. Where does the variation in the grammar of individuals come from? And secondly, why is the variation lost across time, or, more precisely, why should the frequency of the use of the different constructions change from one generation to the next?

GRAMMATICALISATION WITHIN THE IP-DOMAIN

Century Verb-clitic XP-verb- clitic Clitic-verb XP-clitic-verb

117

13th 14th 15th 16th 17th 163 169 140 76 51 76 56 20 15 1 32 41 75 54 155 590 470 546 575 493

Table 2: Clitic placement in finite sentences

This leads to the overall question of why one construction should win the synchronic competition. Both questions are difficult to answer and different views are fiercely discussed among the linguists of the world. In the following, I shall mainly concentrate on the question of why variation gets lost across time, and why one construction should win the synchronic competition. The first question - about how variation gets into a language - seems to be almost impossible to answer without a better knowledge of the origin of language. Two opposing views exist: one approach sees language change as the trigger for variation; the other sees variation as the trigger for language change. In the latter view, variation is an effect of contact between speakers of different languages, dialects, or sociolects. The two views can be compared to the question of which was first: the chicken or the egg? So far, we do not know how and when language arose in humans, and we also do not have an answer to the question of whether there was one first language or several languages. Therefore, it is impossible to answer how the different languages came into existence, i.e. started out. What we do know, however, is that for a long time a lot of different languages have coexisted in this world, and that these languages are in constant contact to each other. Knowing what we know about the development of Romance languages from their beginnings to the present day, I suggest that variation in the grammar of individuals is caused by the social and regional stratification of language, i.e. all kinds of language contact phenomena: standard vs. nonstandard, superstrate vs. substrate, high vs. low, native vs. foreign languages. I do not claim that all linguistic variation needs to be associated with ongoing change but that any change that takes place requires the presence of linguistic variation in at least one of the components: phonology, morphology, syntax, lexicon etc. This seems a reasonable approach with respect to the development of the Old Romance languages since the language situation for these languages has been described from the beginning onwards as diglossic9 – always in 9

The development of the Romance languages under the influence of High Latin in the Middle Ages fits in my opinion Ferguson’s definition of diglossia almost perfectly (cf. Ferguson 1959), i.e. Latin possessed a large body of literature over many centuries combined with a low literacy rate. There was a high variety in the language (church, literature etc.) and there were low varieties which were used for ordinary conversations. With respect to French and Catalan,

118

SUSANN FISCHER

contact and competition with their superstrate language Latin, and always in contact with other substrate languages that either developed or already existed around them (e.g. Basque, Breton). Allowing for syntactic diglossia10 leaves us with the important question of why variation – with respect to some structures – is lost. It used to be argued that the loss could be explained by a failure in the transmission of linguistic features. Such failures, in principle, could occur within a group of adult native speakers of a language, who for some reason substitute one feature for another in their usage, as happens when new words are coined and substituted for old ones (Bickerton 1999). However, in the case of syntactic features, such innovations by adults are largely unattested. Instead, failures of transmission have been shown to occur mainly in first language acquisition. But why should a learner fail to acquire a certain grammatical feature since, obviously, the feature has been part of the grammar of the language in the immediate past? The understanding of transmission failures is very limited. Studies of language acquisition generally take for granted that the evidence to which the learner is exposed is sufficient to ensure their accurate learning by a competent first language learner (cf. Longobardi 2001). This assumption is reasonable, even though language change shows that there are limits to its validity. We do not know what the limits are, and it is not clear how to find them, since experimentally manipulating the evidence presented to the learner is not ethical. Documented cases of change are, therefore, especially interesting as natural experiments in language transmission. Although the interpretation of change is difficult, we do have the advantage that we see which structures existed during which time and which were lost or replaced in favour of other structures. Based on Clark & Roberts’s (1993) view on first language acquisition and markedness I propose that variation does not get lost because of a failure of transmission but because of a preference for simpler structures, i.e. less marked structures. More specifically, I argue that the change in word order from Old to Modern Romance can be accounted for under the assumption of grammaticalisation within the functional categories, i.e. the preference of first language learners for a less marked structure. 4.2 Grammaticalisation within minimalism Diachronic and cross-linguistic variation is accounted for by means of assigning different values to a finite set of options, called parameters that are provided by Universal Grammar (Chomsky 1995, 2000, 2001). The acquisition of syntax is seen as the process of parameter setting. Syntactic change, it was even more complex because they also had to compete in the domain of literature with Provençal which was used for poetry (Àngels Bosch & Puimedon 1985). 10 Syntactic diglossia refers to competing forms that may differ in social register with an unreflecting vernacular variant (cf. Kroch 2001:6).

GRAMMATICALISATION WITHIN THE IP-DOMAIN

119

therefore, is a change in the parametric values specified for a given language. Parameter setting is due to strong vs. weak features on the functional heads. Functional heads (Tense, Determiners, Complementizer etc.) are present in all languages. The parametric variation that is found among different languages does not depend on the functional heads that appear in clause-structure – these are the same in all languages (following Cinque 1999) – but on whether and how these functional heads are phonetically realized (see Roberts 1998 for a similar view). In a minimalist framework, grammaticalisation11 (cf. Roberts & Roussou 2003) is defined as a change with respect to the functional categories in clausestructure, i.e. whether functional categories are spelt out via merge (lexical insertion), via move (material from elsewhere is moved to the functional head) or not spelt out at all (phonologically empty). The obvious pathways of grammaticalisation are defined by the functional hierarchy through which grammaticalised material can travel by means of successive upward reanalysis and by the markedness hierarchy that ‘predicts’ the upward reanalysis. The effects of grammaticalisation, namely phonological reduction and semantic bleaching, are explained by the nature of functional categories. Functional categories are ‘defective’ at the interfaces (Phonological Form and Logical Form). Functional categories compared to lexical categories lack non-logical content and they are prosodically subminimal. Under this assumption, language change is a change in the phonetic realization of functional heads. Change occurs when the trigger experience for a parameter setting, i.e. whether functional heads are phonetically realised by merge or move or not at all, has become obscure, or ambiguous for the learner. As soon as the trigger experience has become obscure, a built-in preference by the learner for simpler representations makes the learner opt for the less marked or default option (Clark & Roberts 1993), the default option being the least marked structure. Structures including moved material are more marked than structures including merge, since movement always creates relatively complex representations, in the sense that these structures include more nodes (cf. Kayne 1994).

11 Grammaticalisation – the process whereby new grammatical material is created out of “autonomous” words (Meillet 1912) – has recently received much attention not only in the field of diachronic linguistics but in all areas of linguistics as witnessed by the considerable amount of conferences and books devoted to this topic (e.g. Campbell 2001, Detges 2001, Heine & Kuteva 2002, Bisang, Himmelmann & Wiemer 2004, and the two proceedings of the conference series “New reflections on Grammaticalization” Wischer & Diewald eds. 2002, and Fischer, Norde & Perridon 2004). Due to limitations on space, I restrict myself to discussing grammaticalisation in terms of a formal theory of syntax, namely in terms of Chomsky’s Minimalist Program (cf. Roberts & Roussou 2003).

SUSANN FISCHER

120 (22)

a. X

b.

X

Y

X

The markedness hierarchy needs to be seen in the following way: structures including moved material are more marked then structures including merge, and structures including merge are more marked than structures that are phonetically not realised. I take syntactic triggers, i.e., word orders alone, to be responsible for the setting of parameters (cf. Fischer 2002, for a different view see Roberts 1998 and Roberts & Roussou 200312). When a piece of data expresses some parameter value, the learner will be under pressure to set that parameter to the value expressed by the piece of data. This leads to the definition of a trigger as follows. (23) Syntactic Trigger: a sentence S is a trigger for parameter P if S expresses P.

Within this theory of markedness, only marked values of parameters need to be expressed; the parameter expression reduces to: (24) a. b.

expression of movement realisations (syntactic displacement). expression of merge (lexical insertion of free functional morphemes)

Unmarked values are triggered by default. The learner will only be under pressure to set the parameter to the value expressed by a piece of data if the trigger is categorical. That means that all sentences the learner hears need to express the strong parameter value, if the learner has the option of choosing between two semantically equivalent representations of a sentence, where one structure includes a moved constituent and another structure lacks such a moved constituent, the learner will always prefer the less marked option. Grammaticalisation can now easily be related to the markedness hierarchy, as a change in the phonetic realisation of functional categories. In a first step, a functional category is realised by lexical material (e.g. a verb) that is moved from elsewhere to the functional material; in a second step, the same functional material is realised by lexical insertion of free functional morphemes (e.g Aux), and in a third step, the functional category is not phonetically realised anymore. In this view, markedness creates ‘basins of attractions’ (in the sense of complexity theory) within the parameter settings and can, thus, explain the ample evidence of pathways or tendencies, i.e., the non-randomness of grammaticalisation. Thus, we retain the explanatory force of the notion of

12 For a discussion and motivation of why syntactic triggers are enough even for the case of English modal verbs, see Fischer (to appear).

GRAMMATICALISATION WITHIN THE IP-DOMAIN

121

parametric change, but make it compatible with the evidence that grammaticalisation follows certain directions. Applying the above approach to the changes we perceive in the different diachronic stages of Romance will provide the right results. 5. Loss of stylistic fronting and postverbal clitics in Romance With respect to Old Romance it has been claimed that postverbal clitics disappeared when they stopped being phonologically enclitic elements and/or when Old Romance stopped being a verb-second language. As was shown in section 2, Old Romance clitics were not inherently enclitics, and Old Romance was not a verb-second language (Fischer 2002, Kaiser 2002). Furthermore, not only the data on Catalan (Fischer 2002), but also the new data on so far less studied Old Romance languages such as Occitan (Hinzelin 2005) clearly show that the former explanations of clitic distribution do not hold. With respect to stylistic fronting, it was claimed that, in the Germanic languages, SF disappeared when the verb stopped moving to I° (cf. Falk 1993, Platzack 1988). In none of the Modern Romance languages is SF available, although the verb still undoubtedly moves to I°. The explanation provided with respect to the loss of SF in Germanic can, therefore, not be extended to the Romance languages. I would like to suggest an account of grammaticalisation within the functional categories that is able to account for the loss of postverbal clitics as well as for the loss of SF in the Romance languages. In the emergence of Old Romance, an additional functional category was active (ΣP), i.e. in need of being phonetically realised. In the emergence of Proto-Romance, different realisations for Σ were available depending on what needed to be expressed: negation vs. emphasis or a neutral affirmation. Σ existed in three variations: with a lexical head Neg°, with a strong [+v] feature and with a weak [-v] feature. In the case of negation a lexical head was merged to Σ yielding the sequence clitic-verb, and never SF. No change can be attested in the available data with respect to negated sentences13. These have always shown the ordering clitic-verb, and no SFed elements. The change obviously took place with respect to the strong [+v] feature. Looking at the change with respect to sentences without negation, we see a clear grammaticalisation effect. Until the end of the 14th century, both realisations of the Σ category are attested with different interpretations in main and subordinate sentences. The canonical word order refers to a Σ category that is phonologically empty. In the non-canonical word order, i.e. SF, postverbal clitics refer to a Σ category that is realized via move. After the 14th century, the 13

Leaving aside the fact that in the beginning Latin in yes-no questions not only used the lexical entry non but also the inflected verb. According to the grammaticalisation theory presented here this is also a change from a marked [merge + move] to a less marked structure [merge].

122

SUSANN FISCHER

sequence verb-clitic was lost in subordinate sentences. From the 15th century onwards, canonical and non-canonical word orders still exist, but they were identified as presenting the same meaning: neutral affirmation. The learner was, thus, exposed to two competing structures, a marked structure (including move) and an unmarked structure (Σ that is phonologically empty) and he subsequently chose the default option: the order without movement to Σ. There are many other effects of this change, two important ones for the approach proposed here. First, the absence of stylistic fronting goes together with the absence of postverbal clitics in finite sentences; second, stylistic fronting rapidly decreased together with the loss of postverbal clitics in embedded sentences.

REFERENCES Àngels Bosch, Maria & Pilar Puimedon. 1985. Iniciació a la Història de la Literatura Catalana. Barcelona: Edhasa. Barnes, Michael. 1986. “Subject, Nominative and Oblique Case in Faroese”. Scripta Islandica 37. 13-46. Benincà, Paola. 1995. “Complement Clitics in Medieval Romance: The ToblerMussafia Law”. Clause Structure and Language Change, ed. by Adrian Battye & Ian Roberts, 325-344. Oxford: Oxford University Press. Bickerton, Derek. 1999. “How to Acquire Language without Positive Evidence: what Acquisitionists Can Learn from Creoles!”. Language Creation and Language Change, ed. by Michel De Graff, 129-157. Cambridge, Mass.: Bradford Book. Bisang, Walter, Nikolaus P. Himmelmann & Björn Wiemer, eds. 2004. What Makes Grammaticalization? A Look from its Fringes and its Components. (= Trends in linguistics: Studies and monographs, 158.) Berlin: De Gruyter. Campbell, Lyle, ed. 2001. Grammaticalization. A Critical Assessment. Language Sciences 23 [special issue]. Cardinaletti, Anna & Ian Roberts. 1991. “Clause Structure and X-second”. Ms. University di Venezia. Clark, Robin & Ian Roberts. 1993. “A Computational Model of Language Learnability and Language Change”. Linguistic Inquiry 24:2. 299-345. Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton. ----------. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. ----------. 2000. “Minimalist Inquiries: the framework”. Step by Step, ed. by Roger Andrew Martin, David Michaels & Juan Uriagereka, 89-155. Cambridge, Mass.: MIT Press. ----------. 2001. “Derivation by Phase”. Ken Hale: A life in language, ed. by. Michael J. Kenstowicz, 1-52. Cambridge, Mass.: MIT Press.

GRAMMATICALISATION WITHIN THE IP-DOMAIN

123

Cinque, Guglielmo. 1999. Adverbs and Functional Projections: A CrossLinguistic Perspective. Oxford: Oxford University Press. Detges, Ulrich. 2001. Grammatikalisierung: Eine kognitiv-pragmatische Theorie, dargestellt am Beispiel romanischer und anderer Sprachen. Habilitation thesis. Universität Tübingen. Falk, Celia. 1993. Non-Referential Subjects in the History of Swedish. PhD Diss.. University of Lund. Ferguson, Charles A. 1959. “Diglossia”. Word 15. 325-40 Fischer, Olga, Muriel Norde & Harry Perridon, eds. 2004. Up and down the Cline: The Nature of Grammaticalisation. (= Typological studies in language, 59.) Amsterdam & Philadelphia: John Benjamins. Fischer, Susann. 2002. The Catalan Clitic System: A Diachronic Perspective on its Syntax and Phonology. Berlin: Mouton de Gruyter. ----------. 2003. “Rethinking the Tobler-Mussafia Law”. Diachronica 20:2. 259-288. ----------. 2004. “Stylistic Fronting: A Contribution to Information Structure”. MLC Effects in Syntax, ed. by Artur Stepanov, Gisbert Fanselow & Ralf Vogel, 125-147. Berlin: Mouton de Gruyter. ----------. to appear. “Review of Robert & Roussous’s Syntactic Change”, Beiträge zur Geschichte der Deutschen Sprache und Literatur. Fontana, Josep. 1993. Phrase Structure and the Syntax of Clitics in the History of Spanish. Ph.D. Diss. University of Pennsylvania. Granberg, Robert Arthur. 1988. Object Pronoun Position in Medieval and Early Modern Spanish. PhD. Diss. University of California. Gessner, Emil. 1893. “Das spanische Personalpronomen”. Zeitschrift für romanische Philologie 17. 1-34. Halpern, Aaron. 1995. On the Placement and Morphology of Clitics. Stanford: Dissertations in linguistics CSLI publications. Henz, Heinrich. 1908. Stellung der Objektspronomina im Verhältnis zum Verbum wie auch unter sich im Altitalienischen. Programm des Kgl. humanistischen Gymnasiums zu Landau. Schuljahr 1907/08. Landau: J. Meixner. Heine, Bernd & Tanja Kuteva. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hinzelin, Marc. 2005. Zur Stellung der Objektpronomina im Gaskognischen. PhD Diss. University of Konstanz. Holmberg, Anders. 1997. “Scandinavian Stylistic Fronting: Movement of Phonological Features in the Syntax”. Working Papers in Scandinavian Syntax 60. 81-124. ----------. 2000. “Scandinavian Stylistic Fronting: how any Category can Become an Expletive”. Linguistic Inquiry 31:3. 445-483. Hrafnbjargarson, Gunnar Hrafn. 2004. Oblique Subjects and Stylistic Fronting in the History of Scandinavian and English. PhD Diss. Aarhus Universitet.

124

SUSANN FISCHER

Jónsson, Jóhannes Gisli. 1991. “Stylistic Fronting in Icelandic”. Working Papers in Scandinavian Syntax 48. 1-43. Kaiser, Georg. 2002. Verbstellung und Verbstellungswandel in den romanischen Sprachen. Tübingen: Narr. Kayne, S. Richard. 1994. The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press. Kok, Ans de. 1985. La place du pronom personnel régime conjoint. Une étude diachronique. Amsterdam: Rodopi. Kroch, Anthony. 2001. “Syntactic Change”. Handbook of Contemporary Syntactic Theory, ed. by Mark Baltin & Chris Collins, 699-730. Malden: Blackwell. Laka, Itziar Miren.1990. Negation in Syntax: On the Nature of Functional Categories and Projections. Cambridge, Mass.: MIT Press. Lightfoot, David. 1991. How to Set Parameters: Arguments from Language Change. Cambridge, Mass.: MIT Press Longobardi, Guiseppe. 2001. “Formal Syntax, Diachronic Minimalism, and Etymology: the History of French chez”. Linguistic Inquiry 32. 275-302. Maling, Joan. 1990. “Inversion in Embedded Clauses”. Modern Icelandic Syntax, ed. by Joan Maling and Annie Zaenen, 71-91. San Diego & London: Academic Press. Martins, Ana Maria. 1995. “Clitic Placement from Old to Modern European Portuguese”. Historical Linguistics, ed. by Henning Andersen, 395-407. Amsterdam & Philadelphia: John Benjamins. Meillet, Antoine. 1912. “L’évolution des formes grammaticales”. Reprint 1958 in A. Meillet, Linguistique historique et linguistique générale, 130-158. Paris: Champion. Mussafia, Alfredo. 1888. “Enclisi o proclisi del pronome personale atono quale oggeto”, Romania 27. 145–146. Pinkster, Harm. 1988. Lateinische Syntax und Semantik. Tübingen: Francke. Platzack, Christer. 1988. “The Emergence of a Word Order Difference in Scandinavian Subordinate Clauses”, MCGill Working Papers in Linguistics, (Special Issue on Comparative Germanic Syntax), 215-238. Poole, Geoffrey. 1997. “Stylistic Fronting in Icelandic: A case study in prosodic X° Movement”. Newcastle and Durham Working Papers in Linguistics 4. 249-283. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, Mass.: MIT Press. Roberts, Ian. 1993. “A Formal Account of Grammaticalisation in the History of Romance Futures”. Folia Linguistica Historica XIII/1-2. 219-258. ----------. 1998. Verb-Movement and Markedness. Ms. University of Wales. ---------- & Anna Roussou. 2003. Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Tobler, Adolf. [1875]. “Besprechung von J. Le Coultre, De l’ordre des mots dans Chrétien de Troyes”. Vermischte Beiträge zur französischen Grammatik by Adolf Tobler, vol. 5, 395-414, reprinted Leipzig 1912.

GRAMMATICALISATION WITHIN THE IP-DOMAIN

125

Wanner, Dieter. 1987. The Development of Romance Clitic Pronouns: From Latin to Old Romance, Berlin & New York: De Gruyter. Wischer, Ilse & Gabriele Diewald, eds. 2002. New Reflections on Grammaticalization. (= Typological studies in language, 49.) Amsterdam & Philadelphia: John Benjamins.

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE* GIAMPAOLO SALVI Eötvös Loránd Tudományegyetem, Budapest Jaj, gyerekek, hány billió és billió véletlenen kellett átbukdácsolnunk, míg ide jutottunk. “Oh, fellows, over how many billions and billions of accidents have we had to stumble to reach this point!” (Kosztolányi Dezsı: Esti Kornél kalandjai, Tanú) Syntactic change consists of (a series of) small, local changes which are the result of chance or are brought about by the possible imperfections of the linguistic system — the impression of globality of the change is due to the sum of the individual changes which, eventually eliminating single imperfections, contribute to the formation of a more coherent system. In Modern Italian, one can identify two independent reflexive (“si”) constructions which syntactically demote the lexical subject: a passive one and an impersonal one, with quite distinct properties. Old Italian only had passive si — the impersonal construction is the result of many small changes in the rules and the domain of application of the passive construction: these changes began in the Old Italian period and lasted for at least five centuries. But this new construction, not being the result of a unitary project, continues to show signs of being imperfectly put together.

If one considers syntactic changes over a long period of time, one might obtain the impression that diachronic change is a global process transforming a complex, well functioning structure into another complex, well functioning structure. I would like to contend that syntactic change does not work this way: syntactic change consists of (a series of) small, local changes which are the result of chance (e.g. contact between languages) or are brought about by the possible imperfections of the linguistic system (which can be eliminated by the generalization of a rule and the consequent simplification of the system) – the impression of globality of the change is due to the sum of the individual changes which, eventually eliminating single imperfections, contribute to the formation of a more coherent system. My way of seeing syntactic change is close to the traditional view of how analogical change works in morphology: the global regularization of a morphological pattern is not the result of a * I would like to thank the anonymous referee for his/her many remarks which greatly helped me to clarify my ideas and (I hope) to render this paper clearer and more readable.

128

GIAMPAOLO SALVI

unique, thorough change, but the fruit of a series of individual (often lexical) changes, as is demonstrated by the exceptions to the general pattern, which are but the ‘imperfections’ of the system not (yet) corrected by the analogy.1 From a more general point of view, although in practice languages are well functioning systems, they are far from perfect: an actual human language is not a system où tout se tient. It may contain perfect subsystems, but it also contains parts where the principles that govern the structures fail to apply in their generality.2 I would guess that this state of affairs is not (always) due to our failure to grasp correctly the rules of the language, but that it is inherent in the nature of language itself (Matthews 1981: ch. 1). This imperfection is what renders linguistic change possible: since languages do not have to be perfect, they can accept innovations that spoil them as well as innovations that improve them, limited only by the restrictions imposed by Universal Grammar. The imperfections of a language are the result of local linguistic changes and are the causes of new local changes.3 In what follows, I will try to show how syntactic change works with a case study taken from the history of the Italian language. Modern Italian can use the reflexive construction to demote the lexical Subject of a verb. On the basis of the general features of Italian sentence structure, most analyses identify two independent constructions: a passive one and an impersonal one, with quite distinct syntactic properties. While the passive construction conforms to the general properties of other Italian passive contructions, the impersonal one shows some imperfections in its functioning (sect. 1). If we consider Old Italian, we find that this stage of the language only had the passive si construction (sect. 2) – the formation of the impersonal construction is an innovation and is the result of many small changes in the rules and the domain of application of the passive si construction: these changes, which began in the Old Italian period and lasted for at least five centuries, eventually resulted in the formation of a new construction (sect. 3). But this construction, which is

1

Saying that syntactic change is non-directional, I would not want to imply that the single changes are not ruled by any general principles. Just like analogical change in morphology, syntactic change, too, is governed by general principles of a formal nature, but these principles rule only the single changes and do not link together series of changes. 2 One may wonder what perfection is. We can provisionally assume that perfection is defined by the theory one accepts, hoping that linguistic theories will provide an ever better model of the functioning of human language. 3 As noted by my referee, the existence of corrective changes logically implies that there could be, at least in principle, something like a perfect system, in which no changes are necessary and in which consequently no new imperfections can arise. But human language is a very complex system composed of various subsystems, each with its own function, often linked to other human skills. On the one hand, improvements in one subsystem may imply damage in another one; on the other hand, imperfections in the linguistic system may also be produced by requirements of some other human capacity. So, the general perfection of the system is prevented by the complexity of the system itself.

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE

129

not the result of a unitary project, continues to show signs of being imperfectly put together. The analysis presented below builds on my previous research into the si construction, in particular Salvi & Vanelli (2004) and Salvi (2004). Further facts and analyses on the si construction and its history may be found in Wehr (1995), a very insightful work, the conclusions of which only partially coincide with mine. 1. The si construction in Modern Italian In Modern Italian the construction in which the verb occurs with a reflexive clitic pronoun has many uses. Specifically, it may serve to signal the demotion of the lexical (‘deep’) Subject of the verb. We can distinguish two constructions: the passive si construction, parallel to the normal (periphrastic) passive contruction (1.1), and the impersonal si construction, parallel to the normal active construction of the sentence (1.2). We follow here the analysis of Salvi & Vanelli (2004: sect. I.6.2), to which we refer the reader for more details. There is some variation in the use of the si constructions in Modern Italian: the description below is based on the variety generally spoken in Northern Italy. 1.1 Passive si The passive si construction displays the general characteristics of the passive construction in Italian: the demotion of the lexical Subject (which is generally eliminated) and the promotion of the Direct Object to Subject, as exemplified in (1) and summarized in (2). The attribution of Subject status to the phrase i diplomi “the diplomas” (1b) is based on its agreement in number and gender with the verbal periphrasis (sono “are” = pl., consegnati “handed over” = pl. masc., as i diplomi):4 Il presidente ha consegnato i diplomi agli the president AUX handed.over the diplomas to.the “The president handed over the diplomas to the students” b. Si sono consegnati i diplomi agli studenti si AUX handed.over the diplomas to.the students “The diplomas were handed over to the students” (2) Subject (il presidente) → Ø Direct Object (i diplomi) → Subject Indirect Object (gli studenti) → Indirect Object (1) a.

studenti students

The demotion and elimination of the lexical Subject is only syntactic: as in the normal passive construction, semantically the actant remains present with a 4

The choice of the auxiliary essere “to be” in (1b) instead of avere “to have”, as in (1a), is automatically determined by the use of the reflexive clitic and has no relevance for the interpretation of the construction.

130

GIAMPAOLO SALVI

generic or indeterminate interpretation depending on the context. In a more formal style the demoted Subject can eventually appear as an adverbial modifier (normally introduced by da parte di “by”): (3) Da parte dell’ opposizione si vorrebbe una maggior trasparenza nelle by.the opposition si would.want a bigger transparency in.the decisioni governative decisions governamental “The opposition would want (there to be) more transparency in the government’s decisions”

As in the normal passive construction, only transitive verbs can appear in this construction, as the only ones which have a Direct Object promotable to Subject (Modern Italian does not permit the passive construction with intransitive verbs). As a specific restriction, this passive construction is only possible if the promoted Object is 3rd person, so the only reflexive clitic admitted is 3rd person si. 1.2 Impersonal si In the impersonal si construction the demotion (and elimination) of the lexical Subject is not accompanied by the promotion of any argument to Subject, as exemplified in (4) and summarized in (5). The verb appears in the unmarked 3rd pers. sg. (the past participle may show plural agreement, while the nominal predicates must show some agreement, as in (10)-(11) below; we will not deal with this problem here): (4) a.

b.

(5)

Il presidente li ha consegnati agli studenti the president them AUX handed.over to.the students “The president handed them over to the students” Li si è consegnati agli studenti them si AUX(sg.) handed.over to.the students “They were handed over to the students / They handed them over to the students” Subject (il presidente) → Ø Direct Object (li) → Direct Object Indirect Object (gli studenti) → Indirect Object

In this construction, too, the elimination of the lexical Subject is only syntactic: although it can never be realized syntactically, semantically this actant is present with a generic or an indeterminate interpretation. Unlike passive si, the impersonal si construction is possible with every kind of verb: transitives (6), intransitives, both unergatives (7) and unaccusatives (8), reflexives (9), the copula (10), and also the periphrastic passive (11):5

5

For the use of ci instead of si in (9b), see ex. (16) and below sect. 3.3.

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE (6) a.

b.

(7) a. b.

(8) a. b.

(9) a.

b.

(10) a. b.

(11) a.

b.

131

Piero guarda solo te Piero is.looking.at only you “Piero is looking only at you” Si guarda solo te si is.looking.at only you “They are looking only at you” Piero dorme Piero is.sleeping Si dorme si is.sleeping “Someone is sleeping” Piero va Piero is.going Si va si is.going “One is going” Piero si addormenta Piero (himself) is.sleeping “Piero is falling asleep” Ci si addormenta (ourselves) si is.sleeping “One is falling asleep” Piero è intelligente Piero is intelligent Si è intelligenti si is intelligent “One is intelligent” Piero viene invitato spesso Piero is invited often “Piero is often invited” Si viene invitati spesso si is invited often “One is often invited”

We have seen that both constructions are possible with transitive verbs: this very fact is one of the main arguments for assuming that we really have two distinct constructions. But the intersection in the domain of application of the two constructions is actually more restricted: the impersonal si construction is only applicable to transitive verbs if the Direct Object is 1st or 2nd person (12a) or is a 3rd person pronominal clitic (12b), but not in the case of a non clitic 3rd person Direct Object (12c); the passive si construction is only applicable to transitive verbs if the Direct Object is 3rd person, both nominal or pronominal (13a,b), but not in the case of a 1st or 2nd person Direct Object (13c). The intersection is therefore limited to the case of 3rd person clitic Direct Objects (12b)/(13b) (we assume that the clitic realization of a subject pronoun is a phonetically null pronoun – cf. Salvi & Vanelli 2004: sect. II.6.1):

132

GIAMPAOLO SALVI

(12) a.

b.

c. (13) a.

b.

c.

Si invita me / Mi si invita si invites me / me si invites “One invites me” Li si invita spesso them si invites often “One often invites them” *Si invita i parenti / solo loro si invites the relatives / only them Si invitano i parenti / solo loro si invite(pl.) the relatives / only they “The relatives / Only they are invited” ø si invitano anche per Pasqua (null Subject) si invite(pl.) too for Easter “They are invited for Easter, too” *si/mi invito io si/myself I.invite I

In Modern Italian the si appearing in the impersonal si construction is a different element from the 3rd person reflexive clitic si. We can see this in two types of construction: a) when co-occurring with a 3rd person accusative clitic, the si of the impersonal construction follows it (14a), while the reflexive clitic preceeds it (14b); b) the si of the impersonal construction can co-occur with the clitic lo which substitutes a predicative AP or NP (15a), while the reflexive (or any other) clitic cannot (15b). As for the si appearing in the passive si construction, we have no arguments for classifying it into one class or the other (but in Old Italian, where only the passive si construction existed (sect. 2), we have no reason for distinguishing it from the reflexive clitic): (14) a.

b.

(15) a.

b.

Lo si mangia it si eats “They eat it” Se lo mangia to.himself it he.eats “He eats it (with enjoyment)” Si sembra imprudenti / Lo si sembra si seems careless / so si seems “One seems careless / One seems so” Si è fatto imprudente / *Se lo (himself) he.turned careless / (himself) so “He turned careless”

è fatto he.turned

In spite of the existence of two si’s, the si of the impersonal construction cannot co-occur with a reflexive-reciprocal clitic (16a); this combination is only possible if one substitutes a 1st pers. pl. clitic (ci) for the 3rd person reflexive (16b). Notice that other languages (e.g. the dialect of Venice – cf.

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE

133

Lepschy 1989: ch. 9) admit the co-occurrence of an impersonal and a reflexive si (16c): (16) a. b.

c.

*Si si vede spesso himself (= each.other) si sees often Ci si vede spesso us (= each.other) si sees often “People often meet” Dialect of Venice El se senta / Se se senta he (himself) sits.down / si (himself) sits.down “He sits down / One sits down”

In the light of the above description, we can see that the impersonal si construction shows at least two ‘imperfections’, i.e. restrictions which are not motivated by the general properties of the language: a) the construction is not possible if the Direct Object is 3rd person and is not a clitic pronoun; b) although in Modern Italian impersonal si and reflexive si are two different morphemes, they cannot co-occur (this second restriction would be motivated if the two si’s were the same morpheme, for in a clitic cluster a clitic can only occur once). 2. Old Italian Old Italian only had the passive si construction (2.1); at that time the impersonal si construction did not exist (2.2). By Old Italian we mean here the language (spoken and) written in Florence in the 13th c. and in the first quarter of the 14th c. The materials presented in this section were collected in the frame of the project ItalAnt, whose aim is a complete synchronic description of the grammar of Old Italian (cf. Renzi 2000). The description of the si construction may be found in Salvi (2004: sect. 3). 2.1 Passive si In Old Italian the construction of the verb (in the 3rd person) with the reflexive clitic could have the function of the passive construction, as is shown by the following examples, where the passive si construction freely alternates with the active (17a) and the periphrastic passive (17b), expressing the same meaning: (17) a.

di ciò non faccia alcuno consciença non volemo che in about this not make(subj.) anyone conscience not we.want that in dire questi pater nostri per alcuno si faccia consciençia saying these Lord’s.prayers by anyone si make conscience “no one should feel any twinges of conscience about this / we do not want any twinge

134

GIAMPAOLO SALVI of conscience to be felt by anyone for not having said the prescribed Lord’s prayer” (Capitoli della Compagnia della Madonna d’Orsammichele (1297), 57 and 66) b. Furono in questa concordia (…) che cierta quantità di pane e they.were in this agreement that certain quantity of bread and d’ altre cose (…) si dovessero vendere, e fuorono vendute a of other things si should(pl.) sell and they.were sold to cierte persone certain people “They agreed (...) that a certain amount of bread and of other things (...) should be sold, and they were sold to certain people” (Ordinamenti della Compagnia di Santa Maria del Carmine, 62.6)

As in the normal passive construction, the demoted Subject could be freely expressed as an adverbial modifier (per alchuno “by anyone” in (17a)). Intransitive unergative verbs could also occur in this construction (18), but this must not be taken as an argument for the existence of an independent impersonal construction (unlike the existence of (7a) for Modern Italian), for in Old Italian unergative verbs could also appear in the periphrastic passive construction (19) (which is impossible in Modern Italian): (18)

(19) a.

b.

Leggesi della bontà del re giovane… reads.si about.the worth of.the king young “One can read about the young king’s worth...” (Novellino, 18.3) fue del mese di settenbre battalglato fortemente con was in.the month of September fought hard with vij difici 7 war.machines “in the month of September they fought hard with seven war machines” (Cronica fiorentina, 125.6) ch’ a llui fosse dato d’ uno bastone that to him should.be given with a staff “that he should be hit with a staff” (Cronica fiorentina, 118.33)

2.2 *Impersonal si In Old Italian the impersonal si construction did not exist. Whenever the reflexive construction is used to signalize the demotion of the lexical Subject, in Old Italian it shows the general properties of the normal passive construction: a) with transitive verbs, the promotion of the Direct Object is obligatory, so that examples such as (12a-b) were not possible. A case like (20) is not a counterexample to this generalization, because the postverbal NP (soldi XL “40 soldos”) is not the Direct Object of the verb (in spite of the lack of agreement), but the Subject: the lack of agreement with the verb was normal in Old Italian with a postverbal Subject in unaccusative constructions (21), including the passive construction (22):

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE

(20)

(21) a.

b.

(22)

135

s’ ordinoe e fermò che per questa pasqua di Natale proxima, che ora si ordered and fixed that for this Christmas next that now dé venire, si dea soldi XL a’ poveri, per l' amore is.to come si gives(subj.) soldos 40 to.the poor for the love di Dio, al modo usato of God in.the manner usual “it was ordered and fixed that for this next Christmas which is now to come 40 soldos be given to the poor for God’s sake in the usual manner” (Ordinamenti della Compagnia di Santa Maria del Carmine, 65.32) In questo tenpo, nel Mccx, si convertio alla fede di Cristo in this time in.the 1210 was.converted to.the faith of Christ lxxm. uomini 70.000 men “In this time, in 1210, 70.000 people were converted to the Christian faith” (Cronica fiorentina, 115.20) Della buona volontà, di cui nasce le quattro virtú cardinali of.the good will from which is.born the four virtues cardinal “Of the good will, from which the four cardinal virtues are born” (Bono Giamboni, Virtú e Vizî, 2(titolo)) sia inposto loro degnia penitentia be imposed(masc.) on.them appropriate penitence(fem.) “Let an appropriate penitence be imposed on them” (Capitoli della Compagnia di S. Gilio, 52.25)

b) of the intransitive verbs, only the unergatives were possible in this construction, as in the periphrastic passive, as discussed in 2.1. The typical unaccusative verbs (simple unaccusatives, as in (8), reflexives, as in (9), copula, as in (10), and the passive periphrasis, as in (11)) were not admitted. The situation described above is characteristic for the 13th c. But we must also notice that in the works of Dante, starting from the last years of the century, the construction begins to be used with (non-reflexive) unaccusative verbs, too, mainly with verbs of movement (but not with the copula): (23) a.

b.

Io tenni li piedi in quella parte de la vita di là da la quale I held the feet in that part of the life beyond which non si puote ire più per intendimento di ritornare not si can go more with intention of returning “I have just set foot on that boundary of life beyond which no one can go, hoping to return” (Dante, Vita Nuova, 14.8, transl. by Mark Musa) non vuol che ’n sua città per me si vegna not he.will that into his city by me si goes(subj.) “[He] wills not (...) that I should make my way into His city” (Dante, Commedia, 1.1.126, transl. by Jean and Robert Hollander)

We would like to suggest that this usage was an innovation introduced by Dante himself, probably as a way of rendering a construction possible in Latin, where unaccusative verbs, too, had a passive form with impersonal

136

GIAMPAOLO SALVI

interpretation (itur “one goes”). In any case, in Old Italian these examples are cases of passive construction, as is shown by the use of the adverbial modifier per me “by me” in (23b) to express the demoted Subject.6 Here we have, therefore, an extension of the domain of the passive si construction, and not (yet) the birth of the impersonal si. An analogous extension of the periphrastic passive construction to unaccusative verbs may be observed in the course of the 14th c.: (24) L’ altra mattina seguente fu andato alla campana da casa Tornaquinci the other morning following was gone to.the bell near house T. “On the following morning someone went to the bell near the Tornaquinci house” (F. Sacchetti, Trecentonovelle, 78.50)

Notice also that in Old Italian the passive si construction with unaccusative verbs was possible only in simple tenses: we had Per me si va... “Through me one goes...”, but not *Per me si è andato... “Through me one went...”. This restriction is perhaps related to the one which is active in Modern Italian prohibiting the use of the impersonal si construction with unaccusative verbs to describe non generic perfective events (Oggi si è andati a pescare cannot mean “Today someone went to fish” – in fact it means “Today we went to fish”, an interpretation to which we come back below; cf. Cinque 1988). 3. Intermediate stages In the previous sections we have seen that Old and Modern Italian differ sharply in the use of the si construction with demotion of the lexical Subject and generic or indeterminate interpretation: while in Old Italian this construction perfectly paralleled the normal (periphrastic) passive construction, Modern Italian distinguishes two separate constructions, one roughly parallel to the passive construction and a new one, with characteristics not shared by the other. This new construction does not appear all at once: in the course of the history of the Italian language many small changes were introduced to the original construction. These changes modified the functioning and the domain of application of the passive si (the first change was the extension of the construction to unaccusative verbs, discussed in 2.2), until a new construction began to take shape. We have shown that this process is not concluded: there continue to be ‘imperfections’ in the system and this is due to the fact that the formation of the new construction is not part of a general project, but the sum 6 Notice that I distinguish impersonally used passive constructions from impersonal ones: both types lack the promotion of an argument to Subject, but in the first one this lack is contingent on the class of the verb, whilst in the latter one it is general. So, the Old Italian si construction is a passive one, but is used impersonally when it applies to an intransitive verb (which has no Direct Object that could be promoted to Subject). The Modern Italian impersonal si construction, on the other hand, is always impersonal.

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE

137

of smaller changes, in part causal and only in part directed to the correction of the existing imperfections. In what follows we will show some of the steps that led to the formation of the impersonal si construction. The materials we will discuss below are the fruit of a rapid survey undertaken of the electronic edition of some literary texts between 16th and 19th c. (available online: www.liberliber.it) and does not have any pretension of completeness. In particular we could not study in any detail the development of the use of the si construction to express the 1. pers. pl. (Noi si parte domani (we si leaves tomorrow) “We are leaving tomorrow”), which presumably played an important role in the shaping of the innovative syntactic properties of the impersonal si construction. In spite of this, this first approach is sufficient to show that the syntactic features of the impersonal si construction appeared only gradually in the course of several centuries.7 3.1 Stage I Some innovations appear at the beginning of the 17th c.: in the Istoria del Concilio Tridentino of Paolo Sarpi (published in 1619) and in the Dialogo sopra i due massimi sistemi del mondo of Galileo Galilei (published in 1630) the si construction is possible with essere “to be” (25) (but apparently not with essere + predicative AP or NP) and with the compounded tenses of unaccusative verbs (26). These two innovations may be related, for the auxiliary of the unaccusatives is essere. Notice that the past participle, in this new combination, is not in the plural, as it would be in Modern Italian: (25) a.

dimostrò (…) già tanti d’ essi aver approvato quel capo senza he.showed already many of them to.have approved that point without contradizzione, et ora venir alcuni con nuovi partiti, quali, contradiction and now to.come some with new proposals which quando si sarà al ristretto, saranno da altri contradetti when si will.be at.the fight will.be by others contradicted “(he) showed (...) that many of them had already approved that point without contradiction and now some came with new proposals which, in the decisive debate, would be contradicted by others” (P. Sarpi, Istoria del Concilio Tridentino, l. VI)

7 In the following sections, after identifying the innovations, I propose a diachronic explanation of the individual changes, mainly based on analogy and language contact. Due to the provisional state of the research and the insufficiency of the material collected, these proposals are no more than guesses waiting for a more profound study. In proposing these explanations, my aim is mainly to give concrete examples of the type of local changes I have in mind for the explanation of apparently more global changes.

138

GIAMPAOLO SALVI

b. Luoghi possibili sono questi de’ quali si è in controversia places possible are these about which si is in controversy “The possible places are those about which there is controversy...” (G. Galilei, Dialogo sopra i due massimi sistemi del mondo) (26) a. Questo luogo ricerca, per le cose dette e che si diranno in this section examines for the things said and that si will.say in varie occasioni circa il modo di dire i pareri in various occasions about the manner of saying the opinions in concilio, (...) come anticamente si faceva e come s’ è pervenuto council how in.old.times si did and how si is come(sg.) all’ usato in questi tempi to.the used in these times “This section examines, because of what has been said and will be said about the manner of expressing one’s opinion in the Council, (...) how this was done in olden times and how one has come to the manner used nowadays” (P. Sarpi, Istoria del Concilio Tridentino, l. II) b. non sarebbe questa la prima cosa che egli, per mancanza di not would.be this the first thing which he for lack of strumenti o per altro difetto, non ha saputa, e pur, fondato instruments or for other deficiency not has known and still based sopra altre saldissime conietture, affermò quello a cui parevano on other very.firm deductions affirmed that to which seemed contrariare le cose non comprese da lui (...); tuttavia si è to.contradict the things not understood by him nevertheless si is poi venuto in certezza, tali mutazioni esservi a capello afterwards come(sg.) in certainty such variations to.be.there to a.hair quali ricercava il sistema copernicano which searched.for the system copernican “This would not be the first thing that he failed to know, either for lack of instruments or from some other deficiency. Yet, grounded upon most solid theories, he affirmed what seemed to be contradicted by things he did not understand (...). Yet since that time it has become certain that such variations are, to a hair, just what the Copernican system required” (G. Galilei, Dialogo sopra i due massimi sistemi del mondo, trasl. by Stillman Drake)

These innovations represent natural extensions of the domain of applicability of the construction: beside the unaccusative verbs of movement we find the main verb of state, essere (only when expressing location); and besides the simple tenses of the unaccusative verbs we also find the compounded ones. 3.2 Stage II New changes appear around the end of the 18th c. and the beginning of the 19th c.: in the autobiography of Vittorio Alfieri (written between 1790 and 1803) and in the first version of Alessandro Manzoni’s novel I promessi sposi (written in 1821-23 with the title Fermo e Lucia) the si construction admits the occurrence of essere + predicative AP/NP (27). Here the adjective shows the plural form, while in the compound tenses the participle continues to show the

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE

139

unmarked singular form (28a); but the presence of an example such as (28b) allows us to suspect that the plural was also a possible option: here the participle of an unergative verb is in the plural (in Modern Italian, as in Old Italian, we would expect the singular), presumably as the result of a faulty extension of the rule applicable to the participles of unaccusative verbs (but not attested in these two works). Another innovation is the use of an accusative clitic of 1st or 2nd person (29), a possibility which is completely excluded in the older language, where the si construction is always passive in the case of transitive verbs. Notice that in (27a) and (28b) the construction has a 1st pers. pl. interpretation: (27) a.

b.

(28) a.

b.

(29)

saliti poi innanzi ai birbi della Municipalità, si era certi gone.up then before.the villains of.the municipality si was certain(pl.) di non poter piú partire, d’ andare anzi prigioni, dove se ci of not being.able more to.leave of going even prisoners where if us trovavano nelle carceri il dí 2 settembre, cioè quindici giorni they.found in.the prison the day 2 September that.is fifteen days dopo, ci era fatta la festa insieme con tanti altri later to.us was made the feast together with so.many other galantuomini che crudelmente vi furono trucidati gentlemen who cruelly there were slaughtered “once we had gone up before the villains of the municipality, we were certain that we could not leave anymore, that we would even be thrown into jail, in which case, if they had found us in the prison on 2nd September, that is fifteen days later, they would have murdered us together with so many other gentlemen who were cruelly slaughtered there” (V. Alfieri, Vita, c. 22) quando si è persuasi d’ una verità bisogna dirla when si is persuaded(pl.) of a truth one.must tell.it “when one is persuaded of the truth of something, one must tell it” (A. Manzoni, Fermo e Lucia, 1.3) allorché finalmente si è venuto ad offenderli involontariamente when in.the.end si is come(sg.) to offend.them... unintentionally “when in the end one has offended them unintentionally...” (V. Alfieri, Vita, c. 13) mi stava seduto in faccia all’ altra parte di quella assai ben to.me he.was sitting facing at.the other part of that very larga tavola dove si era cenati wide table where si was dined(pl.) “(he) was sitting facing me at the other end of that very wide table we had been dining at” (V. Alfieri, Vita, c. 12) Alcuni giorni dopo egli mandò ambasciata a voce, per sapere in some days later he sent message orally in.order.to know in che ore mi si potrebbe trovare which hours me si could find “Some days later he sent an oral message in order to find out at what times I could be found” (V. Alfieri, Vita, c. 12)

The first innovation is an extension of the construction to a further context: not only essere in its locational use, but also in its use as a copula. The second

140

GIAMPAOLO SALVI

one is perhaps tied to the use of the construction as a 1st pers. pl.: initially the si construction may have been used with this new interpretation only in the syntactic forms also admitted by the generic/indeterminate interpretation, but there may soon have arisen a tendency to extend the syntactic possibilities after the model of normal active 1st pers. pl. forms (another, concurrent factor may have been the desire to render the French construction with the generic/indeterminate subject on). This innovative tendency must have been limited by the availability of existing formal models. In our case this formal model may have been the combination of the clitic si with a clitic pronoun of 1st or 2nd person: already in the medieval language these pronouns, which in Italian do not have separate forms for accusative and dative case, might appear in combination with si when they were dative, as in (30): (30) E nel forbire che fece, parve che degli occhi mi si and in.the cleaning that she.made it.seemed that from.the eyes to.me si levasse una crosta di sozzura puzzolente di cose terrene removed a crust of dirt stinking of things earthly “And when she cleaned them, it seemed as if a crust of stinking dirt from mundane things were removed from my eyes” (Bono Giamboni, Vizi e Virtudi, 3.2)

We may therefore imagine the following evolution (for which we so far lack adequate evidence): on the basis of the parallelism between Noi ti/vi diciamo and Noi ti/vi si dice (both “We tell to you”), the model of Noi ti/vi vediamo favoured the creation of Noi ti/vi si vede (both “We see you”); this usage was then extended to the generic/indeterminate construction and to 1st person clitics. Notice that the plural agreement, too, may be due to the influence of the use of the construction to express 1st pers. pl. 3.3 Stage III Around the middle of the 19th c. the modern system is well established in the works of some northern Italian writers (we have examined Giuseppe Rovani’s Manfredo Pallavicino (1845/46) and Cento anni (1856/65), and Ippolito Nievo’s Le confessioni di un italiano (1857/58)). In particular, although the plural agreement in the past participle with unaccusative verbs may be missing (31a), agreement is clearly possible (31b) (this example expresses 1st pers. pl.). Important innovations are the usage with reflexive verbs (32), with the periphrastic passive (33) (ex. (33b) expresses 1st pers. pl.), and with 3rd pers. accusative clitics (34) (the use of the auxiliary avere “to have” instead of essere “to be” in (34a) is a characteristic of the northern

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE

141

Italian dialects, here transferred to the literary language — cf. Mengaldo 1987:75):8 con tante istituzioni e che so io, si è giunto ovunque with so.many institutions and what know I si is come(sg.) everywhere a rendere di un calore uniforme i prodotti delle varie to render of a warmth uniform the products of.the various intelligenze. intelligences “with so many institutions and God knows what else, the result is everywhere that the various minds’ products have been rendered of a uniform temperature.” (G. Rovani, Manfredo Pallavicino, c. 1) b. in sostanza si era tornati alla vita di prima in substance si was come.back(pl.) to.the life of before “we had in substance come back to the former life” (I. Nievo, Le confessioni di un italiano, c. 6) (32) la guerra è come un giuoco, che quanto è più forte la perdita, the war is as a play that how.much is bigger the loss tanto più ci si ostina, e si continua finchè ci so.much more (himself) si is.obstinate and si continues until there è vita is life “the war is as a play, where the bigger the loss, the more obstinacy one shows, continuing until there is life” (G. Rovani, Manfredo Pallavicino, c. 3) (33) a. Se hanno vera vocazione, partano pure; (...) bisogna correre ove if they.have real vocation let.them.leave just one.must run where si è chiamati, altrimenti non si fa nulla di bene. si is called(pl.) otherwise not si does nothing of good “If one has a genuine vocation, let him just leave; (...) one must run where one is called, otherwise one does nothing good” (I. Nievo, Le confessioni di un italiano, c. 20) b. mettendoci a contatto con quei fatti stessi (...), ben si è putting.ourselves to contact with those facts very indeed si is tratti a conchiudere che… led(pl.) to conclude that “considering directly those very facts (...), we are indeed led to conclude that...” (G. Rovani, Cento anni, 12.7) (34) a. un corpicciuolo piú meschino e magagnato del suo, non lo si a little.body more poor and wretched than his not it si avrebbe trovato cosí facilmente would.have found so easily “a poorer and more wretched little body than his would not have been so easy to find” (I. Nievo, Le confessioni di un italiano, c. 1) (31) a.

8

Cf. also, from the letters of Nievo: si ci dimentica (si us forgets) “one forgets”, se lo pesa (si it weighs) “one weighs it”, se li potrà spedire (si them will-be-able to-send) “one will be able to send them” (Mengaldo 1987:67), with the order si + clitic (instead of clitic + si) typical of Venetian (Lepschy 1989).

142

GIAMPAOLO SALVI b. se ognuno sa, almeno lo si diceva da gran tempo, che l' erede if everybody knows at.least it si said for a long time that the heir universale di tutte le sue sostanze era suo fratello, il conte Lodovico universal of all the his properties was his brother the count L. “...if everybody knows (at least they had said it for a long time) that the universal heir of all his properties was his brother, the count Lodovico” (G. Rovani, Cento anni, 1.7)

In the extension of the si construction to reflexive verbs and to the periphrastic passive, too, the usage expressing 1st pers. pl. may have been determinant: the model of Noi laviamo / Noi ci laviamo may have produced Noi si lava / Noi ci si lava (both “We wash / We wash (ourselves)”). This hypothesis would also offer an explanation for the use of the 1st pers. pl. clitic as a substitute for the 3rd person reflexive clitic. The formal model, that is the combination of clitics ci si, existed from the beginning: see (35a), with dative ci, and (35b), with the innovative use of accusative ci (see above sect. 3.2): (35) a.

b.

quello siamo tenuti di volere più sollicitamente, lo quale non ci that we.have to want more eagerly which not to.us si dà per lo gran pericolo che nn’ è si gives for the great danger which from.it is “we have to want more eagerly that which is not given to us because of the great danger that comes from it” (De Amore di Andrea Cappellano volgarizzato, 1.18) Ma la funzione è protratta... e noi ci si manda but the ceremony is delayed and us(ton.) us(cl.) si sends a Malamocco to M. “But the ceremony is delayed... and, as for us, we are sent to Malamocco” (G. Rovani, Valenzia Candiano, c. 12)

We may suppose a parallel process in the case of the periphrastic passive: Noi vediamo / Noi siamo visti may have brought about Noi si vede / Noi si è visti (both “We see / We are seen”). Here the formal model was the structure essere + predicative adjective (see above sect. 3.2). The introduction of 3rd person accusative clitics in the si construction has a different explanation: it is due to a faulty interpretation of a construction of (colloquial) Tuscan by non-Tuscan speakers. In sentences such as (36a), Tuscan has a case of passive si construction with the Subject expressed by a clitic pronoun (the 3rd pers. feminine la); the corresponding masculine is (36b); in the plural the verb agrees with the clitic Subject (36c): (36) a.

b.

La si invita spesso she si invites often “She is often invited” E’ si invita spesso he si invites often “He is often invited”

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE

143

c. E’ / Le si invitano spesso they(masc.) they(fem.) si invite(pl.) often “They are often invited”

But the form la of (36a) is homophonous with the 3rd pers. fem. accusative clitic, and non-Tuscan speakers who already used 1st and 2nd pers. accusative clitics in the si construction, interpreted the pronoun in (36a) as an accusative clitic, and extended this usage to the masculine, creating (37a), and to the plural, creating (37b) (with the verb not agreeing): Lo si invita spesso him si invites often “He is often invited” b. Li / Le si invita spesso them(masc.) them(fem.) si invites often “They are often invited”

(37) a.

4. Conclusions Our reconstruction of the history of the si construction has no pretensions to completeness, and it may be inexact in many points. While we have a fairly good understanding of the Old and the Modern Italian constructions, this is not true for the intermediate stages. The rapid survey we made in sect. 3 did not permit a detailed reconstruction of the different stages which led to the present state of affairs. To achieve this, we would need a description of the functioning of the si construction in every relevant period of time based on a much wider collection of data in texts of different origin. It is only with such a background that we could reconstruct the process which resulted in the formation of an impersonal si construction in Modern Italian. We had to limit our study to the search for the introduction of innovations in the si construction (without supplying the systems which promoted/received these innovations), but even within this more limited frame we have reached some interesting conclusions. In particular, we have found that the formation of the impersonal si construction is the sum of several individual innovations due to different causes: the extension of the original passive si construction to unaccusative verbs of motion (perhaps due to Latin imitation) and of state (via generalization), its extension to other contexts previously excluded as copular constructions, reflexive verbs, periphrastic passive, constructions with accusative clitic pronouns of 1st and 2nd pers. (probably as a consequence of the usage of the construction to express 1st pers. pl.), and finally to constructions with accusative clitic pronouns of 3rd pers. (as a misinterpretation of an existing structure by non-native speakers). This series of

144

GIAMPAOLO SALVI

successive and partly independent innovations created a construction which is not completely coherent – because it is not the result of a coherent project.9

REFERENCES

Cinque, Guglielmo. 1988. “On si constructions and the Theory of Arb”. Linguistic Inquiry 19. 521-581. Lepschy, Giulio C. 1989. Nuovi saggi di linguistica italiana. Bologna: il Mulino. Matthews, Peter H. 1981. Syntax. Cambridge: Cambridge University Press. Mengaldo, Pier Vincenzo. 1987. L’epistolario di Nievo: Un’analisi linguistica. Bologna: il Mulino. Renzi, Lorenzo. 2000. “‘ItalAnt’: come e perché una grammatica dell’italiano antico”. Lingua e stile 35. 717-729. Salvi, Giampaolo. 2004. “La realizzazione sintattica della struttura argomentale”. Grammatica dell’italiano antico, ed. by Giampaolo Salvi & Lorenzo Renzi. Work in progress (http://geocities.com/gpsalvi/).

9 One could look at this whole story in a completely different manner, as my referee notes: Standard Italian has nowadays an impersonal construction which is almost perfect, so one could think that “behind the small, visible syntactic changes, there may also be a ‘big leap’, like the entrenchment of impersonal meaning for an originally passive construction”. My position is that there is no big leap: the local corrections can assume a well-defined direction, but, if this happens, it is because the individual changes (when they correct an imperfection or extend the scope of a rule) are based on constructions which already exist, exactly as in the case of morphological analogy; so the changes can have a cumulative effect. But there is no long-term project and the evolution can be interrupted at any stage. In other words, I do not deny that successive changes can be coherent and move in the same direction, but this coherence is not determinated by a hidden (teleological) plan – it can only be determined by the linguistic state preceding the change. The gradual correction of single imperfections may lead to a perfect subsystem, but the fact that the systems can continue functioning with their imperfections is itself proof that there is no long-term plan in diachronic change. As in the case of the more general concept of grammaticalization, long-term plans do not have the status of explanatory tools: they can at most be descriptions a posteriori of what happened. It is from the above perspective that we must view the fact, noted by the referee, that “in many languages, passive constructions (of all kinds) are either used impersonally (French: cela ne se dit pas, Germ: hier wird nicht gelacht) or acquire impersonal structure (Spanish, se vende coches ). [...But] the reverse is not true: impersonals normally show no signs of turning into passives”. In the unmarked case, passive constructions entail the syntactic demotion of the lexical subject of the construction, but this subject remains semantically active and has often an indefinite (“someone”) or a generic (“people”) interpretation. It is this indefinite/generic interpretation of the lexical subject of a passive construction that is the basis for the diachronic change which eventually creates an impersonal construction. On the other hand, the impersonal construction cannot normally be the basis for a passive construction because in the impersonal construction the other constitutive feature of the passive construction, the promotion of the Direct Object to Subject, is missing, and it is not evident from where the speakers could derive it.

IMPERFECT SYSTEMS AND DIACHRONIC CHANGE

145

---------- & Laura Vanelli. 2004. Nuova grammatica italiana. Bologna: il Mulino. Wehr, Barbara. 1995. SE-Diathese im Italienischen. Tübingen: Narr.

FROM TEMPORAL TO MODAL DIVERGENT FATES OF THE LATIN SYNTHETIC PLUPERFECT IN SPANISH AND PORTUGUESE1

MARTIN G. BECKER University of Stuttgart

This corpus-based study examines the different historical stages the Latin synthetic pluperfect underwent in Spanish and Portuguese. It tries to explain parallel and divergent developments of the morpheme -ara leading to opposite results: the strengthening of the form as an exclusively temporal category in Portuguese and its transformation into a fully-fledged past subjunctive in Spanish. In our analysis we will focus on three main aspects: first, the relevance of the underlying logical-conceptual structures as the enabling condition for the expansion of the category into new domains; second, the role of reanalysis as a listener-based strategy in discourse as the decisive mechanism for functional change; and third, the validity of the prototype concept as a device to explain the organisation of grammatical categories from a diachronic perspective. With our case study we address key questions of this book, namely the locus, the motivation and the nature of linguistic change.

1. A puzzle It is an astonishing fact that the morpheme -ara (as in cantara), the Romance derivative of the Latin pluperfect morpheme -(a)verat (e.g. cantaverat), has evolved so differently in the closely related languages Spanish and Portuguese. In the former, -ara evolved into a fully-fledged subjunctive imperfect form progressively ousting the inherited morpheme -ase, which has derived from the Latin subjunctive pluperfect cantavisset. This evolution reveals itself in the competition of the variants -ara and -ase in final clauses: a query in Davies’ Corpus of Spanish illustrates the swift advance of para que + V_ara during the 19th century, with para que + V_se being finally left behind in the course of the 20th century. Table 1 gives an overview of the evolution and the ratio between the competing forms:2

1

I am grateful to four anonymous reviewers for their useful comments on previous drafts. Davies’ Corpus del español at <www.corpusdelespanol.org> also provides for queries which include grammatical categories. 2

148

MARTIN BECKER

para que V_SE V_RA Ratio

1200 5 0 100:0

1300 19 1 95:5

1400 44 0 100:0

1500 760 41 95:5

1600 345 65 84:16

1700 349 19 95:5

1800 325 241 57:43

1900 131 520 80:20

Table 1: the ratio between V-ase and V-ara in a diachronic perspective

In contrast to the evolution of -ara in Spanish, the pluperfect reading of the morpheme has been stable in Portuguese until the present, leaving aside the fact that in contemporary Portuguese -ara has a markedly literary character, if not an archaic flavour. The Spanish development of -ara was considered as an interesting example for a subjunctive category deriving from an indicative one (Bybee, Perkins & Pagliuca 1994:234-235). This type (or cline) of grammatical change was exemplified for the first time in Klein-Andreu’s study “Losing ground: A discourse-pragmatic solution to the history of -ara in Spanish”, in which the author gives a straightforward account of the reinterpretation of the morpheme as a subjunctive form. In her view, -ara was relegated to “low focus” reference contexts, in which it became associated with the notion of “non assertiveness” and specialised as an exclusive subjunctive marker (Klein-Andreu 1991:173). This view echoes a particular conceptualisation of the subjunctive mood, which has been put forward by Lunn (1989 and 1995) and elaborated into a coherent theory by Haverkate (2002) within the framework of a communicational approach. The mood system – be it described semantically with features like (non-) veridicality or pragmatically by invoking phenomena like backgrounding and defocalising – is traced back to the all-encompassing principle of relevance (Haverkate 2002:198) according to which, the subjunctive can be restated as being the marker of “less-than-optimally relevant information” (Lunn 1989:251). However, we believe that this pragmatic explanation is too coarse-grained for several reasons. First, non-assertiveness is far from being a sufficient criterion for the characterisation of the subjunctive mood. Hypothetical constructions or factive verbs (the latter do not assert, but presuppose the truth of their complement clauses, Kiparsky & Kiparsky 1970:143ss.) can by no means be identified with the subjunctive mood – on the contrary, they are compatible with a broad variety of grammatical categories within the domain of Romance.3

3 A dichotomic view (indicative vs. subjunctive) is also held by Luquet (cp. Luquet 1988:312), whereas Veiga introduces the irrealis as a category sui generis (Veiga 1996:22ss.). The problem with Veiga’s correct intuition is that he does not distinguish between the conceptual and categorical levels, for he uses the label irrealis commensurate with indicative and subjunctive.

FROM TEMPORAL TO MODAL

149

Second, Klein-Andreu’s study is based on a single text (the Libro del Conde Lucanor composed about 1335), which cannot account for the diachronic evolution but only mirrors in synchrony the general distribution of the two pluperfect variants, the -ara-morpheme and the compound past perfect. So her study, as a snapshot of a single linguistic situation in the 14th century, turns out to be less persuasive than it appears at first glance. Above all, it does not attest to a clear rift between the compound past perfect and -ara with respect to the relevant parameters underpinning the main thesis: -ara betrays only a slight inclination towards the stative pole – being less dynamic, less animate and less affirmative. However, these modest tendencies can be easily explained away. As Klein-Andreu’s survey indicates, -ara appears predominantly in relative clauses which, by their very nature, are destined to modify a main-clause argument by adding descriptive content (their so-called adjectival function). The main goal of our article is to give a diachronic analysis of the evolution of -ara in Spanish and Portuguese on the basis of an enlarged corpus comprising key texts in both languages and ranging from the Middle Ages (e.g. El Poema de Mio Cid, a Primeyra Partida) to the beginning of the 17th century in the case of Spanish (Mateo Alemán), and to the end of the 19th century for Portuguese (Eça de Queirós). The comparison will be all the more enlightening as the Spanish evolutions are paralleled in Portuguese within a particular conceptual area but diverge with respect to further developments. Portuguese never crosses the borderline of mood change and even evolves back to a purely temporal form during the 19th century. At the end of our analysis, we will try to elucidate some fundamental questions as to what motivates the changes, what explains common and divergent paths of development, what are the cognitive and pragmatic mechanisms which underlie these developments and finally: does change have an inherent directionality? 2. The historical situation in Old Spanish Texts dating back to the 13th century – el Poema de Mio Cid (circa 1207) and Gonzalo de Berceo’s Milagros de Nuestra Señora – display an ideal landscape for purely temporal pluperfect uses of -ara in characteristic contexts, such as: - in temporal clauses: (1) Miedo á su mugier e quiérel’ quebrar el coraçón, assi ffazie a las dueñas e a sus fijas amas a dos; del dia que nasquieran non vieran tal tremor. (Poema de Mio Cid, 1660-1661) “His wife was scared and it nearly broke her heart; the ladies, both his daughters, felt the same; since the day of their birth (i.e. they had been born), they had never witnessed such a dread.”

- in relative clauses:

150

MARTIN BECKER

(2) murió enna eglesia do fuera visitado, / fue en est logar misme el cuerpo soterrado. (Berceo, Milagros de Nuestra Señora, v. 902) “He died in the same church where he had been visited and he (his body) was buried in the same place.”

- depending on epistemic verbs and verba dicendi: (3) parezió-l a un clérigo de buen entendimiento, dissoli que fizieran en ello fallimento. (Berceo, Milagros de Nuestra Señora, v. 105) “She appeared to a sensible clergyman and told him that they had been wrong in doing so.”

- combined with verbs of perception: (4) Vidieron que viniera esto por la Gloriosa, ca otri non podrié fazer tamaña cosa. (Berceo, Milagros de Nuestra Señora, v. 114) “She saw/understood that this had happened thanks to the Glorious Virgin.”

- after the verb parecer: (5) Issió muy bon omne, en todo mesurado, parecié bien que fuera de bon amo criado. (Berceo, Milagros de Nuestra Señora, v. 578) “He became a very good person, well-balanced in everything; it seemed that he had been brought up by a good master.”

- in causal clauses (introduced by ca): (6) San Peidro el apóstol ovo d’él compassión, ca en su monesterio fiziera professión. (Berceo, Milagros de Nuestra Señora, v. 164) “The apostle St. Peter felt pity for him, because it was in his monastery that he had taken the vows.”

- in indirect questions: (7) Non sabién de cuál guisa fuera ocasionado. (Berceo, Milagros de Nuestra Señora, v. 195) “They didn’t know in which way it had happened.”

Nevertheless, -ara does not only occur in subordinated clauses, but expands after 1250 into the domain of the main clause. As the testimonies of the literary activity at the court of Alfonso el Sabio – the translation of the Arabic book Calila e Dimna (circa 1250) as well as the two conspicuous Chronicles (Estoria de Espanna and the General e Grand Estoria, 1270) – show, -ara alternates with the Indefinido, constituting the “informational focus” of the narrative plot and thus, taking over the function of a simple aoristic past tense as in the following example:

FROM TEMPORAL TO MODAL

151

(8) la primera, que los fallaran usados en razonar, e trobáronlos, (…), para decir encobiertamente lo que querían, et por afirmar buenas razones; e la segunda es, que lo fallaron por buena manera con los entendidos (...); la tercera es, que los fallaron por juglaría a los discípulos e a los niños. (Calila e Dimna, Introducción de Abdalla Ben Almocafa) “First, they gave them the faculty for reasoning and designed them (…) to express their will openly and to state good reasons for it; second, they made them ready to deal with intelligent persons (…); and third, they made them appropriate by good art to pupils and children.”

The aoristic function of -ara in narrative contexts is particularly prominent in the Chronicles as the following example, taken from the Estoria de Espanna, illustrates: (9) E muchos otros iudios de grand logar, que fuxieran otrossi de la uilla et se fueran pora Tito, dizien que sexcientas uezes mil iudios muertos fueran todos contados que echaran por las puertas de la cibdat; e los otros, que por la grand muchedumbre non pudieran seer echados, que eran tantos que no auien cuenta ninguna. (…). Et Tito Cesar mandara por todas sus huestes que les uendiessen quanto mester ouiessen, (...). (Alfonso el Sabio, Prosa histórica: Estoria de Espanna, 78s.) “And many highly esteemed Jews who fled from the town as well and sought refuge with Titus said that, all in all, six hundred thousand Jews had died and had been thrown out of the town; and all the others, due to their large quantity, could not be thrown out, and they were so many that it was impossible to count them all. (…). And Titus Cesar ordered his army to sell them whatever they needed, (…).”

The use of -ara in aorist function, as a variant of the Indefinido, is not restricted to the period and milieu of Alfonso El Sabio alone, but can also be found later on in Pedro López de Ayala’s Libro de Palacio, a 14th century text, which has recourse to both forms, -ara and the Indefinido occurring jointly in the same context: (10) Él las (las cosas) estableció, por que el pobre homne que Él formara e crio allende no pasase de lo que Él mandó. (López de Ayala, Libro Rimado de Palacio, v. 6424-6426) “He made things that way in order to prevent the miserable man he (had) created and brought up from violating the rules he ordered.”

As our short survey illustrates, -ara occurs in a multiplicity of different contexts, which vary individually as to their degree of informational relevance, so that it makes little sense to oppose low and high focus contexts in a categorical way. By the same token, -ara is even up to co-occurring with the indefinido form and fulfilling aoristic functions in the so-called high relevance contexts. It stands to reason that this co-occurrence in ‘high focus contexts’ must have contributed considerably to the weakening of -ara’s profile as a fully-fledged pluperfect form. Apart from that, -ara had to withstand the increasing pressure of the compound past perfect (avía hecho), which advanced

152

MARTIN BECKER

permanently from the end of the 13th century (as can be seen in Davies’ Corpus del Español). However, another development can be grasped in the translation of Calila e Dimna, which should turn out to be even more important, – the expansion of the morpheme to modal readings. A retrospection: Until the 13th century, the apodosis of a counterfactual conditional clause represented the sole non realis domain in which -ara was able to occur. This particular use of the pluperfect was bequeathed by the Latin language where it was regarded as a strongly marked construction: -ara indicated the nearby realisation of an event, which, at the end, was doomed to failure due to the circumstances exposed in the protasis (Hofmann & Szantyr 1965, vol.2:328).4 The stylistic effect arose from the inversion of the syntactic order typically ascribed to a conditional clause: q was stated as being a real event and cancelled retroactively by indicating the blocking effect of p (q if not p), as in the following example: (11) praeclare viceramus nisi spoliatum, inermen, fugientem Lepidus recepisset Antonium. (Cicero, Epistulae ad Familiares, book 12, letter 10, 3) “We had practically won a splendid victory, had Lepidus not given shelter to Antonius who was attempting to flee and left without troops and arms.”

A similar pattern (p, if q) can be detected in Gonzalo de Berceo’s Milagros de Nuestra Señora, where p is anticipated as a result but cancelled ex post by the conditional clause. (12) Tornó a su posada durament engañado, mucho más li valiera si se fuesse quedado. (Berceo, Milagros Milagros de Nuestra Señora, v. 731) “Severely disillusioned, he returned to his house; it would have been much better, if he had stayed at home.”

However, in Berceo we also find the reverse order – the protasis followed by the apodosis in which the ‘anterior’-feature associated with -ara is even stressed by the use of an auxiliary: (13) si ante lo sopiessen lo que depués sopieron, no li ovieran fecho esso que li fizieron. (Berceo, Milagros de Nuestra Señora, v. 148) “If they had known before, what they knew after, they wouldn’t have done what they did.”

Calila e Dimna represents a turning point with respect to the Latin tradition as we encounter for the first time (as confirmed by the CORDE, the historic Corpus of the RAE) the morpheme -ara in the protasis. Still more important: the combination of -ara in both, the protasis and the apodosis of the conditional

4

For modal uses of the imperfect and the pluperfect in Latin see Pinkster 1990:229ss.

FROM TEMPORAL TO MODAL

153

clause should turn out to be the canonical model of counterfactuals until the Siglo de Oro (for a survey see Luquet 1988:252ss). (14) Si yo creyera los dichos de los filósofos de lo que dijeron del poco gradescimiento del homne, non llegara yo a esta tribulación. (Calila e Dimna, cap. XV) “If I had believed what the philosophers told about the meager gratefulness of human beings, I would not have arrived at such despair.”

The remarkable characteristic of this type of conditional clause is the exclusive past counterfactual interpretation of -ara, which prevails until the 15th century. This aspect is all the more important as it points to a conceptual relationship between the pluperfect and the past counterfactual, to be elucidated in the following chapter. 3. The pluperfect as the starting point for the development of irrealis readings Following Reichenbach’s temporal system, the pluperfect involves three different time points, the utterance time t0 of the speaker, a reference time t1, which is situated in the past and an anterior event time t2 at which an expression ϕ, for instance a proposition like λi[creer’ (i)(yo’, los dichos’)] (where λi abstracts over all possible time-world-pairs), holds true. Yet, the relationship of the proposition with regard to t1 is of particular interest. The default assumption concerning t1 is that the proposition does not hold any longer. This is for instance the interpretation for sentences like ‘When Peter came to the party we had already left’. However, in the case of atelic verbs like ‘rely’ as in ‘I had relied on my friends and told them the secret’, we simply ignore whether the speaker continued to rely on his friends from a certain point in the past or whether he lost his faith in them. So with respect to the reference point t1, we can say that the speaker no longer takes the responsibility for the truth of the proposition ϕ, so that its truth value becomes undetermined from that moment on. I claim that this indeterminacy, which is part of the pluperfect scheme is the basis for the modal readings of the pluperfect. An even more far reaching implicature from the pluperfect scheme can be derived in particular contexts, the conditional clauses being the most prominent one. Here, they yield a counterfactual reading based on the inference that the proposition ϕ turns into its contrary (non- ϕ) at the reference point t1. As we saw earlier, the use of the pluperfect in the protasis of counterfactual conditionals was a stilistic device to highlight the imminent, but ultimately frustrated realisation of q due to an intervening factor p. This intervening factor p coincides with the turning point t1, which renderes q ineffective, dividing the zone of evaluation into two subparts. In the left subpart, the realisation of q is still possible (or q is even coming into existence). Beyond the turning point t1, in the right subpart, the proposition turns into ¬q describing an irreversibly

154

MARTIN BECKER

counterfactual situation. This complex constellation can be sketched as in figure 1 (with ◊ representing the possibility operator): Intervening factor p which cancels q definitively (turning it into ¬q)

◊q (zone in which q is still possible) t t2

t1

t0

(turning point)

utterance time

¬q (zone in which ¬q holds) Figure 1: The interpretation of V-ara in a conditional context

The generalisation of the scheme si + cantase in the protasis and cantara in the apodosis must have been accompanied by a reanalysis of the whole structure. The protasis (which we indicate as p for convenience) was no longer considered a mere blocking factor of a fore grounded event q, but it appeared as the fulcrum of a whole system of non real worlds (= p-worlds). So, the protasis p profiles a domain of non real worlds (= indexes) with respect to which the apodosis q – in accordance with a quantifying operator: must, may, generally and so forth – hold(s) true. In conclusion, we can say that the background system of evaluation for the main clause had changed from Latin to Old Spanish passing from a real world reference (in which q was solely rendered ineffective by the blocking factor p cancelling the fore ground proposition) to an evaluation system of non real worlds with respect to which q had to be anchored. The intrusion of -ara in the protasis, which up to then had been restricted to both subjunctive forms, the eventualis future subjunctive -are (cantare) and the irrealis past subjunctive -ase (cantase), meant another decisive change. The reasons for this development are difficult to determine: as we do not find any traces of -ara in the protasis before the middle of the 13th century (with the first occurrences in Calila e Dimna (c.1251) and Sendebar (c.1253)), it is not very likely that the appearance of the pluperfect form in the protasis dates back to a Vulgar Latin tradition. Penny, on the other hand, believes in a general tendency in languages to favour identical coding in conditional clauses (Penny 2 2002:252). However, the typical constellations in Medieval Spanish conditional clauses, e.g. cantase/cantaría, cantare/canta, cantará, do not support the assumption of an analogical adjustment. Yet, the use of -ara in the protasis reveals that the irrealis reading of the pluperfect form has consolidated itself, at least, in a very local context. Until that moment, the past subjunctive form -ase alone fulfilled the function to profile the irrealis domain in which the apodosis q was to be evaluated.

FROM TEMPORAL TO MODAL

155

However, with its insertion into the protasis, -ara had to take over a similar function as -ase within a context that was now exlusively identified by the conjunction si (‘if’). This constellation implies a stronger division of labour: the conjunction ‘if’ functioned as a distance marker establishing the relevant system of non real worlds and -ara indicated within that system a maximal degree of distance from the actual world (the speaker’s world) w0. Apart from that, another crucial aspect contributed to the strengthening of -ara in conditional clauses. The dominant form of irrealis conditionals, cantase, was losing progressively, in the period between 1250 and 1325, its potential to mark a past counterfactual situation (Rojo & Montero Cartelle 1983:153). This must have been a decisive factor for the advance of -ara and it proves, at the same time, that -ara was first of all identified with its particular anterior feature. So, the form -ara included both, a modal potential, which was restricted to particular local contexts and an anterior feature, which implied a certain turning point at which things changed irreversibly to its contrary (in the case of conterfactual clauses: from possible to irreversibly counterfactual). The further developments are determined by two important aspects: in the context of conditionals, -ara appeared predominantly as an irrealis marker so that a modal use of -ara with an irrealis reading was firmly consolidating itself. Yet, there was an even more important move: the modal uses of -ara, which were first confined to the conditional clause had expanded to other related structures with the same underlying conceptual structure. As we already mentioned, conditional clauses can be characterised by an underlying tripartite structure, which yields the representation in figure 2 (see Kratzer 1978, 1991a, 1991b, Partee 1995): S

Quantifier MUST/GEN/ NORMALLY MAY/EXCEPT

Restrictor [IF p ]

(nuclear) focus q

Figure 2: Tripartite quantificational structure

The restrictor (for instance the protasis of a conditional clause) fixes the domain, which is relevant for the evaluation of the nuclear focus q (the consequent). The strength of the relationship, which holds between p and q (e.g. of the implicational relation) is indicated by an overt or (as is normally the case) covert operator, which quantifies over q. Take the clause: If Peter comes, he might be happy.

156

MARTIN BECKER

The domain of evaluation is restricted to worlds in which Peter comes (the p-worlds). For the conditional clause to be true, q must not be true in all pworlds profiled by the antecedent (as indicated by the operator might); it suffices that q is true in at least one of the profiled p-worlds.5 The spreading of -ara follows a remarkable path, which can be grasped in the 15th century treatise Cárcel de amor written by Diego de San Pedro. The morpheme -ara expands to a great variety of contexts, which can be traced back to the basic conditional scheme. It occurs in: (a) implicit conditional clauses in which the protasis has been reduced to a prepositional phrase (as in the first example) or has been completely suppressed as in the second. In the latter case the protasis might be reconstructed easily: (15) Por cierto con su marido, o su padre, o parientes non lo sufriera tal pasar, antes se degollara. (Arcipreste de Talavera o Corbacho, 215) “Of course, with her husband, or father, or family she would not suffer such an evil, and she would rather lay hands upon herself.”

In this example, the PP con su marido can be interpreted as an ellipictal construction for si estuviera con su marido. In the second example, a formula like si pudiera would be appropriate: (16) Por cierto con mejor voluntad caminara para la otra vida que para esta tierra. (Diego de San Pedro, Cárcel de amor, 79) “Of course, I would prefer to pass to the other life, rather than stay in this world.”

These types of contexts paved the way for the use of -ara as a main clause irrealis-conditional. In the last example a more or less concrete background protasis is still available, whereas for the more abstract use of -ara as a main clause irrealis, we have to presuppose a set of conditions (=propositions) which do not hold at the actual index i0, but have to be true at a distant index i* in order to make the apodosis come true (Lohnstein 2000:95ss). In other words, this set of propositions bridges the gulf between the real world w0 and a fictitious world w* by indicating the necessary conditions for p to be true or, to put it a different way, they determine the distant index i* at which p has to be evaluated. For a sentence like ‘I would help you’ we can imagine a set of conditions composed of propositions like “I have time, I am willing, I have the necessary resources”, and so on.

5 For a more fine grained analysis of counterfactual conditional clauses see Kratzer 1978:207ss. and Giorgi & Pianesi 1997:256ss. A cognitive-semantic account is given by Werth 1997:243ss.

FROM TEMPORAL TO MODAL

157

(b) Another characteristic mode of use can be couched in the term illocutionary variation: so, rhetorical questions or exclamations can be derived from elliptical or implicit conditional clauses: (17) ¡Ay! ¿Cómo pudiera yo escusar esto que hoy gané, e esta tristeza en que só entrado? (Calila e Dimna, cap. VI) “Oh ! How could I have avoided what I have gained today and this sadness which has taken hold of me ?”

(c) Another class of contexts consists of comparative clauses introduced by como si: although counterfactual in w0 (¬p in w0), the proposition p serves as a term of comparison, which characterises a state of affairs (represented by q) in w0: q behaves like p, with p not belonging to w0: (18) (...) tenía tal alteración y tan sin aliento la habla como si esperara sentencia de muerte. (Diego de San Pedro, Cárcel de amor, 27) “He was so excited and talked to her out of breath as if he expected the death sentence.”

(d) Indeterminate relative clauses without an antecendent (introduced by quien, los que, quienquiera) can be analysed as free relatives displaying a conditional reading (cf. Quer 1998). There is a slight difference between free relatives and conditionals: the quantificational operator of free relatives only quantifies over individual variables, whereas in conditional clauses it unselectively binds situation and individual variables (Quer 1998:191). An example borrowed from Diego de San Pedro’s Cárcel de amor can be described in the following way: (19) Quienquiera que la oyera pudiera conocer que aquel estudio avié usado poco. (Diego de San Pedro, Cárcel de amor, 27) “Whoever would have listened to her could have understood that she knew little about this art.”

∀x: [persona (x) ∧ oir’(x, ELLA) -> ◊ (conocer’(x, p))] This structure is similiar to the conditional scheme: PODER (◊) [oir’ (x, in s)] Quantifier (may-OP) restriction

conocer’(x, p, in s) nuclear focus

(e) Concessive conditionals, which can appear in three different forms: - as parametric concessives (por muy/más que), which imply a scale representing different degrees of a certain dimension, - as disjunctive concessives (fuera x o fuera y) and

158

MARTIN BECKER

-

as polar concessives, which include a focus particle, for example aun que, aun si, aunque. (For a typology see Haspelmath & König 1998:562ss).

We quote an example for a polar concessive conditional introduced by aunque: (20) que aunque fuera perdidoso del mayor bien desta vida, (…), fuera bienaventurado en no oír ni saber lo que padeces. (Diego de San Pedro, Cárcel de amor, 41) “Even if I had lost the most precious thing in this life, (…), I would (have) be(en) lucky not to discover what you suffer.”

The underlying conditional structure can be represented as: MUST ( ) [ser_perdidoso’ (YO, in s)] ser_aventurado’ (YO, in s) Quantifier (must-OP) restriction nuclear focus Polar concessive conditionals focus on the most adverse cases (worlds), which could affect the realisation of the matrix clause and presuppose that the restriction would normally cancel the consequent, which means: NORMALLY [if p], ¬q. The function of the concessive conditional thus is to state the general validity of q for all contexts (be they normal or exceptional). This is the reason for this type of conditionals being labelled “irrelevance conditionals” (König 1986:635ss.). Another important area for the consolidation of -ara’s irrealis reading is constituted by modal verbs like deber and poder, as well as the desiderative verb querer. Modal verbs are lexical expressions for operators (the necessity and the possibility operator) and they are interpreted on a relevant – epistemic, deontic, dispositional or aletic – modal base. The desiderative querer implies an ideal, a so-called ordering source, which can be described as a preference function that ranks worlds according to their degree of compliance with the ideal (in an ideal wish-world all the wishes of an individual are fulfilled) (Farkas 1985:78, 1992a:86, 1992b:35ss). Modal verbs and querer create non veridical contexts, which means that deber/poder/querer p does not entail p – p being evaluated in a model associated with an individual (Giannakidou’s relativized veridicality). (Giannakidou 1997, 1998:105-115, 1999:367ss.). When combined with the -aramorpheme, these verbs yield equivalent irrealis readings, irrespective of whether a temporal or a modal analysis is given to the morpheme. In a temporal reading, it is understood that the time interval of the deontic situation had already elapsed, while in a modal reading, the deontic worlds had become inaccessible. In any case, the proposition p in the scope of quisiera, debiera and pudiera characterises a state of affairs, which became unrealisable because a point of no return (our famous t1) had been reached.

FROM TEMPORAL TO MODAL

159

To sum up, we can say that in the first stage of developments -ara conquered an important part of the irrealis domain when spreading from the conditional clause to a variety of related structures (implicit conditional clauses, free relatives with a free choice reading, conditional concessives, etc.). Additionally, the morpheme strengthened its irrealis profile when combined with non veridical operators such as the modal verbs querer and poder as well as the desiderative querer. We will now have a look at the developments in Portuguese. 4. The developments in Portuguese This first stage of expansion, comprising the generalisation of -ara’s modal use to other related structures based on the underlying tripartite conditional scheme, has a perfect counterpart in the evolution of the Portuguese language after 1300. An early example, which bears testimony to the intrusion of -ara in the protasis can be found in the 14th century translation of the Royal Legal Codex (Primeyra Partida, first part of the Siete Partidas), which had been published a century before in Castile. (21) (…) dizendolhy q(ue) sse elle hy fora nom morrera seu jrmaao. (Afonso X: Primeyra Partida, Título Lei Fólio 6 4 39b) “(…) telling him that, had he been here, his brother would not have died.” (Cp. John 11.21)

This use of -ara progressively consolidates during the 15th and 16th centuries, as can be seen in official texts such as the Chronicle of the Reis de Bisnaga, and – what is even more important – in popular texts like Gil Vicente’s (1465-1536) plays, which are destined for a large audience. The Auto da India is one example, which reflects the spoken language of this period: (22) AMA Porém vindes vós muito rico? MARIDO Se não fora o capitão, Eu trouxera o meu quinhão. (Gil Vicente, Auto da India, 499-502) “HOUSEWIFE: Do you return as a rich man ? HUSBAND: If the captain had not been with us, I would have made a fortune”.

Taking into account texts from Gil Vicente onwards, we are able to document the process of -ara’s expansion to related conceptual structures: (a) Ellipsis of the protasis and generalisation as a conditional form in the matrix clauses. (23) ANJO Pois cant’eu mui fora estou / de te levar para lá. Essa barca que lá está vai pera quem te enganou. ONZENEIRO Porquê? ANJO Porque esse bolsão / tomara todo o navio. ONZENEIRO Juro a Deus que vai vazio! (Gil Vicente, Barca do Inferno, 210-217)

160

MARTIN BECKER

“ANGEL: I cannot take you with me. Take the boat over there of the person who betrayed you. USURER: Why? ANGEL: because this (your) bag would occupy the whole boat. USURER: I swear by God that it is empty !”

(b) Indeterminate (free) relatives with a conditional reading: Quem pudera (≈ se alguém pudera ver): (24) Quem uma só pudera ver, / vira claramente nela / aquela dor, / aquela pena e padecer, com que choráveis, donzela, / vosso amor. (Gil Vicente, Auto da Alma, 630-641) “Whoever could have seen/see only one tear, would (have) clearly see(n) this pain, this sorrow and suffering with which you mourned your love, my lady.”

in formal representation: ∀x: [persona’ (x) ∧ ver_uma_lágrima’ (x) -> ver_aquela_dor’ (x)] MUST( )[persona’(x in s)∧ver_lágrima’(x in s)][ver_aquela_dor’(x in s )] (c) Illocutionary variation: hyperbolic formula of exclamation (25) Enchi em curta idade e instantes poucos / longa vida de amor, mais longa em penas. / Quem soubera dos túmulos o arcano! (Castilho, Canto IV, p. 37) “In a short lifetime and a few moments, I have lived a long story of love, but even a longer story of pain. Would somebody know the mystery of the tombs !”

(d) The sermons of Jesuit Priest António Viera (1608-1697) include examples showing the use of -ara in comparative clauses introduced by como se: (26) (...) escutando com silêncio, e com sinais de admiração e assenso como se tiveram entendimento o que não entendiam. (António Viera, Sermão aos peixes, II, p. 5) “(…) listening in silence, with signs of admiration and consent, as if they understood what they did/could not hear.”

(e) Evidence for -ara in concessive conditionals is late (but given the limited extend of our Portuguese Corpus we should be cautious with overtly strong generalisations). The following passage taken from the romantic author Herculano includes a parametric concessive introduced by por muito que: (27) (…) de que todo o esforço dos soldados que coroavam os adarves do mosteiro, por muito que houvera sido, não fora bastante para salvar os desgraçados (...) (Alexandre Herculano, Eurico o Presbítero, 73) “(…) that the maximum effort made by the soldiers who occupied the defences of the monastery, however much it would have been, it would not have been sufficient to save the miserable (…).”

FROM TEMPORAL TO MODAL

161

Apart from these extensions of modal -ara uses to a number of related structures, we can also attest to the irrealis reading in combination with modals and the verb querer. We adduce the following example, which expresses a frustrated wish: (28) (...) e pos por terra a fortalleza, e asy outros muytos lugares, d’aquy quisera hir adiante, e na~o no comsemtira~o os do comselho (...). (Crónica dos Reis de Bisnaga, 16th century) “(…) and he dismantled the fortification and, in the same way, many other places; from here he would have liked to advance, but the counsellors didn’t give their consent (…).”

As we see, even in Portuguese -ara was able to cross the irrealis-domain encroaching upon typical conditional and modal contexts normally reserved for the subjunctive -asse (in subordinated clauses) and the form -aria (as the irrealis marker in main clauses). 5. The turning-point in Spanish: from condition-based irreality to generalised counterfactuality The common developments in Spanish and Portuguese we have attested thus far, might be interpreted as a maximal exploitation of -ara’s modal potential in a particular class of contexts (conditional and conditional-like contexts). However, in Spanish the dynamics of change pushes things even further: from Calila e Dimna on, we are able to witness the intrusion of -ara in different types of clauses, which have nothing in common with conditionals and which force the conclusion that a process of reanalysis must have taken place. The first deviant context, which comes up in our corpus is the occurrence of -ara in a prospective temporal clause introduced by fasta que (hasta que). In standard past contexts this kind of temporal clause is associated with a simple past form as in the following sentence: (29) Et fuese Berzebuey por su mandado, et andudo tanto fasta que llegó a tierra de India. (Calila e Dimna, cap. I) “And Berzebuey, following his order, set off and went until he arrived at the territory of India.”

However, in the same text we also discover for the first time: (30) Dijo el rey: ‘Debieras te sofrir fasta que amansara mi ira.’ (Calila e Dimna, cap. IX) “And the king said: ‘You should have endured until my anger (had) subsided.’”

The subordinate clause is in the scope of the modal operator (itself marked by -ara) and it remains temporally indeterminate: this boils down to the fact that the effective index or the set of indexes at which the proposition p (λi (amansar (i)(tu)(mi_ira)) is true (or in other words, the predicated relationship holds) is not specified. And one might even be tempted to interpret the

162

MARTIN BECKER

temporal clause as an abstract background condition for the modalised matrix clause (something like: for the – unspecified – set of indexes at which I did not calm down: you should have endured). Consequently, in this early temporal example the -ara-clause is associated with a modalised matrix frame and conveys the idea of indeterminacy. Another interesting context of change is created by epistemic and speech act verbs which, as we mentioned above, had been typical realis contexts compatible with -ara in the medieval texts. However, in the following quotation – once again from Calila e Dimna – the verbal form -ara selected by an epistemic verb (cuidó < cogitare) indicates an expected event whose indexical anchoring remains totally undetermined. This constellation bears resemblance to the temporal case of hasta que because we do not know at which index p turns out to be true: (31) Et vio el cazador lo que ficieron, e siguiólas por las haber, e non se desfució dellas e cuidó que luego a poca de ora les apesgara la red e cayeran. (Calila e Dimna, cap. V) “And the hunter saw what they were doing; he followed them in order to catch them and he did not lose his hope; he thought that after a while the net would be very heavy and they would fall.”

More than a century later, in Diego de San Pedro’s Cárcel de amor, the same prospective and virtualising reading of -ara re-emerges. Once again, the truth value of the complement clause is suspended until further notice, p being true at some future index, if ever: (32) (...) y también pensé que para ello me ayudaran virtud y conpasión y piedad, porque son acetas a tu condición, (...). (Diego de San Pedro, Cárcel de amor, 61) “and I also thought that for this affair virtue, compassion and piety would (have) help(ed) me because they are in line with your character, (…).”

However, as the process of reanalysis moves on, the morpheme -ara also shows up in contexts presupposing the falsehood of the complement clause or being earmarked explicitly by a negative operator. The earliest example can be witnessed in the famous Calila e Dimna, the -ara containing complement clause is selected by an expression, which unveils a false appearance (fazer muestra que p), thus presupposing the falsehood of p: (33) Et fizo muestra la fembra que le pesara, por toller de sí la mala sospecha de su marido. (Calila e Dimna, cap. XVII) “and the woman pretended to be upset in order to assuage her husband’s suspicion.”

A second example to be quoted, is taken from the Libro de Palacio composed by López de Ayala: the complement clause reports a lie as indicated by the speaker:

FROM TEMPORAL TO MODAL

163

(34) (…) le así vendieron e después a su padre con malicia mintieron: que bestia fiera l' matara falsamente dijeron. (López de Ayala, Libro Rimado de Palacio, v. 386-388) “(…) and they sold him in this way and then they lied maliciously to their father that a wild beast had killed him; they told contrary to the facts.”

In the last example we want to give in this context, the complement clause of parecer unveils a false appearance – as in the case of como si – q bears – superficially – ressemblance to another p, which does not hold in the current situation (p = 0 in w0): (35) (...) que vesiblemente le pareció que le quebrantara la cabeça e le saltara fuego de los ojos, en tanto que del todo la vista perdió e parecióle el mundo todo ser estrellado. (Arcipreste de Talavera o Corbacho, 290) “(…) and it seemed clearly to him that his head would crack and that fire would come out of his eyes.”

Apart from these presuppositional readings, we become aware of the fact that -ara strongly correlates with an explicit negation operator, which marks the proposition p as not holding at some index i (which means that p is counterfactual at i). Take, for instance, the negative polarity item nunca, which excludes the truth of p for all time indexes of a given interval – such that p is marked as counterfactual for every i within that interval: (∀t): λt [¬(pensar (t), (yo), (p) )] , t ∈ T (T being an interval of time indexes) (36) (…) que nunca pensé que así me penara (…). (Nicolás Núñez, Tratado, 101) “(…) that I never thought that it would hurt me so much (…).”

However, the negation operator may also apply to the descriptive content of a relative clause or to a causal clause: in these cases it is stated that p does not hold at a certain index t – because the explanation is not valid (non porque p) or the referent in this case simply does not exist (there is no x such that p). Compare the following two examples: a) non porque p, p = 0 at the evaluation time: (37) Ya oístes como Job, aquel santo varón, muy amigo de Dios, sufrió tribulación, non porque él pecara, mas por que el pregón de la su pacïencia le fuese gualardón. (López de Ayala, Libro Rimado de Palacio, v. 3995-3999) “You have heard what kind of suffering this holy boy, a very good friend of God, had to endure, not because he might have sinned, but in order to be rewarded with the reputation of his proverbial patience.”

b) non x que p: ¬ [(∃ x) ∧ P(x)], P being a predicate (‘descriptive content’) characterising x:

164

MARTIN BECKER

(38) Quien la vido poco tiempo avía e después la vido en tierra vencida e medio muerta, non siento persona tan cruel que de los ojos non llorara (...). (Arcipreste de Talavera o Corbacho, 290) “If someone saw her a short time before and, then, a little afterwards, totally defeated and half-dead – I do not image a person who would be so cruel as not to cry large tears (…).”

As our analysis reveals, the second stage of changes is determined by an expansion of -ara beyond the original conditional or restriction-marking contexts. The morpheme -ara now spreads to a wide range of non veridical contexts, which basically amount to two different classes: in one class of contexts -ara indicates the indeterminacy of the truth value ascribed to p given a virtualising prospective context (basically, temporal clauses and complement clauses selected by epistemic and speech act predicates). In some of these contexts -ara even seems to have lost its characteristic perfective feature, while in other contexts the distinction between perfective and non perfective readings turns out to be irrelevant (as shown by some of our translations). In the second class of contexts -ara appears in the scope of an implicit or explicit negation operator, which marks the sentence (and the underlying proposition p) as not holding in w0 and ascribes p to the reign of counterfactuality. Consequently, the outcome of these evolutions in the second stage is a strengthening and generalising of -ara’s irrealis uses, which goes along with a weakening of the – up until then – inherent perfective feature of the morpheme in virtualising contexts. So until the 14th century the (past) counterfactual reading of -ara must have become deeply entrenched in the functional spectrum of -ara. This fact is confirmed by Ayala’s subtle arrangement of a conspicuous literary effect, which relies on the new functional load of the morpheme. In the Libro de Palacio a long realis sequence reporting facts located at a past time interval is juxtaposed without any explicit marking (!) with a set of counterfactual statements, which describe the absolute reverse: the “antiuniverse”, as Robert Martin (Martin 1983:38) labelled it, of what had really happened. This passage can be understood per se – as due to the counterfactual marking function of -ara: (39) Si yo vi pobre muerto, d'él muy poco curé / de le dar sepoltura, mas los ojos cerré (…). Non hobe pïedat del que vi en prisión, (…). Con mi palabra sola pudiera yo acorrer / algunt cuitado preso: non lo quise fazer (…). Con valor de mis paños a mil pobres vestiera, e grant bien e grant pro de mi alma fiziera; (López de Ayala, Libro Rimado de Palacio, v. 541-561) “When I saw a human being who had deceased I did little to arrange a funeral for him, but I closed my eyes (…). I did not have pity with whom I saw imprisoned (…). With my speech I alone I could have helped any grief-stricken prisoner: (but) I did not want to do so (…). With the money spent on my clothing I would have dressed one thousand poor, and I would have done something good to my soul.”

FROM TEMPORAL TO MODAL

165

However, this state of affairs does not call into question the vitality of -ara’s pluperfect function in relative clauses and its persistence as a diffuse past or aorist form in some temporal contexts in which it co-occurs as a variant of the Indefinido: (40) E por ende se escribe que Dios cató a Abel e catara sus dones, (…). (López de Ayala, Libro Rimado de Palacio, v. 5414-5417) “Therefore it is written that God saw Abel and he saw his gifts (…).” (41) Él las (= las cosas) estableció, por que el pobre homne que Él formara e crio allende non pasase de lo que Él mandó. (López de Ayala, Libro Rimado de Palacio, v. 6423-6426) “He made the things the way they are in order to prevent the miserable man he created and brought up from violating the rules he ordered.”

In the end of this section we would like to comment on some further developments in the Portuguese counterpart. How did -ara evolve in the sister language? There can be no doubt that some spill overs of -ara into the domain of weak intensionals must have taken place. This expectation is borne out, for instance, in a text of the 18th century author António José da Silva who employs -ara in a virtualising context linked to the matrix verb dezir. At the time t1 it is not clear when p will hold: (42) D. FUAS – Para este quintal ou jardim ou o que for me disse Fagundes viera D. Nise a regar a sua manjerona; mas, enquanto ela não vem, me esconderei atrás deste canteiro (...). (António José da Silva, Guerras do Alecrim e Manjerona, 104) “D. Fuas – Fagundes told me that D. Nise would come to this courtyard or garden, or whatever it is, in order to water her marjoram plant; but as long as she doesn’t come, I will hide behind this vegetable bed (…).”

However, in Portuguese the developments are not pushed beyond this line: in the 19th century the irrealis uses of -ara, which remain restricted to conditional contexts are increasingly considered to be stylistic trappings of a literary or very formal official discourse. In the following poem composed by Alexandre Herculano the author gives a maximally solemn diction to the Promethean hero: (43) Eu rodeara, então, o globo inteiro; / Eu sublevara as águas; / Eu dos volcões com raios acendera / Amortecidas fráguas. (Alexandre Herculano, A Harpa do Crente, 56) “I would go around the entire world and stir the waters, I would reignite the ebbing glow of the volcanos.”

On the other hand, some uses of -ara as in tomara que (with the meaning “I would like to”), quisera (“I would like”) and dissera (“I would say”) survive as frozen stereotypical formulas with a particular pragmatic function (e.g. the

166

MARTIN BECKER

modalisation of an assertion in the case of dissera). (See Hundertmark-Santos Martins 1982:192). These uses can be witnessed in the following examples: (44) FAGUNDES – Tomara achar um homem que ma carregasse. (António José da Silva, Guerras do Alecrim e Manjerona, 92) “FAGUNDES – I would like to meet a man who would carry them for me.” (45) D. FUAS – Eu dissera que o óleo de manjerona era excelente remédio. (António José da Silva, Guerras do Alecrim e Manjerona, 92) “D. FUAS – I would say that the oil of the marjoram was an excellent remedy.”

The loss of vitality of -ara’s modal uses does not only exhibit itself in the attrition of the modal uses in the form of lexico-grammatical stereotypes, but also in the vanishing of -ara in late 19th century standard prose. For instance, in Eça de Queirós’ novel Os Maias, our morpheme has not left a single trace in a conditional context. As a counterpart, -ara reaffirms itself in its inherited temporal readings and even functions as a narrative past in the voluminous historical novels of the 19th century, for instance in Herculano’s novel Eurico o Presbítero: (46) O vulto cravara nela um olhar ardente, que a fascinava. Sorriso doloroso lhe pousava nos lábios. Estendeu o braço, segurando a mão de Hermengarda, (…). (Alexandre Herculano, Eurico o Presbítero, 148) “The figure (had) transfixed her with a passionate glance which fascinated her. A sorrowful smile was on his lips. He stretched out his arm, holding Hermengarda’s hand, (…).”

Today’s Grammars on Portuguese like Bechara’s Moderna Gramática Portuguesa (from 2004) relegates the modal use of -ara to an extremely marked linguistic register labelling these instances as “um traço estilístico de linguagem solene” (“a stylistic feature of solemn discourse”) (Bechara 2004:279). To summarise the results, we have to emphasise the fact that -ara in Portuguese, though overstepping the border of the irrealis domain, was unable in the long run to go beyond the delimited area of a restriction-based scheme. Now and then we witness some incursions into the domain of virtuality, but they were tentative in nature and never systematic. For this reason, -ara was not only blocked from evolving into a general marker of non veridicality, but also receded during the 19th century from conditional contexts, converting the use of -ara in irrealis contexts into a highly marked stylistic device in the realm of pompous rhetorics. The temporal reading of -ara was preserved but found itself, through the course of the 20th century, to be relegated to formal written registers. In Brazilian Portuguese the prospects of -ara seem to be even less promising.

FROM TEMPORAL TO MODAL

167

6. The third stage: from counterfactual (irrealis) to subjunctive II In the previous chapter we illustrated the increasing affinity of -ara with negation and counterfactual contexts in general. In virtualising contexts, produced for instance by epistemic verbs, the morpheme also yielded an indeterminate-prospective reading. This decisive development from the second to the third stage is marked by a progressive encroachment upon the complement clauses of strong intensional verbs – represented for example, by commissives and desideratives like mandar or desear. As afore-mentioned, strong intensional verbs create non veridical contexts and they imply both, a set of possible worlds and an ordering-source, which ranks possible worlds according to an ideal or a norm.6 The introduction and, above all, the consolidation of -ara in this class of contexts took a very long time, even centuries, given that -ara, apart from its persisting but shrinking temporal functions, was considered an irrealis form. A very early – and quite exceptional – example for the selection of -ara by a strong intensional verb can be found again in the translated book of fables Calila e Dimna: (47) Et tú quesisteme probar et temiste de muerte, si lo descubrieses, et non mandara Dios que yo así lo feciera, que me has fecho grant servicio et soy tenudo de te lo galardonar. (Calila e Dimna, cap. XI) “And you wanted to test me and you were terribly frightened that you might discover something, and God had not ordered that I should do so/that I had done so.”

In the overwhelming number of cases in which the morpheme appears within a strong intensional context, the intensional operator is itself modalised, either by the negation operator or by an irrealis form. The just-quoted example, for instance, has the logical form: λi: ¬ [mandar’ (i) (Dios’, p)] (i*) (i* being the index at which p holds). An early example of -ara in a volitional context can be seen in Ayala’s Libro de Palacio: again the matrix verb itself has a pluperfect reading and is in the scope of a non veridical operator (Neg). Interestingly enough, a few lines later we find a parallel example, which still abides by the traditional norm of grammar (we quote both examples):

6 The main difference between weak and strong intensional verbs can be described by an entailment relation. Weak intensional verbs like pensar and creer imply the truth of the complement clause at least in the epistemic model (= the reality-model) of the main clause subject (‘relativised veridicality’). Strong intensional verbs like desideratives (querer) or commissives (ordenar) do not entail the truth of their complement clause. In some possible worlds (in which things evolve according to the subject’s preferences) the complement proposition may come true, in other possible worlds the proposition turns out to be wrong.

168

MARTIN BECKER

(48) Maguera qu' el lloraba el su fijo amado e non quisiera él que fuera justiciado (…). (Ayala, Libro de Palacio, 6499-6503) “Although he cried for his beloved son and did not want his son to be executed (…).” (49) E non quisiera Dios que un santo varón, atal como era Job, en desesperación cayese. (Ayala, Libro de Palacio, 6527-6529) “And God did not want a holy boy like Job to fall into despair.”

As we saw in the last section, the morpheme -ara evolved into a marker of irrealis modality. An example taken from the 15th century author Diego de San Pedro contains two -ara forms in strong intensional contexts: fuera ordenado, which is selected by modalised quisiera, and ocuparas, which depends on the final conjunction por que. In both cases the selecting element itself is in the scope of a modalising operator (IRREAL and NEG): (50) Pero quisiera que lo que tu avías de ver fuera ordenado, porque no ocuparas tu saber en cosa tan fuera de tu condición. (Diego de San Pedro, Cárcel de Amor, 26) “But I would like to elucidate things you were obliged to see, lest you should waste your knowledge on questions below your level.”

A further advance of -ara in strong intensional contexts becomes apparent in Columbus’ description of his four expeditions (Los cuatro viajes del almirante y su testamento) published only a few, but decisive, decades later. In his book -ara appears for the first time as a variant of -ase in strong intensional contexts and, what is even more striking, the matrix verb no longer appears in the scope of a non veridical – negative or irrealis-marking – operator. In the following example, -ara and -ase alternate freely in a commissive context: (51) (…) y dado orden que llevasen aquella vuelta fasta medio día, y después que ambas se mudasen las derrotas y se recogieran para mí. (Colón, Los cuatro viajes, 41) “(…) and given the order that they should make this trip by noon, and that afterwards, they should change their course and return to me.”

In Colon’s report, some other strong intensional contexts can be found, which are established by desiderative (desear) and other commissive verbs (rogar), both introducing an ordering source, which discriminates future possible worlds in accordance with a principle or norm. The logical structure of strong intensional verbs requires that the morphologic category it combines with must be able to locate the proposition at possible indexes – at those indexes which make the proposition, the wish, the order, the desire, and so forth, true. Obviously, as the quoted examples prove, the morpheme -ara is considered to respond to this requirement, which boils down to the fact that -ara must have lost its perfective feature by the end of the 15th century. The following examples illustrate this:

FROM TEMPORAL TO MODAL

169

(52) Deseaba que aquello vieran muchas otras personas prudentes y de crédito (…). (Colón, Los cuatro viajes, 73) “He wished that many other intelligent and trustworthy persons should see this (…).” (53) Y bien es verdad que mucha gente de esta que va aquí me habían rogado y hecho rogar que les quisiera dar licencia para quedarse. (Colón, Los cuatro viajes, 112) “And it is true that many people who are around here asked me and made others ask me to give them the permission to stay.”

However, a second fundamental change -ara underwent at the end of the 15th century has to be mentioned: the morpheme lost its pluperfect reading and disappeared from virtually every temporal context. Antonio de Nebrija remembers his own lifetime, in his comment on the fifth chapter of Introductiones Latinae, as the decisive borderline between the old temporal and the new purely modal function of the morpheme.7 We do not intend to elucidate the further developments of -ara, this would take us too far afield (see e.g. Lapesa 92001, § 97.5). We only add to the picture that -ara preserves, in the context of conditionals, the function of distinguishing counterfactual from hypothetical conditionals (compare examples in Lazarillo de Tormes).8 However, in the 17th century this opposition between -ara and -ase is suspended, and at the same time the distinction made between cantara and cantaría in the main clause is given up (see examples and the glosses in Luquet 1988:266 and 279ss.). With these changes -ara turns into a marker of virtual9 and hypothetical contexts, competing with -ase and -aría in their respective domain. To round off the picture, we should mention that -ara occurs regularly in relative clauses of contemporary periodistic prose. There is an ongoing debate10 as to its modal status – is it a re-emergence of the traditional indicative or an 7 “Digo más que el castellano ni tiene el tiempo passado más que acabado, ni el venidero del indicativo. Delos quales el primero dize por rodeo del tiempo passado no acabado del indicativo deste verbo e as ove, que es avia, i del nombre participial infinito, como diziendo quando tu me amaste ia te avia io amado, por lo que dezian hasta aqui io amara i oviera amado.” (Antonio de Nebrija 1495: Glosa al cap. V de Introductiones latinae quoted in Luquet 1988:247). 8 Compare the difference between a possible, but improbable conditional clause and a counterfactual one: (a) Ya, cuando asienta un hombre con un señor de título, todavía por lacería. Pues, por ventura ¿no hay en mí habilidad para servir y contentar a éstos ? Por Dios, si con él topase, muy gran su privado pienso que fuese, y que mil servicios le hiciese, porque yo sabría mentille tan bien como otro y agradalle a las mil maravillas. (Lazarillo de Tormes, 84). (b) Pues estado en esta afligida y hambrienta persecución, un día, no sé por cuál dicha o ventura, en el pobre poder de mi amo entró un real, con el cual él vino a casa tan ufano como si tuviera el tesoro de Venecia, (…). (Lazarillo de Tormes, 79). 9 A 16th century example for the weak intensionals + -ara is the following: “Creo yo que el pecador alfámar pagara por todos.” (Lazarillo de Tormes, 88). 10 For an overview see Veiga 1996:87ss.

170

MARTIN BECKER

instance of the subjunctive? For reasons of space, we cannot enter the debate in this article.11 In the last section, we will turn to the question of how we can explain the developments we have described on a more general level and what kind of lessons we can draw from the particular case of -ara in Spanish and Portuguese. 7. Interpretation and conclusions The history of the verbal morpheme ara shows that the functional change from a pluperfect to a past subjunctive form was brought about by a long-term sequence of small conceptual shifts. By the same token, the path of development from indicative to subjunctive unveiled the intermediate and intertwined conceptual domains which unfold between the two opposite poles – the purely temporal past and the prototypical past subjunctive. As we saw earlier, the conceptual shifts are, by no means, random, but were based on related cognitive schemata underlying the grammatical constructions. At the beginning the cancelling effect of the pluperfect (p becoming ineffective at some past time index t1) was exploited as a very expressive and stylistically marked device to focus on a frustrated action, which was blocked by an intervening factor q. This constellation was generalised to express all situations in which q was prevented by p from coming into existence. During the first stage, the use of -ara in counterfactual conditionals was spread to all related conditional-like structures in which p functions as a domain restrictor for the quantified predication: quantifier [if p] q. The most important outcome of this first stage was the consolidation of the past irrealis reading in main clauses, which must have been a generalisation of implicit conditionals (with a set of conditions unexpressed but inferable by contextual and/or encyclopaedic information). Later on, the irrealis-marking function of -ara was exploited in the context of complement and adverbial clauses, which had originally been common contexts for a pluperfect reading, but which were now placed under the scope of an irrealis operator – such as NEG or a modalised matrix verb (IRR). In these irrealis contexts -ara must have lost progressively its perfective feature (in the course of the 14th and 15th century) and was then able to occur even in prospective contexts (but dependent on a matrix verb in the past). This ‘virtualising’ of -ara was the prerequisite for its 11

I am grateful to an anonymous reviewer who drew my attention to this tricky aspect, which requires a careful in-depth analysis of the modal status of -ara in the relevant contexts. Linguists like Veiga (1996:95-97, but also see Hermerén 1992) claim that -ara has to be interpreted as a traditional indicative in relative clauses, which can be traced back to literary works of the first half of the 19th century. Quer (1998:157-162), on the other hand, interprets the form as a kind of ‘evidential’ subjunctive, which points to a particular model of evaluation different from the speaker’s. Both interpretations might be right synchronically (there are instances of both uses in contemporary Spanish) or diachronically (the second reading can be interpreted as the reanalysis of the first). But cp. Lunn (1995:438-440) for a pragmatic account.

FROM TEMPORAL TO MODAL

171

further expansion into a domain strictly reserved to -ase, the so-called strong intensional contexts. However, also in this domain -ara intruded as an irrealismarker given that the first occurrences in strong intensional contexts were only admissible because the matrix predicates themselves were modalised by non veridical operators. It was not until the early 16th century (in Columbus’ board books) that -ara should appear as a variant of -ase in strong intensional contexts and thereby in the core domain of the subjunctive. Portuguese did not keep pace with the Spanish developments. We have been able to trace some virtualising efforts, but they never lost their deviant and thus marginal status. What is even more crucial, -ara evolving into the conditional domain became progressively relegated to the prototype of conditional predication, the explicit counterfactual clause. However, even in this function it was ostracised from everyday language and elevated to the remotest ranks of ceremonious, solemn and grandiloquent style. We now have to turn to the basic question: what are the lessons to draw with regard to this divergent development of Spanish and Portuguese ? More generally, what does this case study reveal about the ‘paradox of change’, the leitmotif of this volume? As we could observe, the parallel developments of Spanish and Portuguese were motivated by the resemblance of underlying conceptualisations (implicit conditionals, free relatives, concessive conditionals, etc.), which were all related to a basic scheme of domain restriction whose prototype was the canonical conditional clause. These and the later developments turned out not to be necessary, however, because apart from cognitive-semantic motivation, a second aspect had to come into play: language use and its principles. In the particular context of functional change, the focus has to be put on the phenomenon of reanalysis, which has been cogently described as a listenerbased strategy (Croft 2000:118-121, Detges & Waltereit 2002:163ss). In this article we do not intend to enter into the thorny debate about whether the ideal listener who brings about language change is the language-learning child or not.12 The essential aspect for the interpretation of our results is the abductive nature of the general reasoning mechanism, which underlies reanalysis (like other procedures in the domain of folk logic): people (listeners) abduct “the case” – the instantiation of a general rule – starting from a particular, empirically accessible result by invoking a general principle. In our case, language users had to ascribe a systematic status to the morpheme -ara at different stages of its development. We can reconstruct the path of -ara form the original pluperfect to the final subjunctive II reading as a chain of reanalyses, which correspond to a series of small conceptual shifts: 12 For some cornerstones of the ongoing debate: Andersen 1973:767, Roberts 1993:299ss., Haspelmath 1998:317, Croft 2000:44ss., Detges & Waltereit 2002:169ss.

172

MARTIN BECKER

In the first stage, -ara must have been interpreted as a modal extension of temporal -ara in the exclusive domain of counterfactual conditional clauses. So the modal potential of the form, which arose from the characteristic properties of its underlying conceptual scheme (the anteriority feature and the turning point to indeterminacy) only came to bear in an explicitly marked conditional context. However, when the use of -ara was generalised to cognate structures, i.e. other restriction-based constructions like indeterminate relatives and conditional concessives, the link between the irrealis reading of -ara and its particular context of use was becoming looser. The use of -ara in independent clauses such as ‘Te ayudara en este asunto, pero no pude’, which presupposes some set of unspecified, but principally inferable conditions, must have strengthened the autonomy of the morpheme vis-à-vis marked contexts. This generalisation of condition-based (or at least condition implying) uses of -ara led to its reinterpretation as a marker of counterfactuality describing state of affairs contrary to the outcomes of the real world. Thus, the reading of -ara was no longer ‘accommodated’ to the linguistic context marked by the explicit operator SI, but the form itself was seen as a concomitant (i.e. a marker of grammatical well-formedness) of counterfactuality. The gradual autonomy of -ara with respect to particular contexts was the prerequisite for a further expansion of the morpheme into new irrealis contexts. It was shown that the most prominent contexts of expansion were complement clauses of weak intensional verbs (mostly introduced by epistemic verbs) as well as temporal and causal adjunct clauses. However, as we have seen, these constructions were themselves enshrined in an irrealis context, which was created by negation or a modal operator. This development must have contribued to a definite reversal of the traditional interpretation of the morpheme: -ara ceased to be regarded as a (primarily) temporal category and appeared now as a general irrealis marker with some – increasingly marginal – temporal uses. The frequency of temporal uses declined dramatically as -ara was partly supplanted by the competing compound past perfect and its clearcut profile was diluted due to its co-occurrences with the Indefinido in ‘high focus’ contexts of traditional or even archaic narrative prose. In addition, the double reading contexts of -ara (in which it could be interpreted as a present or a past counterfactual) and its prospective readings in epistemic and temporal contexts initiated the ‘bleaching’ of its perfective feature and paved the way for a general virtualising of the form. In ambiguous contexts like the above quoted example fasta que amansara mi ira where it was not apparent whether the situation still subsided or not an either-way-interpretation of -ara must have gained prominence increasing, in this way, the importance of the extralinguistic context and the interlocutors’ world knowledge. It was also as an irrealis form that -ara spread into modalised strong intensional contexts and only after a long transitional period, with -ara ceasing to be associated with irrealis operators, its functional spectrum became

FROM TEMPORAL TO MODAL

173

increasingly identified with that of the competing past subjunctive morpheme -ase with which it often co-occurred in the very beginning (cf. Columbus’ board book). Here again, the listeners must have reinterpreted the use of -ara within a modalised strong intensional context. Quite analogous to the previous steps, the language users shifted once again their attention from the negative or modalising operator, which was in the first moment taken to be the ‘raison d’être’ for the morphological form to the main, i.e. strong intensional, verb. Only gradually, as the examples with modalised strong intensional verbs multiplied, the language users began to recognise the strong intensional verb as the relevant operator for the selection of the verbal form. So -ara started to take over the function of a true subjunctive form competing, from then on, with traditional -ase. It must be stressed that the irrealis and the subjunctive are contiguous domains: the ‘irrealis’ covers a whole domain of non real worlds whereas the subjunctive focusses on a particular subset of non real worlds. This particular subset contains those non real worlds in which a certain requirement (a wish, an intention, a desire and so on) is fulfilled. These worlds represent in a certain way ‘ideal worlds’, which correspond to the matrix subject’s preferences. So the decisive aspect of the second change – form irrealis to subjunctive – is a semantic narrowing from the general notion of ‘irreality’ (which is compatible with a large amount of contexts) to the more specialised concept of ‘preferential possible worlds’. This specialisation should be accompanied by an increasing ousting of -ara from main clause contexts (where a new competitor, -aría, should come up) and a strenghting of the form in strong intensional contexts at the expense of -ase (as is shown by the development of -ara in final contexts). Nevertheless, the process of equalising the functional spectrum of the traditional past subjunctive -ase-form and the intruding -ara-form should take some further centuries as we could see by the data presented in the first table. The final decline of the temporal readings of -ara during the 15th century confronted the listeners with additional data confirming the prominence of the irrealis reading (and the coming up of a nascent subjunctive one) pushing aside its residual uses to the domain of archaic chivalrous prose.13 In Portuguese, the developments took the opposite direction: the available counter-evidence, which could have disclaimed the pluperfect status of -ara was simply insufficient. The modal uses, which were not predominant, could be shrugged off as extensions of the temporal function with the effect that no 13

Juan de Valdés attributes -ara in his 1536 Diálogo de la Lengua to the chivalrous prose of Garcí Ordoñez de Montalvo (as the last characteristic context of occurrence). So he writes: “(...) el autor del libro (el Amadis de Gaula, M. B.) tuvo más razón en usarlos (viniera and passara, M. B.) para acomodar su escritura a lo que en su tiempo se hablava, o por querer mostrar el antigüedad de lo que scrivia, que vos tenéis en reprehendérselos.” (Juan de Valdés, Diálogo de la Lengua, 175, cit. apud Veiga 1996:25).

174

MARTIN BECKER

shift towards an irrealis reinterpretation had taken place. As we have seen, some intrusions into other irrealis contexts were on their way, but these moves remained tentative and were not echoed by the rest of the speech community. So our case study on -ara gives some interesting insights into the nature of language change. As the common sense extensions in Spanish and Portuguese revealed, the path of change did not amount to a mere ‘random walk’, but was motivated cognitively. In our logical-semantic analysis we uncovered an underlying network of relevant conceptual structures, which created the preconditions for the linguistic developments. In this respect, we have to draw a clear distinction between motivation and inherent directionality, implying some kind of telicity: the network of cognitive relations only creates a potential for change, but it neither determines a certain outcome nor does it instantiate a certain dynamic of change. As the case of Portuguese illustrated, developments did not have to continue a possible path until an inherent endpoint – they could and did reverse the direction of linguistic change. The cognitive motivation being the necessary background-condition for change, the language usage of the speakers turns out to be the actual locus of change. It depends on them whether a reading in a particular context is consolidated and turns into a systematic reading in the long run. And, as the divergent outcomes of Spanish and Portuguese made clear, speakers (or rather speech communities) can diverge with respect to the interpretation and usage of linguistic categories. This takes us to another crucial result of our study: as we have seen, not only lexical but also grammatical categories can have different readings at one and the same time, which are related by overlapping zones (creating so-called family resemblances). However, at the same time, a prototypical centre stands out, which gives coherence and identity to the whole grammatical category (for a diachronic prototype account see Geeraerts 1997:60-62). In Spanish the shift from pluperfect to irrealis (and later to past subjunctive) was accompanied by the identification of a new prototypical center, around which the category was reorganized (with its different readings). The first reorganisation had been concluded – after several centuries – around 1500 as is attested by Nebrija’s relabelling of the category. The second reorganisation began at that time and was ongoing during the following centuries. It has not been fully concluded yet, as some irrealis relics of -ara in main clauses, like quisiera (= querría preguntar) and debiera (= debería ayudarme), prove. So the change of our category under scrutiny was transitional, took a very long time and presupposed a prototypical organisation. This prototypical organisaton with central and marginal uses implies a larger period of polysemy. However, this state of affairs did not pose a particular problem for the understanding of the sentences in which the form occurred, given some important mechanisms for the disambiguation of the form. On the one hand, the different readings were associated with particular contexts: the modal reading appeared – in a first moment – only in explicitly marked

FROM TEMPORAL TO MODAL

175

conditional clauses and the temporal reading persisted – in the end almost exclusively – in relative clauses. On the other hand, when it came to resolving the ambiguity of the [+/- perfective]-feature the speech context and the world knowledge of the interlocutors had to come into play. And finally, the speakers must have had an intuition about the prototypical reading of the category at their respective time (which does not exclude that in transitional moments the speakers did not fully coincide as to their assumptions with respect to the prototypical ‘meaning’ of -ara). The crucial moments were, of course, the periods of categorical reorganisation when the categorical center passed from one area to the other. This moment must have been triggered by a critical amount of counterevidence, which was strong enough to invalidate the traditional interpretation of the category. (To be more precise: the irrealis contexts took the upper-hand over the temporal ones). There can be no doubt that the quantity of counterevidence was different in Spanish and Portuguese. Nevertheless, the question remains open whether we must assume a ‘critical threshold’ of data responsible for change or whether we have to accept the simple fact that random also plays its part. To elucidate this point, further research should include studies on a more microscopic level, which focus on the speakers’ choices in particular speech contexts at a given period of time.

REFERENCES Andersen, Henning. 1973. “Abductive and deductive change”. Language 49. 765-93. Bechara, Evanildo. 2004. Moderna Gramática Portuguesa. Rio de Janeiro: Editora Lucerna. Bybee, Joan, Revere Perkins & William Pagliuca. 1994. The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Croft, William. 2000. Explaining Language Change: an Evolutionary Approach. Harlow: Longman. Detges, Ulrich & Richard Waltereit. 2002. “Grammaticalization vs. Reanalysis: a Semantic-Pragmatic Account of Functional Change in Grammar”. Zeitschrift für Sprachwissenschaft 21. 151-195. Farkas, Donka. 1985. Intensional Descriptions and the Romance Subjunctive Mood. New York: Garland. ----------. 1992a. “On the Semantics of Subjunctive Complements”. Romance Languages and Modern Linguistic Theory: papers from the 20th Linguistic Symposium on Romance Languages (LSRL XX), Ottawa, 10-14 April 1990, ed. by Paul Hirschbühler & E. F. Konrad Koerner, 69-104. Amsterdam & Philadelphia: John Benjamins.

176

MARTIN BECKER

----------. 1992b. “Two Types of ‘World-Creating’ Predicates”. The Joy of Grammar: a Festschrift in honor of James D. McCawley, ed. by Diane Brentari et al., 35-63. Amsterdam & Philadelphia: John Benjamins. Geeraerts, Dirk. 1997. Diachronic Protoype Semantics. Oxford: Clarendon Press. Giannakidou, Anastasia. 1997. The Landscape of Polarity Items. Groningen: Groningen dissertations in linguistics. ----------. 1998. Polarity Sensitivity as (Non)veridical Dependency. Amsterdam & Philadelphia: John Benjamins. ----------. 1999. “Affective Dependencies”. Linguistics and Philosophy 22. 367421. Giorgi, Alessandra & Fabio Pianesi. 1997. Tense and Aspect: from Semantics to Morphosyntax. New York: Oxford University Press. Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford: Clarendon Press. ---------- & Ekkehard König. 1998. “Concessive Conditionals in the Languages of Europe”. Adverbial Constructions in the Languages of Europe, ed. by Johan van der Auwera, 563-640. Berlin: Mouton de Gruyter. Haverkate, Henk. 2002. The Syntax, Semantics and Pragmatics of Spanish Mood. Amsterdam & Philadelphia: John Benjamins. Hermerén, Ingrid. 1992. El uso de la forma en RA con valor non-subjuntivo en el español moderno. Lund: Lund University Press. Hofmann, Johann Baptist & Anton Szantyr. 1965. Lateinische Syntax und Stilistik, 2 vol. München: Beck. Hundertmark-Santos Martins, Maria Teresa. 1982. Portugiesische Grammatik. Tübingen: Niemeyer. Kiparsky, Paul & Carol Kiparsky. 1970. “Fact”. Progress in Linguistics, ed. by Manfred Bierwisch & Karl Erich Heidolph, 143-173. The Hague: Mouton. Klein-Andreu, Flora. 1991. “Losing Ground: a Discourse-Pragmatic Solution to the History of -ra in Spanish”. Discourse Pragmatics and the Verb, ed. by Suzanne Fleischman & Linda R. Waugh, 164-178. London: Routledge. König, Ekkehard. 1986. “Conditional, Concessive Conditionals and Concessiveness: Areas of Contrast, Overlap and Neutralization”. On Conditionals, ed. by Elisabeth Closs Traugott et al., 229-246. Cambridge: Cambridge University Press. Kratzer, Angelika. 1978. Semantik der Rede: Kontexttheorie, Modalwörter, Konditionalsätze. Königstein: Scriptor-Verlag. ----------. 1991a. “Modality”. Arnim von Stechow & Dieter Wunderlich 1991. 639-650. ----------. 1991b. “Conditionals”. Arnim von Stechow & Dieter Wunderlich 1991. 650-656. Lapesa, Rafael. 92001. Historia de la lengua española. Madrid: Gredos. Lohnstein, Horst. 1996. Formale Semantik und Natürliche Sprache. Opladen: Westdeutscher Verlag.

FROM TEMPORAL TO MODAL

177

----------. 2000. Satzmodus – kompositionell: zur Parametrisierung der Modusphrase im Deutschen. Berlin: Akademie-Verlag. Lunn, Patricia V. 1989. “The Spanish Subjunctive and Relevance”. Studies in Romance Linguistics, ed. by Carl Kirschner & Janet Decesaris, 249-261. Amsterdam & Philadelphia: John Benjamins. ----------. 1995. “The Evaluative Function of the Spanish Subjunctive”. Modality in Grammar and Discourse, ed. by Joan Bybee & Suzanne Fleischman, 429-449. Amsterdam and Philadelphia: John Benjamins. Luquet, Gilles. 1988. Systématique historique du mode subjonctif espagnol. Paris: Klincksieck. Martin, Robert. 1983. Pour une logique du sens. Paris: Presses universitaires de France. Partee, Barbara. 1995. “Quantificational Structures and Compositionality”. Quantification in Natural Languages, ed. by Emmon Bach et al., 541-601. Dordrecht: Kluwer. Penny, Ralph. 22002. A history of the Spanish language. Cambridge: Cambridge University Press. Pinkster, Harm. 1990. Latin Syntax and Semantics. London: Routledge. Quer, Josep. 1998. Mood at the interface. The Hague: Holland Academic Graphics. Roberts, Ian. 1993. Verbs and Diachronic Syntax: a Comparative History of English and French. Dordrecht: Kluwer. Rojo, Guillermo & Emilio Montero Cartelle. 1983. La evolución de los esquemas condicionales (Potenciales e irreales desde el poema del Cid hasta 1400). Santiago de Compostela: Universidade de Santiago de Compostela. Stechow, Arnim von & Dieter Wunderlich, eds. 1991. Semantik. Ein internationales Handbuch der zeitgenössischen Forschung. (= Handbücher zur Sprach- und Kommunikationswissenschaft, 6.) Berlin & New York: Walter de Gruyter. Veiga, Alexandre. 1996. La forma verbal española “cantara” en su diacronía. Santiago de Compostela: Servicio de Publicaciones e Intercambio Científico. Werth, Paul. 1997. “Conditionality as Cognitive Distance”. On Conditionals Again, ed. by Angeliki Athanasiadou et al., 243-271. Amsterdam & Philadelphia: John Benjamins. Primary texts: Latin: The Perseus Digital Library (http://www.perseus.tufts.edu/cache/perscoll_ Greco- Roman. html) Cicero, Marcus Tullius. Epistulae ad Familiares. Ed. Louis C. Purser. 1961. Oxford: Clarendon.

178

MARTIN BECKER

Spanish: Alemán, Mateo. Guzmán de Alfarache. Ed. de Benito Brancaforte. 1996. Madrid: Akal. Alfonso X el Sabio. Prosa histórica. Ed. de Benito Brancaforte. 1984. Madrid: Cátedra. Berceo, Gonzalo de. Milagros de Nuestra Señora. Ed. de Michael Gerli. 51991. Madrid: Cátedra. Calila e Dimna. Ed. de Juan Manuel Cacho Blecua. 1984. Madrid: Castalia. Colón, Cristóbal. Los cuatro viajes del almirante y su testamento Cristóbal Colón. Ed. de Ignacio B. Anzoátegui. 1946. Buenos Aires: Espasa-Calpe. Davies, Mark: Corpus del Español. Don Juan Manuel. Libro de los enxiemplos del Conde Lucanor e de Patronio. Ed. de Alfonso I. Sotelo. 151991. Madrid: Cátedra. López de Ayala, Pero. Libro Rimado de Palacio. Ed. de Kenneth Adams. 1993. Madrid: Cátedra. Martínez de Toledo, Alfonso. Arcipreste de Talavera o Corbacho. Ed. de Michael Gerli. 41992. Madrid: Cátedra. Nuñez, Nicolas. Tratado que hizo Nicolás Núñez sobre el que Sant Pedro compuso de Leriano y Laureola llamado “Cárcel de Amor”. Ed. in: Carmen Parrilla. Diego de San Pedro. Cárcel de Amor. 1995. Barcelona : Crítica, 83-104. Poema de mio Cid. Ed. de Colin Smith. 171991. Madrid: Cátedra. REAL ACADEMIA ESPAÑOLA: Banco de datos (CORDE) [en línea]. Corpus diacrónico del español. San Pedro, Diego Fernández de. Cárcel de amor. Ed. de Carmen Parrilla. 1995. Barcelona: Crítica. Sendebar. Ed. de María Jesús Lacarra. 1989. Madrid: Cátedra. La vida de Lazarillo de Tormes, y de sus fortunas y adversidades. Ed. de José María Reyes Cano. 1998. Barcelona: Hermes. Portuguese : Universidade Nova de Lisboa: Corpus Informatizado do Português Medieval. (http://cipm.fcsh.unl.pt) Afonso X. Primeyra Partida. Ed. de José Azevedo Ferreira. 1980. Braga: INIC, pp. 3-580. Chronica dos Reis de Bisnaga. Ed. a partir de David Lopes. 1897. Lisboa: Imprensa Nacional. Porto Editora. Biblioteca Digital Colecção Clássicos da litteratura portuguesa. (http://www.portoeditora.pt/bdigital) Castilho, António Feliciano de. Poesias. Herculano, Alexandre. Eurico o Presbítero. ----------. A Harpa do Crente. Queirós, Eça de. Os maias.

FROM TEMPORAL TO MODAL

Silva, António José da. Guerras do Alecrim e Manjerona. Vicente, Gil. Auto da Alma. ----------. Auto da Barca do Inferno. ----------. Auto da Índia. Viera, Antonio. Sermão de Santo António aos Peixes.

179

NON-LEXICAL CORE-ARGUMENTS IN BASQUE, GERMAN AND ROMANCE HOW (AND WHY) SPANISH SYNTAX IS SHIFTING TOWARDS CLAUSAL HEADMARKING AND MORPHOLOGICAL CROSS-REFERENCE1

HANS-INGO RADATZ Catholic University Eichstätt

This article deals with the ways in which non-lexical core arguments can be expressed in various languages. It tries to devise a typological hierarchy for the different types and endeavours to place Romance within this hierarchy. An analysis of Basque verbal markers as cross-reference morphemes introduces the subject with a language radically different from central IE. Using Nichols’ (1986 & 1992) typological differentiation between head-marking and dependent-marking languages as its basis, a typological sub-parameter of “clausal head-marking vs. clausal dependent-marking” is suggested which is shown to correspond to two radically different types of clausal co-reference: (1) agreement (concord) and (2) cross-reference. This terminology is then used to describe and explain an ongoing syntactic change in which Spanish object clitics have evolved into obligatory verbal markers closely resembling those of Basque. Their conventional analysis as “agreement markers” is questioned and Spanish is shown to be moving towards a clausal head-marking language in which all core-arguments of the sentence have to be expressed by verbal affixes, while nominal and pronominal argument realisations become mere appositions outside the sentence core. The traditional concept of an emerging new paradigm of “object conjugation” is rejected.

1. Introduction Empty categories have traditionally been invoked to account for phenomena like the optionality of surface subjects in pro-drop languages like Latin, Italian, or Spanish and it is contended that e.g. in Italian ti amo there is no phonetic realisation of a referential subject. The descriptive device of empty categories has, however, not been limited to cases of ‘empty’ subject arguments but is also used to explain apparently unrealised object arguments in languages like Brazilian Portuguese (Goldbach 1999) and Basque. The latter has been 1 I’d like to thank Harro Stammerjohann for commenting upon an earlier draft of this paper, Martin Haase for checking my Basque examples and data, Ulrich Detges and Rolf Kailuweit for useful comments on my Saarbrücken talk, and an anonymous reviewer for valuable comments on an earlier draft of this paper. John Cleek (Minneapolis) was kind enough to polish my English style and punctuation. Many improvements on this text can be credited to them, any shortcomings, of course, are exclusively due to my own intellectual limitations.

182

HANS-INGO RADATZ

considered to be a ‘null object language’ on the basis that objects need not be realised on the surface, neither by lexical NPs nor by Pronouns. Goldbach writes: Languages like Pashto and Basque, which exhibit agreement with the subject as well as with the object but which do not license null topics, possess null subjects and null objects (translated from Goldbach 1999:64).

While this empty category hypothesis has been widely discussed with respect to the more thoroughly researched IE languages of Europe, I feel, however, that especially the Basque case calls for further discussion. In the second half of this paper, the results of this discussion will then be used to shed light also on the much more well studied languages like e.g. Spanish. Applied to Basque, the null subject and null object hypothesis has always struck me as particularly counterintuitive because it doesn’t seem to account correctly for the facts of this language’s verbal morphology. Basque possesses a complex machinery of verbal inflection that serves only one purpose: to assure the overt surface realisation of every single argument in a sentence. Actually, sentences with effectively unrealised objects are exceedingly rare in Basque and seem to occur only in connection with nominalisations like the following, where there simply are no morphological slots on the nominalised element to which non-lexical objects could possibly be attached: (1)

Zakur-rengana abiatu dog-ALATIVE.ANIM.DET.PL proceed lot-ze-ko.2 tie_up-NOMINALISER-in_order_to “He went to the dogs to tie (them) up.”

da aux.1.SG.PRES.ABS

Basque might therefore be described as having a radical dislike for any kind of non-realisation of object-arguments. However, the general agreement among generative scholars turns out to be diametrically opposed to this analysis and makes Basque a language with far-reaching object-drop capabilities. The question whether Basque either drops its objects regularly or else is reluctant to ever drop them at all is the starting-point for the following considerations: How is it possible to see the same state of affairs so differently? It may prove insightful to set this question in the larger context of the strategies to be found in various languages for the realisation of non-lexical subject and object arguments, i.e. I shall take an onomasiological stance on a universal syntactic problem. In particular, I shall consider the question of drawing a boundary between the concepts of ‘agreeing with’ and ‘instantiating’ an argument in various language types. There appears to be a continuum of 2 This is a sentence I found in a teach-yourself Basque manual (Beaumont & Lazkano 1998:111).

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

183

referential forms used by different languages for the expression of non-lexical arguments which ranges from morphologically free forms like German or English personal pronouns, via clitics (most Romance languages) to the polysynthetic realisations of cross-reference languages like Basque. I shall argue that the features to be found in certain modern Romance languages (essentially Spanish, Catalan, Galician, and European Portuguese) were best explained as a typological shift towards a type quite similar to that of Basque. As this is a generally much less familiar language than the others to be discussed later, I feel that it is important to dwell rather more extensively on a description of those basic facts of Basque syntax I will later cite to analyse the ongoing typological change within Romance. 2. The Realisation of non-lexical arguments in Basque Basque is a language isolate showing ergative morphology, i.e. subjects of intransitive (or rather: unaccusative) verbs as well as direct objects are marked for absolutive case, whereas subjects of transitive verbs take ergative case marking. Basque possesses only very few verbs capable of taking synthetic inflection-affixes: according to the authoritative Basque grammar of the Basque Language Academy Euskaltzaindia from 1987, out of originally over 60 synthetical verbs in the earliest surviving texts, only 26 are still in modern usage and even fewer if we exclude those which are limited to the literary language (cf. Gómez & Sainz (1995:238f.). Most verbal concepts can only be expressed periphrastically with the aid of an auxiliary. The small number of synthetic verbs in combination with a morphology of the agglutinating type, i.e. a verbal morphology with relatively little allomorphy and amalgamations, allows Basque verbs to co-refer not only with their subjects, but also with all other arguments, direct and indirect objects alike. There is even a fourth type of co-reference affix, called ‘allocutive’ referring to second person non-arguments of the ‘ethical dative’-type, which I shall make no further mention of as we will only be concerned with the realisation of corearguments (cf. Gómez & Sainz 1995:236). Moreover, Basque has consistently been classified as a pro-drop language in generative accounts, given that nominal and pronominal subjects appear to be freely elidable. How then do all these elements interact in the construction of intransitive, transitive, and ditransitive sentences?

HANS-INGO RADATZ

184

2.1 Basque, one argument (intransitive, Nor3) By way of exemplification, let us first analyse the intransitive Basque sentence Ni abiatzen naiz “I set out”, Zu abiatzen zara “You set out” etc. as follows: 1 sg.

(ni-ø)i

naizi

2 sg.

(zu-ø)i

zarai

3 sg.

(hura-ø)i

1 pl.

(gu-ø)i

dai

2 pl.

(zuek-ø)i

zaretei

3 pl.

(haiek-ø)i

dirai

abia-tzen

pron-ABS

set_out-PTCP.PRS

garai

aux be.PRS.IND.ABS\1SG …

“[I / you / he she / we / you / they] set(s) out.” Figure 1: Basque, one argument (intransitive)

The example shows absolutive case marking on the facultative subject pronoun (ni, zu, hura etc.), a non-finite verb-form carrying lexical and aspectual information, and the auxiliary izan “to be”, marked for tense and mood and co-referenced with the subject pronoun. Utterances like abiatzen da “he sets out” and abiatzen gara “we set out” without a subject noun phrase or pronoun nevertheless constitute full sentences. Like the pronouns, nominal 3rd person subjects as well carry absolutive case which is marked by a ø-ending:4 (2)

Jon-ø abiatzen da. Jon-ABS parting is “Jon is setting out.”

2.2. Basque, two arguments S-DO (transitive, Nor-Nork) In Basque transitive utterances with two arguments of the type ‘He sees the book’, absolutive case marking goes to the direct object whereas the transitive subject receives ergative case. In this way, subjects of transitive sentences are morphologically distinct from those of intransitive sentences. Again, the verbal concept (here ‘to see’) is realised periphrastically; in transitive sentences, the 3

This is the traditional name for the construction in Basque grammar. Nor, nori and nork are the case-marked forms of the interrogative “who” which are used as the Basque names for absolutive, dative and ergative as well as for the sentence constructions in which they participate. 4 The zero-endings marked as ø in the morphological analysis represent slots in the paradigm in which the absence of an affix is one of a set of possible values within the paradigm; ø therefore contrasts with the presence of other affixes and marks a zero-ending, not the zero-realisation of an argument (= Ø). These morphological facts have nothing to do with the descriptive device in syntactic analysis of positing empty categories like traces or pro and the like.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

185

auxiliary has to be ukan ‘have’, rather than izan ‘be’, as only ukan has the necessary cross-reference capabilities which allow co-reference not only with the ergative-subject but also with the absolutive-direct object. 1 sg.

(ni-k)i

2 sg.

(zu-k)i

3 sg.

(har-k)i

1 pl.

(gu-k)i

dj-u-gui / ditj-u-gui

2 pl.

(zue-k)i

dj-u-zuei / ditj-u-zuei

3 pl.

(haie-k)i

dj-u-tei / ditj-u-ztei

pron-ERG

dj-u-ti / ditj-u-ti liburu-aj / liburu-akj

ikus-ten

dj-u-zui / ditj-u-zui dj-u-øi / ditj-u-øi

book-DET.ABS.SG/

see-

ABS.3SG-aux

have-PRES.IND.ERG.1SG … /

book-DET.ABS.PL

PTCP.PRS

ABS.3PL-aux

have-PRES.IND.ERG.1SG …

“[I / you / he she / we / you / they] see(s) the book(s)” Figure 2: Basque, two arguments (S + DO-3 Sg)

Somewhat simplifying the morphological situation, one might analyse the affix du- as co-referencing the singular direct object while ditu- corresponds to a direct object in the plural. At the same time, du- / ditu- are also marked for 3rd person, as there are special forms for 1st and 2nd person as well: (3)

(Ni-k) ikus-ten I-ERG.SG see-PTCP.PRS “I see you (sg.).”

zai-tu-t. ABS.2SG-aux have.PRS.IND-ERG.1SG

(Zu-k) ikust-en you-ERG.SG see-PTCP.PRS “You (sg.) see me”, etc.

nau-ø. ABS.1SG\aux have.PRS.IND-ERG.2SG

Again, the inflected verb group on its own already constitutes a complete sentence, as all lexical arguments are freely elidable: (4)

Ikust-en see-PTCP.PRS “I see them”.

di-tu-t. ABS.3PL-aux have.PRS.IND-ERG.1SG

Similar affixes – and similar one-word sentences – also occur with the other synthetic verbs like e.g. jakin ‘to know’: (5) a.

Arrazoi-a d-aki-t. reason-DET.ABS.SG ABS.3PL-know.PRS.IND-ERG.1SG “I know the reason.”

186

HANS-INGO RADATZ

Ba-d-aki-t. ENUNC-ABS.3PL-know.PRS.IND-ERG.1SG “I know it.” b.

Arrazoi-ak d-aki-zki-t. reason-DET.ABS.PL ABS.3PL-know.PRS.IND-ABS.3PL-ERG.1SG “I know the reasons.” Ba-d-aki-zki-t. ENUNC-ABS.3PL-know.PRS.IND-ABS.3PL-ERG.1SG “I know them.”

c.

Arrazoi-a d-aki-gu. reason-DET.ABS.SG ABS.3SG-know.PRS.IND-ERG.1PL “We know the reason.” Ba-d-aki-gu. ENUNC-ABS.3SG-know.PRS.IND-ERG.1PL “We know it.”

d.

Arrazoi-ak d-aki-zki-gu. reason-DET.ABS.PL ABS.3PL-know.PRS.IND-ABS.3PL-ERG.1PL “We know the reasons.” Ba-d-aki-zki-gu. ENUNC-ABS.3PL-know.PRS.IND-ABS.3PL-ERG.1PL “We know them”.

It is this type of structures that has given rise to the description of Basque as a null object language. Actually then, Basque is not only considered to be prodrop but has also been dubbed an ‘object-drop-language’ (Landa & Franco 1996:160). Saltarelli goes even one step further and characterises Basque as a ‘null argument language’ in which any type of lexical argument is always optional as far as syntax is concerned: [...] with respect to the issue of empty categories, Basque is a null subject language, as Italian has been defined. Moreover, Basque is a null object language (direct and indirect). If we follow Taraldsen’s original hypothesis that null subject languages are a consequence of the richness of the verb inflectional properties of the language, then Basque is properly a ‘null argument language’ since its verb inflectional parameter marks subject (both thematic and non-thematic), direct object and indirect object (Saltarelli 1988:XIX).

2.3. Basque, three arguments S-IO-DO (transitive, Nor-Nori-Nork) By adding an indirect object as a further argument, we get the ditransitive construction of the type “He gives the man the book” which is rendered in Basque by marking the subject as ergative, the direct object as absolutive, and the indirect object as dative. Again, all arguments are cross-referentially represented on the auxiliary ukan:

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

1 sg.

(ni-k)i

2 sg.

(zu-k)i

3s sg.

(har-k)i

187

dj-i-ok-ti gizon-arik

liburu-aj

ema-ten

dj-i-ok-zui dj-i-ok-øi

etc.

etc. pron-ERG manDET.DAT.SG

bookDET.ABS.SG

givePTCP.PRS

ABS.3SG-AUX have.PRS.IND-DAT.3SG-ERG.1SG …

‘[I / you / he she / we / you / they] give(s) the man the book’ Figure 3: Basque, three arguments (S-IO-DO)

As can be gathered from the gloss, the auxiliary now has to carry an absolutive marker referring to the direct object, a dative marker referring to the indirect object and an ergative marker, securing co-reference with the subject. Again, the table only gives a glimpse of the actual possibilities and intricacies of Basque cross-reference morphology. A sentence like (6) below: (6)

gizon-arik liburu-aj (nik)i (pron-ERG) man-DET.DAT.SG book-DET.ABS.SG dij-ok-ti ABS.3SG-AUX have.PRS.IND-DAT.3SG-ERG.1SG “I give the man the book.”

ema-ten give-PTCP.PRS

shows a reflex of plurality on the direct object and on the auxiliary, if the direct object is in the plural: (7)

(nik)i gizon-arik liburu-akj book-DET.ABS.PL (pron-ERG) man-DET.DAT.SG dizkij-ok-ti ABS.3PL-AUX have.PRS.IND-DAT.3SG-ERG.1SG “I give the man the books.”

ema-ten give-PTCP.PRS

Changing the indirect object to plural as well, we get (8): (8)

gizon-eik liburu-akj (nik)i (pron-ERG) man-DET.DAT.PL book-DET.ABS.PL dizkij-ek-ti ABS.3PL-AUX have.PRS.IND-DAT.3PL-ERG.1SG “I give the men the books.”

ema-ten give-PTCP.PRS

As can be seen in these examples, Basque is markedly different from the central Indo-European languages in that its finite verbs obligatorily reflect in their morphology all basic relationships of argumenthood in a sentence, a syntactic property known as cross-reference in language typology. The status of these object affixes has been described in various ways in the literature:

188 –

–

–

HANS-INGO RADATZ

Schwerteck (1984:9), in talking about the incorporating verb forms, calls the object affixes “pronouns”, “segmental markers” und “pronominal elements”.5 This terminology reflects the intuition that these affixes themselves carry out syntactic functions which go beyond mere agreement; calling them pronouns, however, would endow them with a degree of morphological independence which they definitely do not possess. Franco (1991 and 2000) talks about “verbal agreement morphemes”, Goldbach (1999:10) about a “system of object agreement”6 and “inflectional endings on the verb”.7 This is standard generative terminology which takes into account their stronger morphological integration into the verb, but sees them as mere agreement devices. Saltarelli (1988:XVII) generally uses the unspecific term “marker”; he does, however, call Basque a “highly inflected language” and seems to place less weight on the agglutinative character of Basque morphology.

Most authors seem to concur then in considering these markers ‘agreement affixes’ of some kind and the following analysis by Landa and Franco represents a generally accepted view concerning the structure of Basque sentences. As in English or Spanish, verbal markers are seen as only coindexing the real (lexical or pronominal) arguments. If these happen to be absent, this phenomenon is seen as ‘elision’ and the index goes to an empty category – a trace or a pro: (9)

Jonek aulkia apurtu du eta nik konpondu dut. Jon-ek aulki-ai apurtu d-u-ø eta ni-k Øi Jon-ERG chair-ABS break ABS.3SG-AUX have-ERG.3SG and I-ERG konpondu d-u-t fix ABS.3SG-AUX have-ERG.1SG “Jon has broken the chair and I have fixed (it).” (Landa & Franco 1996:161)

5 “Pronomina [...] segmentale Bedeutungsträger [...] pronominale Elemente” in the German original. As to the historical origin of these polysynthetic verbal complexes of Basque he writes: “It is by no means new that Basque verbal forms, when retraced to former states in their history, have been considered to be entire sentences ([Footnote 21:] Examples for this view can be found e.g. in H. Schuchardt’s Primitiae linguae vasconum)” (translated from Schwerteck 1984:15). 6 “Just as the null-subject possibilities of Italian and Spanish follow from the potential of their subject inflection, the null-object possibilities of Pashto and Basque result from their system of object agreement” (translated from Goldbach 1999:65f.). This is, of course, an example for what Haspelmath (1999:188) has dubbed the “teleological fallacy”: rich agreement morphology may be an important factor in facilitating the described phenomena but can not in itself be seen as causing it. 7 “[...] like for example in Basque, where inflectional endings unequivocally indicate the possible subjects (or objects) of a verb” (translated from Goldbach 1999:10).

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

189

According to this theory, the direct object of the coordinated sentence can be elided easily because the object agreement markers on the verb identify the null object completely. They co-index the direct object aulkia “chair” with the empty constituent in the object position of the coordinated sentence. Note, however, that this is by no means the only possible analysis; after all, both auxiliaries already contain a marker for 3sg. absolutive and we might envisage an alternative view in which the observable verbal affixes, rather than a hypothetical Ø, serve as the locus for co-indexation with the lexical direct object as in (10): (10)

Jonek aulkiai apurtu di-u-ø Jon-ERG chair-ABS break ABS.3SG-AUX have-ERG.3SG eta nik konpondu di-u-t. and I.ERG fix ABS.3SG-AUX have-ERG.1SG “Jon has broken the chair and I have fixed (it)”

Why use an empty constituent if there are actual surface elements available that could be seen either as agreement markers or just as well as pronoun-like representations of the direct object itself? As a consequence of the nonobligatory nature of Basque NP-arguments on the one hand and Du Bois’ Preferred Argument Structure constraint (“Avoid more than one lexical core argument” cf. Du Bois 2003:34ff.) on the other, the great majority of all arguments would then have to be treated as not occurring on the surface, triggering an immense proliferation of empty syntactic categories. This is likely to be an unnecessary violation of any principle of descriptive economy and therefore a classical candidate for an application of Occam’s razor. A further relevant observation about this “empty constituent” can be gained from a comparison with null objects in other languages like e.g. Spanish. A major difference between Basque and Spanish lies in the fact that in Basque the only possible alternative to the null object would be either the full lexical noun phrase aulkia “chair” itself or else an emphatic demonstrative. This is due to the (highly noticeable!) fact that Basque, unlike Spanish, does not possess a class of non-demonstrative object pronouns. Nevertheless, the direct object does not have to be contextually inferred; rather, it is unequivocally specified in the verbal morphology as to case and number, namely absolutive singular. That is, the object affixes are specified with the same or even more precision than the corresponding Romance clitics or the German pronouns. One might ask, therefore, why Basque verbal affixes are usually not considered to be realisations of arguments but are seen rather as mere agreement morphology referring to lexically realised arguments or their respective traces. These elements are clearly different from such well-established agreement markers as e.g. subject agreement morphemes in German or English. In these languages, verbal markers do not only agree with noun phrase subjects but, in

190

HANS-INGO RADATZ

their absence, also with a class of pronominal subjects which in non-pro drop languages as German are even obligatory: (11) a. b.

Peter singt. Er singt. *Singt. Peter sings. He sings. *Sings.

This is prototypical ‘agreement’: morphological co-reference with an overt NP or pronoun. Basque, however, is different. While first and second person indirect objects might still be overtly realised as case-inflected forms of emphatic personal pronouns, no such option is available for third person objects. Basque only has 1st and 2nd person genuine personal pronouns; it is only for the sake of completeness that grammars tend to list the demonstratives hura and haiek for the function of emphatic 3rd person pronouns (cf. Zubiri 2000:47). Therefore, if we consider the verbal markers to be mere agreement morphemes, the least emphatic way of expressing third person pronominal objects on the surface would be demonstratives. This syntactic fact casts further doubt on the analysis of Basque verbal affixes as agreement markers, because if these elements actually were agreement affixes, the question arises: What are they are supposed to be agreeing with? Let’s finally turn to a last example in which the auxiliary is not only marked for cross-reference but also for tense and mood, and which is therefore better suited to show the full possible complexity of Basque finite verbs and the agglutinating character of the language (taken from Saltarelli (1988:XVII)): (12)

Guk gizonei liburuak eman diezazkiekegu. Gu-k gizon-ei liburu-ak we-ERG man-DET.DAT.PL book-DET.ABS.PL d -i -eza -zki ABS.3PL-DAT.3PL -AUX have.PRS -ABS.3PL “We can give the books to the men.”

ema-n give-PRTC.PERF -e -ke -gu -DAT.3PL -MOOD -ERG.1PL

All lexical arguments have to be co-referenced by a marker on the auxiliary: – The subject, i.e. the ergative pronoun by the marker ERG.1PL <-gu>. – The dative indirect object by the discontinuous dative marker for 3rd person and <-e-> for plural. – And finally, the absolutive direct object by the discontinuous absolutive marker for 3rd person and <-zki-> for plural. The complete representation of all arguments on the auxiliary allows for any lexical ergative, absolutive, and dative argument to be optional and the verbal complex alone constitutes a complete grammatical utterance:

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

(13)

Eman diezazkiekegu. Ema-n d -i give-PRTC.PERF ABS.3PL-DAT.3PL -ke -gu -MOOD -ERG.1PL

-eza -zki -AUX have.PRS -ABS.3PL

191

-e -DAT.3PL

“We can give them to them.” The situation in Basque is thus radically different from that in German or, if slightly (and sometimes even considerably) less, from that of the Romance languages. Now, there may be the danger of illicitly projecting our intuitions gained from more familiar Indo-European languages onto the situation in a non-Indo-European language like Basque, for which reason the following two questions are by no means trivial: – Are the ergative, dative and absolutive noun phrases, guk, gizonei and liburuak, really the subject, indirect, and direct object of the sentence, in the same sense in which the nominative, dative, and accusative arguments in the German sentence Wir können den Männern die Bücher geben (“We can give the books to the men”) would be said to instantiate the subject, indirect object, and direct object? – Are the Basque cross-reference affixes really only agreement markers, comparable to verbal agreement in German Wir können (“We can”)? An answer to these questions is not only important to Bascologists but could also aid us in reassessing the well-known facts of the Romance languages, shedding light on the syntactic status of clitics in those constructions that have sometimes been called “clitic-doubling” and sometimes “Romance object conjugation”. 3. The typological parameter head-marking vs. dependent marking In order to do justice to languages like Basque, Johanna Nichols in her 1986 seminal paper on head-marking vs. dependent-marking languages introduced a major typological distinction that may well be relevant to the problem of non-lexical argument realisation in various languages.8 Nichols draws our attention to an important empirical imbalance in linguistic research that existed at that time (and most probably persists until today): while headmarking features are statistically very important and probably even majoritary among the languages of the world (Nichols 1986:89ff.), this language type had received only very little if any attention by linguists – including those who claimed to be working on language universals:

8

In this paper she discusses a sample of over 60 typologically diverse languages, among which there is – quite surprisingly – not one Romance language! The author developed her theory further in Nichols (1992).

HANS-INGO RADATZ

192

Despite the efforts of formal grammarians to take a range of languages into consideration, there is a glaring gap in the typological coverage: the exotic languages that have so far received significant attention have been almost exclusively dependentmarking (Japanese, Korean, Finnish, Malayalam, Australian languages) or doublemarking (Turkish, Arabic, Hebrew, Australian languages). Of the head-marking languages, only Navajo has received significant theoretical attention. In addition, Relational Grammar has investigated a number of head-marking languages (primarily from the Algonkian, Salishan, and Wakashan groups); but these languages have not been used to raise questions of constituency, centricity, syntactic bonds, government etc. (Nichols 1986:115f.).

Nichols’ typological analysis is based on two universal principles, namely headedness and overt morphological marking. Headedness is meant to convey the notion that natural language constituents regularly come as structured into a head, an obligatory constituent determining the overall syntactic status of the whole constituent, and one or more facultative dependents which derive their morphosyntactic accidents from the head. This type of configuration can be shown to exist not only at the phrase but also at the clause and sentence level: Constituent

Noun phrase

Clause Sentence

Head

Dependent

possessed noun

possessor

noun

modifying adjective

adposition

object of adposition

predicate / verb

arguments / adjuncts

auxiliary verb

lexical / main verb

Main-clause predicate

subordinate clause

Figure 4: Head-dependent-configurations (cf. Helmbrecht 2001:1425)

Typical dependent-marking structures can be found e.g. in English possessive constructions like the man’s house, in which the noun house is the head whereas possession is morphologically marked on the dependent by some kind of genitive affix. In Hungarian, on the other hand, the same relationship is expressed by a possessive affix on the head itself (cf. Figure 5). However, while Nichols’ typology considers head-dependent configurations on all syntactic levels, I shall from this point on limit myself to phenomena at the level of sentences, or rather: of non-subordinated or ‘main clauses’. Nichols’ basic claim is that languages throughout the world show a strong tendency to mark grammatical relations consistently either on the head or on the dependent. At the same time, however, she concedes that this is only a statistical regularity and individual languages usually adhere only “predominantly” to one or the other type while any degree of admixture is always possible.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

English

Dependent

Head

the man’s

House

DET

Hungarian

man-POSS

az ember DET

man

193

House ház-a house-3.SG

Figure 5: Dependent vs. head-marking within the NP

If the phenomenon can occur independently on different levels, it should be legitimate to extract typological sub-parameters from the general theory. I therefore suggest a typological parameter clausal head-marking to be derived and abstracted from the general principle that Nichols advocates for. As to the distribution of head and dependent on the clause level, Nichols argues (1986:57) that predicates constitute the head, whereas arguments and adjuncts can be analysed as dependents.9 Clausal head-marking languages, then, are those in which all arguments of the sentence are usually realised by morphological means on their finite verbs or auxiliaries. I contend that the question of non-lexical argument realisation can be seen along the lines of clausal head-marking vs. clausal dependent-marking and that Nichols’ parameter may be used to shed new light on the historical development of clitic systems in Romance in comparison with languages like Basque or German which represent markedly different types. Due to the fact that the tendency for consistently marking either on the head or on the dependent is not absolute, there are actually further types to be found: in double-marking, syntactic relations are redundantly marked on heads and dependents alike, whereas in split-marking languages, the situation may be different, depending on the grammatical subsystem. Basque is actually a good example for both types: at the sentence level, all syntactic relations are reflected on the verb (head) but argument noun phrases are also marked for case so that we get clausal double-marking. Moreover, Basque belongs to the split-type, as double-marking is limited to the clause level in this language, which is otherwise largely dependent-marking on the phrasal level. For an example of a genuine head-marking language we may want to take a look at Abkhaz. Argument noun phrases in Abkhaz bear no case markers whatsoever and the entire morphosyntactic information is polysynthetically realised on the verb, which morphologically cross-references with the lexical arguments for person, number, and gender. In the absence of morphological case, the syntactic relations are encoded through the serialisation of verbal 9 In other cases, auxiliaries and their respective lexical (‘main’) verbs may also stand in a clausal head-dependent relation to each other. I will not dwell further on this interpretation.

HANS-INGO RADATZ

194

affixes. All argument noun phrases are optional and appear only if they are discourse-pragmatically needed to establish the reference of the verbal affixes. (14)

Pure head-marking language: Abkhaz a-xàc’a a-pћ°əs a-š°q°’ə ø-lc-y-te-yt’ the-man the-woman the-book it-to_her-he-gave-FINITE “The man gave the woman the book.” (Helmbrecht 2001:1427)

The inflected verb group again constitutes a complete sentence,10 quite similar to the situation in Basque, as we have already seen: (15)

ø-lc-y-te-yt’ it-to_her-he-gave-FINITE “He gave it to her.”

Now, Nichols observes that this effect seems to be characteristic of all languages with head-marking morphology at the clause level. This is hardly surprising, as these languages code the entire grammatical information about the clausal arguments like case, number, gender etc. on the verb, leaving noun phrases with the sole function of providing lexical information about new discourse participants. Speakers are free to omit lexical or pronominal material if it is referentially recoverable in the discourse universe. In these languages, the surface appearance of lexical noun phrases and pronouns is bound exclusively to discourse motivations and can therefore be completely optional at the syntactic level. When full noun phrases appear at all, they behave like appositions with respect to their coreferential verbal affixes and are realised outside the sentence core.11 In head-marked grammatical relations, the dependent is usually an optional element of the constituent. For instance, in languages with consistently head-marked clauses, the verb itself normally constitutes a complete sentence; full NP’s are included only for emphasis, focus, disambiguation etc. (Nichols 1986:107).

So, in head-marking languages there is a sharing of functions between the verbal affixes and the full noun phrases which is quite comparable to that between full verb and auxiliary in verbal periphrases, in that one element, the auxiliary, specialises in conveying only grammatical information while the lexical information is left entirely to the other element, the full verb. In the same vein, head-marking verbal affixes specialise in the purely grammatical 10

“The appositional status of NPs in Abkhaz can be found in many other non-European languages and may be viewed as a defining feature, among others, of polysynthetic languages” (Helmbrecht 2001:1427). 11 That lexical noun phrases should be analysed as appositions (or ‘adjuncts’ or ‘dislocated elements’ respectively) in these cases has also been suggested by generativist authors. Aoun (1993) proposes such an interpretation for Lebanese Arabic (cf. the critical discussion in Franco 2000:152ff.).

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

195

function while the corresponding appositions carry the lexical information, may receive contrastive stress and can be freely positioned in the sentence periphery according to the requirements of discourse pragmatics. The parallelism is, however, not complete, as in verbal periphrases both elements are obligatory while in head-marking languages only the verb complex is. In Basque and Abkhaz sentences, the verbal affixes already seem to constitute the arguments while the full noun phrases show all the same behaviour of facultative appositions. Nichols was by no means the first to propose this analysis; rather it seems to have been the view of almost any linguist ever to have worked on American Indian languages: The dependent in a head-marked constituent stands in a roughly appositive relation to the head (or, more precisely, to the coreferential marker on the head); the term ‘government’, developed by traditional grammar on the basis of exclusively dependent-marked relations [...], is not appropriate for head-marked constituents. Since the appearance of Boas 1911, descriptions of American Indian languages have insisted that subject and object in these languages are in apposition to the pronominal markers on the verb, rather than (as in Indo-European) being syntactically governed by a verb which agrees with them (Nichols 1986:107).

One of the most prominent advocates for this typological distinction is cofounder of Role and Reference Grammar (RRG) Robert Van Valin, who was drawn to it by his work on the head-marking language Lakhota. This may be the reason why RRG is one of the few grammatical theories on whose very foundations a systematic distinction between head-marking and dependentmarking was implemented. In a short introductory manifesto on RRG, Van Valin writes: In head-marking languages like Lakhota, the bound pronominals on the verb are considered to be the core arguments; overt NPs are within the clause in apposition to them (Van Valin 2005:3)

According to this analysis, Basque verbal affixes would no longer have to be considered agreement markers but rather ‘core arguments’ in themselves; overt noun phrases, on the other hand, would regularly be seen as appositions, which would not only account for their unlimited elidability but also for the lack of syntactic constraints on their position relative to the core-sentence and to each other. The fundamental typological difference between these two marking-types should also be reflected in the terminology that is used to describe the different type of clausal co-reference that can be observed in head-marking and dependent-marking languages. Most authors, however, tend to carry over the notion of ‘agreement’ also to head-marking languages as well, thus obfuscating the fundamental differences between the two types. This is unnecessary as the

HANS-INGO RADATZ

196

terminological clarification has been around for over seventy years and, what is more, was published in Bloomfield’s easily accessible, classical monograph “Language”. Here Bloomfield (1933:191ff.) introduces the term crossreference to refer to the type of clausal co-reference which is typically found in head-marking languages : Agreement (= concord)

Cross-Reference

dependent-marking

head-marking

the argument (= dependent) is realised as a full NP or as a pronoun and is morphologically marked (dependentmarking)

there is co-reference with a non-obligatory full lexical NP which syntactically behaves like a non-core constituent (apposition)

the verb (head) carries a redundant agreement marker

the argument is realised by an obligatory affix on the verb (head-marking

example: subject agreement in German or English

example: Basque, Abkhaz; tendencies in Romance, e.g. French Jean où est-il? (Bloomfield 1933:193)

Figure 6: Two types of clausal co-reference

Thus, whereas clausal co-reference is morphologically realised as ‘agreement’ in languages with clausal dependent-marking, the corresponding morphological process in clausal head-marking languages is ‘cross-reference’ – and therefore radically different. If it is true that in language typology it is not important what can be but rather what must be expressed in a language, then Basque definitely has to be described as a language in which all objects have to be expressed on the surface. And it is as far as can be from a null-subject, nullobject or null-argument language. Rather, null-arguments are impossible in Basque because the verbal morphology requires the specification of all arguments. Lexical noun phrases, which by analogy with dependent-marking languages could be mistaken for subjects and objects, do not belong to the sentence-core and must be viewed as non-obligatory appositions providing additional lexical and referential information about the arguments they are coreferenced with. Rather than being arguments they only partake in the argumenthood of the verbal affixes. It should be noted that the typological importance of the parameter ‘clausal head-marking’ derives largely from the cross-reference morphology on the verb that comes with it. Whether lexical and pronominal arguments carry additional case marking, as e.g. in Basque, or at least ‘differential object marking’ (cf. Bossong 1991), as e.g. in Spanish, is without consequence in this context. On the clause level we can therefore class head- and double-marking languages together and oppose them with the dependent-marking languages and their agreement morphology as the only genuinely different type.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

197

If with cross-reference the traditional terminology of “agreement” vs. “argument” is misleading and inappropriate, which terms should be used then in describing the phenomena adduced above? Du Bois (1987), working on the ergative cross-reference Mayan language Sacapultec, provides us with a terminology which most adequately captures the syntactic facts in languages like Sacapultec and Basque. Where dependent-marking languages tend to have two separate things, i.e. arguments and agreement, head-marking languages only have ‘mentions’ which Du Bois defines as a reference item complex consisting of either a bound form alone (a cross-referencing affix [...]), or an overt free form (full NP or independent pronoun) plus its crossreferencing bound form within the same clause (Du Bois 1987:813).

A ‘mention’ then can consist of an obligatory cross-reference marker alone, in case the lexical noun phrase remains un-expressed. If, however, a lexical NP is present, the NP plus the verbal marker together instantiate one discontinuous argument realisation, bound together by cross-reference. This notion acknowledges the fact that in these languages the grammatical and the lexical aspect of argumenthood are typically split up, only the grammatical element being obligatory. We will have to see whether the notion of ‘mention’ might not be better suited to account for certain Romance languages notorious for the problem of determining the status of clitics and affixes. In this conception, the verbal affixes in Basque constitute actual instances of the arguments with an independent external reference. I have to thank an anonymous reviewer for pointing out to me a number of conventional tests for argumenthood which, apparently, would all have to yield an affirmative result in order to substantiate this hypothesis. These tests include the questions: i. ii. iii. iv. v.

Can these affixes be the answer to a question? Can these affixes carry the main accent? Do these affixes allow for ellipsis of the inflected auxiliary? Do these affixes combine with other DPs or NPs? Are these affixes flexible in their position with respect to each other and the verb?

Quite clearly, the answer to all the above questions is no. This does, however, not necessarily invalidate these elements’ claim to argumenthood. All the above tests are diagnostics designed to detect arguments in dependentmarking languages in which arguments must be realised as free morphemes. But, as detailed above, it is precisely the nature of cross-reference languages that they realise arguments not as free morphemes but rather as ‘mentions’ which, by definition, can consist of just a verbal affix. In their mentions we find a repartition of function between argument-realising verbal affixes and their facultative appositional counterparts; the affixes fail all the above tests because the tests rely on those syntactic circumstances which would call for the

198

HANS-INGO RADATZ

otherwise elided appositions. The failure of Basque verbal affixes to comply with the traditional tests for argumenthood can therefore not be adduced to disprove their status as arguments in a head-marking language. 4. The other end of the hierarchy: German At the other end of the hierarchy of clausal head-marking we find languages like German. Its only head-marking feature on the clausal level is concord, i.e. the typical Indo-European agreement of verbs with their subjects.12 Actually, according to Nichols, pure clausal head-marking seems to be the exception rather than a statistically relevant type, and we may therefore disregard subject agreement and treat German for all practical purposes as a dependent-marking language. In German, only full noun phrases and pronouns can function as arguments: (16)

Ich gebe Günther das Buch. Günther the.ACC book. I.NOM give “I give Günther a book.”

(17)

Ich gebe es ihm. / Ich gebe ihm das. / I give it him / I give him that / Ich gebe dem das. / I give it him. Ich gebe es dem. / I give him that / “I give it to him.”

How free or how bound these pronouns may be in each case, i.e. how far cliticisation has already progressed in German, is not relevant in our context. It is, however, safe to say that German pronouns are syntactically much closer to lexical noun phrases than any ‘weak’ or even ‘strong’ pronoun in the Romance languages (with the possible exception of Brazilian Portuguese, which seems to be moving away from the rest of Romance in this respect). In the context of the head-marking morphology in Basque, I asked whether we wanted to analyse the verbal affixes as actual realisations of or as mere agreement markers with the subject, direct and indirect object. Irrespective of the answer to this question, no one can deny that such a question might at least be raised for Basque. For German, however, it would make no sense at all: – The subject markers on German verbs are clearly agreement affixes referring to an independently realised, overt obligatory subject – a full noun phrase or a subject pronoun. – Object clitics, pronouns, and noun phrases clearly consitute the realisations of the respective arguments and are typically not appositional. Noun 12

In this respect, English would be an even better example, as concord has been reduced to a rather precarious agreement in 3rd person singular verbs. But if English has almost no headmarking, it also has almost no morphological (case-) marking on the dependents, which makes it a completely different type.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

199

phrases and pro-forms are in complementary distribution and there can be no clitic doubling. (18) a.

Ich gebe es ihm. I give it him “I give it to him.” b. *Ich gebe esi ihm dasi. I give it him that c. *Ich gebe es ihmi Peteri. I give it him Peter gefällt das. d. [*Miri / *Ichi] miri [to_me / I] to_me appeals that “This appeals to me / I like this.”

5. The Romance languages – problematic intermediate cases We have now examined one example of a language with full clausal headmarking and another one with full dependent-marking morphology. The best way to characterise the Romance languages would be to locate them on an abstract scale on different points in between these two more extreme examples. This intermediate status of Romance has already been noticed by Nichols, who explicitly singles out the Romance case as an example of ongoing typological change within the group of the otherwise quite consistently dependent-marking Indo-European languages: Indo-European has retained its basic type – dependent-marked with subject inflection on verbs [...] for languages preserving the inherited morphology – for some 6,000 years, with only a recent trend toward head-marked clauses in the pronominal clisis of the Romance languages (a process which occurs only after most of the morphology has been lost) (Nichols 1986:89).

This change is driven by a general Romance tendency to introduce new, alternative head-marking strategies. Historically, most pronouns were reduced to pronominal clitics and these have frequently been further amalgamated into affixes and affix groups which increasingly resemble the cross-reference morphology of Abkhaz, Lakhota or Basque, especially in those Romance languages in which obligatory clitic-doubling is gaining ground. Like most other Indo-European languages, the Romance languages all possess subject co-reference morphology on their verbs, which has traditionally been considered to be agreement just as in the case of languages like German. This view, however, is not undisputed and thus Corbett (1994) finds it disturbing to speak of agreement in a situation where the non-realisation of the purported referent of the agreement is the statistically normal, unmarked case:

200

HANS-INGO RADATZ

Another common problem, and a serious one in some theoretical frameworks, is the existence of so-called ‘pro-drop’ languages. It may be desirable to say of certain languages, for example, Serbo-Croatian, that verbs agree with their subject in person and number, as in ja čitam ‘I read’, ty čitaš ‘you read’. In most circumstances, the more natural utterances would be čitam and čitaš. The problem is that the pronominal agreement controller is normally not present (Corbett 1994:57).

Kailuweit’s (2005) position within the framework of Role And Reference Grammar is more differentiated, as he considers only part of the cases as instances of agreement. Kailuweit no longer analyses the subject markers on Spanish verbs as agreement affixes as has been traditionally done. On his account, at least as long as no lexical or pronominal subject is present, the marker not only agrees with the subject argument but rather realises it directly. Only if a better candidate for subjecthood is present, will this be analysed as the subject, and the same verbal marker is then seen to function as an agreement affix: It is generally assumed that, with the exception of French, Romance languages are socalled ‘pro-drop’-languages. As illustrated under (1) in a Spanish example, they are capable of realising the subject-argument through the morphological categories of Person and Number on the verb – categories which function as agreement markers as soon as a nominal phrase (NP) appears in the position of the subject. (1) a. Maríai cant-ói [agreement] b. Cant-ó [subject realised by means of a verbal suffix] (translated from Kailuweit 2005:1).

This is intuitively appealing on the one hand, because it no longer leaves cant-ó (sing-3SG.PST.IND) without a subject as if it were ‘incomplete’ in any respect; on the other hand it is also problematic because it is unconvincing to treat one and the same affix alternatingly as an argument or as an agreement marker, depending on the presence or absence of a more suitable alternative. This solution has all the air of stopping short halfway. Once we have conceded that Spanish verbal affixes actually realise their subject arguments in some cases, we might then also consider the possibility that Spanish has come to be even more head-marking than that. Like all Romance languages, Spanish has a fully developed subject co-reference morphology on the verb and freely elidable lexical or pronominal subjects,13 a feature that has usually been described as pro-drop or null-subject. This, of course, only makes sense, if one conceives of Spanish subject co-reference in the same terms as of German subject agreement. If, however, the situation in Spanish were to be more analogous to Basque than to German, its subject-markers would have to be seen as an 13

French subject clitics which have traditionally been seen as clitical pronouns are nowadays analysed as agreement affixes by most linguists. In this case, French would be a null subject language, just like the rest of her Romance sister languages (cf. Kaiser 1992:115 et passim).

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

201

entirely different breed of affixes, namely obligatory cross-reference affixes which serve to express the subject argument. Du Bois’ concept of ‘mention’ almost imposes itself here, as the facts of this aspect of Spanish syntax resemble Sacapultec (and Basque) considerably more than those Indo-European languages on which the traditional analysis is based. Spanish is therefore on the one hand similar to German in that it is not a null-subject language; both languages require overt subjects on the surface. The radical difference, on the other hand, lies in the morphological technique of subject realisation which is analytic in German and (poly-) synthetic in Spanish (and in the Romance languages in general). As would be expected for headmarking languages, lexical and pronominal ‘subjects’ are not integrated into the sentence core in the Romance languages and can either appear on the periphery (and only there!) or be elided altogether: (19) (20) (21)

a ese no se=lo=voyi a dar [, yoi] (Spanish) [yoi,] I to that_one not him=it=go.1SG.PRS.IND to give I [moii,] je=ne=vaisi pas le=lui=donner [, moii] (French) it=him=give I I I=not= go.1SG.PRS.IND NEG [*ich,] dem gebe [ich] es nicht [*,ich] (German) I it NEG I I to_that_one give.1SG.PRS.IND “I’m (certainly) not giving it to him!”

While subjects are realised through a head-marking construction in all Romance languages, the situation with objects is far less uniform across Romance. The general picture is complex and varies from language to language; they all seem to have in common, though, that at least certain objects are realised via head-marking on the verb. Spanish (and Catalan) are probably the two most advanced languages on the way towards clausal head-marking and I shall therefore use Spanish to illustrate the point (cf. Figure 7). That table is admittedly an extreme idealisation of the actual situation and is only intended to represent the gross overall picture. Thus, the only dependent-marking solution in the example is only possible as long as the direct object-noun phrase remains in its canonical position. Once it is raised into a left-dislocated topic position, the marker on the verb becomes obligatory: (22) a. Juan conoc-e a María. John know-s to Mary “John knows Mary.” b. A María, Juan la=conoc-e. to Mary John her.ACC=know-s “As for Mary, John knows her.”

HANS-INGO RADATZ

direct objects

202 Head-Marking

Dependent-Marking

NP

( )

* Juan la=conoc-e a María. John her.ACC=know-s to Mary “John knows Mary.”

Juan conoc-e a María. John know-s to Mary “John knows Mary.”

Pron

Juan la=conoc-e a ella. John her.ACC=know-s to she “John knows her.”

*Juan conoc-e a ella. John know-s to she “John knows her.”

indirect objects

Juan la=conoc-e. John her.ACC=know-s “John knows her.” NP

Juan le=habl-a a María. John her.DAT=speak-s to Mary “John is talking to Mary.”

*Juan habl-a a María. John speak-s to Mary “John is talking to Mary.”

Pron

Juan le=habl-a a ella. John her.DAT=speak-s to she “John is talking to her.”

*Juan habl-a a ella. John speak-s to she “John is talking to her.”

Juan le=habla. John her.DAT=speak-s “John is talking to her.” Figure 7: Head-marking vs. dependent-marking in Spanish objects

The tendency in Spanish is clearly towards an increase of clausal headmarking and some varieties are already much further advanced in this than others. Kaiser (1992) reports: In some varieties of Spanish, in particular those from the Rio de la Plata region in Argentina and Uruguay, the object clitic seems to be possible and even obligatory also in those cases where the co-referent direct object NP carries the semantic feature [+animate] and is therefore preceded by the preposition a (translated from Kaiser 1992:57).

In these dialects, Juan la conoce a María is completely grammaticised and we get the full head-marking situation!14 While it is already a noteworthy fact that Romance languages should exhibit clausal head-marking traits at all, the general typological importance of the shift depends heavily on the degree to which each language actually makes use of them. In an empirical study on the related topic of Romance object 14

Van Valin confirms this and acknowledges the existence of Latin American varieties of Spanish in which clitic-doubling has come to be obligatory (personal communication, 2002). Jaeggli (1982:14) reports that, in River Plate Spanish, clitic doubling is impossible with a nonpronominal, inanimate d.o., preferred with animate, specific d.o., highly preferred with a nonpronominal goal i.o., and obligatory in all other cases.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

203

conjugation, Koch (1993, 1994) has undertaken the task of evaluating the statistic relevance of ‘clitic-doubling’ in French, Italian, and Spanish, the main results of which I have conflated into the synoptic table in Fig. 8. It shows a clear ranking, with French being most reluctant to adopt head-marking, Italian as a case in which both strategies seem to coexist on all levels, and Spanish with a clear-cut – and in the case of object pronouns already consummated – tendency towards clausal head-marking. Further evidence in favour of a head-marking analysis of Romance objects comes from the classical examples of segmented sentences which have been particularly popular in the context of the debate over object conjugation in Romance and have quite justly (and following Queneau) led Koch (1993) to a comparison with Chinook and other Amerindian languages: (23) a. [Il=la=lui=a=donnée], à Jean, son père he.NOM=her.ACC=him.DAT=has given to John his father sa moto (Tesnière 1959:175). POSS motorcycle.F.SG “His father gave John his motorcycle.” b. Io in quella casa i piedi [non ce=li=metterò mai] (cf. Koch 1993:172).15 I in that house the feet NEG there=them=put.1SG.FUT never “I shall never set foot in that house.”

In languages without morphological case, the fundamental syntactic relations have to be encoded sequentially, with the ensuing consequence that the order of elements in the sentence becomes increasingly fixed and can no longer be used for discourse-pragmatic functions. While French exhibits a vast array of syntactic constructions to achieve this, of which only some exploit headmarking, these clearly seem to be the preferred structures in Spanish and Catalan. They provide a perfect solution to the discourse pragmatic problem of underlining the informational profile of an utterance by detaching lexical from grammatical information. Syntax proper is condensed into the morphosyntax of the clitic group, and the noun phrases, which carry the lexical information, are set free from the sequential restrictions of the core sentence and can now be used in the sentence periphery, where they can be ordered in any way the distribution of old and new information may require. Even if French shows certain head-marking tendencies as well, they have never come to be fully grammaticised as in Spanish.

15 These traditional examples are obviously and admittedly fabricated by linguists and rather unlikely to ever appear in spontaneous discourse, as they violate the Preferred Argument Structure constraint “Avoid more than one lexical core argument” (Du Bois 2003:34ff.). They are nevertheless grammatically possible utterances in these languages and illustrate their capability of representing all core arguments as affixes on the verb while converting the corresponding NPs into appositions outside the sentence core.

HANS-INGO RADATZ

204

realised as an NP realised as a Pronoun realised as an NP Realised as a Pronoun

I = Indirect object

O = Direct object

French

Italian

VO = object follows verb; no clitic OV = object precedes verb; no clitic oVO = object follows verb; pleonastic clitic precedes verb

92,0

85,5

1,7

1,8

4,0

4,4

OoV = object precedes verb; pleonastic clitic precedes verb VO = object follows verb; no clitic

3,0

6,4

OV = object precedes verb; no clitic oVO = object follows verb; pleonastic clitic precedes verb OoV = object precedes verb; pleonastic clitic precedes verb VI = object follows verb; no clitic IV = object precedes verb; no clitic iVI = object follows verb; pleonastic clitic precedes verb IiV = object precedes verb; pleonastic clitic precedes verb VI = object follows verb; no clitic IV = object precedes verb; no clitic iVI = object follows verb; pleonastic clitic precedes verb IiV = object precedes verb; pleonastic clitic precedes verb

[no data]

Spanish [–hum] [+hum] [–hum] [+hum] [–hum] [+hum] [–hum] [+hum]

25,0

0,0

4,2

0,0

95,0 89,5 0,7 0,0 0,8 7,9 3,6 2,6

25,0 100,0 46,0 95,0

70,1

40,0

0,0

2,4

0,0

0,0

14,0

41,8

5,0

8,7

18,2

5,0

13,7

0,0

0,0

15,7

0,0

23,5 100,0

100,0 45,0

All numbers are percentages. For details about the corpus used cf. Koch (1993:173). O = direct object, o = direct object clitic; I = indirect object, i = indirect object clitic; V = verb; Pron = emphatic, free pronoun; [+/- hum] = object with a human vs. non-human referent. Figure 8: Percentage of pleonastic object clitics in Romance languages (data from Koch 1993)

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

205

This is mainly due to the early development of grammaticalised cleftconstructions in French which constituted an important alternative way of expressing theme-rheme structures, and while clefting never really caught on in Spanish, it eventually won out over clausal head-marking in French. The intermediate position of Italian as seen in the statistics of Fig. 8 is most probably due to the fact that clefting is more important in Italian than in Spanish but at the same time less important than in French. 6. Language change: intermediary stages of a change from dependentmarking towards head-marking The last 1000 years have seen Spanish pronouns change from free forms similar to German pronouns towards increasingly bound forms. In a convincing article from 1990, Rini has tried to reconstruct and date the process in which Old Spanish clitic pronouns developed into the grammaticalised verbal affixes of Modern Spanish (Rini 1990). He does this by analysing the history of three related syntactic phenomena, namely 1. Interpolation, 2. Word order, and 3. Pronominal duplication. Old Spanish clitics, though already phonologically dependent, had not yet developed into bound morphemes (cf. Rivero 1986:775). Their position with respect to their host was not yet as fixed as it is in Modern Spanish (e.g. él díxolo, él lo dixo, lo él dixo “he said it” all can be found in the same document) and the so-called ‘interpolation’ of material between the clitic and its host was quite frequent: (24)

quien te algo prometiere (Corbacho 145) who pron.2SG.DAT something should_promise “whoever should promise you something” para lo mejor conplir que lo ella non mando (Corbacho 91)16 for it better accomplish that it her not order.1SG.PRS.IND “in order to accomplish it better that I shall not order her to do it

While clitic-doubling did already occur in Old Spanish, it was far from being as frequent as today and it was never obligatory. Thus, clitic-doubled examples like: (25)

E tu as me tollido a mí un capellano (Berceo, 229d)17 and you.SG have me taken from me one priest “And you have taken away a priest from me.”

16 Arcipreste de Talavera o Corbacho, ed. Gerli, Michael, Madrid: Cátedra, 1979, quoted from Rini (1990:357). 17 Milagros de Nuestra Señora, ed. García Solalinde, Antonio, Madrid: Espasa-Calpe, 1922, quoted from Rini (1990:361).

206

HANS-INGO RADATZ

occur freely alongside non-doubled structures in the same text: (26)

A ti lo digo, nuera (Corbacho ch. 38, p. 144)18 to you.SG it I_say daughter-in-law “I say it to you, daughter-in-law.”

But even where clitic-doubling did occur, as in (25), the construction is syntactically quite different from the corresponding structures in the modern language: In Modern Spanish [... t]he clitic is indispensable, which suggests that it is the basic pronoun and the tonic is the additional or redundant form, which appears for emphasis. There is no reason to believe, however, that in Old Spanish the emphatic constructions were of this same nature as regards the question of which was the basic and which was the redundant or additional element. In fact, the evidence provided by a diachronic perspective seems to suggest just the opposite: that the tonic pronoun, not the clitic, was the basic element of the emphatic constructions, and the clitic was additional or redundant when the two forms occurred together (Rini 1990:360).

What Rini found out was that, with respect to all three aforementioned parameters, Spanish seems to have reached the modern type of distribution at roughly the same time for each one of them, i.e. around the beginning of the 17th century.19 By then, no further examples of interpolation appear in the texts, pronominal clitics essentially become fixed to the left of finite and to the right of non-finite verbal forms and clitic-doubling becomes obligatory with emphatic pronouns. The fact that all three changes reach their conclusion more or less simultaneously is a strong indication that they have to be seen as three facets of one and the same grammaticalisation process which at this time must have reached a new stage: the stage of affixes. Most interesting, however, is Rini’s observation regarding clitic-doubling. While in Old Spanish clitic-doubling the free pronouns acted as indispensable head of the construction and the clitics were just a redundant reinforcement, the respective roles can be shown to have changed completely in what Rini aptly dubs a ‘cephalic shift’ (Rini 1990:361). From the beginning of the 17th century on it is no longer the free pronouns but rather the clitic affixes that constitute the head of the construction while free pronouns may but need not be added for emphasis. Put in the terms of Bybee (2005), the high token frequency of doubling clitics leads to an increased entrenchment of object clitics in the morphosyntactic structure and ultimately “to the grammaticization of the new 18

Quoted from Rini (1990:360). Other authors date the point for completed grammaticalisation much earlier; thus, for Barry, this state is reached from the thirteenth century onwards (cf. e.g. Barry 1987:219). For further empirical detail cf. Eberenz (2000:175-208). 19

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

207

construction and the creation of grammatical morphemes and changes in constituency” (Bybee 2005:6). Object-clitics were first phonologically reduced in that they lost their capability of being stressed; from now on, stress could only be placed on the accompanying free pronoun or full NP. As verbs and clitics were used together ever more frequently, they came to be perceived as a unit in which the clitics were reinterpreted as obligatory grammatical morphemes. Because free object pronouns lost their obligatory status at about the same time, the new verbal markers began to function as cross-reference morphemes and were now treated as the actual realisations of objects. In modern Spanish, (1) clitic doubling has become obligatory with any (emphatic) object pronoun, (2) is regularly used with indirect object NPs, and sometimes occurs even (3) with direct object NPs (cf. Fig. 8). In all these cases, the verbal affixes co-occur (and morphologically cross-reference with) their respective lexical or pronominal appositive phrases. With nominal objects the extent and obligatoriness of clitic-doubling seems to be in constant flux, and any researcher adducing examples must provide exact information as to which dialect or variety of Spanish is envisaged – all clear indications of a still ongoing process of change. The cephalic shift created a new type of morphology in Spanish which would best be described as ‘cross-reference’. Obviously, there remain large areas in Spanish syntax where cross-reference is not (yet) obligatory or even possible. On the other hand, there are large areas (mainly with emphatic object pronouns) where it has already been fully grammaticalised. While Spanish may still be far from Basque in this respect, it is nevertheless clearly moving in the same direction. Traditionally, grammaticalisation theory envisages the typical development of free (emphatic) pronouns as ‘emphatic personal pronoun > clitic pronoun > agreement affix’ (cf. e.g. Croft 2000:157). In the case of Spanish we can see that this is at least not the only possible development and that we are again faced with the type of shortsighted exclusion of languages with (clausal) headmarking as deplored by Nichols (1986:115f.; see quote above). While this type of grammaticalisation may lead to the establishment of a new type of agreement affixes, it may also, alternatively, create cross-reference affixes thus turning a language into one with clausal head-marking. For centuries, Spanish speakers have chosen head-marking strategies in order to resolve discoursepragmatic problems of information structure and these preferred but still optional patterns are now turning into structural patterns of grammar in just the way Haspelmath (1999:193) envisages it. As “grammars code best what speakers do most” (Du Bois 1987:811), Spanish is accepting ever more clausal head-marking as fixed, obligatory constructions into its syntax and we may be witnessing the emergence of a more and more consistent head-marking syntax. What has frequently been described as a change in which clitic pronouns are grammaticalised into

HANS-INGO RADATZ

208

agreement-morphemes must then be seen as an even more radical change, not into agreement morphemes but into cross-reference affixes. The difference between French and Spanish in this respect is more of degree than of principle, as all Romance languages possess at least some of these head-marking constructions. Clitic-doubled verbal complexes like il la lui a donnée (or sp. se lo ha dado, cat. la hi ha donat etc., all meaning “he has given it to him”) are indeed quite analogous to the corresponding structures in Basque: (27)

Bere

aita-ki Jon-ik bere father-DET.ERG.SG Jon-DAT.SG POSS.3SG moto-aj eman dj-i-ok-Øi . motorcycle-DET.ABS.SG give ABS.3SG-AUXhave.PRS.IND-DAT.3SG-ERG.3SG. “His father gave John his motor-bike.” POSS.3SG

The following examples show the possibilities for expressing differences in functional sentence perspective and the similarity of the French and the Basque solutions: (28) a. b.

(29) a. b.

(30) a.

Il=la=lui=a donnée, à Jean, son père, sa he=it=him=has given to John his father his Eman dio Joni bere aitak bere given he_has_it_to_him John his father his “His father gave John his motor-bike.” À Jean, il=la=lui=a donnée, son to John he=it=him=has given his Joni eman dio bere to_John given he_has_it_to_him his “His father gave John his motor-bike.”

Sa moto, il=la=lui=a donnée, à his bike he=it=him=has given to b. Bere motoa eman dio his bike given he_has_it_to_him “His father gave John his motor-bike.”

père, father aitak father

moto. bike motoa. bike

sa moto. his bike bere motoa. his bike

Jean, son père. John his father Joni bere aitak. to_John his father

The main difference here lies in the fact that this cross-reference marking with appositional complements is the obligatory normal case in Basque while it is a stylistically marked exceptional construction, the appearance of which is bound to a discourse-pragmatic motivation. The French examples are all instances of segmented phrases, a discourse pragmatic repair strategy which justifies the apparition of the object clitics in French. With all lexical arguments in their conventional slots, clitic doubling becomes unacceptable in French:

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

(31) a.

Son père [*il=la=lui=] a donné sa his father he=it=him= has given his b. Bere aitak eman dio bere his father given he_has_it_to_him his “His father gave John his motor-bike.”

209

moto à Jean. bike to John motoa Joni. bike to_John

In Spanish, on the other hand, the corresponding sentence may not be the preferred one, but is at least grammatically possible: c.

Su padre se=la=ha his father him=it=has

dado la moto a Juan. given the bike to John

On this basis we can tentatively formulate a typological hierarchy, where the Romance languages can be situated at an intermediate level:

Dependentmarking

Split marking

Double marking

Head-marking

Type

example

Syntactic consequences

Abkhaz

– all arguments realised polysynthetically on the finite verb; – NPs are facultative appositions without any case marking; – no deictically unmarked object-pronouns (if at all, only demonstratives).

Basque

– all arguments realised polysynthetically on the finite verb (headmarking); – NPs are facultative appositions with case marking (additional, redundant dependent-marking); – no deictically unmarked object-pronouns (if at all, only demonstratives).

Spanish, Catalan, European Portuguese

mixed; – Subject head-marking on the verb with appositional subject-NPs or -pronouns respectively; Objects are realised 1. as affixes on the verb, 2. as affixes + apposition and 3. as NPs (but not as pronouns!)

German

– all arguments realised as NPs pronouns with case marking; – only head-marking feature is the redundant subject agreement on the verb.

Figure 9: Head- vs. dependent-marking-hierarchy

210

HANS-INGO RADATZ

7. Conclusion Syntax, morphology, and phonology are usually so inextricably intertwined in diachrony that trying to establish exactly where a change originated may come down to the classical hen-egg problem. As Romance languages lost their case inflections (morphological phenomenon induced by phonological processes), syntactic relationships increasingly came to be encoded in the sequencing of constituents (syntax). As this sequencing became ever more rigid, new focusing techniques had to arise, as constituents could no longer be arranged freely to the discourse needs of language users (loss of word stress by the 16th century may have aggravated this problem in the case of French). The new techniques have in common that they place a pronominal element (relative or personal pronoun / clitic) in its canonical place such as to leave sequencing unaltered in the sentence core. The corresponding lexical NPs are thus freed to be placed in the periphery where they may be focused according to the conversational needs. The main solutions along these lines are clefts and pronominal reprise (or clitic-doubling), and while French and Spanish both use clefting as well as pronominal reprise techniques, the former are the typical solution in French, whereas Spanish largely depends on the latter. As Dufter (this volume) has shown, the success of c’est-clefts in French let this constructions become so entrenched that their initial motivation as a focusing device didn’t remain their only function; with the advent of the ‘informative-presupposition cleft sentence’ in the 16th century, French clefts could now also carry given or thematic information. What Dufter calls clefting ‘beyond necessity’ can analogously also be observed for Spanish cliticconstructions, the history of which shows a continuous increase of uses that clearly go ‘beyond necessity’, and what once started out as individual speakers’ discourse strategies has, through the centuries, lost its dependence on the original motivations and has now been integrated into the very syntax of the simple sentence in Spanish. The typological drift from a conventional central IE language with agreement towards the ‘exotic’ type of clausal head-marking with cross-reference morphology can thus be argued to derive ultimately from new discourse strategies as a reaction to a syntactic change, namely the progressive fixation of word order towards rigid SVO (even if Langobardi’s Inertia Theory would have us expect otherwise). The cross-reference analysis of Spanish clitics is not a mere terminological reformulation of the traditional analyses. The corresponding language type has long been well established and well described, although it has, until recently, received very little attention within IE linguistics and Romance philology in particular. If it can be shown that Spanish syntax is in a central aspect moving away from its IE sister languages like French, English, or German to become more like Basque, Abkhaz, or Lakhota, this is in itself an important typological insight. For the syntactic analysis of Spanish, the cross-reference hypothesis implies that, in many cases, lexical NPs as well as pronouns must no longer be

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

211

seen as the subject and object arguments, but rather as appositional phrases outside the sentence-core forming ‘mentions’ together with cross-reference markers on the verb. These mentions are discontinuous realisations of arguments in which the verbal markers are obligatory (and therefore the heads), while the lexical NPs or pronouns are syntactically facultative. Spanish does therefore neither tolerate null-subjects nor null-objects. A further consequence of this typological drift is that we can expect clitic-doubling to become ever more obligatory and to be carried over to further new domains. 8. Closing remarks: no “object-conjugation” in Romance (or elsewhere)! The phenomena discussed here are widely identical with those that have been invoked in the debate among Romance linguists over a supposed tendency towards ‘object-conjugation’ in these languages.20 It should be clear from the above, however, that the notion of object-conjugation is misleading here as it evokes the notion of conjugation and with it the notion of agreement. As we have seen, agreement proper is only to be found in pure dependent-marking languages whereas the phenomena in question represent a typological shift precisely in the opposite direction, namely towards head-marking. While allusions to Basque and Hungarian are present in all contributions to the debate, the syntactic facts of these languages are consistently presented within a terminological framework developed for the description of dependent-marking Indo-European languages, treating Basque and Hungarian cross-reference elements as agreement-markers. This type of ‘agreement’ is then projected onto the Romance material. We must therefore reject the notion of ‘objectconjugation’ altogether, not only for the Romance phenomena, but also for the languages in which it purportedly is most fully developed, namely Basque and Hungarian. By failing to differentiate between agreement and cross-reference, the theory of Romance object-conjugation continues to look for the arguments where they are not to be found – in the noun phrases and strong pronouns – and perpetuates a strict distinction between a dependent agreement-marker and its nominal referent. In so doing they miss the very nature of this ongoing process of change: the emergence of cross-reference structures and clausal headmarking.

REFERENCES Ashby, William J. 1977. Clitic inflection in French. Amsterdam: Rodopi.

20 Cf. de la Grasserie (1889:278f.), Lenz (1920:81-3), Heger (1966), Rothe (1966), Llorente &

Mondéjar (1974), Ashby (1977:73-78), Bossong (1980, 1982), Berretta (1989), Schwegler (1990:101-3), and Koch (1993, 1994).

212

HANS-INGO RADATZ

Aoun, Joseph. 1993. “The syntax of clitic-doubled arguments”. International Journal of Basque Linguistics and Philology 27. 709-730. Barry, Anita K. 1987. “Clitic pronoun position in thirtheenth-century Spanish”. Hispanic Review 55. 213-220. Beaumont, Jean-Charles & Ramón Lazkano. 1998. Iniciación al euskara – Euskara baturako hastapenak. Chennevières-sur-Marne: Assimil. Berretta, Monica. 1989. “Tracce di coniugazione oggettiva in italiano”. L'italiano tra le lingue romanze, ed. by F. Foreti et al., 125-150. Roma: Bulzoni. Bloomfield, Leonard. 1933. Language, Chicago: University of Chicago Press. Bossong, Georg. 1980. “Aktantenfunktionen im romanischen Verbalsystem”. Zeitschrift für romanische Philologie 96. 1-22. ----------. 1982. “Historische Sprachwissenschaft und empirische Universalienforschung”. Romanistisches Jahrbuch 33. 17-51. ----------. 1991. “Differential object marking in Romance and beyond”. New Analyses in Romance Linguistics, ed. by Dieter Wanner & Douglas A. Kibbee, 143-170. Amsterdam & Philadelphia: John Benjamins. Bybee, Joan. 2005. “The Impact of Use on Representation: Grammar Is Usage and Usage Is Grammar”. LSA 2005 Presidential Address. Online at http://www.unm.edu/~jbybee/Bybee plenary.PDF. Corbett, Greville G. 1994. “Agreement”. Encyclopedia of Language and Linguistics Vol. 1, 54-60. Oxford: Pergamon Press. Croft, William. 2000. Explaining Language Change. London & New York: Longman. Grasserie, Raoul de la. 1889. “De la conjugaison objective”. MSLP 5. 268-300. Du Bois, John W. 1987. “The discourse basis of ergativity”. Language 63. 805855. ----------. 2003. “Argument structure: Grammar in use”. Preferred Argument Structure, ed. by John W. Du Bois, Lorraine E. Kumpf & William J. Ashby, 11-60. Amsterdam & Philadelphia: John Benjamins. Eberenz, Rolf. 2000. “El pronombre personal átono con función duplicadora”. El español en el otoño de la edad media, ed. by Rolf Eberenz, 175-208. Madrid: Gredos. Franco, Jon. 1991. “Spanish object clitics as verbal agreement morphemes”. MIT Working Papers in Linguistics 14. 99-114. ----------. 2000. “Agreement as a Continuum”. Clitic Phenomena in European Languages, ed. by Frits Beukema & Marcel den Dikken, 147-189. Amsterdam & Philadelphia: John Benjamins. Goldbach, Maria. 1999. Spezifische und arbiträre leere Objekte. Frankfurt am Main: Vervuert. Gómez, Ricardo & Koldo Sainz. 1995. “On the Origin of the Finite Forms of the Basque Verb”. Towards a History of the Basque Language, ed. by José Ignacio Hualde, Joseba A. Lakarra & Robert Lawrence Trask, 235-274. Amsterdam & Philadelphia: John Benjamins.

NON-LEXICAL CORE-ARGUMENTS IN BASQUE AND ROMANCE

213

Haspelmath, Martin. 1999. “Optimality and diachronic adaptation”. Zeitschrift für Sprachwissenschaft 18. 180-205. Heger, Klaus. 1966. “La conjugaison objective en français et en espagnol”. Langages 3. 19-39. Helmbrecht, Johannes. 2001. “Head-marking vs. dependent-marking languages”. Language Typology and Language Universals, ed. by Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible, vol. 2, 1424-1432. Berlin & New York: Walter de Gruyter. Jaeggli, Osvaldo. 1982. Topics in Romance Syntax. Dordrecht & Cinnaminson: Foris. Kailuweit, Rolf. 2005. “‘Pro-drop’, Kongruenz und ‘optimale Klitika’ – Ein Beschreibungsansatz im Rahmen der Role and Reference Grammar”. Morphologie und romanistische Sprachwissenschaft, ed. by Carmen Kelling, Judith Meinschaefer & Katrin Mutz, 87-110. [http://www.ub.unikonstanz.de/kops/volltexte/2006/1812/pdf/AP_120.pdf]. Kaiser, Georg. 1992. Die klitischen Personalpronomina im Französischen und Portugiesischen. Eine synchronische und diachronische Analyse. Frankfurt am Main: Vervuert. Koch, Peter. 1993. “Le ‘chinook’ roman face à l’empirie. Y a-t-il une conjugaison objective en français, en italien et en espagnol et une conjugaison subjective prédéterminante en français?”. Actes du XXe Congrès International de Linguistique et Philologie Romanes, Université de Zurich (6-11 avril 1992), ed. by Gerold Hilty, vol. 3, 169-190. Tübingen & Basel: Francke. ----------. 1994. “L'italiano va verso una coniugazione oggettiva?”. Sprachprognostik und das ‘italiano di domani’. Prospettive per una linguistica ‘prognostica’, ed. by Günter Holtus & Edgar Radtke, 175-194. Tübingen: Narr. Landa, Alazne & Jon Franco. 1996. “Two issues in null objects in Basque Spanish”. Grammatical Theory and Romance Languages, ed. by Karen Zagona, 159-168. Amsterdam & Philadelphia: John Benjamins. Lenz, Rodolfo. 1920. La oración y sus partes. Estudios de gramática general y castellana. Madrid: Revista de Filología Española. Llorente Maldonado de Guevara, Antonio & Mondéjar, José. 1974. “La conjugación objetiva en español”. Revista Española de Lingüística 4. 1-60. Nichols, Johanna. 1986. “Head-marking and dependent-marking grammar”. Language 62. 56-119. ----------. 1992. Linguistic Diversity in Space and Time, Chicago & London: The University of Chicago Press. Rini, Joel. 1990. “Dating the grammaticalization of the Spanish clitic pronoun”. Zeitschrift für Romanische Philologie 106. 354-370. Rivero, María Luisa. 1986. “Parameters in the typology of clitics in Romance, and Old Spanish”. Language 62. 774-807.

214

HANS-INGO RADATZ

Rothe, Wolfgang. 1966. “Romanische Objektkonjugation”. Romanische Forschungen 78. 530-547. Saltarelli, Mario. 1988. Basque, London: Croom Helm. Schwegler, Armin. 1990. Analyticity and Syntheticity. A diachronic perspective with special reference to Romance languages. Berlin & New York: Mouton de Gruyter. Schwerteck, Hans. 1984. Strukturen baskischer Verbformen. Wiesbaden: Franz Steiner. Tesnière, Lucien. 1959. Éléments de syntaxe structurale. Paris: Éditions Klincksieck. Van Valin, Robert D. Jr. 2005. A summary of Role and reference Grammar. [http://linguistics.buffalo.edu/research/rrg/RGsummary.pdf]. Zubiri, Ilari. 52000. Gramática didáctica del euskera. Bilbo: Didaktiker.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE THREE RECENT EVOLUTIONARY APPROACHES*

ESME WINTER-FROEMEL University of Tübingen

This article investigates whether evolutionary accounts can offer new insights into the paradox of language change. Specifically, I will examine three recent influential accounts (Haspelmath 1999, Keller 1994, and Croft 2000). As they contain a broad spectrum of positions on the relations between language and biology, they can be divided into metaphorical, biologistic and generalized views. Cross-cutting these, two types of evolutionary accounts are distinguished, which I call adaptive and two-level views, respectively. I critically evaluate their potential to provide satisfactory explanations for various types of change, drawing on examples from Romance and Germanic. Finally, I propose a revised explanation scheme which brings together the two-level approaches with theoretical distinctions and explanatory factors that have been suggested in earlier non-evolutionary frameworks, so that a more comprehensive view of language change can be obtained.

1. Introduction The paradox of language change (Coseriu 1958:7) can be seen as a fundamental problem which remains essentially unsolved (see the editors’ introduction to this volume). As the theory of biological evolution has highlighted a very similar problem, and given the enormous success of evolutionary theory in other sciences and disciplines (see already Dohrn 1876:120), it is not surprising that various recent approaches to language change refer to biological and evolutionary concepts. What is striking, however, is the fact that evolutionary positions are adopted in both formalist and functionalist frameworks, which makes this kind of approach even more interesting. A general theory of language change would be of great current concern, since many current accounts only focus on specific aspects, such as language universals, systemic linguistics, contact linguistics, pragmatic and sociolinguistic factors * Earlier versions of this paper were presented at the Sprachwissenschaftliches Kolloquium,

Tübingen, June 2005, and the Deutscher Romanistentag, Section VI: Grammatischer Wandel und seine Erklärbarkeit, Saarbrücken, September 2005. I would like to express my gratitude to all participants of both conferences, especially to Richard Waltereit, Ulrich Detges and Peter Koch, for their helpful comments. In addition I would like to thank Sam Featherston for valuable comments and linguistic revision of the manuscript.

216

ESME WINTER-FROEMEL

(Kabatek 2005:155), which each imply specific approaches, so that our understanding of language change remains somewhat fragmented and on the whole unsatisfactory. Therefore I want to examine three recent proposals to interpret language change in evolutionary terms and evaluate their possible contribution to a better understanding of language change: the programmatic paper of Haspelmath (1999) and the works of Keller (1994) and Croft (2000).1 I will start by outlining the assumptions about the relations of language and biology that different evolutionary positions make, and show that there is a broad range of possible positions, which imply different understandings of the term ‘evolution’ proper. Second, evolutionary approaches will be divided into two groups according to their main orientations and key notions. ‘Adaptive views’ and ‘two-level views’ will be discussed in turn, taking into account different phenomena of language change. Then I will take a closer look at some key notions of two-level views in order to determine to what extent these concepts offer new insights into language change. Building on these approaches, I will propose a revised explication scheme which brings together the evolutionary frameworks with a series of more traditional linguistic theories and distinctions. Finally, I will critically review the potential of evolutionary arguments to explain language change. 2. Language and biology: Three positions Evolutionary explanations are often categorically rejected on the grounds that the presumed analogy between biology and linguistics is considered as illegitimate and misguided. Opponents of evolutionary arguments frequently underline that it is fundamentally mistaken to view the evolution of cultural objects as instantiations of processes of biological evolution. Yet such a biologistic view is equally rejected in some evolutionary frameworks, and much weaker positions are adopted. On the whole, we can broadly distinguish three possible attitudes towards the relation of language and biology, namely metaphorical, biologistic and generalized views (see Fig. 1; compare Croft 2000:10-12, who terms them as ‘analogical’, ‘literal’ and ‘generalized’ approaches).

1 While Haspelmath’s (1999) is an article that has been peer-reviewed in the Zeitschrift für Sprachwissenschaft (among others, by Croft), Keller (1994) and Croft (2000) have presented monographs on the subject. I therefore acknowledge that the three approaches are not fully comparable.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

metaphorical views language ~ biology

-

-

The evolution of language is to some extent similar to biological evolution.

Language change is explained by a genuinely cultural model. Keller Haspelmath?

biologistic views biological evolution ↘ evolution of language - The evolution of language is an instantiation of biological evolution. -

The mechanisms of biological evolution also determine the evolution of language. Haspelmath? Newmeyer Givón

217

generalized views evolutionary models ↙ ↘ language biology - Both linguistic and biological systems are subcategories of a general type of historically evolved systems. - There are specific mechanisms of evolution in biology and language. Croft

Fig. 1: Language and biology in evolutionary approaches: Three positions

In the first group of evolutionary accounts, the reference to biological evolution represents a mere metaphor or analogical speaking used for heuristic reasons. Positions that adopt such a metaphorical view explicitly reject the attempt to view the evolution of language in biological terms and claim a genuinely cultural model to explain language change:2 My attempt to conceive the development of language as an evolutionary process is not at all determined by the endeavour to tr ansfer a model of natural sciences to an object of cultural studies. My attempt is rather carried by the endeavour to apply a model which is genuinely from cultural studies, that of the invisible hand, to the consideration of language. […] The theory of the evolution of the living world shall serve us only as a heuristic model. (Mein Versuch, die sprachliche Entwicklung als evolutionären Prozeß zu begreifen, ist überhaupt nicht geprägt von dem Bestreben, ein naturwissenschaftliches Modell auf einen kulturwissenschaftlichen Gegenstand zu üb er tr agen. Mein Versuch ist vielmehr von dem Bestreben getragen, ein genuin kulturwissenschaftliches Modell, das der unsichtbaren Hand, auf die Sprachbetrachtung anzuwenden. […] Die Theorie der Evolution der belebten Natur soll uns nur als heuristisches Modell dienen […].) (Keller 1994:195, emphasis original)3

This metaphorical understanding of ‘evolution’ is in principle relatively unproblematic: the use of models and theories of other scientific domains as 2 The notion of ‘invisible-hand processes’ is in fact introduced by Adam Smith in the context of

his theory of trade. The concept is then taken up by Malthus, some of whose ideas are in turn transferred to evolutionary biology by Darwin (see also Wolgast 1984, who detects some flaws in Darwin’s reference to Malthus). 3 All translations of passages cited in this article are mine, EWF.

218

ESME WINTER-FROEMEL

heuristic devices may in some cases inspire new views of the phenomena at hand. The usefulness of a metaphor for a better understanding of a specific phenomenon, however, entirely depends on whether the assumed analogy is correct or not. In a nutshell, the biological metaphor does not tell us anything about language change unless both the similarities and the dissimilarities between linguistic and biological evolution are made explicit. Furthermore, if evolutionary approaches do not state to what extent the concept of biological evolution is being seen as applicable to the evolution of language, confusions may arise, as the positions may be (mis)interpreted in a much more radical, biologistic way, i. e. as postulating that language change has to be explained by the same mechanisms as biological evolution. Yet certain analogies can be shown to be problematic. Grounded on the basic idea that “(d)iachronic adaptation in language is in many ways analogous to adaptation in biological change” (Haspelmath 1999:204), Haspelmath proposes the following mechanism of evolution:4 “As in biology, observed adaptive patterns in language can be explained through diachronic evolutionary processes, as the unintended cumulative outcome of numerous individual intentional actions.” (Haspelmath 1999, Abstract). This analogy has been strongly contested, and Itkonen (1999:220) detects here a crucial divergence between linguistic and biological developments. While individual intentional actions are certainly relevant to linguistic change, they do not play a role in biological evolution: An analogy between genes and humans could be understood in one of two ways: either genes behave like humans, i.e. they perform abductions and generalizations; or humans behave like genes, i.e. they lack the capacity to perform abductions and generalizations. Both options should be rejected. Why? – because they are false. (Itkonen 1999:220, emphasis original)

The comparison of language and biology is therefore often grounded on other criteria, e. g. on the appearance of design or on the importance of smallscale processes and global results, which will be studied in detail below (see 3.1 and 3.4). Another aspect is put forward by Newmeyer and Givón. They regard the evolution of language and grammar as the result of a biological process of evolution, based on natural selection and adaptation: “[…] innate autonomous grammatical principles were selected for because they allotted a greater evolutionary advantage to populations that had them” (Newmeyer 1991:6), “[…] all linguists should accept the idea that central to language there exists an autonomous grammar, shaped in part by natural selection” (Newmeyer 1991:26), “Language as a biological adaptation” (Givón 2002, Ch. 1). 4 Haspelmath follows here Keller’s formulation of invisible-hand explanations. As will be shown in the next section, however, the main orientations of their accounts differ considerably.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

219

It has to be emphasized, however, that these positions do not focus on language change (understood in the sense of changes in already existing languages), but on the origin of language tout court, which constitutes a different question that lies out of the scope of this paper and will not be examined in detail here. Nevertheless, it is important to keep this related issue in mind, as it may also emerge in theories dealing with language change proper. In addition to the two mentioned ways of defining the relation between language and biology, the metaphorical and the biologistic view, Croft (2000) claims a third possible solution which is characterized as a ‘generalized evolutionary model’:5 First, there is a general theory of selection, most fully developed in evolutionary biology, which shows that evolutionary processes operate at two levels, replication and selection (Hull 1988). Second, this model can be applied to language by taking the speaker as the unit of selection and the lingueme [= unit of linguistic structure, EWF] as the unit of replication; this is the Theory of Utterance Selection […]. (Croft 2000:229)

This third position is often not fully recognized, but will turn out to be of great significance for the concerns of this article. Most importantly, it implies a conceptual shift regarding the term ‘evolution’: whereas in both other views, ‘evolution’ means ‘biological evolution’ (referred to literally or metaphorically), here the term is understood in a new, general sense, comprising biology and language, but independent of specific mechanisms determining evolutionary processes in both domains. Together with the conceptual shift goes a change in the function of other evolutionary concepts for the study of language change: their primary function is no longer to determine specific mechanisms of change, but rather to permit a global understanding of complex historical processes. In a nutshell, besides the different assumptions about the relations between language and biology, evolutionary approaches also differ with respect to their evolutionary contents. Thereby we arrive at a second fundamental distinction between two groups of evolutionary accounts.

5 As Croft points out, a similar, general evolutionary account of conceptual change in the

history of science is given by Hull (1988:20): “I do not propose to extend a gene-based biological theory of evolution to include conceptual development in science. Instead, I provide a general analysis of selection processes which is intended to apply equally to both biological and conceptual change. Neither biological nor conceptual selection is more fundamental than the other.” For another application to the domain of linguistics, see Nettle (1998:447): “Evolutionary processes are not restricted to DNA. They occur wherever there is variation, replication and selection. They are best understood as algorithms for producing efficient design by a series of mindless steps […].”

220

ESME WINTER-FROEMEL

3. Main orientations and key elements of evolutionary accounts In addition to the distinctions made in the preceding section, I propose that we should divide evolutionary approaches into two main groups according to their general orientation of evolutionary reasoning and the key notions that they stress. In a first set of positions, the main emphasis lies on the direction and results of long-scale processes of linguistic evolution, and the ultimate aim is to explain the evolution of language or grammar as a whole. The central question is: how can we explain the long-scale evolution of languages, which exhibits an appearance of functional design? To describe and explain the rise of structures that are claimed to perfectly match the needs of language users, the theories refer to the mechanism of adaptation and the concept of optimization. In general, these approaches assume a metaphorical or a literal view about the relation of language and biology. Another group of evolutionary approaches, in contrast, focuses on the interplay of small-scale processes that entail more general changes. The main goal of these approaches is to integrate different mechanisms governing the creation and spread of linguistic innovations, and replication, variation and selection are key concepts. The central question is: how can we grasp the interplay of individual innovation and social propagation in language change? These latter approaches take a metaphorical or a generalized view about the relation of language and biology. As they distinguish two levels on which linguistic change takes place, they can be labelled ‘two-level views’ and thereby delimited from the first group of ‘adaptive views’.6 3.1. Adaptive views: Adaptation and optimization Adaptive views are taken by various authors in current research (e. g. Haspelmath 1999, Newmeyer 1991, 2001, 2003a, and Givón 2002), but as many approaches are chiefly dedicated to explaining the origin of language, the following discussion will mainly be based on Haspelmath’s paper (1999), which also deals with the explanation of language change. Haspelmath directly addresses the central question: “Why are the [grammatical, EWF] constraints the way they are?” (Haspelmath 1999:183), i. e. he seeks to explain structures of grammar by tracing them back to their origins in the usage of language. Central for achieving this goal is the concept of diachronic adaptation. Grammar is supposed to having been adapted more and more perfectly to the needs of language users in a process of diachronic adaptation (Haspelmath 1999:203-204, see Fig. 2).

6 See also Lass (1990:79), who confronts two senses of ‘evolution’: a “vulgar progressivist

sense of ‘directional change with increasing «fitness»’” (which he rejects) and the evolution of a system as “the story of its change over time, normally the product of variation and differential selection of variants”.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

221

diachronic process of adaptation user constraints

grammatical constraints

Fig. 2: The adaptive origin of grammatical constraints according to Haspelmath (1999)

It is instructive to compare this approach to the one advocated by Newmeyer. While diametrically opposed to Haspelmath with respect to the judgements made about the possibility and legitimacy to describe grammar independently from usage (which is strictly denied by Haspelmath), Newmeyer’s approach also acknowledges that large parts of grammar are accessible to a functional motivation, and he proposes that we should explain them by a process of selection and adaptation (Newmeyer 1991). Newmeyer understands this process as a biological process in the proper sense. It is not assumed to apply to all features of grammar, for in some cases he claims the external motivations have become weakened, indirect or non-existent (Newmeyer 2001). So interestingly, in spite of the differences concerning fundamental assumptions of functionalist and formalist approaches, Haspelmath’s and Newmeyer’s accounts come close to each other with respect to the evolutionary elements alluded to. Can evolutionary accounts therefore achieve the synthesis of formalism and functionalism that has long been claimed for (Newmeyer 1991:6, 2003a:686-687) and offer a comprehensive theory of language change? To give an answer to this question, I will now study different examples of language change as well as theoretical objections to the outlined approaches. 3.2. Critique First of all, a basic merit of adaptive accounts is the non-teleological explanation of language change they offer. The paradox that language changes although speakers (in most cases) do not intend to change it (Saussure 1969:108, Oesterreicher 2001a:1583) has constituted an obstinate problem in the study of language change, which is still not entirely overcome. The notion of adaptation offers a simple solution to this problem: language change is explained by a long process of diachronic adaptation that operates on large scale, and in this respect, language change is independent of the individual motivations or reasons of single uses of language in communication. Yet there are also basic problems which arise with adaptive explanations. They have been criticized for being circular (Newmeyer 2001:119), and it is precisely their strength to explain too much that becomes problematic: “The problem is that appeals to ‘need’ have a post-facto feel to them. One observes a generalization and comes up with a plausible story to account for it.” (Newmeyer 2003a:693, see also Lass 1980:83 and Nettle 1998:450). This critique has been formulated earlier with respect to adaptive explanations in sociobiology by Gould. According to him, the attempt to explain individual adaptations “by reconstructing history and assessing current utility” (Gould

222

ESME WINTER-FROEMEL

1978:530) equals telling “just-so stories” like the poetic stories invented by Rudyard Kipling (1987).7 In Haspelmath’s article, the problem becomes manifest when an earlier stage of pronunciation is reconstructed on the basis of a later, attested stage which is then “derived” from this reconstructed state (Haspelmath 1999:195). To avoid such circularities, the proposed explanations should be derived from independent models (Nettle 1998:451) and the diachronically prior stage should be founded on independent criteria, such as cross-language surveys (Nettle 1998:452), linguistic or metalinguistic historical evidence, comparative data from genetically related languages etc. (some methodological principles for reconstruction are sketched by Lass 1980:50-82 and Koch 2004a). Yet, even if certain explanations can be reformulated in a more sound way, a basic theoretical issue remains problematic. It seems dubious to me that Haspelmath’s general claim (the explanation of the correspondence between grammatical optimality and user optimality in terms of a theory of diachronic adaptation) really “can be falsified easily by showing that a particular synchronically adaptive structure could not have been arisen through a diachronic process of adaptation” (Haspelmath 1999:194). It is in principle always possible to invent an adequate adaptive story, and even if there is no positive evidence confirming this story, it could always be argued that the general claim has to be maintained until the necessary evidence is found (compare Lass 1980:86). Another issue that calls for clarification in adaptive explanations of language change is the concept of the environment. Most biological adaptations presuppose a new or changed (e. g. colder) environment (Campbell 1997:469) that triggers a relatively slow process of adaptation consisting of a series of rather small changes, like in the development of antifreeze proteins in the blood of arctic fish (example from Haspelmath 1999:187). Most processes of language change, in contrast, do not presuppose a changed environment (Dahl 1999:209), but concern universal principles of speaking.8 7 Gould characterizes this mode of reasoning more explicitly as follows: “Even [sic] since

Darwin proposed it, the theory of natural selection has been marred by an uncritical style of speculative application to the study of individual adaptations: one simply constructs a story to explain how a shape, function, or behavior might benefit its possessor. Virtuosity in invention replaces testability and mere consistency with evolutionary theory becomes the primary criterion of acceptance.” (Gould 1980:267). Furthermore, Gould (1978:531, 1980:258) insists on the importance of other evolutionary agents besides natural selection, e. g. genetic drift and the fixation of neutral mutations. In a parallel way, it could be asked if linguistic approaches based on the notion of adaptation are not too one-sided in admitting only this kind of mechanism (see also Lass 1990:80-81). For a general discussion of the problem of just-so stories in biology see also Wolgast (1984:238-245). 8 An example for changes implying a different environment would be lexical changes due to the invention of new technologies and scientific progress (see also Lass 1980:133). Furthermore, a changed environment could be assumed when pragmatic meanings become weakened, or when euphemisms like Fr. imbécile (originally meaning ‘weak’, then ‘stupid’) or Ital. bustarella (‘small envelope’ > ‘bribe’, both examples from Blank 1997) are employed with high frequency

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

223

Furthermore, when referring to the hearer as a part of the linguistic environment, another difference to biological evolution becomes manifest. While the biological environment for certain adaptation processes can be characterized by relatively primitive features (e. g. a certain temperature of water in the case of the development of antifreeze proteins in polar fish, or leaves growing in a certain height in the case of the development of long necks in giraffes, compare Haspelmath 1999:187-190), the linguistic environment is in general conceived in a more sophisticated way. It implies among others an intelligent hearer capable of rapidly discerning new linguistic behaviours and imitating them in turn (Itkonen 1999). In addition to this qualitative difference between the biological and linguistic environment, the interaction of speaker and hearer also differs to some degree from the biological model of an organism interacting with his environment. In biology, the environment sets the frame conditions the organism may react to. Language change and linguistic communication in general, however, imply a direct interaction of speaker and hearer in which both parts are active. For example, the speaker generally tries to anticipate hearer reactions and, on the basis of these reflections, chooses the linguistic formulation s/he considers adequate.9 Moreover, in certain types of language change, the role of the hearer can become even more decisive, as s/he can actively contribute to ratify the change. The different ways of interaction of speaker and hearer can be illustrated by the following examples: (1) a. b. c. d.

Can you pass me the salt? vs. Are you able to pass me the salt? (Eng.) Kannst du mir das Salz geben? vs. Bist du in der Lage, mir das Salz zu geben? (Germ.) Mi puoi dare il sale? vs. Sei capace di darmi il sale? (Ital.) Tu peux me passer le sel? vs. Tu es capable de me passer le sel? (Fr.)

In many languages, expressions of the first kind are used in indirect speech acts to make a request (here, of passing the salt). The speaker tries to anticipate the hearer reaction and assumes that an indirect formulation might be interpreted as less offensive. Moreover, the expressions having this pragmatic implicature (+> Give me the salt.) are regularly opposed to the second, periphrastic expressions, which only convey the “normal” lexical meaning (Levinson 2000:141). Again, to explain this stable opposition of meaning, one has to refer to the intelligent hearer, who is able to recognize the pragmatic

and lose their euphemistic character, so that new euphemisms may be required (e. g. Ital. tangente, Blank 1997:129). 9 Compare, e. g., the third point in Grice’s characterization of conversational implicature: “[…] (3) the speaker thinks (and would expect the hearer to think that the speaker thinks) that it is within the competence of the hearer to work out, or grasp intuitively, that the supposition mentioned in (2) IS required” (Grice 1975:50).

224

ESME WINTER-FROEMEL

differences between the expressions and to draw the adequate inferences. Another example is given in (2): (2) a. b. c. d. e.

während temporal meaning > adversative meaning while ” mientras (que) ” mentre ” tandis que ”

(Germ.) (Engl.) (Span.) (Ital.) (Fr.)

In a broad range of languages, similar expressions with a temporal meaning have acquired an adversative meaning as well. This change has been explained in the following way (Keller 1995a:27, 1995b: 230-231, Detges 2001:40-44):10 At a first stage, when the expressions only have a lexicalized temporal meaning, they can be used in contexts where there is an additional adversative meaning that is realized as a pragmatic inference: (3) a. b.

Während du in der Sonne lagst, habe ich Geschirr gespült. (Keller 1995a:27, 1995b:230) While you were lying in the sun, I washed the dishes.

(Germ.) (Engl.)

The choice of the lexical item während/while in utterances like (3) can be attributed to a strategy of the speaker anticipating a certain hearer reaction: instead of expressing the potentially offensive adversative meaning explicitly, s/he leaves it upon the hearer to draw the pragmatic inference. If, however, this strategy is used more and more often, the pragmatic inference can on the long run become part of the semantic meaning of while/während. Thus, we have here another example in which the hearer plays a fundamental role, as his/her interpretation of the utterance is decisive for the described change. It could be pointed out that Haspelmath by no means overlooks the role of the hearer, as he speaks of user constraints and thus includes both the speaker and the hearer. More precisely, his approach focuses on motor costs and processing costs on the one hand and informativeness and persuasiveness as most important benefits on the other hand (Haspelmath 1999:186). Yet, as the examples have shown, language change may be influenced by further, more complex factors, especially factors related to pragmatic strategies of coveringup and uncovering pragmatic layers of meaning. From this perspective, a more dynamic picture of language change emerges.11 Summing up, the preceding

10 For a similar explanation of the development from concomitant to concessive meaning see

Traugott & König (1991:199-203). 11 Furthermore, in spite of Haspelmath’s basic claim referring to user constraints, there are many passages in his paper in which language change is presented as a process that is essentially controlled by speakers (compare, e. g., the conclusion: “variants are created from which speakers may choose. […] speakers tend to choose those variants that suit them best.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

225

remarks concerning the linguistic environment do not invalidate adaptive explanations in general. Nevertheless, they point to interesting divergences between linguistic and biological evolution, which, to my knowledge, have not been studied in detail in adaptive approaches up to now. More serious problems arise with processes of language change that follow an inflationary logic: (4) a. b. (5) a. b.

Weib “woman” > “(pej.) woman” Frau “mistress (fem. of master)” > “woman” ne … pas: emphatic negation > simple negation no … nada: ”

(Germ.) (Fr.) (Span.)

In all these cases, language change on the long run does not correspond to the original intentions of the speakers (as predicted by many adaptive views), but takes exactly the opposite direction. The fact that Germ. Weib, originally a neutral term, has acquired a pejorative meaning, while Frau can now be considered the “normal” expression (opposed to more polite expressions like Dame) has been explained by an abusive use of the gallant expression Frau (Keller 1994:108, see also the editors’ introduction to this volume). In a parallel way, the changes in (5) can be characterized as an “unintended, random by-product of a strategy of ‘abusive’ relevance-marking” (Detges & Waltereit 2002:185). As such inflationary changes contradict the fundamental thesis of adaptation to speakers’ needs, they pose a major problem for adaptive frameworks.12 This is equally valid for seemingly dysfunctional processes of language change, which appear to counteract the needs of language users by promoting difficult structures. A well-known example for such changes is provided by English word stress (Newmeyer 1999:237, 2001:115). Old English has very simple stress rules (word stress is on the first syllable, except for words containing certain prefixes, in which case the stress is on the root syllable). Borrowings, however, introduce a considerable amount of forms that are accentuated according to the stress rules of Old French and Latin, and this leads to the present, extremely complex system of English word stress. How can this seemingly non-adaptive change be accounted for? One possible answer is to exclude such cases from explanation on the grounds that not all features of grammar are considered as functionally motivated (Newmeyer 2001:115). Yet this means that no explanation for the rise of these structures is offered. In 3.4

[…]”; 1999:203). Bearing in mind the examples discussed above, it seems that such characterizations could at most account for certain kinds of language change. 12 Haspelmath (1999:204, note 10) acknowledges the existence of such “non-adaptive” or “counter-adaptive” changes, but excludes them from the range of his theory and thus offers no explanation for them.

226

ESME WINTER-FROEMEL

and 3.5, I will discuss another possibility of tackling the problem by starting from a different viewpoint, focusing on the use of the single linguistic items. Another fundamental problem adaptive views have to face are variants which coexist within one language, but differ with respect to their grammatical structure: (6) a. b.

Café [ka'fe] “café” – Kaffee [ka'fe] “coffee” – Kaffee ['kafə] “coffee” Pizzas “pizzas” – Pizzen “pizzas”

(Germ.) (Germ.)

According to an adaptive view, such variations should be expected to disappear, the “better” variant winning over the other. In many cases, however, such a process does not take place.13 Furthermore, how can we actually determine which structure is in fact the “better” one? For example, it is customary to consider structures as good if they conform to the regularities of the language system in question (Kaffee ['kafə], Pizzen), but this criterion lacks a thorough linguistic elaboration. In some cases, we find groups of items that differ from the “normal” structures of the language, but nevertheless obey certain regularities or even follow productive patterns. For example, German plural formation in -s constitutes a productive pattern that is applied to subgroups of the German lexicon, like in the case of Pizzas (Wegener 2004). In a general way, it seems questionable that an overall evaluation of linguistic forms of the type “Pizzen is better (adapted to language users’ needs) than Pizzas.” is attainable. (Or would it rather be Pizzas being the “better” form?) A related problem is the explanation of the absence of change (Traugott 1999:239). If languages are supposed to be in a constant process of evolution in which they are adapted ever more perfectly to the needs of language users, it is not to be expected that there are conservative languages or structures which hardly change over certain periods of time. But, as Lass (1980:35) points out, “(l)anguages are apparently quite free not to optimize.” It could be objected that conservative languages or structures might already be in an optimal status, but this would beg the question of why the optimal stage has been reached in some cases, while it is not realized in others. Additionally, even conservative languages are not immune to change, that is, after certain periods of time, changes may occur. This observation represents an instantiation of the general paradox of change outlined in the introduction to this volume, and no easy explanation for this problem seems available in adaptive views.

13 A similar point could be made with respect to the differences between languages (Nettle

1998:454-455). According to the assumption that languages evolve towards communicative optima, they should be expected to converge, which is obviously not the case. Nettle argues that this difficulty can be overcome by accepting different solutions for design problems due to competing motivations and pressures and by introducing the notion of local optima (analogous to the biological notion of local optima).

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

227

Why should fully functioning languages change at all? It has been pointed out that up to now there is no evidence for languages becoming ever more functional (Newmeyer 2003b:593), and the concepts of optimal and nonoptimal languages as well as of linguistic adaptation have been rejected in principle (Lass 1980:35, 79, 133-134, Labov 2001:10, 13). Furthermore, different previous theories of language change posit maxims, principles or constraints that contradict each other, so that there can always be only local progressions in the evolution of languages, but no global optimization (compare Nettle 1998:455, Croft 2000:82).14 For example, Keller formulates contradictory maxims that guide the linguistic behaviour of speakers, such as (7) Speak in such a manner that you do not draw the attention to yourself. vs. Speak in such a manner that you get noticed. (Rede so, dass du nicht auffällst. vs. Rede so, dass du beachtet wirst.) (Keller 1994:137, 139)

Natural approaches to language change equally emphasize the fact that reductions of markedness in one domain (e. g. phonology) may lead to an increase of markedness in other domains (e. g. morphology) due to the inherent contradiction between the principles of markedness. So, it is concluded, there cannot be an overall optimization of the language system, just as there cannot be an “ideal” system of language: […] the improvement of language structure by the reduction of markedness which is achieved by change [is] always only local, not necessarily global as well. Because of the contradictory nature of markedness principles there cannot be a simultaneous optimization of the language system on all sides and therefore no ‘ideal linguistic system’, as by the reduction of markedness new markedness is created again and again. ([…] die vermittels Wandel herbeigeführte Verbesserung der Sprachstruktur durch Markiertheitsabbau [ist] immer nur lokal, nicht notwendigerweise auch global. Aufgrund der Widersprüchlichkeit der Markiertheitsprinzipien kann es keine allseitige Optimierung des Sprachsystems zugleich und daher auch kein ‘ideales Sprachsystem’ geben, da beim Abbau von Markiertheit immer wieder neue Markiertheit entsteht.) (Wurzel 1997:298)

The notion of ‘optimality’ has also been put forward in the framework of Optimality Theory (OT), where it is understood in a local and relative sense as well. In OT, the term is referred to single output forms compared to other possible output forms (Tesar, Grimshaw & Prince 1999:295, 302-302), without being applicable to whole languages or linguistic structures. More specifically, OT distinguishes two groups of constraints which are in contradiction to each other. While markedness constraints call for the simplification of marked forms, faithfulness constraints require that output forms respect the features of the corresponding input forms. The inherent conflict between the groups of 14 Interestingly, this characterization of being only locally progressive holds for biological evolution as well (compare note 13).

ESME WINTER-FROEMEL

228

constraints can be solved differently, which is exemplified by final devoicing in English and Dutch (Kager 1999:14-18). Two constraints are relevant here, a markedness constraint stating that obstruents must not be voiced in coda position, and a faithfulness constraint demanding an identical specification of the feature [voice] in input and output forms: (8) *VOICED-CODA (9) IDENT-IO(voice)

Obstruents must not be voiced in coda position. The specification for the feature [voice] of an input segment must be preserved in its output correspondent.

The two constraints can be ranked differently in single languages. In Dutch, the markedness constraint is given priority, so that devoiced forms are obtained. In English, in contrast, the faithfulness constraint is prior, ruling out a devoiced form and leading to the voiced form [bεd] (Kager 1999:17): (10) a. b.

Dutch: *VOICED-CODA >> IDENT-IO(voice) Engl.: IDENT-IO(voice) >> *VOICED-CODA

→ [bεt] → [bεd]

The general question that seems to lie behind these issues is how universal principles can be incorporated in a theory of language change seeking to explain language-specific developments. Constraints that are assumed to be universal can neither be abolished nor introduced in single languages precisely because of their universality (Wurzel 1999). OT seems to offer a promising solution to this problem, as it incorporates general principles into languagespecific analyses (Hayes 1998:245) and makes a clear distinction between the elements of the theory that are conceived as universal on the one hand (among others, the constraints) and the elements that are supposed to account for the language-specific facts on the other hand (constraint rankings). In evolutionary approaches referring to linguistic adaptation, however, ‘adaptation’ and ‘optimization’ are applied to both universal and language-specific aspects, so that theoretical problems may arise. 3.3. Interim Summary I The objections raised render adaptive explanations of language change problematic in various respects. They have difficulties in accounting for some kinds of change, e. g. inflationary or seemingly dysfunctional changes, or developments leading to a stable coexistence of variants. Even if some of these problems can be solved by excluding certain types of language change from the range of the theories, other theoretical and methodological issues remain problematic. A central issue which seems to lie behind these problems is the failure of adaptive views to adequately distinguish between language-specific and universal aspects. Thereby another distinction of two fundamental questions is obfuscated as well: 1) the question of the origin of language and grammar, and 2) the question of explaining language change.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

229

The first question concerns, according to different grammatical theories, the origin of UG (Newmeyer), of principles and parameters, or the origin of grammatical constraints (OT, Haspelmath). In a nutshell, it refers to those elements of grammar that are assumed to be universal. The second question, in turn, relates to possible differences between fully evolved languages, i. e. to specific settings of parameters or specific rankings of constraints (while the parameters or constraints themselves remain unaltered). Both questions are addressed in Haspelmath’s article. His main concern is the origin of grammatical constraints (Haspelmath 1999:183), but he treats cases of language change as well, in which “variants are created from which speakers may choose” (Haspelmath 1999:203). According to the OT framework of Haspelmath’s article, such variants must be assumed to be generated on the basis of the same universal constraints ranked differently.15 Thus a fundamentally different issue is in question here: the adaptive nature of constraint (re-)rankings. However, this issue is not covered by the original use of the term ‘optimality’ in OT. While it lies out of the scope of the present contribution to discuss whether adaptive arguments may be proposed for the explanation of the origin of language (compare Newmeyer 1991, 2001, 2003a and Givón 2002), it has been shown that the notion of adaptation proves problematic for explaining language change. 3.4. Two-level views: Innovation and propagation Thereby we come to a second group of evolutionary views. I will concentrate on the theories sketched by Keller (1994, 1995a, 2005) and Croft (2000), who can be considered as important advocates of this group of approaches. They focus on the mechanisms that govern the creation of new linguistic forms and their propagation within speech communities, as well as on the interaction of these two processes. It has to be pointed out that this group of positions is often not fully recognized, and many evaluations of evolutionary approaches mainly deal with adaptively orientated accounts. For example, the question whether linguistic change can be functionally explained is treated by Lass (1980, Chapter 3) as the “teleology problem” and subsumed under an 15 Compare Haspelmath’s (1999:190) causal explanation for final devoicing in forms like [kæts], where the author himself admits that the “example is not ideal because [kæts] and [kætz] cannot have occurred as variants side by side for a very long time. [kætz] is very difficult to pronounce, so it was presumably eliminated very soon.” (Haspelmath 1999:190). In fact, several commentators express their doubts whether a form [kætz] may have ever existed (Dahl 1999:209, Dresher & Idsardi 1999:213, Minkova 1999:228). Yet the problem at issue seems of more general nature: either we deal with a valid example of language change, in which case we would have to assume that the form [kætz] has existed at a certain time and is yielded by a possible constraint ranking (such a ranking would be in fact, according to Haspelmath’s formulation of the relevant constraints, IDENT(VOICE) >> SAMEVOICE). On the other hand, if the form [kætz] is assumed to be universally impossible, this would still beg the question of why this form is principally ruled out and why the respective ranking is not possible.

230

ESME WINTER-FROEMEL

adaptationist framework. However, several authors reject both teleological and adaptive assumptions, but nevertheless claim to take an evolutionary view on language change. Contrarily to adaptive accounts like the ones studied in the previous sections, these approaches take a decidedly narrow focus and consider not the overall direction of linguistic change, but single instances of change.16 Although Keller only assumes loose analogies between linguistic and biological evolution (see section 2), he develops a comprehensive evolutionary model for explaining language change. Central to his model is the notion of ‘invisible-hand explanations’, which, according to Keller, constitute the only type of explanation that is adequate for language change: An [invisible-hand, EWF] explanation can generally be characterised as follows: it must show how the communicating individuals’ efforts, which are designed to fulfil their own communicative purposes and not to cause language change, lead or must lead to the structure to be explained, thereby taking into consideration the ecological conditions under which people act. (Keller 1995a:25)

The explanandum, a concrete phenomenon of language change, is viewed as the causal consequence of several individual intentional actions, so that it can be derived from these actions, which in turn are determined by cultural and linguistic conditions as well as general maxims followed by the speakers. As mentioned in (7) above, Keller assumes different maxims which may contradict each other. They can be divided into two groups. While static maxims account for homogeneity and stasis, dynamic maxims are assumed to be responsible for language change. (11) Static maxims: Speak in such a manner that you can be recognized as a group member. Speak in such a manner that you do not draw the attention to yourself. or generally, Speak like the others. (Keller 1994:137) Dynamic maxims: Speak in such a manner that you get noticed. Speak in such a manner that you can be recognized as not being part of the group. Speak in such a manner that you avoid an unnecessary effort. etc. (Keller 1994:139)

Furthermore, Keller conceptualizes language change as a complex process operating on two levels, a micro-level of intentional actions and a macro-level of the global consequences of these actions. According to Keller, a full 16 Similarly, Hull pleads for a narrow focus in his study of the history of science (1988:21): “If biological evolution is to be understood, the specific interactions between small groups of organisms in their particular environments must be studied. A broader focus misses everything of interest. A parallel observation holds for science.”

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

231

explanation of language change has to take into account both levels, so that we obtain the general model of linguistic explanation shown in Fig. 3. ecological

intentional actions

causal consequence invisible-handprocess

explanandum

conditions Fig. 3: The form of an invisible-hand explanation (Keller 1994:125, 1995a:18)

A partly similar evolutionary theory of language change is outlined by Croft (2000). Starting from the basic assumption that language change is initiated in communication, he equally distinguishes two stages of language change in his ‘Theory of Utterance Selection’: […] language change consists of two processes: altered replication of a replicator, i.e. innovation; and differential replication of replicators leading to survival/extinction, i.e. propagation. That is, change = innovation + propagation. (Croft 2000:185)

The first step consists of single replications of linguistic units, which can take the form of ‘normal’ replications (conforming to the linguistic convention) or ‘altered replications’, by which the speaker introduces a new form of a linguistic unit. The new form can then, in a second step, spread throughout the linguistic community, this process being labelled ‘differential replication’ or ‘selection’.17 Thus, according to Croft, an explanation of language change has to comprise both processes, which are guided by distinct kinds of mechanisms. While altered replication is to be explained exclusively by functional factors, only social factors are assumed to be relevant to the explanation of differential replication: One of the central theses of this book is that there are distinct causal mechanisms that bring about the innovation and the propagation of language change […]. Functional factors – the phonetic and conceptual factors appealed to by functional linguists – are responsible only for innovation, and social factors provide a selection mechanism for propagation. (Croft 2000:38)

Croft’s theory can therefore be illustrated by the scheme given in Fig. 4.

17 However, it appears to me that the term ‘selection’ is used in different readings through

Croft’s volume. While it is sometimes referred to individual acts of selection, it has to be understood as an inherently social notion in other passages. The same problematic use can be observed for the more traditional concept of ‘propagation’. I will return to this point in section 4 below.

232

ESME WINTER-FROEMEL

differential replication/ ← social factors propagation/selection
normal replication vs. altered replication of linguistic ← functional factors units/innovation Fig. 4: Language change according to Croft’s (2000) Theory of Utterance Selection

At first glance Keller’s and Croft’s theories seem to come very close to each other. They both take into account two levels in the explanation of linguistic change and thereby integrate individual and social aspects. However, the theories imply a slightly different conception of the two levels. Keller’s account is mainly cumulative, since he studies a series of mutually similar, but independent innovations on the micro-level. In his conception, change on the macro-level is essentially seen as a sum of innovations. Croft’s account, in contrast, is mainly based on replication. He interprets change as a process starting with a single innovation which then spreads throughout the speaker community. Change on the social level is therefore the consequence of a sum of reproductions. As we will see in the following discussion, this difference between Keller and Croft also has repercussions on the factors that are proposed to explain language change. 3.5. Critique A basic merit of this second group of evolutionary accounts becomes immediately clear: as with adaptive accounts, a non-teleological explanation of language change is offered and the Teleological Fallacy is avoided. This time, the paradox that language changes although it is not the intention of the single speakers to change it is solved by distinguishing two levels on which language change operates: individual intentions are restricted to the micro-level of innovations, whereas the process of change on the macro-level is out of the reach of individuals. Yet two-level accounts do not only share this basic merit with adaptive approaches, but also one problematic point. The danger of circular explanations and modes of reasoning is also given in this group of accounts, as the following citation illustrates: Most of the time, we only ascertain post festum on the basis of the existence of the explanandum that the premises were fulfilled. That is, we know the explanandum, know the laws and reconstruct the premises. This is the crucial point in diagnostic explanation. (Meist stellen wir erst post festum auf der Basis der Existenz des Explanandums fest, daß die Prämissen erfüllt waren. Das heißt, wir kennen das Explanandum, kennen die Gesetze und rekonstruieren die Prämissen. Dies ist der Witz der diagnostischen Erklärung.) (Keller 1994:106, emphasis original)

In order to address the possible objection that such explanations are circular, Keller introduces the notion of explanative power. According to him,

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

233

circular explanations are true, but do not have any explanative power at all, because the explanandum and the explanans belong to the same domain. In invisible-hand explanations, in contrast, the concepts by which a certain phenomenon is explained are typically taken from very distant domains (essentially psychology and sociology, Keller 1994:106). However, it seems dubious to me that the problem can be solved in such a manner. The mode of reasoning remains circular if the premises are reconstructed on the basis of the observed explanandum, independently of whether the explaining concepts are taken from a closely related or a more distant domain. (Besides, the traditional explanations for language changes are not exclusively psychological or sociological, but, evidently, also involve linguistic aspects.) Nevertheless, like for adaptive views, it seems in principle possible to remedy such circularities and propose non-circular explanations by pursuing other strategies and basing the explanations on independent or independently confirmed (linguistic, metalinguistic, or extralinguistic) evidence. Besides the common aspects mentioned, two-level accounts differ from adaptive accounts in various important respects. They permit us to grasp a series of phenomena that have proven highly problematic within the other group of theories. The absence of linguistic change, which constitutes a crucial problem for adaptive (and other) theories of linguistic change (Traugott 1999:239), can now be explained in a simple way. Keller provides for static as well as for dynamic mechanisms (recall the maxims cited in (11)), the former fostering linguistic conservatism, while the latter may lead to language change. In a similar way, Croft confronts the options of normal and altered replication and underlines the fundamental role played by linguistic convention. Therefore, both theories can account for change and conservatism. Similarly, they can account for changes which lead to coexisting variants: (6) a.

Café [ka'fe] “café” – Kaffee [ka'fe] “coffee” – Kaffee ['kafə] “coffee”

(Germ.)

According to the two-level views, the explanation has to start by stating the motivations for the use of certain forms on the micro-level. Here, in fact, different motivations can be specified: the forms that are stressed on the ultima keep the original French stress, so that we deal with a form of conservatism, a normal replication to which Keller’s static maxims apply. On the other hand, however, the forms with final stress do not conform to the stress rules of native German words and can thereby carry an additional pragmatic meaning of extravagance, art de vivre, being chic etc. (compare Keller’s maxim “Speak in such a manner that you get noticed.”).18 The form that is stressed on the first 18 The same holds for the spelling of the form Café, which deviates from the rules of German, as the form contains the foreign grapheme <é> and the phoneme-grapheme-correspondence /k/ ↔ . Interestingly, the example hints at the fact that explanations for single forms may be very complex, and both static and dynamic maxims may have to be taken into account.

234

ESME WINTER-FROEMEL

syllable, in turn, is motivated by an adaptation to German word stress rules, and it may be assumed to cause less processing cost. It is frequently found in compounds (Kaffeebohne, Kaffeetasse, Kaffeepause, etc.), so that prosodic integration parallels morphologic integration. Furthermore, the coexistence of the graphic variants Café and Kaffee permits the speaker to make a semantic distinction between the meaning “small restaurant where hot drinks such as coffee are served” and “seeds of the coffee tree”/“hot drink” (compare Eng. café vs. coffee). Thus, each of the three forms can be seen as functional and potentially useful for the speaker, and this may explain why all of the forms continue to be used today. Two-level explanations can also account for inflationary processes of language change. Recall the examples: (4) a. b. (5) a. b.

Weib “woman” > “(pej.) woman” Frau “mistress (fem. of master)” > “woman” ne … pas: emphatic negation > simple negation no … nada: ”

(Germ.) (Fr.) (Span.)

In fact, (4) constitutes a standard example of Keller’s theory, which is lengthily discussed in his book. It is presented as an instantiation of Mandeville’s paradox and resolved in a plausible way by distinguishing micro-level events and their global repercussions (a similar description holds for (5)): precisely because a great number of speakers exaggeratedly use the high, emphatic or expressive forms due to maxims such as “Be expressive”, these forms lose their expressive potential. The change is then the global consequence of a multitude of individual intentional actions. Additionally, even seemingly dysfunctional processes like the evolution of English word stress can be explained. In fact, this change is dysfunctional only on the macro-level. On the micro-level, in contrast, where the two-level explanations start, it can be traced back to a multitude of single acts of borrowing from Old French and Latin, which are certainly not dysfunctional, but serve individual motivations and goals. Here again, the fact that word stress is not assimilated to native English rules in the process of borrowing (which would in principle be a viable option as well) can be motivated by factors such as extravagance, expressiveness etc. (A more detailed listing of possible motives for linguistic borrowing and other types of lexical innovation is given by Tesch 1978:199-216 and Blank 1997:345-405, 2001:95-99.) Summing up, two-level views provide a better explanation for certain kinds of change that have proven problematic for adaptive approaches. However, there are also some possible objections to Keller’s and Croft’s theories. Most of Keller’s remarks focus on changes induced by the speaker and on the role of the speaker in language change. As shown in 3.2, however, the hearer also

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

235

plays a fundamental role in communication and linguistic change (see also Detges & Waltereit 2002). Like most traditional approaches, Keller mainly considers factors that are relevant to innovations, but does not address the question which factors guide language users who take over the innovations of others. Thus, in Keller’s theory, language change is regarded largely as a cumulative phenomenon which takes place when a large number of individual speakers act similarly (compare the analogy Keller draws to the formation of circular geometrical structures by spectators who watch artists in front of Centre Pompidou, Paris; Keller 1994:32-36). Generalizing this point, we can say that in Keller, the propagation of an innovation is not considered as a stage of language change in its own right, which would require to be studied apart from innovation. However, even if the parallel acting of a large number of individuals certainly is an important feature of many examples of language change, in many cases this is not the whole story.19 For instance, the inflationary changes referred to by Keller himself do not simply consist of an accumulation of single acts of being abusively gallant etc., but the accumulation leads to a change in the meaning of the linguistic forms which then induces speakers to behave differently and no longer use certain forms (e. g. Germ. Weib), but other forms instead (e. g. Germ. Frau). With respect to this issue, Croft takes an entirely different position, as he distinguishes the stages of innovation and propagation, and further assumes that these are guided by qualitatively different, that is, functional vs. social factors. Here, his account is more detailed, as he lengthily discusses different factors and mechanisms relevant to both stages (Croft 2000:117-195). In contrast to Keller’s scheme (see Fig. 3), which does not indicate the possibility of macro-level influences on the micro-level (even if Keller, in a later passage, briefly mentions such reverse effects; Keller 1994:205-206), Croft (2000:9599) takes into account this aspect by stressing the importance of linguistic convention. According to him, convention results from individual utterances, but acts back on the individual utterances, since it determines what is to be considered as a normal or an altered replication (in fact, an innovation can only be determined with respect to a given convention). However, as convention is essentially a social concept, this influence seems to undermine Croft’s (2000:166, 175) strict attribution of functional and social factors to the stages of innovation and propagation respectively. And in fact, this assumption has also been criticized by other authors who admit the possibility of both types of factors guiding both innovations and their adoptions and competing with each other (compare the discussion in Croft 2000:184). Recent research suggests that social factors may even rule out functional factors 19 Croft (1999:208, 2000:60) equally criticizes this point and regards Keller’s theory a theory of evolutionary drift rather than selection.

236

ESME WINTER-FROEMEL

during innovation (Thomason & Kaufman 1988:15, Thomason 2003:709, Enfield 2003:7, compare Nettle 1998:453), and some of the examples discussed above equally seem to confirm this possibility (recall the various cases of borrowings). Moreover, just as social factors may be operative during innovation, functional factors may play a role during propagation. Some mechanisms that are relevant here are studied by Enfield, who proposes a further distinction of two subsequent stages (2003:2-19): 1) an individual language user being confronted with an innovation, and 2) the possible adoption of the innovation by this individual. Factors which influence the propagation of an innovation are social network structures, since they contribute to determining which individuals are likely to be confronted with innovations at all (Milroy & Milroy 1985), individual personality factors (innovative vs. more conservative), which can influence the reaction to the innovation, and the social role of the innovative speakers, as their innovations are more likely to be positively perceived if they occupy a model role than if they are social outsiders. Further factors are the cognitive processing effort of the linguistic forms and the social value of each innovation, which can convey high social prestige, but also a ‘risk to face’ run by using an extravagant innovation. As the importance of these factors during the different stages of language change is not well understood up to now, I would plead for a position that considers both functional and social factors as potentially relevant to both innovation and adoption, leaving it open to further research to determine their influence and interaction. Related to this last point, a final observation has to be made. Universal linguistic factors are of fundamental importance to different kinds of language change, as they may imply restrictions on possible innovations. Actual research on grammaticalization and reanalysis even tends to privilege such universal factors (Oesterreicher 2001a:1582). Yet paradoxically, the theoretical status of such universal factors remains problematic not only in adaptive views, but also in two-level views. They are rendered explicit neither in Keller’s nor in Croft’s account, and it is not clear how universal and language-specific factors (e. g. convention) interact. 3.6. Interim Summary II As argued in the preceding section, evolutionary positions based on the distinction of two levels permit us to describe and explain straightforwardly many phenomena of language change. They can account for a series of changes and facts that remain unexplained in adaptive views (e. g. inflationary and seemingly “dysfunctional” changes as well as coexisting variants and the absence of change). Thus, two-level views provide a good starting-point for a general theory of language change.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

237

At the same time, however, two-level views contain certain problematic aspects, some of which they share with adaptive views. The danger of circular explanations is apparent in this group of accounts as well, and there also seems to be a tendency to neglect the role of the hearer. Furthermore, Croft’s theory contains assumptions about the importance of functional and social factors that have been questioned by other authors. Finally and most importantly, both adaptive and two-level views fail to adequately distinguish universal and language-specific facts. This raises the question of how this distinction can be integrated into a model of language change. In order to develop a proposal for a revised explanation scheme, I now want to broaden the perspective and discuss the relationships between two-level views and more traditional theories of language change. 4. Two-level views and traditional approaches to language change First I want to turn to the key element of two-level views: the distinction of a micro- and a macro-level on which language change operates (Keller), or the distinction of innovation and propagation/selection as different stages of language change (Croft). In fact, the distinction of these two aspects is in no way new and is not specific to evolutionary theories. Coseriu (1958:44-46), Weinreich, Labov & Herzog (1968), Milroy & Milroy (1985:347, 381), Traugott (1999) and Janda & Joseph (2003:13) have already proposed similar distinctions between linguistic innovation and its propagation or adoption (see also Oesterreicher 2001a:1583, 2001b:230). But this distinction introduces a traditional problem (see note 17): in various approaches to language change, the concepts of ‘propagation’ and ‘selection’ are used imprecisely, indeed ambiguously. On the one hand, ‘propagation’ and ‘selection’ are conceived as the global results of a series of adoptions of a linguistic innovation: “The propagation of a novel linguistic variant is essentially the adoption of a new linguistic convention by the community.” (Croft 2000:174). On the other hand, however, ‘propagation’ is also considered under its individual aspects, i. e. with respect to the single acts of adoption of the innovation by single speakers: “[…] mechanisms of selection are those that favour a particular form being adopted among members of a speech community.” (Croft 2000:73). Here, the notion refers to individual acts of speech: The hearer later replicates the innovation in another context, intending it to be understood by his interlocutors as having a social value. If his interlocutors understand his intention in this context, then the propagation of the innovation has begun. (Croft 2000:186)

Although both aspects are obviously connected, they refer to different phenomena. In order to make the distinction clear, the term ‘adoption’ will be used for the adoption of an innovation by another speaker (individual aspect),

238

ESME WINTER-FROEMEL

while the term ‘propagation’ will be reserved for processes on the level of the linguistic community in the remainder of this article. Even if the basic idea of distinguishing the two levels of linguistic change is not new, one important point becomes still clearer with the evolutionary accounts. As they stress the heterogeneity of factors that contribute to explaining language change more explicitly, evolutionary frameworks permit us to integrate various traditional explanations of language change. For instance, cognitive, psycholinguistic, pragmatic and sociolinguistic factors are recognized as possibly relevant by Keller and Croft, and Croft emphasizes this potential of his theory by stating that “the generalized theory of selection can in fact subsume any theory of language change” (Croft 2000:42). Yet this does not mean that the evolutionary models are completely devoid of content. On the contrary, when bringing together more traditional explanations, they sometimes permit us to refine, correct or at least complete these explanations. For instance, the evolutionary accounts suggest a precise definition of the criterion of frequency that is referred to in many explanations of language change. According to the evolutionary framework, frequency can no longer be viewed as an abstract or purely statistical matter, but has to be understood in a twofold dimension. First, high or low frequency of linguistic items is a social notion, as it concerns the frequency of their use within a given speech community (that is, frequency can be considered on Keller’s macrolevel). Second, the notion involves cognitive aspects on the micro-level, as it can be assumed that the frequency of occurrence of linguistic items in a speech community has repercussions on the way these items are cognitively stored and processed in the individual: “Repetition and its consequences for cognitive representation are major facts in the creation of grammar.” (Bybee 2003:622, emphasis added, EWF). Likewise, an important methodological principle that is made very clear by evolutionary theories states that explanations of language change should start with an individual language user who employs a certain linguistic form in a new way. This methodological claim, which has been put forward also independently from the evolutionary approaches studied (Coseriu 1962a:287-288, 1988:58, Oesterreicher 2001a:1584, 2001b:237, Janda & Joseph 2003:128, Kabatek 2005:158), can be resumed as follows: “The doctrine of methodological individualism states that the fundamental unit or locus of any social process is the individual, and thus all explanations must be phrased in such terms […].” (Enfield 2003:3). This methodological principle invites us, for example, to reformulate certain kinds of information-theoretical reasoning, such as the explanation given for the existence of faithfulness constraints in Kager (1999:5): “From a functional angle, the importance of faithfulness is clear: to express contrasts of meaning, any language needs a minimal amount of formal contrast.” It would seem interesting to investigate if and how such

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

239

factors can be related more directly to individual innovations and how they can become relevant to individual acts of speech. Related to the principle of methodological individualism, it seems finally interesting to mention the grammar-usage controversy, which divides formal and functional approaches (compare e. g. Newmeyer 1998, 2003a, Bybee 200520). While formal accounts generally stress the significance of grammar, that is, the macro-level results of language change, functional accounts tend to privilege the importance of linguistic utterances on the micro-level. Keller (1994) has been both criticized for not adequately considering the decisive role of language users in language change, and accused of focusing too much on changes in single utterances at the expense of changes in linguistic structures.21 This confirms that the evolutionary approaches take into account both aspects and consider them both essential for explaining linguistic change. With respect to the grammar-usage controversy, evolutionary views therefore seem to offer interesting perspectives. Yet, as the discussion of Keller’s and Croft’s theories has shown, some points in their approaches have also remained problematic. In the following section, I therefore want to propose a revised explanation scheme which is based on the two-level views, but integrates a series of further distinctions. 5. A revised explanation scheme of language change The analysis of various evolutionary as well as more traditional approaches to language change has revealed some points that call for further clarifications and some basic problems that have not been satisfactorily resolved up to now. A first problem concerns the notions of ‘propagation’ and ‘selection’, which are used ambiguously in both traditional and evolutionary approaches. I have proposed to distinguish two aspects, individual adoption and the propagation of linguistic forms in language systems or communities of speakers. However, up to now I have not addressed the question of how this distinction could be integrated in models of language change like the ones proposed by Keller and Croft. Another central point is the interaction of different factors guiding 20 In the meantime, Bybee has revised the original version of her paper and replaced it by a

newer version (Bybee 2006). 21 See on the one hand Kabatek (2005:159), and on the other, Stolz: “Therefore KELLER’s

approach carries things so far that the phenomena of language change themselves become irrelevant: One no longer speaks about changed structures, but about changed acts of speech. Thus we finally get the impression that the book Language change does not at all treat of language change, but at best of change of speaker behaviour.” (“KELLERS Ansatz bringt es daher so weit, daß die Sprachwandelphänomene an sich irrelevant werden: Man spricht nicht mehr über veränderte Strukturen, sondern über veränderte Sprechhandlungen. So ergibt sich letztlich der Eindruck, daß das Buch Sprachwandel gar nicht von Sprachwandel handelt, sondern bestenfalls von Sprecherverhaltenswandel.”, Stolz 1991:554, emphasis original). Interestingly, after this passage Stolz equally claims a comprehensive account of language change which includes both levels.

240

ESME WINTER-FROEMEL

language change, which is widely acknowledged in current approaches to linguistic change (compare Nettle 1998:454-455). Yet these factors are generally mentioned in a rather unsystematic fashion, and their status and possible interactions have not sufficiently been worked out. Finally and most importantly, I have argued that a serious flaw in evolutionary approaches is their failure to adequately distinguish universal and language-specific facts and locate them with respect to each other. In order to solve these problems and shortcomings, I propose that we should introduce a series of further distinctions, by the help of which a revised general model of language change and its explanation can be formulated (see Fig. 5). The scheme, which is inspired by Koch (2001:9, 2005:246-249) and Blank (2005:281-283), is built around an enlarged and partly revised version of Keller’s (1994) and Croft’s (2000) two-level approaches (represented in the central part of the figure and shaded in grey), bringing together these evolutionary approaches with Coseriu’s (1958:25-28, 1988: 70-71) distinction of three linguistic levels of investigation (left column, with the extra-linguistic level added) and factors and mechanisms which have been proposed for the explanation of language change in a series of different, not necessarily evolutionary frameworks, among others Coseriu (1962b), Koch & Oesterreicher (1985, 1990), Milroy & Milroy (1985), Enfield (2003), Croft (2000), Blank (1997), Detges & Waltereit (2002) and Koch (1997, 1999, 2001, 2002, 2004b, 2005) (right column; in fact, most of the cited authors do not advocate evolutionary positions). In the central column we find the alternative between normal replication and altered replication/innovation as well as adoption defined as a single act contributing to the diffusion of an innovation. All these processes are situated on the level of discourse. Language change, however, also contains a further aspect: generally, it is assumed to be completed only when linguistic items have acquired a new status within the linguistic system or the linguistic norm in question, when they have become grammaticalized or lexicalized, or when they have become normative or part of the (new) linguistic convention.22 Thus the two perspectives in which the spread of an innovation can be considered are explicitly distinguished and their methodological differences are made clear by referring to Coseriu’s distinction of three linguistic levels of investigation.

22 The fact that linguistic convention appears twice in the scheme is due to the double role it plays in language change. On the one hand, it represents the result of numerous acts of language use in discourse, and in this respect a new convention is the final stage of language change (central column). On the other hand, linguistic convention itself acts upon language use in discourse and in this respect constitutes a factor that guides language change (compare section 3.5 above, where this reverse effect has already been indicated).

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

LEVELS OF INVESTIGATION

universal level: activity of speaking historical/ structural level: language, discourse traditions individual/actual level: discourse

extra-linguistic level

PROCESSES AND STAGES OF LANGUAGE CHANGE

change in the linguistic system, normativity propagation of a new convention



adoption of the innovation in discourse



innovation/ altered replication vs. normal replication

241

FACTORS AND MECHANISMS GUIDING LANGUAGE CHANGE

cognitive and linguistic factors (e. g. semantic and pragmatic universals) system-related and norm-related factors, intralinguistic and interlinguistic factors (e. g. convention, entrenchment, effort of cognitive processing) (specific factors)

(specific factors) conditions of communication, social and cultural factors (structure of social networks, individual personality, social hierarchy)

Fig. 5: A revised explanation scheme of language change

In the first case, when linguistic innovations and adoptions are studied, we deal with individual utterances, located on the actual level of discourse. When speaking about the status of certain forms in linguistic systems or discourse traditions23, in contrast, the perspective is switched to inherently system-related (be it social or mental) or norm-related facts, situated on the historical level. This level thereby includes both structural features in the proper sense (related to language as a structured system) and normative aspects (“convention”, “normativity”). However, central to a deep understanding of language change are not only the different processes and stages it implies, but also the factors and mechanisms that guide language users in their linguistic behaviour which leads to the change. According to the principle of methodological individualism, all these factors are referred to concrete utterances on the level of discourse. The scheme integrates a series of factors that have been proposed in traditional approaches, e. g. cognitive linguistics, typological, sociological, and pragmatically orientated research, and locates these factors on different levels of investigation. Some factors refer to universal principles of cognition and speech (and thus apply to all languages), others to language-specific, historical and/or 23 Several authors (Blank 2005, Koch 2001, 2005, Oesterreicher 1997, 2001a:1559) have

argued that we should locate on the historical level, contrarily to Coseriu, discourse traditions as well. The controversies that have arisen about this proposal cannot be addressed here for lack of space.

242

ESME WINTER-FROEMEL

structural facts (e. g. linguistic rules and convention), or to extra-linguistic facts. The scheme also includes discourse-specific factors, but their explanatory role is limited (which is represented by the use of the brackets in Fig. 5), as these factors are by definition of episodic character and no reconstructions or generalizations can be made about them.24 Paradoxically, although linguistic change starts in discourse, other factors of more general value are central to its explanation. Thus, founding the explanation on processes on the level of discourse does by no means imply that the explanation itself becomes infinitely particular. Furthermore, by including universal factors, but locating them on a specific level and thus separating them from language-specific factors, the scheme fulfils the requirement that 1) universal factors should be explicitly integrated and located in an explanation scheme, but 2) the importance of other, e. g. historical, factors should not be neglected (Oesterreicher 2001a). The broad array of mechanisms that are considered stands in contrast to more restrictive approaches which limit, like Croft does, single factors to specific stages of linguistic change, or explicitly deny the importance of some factors, like Labov, who rejects universal factors altogether (2001:503). The scheme in Fig. 5 deliberately adopts a more general view, leaving open the possibility, however, to make subsequent refinements, concerning, e. g. the relative processing costs of linguistic forms (compare Haider 1999:216) or the range and interaction of different kinds of factors, as it has been proposed that some factors might only apply to innovation or adoption respectively (Croft 2000, compare Coseriu 1983:58, Enfield 2003:16, Koch 2004b). Let us now consider whether the scheme can provide a more comprehensive understanding of the changes discussed above. (1) a. b.

(2) a. b. (3) a. b. (4) a. b. (5) a. b.

Can you pass me the salt? vs. Are you able to pass me the salt? (Eng.) Kannst du mir das Salz geben? vs. Bist du in der Lage, mir das Salz zu geben? (Germ.) etc. während temporal meaning > adversative meaning (Germ.) while ” (Engl.) etc. Während du in der Sonne lagst, habe ich Geschirr gespült. (Germ.) (Keller 1995a:27, 1995b:230) While you were lying in the sun, I washed the dishes. (Engl.) Weib “woman” > “(pej.) woman” (Germ.) Frau “mistress (fem. of master)” > “woman” ne … pas: emphatic negation > simple negation (Fr.) no … nada: ” (Span.)

24 Compare Koch (1997:45-46), who asserts the logical impossibility of formulating rules and norms on the individual/actual level.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

243

These changes are guided by pragmatic inferencing ((1) and (3)) and pragmatic speaker strategies (avoidance of expressing potentially offensive meanings explicitly in (3), abusive politeness or abusive relevance marking in (4) and (5)), factors which can be located on the universal level.25 At the same time, extra-linguistic aspects (e. g. the conditions of communication) play an important role. Furthermore, the examples show that a speaker may deliberately choose to deviate from linguistic convention and use non-conventional means of expressing a given content. Thus, the historical level is concerned as well. The examples therefore illustrate the great complexity of certain types of change, and the scheme represents a framework which permits us to consider the effects and interaction of various factors of change. In seemingly dysfunctional changes like the evolution of English word stress, no explanation on the structural level has seemed available. Yet if we include on the historical level not only structural factors in the proper sense, but also norm-related facts, and further integrate interlinguistic aspects, it seems possible to give an explanation: the conservation of non-native stress in numerous borrowings, which has lead to the change in question, can be treated as an example of interlinguistic ‘conservatism’. At the same time, social and cultural factors related to the situation of language contact are paramount to this change (see 3.5). Similarly, a broad array of factors accounts for the coexistence of variants like in (6): the effort of cognitive processing and the entrenchment of the different forms, their interlinguistic and intralinguistic status, social and cultural factors, additional pragmatic meanings, etc. (6) a. b.

Café [ka'fe] “café” – Kaffee [ka'fe] “coffee” – Kaffee ['kafə] “coffee” Pizzas “pizzas” – Pizzen “pizzas”

(Germ.) (Germ.)

For reasons of space, the different factors and groups of factors cannot be discussed in more detail here. The scheme does not claim to provide a complete list of all factors that are relevant to different kinds of language change, and it might prove necessary to add other factors.26 However, I hope to have shown that the scheme may contribute to a better understanding of different cases of language change by integrating various factors that play a role in the behaviour of language users. Summing up, the main aim of the scheme presented in Fig. 5 above is to propose some general methodological distinctions and principles of 25 ‘Pragmatics’ is understood here as the study of the aspects of meaning that are not covered by semantics (Levinson 1983:12, 32). In a more general understanding of ‘pragmatics’ as the study of language use, the term covers a broad range of primarily extra-linguistic factors which are related to the speaker, the hearer, and the situation of communication. 26 Additionally, it would seem interesting to ask how reductive change, understood as a process which results in the loss of certain items of the lexicon or of grammatical structures (Lüdtke 1989:135, Blank 2001:70), can be incorporated in the scheme.

244

ESME WINTER-FROEMEL

explanation which might permit us to integrate, enlarge and in some cases refine traditional explanations. The framework stresses the complexity of language change and invites us to look for different types of explanatory factors. Moreover, it invites us to make clear how the factors and mechanisms become operative during the single acts of innovation and adoption, and how they interact. Furthermore, it permits us to (re)formulate explanations according to the distinction of different linguistic levels, observing the methodological corollaries that derive from this distinction. 6. Conclusions The investigations of the present article have been directed to the question to what extent evolutionary frameworks might contribute to a better understanding and explanation of processes of language change. In order to answer this general question, it has proven necessary to introduce several distinctions. I have argued that we should distinguish various kinds of evolutionary approaches by 1) the relations between language and biology they assume (metaphorical vs. biologistic vs. generalized views) and 2) the notions that they consider as key elements (adaptive vs. two-level views). Approaches that are mainly focused on adaptation and optimization (e. g. Haspelmath 1999) have proven problematic for explaining language change, while they might be useful for studying the origins of language, an issue that could not be treated here. I have argued that the potential of evolutionary approaches to explain language change rather lies in a second group of views, which put the main emphasis on the introduction of variants and their adoption and propagation. These approaches can provide satisfactory explanations for different kinds of language change, but certain issues have remained problematic. Basing on the two-level approaches, I have therefore developed a revised scheme for explaining language change. The scheme contains the key elements of the evolutionary approaches, but introduces a distinction of two perspectives from which the spread of an innovation can be considered, individual adoption vs. social propagation. Furthermore, the evolutionary models are broadened by combining them with the distinction of four aspects (universal, historical/structural, individual/actual and extra-linguistic) from which linguistic facts and language change may be considered. Finally, it integrates various factors and mechanisms that have traditionally been proposed to explain language change and assigns them to the different levels of investigation. Bringing together different approaches to language change, the scheme can account for a broad array of phenomena, and it may therefore also offer new perspectives on the general paradox of language change. The explanatory potential of evolutionary approaches that adopt a two-level view can be resumed in the following theses: they can serve as a valid point of departure in developing a comprehensive account of language change, which might provide new insights into the paradox of change. They offer a non-

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

245

teleological explanation of language change and integrate various traditional approaches to language change. Yet, as the evolutionary models do not propose genuinely new explanations, they essentially possess a metatheoretical value in their capacity to bring together and systematize previous explanations. In addition, they provide a heuristics that permits us in some cases to refine previous explanations, so that they can serve methodological functions as well. As a final point, however, it seems legitimate to raise the question whether such a framework is still evolutionary at all. In fact, most of the distinctions made by the evolutionary approaches have already been proposed in earlier works, and the language of evolution does not seem strictly necessary on theoretical grounds, so that it would equally be possible to formulate the models independently of biological vocabulary. This would, on the one hand, mean a renunciation of the prestige evolutionary ideas seem to have, but on the other hand, biologistic misunderstandings of evolutionary accounts could perhaps be avoided.

REFERENCES Blank, Andreas. 1997. Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. Tübingen: Niemeyer. ----------. 2001. Einführung in die lexikalische Semantik für Romanisten. Tübingen: Niemeyer. ----------. 2005. “Wie von unsichtbarer Hand geleitet: Der Prozeß des Bedeutungswandels”. Stehl 2005. 277-298. Bybee, Joan L. 2003. “Mechanisms of Change in Grammaticization: The Role of Frequency”. Joseph & Janda 2003. 602-623. ----------. 2005. “The impact of representation: grammar is usage and usage is grammar”. LSA Presidential Address, LSA Annual Meeting 2005. (12.10.2005). ----------. 2006. “From Usage to Grammar: The Mind’s Response to Repetition”. Language 82. 711-733. Campbell, Neil A. 1997. [Original edition 1996.] Biologie. [Original title: Biology.] Heidelberg, Berlin & Oxford: Spektrum. Coseriu, Eugenio. 1958. Sincronía, diacronía e historia. El problema del cambio lingüístico. Montevideo: Universidad de la República (Reimpresión fotomecánica Tübingen 1969). ----------. 1962a. “Determinación y entorno”. Coseriu 1962c. 282-323. ----------. 1962b. “Sistema, norma y habla”. Coseriu 1962c. 11-113. ----------. 1962c. Teoría del lenguaje y lingüística general. Cinco Estudios. Madrid: Gredos. ----------. 1983. “Linguistic Change Does Not Exist”. Linguistica nuova ed antica 1. 51-63.

246

ESME WINTER-FROEMEL

----------. 1988. Sprachkompetenz. Grundzüge einer Theorie des Sprechens. Tübingen: Francke. Croft, William. 1999. “Adaptation, Optimality and Diachrony”. Zeitschrift für Sprachwissenschaft 18. 206-208. ----------. 2000. Explaining Language Change. An Evolutionary Approach. Harlow: Longman. Dahl, Östen. 1999. “Does adaptation really help us to explain language change?”. Zeitschrift für Sprachwissenschaft 18. 209-211. Darnell, Michael et al., eds. 1998. Functionalism and Formalism in Linguistics. (= Studies in Language Companion Series, 41.) Amsterdam & Philadelphia: John Benjamins. Detges, Ulrich. 2001. Grammatikalisierung. Eine kognitiv-pragmatische Theorie, dargestellt am Beispiel romanischer und anderer Sprachen. Habilitation thesis, University of Tübingen. ---------- & Richard Waltereit. 2002. “Grammaticalization vs. Reanalysis: a Semantic-Pragmatic Account of Functional Change in Grammar”. Zeitschrift für Sprachwissenschaft 21. 151-195. Dohrn, Anton. 1876. “Ueber die Bedeutung der zoologischen Station in Neapel für die Lösung zoologischer Probleme”. Deutsche Rundschau 6. 120-127. Dresher, B. Elan & William J. Idsardi. 1999. “Prerequisites for a Theory of Diachronic Adaptation”. Zeitschrift für Sprachwissenschaft 18. 212-215. Enfield, Nick. 2003. Linguistic epidemiology: Semantics and grammar of language contact in mainland Southeast Asia. London: Routledge Curzon. 1-44. Frank, Barbara, Thomas Haye & Doris Tophinke, eds. 1997. Gattungen mittelalterlicher Schriftlichkeit. (= ScriptOralia, 99.) Tübingen: Narr. Givón, Talmy. 2002. Bio-Linguistics. The Santa Barbara Lectures. Amsterdam & Philadelphia: John Benjamins. Gould, Stephen Jay. 1978. “Sociobiology: the art of storytelling”. New Scientist 80 (16 November 1978). 530-533. ----------. 1980. “Sociobiology and the Theory of Natural Selection”. Sociobiology: Beyond Nature/Nurture? Reports, Definitions and Debate, ed. by George W. Barlow & James Silverberg, 257-269. Boulder, Colo.: Westview Press. Grice, H. P. 1975. “Logic and Conversation”. Syntax and Semantics, ed. by Peter Cole & Jerry L. Morgan, Vol. 3: Speech Acts, 41-58. New York, San Francisco & London: Academic Press. Haider, Hubert. 1999. “On the survival of the fittest grammar (theory)”. Zeitschrift für Sprachwissenschaft 18. 216-218. Haspelmath, Martin. 1999. “Optimality and diachronic adaptation”. Zeitschrift für Sprachwissenschaft 18. 180-205. Hayes, Bruce P. 1998. “Phonetically Driven Phonology: The Role of Optimality Theory and Inductive Grounding”. Darnell et al. 1998. Vol. 1, 243-285.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

247

Hull, David L. 1988. Science as a Process. An Evolutionary Account of the Social and Conceptual Development of Science. Chicago & London: The University of Chicago Press. Itkonen, Esa. 1999. “Functionalism yes, biologism no”. Zeitschrift für Sprachwissenschaft 18. 219-221. Janda, Richard D. & Brian D. Joseph. 2003. “On Language, Change, and Language Change – Or, Of History, Linguistics, and Historical Linguistics”. Joseph & Janda 2003. 2-180. Joseph, Brian D. & Richard D. Janda, eds. 2003. The Handbook of Historical Linguistics. Malden, Mass.: Blackwell. Kabatek, Johannes. 2005. “Über Trampelpfade, sichtbare Hände und Sprachwandelprozesse”. Stehl 2005. 155-174. Kager, René. 1999. Optimality Theory. Cambridge: Cambridge University Press. Keller, Rudi. 1994 [11990]. Sprachwandel. Von der unsichtbaren Hand in der Sprache. 2nd ed. Tübingen: Francke. ----------. 1995a. “The epistemic weil”. Subjectivity and subjectivisation. Linguistic perspectives, ed. by Dieter Stein & Susan Wright, 16-30. Cambridge: Cambridge University Press. ----------. 1995b. Zeichentheorie. Zur einer Theorie semiotischen Wissens. Tübingen & Basel: Francke. ----------. 2005. “Sprachwandel als invisible-hand-Phänomen”. Stehl 2005. 2742. Kipling, Rudyard. 1987 [11902]. Just So Stories. London: Penguin. Koch, Peter. 1997. “Diskurstraditionen: zu ihrem sprachtheoretischen Status und ihrer Dynamik”. Frank, Haye & Tophinke 1997. 43-79. ----------. 1999. “Court Records and Cartoons. Reflections of Spontaneous Dialogue in Early Romance Texts”. Historical Dialogue Analysis, ed. by Andreas H. Jucker, Gerd Fritz & Franz Lebsanft, 399-429. Amsterdam & Philadelphia: John Benjamins. ----------. 2001. “Bedeutungswandel und Bezeichnungswandel. Von der kognitiven Semasiologie zur kognitiven Onomasiologie”. Zeitschrift für Literaturwissenschaft und Linguistik 31 (121). 7-36. ----------. 2002. “Diachronische Varietätenlinguistik: extern und intern”. Sprachgeschichte als Varietätengeschichte. Beiträge zur diachronen Varietätenlinguistik des Spanischen und anderer romanischer Sprachen. Anläßlich des 60. Geburtstages von Jens Lüdtke, ed. by Andreas Wesch et al., 3-15. Tübingen: Stauffenburg. ----------. 2004a. “Diachronic cognitive onomasiology and semantic reconstruction”. Lexical Data and Universals of Semantic Change, ed. by Wiltrud Mihatsch & Reinhild Steinberg, 79-106. Tübingen: Stauffenburg. ----------. 2004b. “Sprachwandel, Mündlichkeit und Schriftlichkeit”. Zeitschrift für romanische Philologie 120. 605-630.

248

ESME WINTER-FROEMEL

----------. 2005. “Ein Blick auf die unsichtbare Hand: Kognitive Universalien und historische romanische Lexikologie”. Stehl 2005. 245-275. ------- & Wulf Oesterreicher. 1985. “Sprache der Nähe – Sprache der Distanz. Mündlichkeit und Schriftlichkeit im Spannungsfeld von Sprachtheorie und Sprachgeschichte”. Romanistisches Jahrbuch 36. 15-43. ------- & Wulf Oesterreicher. 1990. Gesprochene Sprache in der Romania: Französisch, Italienisch, Spanisch. Tübingen: Niemeyer. Labov, William. 2001. Principles of Linguistic Change. Vol. 2: Social Factors. Malden, Mass.: Blackwell. Lass, Roger. 1980. On Explaining Language Change. Cambridge et al.: Cambridge University Press. ----------. 1990. “How to do things with junk: exaptation in language evolution”. Journal of Linguistics 26. 79-102. Levinson, Stephen C. 1983. Pragmatics. Cambridge: Cambridge University Press. ----------. 2000. Presumptive Meanings. The Theory of Generalized Conversational Implicature. Cambridge, Mass. & London: The MIT Press. Lüdtke, Helmut. 1989. “Invisible-hand processes and the universal laws of language change”. Language Change. Contributions to the Study of Its Causes, ed. by Leiv Egil Breivik & Ernst Håkon Jahr, 131-136. Berlin; New York: Mouton de Gruyter. Milroy, James & Lesley Milroy. 1985. “Linguistic change, social network and speaker innovation”. Journal of Linguistics 21. 339-384. Minkova, Donka. 1999. “Proximate mechanisms vs. causal explanations”. Zeitschrift für Sprachwissenschaft 18. 226-229. Nettle, Daniel. 1998. “Functionalism and Its Difficulties in Biology and Linguistics”. Darnell et al. 1998. Vol. 1, 445-467. Newmeyer, Frederick J. 1991. “Functional Explanation in Linguistics and the Origins of Language”. Language & Communication 11. 3-28. ----------. 1998. “Some Remarks on the Functionalist–Formalist Controversy in Linguistics”. Darnell et al. 1998. Vol. 1, 469-486. ----------. 1999. “Adaptation, optimality and functional explanation: Two serious problems”. Zeitschrift für Sprachwissenschaft 18. 235-237. ----------. 2001. “Where is functional explanation?”. CLS/Chicago Linguistic Society 37: The Panels. 99-122. ----------. 2003a. “Grammar is grammar and usage is usage”. Language 79. 682-707. ----------. 2003b. “Review article. On nature and language. By Noam Chomsky [...]; The language organ: Linguistics as cognitive psychology. By Stephen R. Anderson and David W. Lightfoot [...]; Language in a Darwinian perspective. By Bernard H. Bichakjian [...]”. Language 79. 583-599. Oesterreicher, Wulf. 1997. “Zur Fundierung von Diskurstraditionen”. Frank, Haye & Tophinke. 1997. 19-41.

TOWARDS A COMPREHENSIVE VIEW OF LANGUAGE CHANGE

249

----------. 2001a. “Historizität – Sprachvariation, Sprachverschiedenheit, Sprachwandel”. Language Typology and Language Universals. An International Handbook, ed. by Martin Haspelmath et al., Vol. 2, 15541595. Berlin & New York: de Gruyter. ----------. 2001b. “Sprachwandel, Varietätenwandel, Sprachgeschichte. Zu einem verdrängten Theoriezusammenhang”. Varieties and Consequences of Literacy and Orality. Formen und Folgen von Schriftlichkeit und Mündlichkeit. Franz H. Bäuml zum 75. Geburtstag, ed. by Ursula Schaefer & Edda Spielmann, 217-248. Tübingen: Narr. Saussure, Ferdinand de. 1969 [11916]. Cours de linguistique générale. Publié par Charles Bally et Albert Sechehaye. Paris: Payot. Stehl, Thomas, ed. 2005. Unsichtbare Hand und Sprecherwahl. Typologie und Prozesse des Sprachwandels in der Romania. Tübingen: Narr. Stolz, Thomas. 1991. “[Review] Rudi Keller, Sprachwandel: Von der unsichtbaren Hand in der Sprache […] 1990 […]”. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 44. 550-554. Tesar, Bruce, Jane Grimshaw & Alan Prince. 1999. “Linguistic and Cognitive Explanation in Optimality Theory”. What is Cognitive Science?, ed. by Ernest Lepore & Zenon Pylyshyn, 295-326. Malden, Mass. & Oxford: Blackwell. Tesch, Gerd. 1978. Linguale Interferenz. Theoretische, terminologische und methodische Grundfragen zu ihrer Erforschung. Tübingen: Narr. Thomason, Sarah Grey. 2003. “Contact as a Source of Language Change”. Joseph & Janda 2003. 687-712. ------- & Terrence Kaufman. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley, Los Angeles, Oxford: University of California Press. Traugott, Elizabeth Closs. 1999. “From a diachronic perspective”. Zeitschrift für Sprachwissenschaft 18. 238-241. ------- & Ekkehard König. 1991. “The Semantics-Pragmatics of Grammaticalization Revisited”. Approaches to Grammaticalization. Vol. 1: Focus on Theoretical and Methodological Issues, ed. by Elizabeth Closs Traugott & Bernd Heine, 189-218. Amsterdam & Philadelphia: John Benjamins. Wegener, Heide. 2004. “Pizzas und Pizzen – die Pluralformen (un)assimilierter Fremdwörter im Deutschen”. Zeitschrift für Sprachwissenschaft 23. 47112. Weinreich, Uriel, William Labov & Marvin I. Herzog. 1968. “Empirical Foundations for a Theory of Language Change”. Directions for Historical Linguistics. A Symposium, ed. by Winfried P. Lehmann & Yakov Malkiel, 95-195. Austin & London: University of Texas Press. Wolgast, Elizabeth. 1984. “The invisible paw”. The Monist 67. 229-250. Wurzel, Wolfgang Ullrich. 1997. “Natürlicher Grammatischer Wandel, ‘unsichtbare Hand’ und Sprachökonomie – Wollen wir wirklich so Grundverschiedenes?”. Vergleichende germanische Philologie und Skandina-

250

ESME WINTER-FROEMEL

vistik. Festschrift für Otmar Werner, ed. by Thomas Birkmann et al., 295308. Tübingen: Niemeyer. ----------. 1999. “Principles of Evaluation, Change and Related Issues”. Zeitschrift für Sprachwissenschaft 18. 242-250.

SUBJECT INDEX

A. absence of change, conservatism 228, 235, 239 acquisition 108, 116, 118, 119 acquisition-based change 1-2, 14, 15, 22-23 agreement 129, 130, 134, 140, 142, 182, 188-191, 195-200, 207, 209-211 Agreement (Minimalist Program) 64, 6 analogy 128, 140, 142 aorist/aoristic past tense 150, 151 apprehension 60 auxiliary verb 137, 140

B. bare noun

58, 63

C. categorial reorganisation 174, 175 circularity of explanation 223, 234–35, 239 classification 60, 62, 69 cleft 31-52, 204, 209, 210 cleft, informative-presupposition 31, 33, 49-52 clitic doubling 198, 202n, 206, 208 clitic pronoun 92-94, 129, 130, 131, 139, 140, 142, 143 cognitive-semantic motivation 171, 174 competing grammars 2, 116-117, 122 conditional clause 152, 154, 155 construction Grammar 32 contrast 31, 48-49 convention 233, 235, 237, 238, 243, 245 copula verb 130, 135, 137, 138, 142, 143 correlations 102 countable noun 63, 67, 78 counterfactual 156, 164, 169 cross-reference 183, 184, 187, 190, 191, 193, 195-197, 206-211 D. dependent-marking 191-201, 204, 20607 diglossia 118 dislocation 204, 208

dysfunctional changes 239

227, 230, 236,

E. Emerging Grammar 2 empty category 181, 182, 186, 188, 189 entrenchment 20, 21-22, 22-23, 28, 246 ergative case (Basque) 183, 184, 186, 187, 190., 191, 196 est-ce que interrogative particle 16-23, 51 expletive subject 34 F. focus 31-33, 39-40, 42-44, 47-49, 51-52 formalist-functionalist debate 1, 223, 241 frequency 3, 22, 27, 28, 240 G. grammar-usage controversy 241 grammatical reconstruction 103 grammaticalization 35, 52, 78 H. head-marking

191-207

I. impersonal 127-145 indefiniteness 57, 58 Inertia Theory 2, 13-15, 21, 22-23, 2728 inferential (sentence) 36-37, 48 infinitival complement 94, 97, 100-102 Infl(ection) 89, 91, 98, 101, 103 information structure 2 see also contrast, focus, topiccomment innovation 222, 231, 233, 234, 237, 238, 239, 243, 245, 246, 247 inversion, subject-verb 35-36, 44 Invisible-Hand explanations 3, 219, 220, 232, 233 Irrealis 156, 165, 173 L. language contact 127, 135, 140, 143 linking Rules 28

SUBJECT INDEX

252

M. markedness 2, 108, 118-120, 229 mass noun 58, 67 mention 196, 197, 200, 210 Merge (Minimalist Program) 108, 119121 Minimal Link Condition 110 mood system 147-179 Move (Minimalist Program) 108, 119122 N. null object language 196, 210 O. object conjugation Optimality Theory

182, 186, 188, 189,

191, 202, 204 229, 231

P. paradigm alteration 86 paradigm coverage 87 paradox of change 1, 102, 217, 228, 234, 247 parameter 103 parameter 107, 116, 119, 120 partitive article 58, 63 passive construction 127-145 pluperfect 147-179 predictability of change 86, 102 Preferred argument structure constraint 189 presentational construction 23-28 Principle of reference (reanalysis) 20-23 Principle of transparency (reanalysis) 26-27 Probe 65, 66 pro-drop language 183, 186, 196, 199, 200

propagation of an innovation 222, 231, 233, 234, 237, 238, 239, 240, 242, 247 prosodic change 42-43 prototype 171, 174, 175 R. reanalysis 20-23, 26-28, 77, 107, 119, 162, 171 reflexive pronoun 127-145 rule generalization 127, 128, 140, 142, 143 S. semilattice structure 65 spoken vs. written language 39 stressed-focus cleft see cleft, stressedfocus stylistic fronting 107, 108, 114, 115, 121, 122 subject, expletive see expletive subject syntactic change 43-48 T. teleology problem 223, 231, 234 temporal system, Reichenbach 147-179 topicalisation 108, 109, 111 topic-comment 49 transitive verb 130, 131, 134, 139 trigger of change 2, 14, 15, 23, 107, 116, 117, 120 U. unaccusative verb 130, 134, 135, 136, 137, 138, 140, 143 unergative verb 130, 134, 135, 139 usage-based change 14-15, 22-23 V. verb-third

107

CURRENT ISSUES IN LINGUISTIC THEORY

E. F. K. Koerner, Editor

Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin [email protected] Current Issues in Linguistic Theory (CILT) is a theory-oriented series which welcomes contributions from scholars who have significant proposals to make towards the advancement of our understanding of language, its structure, functioning and development. CILT has been established in order to provide a forum for the presentation and discussion of linguistic opinions of scholars who do not necessarily accept the prevailing mode of thought in linguistic science. It offers an outlet for meaningful contributions to the current linguistic debate, and furnishes the diversity of opinion which a healthy discipline must have. A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 294 Frellesvig, Bjarke and John Whitman (eds.): Proto-Japanese. vii, 231 pp. + index. Expected February 2008 293 Detges, Ulrich and Richard Waltereit (eds.): The Paradox of Grammatical Change. Perspectives from Romance. 2008. vi, 252 pp. 292 Nicolov, Nicolas, Kalina Bontcheva, Galia Angelova and Ruslan Mitkov (eds.): Recent Advances in Natural Language Processing IV. Selected papers from RANLP 2005. 2007. xii, 307 pp. 291 Baauw, Sergio, Frank Drijkoningen and Manuela Pinto (eds.): Romance Languages and Linguistic Theory 2005. Selected papers from ‘Going Romance’, Utrecht, 8–10 December 2005. 2007. viii, 338 pp. 290 Mughazy, Mustafa A. (ed.): Perspectives on Arabic Linguistics XX. Papers from the twentieth annual symposium on Arabic linguistics, Kalamazoo, Michigan, March 2006. 2007. xii, 247 pp. 289 Benmamoun, Elabbas (ed.): Perspectives on Arabic Linguistics XIX. Papers from the nineteenth annual symposium on Arabic Linguistics, Urbana, Illinois, April 2005. 2007. xiv, 304 pp. 288 Toivonen, Ida and Diane Nelson (eds.): Saami Linguistics. 2007. viii, 321 pp. 287 Camacho, José, Nydia Flores-Ferrán, Liliana Sánchez, Viviane Déprez and María José Cabrera (eds.): Romance Linguistics 2006. Selected papers from the 36th Linguistic Symposium on Romance Languages (LSRL), New Brunswick, March-April 2006. 2007. viii, 340 pp. 286 Weijer, Jeroen van de and Erik Jan van der Torre (eds.): Voicing in Dutch. (De)voicing – phonology, phonetics, and psycholinguistics. 2007. x, 186 pp. 285 Sackmann, Robin (ed.): Explorations in Integrational Linguistics. Four essays on German, French, and Guaraní. ix, 217 pp. Expected January 2008 284 Salmons, Joseph C. and Shannon Dubenion-Smith (eds.): Historical Linguistics 2005. Selected papers from the 17th International Conference on Historical Linguistics, Madison, Wisconsin, 31 July - 5 August 2005. 2007. viii, 413 pp. 283 Lenker, Ursula and Anneli Meurman-Solin (eds.): Connectives in the History of English. 2007. viii, 318 pp. 282 Prieto, Pilar, Joan Mascaró and Maria-Josep Solé (eds.): Segmental and prosodic issues in Romance phonology. 2007. xvi, 262 pp. 281 Vermeerbergen, Myriam, Lorraine Leeson and Onno Crasborn (eds.): Simultaneity in Signed Languages. Form and function. 2007. viii, 360 pp. (incl. CD-Rom). 280 Hewson, John and Vit Bubenik: From Case to Adposition. The development of configurational syntax in Indo-European languages. 2006. xxx, 420 pp. 279 Nedergaard Thomsen, Ole (ed.): Competing Models of Linguistic Change. Evolution and beyond. 2006. vi, 344 pp. 278 Doetjes, Jenny and Paz González (eds.): Romance Languages and Linguistic Theory 2004. Selected papers from ‘Going Romance’, Leiden, 9–11 December 2004. 2006. viii, 320 pp. 277 Helasvuo, Marja-Liisa and Lyle Campbell (eds.): Grammar from the Human Perspective. Case, space and person in Finnish. 2006. x, 280 pp. 276 Montreuil, Jean-Pierre Y. (ed.): New Perspectives on Romance Linguistics. Vol. II: Phonetics, Phonology and Dialectology. Selected papers from the 35th Linguistic Symposium on Romance Languages (LSRL), Austin, Texas, February 2005. 2006. x, 213 pp. 275 Nishida, Chiyo and Jean-Pierre Y. Montreuil (eds.): New Perspectives on Romance Linguistics. Vol. I: Morphology, Syntax, Semantics, and Pragmatics. Selected papers from the 35th Linguistic Symposium on Romance Languages (LSRL), Austin, Texas, February 2005. 2006. xiv, 288 pp.

274 Gess, Randall S. and Deborah Arteaga (eds.): Historical Romance Linguistics. Retrospective and perspectives. 2006. viii, 393 pp. 273 Filppula, Markku, Juhani Klemola, Marjatta Palander and Esa Penttilä (eds.): Dialects Across Borders. Selected papers from the 11th International Conference on Methods in Dialectology (Methods XI), Joensuu, August 2002. 2005. xii, 291 pp. 272 Gess, Randall S. and Edward J. Rubin (eds.): Theoretical and Experimental Approaches to Romance Linguistics. Selected papers from the 34th Linguistic Symposium on Romance Languages (LSRL), Salt Lake City, March 2004. 2005. viii, 367 pp. 271 Branner, David Prager (ed.): The Chinese Rime Tables. Linguistic philosophy and historicalcomparative phonology. 2006. viii, 358 pp. 270 Geerts, Twan, Ivo van Ginneken and Haike Jacobs (eds.): Romance Languages and Linguistic Theory 2003. Selected papers from ‘Going Romance’ 2003, Nijmegen, 20–22 November. 2005. viii, 369 pp. 269 Hargus, Sharon and Keren Rice (eds.): Athabaskan Prosody. 2005. xii, 432 pp. 268 Cravens, Thomas D. (ed.): Variation and Reconstruction. 2006. viii, 223 pp. 267 Alhawary, Mohammad T. and Elabbas Benmamoun (eds.): Perspectives on Arabic Linguistics XVII–XVIII. Papers from the seventeenth and eighteenth annual symposia on Arabic linguistics. Volume XVII–XVIII: Alexandria, 2003 and Norman, Oklahoma 2004. 2005. xvi, 315 pp. 266 Boudelaa, Sami (ed.): Perspectives on Arabic Linguistics XVI. Papers from the sixteenth annual symposium on Arabic linguistics, Cambridge, March 2002. 2006. xii, 181 pp. 265 Cornips, Leonie and Karen P. Corrigan (eds.): Syntax and Variation. Reconciling the Biological and the Social. 2005. vi, 312 pp. 264 Dressler, Wolfgang U., Dieter Kastovsky, Oskar E. Pfeiffer and Franz Rainer (eds.): Morphology and its demarcations. Selected papers from the 11th Morphology meeting, Vienna, February 2004. With the assistance of Francesco Gardani and Markus A. Pöchtrager. 2005. xiv, 320 pp. 263 Branco, António, Tony McEnery and Ruslan Mitkov (eds.): Anaphora Processing. Linguistic, cognitive and computational modelling. 2005. x, 449 pp. 262 Vajda, Edward J. (ed.): Languages and Prehistory of Central Siberia. 2004. x, 275 pp. 261 Kay, Christian J. and Jeremy J. Smith (eds.): Categorization in the History of English. 2004. viii, 268 pp. 260 Nicolov, Nicolas, Kalina Bontcheva, Galia Angelova and Ruslan Mitkov (eds.): Recent Advances in Natural Language Processing III. Selected papers from RANLP 2003. 2004. xii, 402 pp. 259 Carr, Philip, Jacques Durand and Colin J. Ewen (eds.): Headhood, Elements, Specification and Contrastivity. Phonological papers in honour of John Anderson. 2005. xxviii, 405 pp. 258 Auger, Julie, J. Clancy Clements and Barbara Vance (eds.): Contemporary Approaches to Romance Linguistics. Selected Papers from the 33rd Linguistic Symposium on Romance Languages (LSRL), Bloomington, Indiana, April 2003. With the assistance of Rachel T. Anderson. 2004. viii, 404 pp. 257 Fortescue, Michael, Eva Skafte Jensen, Jens Erik Mogensen and Lene Schøsler (eds.): Historical Linguistics 2003. Selected papers from the 16th International Conference on Historical Linguistics, Copenhagen, 11–15 August 2003. 2005. x, 312 pp. 256 Bok-Bennema, Reineke, Bart Hollebrandse, Brigitte Kampers-Manhe and Petra Sleeman (eds.): Romance Languages and Linguistic Theory 2002. Selected papers from ‘Going Romance’, Groningen, 28–30 November 2002. 2004. viii, 273 pp. 255 Meulen, Alice ter and Werner Abraham (eds.): The Composition of Meaning. From lexeme to discourse. 2004. vi, 232 pp. 254 Baldi, Philip and Pietro U. Dini (eds.): Studies in Baltic and Indo-European Linguistics. In honor of William R. Schmalstieg. 2004. xlvi, 302 pp. 253 Caffarel, Alice, J.R. Martin and Christian M.I.M. Matthiessen (eds.): Language Typology. A functional perspective. 2004. xiv, 702 pp. 252 Kay, Christian J., Carole Hough and Irené Wotherspoon (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume II: Lexis and Transmission. 2004. xii, 273 pp. 251 Kay, Christian J., Simon Horobin and Jeremy J. Smith (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume I: Syntax and Morphology. 2004. x, 264 pp. 250 Jensen, John T.: Principles of Generative Phonology. An introduction. 2004. xii, 324 pp. 249 Bowern, Claire and Harold Koch (eds.): Australian Languages. Classification and the comparative method. 2004. xii, 377 pp. (incl. CD-Rom). 248 Weigand, Edda (ed.): Emotion in Dialogic Interaction. Advances in the complex. 2004. xii, 284 pp. 247 Parkinson, Dilworth B. and Samira Farwaneh (eds.): Perspectives on Arabic Linguistics XV. Papers from the Fifteenth Annual Symposium on Arabic Linguistics, Salt Lake City 2001. 2003. x, 214 pp.

246 Holisky, Dee Ann and Kevin Tuite (eds.): Current Trends in Caucasian, East European and Inner Asian Linguistics. Papers in honor of Howard I. Aronson. 2003. xxviii, 426 pp. 245 Quer, Josep, Jan Schroten, Mauro Scorretti, Petra Sleeman and Els Verheugd (eds.): Romance Languages and Linguistic Theory 2001. Selected papers from 'Going Romance', Amsterdam, 6–8 December 2001. 2003. viii, 355 pp. 244 Pérez-Leroux, Ana Teresa and Yves Roberge (eds.): Romance Linguistics. Theory and Acquisition. Selected papers from the 32nd Linguistic Symposium on Romance Languages (LSRL), Toronto, April 2002. 2003. viii, 388 pp. 243 Cuyckens, Hubert, Thomas Berg, René Dirven and Klaus-Uwe Panther (eds.): Motivation in Language. Studies in honor of Günter Radden. 2003. xxvi, 403 pp. 242 Seuren, Pieter A.M. and Gerard Kempen (eds.): Verb Constructions in German and Dutch. 2003. vi, 316 pp. 241 Lecarme, Jacqueline (ed.): Research in Afroasiatic Grammar II. Selected papers from the Fifth Conference on Afroasiatic Languages, Paris, 2000. 2003. viii, 550 pp. 240 Janse, Mark and Sijmen Tol (eds.): Language Death and Language Maintenance. Theoretical, practical and descriptive approaches. With the assistance of Vincent Hendriks. 2003. xviii, 244 pp. 239 Andersen, Henning (ed.): Language Contacts in Prehistory. Studies in Stratigraphy. Papers from the Workshop on Linguistic Stratigraphy and Prehistory at the Fifteenth International Conference on Historical Linguistics, Melbourne, 17 August 2001. 2003. viii, 292 pp. 238 Núñez-Cedeño, Rafael, Luis López and Richard Cameron (eds.): A Romance Perspective on Language Knowledge and Use. Selected papers from the 31st Linguistic Symposium on Romance Languages (LSRL), Chicago, 19–22 April 2001. 2003. xvi, 386 pp. 237 Blake, Barry J. and Kate Burridge (eds.): Historical Linguistics 2001. Selected papers from the 15th International Conference on Historical Linguistics, Melbourne, 13–17 August 2001. Editorial assistance Jo Taylor. 2003. x, 444 pp. 236 Simon-Vandenbergen, Anne-Marie, Miriam Taverniers and Louise J. Ravelli (eds.): Grammatical Metaphor. Views from systemic functional linguistics. 2003. vi, 453 pp. 235 Linn, Andrew R. and Nicola McLelland (eds.): Standardization. Studies from the Germanic languages. 2002. xii, 258 pp. 234 Weijer, Jeroen van de, Vincent J. van Heuven and Harry van der Hulst (eds.): The Phonological Spectrum. Volume II: Suprasegmental structure. 2003. x, 264 pp. 233 Weijer, Jeroen van de, Vincent J. van Heuven and Harry van der Hulst (eds.): The Phonological Spectrum. Volume I: Segmental structure. 2003. x, 308 pp. 232 Beyssade, Claire, Reineke Bok-Bennema, Frank Drijkoningen and Paola Monachesi (eds.): Romance Languages and Linguistic Theory 2000. Selected papers from ‘Going Romance’ 2000, Utrecht, 30 November–2 December. 2002. viii, 354 pp. 231 Cravens, Thomas D.: Comparative Historical Dialectology. Italo-Romance clues to Ibero-Romance sound change. 2002. xii, 163 pp. 230 Parkinson, Dilworth B. and Elabbas Benmamoun (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XIII-XIV: Stanford, 1999 and Berkeley, California 2000. 2002. xiv, 250 pp. 229 Nevin, Bruce E. and Stephen B. Johnson (eds.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 2: Mathematics and computability of language. 2002. xx, 312 pp. 228 Nevin, Bruce E. (ed.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 1: Philosophy of science, syntax and semantics. 2002. xxxvi, 323 pp. 227 Fava, Elisabetta (ed.): Clinical Linguistics. Theory and applications in speech pathology and therapy. 2002. xxiv, 353 pp. 226 Levin, Saul: Semitic and Indo-European. Volume II: Comparative morphology, syntax and phonetics. 2002. xviii, 592 pp. 225 Shahin, Kimary N.: Postvelar Harmony. 2003. viii, 344 pp. 224 Fanego, Teresa, Belén Méndez-Naya and Elena Seoane (eds.): Sounds, Words, Texts and Change. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 2. 2002. x, 310 pp. 223 Fanego, Teresa, Javier Pérez-Guerra and María José López-Couso (eds.): English Historical Syntax and Morphology. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 1. 2002. x, 306 pp. 222 Herschensohn, Julia, Enrique Mallén and Karen Zagona (eds.): Features and Interfaces in Romance. Essays in honor of Heles Contreras. 2001. xiv, 302 pp. 221 D’hulst, Yves, Johan Rooryck and Jan Schroten (eds.): Romance Languages and Linguistic Theory 1999. Selected papers from ‘Going Romance’ 1999, Leiden, 9–11 December 1999. 2001. viii, 406 pp.

220 Satterfield, Teresa, Christina M. Tortora and Diana Cresti (eds.): Current Issues in Romance Languages. Selected papers from the 29th Linguistic Symposium on Romance Languages (LSRL), Ann Arbor, 8–11 April 1999. 2002. viii, 412  pp. 219 Andersen, Henning (ed.): Actualization. Linguistic Change in Progress. Papers from a workshop held at the 14th International Conference on Historical Linguistics, Vancouver, B.C., 14 August 1999. 2001. vii, 250 pp. 218 Bendjaballah, Sabrina, Wolfgang U. Dressler, Oskar E. Pfeiffer and Maria D. Voeikova (eds.): Morphology 2000. Selected papers from the 9th Morphology Meeting, Vienna, 24–28 February 2000. 2002. viii, 317 pp. 217 Wiltshire, Caroline R. and Joaquim Camps (eds.): Romance Phonology and Variation. Selected papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000. 2002. xii, 238 pp. 216 Camps, Joaquim and Caroline R. Wiltshire (eds.): Romance Syntax, Semantics and L2 Acquisition. Selected papers from the 30th Linguistic Symposium on Romance Languages, Gainesville, Florida, February 2000. 2001. xii, 246 pp. 215 Brinton, Laurel J. (ed.): Historical Linguistics 1999. Selected papers from the 14th International Conference on Historical Linguistics, Vancouver, 9–13 August 1999. 2001. xii, 398 pp. 214 Weigand, Edda and Marcelo Dascal (eds.): Negotiation and Power in Dialogic Interaction. 2001. viii, 303 pp. 213 Sornicola, Rosanna, Erich Poppe and Ariel Shisha-Halevy (eds.): Stability, Variation and Change of Word-Order Patterns over Time. With the assistance of Paola Como. 2000. xxxii, 323 pp. 212 Repetti, Lori (ed.): Phonological Theory and the Dialects of Italy. 2000. x, 301 pp. 211 Elšík, Viktor and Yaron Matras (eds.): Grammatical Relations in Romani. The Noun Phrase. with a Foreword by Frans Plank (Universität Konstanz). 2000. x, 244 pp. 210 Dworkin, Steven N. and Dieter Wanner (eds.): New Approaches to Old Problems. Issues in Romance historical linguistics. 2000. xiv, 235 pp. 209 King, Ruth: The Lexical Basis of Grammatical Borrowing. A Prince Edward Island French case study. 2000. xvi, 241 pp. 208 Robinson, Orrin W.: Whose German? The ach/ich alternation and related phenomena in ‘standard’ and ‘colloquial’. 2001. xii, 178 pp. 207 Sanz, Montserrat: Events and Predication. A new approach to syntactic processing in English and Spanish. 2000. xiv, 219 pp. 206 Fawcett, Robin P.: A Theory of Syntax for Systemic Functional Linguistics. 2000. xxiv, 360 pp. 205 Dirven, René, Roslyn M. Frank and Cornelia Ilie (eds.): Language and Ideology. Volume 2: descriptive cognitive approaches. 2001. vi, 264 pp. 204 Dirven, René, Bruce Hawkins and Esra Sandikcioglu (eds.): Language and Ideology. Volume 1: theoretical cognitive approaches. 2001. vi, 301 pp. 203 Norrick, Neal R.: Conversational Narrative. Storytelling in everyday talk. 2000. xiv, 233 pp. 202 Lecarme, Jacqueline, Jean Lowenstamm and Ur Shlonsky (eds.): Research in Afroasiatic Grammar. Papers from the Third conference on Afroasiatic Languages, Sophia Antipolis, 1996. 2000. vi, 386 pp. 201 Dressler, Wolfgang U., Oskar E. Pfeiffer, Markus A. Pöchtrager and John R. Rennison (eds.): Morphological Analysis in Comparison. 2000. x, 261 pp. 200 Anttila, Raimo: Greek and Indo-European Etymology in Action. Proto-Indo-European *aǵ-. 2000. xii, 314 pp. 199 Pütz, Martin and Marjolijn H. Verspoor (eds.): Explorations in Linguistic Relativity. 2000. xvi, 369 pp. 198 Niemeier, Susanne and René Dirven (eds.): Evidence for Linguistic Relativity. 2000. xxii, 240 pp. 197 Coopmans, Peter, Martin Everaert and Jane Grimshaw (eds.): Lexical Specification and Insertion. 2000. xviii, 476 pp. 196 Hannahs, S.J. and Mike Davenport (eds.): Issues in Phonological Structure. Papers from an International Workshop. 1999. xii, 268 pp. 195 Herring, Susan C., Pieter van Reenen and Lene Schøsler (eds.): Textual Parameters in Older Languages. 2001. x, 448 pp. 194 Coleman, Julie and Christian J. Kay (eds.): Lexicology, Semantics and Lexicography. Selected papers from the Fourth G. L. Brook Symposium, Manchester, August 1998. 2000. xiv, 257 pp. 193 Klausenburger, Jurgen: Grammaticalization. Studies in Latin and Romance morphosyntax. 2000. xiv, 184 pp.

192 Alexandrova, Galina M. and Olga Arnaudova (eds.): The Minimalist Parameter. Selected papers from the Open Linguistics Forum, Ottawa, 21–23 March 1997. 2001. x, 360 pp. 191 Sihler, Andrew L.: Language History. An introduction. 2000. xvi, 298 pp. 190 Benmamoun, Elabbas (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XII: Urbana-Champaign, Illinois, 1998. 1999. viii, 204 pp. 189 Nicolov, Nicolas and Ruslan Mitkov (eds.): Recent Advances in Natural Language Processing II. Selected papers from RANLP ’97. 2000. xi, 422 pp. 188 Simmons, Richard VanNess: Chinese Dialect Classification. A comparative approach to Harngjou, Old Jintarn, and Common Northern Wu. 1999. xviii, 317 pp. 187 Franco, Jon A., Alazne Landa and Juan Martín (eds.): Grammatical Analyses in Basque and Romance Linguistics. Papers in honor of Mario Saltarelli. 1999. viii, 306 pp. 186 Mišeska Tomić, Olga and Milorad Radovanović (eds.): History and Perspectives of Language Study. Papers in honor of Ranko Bugarski. 2000. xxii, 314 pp. 185 Authier, Jean-Marc, Barbara E. Bullock and Lisa A. Reed (eds.): Formal Perspectives on Romance Linguistics. Selected papers from the 28th Linguistic Symposium on Romance Languages (LSRL XXVIII), University Park, 16–19 April 1998. 1999. xii, 334 pp. 184 Sagart, Laurent: The Roots of Old Chinese. 1999. xii, 272 pp. 183 Contini-Morava, Ellen and Yishai Tobin (eds.): Between Grammar and Lexicon. 2000. xxxii, 365 pp. 182 Kenesei, István (ed.): Crossing Boundaries. Advances in the theory of Central and Eastern European languages. 1999. viii, 302 pp. 181 Mohammad, Mohammad A.: Word Order, Agreement and Pronominalization in Standard and Palestinian Arabic. 2000. xvi, 197 pp. 180 Mereu, Lunella (ed.): Boundaries of Morphology and Syntax. 1999. viii, 314 pp. 179 Rini, Joel: Exploring the Role of Morphology in the Evolution of Spanish. 1999. xvi, 187 pp. 178 Foolen, Ad and Frederike van der Leek (eds.): Constructions in Cognitive Linguistics. Selected papers from the Fifth International Cognitive Linguistics Conference, Amsterdam, 1997. 2000. xvi, 338 pp. 177 Cuyckens, Hubert and Britta E. Zawada (eds.): Polysemy in Cognitive Linguistics. Selected papers from the International Cognitive Linguistics Conference, Amsterdam, 1997. 2001. xxviii, 296 pp. 176 Van Hoek, Karen, Andrej A. Kibrik and Leo Noordman (eds.): Discourse Studies in Cognitive Linguistics. Selected papers from the 5th International Cognitive Linguistics Conference, Amsterdam, July 1997. 1999. vi, 187 pp. 175 Gibbs, Jr., Raymond W. and Gerard J. Steen (eds.): Metaphor in Cognitive Linguistics. Selected papers from the 5th International Cognitive Linguistics Conference, Amsterdam, 1997. 1999. viii, 226 pp. 174 Hall, T. Alan and Ursula Kleinhenz (eds.): Studies on the Phonological Word. 1999. viii, 298 pp. 173 Treviño, Esthela and José Lema (eds.): Semantic Issues in Romance Syntax. 1999. viii, 309 pp. 172 Dimitrova-Vulchanova, Mila and Lars Hellan (eds.): Topics in South Slavic Syntax and Semantics. 1999. xxviii, 263 pp. 171 Weigand, Edda (ed.): Contrastive Lexical Semantics. 1998. x, 270 pp. 170 Lamb, Sydney M.: Pathways of the Brain. The neurocognitive basis of language. 1999. xii, 418 pp. 169 Ghadessy, Mohsen (ed.): Text and Context in Functional Linguistics. 1999. xviii, 340 pp. 168 Ratcliffe, Robert R.: The “Broken” Plural Problem in Arabic and Comparative Semitic. Allomorphy and analogy in non-concatenative morphology. 1998. xii, 261 pp. 167 Benmamoun, Elabbas, Mushira Eid and Niloofar Haeri (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XI: Atlanta, Georgia, 1997. 1998. viii, 231 pp. 166 Lemmens, Maarten: Lexical Perspectives on Transitivity and Ergativity. Causative constructions in English. 1998. xii, 268 pp. 165 Bubenik, Vit: A Historical Syntax of Late Middle Indo-Aryan (Apabhraṃśa). 1998. xxiv, 265 pp. 164 Schmid, Monika S., Jennifer R. Austin and Dieter Stein (eds.): Historical Linguistics 1997. Selected papers from the 13th International Conference on Historical Linguistics, Düsseldorf, 10–17 August 1997. 1998. x, 409 pp. 163 Lockwood, David G., Peter H. Fries and James E. Copeland (eds.): Functional Approaches to Language, Culture and Cognition. Papers in honor of Sydney M. Lamb. 2000. xxxiv, 656 pp. 162 Hogg, Richard M. and Linda van Bergen (eds.): Historical Linguistics 1995. Volume 2: Germanic linguistics.. Selected papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995. 1998. x, 365 pp.

161 Smith, John Charles and Delia Bentley (eds.): Historical Linguistics 1995. Volume 1: General issues and non-Germanic Languages.. Selected papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995. 2000. xii, 438 pp. 160 Schwegler, Armin, Bernard Tranel and Myriam Uribe-Etxebarria (eds.): Romance Linguistics: Theoretical Perspectives. Selected papers from the 27th Linguistic Symposium on Romance Languages (LSRL XXVII), Irvine, 20–22 February, 1997. 1998. vi, 349 pp. + index. 159 Joseph, Brian D., Geoffrey C. Horrocks and Irene Philippaki-Warburton (eds.): Themes in Greek Linguistics II. 1998. x, 335 pp. 158 Sánchez-Macarro, Antonia and Ronald Carter (eds.): Linguistic Choice across Genres. Variation in spoken and written English. 1998. viii, 338 pp. 157 Lema, José and Esthela Treviño (eds.): Theoretical Analyses on Romance Languages. Selected papers from the 26th Linguistic Symposium on Romance Languages (LSRL XXVI), Mexico City, 28–30 March, 1996. 1998. viii, 380 pp. 156 Matras, Yaron, Peter Bakker and Hristo Kyuchukov (eds.): The Typology and Dialectology of Romani. 1997. xxxii, 223 pp. 155 Forget, Danielle, Paul Hirschbühler, France Martineau and María Luisa Rivero (eds.): Negation and Polarity. Syntax and semantics. Selected papers from the colloquium Negation: Syntax and Semantics. Ottawa, 11–13 May 1995. 1997. viii, 367 pp. 154 Simon-Vandenbergen, Anne-Marie, Kristin Davidse and Dirk Noël (eds.): Reconnecting Language. Morphology and Syntax in Functional Perspectives. 1997. xiii, 339 pp. 153 Eid, Mushira and Robert R. Ratcliffe (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume X: Salt Lake City, 1996. 1997. vii, 296 pp. 152 Hiraga, Masako K., Christopher Sinha and Sherman Wilcox (eds.): Cultural, Psychological and Typological Issues in Cognitive Linguistics. Selected papers of the bi-annual ICLA meeting in Albuquerque, July 1995. 1999. viii, 338 pp. 151 Liebert, Wolf-Andreas, Gisela Redeker and Linda R. Waugh (eds.): Discourse and Perspective in Cognitive Linguistics. 1997. xiv, 270 pp. 150 Verspoor, Marjolijn H., Kee Dong Lee and Eve Sweetser (eds.): Lexical and Syntactical Constructions and the Construction of Meaning. Proceedings of the Bi-annual ICLA meeting in Albuquerque, July 1995. 1997. xii, 454 pp. 149 Hall, T. Alan: The Phonology of Coronals. 1997. x, 176 pp. 148 Wolf, George and Nigel Love (eds.): Linguistics Inside Out. Roy Harris and his critics. 1997. xxviii, 344 pp. 147 Hewson, John: The Cognitive System of the French Verb. 1997. xii, 187 pp. 146 Hinskens, Frans, Roeland van Hout and W. Leo Wetzels (eds.): Variation, Change, and Phonological Theory. 1997. x, 314 pp. 145 Hewson, John and Vit Bubenik: Tense and Aspect in Indo-European Languages. Theory, typology, diachrony. 1997. xii, 403 pp. 144 Singh, Rajendra (ed.): Trubetzkoy's Orphan. Proceedings of the Montréal Roundtable on “Morphonology: contemporary responses” (Montréal, October 1994). In collaboration with Richard Desrochers. 1996. xiv, 363 pp. 143 Athanasiadou, Angeliki and René Dirven (eds.): On Conditionals Again. 1997. viii, 418 pp. 142 Salmons, Joseph C. and Brian D. Joseph (eds.): Nostratic. Sifting the Evidence. 1998. vi, 293 pp. 141 Eid, Mushira and Dilworth B. Parkinson (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume IX: Washington D.C., 1995. 1996. xiii, 249 pp. 140 Black, James R. and Virginia Motapanyane (eds.): Clitics, Pronouns and Movement. 1997. 375 pp. 139 Black, James R. and Virginia Motapanyane (eds.): Microparametric Syntax and Dialect Variation. 1996. xviii, 269 pp. 138 Sackmann, Robin and Monika Budde (eds.): Theoretical Linguistics and Grammatical Description. Papers in honour of Hans-Heinrich Lieb. 1996. x, 375 pp. 137 Lippi-Green, Rosina L. and Joseph C. Salmons (eds.): Germanic Linguistics. Syntactic and diachronic. 1996. viii, 192 pp. 136 Mitkov, Ruslan and Nicolas Nicolov (eds.): Recent Advances in Natural Language Processing. Selected Papers from RANLP ’95. 1997. xii, 472 pp. 135 Britton, Derek (ed.): English Historical Linguistics 1994. Papers from the 8th International Conference on English Historical Linguistics (8 ICEHL, Edinburgh, 19–23 September 1994). 1996. viii, 403 pp. 134 Eid, Mushira (ed.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume VIII: Amherst, Massachusetts 1994. 1996. vii, 261 pp.