Cyclical Change (Linguistik Aktuell Linguistics Today)

Cyclical Change

Linguistik Aktuell/Linguistics Today (LA) Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective.

General Editors Werner Abraham

University of Vienna / Rijksuniversiteit Groningen

Elly van Gelderen

Arizona State University

Advisory Editorial Board Cedric Boeckx

Christer Platzack

Guglielmo Cinque

Ian Roberts

Günther Grewendorf

Lisa deMena Travis

Liliane Haegeman

Sten Vikner

Hubert Haider

C. Jan-Wouter Zwart

ICREA/Universitat Autònoma de Barcelona University of Venice

J.W. Goethe-University, Frankfurt University of Lille, France University of Salzburg

Terje Lohndal

University of Maryland

Volume 146 Cyclical Change Edited by Elly van Gelderen

University of Lund

Cambridge University McGill University

University of Aarhus University of Groningen

Cyclical Change

Edited by

Elly van Gelderen Arizona State University

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Cyclical change / edited by Elly van Gelderen. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166-0829 ; v. 146) Includes bibliographical references and index. 1.  Grammar, Comparative and general--Syntax. 2.  Linguistic change.  I. Gelderen, Elly van. P291.C885

2009

417'.7--dc22 isbn 978 90 272 5529 7 (hb; alk. paper) isbn 978 90 272 8921 6 (eb)

2009017906

© 2009 – 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

Table of contents List of contributors chapter 1 Cyclical change, an introduction Elly van Gelderen

vii

1

part i.  Negatives chapter 2 Jespersen recycled Jack Hoeksema

15

chapter 3 The Jespersen Cycles Johan van der Auwera

35

chapter 4 The negative cycle in early and modern Russian Olena Tsurska

73

chapter 5 Jespersen off course?:The case of contemporary Afrikaans negation Theresa Biberauer

91

part ii.  Pronouns, agreement, and topic markers chapter 6 Weak pronouns in Italian:Instances of a broken cycle? Diana Vedovato

133

chapter 7 The subject cycle of pronominal auxiliaries in old North Russian Kyongjoon Kwon

157



Cyclical Change

chapter 8 Two instances of a broken cycle:Sentential particles in Old Italian Cecilia Poletto

185

part iii  Copulas, auxiliaries, and adpositions chapter 9 The copula cycle Terje Lohndal

209

chapter 10 Rather: On a modal cycle Remus Gergel

243

chapter 11 Cycles of complementation in the Mayan languages Clifton Pye

265

chapter 12 The preposition cycle in English Cathleen Waters

285

part iv  An experiment chapter 13 The study of syntactic cycles as an experimental science Roeland Hancock and Thomas G. Bever

303

Author index

323

Subject index

327

List of contributors Johan van der Auwera Department of Linguistics Antwerp University Rodestraat 14  2000 Antwerpen, Belgium [email protected] Thomas G. Bever Department of Linguistics University of Arizona Tucson, AZ 85721, USA [email protected] Theresa Biberauer Faculty of Modern and Medieval Languages University of Cambridge Sidgwick Avenue Cambridge CB3 9DA, U.K. [email protected] Elly van Gelderen Department of English Arizona State University Tempe, AZ 85287–0302, USA [email protected] Remus Gergel University of Tübingen English Department Wilhelmstr. 50, R. 407 D-72074 Tübingen, Germany [email protected]

Roeland Hancock Department of Linguistics University of Arizona Tucson, AZ 85721, USA [email protected] Jack Hoeksema Department of Linguistics University of Groningen P.O. Box 716 9700 AS Groningen, The Netherlands [email protected] Kyongjoon Kwon Harvard University Department of Slavic Languages & Literatures 12 Quincy Street, 3rd Floor, Barker Center Cambridge, MA 02138, USA [email protected] Terje Lohndal Department of Linguistics 1401 Marie Mount Hall University of Maryland College Park, MD 20742, USA [email protected] Cecilia Poletto Dipartimento di Scienze del Linguaggio Università di Venezisa Ca’ Bembo Dorsoduro 1075, Venezia, 30123, Italy [email protected]

 Cyclical Change

Clifton Pye University of Kansas Department of Linguistics The University of Kansas Lawrence, KS 66045, USA [email protected] Olena Tsurska Department of English Arizona State University Tempe, AZ 85287–0302, USA [email protected]

Diana Vedovato via Beato Pellegrino, 1 35137 Padova, Italy [email protected] Cathleen Waters Department of Linguistics University of Toronto 130 St. George Street, room 6076 Toronto, Ont. M5S 3H1, Canada [email protected]

chapter 1

Cyclical change, an introduction Elly van Gelderen

Arizona State University This chapter provides a general background to the linguistic cycle and cyclical change. It reviews some of the cycles that we know, the steps that are typical in a cycle, some differences between various cycles, and examines where cycles start. The chapter also considers a theoretical account. While reviewing the types of cycles, it sets the stage for the issues discussed in the chapters that follow.

The chapters in this collection present an excellent overview of work on cyclical processes relevant to synchronic and diachronic syntax. Most recent research on the cycle has focused on the negative cycle and the papers in this volume reflect that interest. As is obvious from the other papers, some linguists have also started to examine agreement phenomena, and that involves pronouns as well as auxiliaries and copulas, in the light of cyclical change. Looking at adpositions, modals, and elements in the left periphery through a cyclical lens can also provide a new perspective and analysis, both in the syntax and semantics. The current volume is based on papers presented during the Workshop on the Linguistic Cycle that took place at Arizona State University in April of 2008. The discussions during the workshop were lively and very focused and emphasized the variation in the cycles. Crucial questions in relation to the linguistic cycle are the following, with (a) to (d) being descriptive, and (e) and (f) asking why language is the way it is. The latter kind of question is currently the focus of much minimalist inquiry (e.g. Chomsky 2005). (1)

a. b. c. d.

Which cycles exist and why? Are there typical steps in a cycle? What are the differences between cycles? What are the sources of renewal once a cycle has desemanticized a lexical item? e. What insights or explanations can certain theoretical frameworks provide? f. Why is there cyclical change?



Elly van Gelderen

Sections two to seven of this introduction will discuss how the papers contained in this volume relate to these six questions. In section one, some brief background to cyclical change is provided first.

1. The linguistic cycle and cyclical change The Linguistic Cycle is a name for changes where a phrase or word gradually disappears and is replaced by a new linguistic item. The most well-known cycles involve negatives, where an initial single negative, such as not, is reinforced by another negative, such as nothing, or replaced by never, and subjects, where full pronouns are reanalyzed as endings on the verb. The former is often called Jespersen’s Cycle, after the Danish linguist Otto Jespersen, but as van der Auwera & de Vogelaer (2008) point out, the Egyptologist John Gardiner was earlier in identifying this process. Clausal markers, demonstratives, aspect markers, and copula verbs also undergo cycles of internal change. There are early advocates of the view that language change is cyclical, e.g. de Condillac (1746), Tooke (1786–1805), von Humboldt (1822), and Bopp (1816). The oftcited passage in von der Gabelentz (1901: 256) uses ‘spiral’ to indicate new cycles are not identical to the old ones: immer gilt das Gleiche: die Entwicklungslinie krümmt sich zurück nach der Seite der Isolation, nicht in die alte Bahn, sondern in eine annähernd parallele. Darum vergleiche ich sie der Spirale. (von der Gabelentz 1901: 256)1

In the 1950s, Tauli (1958) provides many examples of cyclical change, but apart from sporadic work, e.g. by Hodge (1970), Greenberg (1978), Givón (1976), and Tauli (1966), not much research had been done up to very recently. The renewed interest in grammaticalization starting in the 1980s was, of course, essential to understanding the linguistic cycle, with work such as Givón (1976), Lehmann (1985, 1995), Traugott & Heine (1991), Abraham (1993), and others. Recently, work on the negative cycle has started to appear. For instance, from June 2008 to May 2009, one-day events on the negative cycle took place in Birmingham (http://www.lhds.bcu.ac.uk/english/cycles-of-grammaticalization), but other cycles have not been given as much attention. Cycles of language change have not been studied in generative linguistics (apart from again the negative cycle), and only sporadically in other formal frameworks. The workshop in Arizona in April 2008 was an attempt to bring together linguists who are interested in cyclical change from a variety of frameworks and to contribute to new directions in work on language change.

1. “always the same: the development curves back towards isolation, not in the old way, but in a parallel fashion. That’s why I compare them to spirals” (my translation, EvG).



Chapter 1.  Cyclical change, an introduction

Table 1.  Examples of Cyclical Change Negative   negative argument > negative adverb > negative particle > zero   negative verb > auxiliary > negative > zero Subject (and Object) Agreement Cycle   demonstrative/emphatic > pronoun > agreement > zero Copula Cycle   demonstrative > copula > zero   verb/adposition > copula > zero Definiteness2   demonstrative > definite article > ‘Case’ > zero Future and Aspect Auxiliary   adverb/adposition > modal > complementizer Place/time   noun > adposition > complementizer

A partial list of cycles is presented in Table 1 and Section 2 outlines which chapters of the book deal with which change. The examples in Table 1 are, of course, perfect instances of grammaticalization clines. Yet, the current literature on grammaticalization has been reluctant in embracing the cycle. Cyclicity is mentioned only three times in Hopper & Traugott (2003). They talk about grammaticalization as a “continuously occurring phenomenon” (p. 124) and point out that the cyclical model is “extremely problematic because it suggests that a stage of a language can exist when it is difficult or even impossible to express some concept” (p. 124). They see renewal not as replacing a weakened lexical item but as competing with it. The clines in Table 1 could be adapted to show overlap between the reduced form and the start of a new cycle. A number of chapters in this volume adopt such an approach explicitly, e.g. all four chapters in Part I. The model of Feature Economy that some authors adopt also accounts for the overlap: once a feature has been reanalysed as uninterpretable, it becomes a probe and needs new semantic features (more on this in section six). Thinking of change as cyclical assumes that it is unidirectional. Some generative linguists have argued against this unidirectionality, for instance, Newmeyer (1998: 263– 275) and Lightfoot (e.g. 2006: 38). However, Traugott & Dasher (2002: 87) make the point that the number of real counterexamples to unidirectionality is small and not systematic. This volume takes that approach.

2. The definiteness cycle is not represented in this volume, but see Lyons (1999) and van Gelderen (2007)





Elly van Gelderen

2. The cycles discussed in this volume The cycles typically discussed nowadays affect a minor part of the grammar of a language, e.g. negation or modality. There are people who look at the shifts in typological character, e.g. Hodge (1970) and Baker (2001), but most people are more conservative. For instance, Heine, Claudi & Hünnemeyer (1991: 246) argue that there is “more justification to apply the notion of a linguistics cycle to individual linguistic developments” rather than to changes from analytic to synthetic and back to analytic. The papers in this book follow this trend, but section six mentions some ways of looking at cyclical change from a typological perspective. Part I in this book contains four chapters on the Negative Cycle. Two typical sources for negatives, or starting points of cyclical change, are listed in Table 1. They are full phrases, such as English no thing, and verbal heads, such as Chinese bu ‘not’ (from a verb meaning ‘to die’) and Lewo toko ‘not’ (from a prohibitive verb, see below). The majority of the data in the volume deals with the first of these sources. Biberauer examines negative concord elements in Afrikaans, Tsurska does so in various stages of Russian, and Hoeksema focuses on the development of a polarity item into a negative marker in Dutch. The main trend here is “polarity items turning into negative quantifiers [and then turning] into adverbial elements” (Hoeksema, this volume). Van der Auwera mainly examines the phrasal type of the negative cycle, providing detailed information on phrasal negation from varieties of Dutch. He also adds data on the verbal/non-phrasal origins of negatives in languages such as the Austronesian language Lewo. In (2a), toko is a verb but (2b) shows it is also used as negative. (2) a. Na-kan-ena toko! nom-eat-nom desist ‘Desist from eating!’

Lewo

b. Ve a-kan re toko! neg 2sg-eat neg neg ‘Don’t eat it!’ (Early 1994: 76; see van der Auwera, this volume)

Lewo

The three chapters in Part II examine agreement, subject marking, and the left periphery. They all provide evidence for the intricacies of cycles. The traditional agreement cycle can be represented as in Table 1, namely as having a demonstrative/pronominal source. In many languages, the agreement affix resembles the emphatic pronoun and derives from it. The most well-known case is, of course, French subject pronouns. In the history of French, the subject jo ‘I’ is reanalyzed from emphatic pronoun to subject pronoun to clitic je and is currently regarded by many (Lambrecht 1981; Zribi-Hertz 1994) as an agreement marker, e.g. (3a) shows an additional emphatic typically present, (3b) and (3c) show that subject marker is obligatory. The same is true for the second person and in many dialects even for third person, as in (4).



Chapter 1.  Cyclical change, an introduction

(3) a. Moi, j’ai vu ça. me I-have seen that ‘I’ve seen that.’

Colloquial French

b. *Je lis et écris I read and write

Colloquial French

c. *Je probablement ai vu ça I probably have seen that (4) Personne il a rien dit person he has nothing said ‘Nobody said anything.’ (Zribi-Hertz 1994: 137)

French Colloquial French

The three chapters in Part II add complexity to this picture. Vedovato’s chapter carefully examines the Italian pronominal paradigm and provides evidence of emphatic pronouns being reanalyzed as weak pronouns. She then argues that prescriptive forces interfered in this agreement cycle and stopped the weak pronouns from further development in standard Italian. Vedovato coined the term ‘broken cycle’ in her original contribution to the workshop in April. Kwon’s chapter provides some evidence for the reanalysis of a be-verb as a pronoun in earlier stages of Russian. This is a change not represented in Table 1, but corroborates Katz (1996) who indicates a possibly similar development in the history of Hebrew and of Turkish. Poletto’s paper examines the development of Italian e and sì, topic and focus markers in the left periphery. She also shows that the grammaticalization cycle can be ‘broken’, i.e. stopped or changed by other changes going on in the language. Part III contains studies on the various cycles that heads such as copulas, adverbs, modals, auxiliaries, and adpositions take part in. They are, of course, the prototypical instances of grammaticalization and follow a path of being reanalyzed as higher, more abstract elements. Copulas and auxiliaries can also be derived from (phrasal) demonstratives and in that sense they bridge Parts II and III. Copulas have two main sources, demonstratives and verbs. Lohndal examines these two paths giving a uniform analysis through Feature Economy. Auxiliaries and prepositions are involved in many kinds of cyclical change. The most well-known data show them as originating from more lexical categories. Gergel’s work shows that grammaticalization also affects semantic structure. His data and Pye’s examine modal and aspectual auxiliaries and their more adverbial and verbal origins whereas Waters examines the periodic reinforcement of spatial prepositions through a noun.

3. The steps in a cycle All papers discuss steps in a particular cycle. Hoeksema outlines four stages: from single negation, to optionally reinforced, to obligatorily reinforced, to optional again.





Elly van Gelderen

The last stage leads back to the initial one but with a new negative. Van der Auwera argues for an alternative to the Jespersen Cycle, with his account describing “exactly eight possible trajectories or ‘cycles’”. Both van der Auwera and Biberauer emphasize the repetitive clause-final negation, as in (5). Its role is not to emphasize, so it cannot be pragmatic renewal. (5) Ik heb niets gekregen niet Variety of Dutch I have nothing received neg ‘I haven’t received anything.’ (van der Auwera, this volume) Biberauer argues that “structural height and deficient featural properties can disqualify a concord element … from being reanalysed as a ‘real’ negator”. Tsurska examines the steps from a non-strict Negative Concord language, in which the pre-verbal nwords are used without the preverbal negative marker to express negation, to a strict Negative Concord language, where n-words are unable to express sentential negation alone. The steps she observes can be put in terms of a change in features. As mentioned above, the steps in the cycles discussed in Part II are less uniform, especially since some are ‘interfered’ with. Those in Part III typically involve a reanalysis in a higher position. Thus, Waters finds nouns being reanalyzed in higher positions, but in a way that makes use of the expanded PP, and Gergel’s analysis of rather shows intricate steps from lower adverb to higher modal. A factor that is important in the cycles is the structural position of the element. As Biberauer and Lohndal show, the element that would be a candidate for reanalysis might be out of reach.

4. The differences between cycles Heine, Claudi & Hünnemeyer (1991: 244) note that “little is known about the time span of grammaticalization processes”. Some change is fast, e.g. the verb to auxiliary and noun to preposition reanalyses. Other change, such as the Chinese verb ba ‘to grasp’ being reanalysed as object marker, takes much longer. The papers in this volume show that some change is faster than other change and most changes can be ‘interrupted’. They also show that the negative cycle is perhaps the most uniform. Possibly because negation is pragmatically so important, it can be subject to extremely fast change. For instance, the relatively conservative Athabascan language family of North America displays enormous variation where negation is involved (see van Gelderen 2008a). The archaic varieties, as in (6), show evidence of an incorporated verb whereas more innovative varieties, as in (7), show renewal. (6) tendhghaaghetltenęę fut-qua-neg-qua-qua-1s-cause-ice-neg ‘I won’t freeze it solid.’ (from Kari 1993: 55)

Lower Tanana



Chapter 1.  Cyclical change, an introduction

(7) Doo dichin nishłįį da neg hungry 1s-be neg ‘I’m not hungry.’ (Young & Morgan 1987; 350)

Navajo

Change in agreement markers is uniform since the source is pronominal, as indicated in Table 1. There is, however, some disagreement as to which pronoun sets the cycle in motion, so to speak. Givón’s (1976) work suggests that it is the third person but data from French (see (3)) suggests that it is first person. The object cycle, not discussed in this volume, shows that object agreement typically starts with animate definite objects. Markers of mood and modality experience fast change like negatives (see, for instance, Traugott & Dasher 2002) as do aspect markers (see Brinton 1988 and Sims 2008 ). Their source is typically a more lexical head, but can on occasion be a phrase as well (not indicated in Table 1). Well-known examples of verbs reanalyzed as auxiliaries can be found as part of future cycles in Romance, Germanic, and Urdu/Hindi. Well-known cases of Aspect Cycles involve adverbs incorporated into phrasal verbs and become aspect markers.

5. The sources of renewal As indicated in Table 1, sources of renewal can be full phrases as well as single lexical items. The sources are familiar from the vast literature on grammaticalization, e.g. Heine & Kuteva (2002). A few are listed in Table 2. The reason for these sources is that they provide new semantic features for what was ‘grammaticalized away’: person and number (phi-features) in the case of agreement and copulas, negative features in the case of negatives, spatial features in the case of prepositions, and so on. Once the functional element has lost its semantic and interpretable features, this would be formulated within a Minimalist framework as becoming a probe looking for an element to value its features. Some elements are straightforward renewers: demonstratives have phi-features and can renew agreement and adverbs have temporal or spatial features and renew prepositions and complementizers. Table 2.  Examples of Renewal Agreement: Copula: Modal: Negative: Preposition:

Emphatic pronoun/noun Demonstrative/verb Verb/adverb Minimizer/Negative nominal/Negative adverb Noun





Elly van Gelderen

6. Insights of certain theoretical frameworks The grammaticalization literature has been a wonderful resource in outlining the clines of change, but has been less eager to look at the linguistic cycle (see e.g. Hopper & Traugott 2003, mentioned above). As indicated in Section 1, one can think of a Minimalist account that incorporates the clines in Table 1 as well as the necessary renewal. Starting with Chomsky (1995), the features relevant for and accessible during the derivation are formal. Formal features can be interpretable (relevant to the semantic interface) or uninterpretable (only relevant to move elements to certain positions). Interpretable features are acquired before uninterpretable ones, as argued in Radford (2000), but are later reinterpreted as uninterpretable, triggering the functional/grammatical system. The same happens in language change. Changes in negatives can be explained by arguing that their (initially) semantic features are reanalyzed as interpretable and then as uninterpretable, as in (8), from van Gelderen (2008b). Phrases like never have interpretable negative and phi-features that are probed by a probe in a functional category. Once the phrase is reanalyzed as a head (e.g. Old English ne ‘not’), another element is required. (8)

Feature Economy Minimize the interpretable features in the derivation, e.g: Adjunct Specifier Head affix semantic > [iF] > [uF] > [uF]

Hicks (2009: 204) characterizes Feature Economy as “establish[ing] dependencies where possible” and calls this principle Maximize Featural Economy. An early version is perhaps Schütze (1997) who uses an agreement maximization principle. The same change as in (8) occurs with the subject cycle: the interpretable person (and gender) features of a full pronoun are reanalyzed as uninterpretable when they become agreement3. (9) emphatic > full pronoun > head pronoun > agreement [i-phi] [i-phi] [u-1/2] [i-3]4 [u-phi] The clines in Table 1 should therefore be seen as having the renewing element on the left side of a cline overlap with the item on the right side of the older cline. Feature Economy requires such a renewal.

3. This is compatible with Chomsky’s (1995: 230; 381) views on features: “formal features have semantic correlates and reflect semantic properties (accusative Case and transitivity, for example)”. 4. In van Gelderen (in progress), I argue that third person pronouns have deictic features in addition to phi-features and are therefore harder to reanalyze as the uninterpretable phi-features of T/agreement.



Chapter 1.  Cyclical change, an introduction

7. Why is there change and why are there cycles? Many historical linguists see language change as determined by two kinds of factors. There are internal factors, such as those instigated by the Economy Principles or by ‘Ease’, as in Jespersen (1922). These approaches deal mainly with the articulatory ease of pronunciation. Children acquiring a language use these principles to analyze their input. Principles such as (8) above and (10) below are examples of that. There are also external factors for language change such as a need on the part of speakers to be innovative and creative or conservative. I will discuss both briefly in this section since some of the contributions (e.g. Vedovato) mention external factors. External factors include pragmatic ones. The urge of speakers to be innovative may introduce new, loosely adjoined elements into the structure. Hagège (1993: 153) uses the term Expressive Renewal. Speakers may want to be explicit and therefore choose full phrases rather than single words. One source of new specifiers and words is borrowing. Heine & Kuteva (2005: 3) give examples of wh-interrogatives being expanded to relatives in Tariana under the influence of Portuguese, where interrogatives and relatives share the same form. The same pattern occurred in the history of English: whpronouns were used in questions but were later extended to relative contexts under the influence of French (see van Gelderen 2004). Heine & Kuteva (2005: 73) give many other examples, e.g. Tariana speakers renewing their evidentials by using Portuguese expressions such as eu vi ‘I saw’. Another external factor is the ‘need’ to be conservative and prescriptive. This may stop change altogether. In the chapters that follow, we will examine some examples. For now, I will mention stranding and negatives, where prescriptive rules are very strong. Considering Economy, a principle such as (10) is expected. (10) Stranding Principle Move as little as possible. This principle has been used to explain why speakers in English typically front the DP, as in (11) and (12), rather than the full PP in (13). (11) Who did you talk to who? (12) Quilc men mai get wundren on ‘which men may yet wonder about’ (Genesis & Exodus 3715, from Denison 1993: 132) (13) To whom did you talk to whom? Preposition stranding in English, as in (12), starts in the 13th century (Denison 1993: 125 ff.). It is preferred under (10) and it is estimated that in speech 86% of prepositions are stranded while in writing only 7% are. This difference between spoken and written data points towards strong prescriptive pressure. Bullokar’s grammar from 1586 contains stranded prepositions, but one century later, most grammarians pre-





Elly van Gelderen

scribe against its use. Yáñez-Bouza (2007) finds that these prescriptivists indeed had an influence on the language. Other languages may experience prescriptive pressure as well. As chronicled in great depth in Fleischer (2002), in many varieties of German preposition stranding is frequent though some describe it as “älter oder umgangssprachlich” (137). It is also common in North-American varieties of French.

8. Conclusion If cycles are real, we need an explanation. This volume hopes to contribute to both the description of cyclical change and an account (or more than one). The book therefore ends with the description by Hancock & Bever of a proposed experiment to test the psychological reality of cycles.

Acknowledgements and practical notes Many thanks to the internal and external reviewers, as well as to Henning Andersen, Aryeh Faltz, and David Ingram. Thanks to Olena Tsurska for helping to organize and think about the workshop, to Werner Abraham and Terje Lohndal for thinking through some of the issues in this introduction, and to Harry Bracken and Olena Tsurska for editorial assistance. Since there are a number of languages represented in this book, abbreviations to glosses and languages are provided in the first (or second) footnote to each chapter. This makes it easier to read chapters on their own. References are also listed after each chapter. Names starting with van or van der or de are listed in Dutch publications under the letter that follows these connectors but in Belgian publications under v or d. US and other usage is mixed. The reference lists at the end of each chapter keep to the Dutch use (except for Belgian authors). The exception is in the references to van der Auwera’s chapter.

References Abraham, Werner. 1993. Einleitung zum Thema dieses Bandes. Grammatikalisierung und Reanalyse: Einander ausschließende oder ergänzende Begriffe? Folia Linguistica Historica 13(1–2): 7–26. Baker, Mark. 2001. The Atoms of Language. New York NY: Basic Books. Bopp, Franz. 1816. Über das Conjugationssystem der Sanskritsprache in Vergleichung mit jenem der griechischen, lateinischen, persischen und germanischen Sprachen. Frankfurt am Main: In der Andreäischen Buchhandlung. Brinton, Laurel. 1988. The Development of English Aspectual Systems. Cambridge: CUP.



Chapter 1.  Cyclical change, an introduction Bullokar, William. 1586 [1970]. Bref Grammar. In The Works of William Bullokar II, John Raymond Turner (ed.). Leeds: Leeds University. Chomsky, Noam. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky, Noam 2005. Three factors in language design. Linguistic Inquiry 36(1): 1–22. Condillac, Etienne Bonnot de. 1746. Essai sur lórigine des connaissances humaines. Paris. Denison, David. 1993. English Historical Syntax. London: Longman. Early, Robert. 1994. Lewo. In Typological Studies in Negation [Typological Studies in Language 29], Peter Kahrel and René van den Berg (eds), 65–92. Amsterdam: John Benjamins. Fleischer, Jürg. 2002. Die Syntax von Pronominaladverbien in den Dialekten des Deutschen. Wiesbaden: Franz Steiner Verlag. Gabelentz, Georg von der. 1891. Die Sprachwissenshaft. Ihre Aufgaben, Methoden und bisherigen Ergebnisse. Leipzig: Weigel. (Reprint Tübingen: Narr 1972). Gelderen, Elly van. 2004. Grammaticalization as Economy [Linguistik Aktuell-Linguistics Today 71]. Amsterdam: John Benjamins. Gelderen, Elly van. 2007. The definiteness cycle in Germanic. Journal of Germanic Linguistics 19(4): 275–305. Gelderen, Elly van. 2008a. Cycles of negation in Athabaskan. Working Papers in Athabaskan Languages 7: 49–64. Gelderen, Elly van. 2008b. The negative cycle. Linguistic Typology 12(2): 195–243. Gelderen, Elly van. In progress. The Linguistic Cycle. ms. Givón, Talmy. 1976. Topic, pronoun, and grammatical agreement. In Subject and Topic, Charles Li (ed.), 151–188. New York NY: Academic Press. Greenberg, Joseph. 1978. How does a language acquire gender markers? In Universals of Human Language 3, Joseph Greenberg (ed.), 47–82. Stanford CA: Stanford University Press. Hagège, Claude. 1993. The Language Builder. Amsterdam: John Benjamins. Heine, Bernd, Claudi, Ulrike & Hünnemeyer, Friederike. 1991. Grammaticalization: A Conceptual Framework. Chicago IL: University of Chicago Press. Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Heine, Bernd & Kuteva, Tania. 2005. Language Contact and Grammatical Change. Cambridge: CUP. Hicks, Glyn. 2009. The Derivation of Anaphoric Relations. Amsterdam: John Benjamins. Hodge, Carleton. 1970. The linguistic cycle. Linguistic Sciences 13: 1–7. Hopper, Paul & Traugott, Elizabeth. 2003. Grammaticalization. Cambridge: CUP. Humboldt, Wilhelm von. 1822. Über die Entstehung der grammatischen Formen und ihren Einfluss auf die Ideenentwicklung. Abhandlungen der Akademie der Wissenschaften zu Berlin. (Reprint: Humboldt (1972): 31–63). Jespersen, Otto. 1922. Language. London: Allen & Unwin. Kari, James. 1993. Diversity in morpheme order in several Alaskan Athabaskan languages. BLS Proceedings 50–56. Katz, Aya. 1996. Cyclical Grammaticalization and the Cognitive Link between Pronoun and Copula. PhD dissertation, Rice University. Lambrecht, Knud. 1981. Topic, Antitopic, and Verb Agreement in Non Standard French [Pragmatics & Beyond II:6]. Amsterdam: John Benjamins. Lehmann, Christian. 1985. Grammaticalization: Synchronic variation and diachronic change. Lingua e stile 20(3): 303–318. Lehmann, Christian. 1995. Thoughts on Grammaticalization. Munich: Lincom. Lightfoot, David. 2006. How New Languages Emerge. Cambridge: CUP. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP.





Elly van Gelderen Newmeyer, Frederick. 1998 Form and Function. Cambridge MA: The MIT Press. Radford, Andrew. 2000. Children in search of perfection: Towards a minimalist model of acquisition. Essex Research Reports in Linguistics 34. (http://privatewww.essex.ac.uk/~radford/ PapersPublications/perfection.htm). Schütze, Carson. 1997. INFL in Child and Adult Language. PhD dissertation, MIT. Sims, Lynn. 2008. The Grammaticalization of Ingressive Aspect in Early English. ASU PhD. Tauli, Valter. 1958. The Structural Tendencies of Languages. Helsinki. Tauli, Valter. 1966. Structural Tendencies in Uralic Languages. Den Haag: Mouton. Tooke, John Horne. 1786–1805. The Diversion of Purley. London. Traugott, Elizabeth & Heine, Bernd. 1991. Grammaticalization. Amsterdam: John Benjamins. Traugott, Elizabeth & Dasher, Richard. 2002. Regularity in Semantics Change. Cambridge: CUP. van der Auwera, Johan & De Vogelaer, Gunther. 2008. Negation and quantification. In Syntactic Atlas of the Dutch Dialects, Vol. II, Sjef Barbiers, Johan van der Auwera, Hans Bennis, Eefje Boef, Gunther De Vogelaer & Margreet van der Ham (eds), 58–72. Amsterdam: Amsterdam University Press. Yáñez-Bouza, Nuria. 2007. Preposition Stranding and Prescriptivism in English from 1500 to 1900. PhD dissertation, University of Manchester. Young, Robert & Morgan, William. 1987. The Navajo Language. Albuquerque NM: University of New Mexico Press. Zribi-Hertz, Anne. 1994. La syntaxe des clitiques nominatifs. Travaux de Linguistique et Litterature 131–147.

part 1

Negatives

chapter 2

Jespersen recycled Jack Hoeksema

University of Groningen The Jespersen cycle, the series of changes in which single negation is replaced by double negation, which in turn is reduced to single negation, is fleshed into its component parts: the various changes which conspire to feed the cyclical change. Special attention is given to the process by which a negative polarity item is reinterpreted as the bearer of negation. It is argued that this process only applies when the likelihood of negation, given the presence of the polarity item, is very high. Items such as English ever or any do not fall into this category, since they appear in a wide variety of contexts. Another change, sometimes overlooked in the literature, is the one from DP-like polarity item (such as English naught) to adverbial negation element (English not). Some contexts are identified where the difference between DP and Adverb appears to be neutralized, and it is speculated that the reinterpretation process starts precisely in such contexts.

1. Introduction1 In this paper, I take a look at certain aspects of the so-called Jespersen cycle. In particular, I want to consider more closely the role of negative polarity items in the renewal of negation, and the grammatical interpretation of this renewal in terms of the NegP hypothesis. The Jespersen Cycle is a series of processes by which negation markers get renewed in the manner depicted in Figure 1. By ‘double negation’, I am referring here to formal features only, since the usual interpretation of such double negation is single negation. The phenomenon of double negation serving to express a single negation is also known as Negative Concord (Labov 1972, Ladusaw 1992, 1993).

1. The material in the paper was presented at the Linguistic Cycles workshop, organized by Elly van Gelderen at the Arizona State University, in Tempe, April 2008. I am grateful to Elly and to the audience at this meeting for feedback and discussion, as well as to Johan van der Auwera for his comments on an earlier draft.



Jack Hoeksema

I: Solitary negation: A

II: Optional double negation: A+(B)

IV: Optional double negation: (A)+B

III: Obligatory double negation: A+B

Figure 1.  The Jespersen Cycle

The cycle in Figure 1 turns out to be quite common and has been attested in Latin (Jespersen 1917, Bernini & Ramat 1996), French (Jespersen 1917, Rowlett 1998, Deprez 2000, Roberts and Roussou, 2003, among others), Greek (Kiparsky and Condoravdi 2006), German (Jäger 2008, Breitbarth 2008), English (van Kemenade 2000, van Gelderen 2004a), Dutch (Hoeksema 1997, Postma 2002, Zeijlstra 2004), Welsh (Willis 2008), Arabic and Berber (Lucas 2007), Chinese, and Athabaskan (van Gelderen 2008). As van der Auwera & de Vogelaer (2008) point out, Jespersen was not even the first to discuss the cycle of negation, as he was preceded by Gardiner (1904) and Meillet (1912), at the very least. The structure of this paper is as follows. I will start with a discussion of Stage I, and then work my way around to Stage IV and back to Stage I. Along the way, I will be looking at the changes involved, their nature, what might prompt them, as well as the question to what extent the changes are unidirectional.

2. Stage I Stage I would seem not to require a whole lot of attention. It is the starting point of the cycle, and has what one might call the most parsimonious and simple system of negation. Indeed, one might wonder why a language would ever want to drift away from this stage. At the same time, one may wonder which language ever completely conformed to this rosy picture. Simple, solitary negation is often found in sentences, but rarely in a whole language. First of all, languages tend to have multiple ways to express negation.



Chapter 2.  Jespersen recycled 

A language like modern standard Dutch, which many would probably classify as being in Stage I, usually expresses negation by a single negative marker: (1) Hij is niet sterk He is not strong ‘He is not strong.’ However, there are several optional ways to strengthen negation by adding modifiers, as well as a number of expressions which serve as occasional, marked, alternatives to niet:2 (2) Hij is allerminst sterk he is all-least strong ‘He is not strong at all.’ (3) Hij is geenszins sterk He is no-way strong ‘He is not strong at all.’ (4) Hij is allesbehalve sterk he is all-but strong ‘He is not strong at all.’ Of course I have not even begun to discuss negative polarity items. In this respect, modern Dutch is not in any way unusual, but probably quite typical of languages in general: they have multiple means to express negation, even at Stage 1, and these means may differ not only in syntactic properties, such as whether they appear in finite or non-finite clauses or in indicative, interrogative or imperative constructions, but also in their semantics and pragmatics. Some negation markers are emphatic in nature, leading to stronger, more expressive statements; others may have some special discourse function. Schwenter (2006) has pointed out that Catalan pas and Italian mica, for instance, are used for denying discourse-old propositions. I do 2. The negative character of these expressions is not just indicated by the fact that they can be paraphrased by regular negation, but also by the fact that they trigger items which are otherwise only triggered by negation (cf. Van der Wouden 1994, 1997, Zwarts 1998), such as the Dutch polarity-sensitive predicate mals ‘mild’, which fails to be triggered by weaker forms of negation, such as n-words and quantifiers like weinig ‘few/little’: (i) a. Zijn oordelen waren niet / allesbehalve / geenszins mals his judgments were not / all-but no-way tender ‘His judgments were not (at all) mild.’ b. *Geen van zijn oordelen was mals none of his judgments was tender ‘None of his judgments was mild.’ c. *Slechts weinige oordelen waren mals only few judgments were tender ‘Only few judgments were mild.’



Jack Hoeksema

not know whether there are any languages (with the exception of artificial languages such as Esperanto or predicate logic, and impoverished languages such as pidgins or early stages of child language) which have only one marker for negation.

3. From stage I to stage II Given the availability of alternative negation signs, one may wonder why lexical renewal of negation does not simply involve some process of competition, at the end of which one of the alternative signs wins out and replaces the old unmarked negation sign. That would turn the Jespersen cycle into a one step process, and clearly take all the fun out of it. But to be frank, it is not entirely clear to me, why this is not, as far as I can tell, the most common way for the negation system to change. Presumably, the thing that makes the Jespersen Cycle such a common process is its conservative nature. Rather than replacing one negation sign by another, we keep the old one, but add another one, and only when the new sign is as familiar as the old one, do we drop the old sign. For Jespersen, the driving force behind the cycle was phonetic erosion. Grammatical markers such as negation often undergo phonetic reduction, leading to a point where they become almost inaudible. Given that negation is too important to let go unnoticed, something has to be done. A polarity item is enlisted to prop up the faltering negation marker. Now this kind of process can be witnessed in many areas of the grammar, not just negation. Grammatical markers often undergo reduction, leading to changes in the grammar. Sometimes they are renewed, and sometimes they are not. One may think of the loss of case marking in English or Dutch, a complex process in which a great many factors are at play, but reduction is clearly one of them. However, here there seems to be no sign of a cyclical process. Case marking was lost, and whether it will ever be renewed by some future process is very doubtful. Clearly, there is a difference between case marking and negation. While case marking is not essential for the expression of meaning, given that there are other means to encode grammatical functions, negation is essential. It would be very hard to do away with negation. One could, of course, imagine a language where every verb and every adjective has a negative counterpart. Not just an antonym expressing contrary negation, but a true contradictory counterpart. Such a language would seem possible, given that contradictory antonyms exist. For example, present and absent seem to be perfectly contradictory in the sense that everything is either present or absent, and nothing is both. However, a grammar without the category of negation would entail doubling large parts of the vocabulary, a very costly move. Natural languages are not optimally parsimonious in their vocabulary, but neither are they utterly wasteful. Moreover, it would seem that some negative verbs might be exceedingly hard to learn. Just consider the problems a child may have in mastering a verb that means ‘not to hiccup’, if, first of all, that verb is not related in its form to the verb for hiccup, second, is probably not used a whole lot, and third, there



Chapter 2.  Jespersen recycled 

is no negation in the language that might help the parent to explain the meaning of the verb. The evolutionary advantages of having negation in a grammar are so strong that no language would want to do without it. So even without assuming that the presence of negation is an innate property of natural language, one may, for purely functional reasons, expect it to be universal. Clearly, when erosion threatens to destroy the expression of negation, something has to be done. Negative polarity items to the rescue! For Jespersen himself, this is pretty much where the story ends. The cycle is driven by phonetic erosion and the desire to keep the category of negation intact. The rest is details. Now one thing where Jespersen may well be wrong is in his insistence on the causal role of reduction. Jespersen argued that the reduction of French negation to ne inevitably led to the emergence of double marking as a way to reinforce the weakened sign of negation. This in turn rendered ne entirely superfluous, leading ultimately to its disappearance in the spoken vernacular. However, one might also argue the other way around. If negation is so important, and clearly it is, why would one want to reduce it at all? Ease of articulation is certainly not of such paramount importance that speakers should let it prevail over the clarity of their message. Perhaps it is the fact that negation is often predictable, because of double marking or because of constructional features, that makes it easy prey for phonetic reduction. Consider in this connection English. The contracted forms of negation only came about after do-support had made the presence of negation easier to detect. I would not want to say that do is a marker of negation, but it certainly helps identify a sentence as negative. So it is a construction feature that made it relatively unproblematic for not to turn into n’t. In addition to this point, I might note that not every instance of the Jespersen cycle seems to necessarily involve phonetic reduction (cf. Kiparsky and Condoravdi 2006 on Greek). It is arguable that a more essential characteristic of negation driving the Jespersen cycle is its double role as a device for the expression of logical polarity and a rhetorical device. Negation is first and foremost a logical operator, changing the truth-value of a proposition to its opposite. This basic function is what makes it indispensible. However, negative sentences often express more than just the negation of their positive counterpart. They may develop special pragmatic uses such as understatement or emphatic denial. For these pragmatic functions, regular negation may be used, but various colorful alternatives are on offer as well. If someone tells you it will rain tonight, you might respond with it won’t rain tonight, if you want to contradict him, but also with Like hell it will, or no way will it rain. Negative polarity items often start out as colorful terms intended to strengthen or weaken the force of negation. From what I can tell, every documented language has such polarity items. Apparently, there is great need for expressions that serve to boost the rhetorical effect of negation. And it is precisely the availability of negative polarity items which is essential in setting up a system of double marking which eventually leads to complete lexical replacement of the original marker of negation. In the course of this process of replacement, the rhetorical character of the polarity item gets lost (Kiparsky and Condoravdi 2006, Schwenter 2005).



Jack Hoeksema

4. Polarity items Polarity items come in a great many varieties. Only some are ever chosen to become the main sign of negation in a language, like pas in French or not in English. However, some of the typical changes involved in the Jespersen cycle may be witnessed among the less important polarity items as well. One such change is the one from nonnegative item in the scope of negation to a negative item that no longer requires licensing. Horn (2001), referring to as-of-then unpublished work by Ross and Postal (but see Postal 2004), discussed a set of English taboo items which still take, but no longer require, negation: (5) a. He didn’t tell me fuck all about the car. b. He told me fuck all about the car. (6) a. Fred doesn’t know jack shit about the car. b. Fred knows jack shit about the car. (7) a. There isn’t diddly squat in the fridge. b. There’s diddly squat in the fridge. (8) a. The cops didn’t tell me dick. b. The cops told me dick. What is striking about these items is their uniform behavior and their lack of a determiner. Most minimizers in English, such as an iota, a word, a thing, a hope in hell etc. are preceded by a(n) or one. The reason might be that the items in (5) to (8) are mass nouns, but that in itself is rather unusual for minimizers. In German, various words for bowel movements (for which Horn 2001 has coined the term drecative) show a very similar behavior: (9) a. Google schert sich keinen feuchten Dreck um den Datenschutz. Google bothers self no moist shit about the data protection ‘Google does not give a damn about data protection.’ b. Google schert sich einen feuchten Dreck um den Datenschutz Google bothers self a moist shit about the data protection ‘Google could care less about data protection.’ Dutch has an expression that underwent a quick change from polarity item to negative idiom within a short period (Hoeksema 2002). It is likewise a taboo term: (10) a. Hij begrijpt er de ballen niet van. He understands there the bollocks not of ‘He doesn’t understand jack shit (about it).’ b. Hij begrijpt er de ballen van. He understands there the bollocks of ‘He understands jack shit (about it).’



Chapter 2.  Jespersen recycled 

Semantically, the expression belongs to the group of taboo expressions with minimizing properties, but syntactically it stands out as unusual within that set because it is formally definite, rather than indefinite. I cannot help but think that it is the exceptional features of these expressions which allowed them to split off from the pack and to undergo a turn of the Jespersen cycle which other minimizers in English and Dutch do not seem ready yet to submit to. In addition, one may note that the taboo items in question belong to the domain of substandard usage, where some form or other of negative concord is common enough. In order to get at sentences like (10b), we first need to have a reanalysis of (10a) where the taboo expletive is interpreted as a negative quantifier, a rude counterpart to nothing, and not as semantically vacuous. Such a reanalysis is natural enough in varieties of English or Dutch that have negative concord, but would be odd in languages that otherwise have no traces of it. After this reanalysis, dick or bugger all have taken on many of the characteristic properties of English nothing, including some of the typical collocates of that word, like sweet or next to, as the examples in (11) and (12) show: (11) a. I know sweet bugger all about politics.3 b. When all’s said and done, so-called concurrent processing means sweet diddly squat in terms of saving time.4 c. Your advice on Australian Idol means sweet jack shit.5 (12) a. Admittedly, I know next to jack shit about electronica music.6 b. He won us nothing in the last 2 season and sold our best players for next to fuck all.7 c. Sadly, he’s recently admitted he knows next to squat about how the economy works.8 I did a small corpus study of these taboo expressions, by informally collecting occurrences of these items from printed sources, Internet, as well as TV shows and putting them in a database. The sentences were classified according to type of environment. After analyzing the data set, it emerged that these taboo terms appear pretty much only in strictly negative contexts: see Table 1.

3.

From: http://forums.vault9.net/..

4. From: http://ungratefulimmigrant.blogspot.com 5.

From: http://www.defamer.com.au/2007/08/144

6. From: http://www.byroncrawford.com/2005/11/the_best_songs.html 7. From: http://arseblog.com/columns/2007/07/09/gallas-says-players-are-questioning-arsenals-future/ 8. From: http://oncommonground.blogspot.com/2008/02/weekly-poll-it-takes-woman-tostimulate.html



Jack Hoeksema

Table 1.  Distribution of English taboo expressions Shit, diddly squat, dick, fuck all etc Environment

N

%

Not N-word Without

150   12    1

92% 67.3%   0.7%

Total

163

100%

In this small sample, nearly all occurrences were from negative contexts.9 None were found in questions, conditionals, comparatives, in complements of negative or adversative predicates, etc. There are two lessons to be learned from these items. First of all, they clearly show that polarity items do not just get reinterpreted as negative elements after they have become virtually obligatory in negative sentences. On the contrary, these taboo items are fairly infrequent in most people’s speech, with the possible exception of the likes of Tony Soprano, and still they got reinterpreted. Second, it may be more important for this semantic change that the items that undergo it only appear in negative sentences. From the historical record, it emerges that the types of polarity items that undergo grammaticalization as new negation markers are typically minimizers, indicating some minimal quantity or extent, polarity sensitive indefinites, meaning ‘something/ anything’, or generic nouns, meaning thing. In French and Catalan, the markers of negation are derived from minimizers. In Welsh (Willis 2008), the marker ddim derives from a generic word meaning ‘thing’ that had turned into a polarity-sensitive indefinite in the Middle Welsh period. The same can be said about the Arabic postverbal marker of negation ši (Lucas 2007). In Germanic, an indefinite pronoun gets merged with an extra copy of clitic negation, and becomes the new marker of negation (Jäger 2008). Given the wide variety of polarity items, it may come as a bit of a surprise that negation markers are selected from such a small subset. In (13), a partial list of English types of polarity items is given. In Dutch and German the list of polarity items, in spite of some minor differences here and there, is much the same. 9. The fact that the only not strictly negative environment is a PP headed by without does not come as a surprise. Of all the environments in which polarity items may appear, without-PPs seem to be most strongly akin to regular negation. Thus in Greek, emphatic elements of the kanenas-series, the n-words of that language, appear in negative clauses and clauses initiated by xoris ‘without’ (cf. Giannakidou 1998, 2000). In French, aucun and other negative elements appear with negative ne and in PPs introduced by sans (cf. de Swart and Sag, 2002, Deprez and Martineau 2004). In Giannakidou’s (1998) terms, without is antiveridical, in the sense that it implies negation: p without q entails ¬p. Other types of environments maybe downward entailing, or nonveridical, but antiveridicality is a stronger property, since it entails the others.



Chapter 2.  Jespersen recycled 

(13)

Types of polarity items in English – Minimizers: a word, a thing, a syllable, a moment, an inch – Adverbial minimizers: in the least, in the slightest, one bit, the least bit – Taboo items I: a fucking thing, a bloody word, a damn thing, a blasted thing – Taboo items II: shit, jack shit, diddly squat, squat, dick, fuck all, bugger all – Minimizing predicates: say boo to a goose, lift a finger, sleep a wink, bat an eyebrow, know the first thing about, have a clue, have a prayer – Particles: anymore, yet, as yet, either – Indefinite pronouns: any, anybody, anything, anywhere, anyone – Domain wideners: whatsoever, on earth, in the world, at all, in years, in ages, in decades – Domain restrictors: in his right mind, self-respecting, worth his salt – Modal strengtheners: for the life of me, if my life depended on it, for the world, for love or money – Downtoners: all that, exactly, the sharpest knife in the drawer – Modal verbs and idioms: need, humanly possible, strictly necessary – Verbs: budge, faze, mind – Verbal idioms: can be bothered, can care less, can stand/abide, take long, make bones about, give the time of day, would be caught dead in – Litotes: take no for an answer, miss a beat, can deny, a day goes by without – Scalar items: so much as, much less, least of all

The distributional characteristics for a number of minimizers are given in Table 2 (data from the same corpus as data in Table 1). Similar distributional properties can be observed for Dutch and German, as shown in Table 3. Table 2.  Distributional properties of some English minimizers environment

a thing

%

a word

%

a damn thing

%

one bit %

Not N-word Without Question Other

78 10  2 –  3

  84%   11%    2% –    3%

120   36   40    4   18

55% 16% 19%   2%   8%

35  4  1 – –

  88%   10%    2% – –

41  2 –  1 –

93%   5% –   2% –

Total

93

100%

218

100%

40

100%

44

100



Jack Hoeksema

Table 3.  Some Dutch and German minimizers Environment

een woord (D)

%

ein Wort (G)

%

Een bal

%

Negation N-word Without Other

   55 1256   358   121

   3%   70%   20    7%

1 43  6  1

   2%   84%   12%    2%

– 440    5    6

–   98%    1%    1%

Total

1790

100%

51

100%

451

100%

For minimizers and taboo expressions, we see a very strong affinity with negation. Only the items meaning ‘word’ show up in a different environment as well, the complement of without. This is due to a special idiomatic use of the items. Compare the examples in (14): (14)

a. Jones left without saying a word. b. Jones left without a word. c. Jones left without saying a thing. d. *Jones left without a thing. [* on the interpretation of 12c] e. Jones left without a coat.

Normally, without NP means without having an NP, as in (14e). Only without a word has the special interpretation without saying a word (not: without understanding a word, without regretting a word, without receiving a word, etc.). Other minimizers, like a thing, do not have this special interpretation, as you see in (14d). If you take out the idiomatic without a word, but keep cases like without saying a word and so on, the percentage of occurrences in without-clauses drops to about 10%, and you have a distribution rather like that of the other minimizers. Comparing minimizers with other polarity items, such as ever or any, you will notice a sharp contrast. In Table 4, I have listed the results of a corpus study of English ever and its counterparts in Dutch and German, ooit and je(mals). The main thing to notice here is how these expressions are all over the map, showing up in all sorts of environments, not merely in negative sentences. The same can be noted for English any. Occurrences of free choice any, by the way, are excluded from this table. Perhaps you have wondered why English any or ever have not undergone negative reinterpretation. If words like shit or squat can develop into negative quantifiers, why not their more respectable cousins any and ever? Part of the explanation here might be prescriptive grammar, which has fought hard to keep negative concord out of the standard language. If a sentence like I will ever love you is to develop the meaning



Chapter 2.  Jespersen recycled 

Table 4.  Ooit, ever, je(mals) Item→ environment↓ Comparative Conditional Hardly Negation Negative Predicate Superlative Question Without Other

ooit N=17.304

ever N=3082

je(mals) N=792

20 10  1 22  3 10 24  5  5

13  8  2 25  5 17 19  1 10

21  5  5 18  4 13 21  6  5

Table 5.  Distribution of polarity-sensitive any item→ environment↓ Comparative Conditional Hardly Negation Negative predicate Superlative Question Without Other

any N=3718

%

  216   401    20 1736   343    45   699   141   117

 6 11 0.5 36  9  1 19  4  3

I will never love you, there must be a prior stage in which I won’t ever love you is reinterpreted as I won’t never love you, with negative concord: (15) Stage I: I won’t ever love you [standard interpretation] Stage II: I won’t ever love you [ever = never; due to negative concord equivalent in interpretation with Stage I] Stage III: I will ever love you [ever = never; loss of redudant negation] This is probably why the taboo expressions, which are not subject to prescriptive grammar to the same degree, are more prone to undergo semantic reinterpretation than other polarity items. However, we may also consider the distributional characteristics of the items in question as important preconditions. It would seem to be much harder to reinterpret an item like ever as a negative quantifier, if the majority of its occurrences



Jack Hoeksema

are not even in negative sentences. My expectation, therefore, is that any and ever are unlikely to undergo the kind of reinterpretation needed to partake in the Jespersen Cycle, and that one will be hard-pressed to find a dialect or variant of English where it does, unless, of course, it is from a dialect where these words have come to have a rather more restrictive distribution, comparable to the minimizers we have looked at. At the same time, we have a little bit more grip on the issue why languages typically employ minimizers in the Jespersen Cycle. I take semantic reinterpretation of indefinites as negative quantifiers to always involve, at least initially, an interpretation error, not unlike the errors in noisy channels studied by information theory. The overall interpretation of an utterance is correctly computed, but not in a compositionally-correct manner. The wrong item is taken to express negation. A notion from information theory might come in handy here, namely mutual information, which is a measure of redundancy. When two items tend to co-occur, like negation and polarity items, their relative informational value decreases. If two variables x and y are completely independent, the chance of the two co-occurring is p(x), the probability of x, times p(y), the probability of y. If the two items are not independent, the probability of the two co-occurring, p(x,y), is greater than or smaller than p(x) times p(y). So if you divide p(x,y) by (p(x) times p(y)), you get a measure for whether the two are more likely to co-occur or less likely to co-occur than would be predicted on the basis of their individual probabilities. By taking the log of that division, you get a measure, called specific pointwise mutual information. If it is 0, there is no effect of one variable on the other, if it is below zero, the two variables are less likely to co-occur than their individual probabilities would predict, and if it is positive, the two are more likely to co-occur than you might have guessed given their individual probabilities. See the formula in (16): p(x,y) (16) MI(x,y) = log _________ p(x)×p(y) The notion of mutual information is used in corpus-based study of collocations (Church and Hanks 1990, Manning and Schütze 1999). Clearly, minimizers are prime examples of expressions with high mutual information with respect to negation, other polarity items less so. Since we want to look at the probabilities of various polarity items co-occurring with negation, one of the two variables x and y in (15) is given, namely the probability of negation itself. If we likewise forget about the logs, we can simplify the formula to the one in (17): p(x,y) (17) ______ = p(x|y) p(y)) which is the conditional probability of x given y. In our case, the conditional probability of negation, given some polarity item. The larger this probability, the greater are the



Chapter 2.  Jespersen recycled 

chances of misconstrual and reinterpretation. For minimizers and taboo terms, this conditional probability is well in the 90% range. For other items, it is a lot lower. Note that conditional probability is not a symmetric notion. The probability of negation, given the presence of some minimizer, may be very high, while the probability of that minimizer, given the presence of negation, is very low. As we see from the example of the English taboo NPIs, the latter probability does not seem to matter much. They underwent reinterpretation, regardless of their low overall frequency. Unfortunately for the historical study of the Jespersen Cycle, we usually lack good corpus data that might help inform us about conditional probabilities. This is why it is important to study these changes in contemporary stages of languages, using the present to explain the past, to quote Bill Labov. Another point of interest is the reversibility of the change. If we can reinterpret polarity items as negative quantifiers, can we also reinterpret negative quantifiers as nonnegative polarity items in negative concord languages? The answer to this question appears to be affirmative (cf. Hoeksema 1997). In languages like Middle Dutch, we see negative quantifiers, clearly marked as such by the presence of the negative prefix n-, showing up in polarity contexts without a negative meaning: (18)

Die bliidste soudic wesen dan the happiest would-I be then Die nie man sach op erterike; That never man saw on earth10 `I would be the happiest [woman] that one ever saw on earth.’

The lack of concord marking on the finite verb, typical for such cases, shows that we are no longer dealing with an n-word, but with a homophonous polarity item.

5. Adverbials from minimizers Now that we have seen how minimizers might be prone to reanalysis as negative quantifiers, we are still at a loss how these negative quantifiers may become negative adverbs. Clearly, the change from the quantifier naught to the adverb not is a big one. There are several scenarios describing how this might have happened. One is misanalysis of negative quantifiers that serve as objects to optionally transitive verbs as adverbial modifiers (cf. Jäger 2008, Lucas 2007, Bayer 2007 for suggestions along these lines): (19) Jones [neg ate naughtDP] → Jones [[neg ate] naughtADV] This adverb then spreads to other types of VP, on its route to becoming the main exponent of negation. 10. Example taken from Hein van Aken’s medieval Dutch romance Roman van Heinric ende Margriete van Limborch.



Jack Hoeksema

Table 6.  Some English verbs of indifference Environment

Care (N = 792)

Matter (N = 406)

Negation Other negative Interrogative Affirmative

53% 12% 15% 20%

57%   7% 13% 20%

A slightly different scenario is one where the n-word is used as a measure-like argument with verbs like matter or care. These verbs take on a variety of nominal adjuncts which measure the degree of indifference felt by the human experiencer: (20)

a. b. c. d. e. f. g.

Jones did not care much. Jones did not care a whit. It does not matter a jot to Jones. It does not matter much to Jones. It matters nothing to Jones. Jones cares nothing about it. It matters very little to Jones.

Note that these nominal adjuncts are entirely optional: (21) a. Jones did not care. b. It does not matter to Jones. Dutch also has a largish class of similar verbs and verbal expressions. What is interesting for our purposes about these particular verbs is that they are quasi-polarity items, in the sense that 80% or more of their occurrences are in environments that are wellknown contexts for negative polarity items (cf. Hoeksema 1994). Table 6 lists some data from that paper. While these verbs are probably not as frequent as some of the most common optionally intransitive verbs, they are much more likely to occur with negative quantifiers. After all, how often do we say ‘John ate nothing’ or something of the kind? And so these verbs of indifference might deserve some special scrutiny as possible hosts for the change from nominal argument to adverbial modifier. However, as soon as this change has taken place, there is likely to be a quick spread to other contexts, and so it may be impossible to tell, with any degree of certainly, where the change actually originated. Adverbial uses are by no means restricted to negative quantifiers turning into negation proper, like English naught, Dutch niet, or German nicht. We see similar changes



Chapter 2.  Jespersen recycled 

among minimizers. For instance, one bit is clearly adverbial in English. When it is used as a nominal complement, the result is degraded: (22) a. Jones did not like her one bit. b. ?Jones did not feed her one bit. c. *Jones did not eat one bit. Even more striking is the case of adverbial any: (23)

a. b. c. d.

That did not help the soldiers any. It hasn’t changed him any. Would that bother her any? Don’t you worry about it any.

So in some cases, the indefinite that became the marker of negation in the course of a Jespersen Cycle may well have been adverbialized even before it got reinterpreted as a negative quantifier. We will need extensive and refined corpus data to see for each case in what way a nominal quantifier may have developed into a negative head.

6. The Neg-P hypothesis This leads me to the final part of this paper. We have looked at polarity items turning into negative quantifiers which turned into adverbial elements. At some point, there usually is a system of double negation, involving some kind of negative head and the new adverb of negation. Commonly, this situation is given a grammatical treatment in terms of some version of the NegP-hypothesis (cf. Pollock 1989, Haegeman and Zanuttini 1991, Ouhalla 1990, Zanuttini 1997, among others). The old negator is typically the head of NegP, and the new adverb is a specifier of that head. The final stage of the Jespersen Cycle is the one which involves the disappearance of the old negator and its replacement by the new item. To make the Cycle truly circular, one needs to place the specifier in the position of the head. Under most people’s understanding of modern generative syntax, this cannot be done by regular rules moving elements from SpecNeg to Neg0. The Spec position is a phrasal position, and Neg0 is a lexical head. So how do we get from Spec to head? The only remaining road seems to be syntactic reanalysis, by which the former specifier is reanalyzed as the head of the construction. This is an option which is especially attractive when the specifier is already a single word, such as is the case with French pas. Elly van Gelderen has proposed a principle called Head Preference Principle: (24) Head Preference Principle Be a head, rather than a phrase

(Van Gelderen 2004b)



Jack Hoeksema

For items in the Spec of NegP that are small enough to look like heads, rather than phrases, this principle is a call to arms, to rise from the ranks to assume a commanding position as the head of NegP when this position is no longer occupied by another element. In part, as van Gelderen points out, the Head Preference Principle is a restatement of an old observation from grammaticalization theory, namely that function words tend to originate as nonfunctional words from open classes. However, by combining this older insight with notions from modern syntax, in particular the complex array of mechanisms surrounding the NegP hypothesis, with its associated verb movements to various head positions, we reach a stage where a principle such as the HPP actually makes some predictions. For modern spoken French, for instance, it predicts imminent reanalysis of pas as a head. This will entail that it changes position with respect to the verb, either by becoming a clitic to it, like ne used to be, or, if it is not, by blocking movement of the verb to positions higher than NegP due to the Head Movement Constraint (Travis 1984). Whether this prediction is likely to ever come true is of course mere speculation at this point. However, I submit that in languages with stable word order patterns, such as modern French or English, such changes are unlikely to occur. Word order patterns of functional elements are especially resistant to change. We see this clearly in a language like Dutch, which has had simple negation by a single adverb ever since the demise of negative concord in the 17th century. During this period, the position of the negation element niet has never changed. Well, actually, there is one example in the literature that there was a change, a recent one in fact. In their history of Dutch in the 20th century, Van der Horst and Van der Horst (1999: 286) state: “Het ziet ernaar uit dat het woord niet in de afgelopen eeuw een andere plaats gekregen heeft, namelijk meer naar voren.” [It appears that the word niet has received a different position in the course of the last century, namely more to the left].

They illustrate this with examples such as (25): (25) ge hoeft dien brief zoo stevig niet vast te houden you need that letter so tight not to-hold-on-to ‘You need not hold on to that letter so tightly.’ (L. van Deyssel, 1889) where more modern writers would prefer to write (26): (26) ge hoeft die brief niet zo stevig vast te houden. However, these examples are not so much evidence for a different position of negation as they are of the decline of scrambling in the 20th century. Example (24) is an instance of scrambling of the adverbial phrase zo stevig across negation. As I have shown elsewhere (Hoeksema 2003, 2006), scrambling has been declining since the 18th century, at differing rates for different types of expressions. Scrambling, of course, does not involve heads changing position, but constitutes an optional movement process of phrasal material in the middle field, both in Dutch and German. When you focus on the position



Chapter 2.  Jespersen recycled

of negation in sentences like 25, it may seem that the position of negation has changed, but actually, it is the position of the scrambled element. This is clear from the fact that scrambling across other elements than negation is also on the decline, something that would be unaccounted for if only the position of negation were to have changed. But that leaves us with a bit of a puzzle. If the position of negation in Dutch has not changed at all since the Middle Ages, in spite of the disappearance of negative concord in early modern Dutch, what does that tells us about the Head Preference Principle? At this point, a number of options suggest themselves. One might suppose that the Head Preference Principle is held in check by countervailing forces. One might postulate an Inertia Principle for historical change that prefers changes which do not affect the surface order of constituents. In the absence of catastrophic changes due to heavy language contact, or severe paradigmatic pressure, the position of functional elements is quite fixed. Alternatively, it might be worthwhile to consider abandoning the NegP hypothesis, and to view negation as either adverbial in nature, for languages such as Dutch, German or Norwegian, or part of the inflectional system, in languages such as English. Treating negation as an adverbial adjunct in Dutch or German has the advantage that we do not have to say anything special about constituent negation in sentences such as (27), where the negative adverb acts like any other focus adverb, being adjoined to a phrasal projection: (27) a. Niet in alle landen sneeuwt het in de winter. not in all countries snows it in the winter ‘It does not snow in all countries in the winter.’ b. Niet eens zo lang geleden sneeuwde het nog. Not even so long ago snowed it yet ‘It snowed not even that long ago.’ The fact that verbs move to C in Germanic without any hinder from negation is then simply what is expected, and not something that could change as soon as negation gets reassigned from Spec of NegP to being a Neg-head. Of course I don’t want to claim to have found the solution to all or even many syntactic problems surrounding negation. To the contrary. However, it does not seem right to treat constituent negation as completely divorced from sentential negation, as the NegP theory requires one to assume, even for languages which employ the same expression for both kinds of negation, such as English or French. To return to the Jespersen Cycle, let me conclude this section with the following remarks. For Dutch or German, my account no longer assumes a cyclical change from negative head to double negation by Spec-Head agreement, followed by Spec-only negation, possibly to be turned into head only negation. Rather, we have clitic negation on an inflectional head in the first stage, let us say early Old Dutch, then n-word + clitic negation in Middle Dutch, and finally adverbial or n-word negation in modern Dutch. It is unclear whether or how the current system might ever turn back into one of clitic negation. I suspect the system is stable, and that the new millennium won’t see





Jack Hoeksema

too much change in this area, assuming, of course, that Dutch and German do not disappear under the onslaught of English. But that is a different story altogether.

7. Conclusions In this paper, I review the various linguistic changes that together constitute the Jespersen cycle. My focus in this paper has been on the European languages mainly, and so certain types of change, say from prohibitive verb to marker of negation, have not been discussed here. I have argued that one particular change, from negative polarity item without negative import to negative quantifier, is limited to those items which occur primarily in strictly negative environments, rather than the much larger set of contexts where e.g. English any may show up. I have motivated this claim with data from English taboo terms such as diddly squat, and I hope that others will feel inspired by this hypothesis to test it further. I have also made a suggestion regarding the change from nominal quantifier to adverbial negation that takes place in many (though certainly not all) languages as part of the Jespersen cycle. The suggestion is that the change may be the result of reanalysis in constructions with polarity-sensitive verbs that take measure-like complements, such as matter and care. At the moment, this suggestion is very speculative, as are alternative suggestions. The paper ends with a brief discussion of the Head Preference Principle (van Gelderen 2004b) and the status of NegP. I am skeptical about the possibility that negation will change linear position as a result of reanalysis when Spec of NegP turns into Neg0. This may happen whenever verb movement through NegP interferes with the position of negation. Here, too, more evidence needs to be gathered, in this case for sudden jumps in linear position, to ascertain the validity of such an account.

References Bayer, Josef. 2009. Nominal negative quantifiers as adjuncts. Journal of Comparative Germanic Linguistics 12(1): 5–30. Bernini, Giuliano & Ramat, Paolo. 1996, Negative Sentences in the Languages of Europe: A Typological Approach. Berlin: Mouton de Gruyter. Breitbarth, Anne. 2008. A hybrid approach to Jespersen’s Cycle in West Germanic. Ms, University of Cambridge. Church, Kenneth Ward & Hanks, Patrick. 1989. Word association norms, mutual nformation, and lexicography. Proceedings of the 27th Annual Meeting of the Association for Computational Linguistics. Deprez, Viviane. 2000. Parallel (a)symmetries and the structure of negative expressions. Natural Language and Linguistic Theory 18(2): 253–342.



Chapter 2.  Jespersen recycled  Deprez, Viviane & Martineau, France. 2004. Microparametric variation and negative concord. In Contemporary Approaches to Romance Linguistics. Selected Papers from the 33rd Linguistic Symposium on Romance Languages, Julie Auger, Clancy Clements & Barbara Vance (eds),139–158. Amsterdam: John Benjamins. . Zeitschrift für Ägyptische Sprache und Altertumskunde Gardiner, Alan H. 1904. The word 41:130–135. Gelderen, Elly van. 2004a. Economy, innovation and prescriptivism. Journal of Comparative Germanic Linguistics 7: 59–98. Gelderen, Elly van. 2004b. Grammaticalization as Economy [Linguistik Aktuell-Linguistics Today 71]. Amsterdam: John Benjamins. Gelderen, Elly van. 2008. Negative cycles. Linguistic Typology 12:195–243. Giannakidou, Anastasia. 1998. Polarity Sensitivity as(Non)veridical dependency [Linguistik Aktuell-Linguistics Today 23]. Amsterdam: John Benjamins. Giannakidou, Anastasia. 2000. Negative...concord? Natural Language and Linguistic Theory 18: 457–523. Haegeman, Liliane & Zanuttini, Raffaela. 1991. Negative heads and the NEG criterion. The Linguistic Review 8(2–4): 233–251. Hoeksema, Jack. 1994. On the grammaticalization of negative polarity items. Proceedings of the Twentieth Annual Meeting of the Berkeley Linguistics Society: 273–282. Hoeksema, Jack. 1997. Negation and negative concord in Middle Dutch. In Negation and Polarity: Syntax and Semantics [Current Issues in Linguisic Theory 155], Danielle Forget, Paul Hirschbühler, France Martineau & María Luisa Rivero (eds), 139–158. Amsterdam: John Benjamins. Hoeksema, Jack. 2002. Minimaliseerders in het standaard Nederlands. Tabu 32(3–4): 105–174. Hoeksema, Jack. 2003. In het minst: Eigenschappen en ontwikkeling van een negatief-polaire uitdrukking en de neergang van focus scrambling. Tabu 33(1–2): 27–50. Hoeksema, Jack. 2006. Dutch focus scrambling and the constant rate hypothesis. Talk given at the University of Pennsylvania. Horn, Laurence R. 2001. Flaubert triggers, squatitive negation, and other quirks of grammar. In Perspectives on Negation and Polarity Items [Linguistik Aktuell-Linguistics Today 40], Jack Hoeksema, Hotze Rullmann, Víctor Sánchez-Valencia & Ton van der Wouden (eds), 173– 200. Amsterdam: John Benjamins. van der Horst, Joop & van der Horst, Kees. 1999. Geschiedenis van het Nederlands in de Twintigste Eeuw. The Hague/Antwerp: Sdu/Standaard. Jäger, Agnes. 2008. History of German negation [Linguistik Aktuell-Linguistics Today 118]. Amsterdam: John Benjamins. Jespersen, Otto. 1917. Negation in English and Other Languages. Copenhagen: Host. Kemenade, Ans van. 2000. Jespersen’s cycle revisited: Formal properties of grammaticalization. In Diachronic Syntax. Models and Mechanisms, Susan Pintzuk, George Tsoulas & Anthony Warner (eds), 51–74. Oxford: OUP. Kiparsky, Paul & Condoravdi, Cleo. 2006. Tracking Jespersen’s cycle. In Proceedings of the 2nd International Conference of Modern Greek Dialects and Linguistic Theory, Mark Janse, Brian Joseph & Angela Ralli (eds),172–197. Patras: University of Patras. Labov, William A. 1972. Negative attraction and negative concord in English grammar. Language 48: 773–818. Ladusaw, William A. 1992. Expressing Negation. Proceedings of the conference on Semantics and Linguistic Theory 2: 237–259. Columbus OH: The Ohio State University.



Jack Hoeksema Ladusaw, William A. 1993. Negation, indefinites, and the Jespersen cycle. BLS 19: 437–446. Lucas, Christopher. 2007. Jespersen’s cycle in Arabic and Berber. Transactions of the Philological Society 105(3): 398–431. Manning, Christopher D. & Schütze, Hinrich. 1999. Foundations of Statistical Natural Language Processing. Cambridge MA: The MIT Press. Meillet, Antoine. 1912. L’évolution des formes grammaticales. Scientia 12:384–400. (Reprinted in Meillet, Antoine. 1926. Linguistique historique et linguistique générale. 130–148. Paris: H. Champion). Ouhalla, Jamal. 1990. Sentential negation, relativized minimality and the aspectual status of auxiliaries. The Linguistic Review 7: 183–231. Pollock, Jean-Yves. 1989. Verb Movement, UG and the structure of IP. Linguistic Inquiry 20: 365–424. Postal, Paul M. 2004. Skeptical linguistic essays. Oxford: OUP. Postma, Gertjan. 2002. De enkelvoudige clitische negatie in het Middelnederlands en de Jespersen-cyclus. Nederlandse Taalkunde 7(2): 44–82. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change. A Minimalist Approach to rammaticalization. Cambridge: CUP. Rowlett, Paul. 1998. Sentential Negation in French. Oxford: OUP. Schwenter, Scott A. 2006. Fine-tuning Jespersen’s cycle. To appear in a festschrift for Larry Horn. Swart, Henriëtte de & Sag, Ivan A. 2002. Negation and negative concord in Romance. Linguistics and Philosophy 25: 373–417. Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. PhD dissertation, MIT. van der Auwera, Johan & De Vogelaer, Gunther. 2008. Negation and quantification. In Syntactic atlas of the Dutch dialects, Vol. II, Sjef Barbiers, Johan van der Auwera, Hans Bennis, Eefje Boef, Gunther De Vogelaer & Margreet van der Ham 58–72. Amsterdam: Amsterdam University Press. Willis, David. 2008. A minimalist approach to Jespersen’s cycle in Welsh. Ms. Wouden, Ton van der. 1994. Negative Contexts. PhD dissertation, University of Groningen. Wouden, Ton van der. 1997. Negative Contexts: Collocation, Polarity and Multiple Negation. London: Routledge. Zanuttini, Raffaella. 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford: OUP. Zeijlstra, Hedde. 2004. Sentential Negation and Negative Concord. PhD dissertation, University of Amsterdam. Zwarts, Frans. 1998. Three types of polarity. In Plurality and Quantification, Fritz Hamm & Erhard Hinrichs (eds), 177–238. Dordrecht: Kluwer.

chapter 3

The Jespersen Cycles*1 Johan van der Auwera University of Antwerp

This chapter analyzes the kind of renewal of clausal negators referred to with the term of ‘Jespersen Cycle’. It describes how a negator may collocate with a strengthener, which may later become an additional exponent of negation and possibly the only one. Through an analysis of a century’s worth of scholarship, not only giving pride of place to Jesperen (1917) but also Gardiner (1904) and Meillet (1912), the paper sets out to describe parameters such as the role of emphasis, the identity or difference of the old and the new negator, and the question whether or not the stage of the two exponents simplifies into a single exponence stage or takes us to a three negator stage. In so doing, the paper also advocates taking Jespersen cycle research beyond the confines of Europe and the Mediterranean.

Introduction This paper deals with Jespersen’s Cycle, also known as the ‘Jespersen Cycle’ and ‘Negative Cycle’. I will offer an account that is more general than the one envisaged by * This paper was presented at the April 2008 Tempe conference on Linguistic Cycles and at the May 2008 Helsinki NORMS Workshop on dialectology (NORMS = Nordic Center of Excellence in Microcomparative Syntax). I have also profited from attending the June and August 2008 Leverhulme Trust Seminars on cycles of grammaticalization (Birmingham). Special thanks are due to Theresa Biberauer (Cambridge), Ludovic De Cuypere (Ghent), Gunther de Vogelaer (Ghent), Robert Early (Port Vila), Franck Floricic (Paris), Eitan Grossman (Jerusalem), Liliane Haegeman (Lille), Jack Hoeksema (Groningen), Mark Janse (Ghent), Mena Lafkioui (Ghent), Pierre Larrivée (Birmingham), Mair Parry (Bristol), Amina Mettouchi (Nantes), Matti Miestamo (Helsinki), Samia Naïm (Paris), Lene Schøsler (Copenhagen), Joop van der Horst (Leuven), Elly van Gelderen (Tempe), Willy Van Langendonck (Leuven), and Klaas Willems (Ghent). Special thanks are also due to Jean Winand (Liège), who drew my attention to his 1997 paper, which then directed me to Gardiner (1904). I also acknowledge project P6/44 on Grammaticalization and (Inter)Subjectification of the Belgian Federal Government for financial support. The paper uses the following abbreviations: COP ‘copula’, NEG ‘negation’, NOM ‘nominalization’, PL ‘plural’, R ‘realis’, REL ‘relative’, SG ‘singular, SUBJ ‘subjunctive’, X ‘element other than negation’, 1 ‘first person’, 2 ‘second person’, and 3 ‘third person’.



Johan van der Auwera

Jespersen (1917) and others. This paper has four sections. In Section 1, I sketch what Jespersen (1917) meant, how his view has been represented, and how it is partially wrong, at least for the data that he had in mind. I discuss the alternative view, also going under the label of the ‘Jespersen Cycle’ and spell it out in some detail. I will also argue that the alternative view has room for the original Jespersenian idea, in more than one way, and the resulting account will describe exactly eight possible trajectories or ‘cycles’. In Section 2, I discuss an additional type of Jespersenian negation renewal, which is worthy of being called a ‘Jespersen Cycle’ as well, not least because it fits the system described in Section 1. Section 3 develops the typology even more. The typology allows at least four questions, the answers to which will be positive and show yet more variation. Section 4 is the conclusion. For the analysis of Jespersen’s original idea, I will mostly rely on French, which is one of the languages that Jespersen (1917) had in mind and for which we possess an enormous amount of research. For the development of the typology, I will primarily tap micro-variational and macro-variational sources. For micro-variation, I will use materials and analyses concerning Dutch and especially Belgian Dutch negation, which have recently come to the foreground (Barbiers et al 2009, Neuckermans 2008).1 For macrovariation, I will use and partially reinterpret the facts of the Vanuatu language Lewo, which have stood as a challenge for Jespersen cyclists since Early (1994a, 1994b). Despite the goal of reaching a general account, let me point to some restrictions. This paper is only about the development of negative strategies that involve something like a doubling stage. To take the text book example and to present it in a simplified way, French once had a ne negator, it is heading for pas, but there is also a middle stage with both ne and pas. Pas is the newer strategy and, to reach that stage, the language went through a doubling ne… pas stage. Negative strategies need not pass through any such stage, however. Negative markers may directly develop from verbs (e.g. from a verb meaning ‘not exist’) or nouns (e.g. from a noun meaning ‘taboo’) in ways that have been described by Croft (1991), van der Auwera (2006), van Gelderen (2008), and van der Auwera (In print b). Furthermore, this paper focuses on clausal negation, the negation that has scope over an entire clause or proposition. So not much will be said about the development of ‘negative quantifiers’, such as pronouns meaning ‘nobody’ or adverbs meaning ‘never’, even though the development of clausal negation and quantificational negation are interconnected. There will be nothing on negative concord either. Negative polarity will be mentioned quite often, but again this matter will not be focused on. Finally, there will be nothing on language contact or areal typology, even though this dimension has been crucial in quickening the interest in the 1. Barbiers et al (2009) and Neuckermans (2008) emanate from a Flemish Dutch dialect syntax project called ‘SAND’ (‘Syntactische Atlas van de Nederlandse Dialecten’), which received funding from 2000 to 2003, but continues to spawn and influence publications and new projects. Of course, Belgian Dutch negation had become important in theories of negation before, esp. thanks to Haegeman (e.g. Haegeman 1995). Other publications generated by SAND referred to in this paper are van der Auwera & Neuckermans (2004) and Zeijlstra (2004).



Chapter 3.  The Jespersen Cycles 

Jespersen cycle through the pivotal work of Bernini and Ramat (1992, 1996) (see also van der Auwera In print a) and remains important up to today (e.g. Lucas 2008). As already adumbrated in the above, I prefer the term ‘Jespersen Cycle’ to ‘Jespersen’s Cycle’ (because there is too much variation to warrant the definitizing genitive) and to ‘Negative Cycle’ (to single out the cycles with a doubling stage from all other negative cycles).

1. The two Jespersen accounts 1.1

What did Jespersen (1917) mean?

The paragraph from Jespersen (1917) that has been most influential, not least because Dahl (1979: 88) drew attention to it and coined the phrase ‘Jespersen’s Cycle’, is the following: The history of negative expressions in various languages makes us witness the following curious fluctuation: the original negative adverb is first weakened, then found insufficient and therefore strengthened, generally through some additional word, and this in turn may be felt as the negative proper and may then in course of time be subject to the same development as the original word. (Jespersen 1917: 4)

This scenario is usually represented as involving three, four, or five stages. For French, the three-stage model has slots for ne, ne... pas and pas. The five stage model is essentially the same but it makes explicit that the changes from ne to ne... pas and from ne... pas to pas involve the in-between stages of pas, resp. ne being optional. The four stage models typically2 include the three stage model but either add non as the phonetically stronger ancestor to ne or make clear that the pas that is added to ne was not originally negative—I will symbolize this with the subscript X instead of NEG. The representation in (1) is limited to the French cycle. One could, of course, go back to Latin, and list ne—another ne, i.e. a Latin ne different from the French one—as the ancestor to non, as Jespersen (1917:  7) did, and recently also Muller (1991:  206) and Lenz (1996: 183). This is not represented in the schemas in (1). I have chosen French instead of English, for English is more complicated in that it has arguably started a new cycle or, at least, a new development. English has had a development from ne to not via ne... not, not unlike what we have in French, but English now also has a weakened n’t cliticized to a preceding auxiliary and it also saw the development of do periphrasis. Lenz (1996: 184–5), Anderwald (2002: 19) and Zeijlstra (2004: 56) thus have the three or five stages but they add one or more stages with a not or n’t in front of the lexical verb, as in I do not say and I don’t say. When Jespersen (1917: 9–11) discusses English, he also lists I do not say and I don’t say, even as separate stages, and adds them onto a three 2. An example of an untypical case is Van Kemenade (1999: 148; also in Fischer et al. 2000: 305). Here the four stages are stages 1, 2, 4, and 5 of the five stage model.



Johan van der Auwera

stage model. I will leave such developments out of account, as they follow the cyclical end point of the cycle starting with ne and ending with not. (1) Three stages

1

neNEG

Four stages A 1

nonNEG

2

neNEG

Five stages B 1

neNEG

2

neNEG... pasX

1

neNEG

2

neNEG (… pasNEG)

2 neNEG … pasNEG 3 neNEG …pasNEG 3 neNEG … pasNEG 3 neNEG … pasNEG 4 (neNEG …) pasNEG 3

pasNEG

4

pasNEG

4

pasNEG

5

pasNEG

(2) shows where these models can be found; the list is far from exhaustive. (2) Three stages

Burridge (1983: 36); Bernini & Ramat (1996: 33), Haspelmath (1997: 203), Zanutttini (1997: 11–14), Horn (1989: 455), Hoeksema (1997: 140), Horn (2001: 190), Roberts and Roussou (2003: 154–155), van der Auwera & Neuckermans (2004: 458), Mazzon (2004: 5), Willis (2005), Lucas (2007), Jäger (2008) Four stages – A Dahl (1979: 88), Muller (1991: 206), Lenz (1996: 183–4), Larrivée (2004: 18–19), van Gelderen (2008: 210) Four stages – B Schwegler (1988), Schwegler (1990: 158), Schwenter (2006: 327) Five stages Donhauser (1996), Honda (1996: 207), Beukema (1999), Anderwald (2002), van der Auwera & Neuckermans (2004: 458), Zeijlstra (2004), Willis (2005)

The five stage representation has two stages with two strategies. ‘neNEG (… pasNEG)’, for instance, in effect says that the language has both the neNEG and the neNEG … pasNEG strategy. This is a good way to visualize that two strategies are in competition. But this method obviously makes for more complex representations, and one can avoid quite of bit of complication by simply stating that at one period a language could be at more than one stage, not least because some constructions may be ahead of other constructions. For instance, in Brabantic Dutch of 1650 declaratives barely allowed nietNEG (the counterpart of to pasNEG) and still overwhelmingly used enNEG … nietNEG (the counterpart to neNEG... pasNEG), whereas the statistics are exactly the opposite for prohibitives (Burridge 1983:  33, see also Hoeksema 1997:  145, van der Auwera 2006:  18).



Chapter 3.  The Jespersen Cycles 

Another point to note in this respect is that the two transition stages of the five stage model each only have two strategies. There is no reason why a construction could not at any one period allow more than two strategies. Consider the example in (3). (3) French a. Il ne peut venir ce soir. b. Il ne peut pas venir ce soir. c. Il peut pas venir ce soir. he neg can neg come this evening ‘He can’t come tonight.’ (3a) exemplifies the archaic strategy, (3b) is standard written French, and (3c) is typical for a casual spoken register. So the three stages may actually be said to coexist, at least for some constructions (like with the verb pouvoir ‘can’ illustrated in (3)). And yet there is no doubt that neNEG came first, that neNEG … pasNEG followed and that solitary pasNEG was the last one. While admitting the value of a model in which the stages show competing variants (for a plea to that effect, see e.g. Martineau and Mougeon 2003: 146), I will use the simpler representations in this study, i.e. the ones without transition stages (except when in (13) below the discussion crucially concerns the nature of transitional stage). In (4), I reconstruct the cycle in six stages, again using French, and I then paraphrase it using the words of Jespersen (1917) (between single quotation marks). The reason I choose six stages is that it captures Jespersen (1917) better than the simpler schemes. (4)

Stages 1 2 3 4 5 6

Strategies nonNEG neNEG neNEG … pasX neNEG … pasNEG neX … pasNEG pasNEG

NeNEG is the formally ‘weakened’ form of the ‘original negative adverb’ nonNEG. At stage 2 only the weakened form occurs, at stage 1 only the original form and the weakening takes place in between. Between stages 2 and 3 neNEG ‘is found insufficient’ and ‘strengthened … through some additional word’. The strengthening is both formal and semantic: the addition of the word makes for a formally new strategy, and the additional word has a meaning of its own, but it is not that of clausal negation yet—that is why the subscript ‘X’ is used and not ‘neg’. At stage 3 this strengthened strategy has ousted the old one. At stage 4 the ‘additional word’ partakes in the negative meaning as such and the strengthening effect is lost. Between stages 4 and 5 the second part of the negator is beginning to ‘be felt as the negative proper’, a process that is completed at stage 5 at the semantic level. Now neX is an ‘additional word’; I again mark the absence of the negative function with the subscript ‘X’. At stage 6, the process is finished



Johan van der Auwera

at the formal level: the non-functional additional word has disappeared. The representation in (5) repeats that of (4) and it adds the notions of formal vs. semantic weakening and strengthening.

(5) Stages 1



Strategies nonNEG

2

neNEG

3

neNEG … pasX

4

neNEG … pasNEG

5

neX … pasNEG

6

formal weakening formal & semantic strengthening semantic weakening semantic weakening formal weakening

pasNEG

The representations in (4) and (5) are a little bit more detailed than the usual ‘reconstructions’. It represents essentially the three stage model, enriched with the nonNEG stage of the A type of the four stage model, the neNEG pasX stage of the B type of the four stage model, and a parallel neX pasNEG stage. But (4)/(5) is still a simplification and there are several issues that demand a comment. Let me mention two. First, it is generally assumed that pas was introduced in the context of movement verbs at a stage predating the earlier texts (e.g. Buridant 2000:  60) and then spread to other verbs. Jespersen’s (1917) quote does not deal with this and for this reason (4)/(5) does not show this stage either, nor will I go into this matter later. Second, the schema abstracts from word order properties. The citation from Jespersen (1917) does that too. Nevertheless, it is obvious that word order does matter, as the positions of French neNEG and pasNEG are markedly different: whereas neNEG precedes the finite verb, pasNEG follows it. It is less clear, however, how exactly word order matters, at what stage, whether it is a consequence or a cause, and to what extent there could be cross-linguistic generalizations. In any case, I will leave these issues out of consideration, not least also because proposals positing strong correlations between the placement of negation and general word order properties and aiming to explain the cycle (e.g. Vennemann 1974, Harris 1978, van der Horst and van der Wal 1979) have been criticized and largely abandoned (e.g. Schwegler 1983).



Chapter 3.  The Jespersen Cycles 

1.2

What do many other linguists mean?

The scenario sketched in (4)/(5) is not in fact the only one that goes under the name of ‘Jespersen cycle’. In an alternative view (see (8) for references and see also Breitbarth In print for a similar and independently arrived at classification), the claim is not that neNEG is weak and that it is therefore strengthened. Instead, the claim is that the language has both a neutral and an emphatic negative construction, and that the latter may lose its emphatic flavor, become a competitor to the erstwhile negation, and eventually replace it. So what starts off the process is not the weakness of the original negator, but the general process of the inflation of an emphatic use and the consequent bleaching. (6) is a first attempt to represent the crucial stages of the alternative scenario, again illustrated with French. (6) Stages n n+1

Emphatic strategies Neutral strategies neNEG … pasNEG neNEG neNEG... pasNEG

At stage ‘n’, the language has two strategies, an emphatic neNEG … pasNEG and a neutral neNEG. Then neNEG … pasNEG bleaches and loses the emphatic meaning, thereby becoming a competitor to neutral neNEG. At stage n+1, the fully bleached neNEG … pasNEG has replaced the earlier and simpler neutral neNEG strategy. (7) repeats (6) and identifies the changes. The loss of the emphatic meaning is semantic weakening and the simultaneous replacement of the neutral simple neNEG by the more complex but also neutral neNEG … pasNEG is formal strengthening.

(7) Stages Emphatic strategies n neNEG … pasNEG n+1



Semantic weakening

Neutral strategies neNEG neNEG …. pasNEG

Formal strengthening

Somewhat paradoxically, the appearance of neNEG … pasNEG in the domain of neutral negation instantiates both weakening and strengthening: the weakening is semantic and relates to the emphatic neNEG … pasNEG construction, and the strengthening is formal and relates to the simple neNEG construction. I stressed that under the alternative scenario neNEG is not claimed to have been weak and in need of strengthening, but this claim only concerns its function as a neutral strategy. Of course, we see that neNEG is a component of the emphatic strategy as well, and one can grant that for emphatic purposes neNEG, as an unstressed preverbal clitic, was indeed weak, and it could do with formal and semantic strengthening. So something like stages 1 to 4 of scheme (4)/(5) are indeed relevant for the full picture of French negation, even under the alternative scenario, but, crucially, this component of formal and semantic strengthening of neNEG with an ‘additional word’ pasX only



Johan van der Auwera

features in the history of emphatic negation. When Latin non formally weakened to ne, it remained perfect for neutral negation, but not for emphatic negation, and it is here that it accepted formal and semantic strengtheners of various types, one of them with the lexical element originally meaning ‘step’. As implied in the preceding lines, I believe that for the development of pasNEG from neNEG over neNEG … pasNEG, the alternative account is correct and that the original Jespersenian account, which has the process start off with phonetic erosion, is wrong. In the words of Kiparsky & Condoravdi (2006:  175), typical for a growing number of linguists, `The role of phonetic weakening […], however plausible it might seem, is not backed up [by] any data as far as we know.’ And interestingly, more or less clear endorsements of the alternative account even antedate the formulation by Jespersen, with Gardiner (1904: 134), and with Meillet (1912: 393–394 [1926 139–140]). Gardiner (1904) discusses Egyptian and Coptic and notes ‘the often noted likeness’ to French and then writes about pas and point: These words, from the Latin passum and punctum, were originally adverbial accusatives placed at the end of negative sentences for the purpose of emphasis; just like the English “not a jot”, “not a straw”. […] Pas and point, and like them the Demotic , Coptic ⲁⲛ, next lose their emphasizing force, and become mere adjuncts of the negative words (French ne, Coptic = ⲛ̅). Last of all, they come themselves to be looked upon as negative words. (Gardiner 1904: 134)

In the famous paper that seems to introduce the term ‘grammaticalization’ Meillet (1912: 393 [1926: 140]) discusses Latin, French, and German and writes the following: Là où l’on avait besoin d’insister sur la négation […] on a été conduit à renforcer la négation ne... par quelque autre mot. […] On sait comment pas a perdu, dans les phrases où il était un accessoire de la négation, tout sons sens propre—sens conservé parfaitement dans le mot isolé pas—, comme dès lors, pas est devenu à lui seul un mot négatif, servant à exprimer la négation […]. (Meillet 1912: 393 [1926: 140])3 3. On the next page, he writes in a way that could have seduced Dahl (1979) to honor Meillet rather than Jespersen: Les langues suivent ainsi une sorte de développement en spirale : elles ajoutent des mots accessoires pour obtenir une expression intense : ces mots s’affaiblissent, se dégradent et tombent au niveau de simples outils grammaticaux ; on ajoute de nouveaux mots ou des mots différents en vue de l’expression ; l’affaiblissement recommence et ainsi sans fin. (Meillet 1912 : 394 [1926 : 139–140])  ote that when Dahl (1979: 88) coins the phrase ‘Jespersen’s cycle’, he didn’t say that Jespersen N (1917) was the first to make the relevant observation. This has been misunderstood later (e.g. Burridge 1983: 36, Postma 2002: 44, Mazzon 2004: 5, Roberts 2007: 142). Note that I do not claim that Gardiner (1904) was the first either. For French, for instance, the idea that the weakness of neNEG plays a causal role is nicely expressed in Perle (1878: 407). So the historiography of



Chapter 3.  The Jespersen Cycles 

At this moment, both accounts find their supporters since both are after all very similar and the difference might go unnoticed or be irrelevant. There are also statements that are neutral with respect to this difference. (8) categorizes some of the literature.4 (8) Hypotheses Weakness Strengthening - Bleaching Emphasis Bleaching Neutral

Supporters Jespersen (1917), Wackernagel (1926), Dahl (1979), Pohl (1968), Horn (1989: 454–457), Lenz (1996), Beukema (1999: 10), van Kemenade (1999), Breitbarth and Haegeman (2008), Jäger (2008) Gardiner (1904); Meillet (1912 [1926]), Hock (1991), Hopper and Traugott (2003); Dahl (2001: 473); Detges & Waltereit (2002), Kiparsky and Condoravdi (2006) Marchello-Nizia (1979), Bernini and Ramat (1996); Haspelmath (1997); Schwenter (2006)

(9) summarizes the discussion. It integrates the Jespersenian formal and semantic strengthening, but only for emphatic negation, and the ‘alternative’ idea of a transfer of an emphatic strategy to the realm of neutral negation, as represented in (7).

the idea of the Jespersen cycle remains to be written, and an important part will concern scholars of Egyptian, Coptic, Berber and Arabic working independently of Jespersen (1917) and of the name giving Dahl (1979) (e.g. Gardiner 1904, Winand 1997 for Egyptian and Coptic and Chaker & Caubet eds 1996 for Arabic and Berber). 4. A full classification would also categorize accounts that neither follow Jespersen or Gardiner nor count as neutral but offer yet a different analysis. A recent account that belongs here is Breitbarth (In print), followed by Breitbarth & Haegeman (2008). Breitbarth (In print) claims that in West Germanic there was never a stage in which negation had two exponents. She argues (again only for West Germanic) that as soon as the strengthener became compulsory, it was the only negator. The reason for this is that by then the old negator had expanded its territory to non-negative uses and it was already used with negative pronouns and adverbs. I find the argument problematic, for the expansion of negators to non-negative yet still negatively polar contexts is a general property of negators (see the discussion around examples (10) and (11) below) and so is the collocation of a clausal negator with a negative pronoun or adverb (‘negative concord’) and it is at least not obvious how these general properties imply that the negativeness of the negator is damaged. With respect to the negative concord property, it is suggestive that in present-day Belgian Dutch dialects all allow negative concord: if they don’t have it with the old en, they do it with the new niet¸ and in the middle area they allow both en… niet, (thus combining with e.g. niemand ‘nobody’ and yielding en… niemand, niemand … niet, and en … niemand … niet (van der Auwera and Neuckermans 2004). This suggests that the old en is no less negative than the new niet. Furthermore, the arguably most progressive dialect has denegativized the pronoun niemand but kept the clausal negator niet, thus yielding iemand niet ‘nobody’, literally ‘somebody not’ (van der Auwera, De Cuypere & Neuckermans 2006).



Johan van der Auwera



(9) Stages

Emphatic strategies

Neutral strategies Formal weakening

1

nonNEG neNEG

2



Formal & semantic strengthening

3

neNEG … pasX

neNEG

4

neNEG … pasNEG

neNEG

5

neNEG … pasNEG

6

neX … pasNEG

7

pasNEG

Semantic reanalysis Formal strengthening Semantic weakening Semantic weakening Formal weakening

Like all other representations given in this paper (9) is only a skeleton sketch. First, the sketch focuses on the neutral strategies. It brings in just one emphatic strategy and then only to the extent that it provides a source for a neutral strategy. While (9) does claim that at stage 1 the negator nonNEG allowed both neutral and emphatic uses, the latter, I assume, requiring stress (see the discussion of example (12) below), it does not claim that there were no other emphatic strategies then, or of course at any later stage. (9) also does not sketch how the noun step, a ‘minimizer’ (i.e. a noun referring to a small quantity) developed into the emphatic pasNEG, from the ‘not even a step’ to ‘not at all’, a phase of progressive specialization for negation going through a more general phrase of negative polarity (e.g. Muller 2004). In the Jespersenian account, this step was considered to be weakening, for it was taken to involve the loss of the emphatic effect of the ‘additional word’. But in this account, the emphatic effect remains. It is correct that step loses its lexical meaning, but then it gains a pragmatic meaning. So something is lost, and something else is gained: for this situation (9) uses the term ‘semantic reanalysis’. Second, (9) does not sketch the competition that pas had from other minimizers, most prominently point ‘point’ and mie ‘crumb’ (see Möhren 1980 and Kawaguchi In print). Third, (9) is restricted to declarative negation and does not take into account the influence of sentence type, such as declarative vs. interrogative or imperative, a factor that is indeed important for French (Martineau and Mougeon 2003: 119–120, Martineau and Vinet 2005) no less than for Dutch (as alluded to in the paragraph just before (1)). Fourth, a full account of the history of neNEG, pasNEG and neNEG... pasNEG will have to provide for the fact that even when pasNEG was established



Chapter 3.  The Jespersen Cycles 

as the main exponent of negation and left the pure negative polarity stage, it still allowed negative polarity uses, as did and do neNEG and neNEG... pasNEG.(10) gives some older and recent examples. (10) a.

17th century ne (Muller 1991: 375) Je n’ignore pas qu’il n’ait voulu me nuir. I neg ignore neg that he X would.have wanted me damage ‘I don’t ignore that he would want to damage me.’

b.

17th century ne... pas (Muller 1991: 24) Vous ferez plus que vous ne estes pas proposé. you will.do more than you X are X proposed ‘You will do more than you have proposed.’

c.

18th century pas (Muller 1991: 25) C’est la plus jolie fille qu’ y a pas dans le canton. This is the more pretty girl that there has X in the canton ‘This is the prettiest girl there is in the canton.’

(11) a.

Present-day ne (Muller 1991: 367) Ils estiment que leur situation militaire est they think that their situation military is suffisamment solide pour empêcher que cet assaut sufficiently solid for prevent that this attack ne devienne une promenade militaire. X become.subj a walk military ‘They think that their military situation is solid enough to prevent that their attack would become a military walk.’

b.

Present-day French ne pas (Larriveé 2004: 28) La Grande-Bretagne déconseille à ses ressortissants the Great Britain discourages to its citizens de ne pas se rendre à Bali. of X X oneself go to Bali ‘Great Britain advises its citizens not to travel to Bali.’

c.

Present-day French pas (Larriveé 2004: 27) J’imagine que je désire plus que j’peux pas obtenir. I imagine that I desire more than I can X obtain ‘I imagine that I want more than I can obtain.’

The fact that each of the negators had and has negative polarity uses is a reflection of the fact that the negative polarity uses of negators, often called ‘expletive negation’, are a universal potentiality (see also Joly 1981, van der Wouden 1997: 196–204, van der Wurff 1999 for both European and non-European examples). Of course, this potentiality can be grammaticalized or not and, if so, in different ways (i.e. in different negative



Johan van der Auwera

polarity contexts) and with different strengths. The link with the Jespersen cycle seems double. First, chances that a negator has negatively polar uses are high if it has in fact developed from a negative polarity item, as is the case with pas. From this perspective, the older uses of pas and ne... pas, illustrated in (10b-c) can be seen as relics of the negative polarity past of pas (this is also the hypothesis offered in Muller 2004 and in Eckardt 2003, 2006:  128–170). Second, a point already made by Jespersen (1917), chances that a negator has negative polarity uses are also high ‘if the negative employed no longer has its full force’ (Jespersen 1917: 75). What he had in mind were the modern uses of ne as in (11a) (see also Breitbarth & Haegeman 2008 for Belgian Dutch and Wallage 2008 for Middle English).5 Another point worth observing is that (9) implicitly takes solitary neNEG to disappear because of the introduction of the newcomer neNEG... pasNEG. This makes sense. The renewal of grammatical markers, with originally expressive markers becoming neutral and ousting the original neutral markers, is the essence of grammaticalization, with lots of examples in classic texts on grammaticalization. Of course, the fact that other domains of the grammar have new constructions oust old constructions does not predetermine the outcome of the competition between the old neNEG and the new neNEG … pasNEG. For one thing, lots of old constructions do successfully hold competitors or potential competitors at bay and, even in the domain of French negation, newcomers like neNEG... pointNEG and neNEG... mieNEG did not make it. And for another thing, neNEG... pasNEG is less economical than neNEG, and so for reasons of economy, one would actually expect neNEG to hold ground. So why did neNEG... pasNEG gain supremacy over neNEG? I suppose that there is no single answer. Part of the explanation will have to do with dialect and register variation and the competition between dialects and registers, a competition which did not merely involve neNEG and neNEG … pasNEG, but also neNEG … pointNEG and neNEG … mieNEG and yet other constructions. I suspect that the development of concord structures such as neNEG … personne ‘not nobody’ promoted ne … pas as well, in a mutual support constellation. Let me offer two additional considerations. First, a complex construction such as neNEG... pasNEG holds one advantage over neNEG. Even though neNEG... pasNEG is no longer compositionally dedicated to emphasis (with pasNEG having lost its emphatic effect), it does have a component that can be stressed, viz. pasNEG, and thus serve emphatic uses anyway, in contrast with solitary neNEG, which cannot be stressed, at least not since the end of Middle French (Martineau & Mougeon 2003: 124). There is a similar contrast in English: free standing not can be stressed, but cliticized n’t cannot. 5. The negative that no longer has its full force is always an old negative. I venture the hypothesis that the likelihood of negative polarity is also partially a function of age: the older the marker the longer it had a chance of developing negative polarity uses. Thus, all (?) the old Romance and Slavic markers corresponding to French ne also have negative polarity uses, whether or not the languages have set on a Jespersen trajectory.



Chapter 3.  The Jespersen Cycles 

(12) a. Je ne... veux pas! I neg want neg ‘I do not want it!’ b. *Je ne veux! I neg want *‘I doN’T want it!’ Now economy is in favor of neNEG... pasNEG. With neNEG... pasNEG, speakers have a construction that allows both a neutral use (without emphatic stress) and an emphatic use (with emphatic stress on pasNEG). For this double use, solitary neNEG is indeed too weak, so here Jespersen (1917) is right after all. The second consideration that helps explain why neNEG … pasNEG replaced neNEG concerns the timing when pasNEG in neNEG … pasNEG was beginning to ‘be felt as the negative proper’. In (9), this is represented as happening after the demise of neNEG. But this is probably a simplification, due to the decision taken in 1.1 to avoid to modeling competing variants. In fact, when neNEG … pasNEG becomes a neutral construction, it alternates with neNEG for some time and one may assume that the reinterpretation of the neNEG component of neNEG … pasNEG as an exponent of negation to the status of ‘additional word’ took place before the disappearance of solitary neNEG, and that this reinterpretation in fact proved a(n additional) factor in the demise of solitary neNEG. I represent this alternative scenario in (13). Stages 4 and 5 correspond to those of (9). (13) Stages Emphatic strategies ... ... 4 neNEG... pasNEG 5 ...

Neutral strategies ... neNEG neNEG … pasNEG & neNEG neX … pasNEG & neNEG neX … pasNEG & neX neX … pasNEG ...

When neNEG bleaches to a mere ‘additional word’, one may assume that this happened first in the doubling construction, for it is only here that there was another exponent of negation, viz. pasNEG. The ongoing bleaching from neNEG into neX could have infected solitary neNEG as well, and thus added to its downfall.6 6. Note that this period of the ‘cohabitation’ of neNEG ... pasNEG and neNEG was essentially all of Old and Middle French and a bit beyond (9th to 17th century). There has been quite a bit of work, most recently Hansen (In print), Hansen and Visconti (In print), largely inspired by work of Schwenter (2006 on Catalan, Italian and Brazilian Portuguese) aiming to show whether for some of this time the two strategies were different in terms of discourse status. Simplifying these accounts, one could say that the neNEG ... pasNEG strategy was more ‘presuppositional’ (or ‘discourse-old’). I find this rather plausible given that typical ‘not even a step’ and ‘not at all’ uses are also more presuppositional than the simple ‘not’. Bleaching then would first take away emphasis



Johan van der Auwera

A final point concerns the fact that in all scenarios so far the deemphasizing of neNEG... pasNEG precedes the demise of neNEG as a part of neNEG... pasNEG, i.e. the formal weakening of neNEG... pasNEG to pasNEG only happens to neutral neNEG... pasNEG, and not to emphatic neNEG... pasNEG. This is indeed the case in French and in fact in all other cases I know of, yet it might not be a universal. The facts of Horn’s (2001) ‘squatitive’ negation (see also Postal 2004:  159–172, Hoeksema 2009) are suggestive. The term ‘squatitive’ refers to minimizers that employ taboo expressions, like English squat, shit, or fuck. In a negative context, they can mean ‘nothing at all’. Here is an attested example. (14) He … discovered writing didn’t pay squat.

(Horn 2001: 186)

This is emphatic, just like ‘not even a step’ has been emphatic, but different from the French ‘step’ notNEG squat didn’t neutralize, and yet it did simplify. (15) is an attested solitary squat in which squat by itself means ‘nothing at all’. (15) Students complain that they are learning squat.

(Horn 2001: 187)

Of course, these squat uses are still nominal or, better pronominal, for in (15) squat has become a negative pronoun meaning ‘nothing whatsoever’, but we know from French pas and English not that minimizing and negative indefinites can turn into adverbs. If that happened for squat, we would have a clausal negator that did not neutralize at the doubling stage. It would arrive at the new solitary stage as an emphatic negator, possibly, of course, neutralizing still later. Another illustration would be a variety of French (probably an imaginary one) that keeps both neNEG … pointNEG and neNEG … pasNEG, with neNEG … pointNEG as the emphatic variant and neNEG … pasNEG as the neutral one, and in which neNEG … pointNEG is turning into pointNEG just like neNEG … pasNEG is turning into pasNEG. Interestingly, no good example of an emphatic solitary clausal negator that derives from a minimizer has yet been found. However, in principle, it could come about this way, not least also because we see that a negative indefinite can originate this way (as in the squat example). (16) sketches the imaginary clausal development from an emphatic notNEG…squatX to emphatic notNEG…squatNEG to emphatic squatNEG to neutral squatNEG as well as the attested development from emphatic neNEG…pasX to neNEG…pasNEG to pasNEG. The attested developments are symbolized with a full line, and the imaginary one with a dotted one.

but leave the higher presuppositionality. Note though that the Schwenter approach has been criticized for French (Larrivée 2008, Schøsler 2008).



Chapter 3.  The Jespersen Cycles 

(16)

a.

neg1 X

neg1 neg2 –emphatic neNEG … pasNEG

+emphatic neNEG … pasX

+emphatic neNEG … pasNEG –emphatic squatNEG

b.



1.3

Ø neg2 –emphatic pasNEG

+emphatic neNEG … squatX

+emphatic notNEG … squatNEG

+emphatic squatNEG

Clause-final repeated negators

In the preceding two sections, I rejected Jespersen’s idea (1917) that it was the weakness of neutral neNEG that called for neNEG … pasNEG. I did not, however, reject this idea completely. First, I granted that neNEG was indeed too weak for emphatic negation. Second, as a marker allowing stress to mark both neutral negation (no stress) and emphatic negation (stress) it was also too weak. Let me now come to a third way in which Jespersen’s (1917) point about weakness has some value, and again, not in the way that he imagined. The matter concerns the clause-final doubling of the clausal negator (symbolized as ‘NEG#’). This kind of doubling is perhaps best known from Brazilian Portuguese (e.g. Schwegler 1991, Schwenter 2006) and Afrikaans (e.g. Roberge 2000, Biberauer In print, Biberauer 2008, this volume) but it occurs more widely in both Romance (e.g. Schwegler 1990, 1996 for Spanish and Ramat 2006 and Floricic and Molinu 2008 for Italian) and Germanic (e.g. Pauwels 1958 for Dutch, and for Swedish, we can again go back to Jespersen 1917: 72). (17) shows attestations from Belgian Brabantic dialects of Dutch (Neuckermans 2008: 210–226; Barbiers et al. 2009). In (17a) the negator nietNEG is repeated, in (17b) the repeated negator is flanked by the additive particle ook ‘also’. (17c) and (17d) shows that the clause-final negator is also allowed when the earlier part of the sentence contains a quantificational negator, whether or not the latter uses nietNEG (niemand niet ‘nobody not’ vs. niets ‘nothing’).



Johan van der Auwera

(17) Belgian Brabantic (Neuckermans 2008: 215, 213, Pauwels 1958: 443, 440)7 a. Hij wil geen soep niet meer eten niet. he wants no soup neg more eat neg ‘He doesn’t want to eat any more soup.’ b. Els wilt niet dansen en ze wil niet zingen ook niet. Els wants not dance and she wants neg sing also neg ‘Els doesn’t want to dance and she doesn’t want to sing either.’ c. Ik heb niemand niet gezien niet. I have nobody neg seen neg ‘I haven’t seen anyone.’ d. Ik heb niets gekregen niet. I have nothing received neg ‘I haven’t received anything.’ In these constructions, the addition of the clause-final negator does not have any emphatic effect and it cannot be emphasized (Pauwels 1958: 444–445), even though its formally identical ‘clause-earlier’ negator can. At most, it helps to secure the correct uptake, a reminder to the hearer that the proposition is negative, but this functionality is not necessary and its use may be fully bleached. Then, the clause-final negator is simply a partial exponent of neutral negation, a part of nietNEG … niet#NEG, just like pasNEG as a part of neNEG pasNEG. niet#NEG has no additional semantic effect either: the sentence is negative already. The lack of any special meaning is corroborated by the fact that speakers are often not aware that they are using it, a fact specifically mentioned both by Pauwels (1958: 445), the linguist responsible for the classic description of niet#, and by Blancquaert (1923: 65), the linguist who first wrote about it. So much for the synchrony of Belgian Brabantic clause-final negation! Let us now relate it to the Jespersen cycle. Can the appearance of the repetitive clause-final negation be explained with the Jespersen cycle hypothesis, in the variant that posits emphasis and bleaching? In that case, the clause-final negation must have had an emphatic effect before, even if it does not have one now. Cross-linguistically, this indeed seems the most common analysis. For Brazilian Portuguese (but also Jespersen 1917 and Bernini & Ramat 1996: 43 for other languages), Schwegler (1991: 209; 1990: 169–173), for instance, derives (18b) from (18a), a structure with a reinforcing intonationally separate pragmatic particle, which is crucially identical in form to the clausal negator. (18) Brazilian Portuguese a. Eu não quero, não! I neg want no ‘I don’t want to, absolutely not!’

(Schwegler 1991: 209)

7. Here and elsewhere, no attempt is made to represent to faithfully represent the sounds or forms of the Dutch dialect facts.



Chapter 3.  The Jespersen Cycles

b. Eu não quero não. I neg want neg ‘I don’t want to!’ This analysis exists for Dutch too, most clearly presented by Roberge’s (2000: 146–147), supported by Biberauer (2008), with structures such as (19) as the source of clause-final negator doubling. (19) Ik ga niet, nee! I go neg no ‘I am not going, no!’ The context in which this claim figures actually concerns Afrikaans clause-final negation, which Roberge traces back to Dutch. In the Dutch case, the analysis is problematic in that the pragmatic particle in (19) is nee and not niet. Bernini & Ramat (1996: 78) therefore reject the pragmatic hypothesis, at least for Dutch and Afrikaans (but interestingly, not for Berbice Creole Dutch, in which the clause-final negator contains the morpheme -nε, reconstructed from the Dutch pragmatic particle neen (Kouwenberg 1994: 264)). Roberge (2000: 147), in turn, rejects Bernini & Ramat’s rejection pointing out that early Modern Dutch did at least allow niet as an answer particle, as well, next to neen. However, this use of niet was very rare and that of neen very common (Joop van der Horst, p.c.). Hence, it is not plausible to assume that it is the rare use that grammaticalized the way Brazilian Portuguese não and markers in other languages did. So I, in turn, reject Roberge’s rejection. In fact, there are two hypotheses that trace Belgian Brabantic clause-final negator to a non-emphatic origin, both due to Pauwels, viz., Pauwels (1958) and Pauwels (1974). Actually in both publications Pauwels is sympathetic to the pragmatic hypothesis as well, but he offers two different hypotheses in addition. First, Pauwels (1958: 457) thinks that clause-final doubling could derive from a negative concord construction in which a negative quantifier is followed by nietNEG. This construction is illustrated in (20). Like clause-final negation, it is typically Belgian Brabantic. This construction is not emphatic and nobody has hypothesized that it was. (20) Belgian Brabantic a. Ik heb niemand niet gezien. I have nobody neg seen ‘I haven’t seen anybody.’ b. Ik zag niemand niet. I saw nobody neg ‘I didn’t see anybody.’ The idea is that the nietNEG of the negative concord construction may come at the clausal end, as illustrated in (20b), and that it became associated with that position. Second, in his last public pronouncement on this matter, Pauwels (1974: 76) supports





Johan van der Auwera

the idea that clause-final negation is essentially just a repetition strategy, devoid of emphasis and devoid of additional semantics. He compares it with preposition doubling as in (21). (21)

Belgian Brabantic Ik kan nie aan het plafond aan. I can neg at the ceiling at ‘I can’t reach the ceiling.’

(Pauwels 1974: 76)

I suppose that this kind of repetition at best makes the meaning clearer, a locational one in (21), and a negative one in (17). To the extent that repetition can be said to ‘strengthen’ a meaning, the non-repeated meaning can be said to be weaker than the doubly expressed meaning, but whether expressed with just one negator or with two, the meaning is just a neutral, non-emphatic negation. This then brings us back to the Jespersenian weakness idea: the simple expression is not exactly ‘too weak’, but it is a least ‘weaker’ (i.e. weaker than the doubling expression). Belgian Brabantic clause-final doubling has not been simplified into a construction with a solitary clause-final negator. This is different from Brazilian Portuguese não#NEG, though it is not clear whether the solitary Brazilian Portuguese não#NEG is emphatic or not. Schwegler (1991: 206) thinks that this construction ‘has by and large lost its emphatic character’. Biberauer (2008), however, compares it with English (22), which would seem to imply that the construction is emphatic. (22) She came to the party … not! So perhaps we should keep both hypotheses open, just like we did with the analyses of clause-final doubling. (23) summarizes the facts and hypotheses about clause-final negator doubling for Dutch and Brazilian Portuguese. Full lines represent the scenarios proposed in the literature and retained as plausible: (i) the Dutch doubling is not emphatic and it never was, and it didn’t develop a new solitary stage; (ii) Brazilian Portuguese doubling is emphatic and the solitary neg1# could have been emphatic or not. Dotted lines show hypotheses that I add myself: if solitary Brazilian Portuguese neg1# is non-emphatic, it must have bleached, either at the doubling stage or only at the final, solitary stage.



Chapter 3.  The Jespersen Cycles 

(23)

neg1 a.

nietNEG

b.

neg1 neg1# –emphatic

neg1#

nietNEG … niet#NEG –emphatic nãoNEG … não#NEG

–emphatic não#NEG

nãoNEG +emphatic

1.4

nãoNEG … não#NEG

+emphatic não#NEG

Intermediate conclusion

I hope to have shown that the scholarship on classical cases of the Jespersen cycle such as the progression from French neNEG to pasNEG via neNEG … pasNEG has been interpreted in two different ways, both involving a notion of strengthening. In the approach associated with Jespersen (1917), a neutral negator is claimed to be too weak and to be ‘strengthened’ as a result. In the alternative approach, the neutral negator gets competition from a construction that is originally emphatic—‘strengthened’—and undergoes bleaching in a natural process of grammaticalization. I have argued that the second approach is correct for the analysis of neutral negation. However, Jespersen was still right and even in three ways. First, for emphatic negation neNEG was indeed too weak. Second, neNEG was also too weak as a multifunctional negator, serving both neutral and emphatic negation.8 Third, although a neg … neg# construction may have its origin in emphasis, it may also have a non-emphatic origin, and in that case one can compare, as strategies for expressing neutral negation, the single neg and the double neg … neg# construction, and then the former is weaker, in a trivial sense, than the latter. The general picture that emerges is represented in (24). First, one can get from a single negator to a doubling pattern by either reanalyzing an element X, a minimizer or a negative quantifier (but also a negative existential; Croft 1991: 10, 13–14), and this construction is necessarily emphatic, or can simply repeat the negator at the end of the clause, a process which may be emphatic or not. Second, doubling constructions may turn into solitary constructions, most clearly when they are non-emphatic, but the squat and não evidence suggests that emphatic negators may also turn single. Third, emphatic constructions may bleach, most clearly at the doubling stage, but on account of the não evidence, one can also hypothesize bleaching at the solitary stage. For ease of understanding, in (24), neg1# will be represented as ‘neg2’, and I do not distinguish 8. This claim follows from the first one, but it is worthwhile separating both claims, for the second one, and not the first one, was argued to help explain the demise of the neutral neNEG use.



Johan van der Auwera

between full and dotted lines anymore—at this level of abstraction all lines except the one from neg2 +emphatic to neg2 –emphatic have in fact been proposed, either in a neNEG … pasNEG type scenario or in a nietNEG … niet#NEG type scenario, or in both. Like in all other schemas, negative polarity uses such as illustrated in (10) and (11) are not taken into account, essentially because they can occur at all stages of the life cycle of a negator. (24)

neg1 neg2

neg2

–emphatic

–emphatic

neg1 X

neg1 neg2

neg2

+emphatic

+emphatic

+emphatic

neg1



2. Any more variation? It is implied in (24) that the X source is always emphatic. But is that really the case? And relatedly, does this element have to be a minimizer or a negative quantifier (or, if Croft 1991: 10, 13–14 is right, a negative existential)? A negative answer, an implicit one because it did not relate the issue to Jespersen cycles, was given for Arizona Tewa (Kiowa-Tanoan) by Kroskrity (1984), supported by Honda (1996: 41–44), who provides additional data (Honda 1996: 207) and does relate them to the Jespersen cycle. In 2007, two further negative answers appeared, one by Lucas (2007) and the other by De Cuypere et al. (2007: 309–312, repeated in De Cuypere 2008: 238–245). I will restrict the discussion to the latter two answers. Lucas’ (2007) answer is this: The evidence from Spoken Arabic suggests that the only condition for JC [Jespersen’s Cycle] to be set in motion […] is that some postverbal element could potentially be reanalyzed as a marker of negation. (Lucas 2007: 427)

The evidence concerns the spoken Arabic form -š, which is argued to have had negative contexts in which its original indefinite pronoun meaning ‘anything’ (itself deriving from the noun ‘thing’) could have been interpreted as an exponent of negation. The case is far from clear, I find, for after all ‘anything’ comes very close to a minimizer, and the original noun ‘thing’ is arguably even closer to e.g. ‘step’, and would thus invite an account in terms of emphasis and bleaching, an invitation taken up by Bernini & Ramat (1996: 47), Roberts and Roussou (2003: 155), François (2003: 317), and Miestamo (2005: 225) for Arabic, and by Willis (2005) for Welsh. Lucas (2007: 417) is aware of this kind of approach, yet does not find ‘this explanation entirely convincing’ (Lucas



Chapter 3.  The Jespersen Cycles 

2007: 420). His phrasing still means, however, that he finds it fairly convincing anyway. I conclude that the case is not settled yet. The other negative answer is due to De Cuypere et al. (2007: 309–312 (also De Cuypere 2008:  223–249).9 Their starting point is Miestamo’s (2005) work on asymmetric negation. Miestamo notes that positive and negative declarative sentences often differ from another in more than just the absence vs. the presence of a negative marker. Consider negation in the Salishan language Bella Coola. Third person singular positive declarative verb forms mark the information as either old or new. However, in the corresponding negative declaratives there is no choice: the verb form is obligatorily marked for old information. (25) Bella Coola (Miestamo 2005: 136, based on Nater 1984: 36; also in De Cuypere et al. 2007: 312 and De Cuypere 2008: 242) a. ksnmak-Ø work-3sg.new ‘He is working.’ b. ksnmak-s work-3sg.old ‘He is working.’ c. ?axw ksnmak-Ø neg work-3sg.old ‘He is not working.’ This asymmetry has a functional motivation: in negatives one tends to express a negative evaluation of something that is presupposed (given, old) in the context (in a way that needs to be more precise—see note 6), and what Bella Coola has done is to grammaticalize this tendency (Miestamo 2005: 213). The relevance of this asymmetry is the following: imagine that the marker for old information disappears in the positive construction and that it remains in the negative one. This would easily allow this marker to be reanalyzed as an exponent of negation, and would thus offer a good case for the reanalysis of a marker that is not a negative quantifier or minimizer (nor of course a repeated clause negator) and mostly importantly, not emphatic. The Bella Coola case is an imaginary case: the old information marker has not left the positive declarative. It is problematic that De Cuypere et al. (2007) do not offer any convincing case (and neither does De Cuypere 2008: 223–249). Let me discuss one of their examples. In some Vanuatu languages, a partitive marker, which at least optionally appears in negative declaratives, in which negation already has one exponent, can be interpreted as an additional exponent of negation, and one which may oust the first one. This has been argued for Lewo (Early 1994a) and for Motlav (François 2003: 317–318). 9. A terminological difference between Lucas (2007) and De Cuypere et al (2007) and De Cuypere (2008) is that the latter do not consider the scenario described as a ‘Jespersen cycle’ but rather as an alternative.



Johan van der Auwera

(26) Motlav a. Et igni-k te. neg wife-my neg ‘This is not my wife.’

(François 2003: 313, 318)

b. Ino te, ikē! I neg he ‘It is not me, it is him!’ For both languages, the grammarians remark on the similarity with the classical Jespersen cases of the French neNEG... pasNEG type. The key question is whether the partitive once had an emphatic function or whether it was just a concomitant of a negative sentence, a marker of asymmetry, which again has a functional explanation (in this case relating to a tendency of negatives having non-referential objects—see Miestamo 2005: 232). The answer may be different from one language to the next, but at least for some Vanuatu languages it seems that the partitive was in fact used for emphasis, just like a minimizer. For Motlav, François (2003: 317) is fairly explicit: the combination of the negator with the partitive must have served to deny the event that didn’t even concern a (small) part of something. Paamese is interesting too. In this language, the element corresponding to te is not obligatory yet, and maybe for this reason Crowley (1982: 140–142) still considers it to have partitive meaning, but his gloss is more than just partitive. The combination of the negation and the partitive is not just to mean that something is partially unaffected, but rather that it is not even partially affected, i.e. not even a bit, which can surely be called emphatic in the same way as ne … pas ‘not even a step’ is.10 So at least in some Vanuatu languages, the partitives would rather plausibly have lent their service in the emphasis-cum-bleaching cycle. Interestingly, in Avava, yet another Vanuatu language, the second element is not related to a partitive marker, but to a marker meaning ‘first’ (Crowley 2006: 84–85, 99). Crowley (2006: 99) is uncertain as to how to relate to two meanings, but an emphasis-cum-bleaching cycle would again seem possible, with the original meaning of the now bipartite negative being that some event did not even affect the first part of something. Consider English know the first thing about in this respect (J. Hoeksema, p.c.). (27) He didn’t even know the first thing about checkers. Yet our search for a Jespersen cycle that concerns neither emphasis with bleaching nor non-emphatic repetition is not in vain. The Vanuatu language Lewo may have an emphasis-driven partitive element re, but the element that is of interest here is the first marker, which is either pe (for realis) or ve (for irrealis). 10. When something is not even partially unaffected, it means that it is completely unaffected. So one might imagine a completeness marker to slide in the pragmatic Jespersen path and De Cuypere et al. (2007: 310) actually mention that the second of two negation exponents of the Barbacoan language Awa Pit is diachronically linked to a completive aspect marker (Curnow 1997: 332).



Chapter 3.  The Jespersen Cycles 

(28)

Lewo Pe ne-pisu-li re Santo poli. neg.r 1sg-see-try neg Santo neg11 ‘I’ve never seen Santo.’

(Early 1994a: 69)

This pe/ve marker derives from the positive copula, which also comes in both a realis and an irrealis form. One might venture the hypothesis (see Dahl 1995: 80; Miestamo 2005: 79) that pe/ve is in fact the copula12, and indeed, as Miestamo (2005: esp. 75–80) makes clear, copulas are frequently used as the finite verbs in negative clauses. Yet, though the copula will indeed be the highly plausible origin of the pe/ve marker, at present the pe/ve marker is not a copula anymore. Different from a real copula, illustrated in (29), the negatively dedicated pe/ve markers do not carry subject markers. In fact, prefixal subject marking is attached to the lexical verbs. (29) furthermore shows that the negative unconjugated pe and the conjugated pe copula can occur together. (29)

Lewo Teras la na˜pa pe a-pe praktis re poli … youth pl rel neg 3pl-be practise neg neg ‘The young guys who haven’t practised …’

(Early 1994a: 72)

And interestingly, there is one construction in which the pe/ve marker does carry subject prefixes, viz., in second person imperatives, and here the lexical verbs lack prefixes. This use is restricted to older speakers, so from a grammaticalization point of view, this construction could be considered a relic, with the original copula construction. (30)

Lewo O-pe tag re! 2sg-cop cry neg ‘Don’t cry!’

(Early 1994b: 422)

The upshot of the discussion is that an element involved in the asymmetry that characterizes negation in the sense of Miestamo (2005), such as a copula, can indeed 11. Note that there is even a third neg marker poli. I will discuss tripling in 3.3. 12. The interpretation by Dahl (1995) and Miestamo (2005) is supported by some of Early’s phrasing. Early (1994a: 66) says that the pe/ve marker is ‘identical in form to that of the copula’, but it is clear from the rest of the analysis that it is really only the stem that is identical. Early (1994a: 80) also asked speakers whether they felt the negative markers and the copula to be the same and the answer ‘was somewhat confirmatory, and it certainly was not the same as some of the other homophones of ve (verbs meaning ‘to weave’ and a commonly reduced from of one of the verbs ‘to say.’ This somewhat confirmatory stand can easily be interpreted as meaning that speakers still felt the negative markers and the copulas to be related. Or as Robert Early (p.c.) puts it: ‘while speakers are readily aware of the phonological identity of the two forms, they generally understand the negative pe/ve as meaning ‘no’ and are somewhat bemused, and certainly don’t show any clear agreement, when asked if they think that it is the “same” as the copula pe/ve.’



Johan van der Auwera

provide a non-emphatic source for a second negator. It is time, therefore, to update the figure in (24). All we have to do is to add one source construction (marked in (31) with shading): next to the emphasis neutral neg1 source and the +emphatic neg1 X source, there is a –emphatic neg1 X source, which can feed directly into an equally non-emphatic neg1 neg2 construction. (31)

neg X

neg1 neg2

neg2

–emphatic

–emphatic

–emphatic

neg1 X

neg1 neg2

neg2

+emphatic

+emphatic

+emphatic

neg1



If one wants to count cycles, one will find that there are eight possible cycles, i.e. eight different trajectories to get from one of three possible single negator source constructions to either of the two possible single negator outcome constructions. The basic idea remains relatively simple though: either through the repetition of a clause negator or through the reanalysis of something else, one arrives at a bipartite negative construction, which is either emphatic or neutral. In the bipartite construction, the new negator can oust the old one, and at both the bipartite or the new simple stage the emphasis can disappear.

3. Even more variation? The schema in (31) is also useful as a starting point. First, in (31) the doubling strategy cannot stay emphatic: it either neutralizes at the doubling stage (as with neNEG... pasNEG) or it loses one exponent of negation, and then it may stay emphatic, but it is not of course a doubling construction any more. So the question is whether a doubling strategy cannot in some way become emphatic (again). I will discuss this in 3.1. Secondly, in the schema in (31) the two solitary stages are interestingly different: the first solitary stage has a negator (neg1) and either some marker X or nothing, whereas the second solitary stage has a negator (neg2) or nothing, i.e. all the first negator seems to be able to do is to disappear. I will discuss this in 3.2. Thirdly, in (31) the reinterpretation of a marker X into a negator (neg2) or the mere addition of this negator is only allowed after a stage of a solitary negator. But would it be ruled out that a doubling construction can also receive a new exponent of negation, either through reinterpretation or addition, thus resulting in a tripling construction? This will be discussed in 3.3. Fourth, when neg1 X is followed by neg1 neg2 there are two bipartite stages, the difference is, of course, that the second part of the first stage is not negative yet. One



Chapter 3.  The Jespersen Cycles 

wonders whether the second bipartite stage is absolutely essential, i.e. whether a language cannot also go directly from neg1 X to Ø neg2. I will discuss this in 3.4.

3.1

A new emphatic strategy at the doubling stage

The first question that I asked in the introduction to this section is whether a doubling construction is really fated to weakening, either semantically (reinterpretation of emphatic negation as neutral negation) or formally (with a double negator turning into a solitary one). That the situation is more complex has been claimed by Haegeman (2002: 181) and again by Breitbarth and Haegeman (2008) relative to contemporary West and East Flemish dialects. The hypothesis is this: at the stage where enX … nietNEG is in competition with nietNEG, speakers could find the enX … nietNEG strategy useful for emphasis. Haegeman (2002: 181) provides a minimal pair: in (32) the conditional without enX is claimed to be an open conditional, but with enX it is emphatic and presuppositional: the speaker expects rain.13 (32) West Flemish (Haegeman 2002: 181) a. Oat nie regent, moe-j de blommen woater geven. if.it neg rains must-you the flowers water give ‘If it doesn’t rain, you must water the flowers.’ b. Oat nie en regent, moe-j de blommen woater geven. if.it neg1 X rains must-you the flowers water give ‘If it does not rain, you must water the flowers.’ Haegeman (and Breitbarth) do not explain why the contrast is the way it is, maybe in part because they consider it obvious. To me, it would indeed seem obvious: if the contrast sketched in (23) is real, then the motivation will be iconic: stronger, heavier, marked form is for the stronger meaning (see Horn 1991: 86 for many examples). (33) schematizes the possible reanalysis. (33) Stages

Emphatic strategies

enX … nietNEG & nietNEG

N



n+1

Neutral strategies

enX … nietNEG

Semantic strengthening

enX … nietNEG & nietNEG

It is important that this emphatic use of Flemish enX … nietNEG has not become the only use, and, to be judged from the large dialectological literature (see Neuckermans 2008 for an overview), which does not mention any emphatic use of enX … nietNEG, the neutral enX … nietNEG use must be considered the more important use. Also, if enX … 13. Note that in subordinate clauses nietNEG precedes enX. It will be remembered from 1.1 that word order matters are not dealt with in this paper.

 Johan van der Auwera

nietNEG indeed allows a secondary emphatic use now, one would expect this to have been possible earlier as well, and thus there might even be a continuity of emphatic enX … nietNEG uses (weakly suggested for subordinate clauses in the early 20th century South-Holland dialect of Katwijk aan Zee by Overdiep 1933a: 22—repeated in Overdiep 1937:  453, 1940:  203). If there was continuity, note that the nature of the emphatic use will have changed. Emphatic en … niet started out as combination of the neutral negator en and the ‘additional word’ niet, but now emphatic enX … nietNEG does not contrast with solitary enNEG but with solitary nietNEG and thus the ‘additional word’ is enX now. And in between both en and niet were exponents of negation. (34) Emphatic strategies enNEG... nietX → enNEG... nietNEG →

3.2

enX... nietNEG

Denegativizing

In the Jespersen cycles, the new solitary stage does not have the old neg1 anymore. It is legitimate to say that it actually already disappears as the doubling stage whenever it is no longer a true exponent of negation but only an ‘additional word’ X. We have already mentioned that when the old neg1 marker has reached this X-stage, it is particularly liable to allow negative polarity uses. In this connection, I will discuss two questions. First, is it possible for an old neg1 marker to have negative polarity uses and to have lost the negative proper use? For example, would it be possible for a future variety of French to allow (11a), repeated as (35a), yet not a simple negative declarative like (35b)? (35) French a. Ils estiment que leur situation militaire est they think that their situation military is suffisamment solide pour empêcher que cet assaut sufficiently solid for prevent that this attack ne devienne une promenade militaire. X become.subj a walk military ‘They think that their military situation is solid enough to prevent that their attack would become a military walk.’ b. Cet assaut ne devient pas une promenade militaire. This attack neg becomes neg a walk military ‘This attack does not become a military walk.’ This scenario has not materialized for French, obviously, and it has not happened for Dutch either. Of particular interest here are dialect data for Belgian Dutch. Belgian Dutch has the enNEG … nietNEG counterpart for neNEG …pasNEG, and like French neNEG, Belgian Dutch enNEG allows negative polarity uses. It appears from Neuckermans (2008: 181) that whenever a dialect allows en with negative polarity uses, it also allows



Chapter 3.  The Jespersen Cycles 

the strictly negative uses, but not vice versa.14 For Middle Low German and Middle English, however, the literature does contain the claim that an old negator may lose its negative uses and become entrenched in a negative polarity use. First, the Middle Low German case concerns ‘exceptive clauses’, i.e. ‘unless clauses’, exemplified in (36) in which en is hypothesized to have survived longer than in simple negative clauses (Breitbarth In print). (36)

Middle Low German (Breitbarth In print) Vnde dar moste nummentyn, he ne gheue V mark vp dat minste and there must nobody.in he neg would.give five mark up that least ‘and nobody must be admitted unless he gives at least five marks.’

The problem is though, in my view, that the ne marker still carries negative force, as we see when we paraphrase (36) as follows ‘if nobody is admitted, then any potential person trying to be admitted did not give at least 5 marks’. Of course, the point remains interesting, because the exceptive clause is a very special type of negative clause. Second, the claim about Middle English is found in Wallage (2008: 666–668), with Breitbarth (In print) in support. It says that when in (Late) Middle English neNEG did not exist anymore, it still had a negative polarity use in complements of negated doubt. (37)

Middle English (Wallage 2008: 666) No man douteth that he ne is strong in whom he seeth strengthe no man doubts that he X is strong in whom he sees strength ‘No one doubts that that person is strong in whom he sees strength.’

I now come to a second question: can a neg1 marker develop uses that are not even negatively polar anymore? Here the answer is positive, and for Dutch it has been claimed twice. The first case concerns the South Holland Katwijk dialect, again, and its grammarian Overdiep (Overdiep 1933a:23, Overdiep 1933b:45, Overdiep 1937: 455, Overdiep 1940). (38) is a case in point.

14. Breitbarth & Haegeman (2008) argue that the fact that en allows negative polarity uses helped maintain en in negative en … niet contexts. Given the dependency of negative polarity uses on negative uses, one would rather say the opposite. As already adumbrated by Jespersen (1917): though negative polarity uses of negative markers are always possible, they are particularly likely when a language is at the neNEG … pasNEG stage. A further critical remark on the claim that negatively polar en helped maintain en … niet concerns that the fact negatively polar en was common in Middle Dutch in areas where it no longer is now (i.e. whether it is either gone completely or where its use is very restricted) (Neuckermans 2008: 316–318; cp. also Postma 2002, Van der Horst 2008: 519, 753). So for these areas the negatively polar uses did not in fact save en … niet. This remark is the diachronic counterpart to the synchronic one that there are currently no dialects with negative polar en without negative en … niet.



Johan van der Auwera

(38)

Katwijk Toen ze bij de poort en kwamme ... when they at the gate X came ‘When they arrived at the gate …’

This en was restricted to subordinate clauses, it was optional, but under certain prosodic conditions15 quite common, and because of the relevance of these prosodic conditions, Overdiep calls this use of en its ‘rhythmical’ use. A similar case has also been documented for present-day Belgian Brabantic dialects. (39)

Belgian Brabantic (Neuckermans 2008: 176) Ze pakte eu portefueille waar dase eu sleutel in en doet. She took her wallet where that.she her key in en does ‘She took the wallet of hers in which she puts her key.’

Here too, the use is optional and restricted to subordinate clauses. The restriction of this ‘meaningless’ en to subordinate clauses makes sense. It has long been known that en survives best in subordinate clauses. Perhaps Overdiep was the first to make this observation and it has been confirmed many times including also for the current dialects (Barbiers et al. 2009; see also for references). What we see in (39) can be considered as a final stage in the demise of the en marker: en still optionally occurs in its last stronghold, the subordinate clause, but it has totally lost its negative and its negative polarity meaning. At most, it can still be considered a marker of embeddedness, and an optional and rare one, for that matter. The fact that it currently only survives in Belgian Brabantic makes sense, too, for here enNEG only survives in subordinate clauses. So we don’t have the subordinateness marker in dialects which have completely lost the negative use.

3.3

Tripling

The third question addressed at the beginning of Section 3 is whether Jespersenian accumulation can only set in at a stage with a solitary negator. If it can also set in at the doubling stage, then we will effectively have tripling. The question was already asked by Blancquaert (1923:  68): he specifically asked whether Dutch niet#NEG could be combined with en NEG … niet NEG. It was a real question, in the sense that Blancquaert did not himself know the answer. We have to wait until Pauwels (1958: 454) for an answer, and it is positive. He claims that in the Belgian Brabantic dialect of Aarschot of the first part of the twentieth century, niet#NEG could be added to subordinate enNEG … nietNEG.

15. With the trisyllabic alternative past tense kwaeme, for instance, the en particle is impossible (Overdiep 1933b: 45).



Chapter 3.  The Jespersen Cycles 

(40) Pas op dat ge niet en valt nie. fit on that you neg neg fall neg ‘Take care that you don’t fall.’ Note that each of the negators is clausal: we are not dealing with a clausal negator in concord with some negative pronouns or negative adverbs of time, place or manner. There is no information as to whether this construction carries emphasis, but since at the time of the study neither enNEG …nietNEG nor niet#NEG carried emphasis, I suspect that the tripling didn’t carry emphasis either. A second positive answer takes us back to Lewo. According to Early (1994a, 1994b) reinterpreted a little by van der Auwera (2006), the language has two tripling strategies, one realis and the other irrealis. I will here focus on the irrealis strategy, because it is more transparent than the realis one (but see (28) for an example of realis tripling). With a second person subject prefix, the irrealis strategy has a prohibitive use. It comes in both a doubling and a tripling pattern. (41) Lewo a. Ve a-kan re! neg 2sg-eat neg ‘Don’t eat it!’

(Early 1994a: 76)

b. Ve a-kan re toko! neg 2sg-eat neg neg ‘Don’t eat it!’ c. Na-kan-ena toko! nom-eat-nom desist ‘Desist from eating!’ The tripling structure is an innovation. The marker toko derives from a verb meaning ‘desist’. As a verb, it combines with a nominalization, which also expresses prohibition. In the pattern in (41b), toko is therefore not a verb with a complement ‘eating’: the form of the verb ‘eat’ is wrong and if it were the complement, the construction as a whole would have to mean ‘desist from not eating’, i.e. the opposite from what is actually meant. Early (1994a: 77) stresses that (41b) has no phonological break between ve akan re and toko and that it does not therefore mean ‘Don’t eat it, desist!’ or ‘Don’t it, don’t!’ However, I find it rather plausible to suppose that this double clause structure was the point of origin, an emphatic prohibition, which then, in good Jespersenian fashion, bleached to a neutral prohibition. (42) and (43) schematize the two cases of tripling discussed above.



Johan van der Auwera

(42)

neg1 neg2

Illustration neg1 neg2 neg2#

neg2 neg2#

– emphatic

– emphatic

(43) neg1 neg2 X

+emphatic

neg1 neg2 neg3 +emphatic

enNEG …nietNEG …niet#NEG Illustration

–emphatic veNEG … reNEG … tokoNEG

Note, finally, that what happens in Lewo tripling is a bit like Flemish reemphasizing. In both languages we had a –emphatic doubling strategy which became emphatic. In Flemish it happened through reanalysis of the doubling construction, and Lewo through the addition of yet a third negator.

3.4

Skipping the reanalysis of doubling

Leaving out of account the tripling just discussed, we see that the scenarios discussed so far have a bipartite construction with two markers both of which are or become negative. If the new marker is not negative already (like Belgian Brabantic niet#), the reanalysis crucially takes place at a bipartite stage through the influence from the old negative marker. However, there is at least one situation where the old marker can be absent and the context can have a negation-inducing effect. Interestingly, the cases documented concern emphatic negation only. The example in (44) (discussed in Detges & Waltereit 2002: 187) concern French pas du tout ‘not at all’. In the context of pas du tout occurring as an answer to a polar question, it can be shortened to du tout. (44) - Est-ce votre avis? Is-this your opinion ‘Is this your opinion?’ - Du tout! of all ‘Not at all!’ The pas that could have been present is still fully negative, of course. It is not the case that du tout is taking over from pas. Rather, du tout is just an economical, i.e. elliptical version of pas du tout. It is hard to imagine how this phenomenon could ever make du tout a general negator and thus this phenomenon does not really deserve the ‘Jespersen cycle’ label. Yet these facts are of course related to the Jespersen facts, and the use of du tout should not be relegated to pragmatics. The elliptic use of du tout has been conventionalized, in a way that distinguishes it from English at all, which does not allow this elliptic use.



Chapter 3.  The Jespersen Cycles 

3.5

Another intermediate conclusion

The schema in (31) served as the intermediate conclusion of Section 3. It already showed more Jespersen cycles than usually accepted. But this section pointed at even more action. (31) showed that if constructions change in emphasis, they invariably lose emphasis. But Section 3.1 showed that if Overdiep, Haegeman and Breitbarth are right, constructions might also gain an emphatic value, at least to a limited degree. (31) showed that an old negator will disappear. Section 3.2 did not contest this, but it documented a stage in which an old negator may survive without negative and even without negative polarity meanings; in the case at hand, it survives as a marker of embeddedness. (31) embodies the claim that double negators can only simplify: this is not correct, after doubling one may get tripling. Finally, Section 3.4 reminds us of an ellipsis phenomenon that allows an emphasizing element to function as an exponent of negation without any intermediate doubling stage.

4. Conclusion By now there are too many possible Jespersen cycles to try to represent them all in one schema. I will use two schemas. The schema in (45) shows what happens to the negative constructions. In (45a), I describe the development of the combinations of NEG1 and either X (a minimizing or negative quantifier or any element of Miestamo (2005) asymmetry) or Ø (in case the new negator is the clause-final repetition of the existing one). The interesting stage is the third one: we either get a simplified solitary negator or we get negator tripling. In (45b-d), I show what happens to the components of the two source constructions. (45b) shows what happens to the old negator; the X in the third stage stands for the Middle English expletive and for the subordinateness marker of Belgian Brabantic. Sentences (45c-d) show how the new negators, arising through either reanalysis ((45c)) or addition ((45d)), become second negators in the doubling constructions, and then either new solitary ones or third negators. (45) a.

neg1 X or neg1 Ø

b. c. d.

neg1

neg1

X Ø

Ø neg2 or neg1 neg2 neg3

neg1 neg2 X

Ø

neg2

neg2

or

neg3

neg2

neg2

or

neg3

Diagram (46) describes the evolution of emphasis marking. Source constructions can be either –emphatic or +emphatic, and both reanalysis of an X element and the



Johan van der Auwera

repetition of a negator can yield –emphatic and +emphatic constructions, independent of whether the results are doubling or tripling constructions. Constructions can more easily change from +emphatic to –emphatic (this is simply inflationary bleaching), but a change from –emphatic to +emphatic is not to be ruled out (the Flemish reemphasizing scenario discussed in 3.1.) (46) neg1 X –emphatic

neg1 neg2 –emphatic

Ø neg2 –emphatic neg1 neg2 neg3 –emphatic

neg1 Ø

neg1 neg2 neg3 +emphatic

neg1 X +emphatic

neg1 neg2 +emphatic

Ø neg2 +emphatic

I hope to have shown that pluralizing the label ‘Jespersen cycle’ makes sense. The Jespersen cycle is more than a straightforward progression from elements like neNEG to pasNEG over neNEG … pasNEG, and even this cycle is far from straightforward. I am not, of course, claiming that every variation will be equally frequent in the world’s languages. Most certainly not: a tripling cycle might be rare, as might be the cycle that makes clause-final negators that start as repetitions of clause-earlier negators. But just how rare these phenomena are we don’t know. Tripling, for example, is at least not exceedingly rare: it is also attested in a few Italian dialects (Parry 1997: 251) and more widely in Bantu (Devos, Kasombo Tshibanda and van der Auwera 2008). I agree with van Gelderen (2008) that to get the whole picture, we have to look at the whole world.

References Anderwald, Lieselotte. 2002. Negation in Non-standard British English. Gaps, Regularizations and Asymmetries. London: Routledge. Barbiers, Sjef, van der Auwera, Johan, Bennis, Hans, Boef, Eefje, de Vogelaer, Gunther & van der Ham, Margreet. 2009. Syntactische Atlas van de Nederlanse Dialecten, Deel II (Syntactic Atlas of the Dutch Dialects, Vol. II). Amsterdam: Amsterdam University Press. Bernini, Giuliano & Ramat, Paolo. 1992. La frase negativa nelle lingue d’Europa. Bologna: Il Mulino. Bernini, Giuliano & Ramat, Paolo. 1996. Negative Sentences in the Languages of Europe. Berlin: Mouton. Beukema, Frits. 1999. Five ways of saying no: The development of sentential negation in English in a Government and Binding perspective. In Negation in the History of English, Ingrid Tieken-Boon van Ostade, Gunnel Tottie & Wim van der Wurff (eds), 9–26. Berlin: Mouton.



Chapter 3.  The Jespersen Cycles  Biberauer, Theresa. In print. Doubling vs omission: Insights from Afrikaans negation. In Microvariations in Syntactic Doubling, Sjef Barbiers, Margreet van der Ham, Olaf Koeneman & Maria Lekakou (eds). London: Emerald. Biberauer, Theresa. 2008. Multiple Jespersen’s cycles in Afrikaans negation and the future of Afrikaans negative concord. Handout Cycles Workshop Tempe. Blancquaert, Edgard. 1923. Over de dubbele ontkenning en nog wat. Handelingen van het Vlaamse filologencongres 6: 60–69. Breitbarth, Anne. In print. A hybrid approach to Jespersen’s cycle in West Germanic. Journal of Comparative Germanic Linguistics. Breitbarth, Anne & Haegeman, Liliane. 2008. Not continuity, but change: Stable change II in Jespersen’s cycle. Handout NORMS Workshop on Negation, Oslo. Buridant, Claude. 2000. Grammaire nouvelle de l’ancien français. Paris: Sedes. Burridge, Kate. 1983. On the development of Dutch negation. In Linguistics in the Netherlands 1983, Hans Bennis & W.U.S. van Lessen Kloeke (eds), 31–40. Dordrecht: Foris. Chaker, Salem & Caubet, Dominique. 1996. La négation en berbère et en arabe maghrébin. Paris: L’Harmattan. Croft, William. 1991. The evolution of negation. Journal of Linguistics 27:1–27. Crowley, Terry. 1982. The Paamese Language of Vanuatu [Pacific Linguistics, Series B, No. 87]. Canberra: ANU Dept. of Linguistics–Research Unit of Pacific Studies. Crowley, Terry (edited by John Lynch). 2006. The Avava Language of Central Malakula (Vanuatu). Canberra: Australian National University. Curnow, Timothy J. 1997. A Grammar of Awa Pit (Cuaiquer): An Indigenous Language of South-western Colombia. PhD dissertation, Australian National University. Dahl, Östen. 1979. Typology of sentence negation. Linguistics 17: 79–106. Dahl, Östen. 1995. Review of Kahrel & van den Berg. Journal of Linguistics 31: 476–478. Dahl, Östen. 2001. Inflationary effects in language and elsewhere. In Frequency and the Emergence of Linguistic Structure, Joan Bybee & Paul Hopper (eds), 471–480. Amsterdam: John Benjamins. De Cuypere, Ludovic, van der Auwera, Johan & Willems, Klaas. 2007. Double negation and iconicity. In Insistent images, Elżbieta Tabakowska, Christina Ljungberg & Olga Fischer (eds), 301–320. Amsterdam: John Benjamins. De Cuypere, Ludovic. 2008. Limiting the Iconic. From the Metatheoretical Foundations to the Creative Possibilities of Iconicity in Language [Iconiciy in Language and Literature 6]. Amsterdam: John Benjamins. Detges, Ulrich & Waltereit, Richard. 2002. Grammaticalization vs. reanalysis: A semantic-pragmatic account of functional change in grammar. Zeitschrift für Sprachwissenschaft 21: 151–195. Devos, Maud, Kasombo Tshibandi, Michael & van der Auwera, Johan. 2008. Double and triple negation in Kanincin (L53). Paper 3rd International Bantu Conference, Tervuren. Donhauser, Karin. 1996. Negationssyntax in der deutschen Sprachgeschichte. In Deutsch - typologisch, Ewald Lang & Gisela Zifonun (eds), 201–217. Berlin: Mouton de Gruyter. Early, Robert. 1994a. Lewo. In Typological Studies in Negation [Typological Studies in Language 29], Peter Kahrel & René van den Berg (eds), 65–92. Amsterdam: John Benjamins. Early, Robert. 1994b. A Grammar of Lewo, Vanuatu. PhD dissertation, Australian National University. Eckardt, Regine. 2003. Eine Runde im Jespersen-Zyklus. Negation, emphatische Negation und negativ-polare Elemente im Altfranzösischen. Konstanzer Online-Publikations-System. http://kops.ub.uni-konstanz.de/volltexte/2003/991/, accessed on Feb 17, 2009.



Johan van der Auwera Eckardt, Regine. 2006. Meaning Change in Grammaticalization. An Enquiry into Semantic Reanalysis. Oxford: OUP. Fischer, Olga, van Kemenade, Ans, Koopman, Willem & van der Wurff, Wim. 2000. The Syntax of Early English. Cambridge: CUP. Floricic, Franck & Molinu, Ludia. 2008. L’Italie et ses dialectes. Lalies 28. Actes des sessions de littérature et linguistique (La Beaume-les-Aix, 28–31  août 2007), 5–107. Paris: Presses de l’École Normale Supérieure. François, Alexandre. 2003. La sémantique de prédicat en mwotlap (Vanuatu). Leuven: Peeters. . Zeitschrift für Ägyptische Sprache und Altertumskunde Gardiner, Alan H. 1904. The word 41: 130–135. Haegeman, Liliane. 1995. The syntax of Negation. Cambridge: CUP. Haegeman, Liliane. 2002. West Flemish negation and the derivation of SOV-order in West Germanic. Nordic Journal of Linguistics 25: 154–189. Hansen, Maj-Britt Mosegaard. In print. The grammaticalization of negative reinforcers in Old and Middle French: A discourse-functional approach. In Current trends in diachronic semantics and pragmatics, Maj-Britt Mosegaard-Jensen & Jacqueline Visconti (eds). Hansen, Maj-Britt Mosegaard & Visconti, Jacqueline. In print. On the diachrony of “reinforced” negation in French and Italian. In Grammaticalization and Pragmatics: Facts, Approaches, Theoretical Issues, Corinne Rossari, Corina Cojocariu, Claudia Ricci & Adriana Spiridon (eds), Oxford: Elsevier. Harris, Martin. 1978. The Evolution of French Syntax. London: Longman. Haspelmath, Martin. 1997. Indefinite Pronouns. Oxford: OUP. Hock, Hans Henrich. 1991. Principles of Historical Linguistics. Berlin: Mouton. Hoeksema, Jack. 1997. Negation and negative concord in Middle Dutch. In Negation and polarity. Syntax and Semantics [Current Issues in Linguistic Theory 155], Danielle Forget, Paul Hirschbüller, France Martineau & María-Luisa Rivero (eds), 139–156. Amsterdam: John Benjamins. Honda, Isao. 1996. Negation: A Cross-linguistic Study. PhD dissertation, University of New York at Buffalo. Hopper, Paul & Closs Traugott, Elizabeth. 2003. Grammaticalization. Cambridge: CUP. Horn, Laurence R. 1989. A Natural History of Negation. Chicago IL: The University of Chicago Press. Horn, Laurence R. 1991. ‘Duplex negatio affirmat.’: The economy of double negation. In Papers from the 27th Regional Meeting of the Chicago Linguistic Society 1991. Part 2: The Parasession on Negation, Lise M. Dobrin, Lynn Nichols & Rosa M. Rodriguez (eds), 80–106. Chicago IL: Chicago Linguistic Society. Horn, Laurence R. 2001. Flaubert triggers, squatitive negation, and other quirks of grammar. In Perspectives on Negation and Polarity Items [Linguistik Aktuell-Linguistics Today 40], Jack Hoeksema, Hotze Rullmann, Victor Sanchez-Valencia & Ton van der Wouden (eds), 173– 200. Amsterdam: John Benjamins. Jäger, Agnes. 2008. History of German Negation [Linguistik Aktuell-Linguistics Today 118]. Amsterdam: John Benjamins. Jespersen, Otto. 1917. Negation in English and Other Languages. København: A. F. Høst & Søn. Joly, André. 1981. Structure psychique et structure sémiologique de la négation nexale dans les langues indo-européennes. Bulletin de la Société Linguistique de Paris 76: 99–154. Kawaguchi, Yuji. In print. Particules négatives du français: ne, pas, point et mie - Un aperçu historique. In Mélanges Yves-Charles Morin. Montreal: Presses de l’Université Laval.



Chapter 3.  The Jespersen Cycles  Kiparsky, Paul & Condoravdi, Cleo. 2006. Tracking Jespersen’s cycle. In Proceedings of the 2nd international conference of Modern Greek dialects and linguistic theory, Mark Janse, Brian Joseph & Angela Ralli (eds), 172–197. Patras: University of Patras. Kouwenberg, Sylvia. 1994. Berbice Dutch. In Typological Studies in Negation [Typological Studies in Language 29], Peter Kahrel & René van den Berg, 237–266. Amsterdam: John Benjamins. Kroskrity, Paul V. 1984. Negation and subordination in Arizona Tewa: Discourse pragmatics influencing syntax. International Journal of American Linguistics 50: 94–104. Larrivée, Pierre. 2004. L’association négative. Depuis la syntaxe jusqu’à l’interprétation. Genève: Droz. Larrivée, Pierre. 2008. The pragmatic triggers of the Jespersen cycle: Default, activation, and the history of negation in French. Ms. Lenz, Barbara. 1996. Negationsverstärkung und Jespersens Zyklus im Deutschen und in anderen europäischen Sprachen. In Deutsch - typologisch, Ewald Lang & Gisela Zifonun (eds), 183–200. Berlin: Mouton de Gruyter. Lucas, Christopher. 2007. Jespersen’s cycle in Arabic and Berber. Transactions of the Philological Society 105: 398–431. Lucas, Christopher. 2008. Contact as catalyst: The case for Coptic influence in the development of Arabic negation. Ms, Cambridge University. Marchello-Nizia, Christiane. 1979. Histoire de la langue française aux XIVe et XVe siècles. Paris: Bordas. Martineau, France & Mougeon, Raymond. 2003. A sociolinguistic study of the origins of ne deletion in European and Quebec French. Language 79: 118–152. Martineau, France & Vinet, Marie-Thérèse. 2005. Microvariation in French negation markers: A historical perspective. In Grammaticalization and Parametric Variation, Montserrrat Batllori, Maria-Lluïsa Hernanz, Carme Picallo & Francesc Roca (eds), 194–205. Oxford: OUP. Mazzon, Gabriella. 2004. A History of English Negation. Harlow: Pearson Longman. Meillet, Antoine. 1912. L’évolution des formes grammaticales. Scientia 12:384–400. (Reprinted in Meillet, Antoine. 1926. Linguistique historique et linguistique générale. 130–148. Paris: H. Champion). Miestamo, Matti. 2005. Standard Negation. The Negation of Declarative Verbal Main Clauses in a Typological Perspective. Berlin: Mouton de Gruyter. Möhren, Frankwalt. 1980. Le renforcement affectif de la négation d’une valeur minimale en ancien français. Tübingen: Niemeyer. Muller, Claude. 1991. La négation en français. Syntaxe, sémantique et éléments de comparaison avec les autres langues romanes. Genève: Droz. Muller, Claude. 2004. Sur quelques emplois particuliers de “pas” et “point” à l’aube du français classique. Langue française 143:19–32. Nater, Hank F. 1984. The Bella Coola Language. Ottawa: National Museums of Canada. Neuckermans, Annemie. 2008. Negatie in de Vlaamse dialecten volgens de gegevens van de Syntactische Atlas van de Nederlandse dialecten (SAND). PhD dissertation, University of Ghent. Overdiep, Gerrit S. 1933a. Dialectstudie en syntaxis. Onze Taaltuin 2: 18–23. Overdiep, Gerrit S. 1933b. Dialectstudie en syntaxis: een overgangsklank. Onze Taaltuin 2: 44–45. Overdiep, Gerrit S. 1937. Stilistische Grammatica van het Moderne Nederlandsch. Zwolle: Tjeenk Willink. Overdiep, Gerrit S. 1940. De Volkstaal van Katwijk aan zee. Antwerpen: Standaard. Parry, M. Mair. 1997. Preverbal negation and clitic ordering, with particular reference to a group of North-West Italian dialects. Zeitschrift für romanische Philologie 113: 243–270.



Johan van der Auwera Pauwels, Jan L. 1958. Het Dialect van Aarschot en omstreken. Brussel: Belgisch interuniversitair centrum voor Neerlandistiek. Pauwels, Jan L. 1974. Expletief nie en andere herhalingswoorden als zinsafsluiters. In Taalkunde - ’n lewe. Studies opgedra aan prof. W. Kempen by geleentheid van sy 65st verjaardag, F.F. Oldendal (ed.), 73–76. Kaapstad: Tafelberg. Perle, Friedrich. 1878. Die Negation im Altfranzösichen. Zeitschrift für romanische Philologie 2: 2–24, 407–418. Pohl, Jacques. 1968. Ne dans le français parlé contemporain: Les modalités de son abandon. Actas del XI congresso internacional de linguistica y filología Romanicas, Madrid. Postal, Paul. 2004. Skeptical Linguistic Essays. Oxford: OUP. Postma, Gertjan. 2002. De enkelvoudige clitische negatie in het Middelnederlands en de Jespersen-cyclus. Nederlandse Taalkunde 7: 44–82. Ramat, Paolo. 2006. Italian negatives from a typological/areal point of view. In Scritti in onore di Emanuele Banfi in occasione del suo 60º compleanno, Nicola Grandi & Gabriele Iannàccaro (eds), 355–370. Cesena: Caissa Italia. Roberge, Paul. 2000. Etymological opacity, hybridization, and the Afrikaans brace negation. American Journal of Germanic Linguistics & Literatures 12: 101–176. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: CUP. Roberts, Ian. 2007. Diachronic Syntax. Oxford: OUP. Schøsler, Lene. 2008. Manuscrits du Charroi de Nimes. Handout Leverhulme Trust seminar on cycles of grammaticalization (Birmingham). Schwegler, Armin. 1983. Predicate negation and word-order change: A problem of multiple negation. Lingua 61: 297–334. Schwegler, Armin. 1988. Word-order changes in predicate negation strategies in Romance languages. Diachronica 5: 21–58. Schwegler, Armin. 1990. Analycity and Syntheticity. A Diachronic Perspective with Special Reference to Romance Languages. Berlin: Mouton de Gruyter. Schwegler, Armin. 1991. Predicate negation in contemporary Brazilian Portuguese. Orbis 34: 187–214. Schwegler, Armin. 1996. La doble negación dominicana y la génesis del español caribeño. Hispanic Linguistics 8: 247–315. Schwenter, Scott A. 2006. Fine-tuning Jespersen’s cycle. In Drawing the boundaries of meaning: Neo-Gricean studies in pragmatics and semantics in honor of Laurence R. Horn., Betty Birner & Gregory Ward (eds), 327–344. Amsterdam: John Benjamins. van der Auwera, Johan. 2006. Why languages prefer prohibitives. 外国语 (Wai guo ju – Journal of Foreign Languages) 161:2–25. van der Auwera, Johan In print a. Deutsch als eine/die durchschnittseuropäische Sprache. In Unsere sprachlichen Nachbarn in Europa. Die Kontaktbeziehungen zwischen Deutsch und seinen Grenznachbarn, Christol Stolz (ed.). Bremen: Brockmeyer. van der Auwera, Johan. In print b. On the diachrony of negation. In The Expression of Negation, Laurence R. Horn (ed.), Berlin: Mouton de Gruyter. van der Auwera, Johan, de Cuypere, Ludovic & Neuckermans, Annemie. 2006. Negative indefinites: A typological and diachronic perspective on a Brabantic construction. In Types of Variation. Diachronic, Dialectal and Typological Interfaces, Terttu Nevailanen, Juhani Klemola & Mikko Laitinen (eds), 305–319. Amsterdam: John Benjamins.



Chapter 3.  The Jespersen Cycles  van der Auwera, Johan & Neuckermans, Annemie. 2004. Jespersen’s cycle and the interaction of predicate and quantifier negation. In Dialectology meets Typology. Dialect Grammar from a Cross-linguistic Perspective, Bernd Kortmann (ed.), 453–478. Berlin: Mouton de Gruyter. Van der Horst, Joop M. 2008. Geschiedenis van de Nederlandse Syntaxis 1. Leuven: Universitaire Pers Leuven. Van der Horst, Joop M. & van der Wal, Marijke. 1979. Negatieverschijnselen en woordvolgorde in de geschiedenis van het Nederlands. Tijdschrift voor Nederlandse Taal-en Letterkunde 95: 6–37. Van der Wouden, Ton. 1997. Negative Contexts, Polarity and Multiple Negation. London: Routledge. Van der Wurff, Wim. 1999. On expletive negation with adversative predicates in the history of English. In Negation in the History of English, Ingrid Tieken-Boon van Ostade, Gunnel Tottie & Wim van der Wurff (eds), 295–327. Berlin: Mouton. Van Gelderen, Elly. 2008. Negative cycles. Linguistic Typology 12: 195–243. Van Kemenade, Ans. 1999. Sentential negation and clause structure in Old English. In Negation in the History of English, Ingrid Tieken-Boon van Ostade, Gunnel Tottie & Wim van der Wurff (eds), 147–165. Berlin: Mouton. Vennemann, Theo. 1974. Topics, subjects, and word order. From SXV to SVX via TVX. In Historical Linguistics, John M. Anderson & Charles Jones (eds), 339–376. Amsterdam: North Holland. Wackernagel, Jacob. 1926. Vorlesungen über Syntax. 2. Reihe. Basel: Birkhäuser. Wallage, Phillip. 2008 Jespersen’s cycle in Middle English: Parametric variation and grammatical competition. Lingua 118: 643–674. Willis, David. 2005. A minimalist approach to Jespersen’s cycle in Welsh. Ms. Winand, Jean. 1997. La négation bn … iwⳍ en néo-égyptien. Lingua Aegyptia 5: 223–236. Zanutttini, Raffaella. 1997. Negation and Clausal Structure. A Comparative Study of Romance Languages. Oxford: OUP. Zeijlstra, Hedde. 2004. Sentential Negation and Negative Concord. PhD dissertation, University of Amsterdam.

chapter 4

The negative cycle in Early and Modern Russian*2 Olena Tsurska

Arizona State University In this paper, I examine changes in the negative structures in the history of Russian. The data indicate that both in Early and Modern Russian, multiple negative elements could be used simultaneously to express sentential negation. However, in Early Russian, the main preverbal marker ne is sometimes optional when preverbal n-words are present, while in Modern Russian, the use of ne is mandatory when both pre- and postverbal n-words are used. Using the Minimalist framework, I show that Russian sentential negation has undergone cyclical changes from Non-Strict to Strict Negative Concord, which involved changes in the feature interpretability of Russian negative elements, more specifically the loss of the [ineg] feature of the preverbal negative marker.

1. Introduction13 It has been suggested in the literature that linguistic change is cyclical, that is languages tend to change from synthetic to analytic to synthetic. Various researchers have studied different types of linguistic cycles such as subject cycle, aspect cycle, etc (Hodge 1970; Givón 1984; van Gelderen 2008, in progress). One of the most famous studies on the * Parts of this paper were presented at the Linguistic Cycles Workshop at Arizona State University and LASSO XXXVII in 2008. I thank these audiences for their comments on the paper. I am deeply indebted to Elly van Gelderen for fruitful theoretical discussions, insightful comments, helpful suggestions, and constant inspiration. I am also grateful to Jack Hoeksema, Theresa Biberauer, Diana Vedovato, Cathleen Waters, and Terje Lohndal for helpful feedback. All remaining errors are mine. 1. List of abbreviations used in this article: CE – common era, FUT – future tense, iNEG – interpretable negative feature, iPOL – interpretable polarity feature, LF – logical feature, LI – lexical item, M – masculine, NC – negative concord, NEG – negative marker, NegP – negative phrase, NPI – negative polarity item, NQ – negative quantifier, PART – particle, PF – phonological feature, PL – plural, PST – past tense, SG – singular, uNEG – uninterpretable negative feature, uPOL – uninterpretable polarity feature.



Olena Tsurska

negative cycle is the monograph by Otto Jespersen, Negation in English and Other Languages (1917), in which he argued that negative structures in languages tend to change in the following manner. First, negation in a language is expressed by a negative adverb or some adverbial-like expression (e.g. Old English na wiht ‘no creature’, which later becomes not). With time, the adverb loses its semantic power and weakens becoming a negative particle, and it may further become too weak, turn into a clitic, and get attached to the verb it is negating (e.g. English -n’t ). The negative clitic then would require some extra reinforcement and an additional negative element may start being used together with the clitic to express negation (double negation in English as in I don’t see nothing.) In many cases, the clitic may get completely deleted, the reinforcing negative element becomes the only expression of negation, and the cycle begins again. The negative cycle in Slavic has not been studied very extensively. The negative structure in Russian, in particular, has not undergone the process of grammaticalization that is usually involved in a linguistic cycle. However, the structure of the Russian negative concord and the interaction between the preverbal negative marker and other negative elements in a sentence has changed over time. In this study, I examine the data from several Early Russian texts as well as Modern Non-Standard Russian dialects, which reveal the ways for expressing sentential negation that are not present in Modern Standard Russian. Using the Minimalist program and the analysis of the negative concord by Zeijlstra (2004, 2008), I show that the Russian negative concord has been undergoing some changes. The data indicate that Early Russian had features of a Non-Strict Negative Concord language, in which the preverbal n-words could be used without the preverbal negative marker to express negation. Modern Standard Russian is a Strict Negative Concord language, and no n-words are able to express sentential negation alone. The corpora data from Modern Russian also indicate the frequent use of minimizers, e.g. ni kapli (‘not a drop’), that strengthen the negative marker ne. I argue that this indicates the weakening of the [neg] features of ne, which might lead to the continuation of the cycle and future reanalysis (if not disappearance) of ne. The paper is organized as follows. Section 2 presents the theoretical framework used in this study. Section 3 includes a brief history of Russian and the data from a variety of Early Russian and Modern Russian sources. Section 4 presents an interpretation of the data described in Section 3 and the analysis of the NC cycle in Russian. Section 5 is a conclusion.

2. Theoretical considerations In this section, I provide a brief overview of the main theoretical assumptions that I use in my analysis of the Russian data.



Chapter 4.  The negative cycle in Early and Modern Russian 

Feature checking in Minimalism

2.1

According to Chomsky (2001, 2004, 2005, 2007), syntactic structures are formed from the projections of lexical items (LIs) that are taken out of the lexicon. Each LI is considered to be a bundle of features: phonological, semantic, and formal. Phonological features are interpreted at PF, and semantic features are interpreted at LF. Formal features are categorical features and phi-features, and they carry the information for the syntactic component. Formal features can be either interpretable or uninterpretable. Uninterpretable features need to be checked and deleted during the syntactic derivation, otherwise they cannot be interpreted at LF, and the derivation crashes. Feature checking is a process when a category with uninterpretable features can check them against the same interpretable features of another category. Chomsky (2001) argues that this takes place during the operation Agree when the category with an uninterpretable feature called a Probe looks down for another category with the same interpretable feature (Goal) and gets the uninterpretable feature checked, or valued. As summarized in Zeijlstra (2008), some researchers have slightly modified the Chomskyan Probe-Goal relationship and have suggested that Agree can take place from bottom up, i.e. the Probe with the uninterpretable feature can be looking up for the corresponding Goal that c-commands the Probe. According to Zeijlstra (2008: 21), this idea can be “traced back to Rizzi’s (1989) proposal for criteria (where semantically active operators always had to occupy specifier positions whilst agreeing with their respective heads)”. Several researchers (Ura 1996; Hiraiwa 2001) have suggested the notion of Multiple Agree. This phenomenon refers to cases when a single interpretable formal feature enters into an Agree relationship with multiple uninterpretable formal features given that the proper locality conditions are fulfilled:

(1) Multiple Agree (multiple feature checking) with a single probe is a single simultaneous syntactic operation; Agree applies to all matched goals at the same derivational point derivationally simultaneously. (Hiraiwa 2001: 69)

In this study, I consider the notion of Multiple Agree2 as well as the bottom up ProbeGoal relationship plausible for my analysis of the feature interpretation of multiple negative elements in Russian negative sentences.

2.2

Negative Concord

Russian is a Negative Concord (NC) language. NC is a linguistic phenomenon that describes linguistic environments, in which multiple negative elements in a syntactic

2. Please see Haegeman & Lohndal (forthcoming) for an alternative account not using Multiple Agree.



Olena Tsurska

structure of the sentence express only one semantic negation. Italian is traditionally given as an example of an NC language (from Zanuttini 1997): (2) Non ha detto niente. neg has said nothing ‘S/he hasn’t said anything.’

Italian

As shown in (2), in addition to the negative marker, negative sentences in NC languages can also contain other negative elements, called n-words (Laka 1990; Giannakidou 1998, 2000). N-words are negative indefinites that, on the one hand, behave like inherently negative elements, e.g. they can be used as one-word answers, as in (3), and on the other hand, require negative licensing in other contexts, as in (4): (3) Kogo ty videl? Nikogo. who you saw nobody ‘Who did you see?’ ‘Nobody.’

Russian

(4) No vi ninguno. neg 1sg-pst-see none ‘I didn’t see any.’

Spanish (from Vallduví 1994: 266)

In Russian, Czech, and other Slavic languages, the preverbal negative marker is obligatory whenever n-words are used in the sentence. In Italian, however, the negative marker is mandatory only if n-words are in the postverbal position, as in (2) above. Giannakidou (1998), following den Besten (1986), calls the type of NC in which nwords in the postverbal position require the presence of the negative marker, NC proper (which can be further divided into Strict and Non-Strict). There are some languages where NC involves co-occurrence of two or more n-words without the negative marker, and this kind of NC is called negative spread, however, this phenomenon will not be of relevance in this study. Two other types of elements that may occur in negative environments are negative polarity items (NPIs) and negative quantifiers (NQs). Unlike n-words, NPIs do not occur only in negative sentences, but can also appear in questions (Hoeksema 1997). NPIs require a licensor, which is often a negative or interrogative operator. N-words are often argued to be self-licensing (Ladusaw 1992). NQs require no trigger as well, and unlike n-words and NPIs, NQs are semantically negative, i.e. they carry independent negative force and can express negation without the presence of any other negative element:

(5) I saw nobody.

If two or more NQs occur in the same clause, they yield a double negation reading. According to Giannakidou (1998, 2000), there are two types of NC proper: Strict and Non-Strict. Strict NC languages, like modern Slavic languages, require n-words to



Chapter 4.  The negative cycle in Early and Modern Russian 

be always accompanied by a preverbal negative marker, even if the n-word is in the preverbal position. The following examples illustrate that: (6) Nikdo nic neviděl. nobody nothing neg-saw ‘Nobody saw anything.’ (7) Neviděl nic. neg-saw nothing ‘He didn’t see anything.’ (8) Nic nikomu nie powiedziałem. nothing to- nobody neg said-1sg ‘I didn’t tell anybody anything.’

Czech

(from Rechzieglova 1995: 119) Polish

(9) Marysia nie dała nikomu książki. Mary neg gave nobody book ‘Mary didn’t give anyone a/the book.’ (from Przepiórkowski and Kupść 1997: 8) Non-Strict NC languages, such as Italian and Spanish, do not allow n-words to precede the preverbal negative marker, i.e. if the n-word occurs in a subject position, the addition of the preverbal negative marker makes the sentence ungrammatical (unless the n-word is emphasized, in which case the sentence will render a double negation reading (from Zanuttini 1997: 8, 1997: 10 and Vallduví 1994: 266): (10) Non ho visto nessuno. neg have seen nobody ‘I haven’t seen anybody.’

Italian

(11) Non ha ditto niente. neg has said nothing ‘S/he hasn’t said anything.’ (12) Nessuno (*non) ha detto niente. (13) No he visto a nadie. neg has seen to nobody ‘I haven’t seen anybody.’ (14) No vi ninguno. neg saw none ‘I didn’t see any.’ (15) Nadie (*no) ha dicho nada. nobody neg has said nothing

Spanish



Olena Tsurska

There are different varieties of Catalan with respect to the NC type. One variety requires n-words to be always used with preverbal negative markers, thus exhibiting a Strict NC, as shown in (16) (from Zeijlstra 2004: 147): (16) a. No ha vist (pas)3 ningu. neg has.3sg seen neg n-body ‘He didn’t see anybody.’

Catalan (I)

b. No functiona (pas) res. neg works neg n-thing ‘Nothing works.’ c. Res *(no) functiona (pas). In a Non-Strict NC variety of Catalan, preverbal n-words cannot be used with the preverbal negative markers, as in (17): (17) a. No ha vist (pas) ningu. neg has.3sg seen neg n-body ‘He didn’t see anybody.’

Catalan (II)

b. No functiona (pas) res. Neg works neg n-thing ‘Nothing works.’ c. Res (*no) functiona (pas).

2.3

NEG feature interpretation in NC languages

In this study, I adopt Zeijlstra’s (2004, 2008) proposal that NC is a case of syntactic agreement between the interpretable negative feature of the negative operator and uninterpretable features of other negative elements. Zeijlstra argues that “NC is an Agree relation between a single feature [ineg] and one or more features [uneg]” (2008: 20). N-words in NC languages are indefinites that are formally, but not semantically negative, i.e. they have a syntactic [uneg] feature. According to Zeijlstra, the negative operator carrying [ineg] feature in NC languages can be covert or overt. The negative marker in Non-Strict NC languages (like Italian non) has [ineg] and is an overt negative operator. Contrary, in Strict NC languages, the negative marker carries [uneg]. Zeijlstra presents a number of supporting arguments for this proposal, including the evidence that the negative marker in Italian is a phonological realization of the negative operator because it falls in the scope of the quantifying DP like molto, which is not the case in the Strict NC language, like Czech.

3. Pas is the optional negative adverb that maybe used in NC chains in both Catalan varieties (Zeijlstra 2004: 147).



Chapter 4.  The negative cycle in Early and Modern Russian 

In Strict NC languages, the covert negative operator “immediately c-commands the highest instance of [uneg]”, as shown in the following examples from Czech (Zeijlstra 2008: 25): (18) Dnes nikdo *(ne)volá. Today n-body neg.calls ‘Today nobody is calling.’

Czech

(19) Milan nevidi nikoho. Milan neg.sees n-body ‘Milan doesn’t see anybody.’

Czech

(20) [Dnes Op¬[iNEG] [TP nikdo[uNEG] nevola[uNEG]]] (21) [TP Milan Op¬[iNEG] [Neg° nevidi[uNEG]i [vP nikoho[uNEG] ti]]] In Italian, as mentioned earlier, the negative marker is a phonological realization of the negative operator, and the postverbal n-words can check their [uneg] features against it, as in (23) (Zeijlstra 2008: 26): (22) Gianni non telefona a nessuno. Gianni neg calls to n-body ‘Gianni doesn’t call anybody.’

Italian

(23) [TP Gianni [NegP non[iNEG] telefona [vP a nessuno[uNEG]]]]. When an n-word precedes the verb in a sentence without the preverbal negative marker, then there is no overt negative operator to check the [uneg] features, therefore an abstract operator, immediately c-commanding the n-word is assumed (Zeijlstra 2008: 27): (24) Ieri nessuno ha telefonato a nessuno. Yesterday n-body has called to n-body ‘Yesterday nobody called anybody.’

Italian

(25) [Ieri Op¬[iNEG] [TP nessuno ha telefonato a nessuno]]. This analysis is suitable for my examination of the negative cycle in Russian. The cycle involves the weakening of the negative elements, and therefore weakening and loss of their features. My data indicate that Russian has changed from the Non-Strict to Strict NC language, and Zeijlstra’s analysis of NC as syntactic agreement allows me to interpret the change of the NC type in terms of feature change (e.g. the loss of the interpretable [ineg] feature of the sentential negative marker ne). The use of minimizers in Modern Russian also proves that ne continues to lose its strength and requires reinforcement. This analysis is also consistent with the Feature Economy principle, which states that it is more economical to minimize semantic and interpretable features (van Gelderen in progress) than not to. In this section, I have outlined the main theoretical considerations relevant to my analysis of the Russian data, to which I now turn.



Olena Tsurska

3. Negation in Russian 3.1

A brief note on the history of Russian

Russian, like other Slavic languages, comes from a Proto-Slavic, or Common Slavic/Slavonic language that existed around two thousand years ago (Press 2007). By the ninth century CE, the differences among numerous dialects of the Proto-Slavic language become quite significant due to the migration and formation of individual Slavic states. According to Press (2007, 2008), in the 9th century, the books were written in the ‘Slavonic’ language to be used in the Eastern Christian Church. A linguist Constantine (who before his death took a name Cyril) created a written Slavonic alphabet based on the language spoken in the area where he lived (modern Macedonia). This language was Old Church Slavonic, and the alphabet is called ‘Glagolitic’. This alphabet was widely used in Bulgaria, but some scholars who came to the Balkans in the tenth century began using another alphabet closer to Greek, which is called ‘Cyrillic’. This language and Cyrillic alphabet came to Russia in 988, when it took on Christianity. The language was very close to languages spoken in Russia, and over the time it “became more localized/Rusified, though it retained its ‘high style’ character, and may be referred to as the ‘Rus(s)ian recension of Old Church Slavonic’, or ‘Rus(s)ian Church Slavonic’” (Press 2008: 7). In the 14–15th centuries, some scholars came to Russia from the Balkans and brought with them an older form of Church Slavonic with many features, which most likely had ceased to exist at that time in somewhat isolated Russia. As Press points out, these features were adopted by the Russians because of their desire to be “the bastion of Christianity in the East” (2008: 8). But only the written language was influenced, and over the next couple of centuries it started to be used as purely ‘Church language’, or Russian Church Slavonic, and not in everyday speech. When Peter the Great implemented linguistic reforms, he wanted the ‘Standard Russian’ to reflect the ‘real’ speech. The language that is now Russian fully emerged by the middle of the nineteenth century.

3.2

Early Russian

In Early Russian (11th -12th cc.), there appear to be two competing strategies for expressing sentential negation4. The first strategy, an indicator of its Indo-European roots, is the use of the preverbal negative marker ne: (26) …Koli sokol… ne dast gnezda svojego v obidu … if falcon neg give nest his to harm ‘If a falcon doesn’t let his nest be harmed’…

(The Igor Tale, 12th c.)

4. The examples of sentences in Early Russian are taken from the Anthology of Old Russian Literature by Stender-Petersen(1954) and from the Historical Grammar of Russian by Borkovskij (1978).



Chapter 4.  The negative cycle in Early and Modern Russian 

Occasionally, n-words can be also used in the presence of ne (preceding or following it) to render a single instance of negation: …i Xam i Afet, razdelivše zemlju, i žrebii metavše… ne and Xam and Afet divide lands and coins toss neg prestupati nikomuže v žrebii bratene… cross no one part in deal taken ‘… and Xam and Afet divided the lands and tossed coins [to determine what land belongs to whom] … for anyone not to cross the brother’s lands…’ (The Tales of Bygone Years, 12th c.) (27)

(28) On že reče mi nikomou že bedy ne stvoriti he part tell to-me to-no-one part harm neg cause ‘He tells me to cause harm to nobody.’ (Sinaj Paterik, 11th-12th c.) (29) Nikoli že ne prezeri ouboga plačušta… never part neg despise to-God crying ‘Never despise those crying to God…’

(Izbornik, 1076)

In addition to the use of the preverbal marker ne, n-words preceding the main verb also serve as markers of sentential negation without the additional help of ne: (30) … i nikto že mene vidit. and no-one part me sees ‘…And nobody sees me.’ (31) I nikto že prixodil k nim… and no-one part came to them ‘And nobody came to them...’

(Izbornik, 1076)

(Life of Feodosij Pečersky, 12th c.)

By the 15th -17th centuries, the use of preverbal n-words without the preverbal negative marker ne became less and less frequent, and according to Borkovskij (1978), sentences such as (32) below became more prevalent in formal written documents and not in colloquial speech: (32) Nikto že bez truda venčan budet. no-one part without work wed fut ‘Nobody will be wed without labor.’

(Domostroj, 16th c.)

N-words used together with the preverbal marker are quite prevalent in the texts of that period: (33) … I ja is požaru ničevo ne pospel vynest’ and I from fire nothing neg managed take ‘And I didn’t have time to take anything out of the fire.’ (An Appeal of V. Krečatnikov, 17th c.)



Olena Tsurska

(34) Slugi že mimo xodašče ne daša jemu ničto že servants part by passed neg gave to-him nothing part ‘The servants that passed by didn’t give him anything.’ (The Tale about Saints Boris and Gleb, 14th c.) (35)

Takovago gromu skazyvajut nigde nikto ne slyxal ot načala such thunder say nowhere nobody neg heard from start miru world ‘They say that no one has heard such a thunder anywhere since the beginning of the world.’ (Second Novgorod Chronicle, 14th c.)

Examples (33)-(35) show that, by the 17th century, Russian had the features of the Strict NC language.

3.3

Modern Standard Russian

In Modern Standard Russian, sentential negation can be expressed by means of the negative marker ne that always precedes the verb and nothing can typically be placed between the verb and ne: (36) Ja ne pročitala etu knigu. I neg read this book ‘I have not read this book.’ (37) My ne xodili v kino. we neg went to movies. ‘We did not go to the movies.’ Since Russian is an NC language, n-words occur together with ne in a sentence. Together with the negative marker, n-words yield one instance of semantic negation (examples (42) and (43) below are from Brown (1999)): (38) On nikogo ne videl. he no-one neg see.pst.m ‘He did not see anyone.’ (39) *On nikogo videl. (40) Ja včera nikuda ne xodila. I yesterday nowhere neg went ‘I did not go anywhere yesterday.’ (41) *Ja včera nikuda xodila.



Chapter 4.  The negative cycle in Early and Modern Russian 

(42) Nikto nikogda nigde ni s kem ne tancuet. no-who no-when no-where no with who neg dance ‘No one ever dances anywhere with anyone.’ (43) *Nikto nikogda nigde ni s kem tancuet. As shown in (39), (41), and (43) above, the presence of n-words in a sentence is ungrammatical without ne, thus Russian, as mentioned earlier, is a Strict NC language.

3.4

Modern non-standard Russian dialects

Most non-standard dialects of Russian exhibit similar NC patterns, i.e. n-words are used only in the presence of the negative marker ne5: (44) Nicevo-to nigde ne bylo nothing nowhere neg was ‘There wasn’t anything anywhere.’

(Arhangelsk region)

(45) Ničevo my ne pokupali takova nothing we neg bought of-a-sort ‘We bought nothing of a sort.’

(St. Petersburg region)

(46) Jona ni budo nikogda slobodna she neg will-be never free ‘She will never be free.’

(Novgorod region)

(47) Naverno ni odin i ne voz’m’ot – srodu taka v’ortuška! probably not-one and neg take from-birth such flirt ‘Probably no one will take [her] – she has always been such a flirt!’ (Pskov region) (48) Nečovo ne podelyš nothing neg do ‘There is nothing you can do.’ (49) Nikto uxu u nas ne pominat nobody fish-soup with us neg mention ‘Nobody mentions ‘uha’ (fish soup) at our place.’

(Tver’ aka Kalinin region)

(Pskov region)

(50) Laptej ni najdeš ni u kogo lapti neg find not with anybody ‘Nobody has lapti (‘handmade shoes from bark’) (you won’t find lapti).’ (Kaluga region)

5. The examples in (44)-(51) come from the collection of texts in non-standard Russian dialects by Mel’ničenko (1985).



Olena Tsurska

Table 1.  History of Russian sentential negation6 11th - 12th c.

14th - 17th c.

Present Standard Russian

ne +V ne +V Pre- and postverbal Pre- and postverbal n-words + ne +V n-words + ne +V (frequent) Preverbal n-words Preverbal n-words without ne without ne (less frequent)

Non-Standard Russian

ne +V ne +V ne +V Pre- and postverbal Pre- and postverbal Preverbal n-words + ne +V n-words + ne +V n-words without ne Ne +V + minimizers ni kapli/ni čut’6

(51) … Ničavo iz laxmotkaf v lafkax n’a brali nothing from rags at stores neg take ‘We didn’t take any rags from the stores.’

(Ryazan region)

However, certain North Russian dialects (including the ones yielding the aforementioned NC examples) still use the single negative marker strategy, i.e. they use n-words without the mandatory preverbal ne (examples from Gorškova and Xmelevskaja 1986): (52) Za vsju žys’ mne nikudy prixodilos’ xodit’ for whole life to-me nowhere had to go ‘I didn’t have to go anywhere all my life.’ (53) Mne ništo zdilajetsja to-me nothing happen ‘Nothing will happen to me.’ (54) Nikto tut bol’še xoronenyj, kromja etix voinof nobody here more buried except these soldiers ‘Nobody is buried here except these soldiers.’ The strategies for expressing sentential negation in the history of Russian are summarized in Table 1 above.

4. Negative/ NC cycle in Russian 4.1

Feature interpretation in NC structures in the history of Russian

The data from the Russian texts of the 11th-16th cc. show several patterns in expressing sentential negation. The first pattern is the use of the preverbal negative marker ne without any n-words accompanying it. The second pattern involves the use of the 6. See the discussion on the use of minimizers in Russian in 4.2 below.



Chapter 4.  The negative cycle in Early and Modern Russian 

preverbal n-words, functioning as subjects and objects, and the absence of the preverbal ne. Finally, the last pattern is the use of both preverbal ne together with preverbal and postverbal n-words. The aforementioned patterns mirror different strategies employed by Negative Concord (NC) languages. On the one hand, Early Russian resembles Italian and thus appears to employ the Non-Strict NC strategies. On the other hand, however, there are examples of sentences, though not very frequent, such as (27)–(29) and (33)–(35), that look like the Modern Russian Strict NC. Early Russian appears to be patterning the varieties of Catalan described in Zeijlstra (2004, 2008), in a way that it allows the preverbal n-words to be either sole markers of the sentential negation (like Catalan II) or be happily co-existing with the preverbal negative marker ne to render one instance of sentential negation (Catalan I). Brown (2003) has looked at the negation strategies in Old Church Slavonic texts and Modern Russian and analyzed sentential negation in those languages within the Minimalist framework. Brown claims that all NC languages differ with respect to which features n-words possess. She argues that all n-words need to check some sort of a feature, and in some languages this uninterpretable feature is [u pol] and in others, [uneg]. [neg] feature in the Brown’s analysis is an abstract feature that expresses sentential negation (the head of the NegP has this feature that is interpretable); [upol] feature is also something that the head of the NegP has, and it is a feature that “ensures that a negated clause contain some overt expression of negation” (2003: 169). Therefore, “which of these two features [[u pol] or [u neg]] occurs in the sub-label of nwords in a given language, in conjunction with the requirements of NegP, will determine whether or not a sentential negation marker is obligatory, optional, or disallowed with preverbal n-words” (2003: 170). Brown further argues that, in Italian, n-words are not inherently negative, and they have [u pol] in their sublable. The negative marker non also has the same feature, while Neg has the interpretable [neg] feature. Either the n-word or non move overtly to the NegP to check off the [upol] feature of the head and also check their own [upol] feature by the interpretable [neg] feature of the NegP. If the n-word is preverbal, it checks the [upol] feature of the head, making it unnecessary to have non in the sentence. In sentences with no n-words or postverbal n-words, non is inserted into the head position of NegP to check off the [upol] feature, and the postverbal n-words check their [upol] features covertly (Brown 2003). Modern Russian n-words, according to Brown (2003), are inherently negative, i.e. they have an unintepretable negative feature [uneg] instead of [upol]. The sentential marker ne, however, has a [upol] feature, thus only ne and not n-words can check the [upol] feature of the NegP. In languages, like Catalan and Old Church Slavonic, the two NC patterns coexist rendering different negative sentences. Brown points out that the NC system in Old Church Slavonic was in the “state of flux”, which was a result of the on-going shift from the Italian-type to the Modern Russian-type NC. “This shift involves a change in the uninterpretable feature in the label of n-words from [u pol]



Olena Tsurska

to [u neg]” (Brown 2003: 177). Brown notes that the Catalan-type NC appears to be an intermediate stage in such a shift. Brown’s account raises several questions. First, she does not completely explain the motivation for the introduction of the [u pol] feature into the NC configuration. Second, the checking process is also very vague and does not seem to correspond to the feature checking procedures described in the Minimalist program. For example, it is not clear why the [u pol] feature of Neg is checked off by the [upol] feature of the n-word or negative marker, and how the [upol] feature of the latter negative elements is checked by the [neg] feature of Neg. Zeijlstra’s (2004, 2008) analysis of the negative concord as syntactic agreement seems more plausible because it is simpler in a sense that it does not require the introduction of an additional feature for the analysis of the changes in the Russian NC structures, and it is more consistent with the current Minimalist account of feature checking and interpretation. Therefore, I use this analysis in my examination of the changes in Russian negative structures. In Modern Standard Russian, both n-words and the negative marker ne have [uneg] that is checked by the negative operator with the [ineg] feature via Multiple Agree: (55) [TP On [NegP Op¬[iNEG] [Neg° ne[uNEG] videl [VP nikogo[uNEG] ]]]] Early Russian exhibits the signs of change in the negative concord structures, and most examples indicate that the feature checking process was similar to that in modern Italian NC structures. The following Italian examples are repeated here (Zeijlstra 2008): (56) Gianni non telefona a nessuno Gianni neg calls to n-body ‘Gianni doesn’t call anybody.’

Italian

(57) [TP Gianni [NegP non[iNEG] telefona [vP a nessuno[uNEG]]]] (58) Ieri nessuno ha telefonato a nessuno Yesterday n-body has called to n-body ‘Yesterday nobody called anybody.’

Italian

(59) [Ieri Op¬[iNEG] [TP nessuno ha telefonato a nessuno]] The feature checking in NC sentences in Early Russian is illustrated below. Examples (31) and (34) are repeated here as (60) and (62): (60) I nikto že prixodil k nim… and no-one part came to them ‘And nobody came to them...’ (61) [NegP Op¬[iNEG] [Neg° nikto že[uNEG] [VP prixodil k nim]]]



Chapter 4.  The negative cycle in Early and Modern Russian 

(62) Slugi že mimo xodašče ne daša jemu ničto že servants part by passed neg gave to-him nothing part ‘The servants that passed by didn’t give him anything.’ (63) [TP Slugi [NegP ne[iNEG] daša [vP jemu ničto[uNEG]]]] The data from the Early Russian texts indicate that the n-words are able to be sole markers of the sentential negation in the preverbal positions (when they function as subjects or objects). It is possible that at some point they were both formally and semantically negative, i.e. they had an interpretable negative feature [iNeg], just like the negative marker ne. Earlier n-words used to be followed by the particle že: nikto že, nichego zhe, and this form was often used in the texts from which the examples in Section 3.2 were taken. It is hard to determine precisely when the feature loss occurred since the earliest Russian texts available to us already show a transition period in the use of negation strategies, i.e. two competing forms of NC are used, and according to my analysis, n-words already had [uneg] feature. In later Russian texts, the instances of the n-words being the sole markers of the sentential negation (features of a Non-Strict NC) become less and less frequent, and in fact, it has been mentioned in the literature on the history of Russian that by the 16th c. such use of n-words is found primarily in written literary texts, while the structures with Strict NC become widely used in colloquial, informal speech (Borkovskij 1978). This colloquial use was taken into consideration during the first language reforms by Peter the Great; therefore, it is safe to assume that NC patterns in the Russian texts of 17th – 18th cc. truly reflected the state of the language at that time. Modern Standard and Non-Standard Russian dialects exhibit the Strict NC behavior. But a few words need to be said with respect to the use of n-words as sole markers of negation in some of the non-standard dialects of Russian. Two scenarios are possible. On the one hand, one may assume that the preverbal negative marker has weakened so much that it wasn’t able to express sentential negation anymore and the nwords regained their semantic negative features and became new markers of negation. On the other hand, a more plausible explanation is that certain regions of Russia preserved the two types of the NC, Strict and Non-Strict, prevalent in Early Russian, and thus they continue to pattern NC structures that exist in Modern Catalan. In summary, the examples from the Early and Modern Russian discussed above show that Russian has undergone a change from a Non-Strict NC to a Strict NC language.

4.2

Emphatic negation in Modern Standard Russian

An additional piece of evidence for the weak status of the negative marker in Modern Russian comes from the frequent use of minimizers in negative sentences. Minimizers are markers of minimal quantity or extent (Vallduví 1994). When minimizers appear in negative contexts, “the negation denotes the absence of a minimal quantity, and hence the presence of no quantity at all” (Horn 1989: 400). Thus, in negative contexts,



Olena Tsurska

minimizers add negative reinforcement. In English, examples of minimizers are a word, an inch, a clue, etc: (64) a. He didn’t say a word. b. The car didn’t move an inch. c. I don’t have a clue. In Modern Standard Russian, several minimizers occur only in negative sentences, namely ni kapli/ni kapel’ki (‘not a drop’) and ni čut’/ ni čutočki (‘not a bit’). Interestingly, the use of these minimizers without ni is ungrammatical, as shown in (66): (65) Mne etogo ni kapli ne xotelos’! to-me this not drop neg wantpast ‘I didn’t want this one bit’

(Russian National Corpus)

(66) *Mne etogo kapli ne xotelos’! (67) U nas net ni kapli somnen’ja… with us neg-be not drop doubt ‘We don’t have a bit of doubt.’

(Russian Internet Corpus)

(68) Četyre tabletki analgina ni čut’ ne pomogli four tablets analgesic not bit neg help ‘Four tablets of the analgesic didn’t help one bit.’ (Russian Internet Corpus) (69) Ja tebe ni čutočki ne verju I to-you not bit neg believe ‘I don’t believe you one little bit.’

(Russian Internet Corpus)

These examples suggest that, in Modern Standard Russian, ne is weak and requires reinforcement from the minimizers. This claim is further supported by the example in (70), in which the negative existential verb net is missing, and the negative meaning of the second clause is rendered by the minimizer ni kapli: (70)

Postojanno odni i te že gnusnyje roži, constantly one and same part disgusting faces v glazax ni kapli intellekta. in eyes not drop intelligence ‘Constantly, there are the same faces; there is not even a bit of intelligence in their eyes.’ (Russian Internet Corpus)

Changes in negative constructions in Russian not only show the shift in the NC patterns, but might also predict the future disappearance or reanalysis of the main negative marker ne in Russian.



Chapter 4.  The negative cycle in Early and Modern Russian 

5. Conclusions In this paper, I have provided evidence that sentential negation in Russian has undergone major changes with respect to the NC patterns. I have shown that Russian has changed from a Non-Strict NC to a Strict NC language. Zeijlstra’s analysis of NC as a case of syntactic agreement allowed me to argue that the changes of the NC type in Russian involved changes in the feature interpretability of Russian negative elements, i.e. the loss of the [ineg] feature of the preverbal negative marker. The examples of negative sentences in Modern Russian listed above have also shown that [uneg] feature of the preverbal negative marker continues to weaken, which might lead to a potential reanalysis, if not disappearance, of ne. This study shows that the history of Russian sentential negation deserves attention and provides interesting data for further studies of the NC changes in other languages.

References A Collection of Russian Corpora. Retrieved 28 February, 2008. http://corpus.leeds.ac.uk/ruscorpora.html. Besten, Hans den. 1986. Double negation and the genesis of Afrikaans. In Substrata versus Universals in Creole Languages. Papers from the Amsterdam Creole Workshop, April 1985 [Creole Language Library 1], Pieter Muysken & Norval Smith (eds), 185–230. Amsterdam: John Benjamins. Borkovskij, Viktor Ivanovič (ed.). 1978. Istoričeskaja Grammatika Russkogo Jazyka. Moscow: Nauka. Brown, Sue. 1999. The Syntax of Negation in Russian: A Minimalist Approach. Stanford CA: Center for the Study of Language and Information. Brown, Sue. 2003. A Minimalist approach to negation in Old Church Slavonic: A look at the Codex Marianus. In Formal Approaches to Slavic Linguistics XI: The Amherst Meeting, Wayles Browne, Ji-Yung Kim, Barbara H. Partee & Robert A. Rothstein (eds), 159–178. Ann Arbor MI: Michigan Slavic Publications. Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.), 1–52. Cambridge MA: The MIT Press Chomsky, Noam. 2004. Beyond explanatory adequacy. In Structure and Beyond, Adriana Belletti (ed.), 104–131. Oxford: OUP. Chomsky, Noam. 2005. Three factors in language design. Linguistic Inquiry 36(1):1–22. Chomsky, Noam. 2007. Approaching UG from below. In Interfaces + Recursion = Language, Uli Sauerland & Hans-Martin Gärtner (eds), 1–29. Berlin: Mouton de Gruyter. Gelderen, Elly van. 2008. Linguistic cycles and economy principle: The role of Universal Grammar in language change. In Grammatical Change and Linguistic Theory: The Rosendal Papers [Linguistik Aktuell-Linguistics Today 113], Thórhallur Eythórsson (ed.), 245–264. Amsterdam: John Benjamins. Gelderen, Elly van. In progress. Linguistic Cycles. ms.

 Olena Tsurska Giannakidou, Anastasia. 1998. Polarity Sensitivity as (Non)veridical Dependency [Linguistik Aktuell-Linguistics Today 23]. Amsterdam: John Benjamins. Giannakidou, Anastasia. 2000. Negative... concord? Natural Language and Linguistic Theory 18: 457–523. Givón, Tom. 1984. Syntax: A Functional-Typological Introduction. Amsterdam: John Benjamins. Gorškova, Ol’ga Vasil’evna & Xmelevskaja, Tatiana Aleksandrovna. 1986. Sbornik Zadač i upražnenij po russkoj Dialektologii. Moscow: Prosveščeniye. Haegeman, Liliane & Lohndal, Terje. Forthcoming. Negative Concord is not Multiple Agree. Hiraiwa, Ken. 2001. Multiple agreement and the defective intervention effect. MITWPL 40: 67–80. Hodge, Carleton. 1970. The linguistic cycle. Linguistic Sciences 13: 1–7. Hoeksema, Jack 1997. Negation and negative concord in Middle Dutch. In Negation and Polarity: Syntax and Semantics [Current Issues in Linguistic Theory 155], Danielle Forget, Paul Hirschbühler, France Martineau & María Luisa Rivero (eds.) 139–158. Amsterdam: John Benjamins. Horn, Lawrence R. 1989. A Natural History of Negation. Chicago IL: The University of Chicago Press. Jespersen, Otto. 1917. Negation in English and Other Languages. Copenhagen: A.F. Høst. Ladusaw, William A. 1992. Expressing negation. In SALT II. Proceedings from the Second Conference on Semantics and Linguistic Theory, Chris Barker & David Dowty (eds.), 237–259. Columbus OH: Ohio State University. Laka, Itziar. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections. PhD dissertation, MIT. Mel’ničenko, G.G. 1985. Xrestomatija po Russkoj Dialektologii. Moscow: Prosveščenije. Press, Ian. 2007. A History of the Russian Language and its Speakers. Munich: Lincom. Press, Ian. 2008. Topics in the History of Russian. Munich: Lincom. Przepiórkowski, Adam & Kupść, Anna. 1997. Negative concord in Polish. Research Report 828 of IPI PAN (Institute of Computer Science, Polish Academy of Sciences). Rechzieglova, Adela. 1995. On Negation in Czech [Slavisticke monografie: Reda lingvisticka 2]. Praha: Euroslavica. Rizzi, Luigi. 1989. Relativised Minimality. Cambridge MA: The MIT Press. Stender-Petersen, Adolf. (ed.) 1954. Anthology of Old Russian Literature. New York NY: Columbia University Press Ura, Hiroyuki. 1996. Multiple Feature-checking: A Theory of Grammatical Function Splitting. PhD dissertation, MIT. Vallduví, Enric. 1994. Polarity items, n-words and minimizers in Catalan and Spanish. Probus 6: 263–294. Zanuttini, Raffaella. 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford: OUP. Zeijlstra, Hedde. 2004. Sentential Negation and Negative Concord. PhD dissertation, University of Amsterdam. Zeijlstra, Hedde. 2008. Negative concord is syntactic agreement. Ms, lingBuzz/000645.

chapter 5

Jespersen off course? The case of contemporary Afrikaans negation*1 Theresa Biberauer

University of Cambridge and Stellenbosch University This paper takes a closer look at how the properties of the Negative Concord (NC) system in contemporary Afrikaans, both the standard and colloquial varieties, relate to Jespersen’s Cycle (JC). It shows that Afrikaans differs from familiar NC systems in respect of the source of its concord item (clause-final nie) and that this difference has various important consequences as far as further JC developments are concerned. Specifically, it is proposed that structural height and deficient featural properties can disqualify a concord element from being reanalysed as a “real” negator. In essence, then, “grammaticalisation beyond the Cycle” is possible. The paper also considers JC-like developments in the domain of quantifier negation, seeking to highlight how these may interact with developments in the domain of sentential/predicate negation.

1. Introduction This paper focuses on the cyclic developments that can be observed in the domain of Afrikaans negation. Afrikaans is standardly cited as an example of an obligatory

* Grateful thanks to the audience at the Conference on Cycles held at Arizona State University (25–26 April 2008), particularly Johan van der Auwera, Elly van Gelderen, Jack Hoeksema, Terje Lohndal and Clifton Pye; to Anne Breitbarth, Chris Lucas and Sonia Cyrino for their stimulating questions following presentation of similar material at the Cambridge SyntaxLab (29 April 2008); to the CGSW23 audience (Edinburgh, 11 June 2008), particularly Doris Penka and Hedde Zeijlstra; and to the audience at DiGSX (Cornell, 8 August 2008), particularly Montse Batllori, Tony Kroch and John Whitman; to Liliane Haegeman and Anne Breitbarth for constructive comments on an earlier version of this text; and, finally, to an anonymous reviewer for an exceptionally helpful review. Thanks too to Elly van Gelderen for her lenient interpretation of the notion ‘deadline’. This work-in-progress is supported by AHRC Grant No. AH/E009239/1 (“Structure and Linearisation in Disharmonic Word Orders”).



Theresa Biberauer

Negative Concord (NC)1 language which differs from its Germanic parent, non-NC Dutch, in requiring not only a negation element (e.g. a sentential negator or some negative quantifier2), but also, necessarily, a clause-final concord element, nie. The relevant difference is illustrated in (1–2): (1) a. Ik ben niet rijk I am not rich ‘I am not rich.’

Dutch

b. Ek is nie1 ryk nie2 I is not rich neg ‘I am not rich.’ (≠ ‘I am not not rich)

Afrikaans

(2) a. Zij hebben nooit een auto gehad they have never a car had ‘They never had a car.’

Dutch

b. Hulle het nooit ‘n motor gehad nie2 they have never a car had not ‘They never had a car.’

Afrikaans

If we – rather simplistically, as will become clear below – divide up Jespersen’s Cycle (JC; cf. Jespersen 1917) as indicated in (3), Afrikaans would, then, seem to be a Stage III language, while its parent is synchronically stage I, i.e. a Double Negation language:3 (3) I.

neg1 … II. dn

neg1 … (neg2) III. neg1 … neg2 Optional nc Obligatory nc

IV.

neg3 dn

What this paper aims to show is, first, that the characterisation of Afrikaans as a Stage III language is only partially correct; secondly, that modern Afrikaans does not, as is sometimes assumed, appear to be undergoing the “next step” in JC by developing into a Stage IV language, but that also does not mean that it is “inert” in the context of this cycle; and finally, that there appear to be good syntactic reasons why Afrikaans is behaving as it is. Our central contention, then, is that Afrikaans, for readily identifiable 1. Abbreviations used in this chapter are the following: NC = Negative Concord; DN = Double Negation (i.e. non-NC); JC = Jespersen’s Cycle; NEG = negative concord element; SA = Standard Afrikaans; CA = Colloquial Afrikaans; NPI = negative polarity item; 2. We avoid the term n-word, which is very commonly used to refer to negative quantifiers in the context of NC languages (cf. Laka 1990/1994, Giannakidou 2000, 2005, Herburger 2001 and Zeijlstra 2004). The reasons for this will become clear in Section 4. 3. Roberge (2000) shows that the varieties of 17th century Dutch spoken by settlers at the Cape were also Stage II varieties, lacking the NC possibilities available in earlier Dutch (cf. also Ponelis 1993 and Zeijlstra 2004). As such, it does not seem plausible to assume that Afrikaans’s NC property, which only emerged strongly during the late 19th century, in fact originated in one of the 17th century dialects introduced to the Cape. We return to this point in Section 4.



Chapter 5.  Jespersen off course 

reasons, differs in JC terms from better-studied languages and that its future progression in respect of the (simplified) schema in (3) can be schematised in the manner given below: (4) a. “Traditional” JC expectation for Afrikaans: Stage III → Stage IV, i.e. nie1 will become optional, ultimately to be replaced by nie2 Cf. French: ne … pas → (ne) … pas → pas Std French Colloquial French English: ne …(noht) → (ne) … no(h)t → not Old English Modern English *Afrikaans: nie1 … nie2 → (nie1) … nie2 → nie2 Modern Future Afrikaans Afrikaans b. Actual JC development in Afrikaans: nie1 → nie1 … (nie2) → nie1 … nie2 Dutch Early Afrikaans Modern Afrikaans The paper is organised as follows: Section 2 outlines the relevant background on Afrikaans negation; Section 3 focuses on current developments affecting the language’s concord element (nie2 in (1) and (2) above), while Section 4 considers developments affecting other negative elements; Section 5 then suggests a (partial) generative analysis of what appear to be the core developments, with Section 6 concluding by spelling out some predictions and extensions of the account offered here.

2. Afrikaans negation: The basic facts Arguably the most well-known fact about Afrikaans negation is the fact that it requires two identical negators in the absence of specific negative quantifiers/indefinites – cf. (1) above, which clearly illustrates this property. These two nies are, however, rather different, a fact that has also frequently been observed in both the generative and the descriptive literature (cf. Waher 1978, Robbers 1992, 1997, Ponelis 1993, Donaldson 1993, Oosthuizen 1998, Bell 2004a, and Biberauer 2007, 2008a; see, however, Molnárfi 2002 for a contrary view in terms of which these elements are viewed as “multiple spellouts” of a single lexical item). Specifically, the difference is that while nie1 is a “real” negator, contributing negative semantics to structures in which it appears, nie2 is a concord element, which does not contribute its own negative meaning (we will return to the matter of the features associated with Afrikaans negation elements in



Theresa Biberauer

Section 5.1 below).4 The difference between the two elements emerges clearly when we consider their behaviour in respect of: a. omissibility b. modifiability c. reinforcement; and d. stressability (cf. Biberauer 2007 and 2008a for more detailed discussion) Taking omissibility first, we see that omitting nie1 results in ungrammaticality (5), whereas omission of nie2 just sounds as if it is was forgotten (6a), being a particularly common feature of L2 Afrikaans, but also surfacing in colloquial L1 Afrikaans; further, it is also an obligatory feature of “headlinese” (6b):5 (5) a. Hy kom nie1 in nie2 he come not in not ‘He doesn’t come in/He isn’t coming in.’ b. *Hy kom in nie2 he come in not (6) a. Hy kom nie1 in he come not in ‘He doesn’t come in/He isn’t coming in.’

4. Worth noting here is that the syntactic properties of the two nies in Afrikaans may be distinct, but that this need not be true of their semantic properties. Thus Biberauer & Zeijlstra (2009), for example, argue that both nies in standard Afrikaans are in fact semantically non-negative in the way that negative markers in strict NC languages more generally are (cf. Zeijlstra 2004 and following). The syntactic and phonological differences between the two nies then follow from the fact that these semantically identical elements are inserted into different syntactic slots (a lower vP-related adjunct slot in the case of nie1 and a left-peripheral head slot in the case of nie2). 5. The extent to which nie2 is an entrenched feature of L1 Afrikaans generally, irrespective of the variety concerned, is clearly illustrated by the fact that this element still features prominently even in the matrilectal variety that has undoubtedly been most affected by English, socalled Kaaps spoken in the Cape by people of mixed-race origin (cf. Roberge 2002 and McCormick 2002). This is clearly illustrated in the following attested utterance: (i) Õs praatie yntlik regte Afrikaans nie; Õs praat Kaapse Afrikaans, us speak-not actually real Afrikaans neg us speak Cape Afrikaans, Engels en Afrikaans deu’mekaar. English and Afrikaans mixed-up ‘We don’t actually speak real Afrikaans; we speak Kaaps, English and Afrikaans all mixed up’ [source: Dyers, C. 2007. An investigation into the role of the Afrikaans language in indexing the individual and collective self among some township youth in South Africa. Uluslararası Sosyal Aratırmalar Dergisi The Journal Of International Social Research 1(1): 83–101]



Chapter 5.  Jespersen off course 

b. See ‘is nie verkoop’ Sea is not sold ‘Sea wasn’t sold.’

(Die Burger – 26/08/2008)

Comparison of the modification possibilities exhibited by the two nies reveals that nie1 can be modified in various ways (7); this is not possible with nie2. (7) a.

Jy let glad/ hoegenaamd/ absoluut/ geheel en al nie1 you attend altogether/ at-all/ absolutely/ whole and all not op nie2 up neg ‘You aren’t remotely paying attention’

b. *Jy let nie1 op glad/hoegenaamd/absoluut/geheel en al nie2 Further, nie1 can also be replaced by a more emphatic negator, in contrast to nie2. (8) illustrates the type of reinforcement one might encounter in formal registers, while (9) is typical of spoken registers (we return to structures of this type in Section 4 below): (8) a. Ons is nie1 beïndruk nie2 us is not impressed not ‘We are not impressed.’ b. Ons is geensins beïndruk nie2 us is not-remotely impressed not ‘We are not remotely impressed.’ c. *Ons is nie1 beïndruk geensins (9) a. Hy wil nie1 luister nie2 he will not listen not ‘He won’t listen.’ b. Hy wil g’n niks luister nie2 he will no(ne) nothing listen not ‘He won’t listen at all.’ c. *Hy wil nie1 luister g’n niks Finally, as (10) shows, nie1 differs from nie2 in terms of stressability: (10) a. Ek weet nooit/nie1 wat hy doen nie2 I know never not what he do neg ‘I never know what he’s doing/I don’t know what he’s doing/he does.’ b. *Ek weet nooit/nie1 wat hy doen nie2 The differences between the two nies are summarised in Table 1:



Theresa Biberauer

Table 1.  Summary comparison of the properties of nie1 and nie2 in Afrikaans Property

nie1

nie2

1. Omission → ungrammaticality 2. Modifiability 3. Substitution by emphatic negator 4. Stressability

Yes Yes Yes Yes

No No No No

What we see, then, is that nie1 and the position associated with the “real” negator can be strengthened in various ways, whereas nie2 cannot. In the context of JC, this is particularly significant as it unambiguously indicates that the “real” negator in Afrikaans, i.e. neg1 in (3), is not a weak element (cf. Cardinaletti & Starke 1999 for discussion of strong vs weak elements). The significance of this fact becomes clear if we consider Jespersen’s own description of the Cycle that now bears his name: (11) “The history of negative expressions in various languages makes us witness the following curious fluctuation: the original negative adverb is first weakened [my emphasis – TB], then found insufficient and therefore strengthened, generally through some additional word, and in its turn may be felt as the negative proper and may then in course of time be subject to the same development as the original word.” (Jespersen 1917: 4) According to Jespersen, then, introduction of nie2 should have been preceded by weakening of nie1, whereafter nie1 should remain weak, with nie2 progressively assuming the role of “real” negator. In reality, however, it appears to be nie2 that is weak, with nie1 behaving in a way that resembles the single sentential negators in other languages. Consider the comparison with English given in (12): (12) a. Omission: leads to a change in meaning – cf. I will (not) give up running b. Modifiability: possible in both formal – cf. I will absolutely/definitely not give up running – and informal registers – cf. I will so not give up running c. Substitution by emphatic negator: once again possibly in both registers – cf. I will not remotely/never/not on your life/no ways give up running! d. Stressability – I will not give up running Taking this into account, we see that Afrikaans is not a typical Stage III language in Jespersonian terms: functionally, nie2 is clearly the new negator, just as English noht, French pas and Dutch niet were at one stage in the history of those languages; formally, however, nie2 more closely resembles the original negators in the other languages – i.e. ne in earlier English and Modern French, and ne/en in Middle Dutch – all of which undeniably underwent weakening during the course of their history (although see Kiparsky & Condoravdi 2006 on the question of cause-and-effect here).



Chapter 5.  Jespersen off course 

Given that Afrikaans is clearly different in respect of the formal properties associated with its negators, it is justified to ask whether this language will proceed to the next stage of JC in the way that familiar Western European languages have. If phonological weakening of the original negator is a necessary prerequisite for this progression, the answer would appear to be no: even if – and there is no evidence to suggest this – nie1 had at some point in its prior history undergone phonological weakening so as to create the appropriate conditions for the introduction of nie2, it is clear that nie1 is synchronically strong, while nie2 is currently weak. From a phonological perspective, then, there would appear to be no motivation for progression to Stage IV. Alternatively, it may be the case that semantico-pragmatic rather than phonological considerations drive the weakening of the initial negator, leading to the possibility of its ultimate replacement by an originally reinforcing negation element (cf. Dahl 2001, Schwenter 2002, 2005, 2006, Kiparsky & Condoravdi 2006 and Condoravdi 2007 for discussion of this possibility). On this view, then, semantic bleaching is a crucial ingredient in the move from Stage II to III. For Kiparsky & Condoravdi, an originally emphatic optional NC negation pattern at Stage II becomes the neutral (i.e. obligatory) one at Stage III, with the functional consideration that languages must supply their users with a means to express the distinction between neutral and emphatic negation then triggering further developments, which may include the loss of the original negation and the rise of a new reinforcer; the possibility of retaining the original negation and adding a further reinforcer is, however, not excluded (cf. also van der Auwera this volume, 2009). Schwenter (2002, 2005, 2006), in turn, points out that the function of the originally reinforcing negator can also be to distinguish negated clauses on the basis of their discourse status (e.g. those expressing discourse-old or inferable propositions vs those which do not). Once again, subsequent bleaching, leading from Stage II to Stage III, is a possibility on this view, although it need not occur: the newly introduced concord element may simply retain its specific semantic/pragmatic function, enabling the language in question to systematically distinguish different negation functions that are not explicitly encoded in all languages (cf. Miestamo 2005 and van der Auwera 2009 for overview typological discussion). Importantly, then, semanticopragmatic accounts of the course of JC highlight various points at which the cycle may “stall”: continuation to Stages III and IV depends on speakers’ lexical choices in respect of the elements they employ in neutral vs emphatic contexts. Looking at standard Afrikaans from this type of perspective, it would seem that nie2, as an obligatory element6, and moreover a phonologically weak one, is ill-suited for reanalysis as an element serving emphatic or discourse-sensitive functions. The reality is, however, somewhat different. As we shall see below, nie2 is not as uninvolved in emphatic structures as one might naturally suppose. At the same time, it is, however, 6. Cf. Reinhart (2006), Fox (2000), Dahl (2001), Kiparsky & Condoravdi (2006) and Biberauer & Richards (2006) for discussion of the inverse relationship between obligatoriness and “extra/special interpretations”.



Theresa Biberauer

clear that nie1’s phonological strength has not rendered it immune to the kind of strengthening that JC leads us to expect in respect of “real” negators; strikingly, however, the observed strengthening does not involve nie2, i.e. Afrikaans shows no signs of progressing to Stage IV. In the following sections, we will consider two types of synchronic variation in the domain of Afrikaans negation in order to establish more precisely what course this language appears to be on in the context of JC. Section 3 specifically considers nie2related developments, whereas Section 4 focuses on developments in the domain of negative quantifiers.

3. Current developments I: The distribution of nie2 In the preceding section, we characterised nie2 as an element which obligatorily surfaces in all negation structures in standard Afrikaans, thereby underlining Afrikaans’s status as a Stage III language. Before we consider the behaviour of this element in more detail, it is worth noting an important context in which nie2 consistently fails to surface. Consider (13): (13) a. Hy kom waarskynlik nie1 (*nie2) he come probably not neg ‘He probably isn’t coming’ b. Sy ken hom nie1 (*nie2) she know him not neg ‘She doesn’t know him’ c. Wie ken sy nie1 (*nie2)? who know she not neg ‘Who doesn’t she know?’ In each of these structures, then, nie2 is barred. That it is indeed nie1 which occupies clause-final position in these examples is clear from the properties of this final nie: as (14) shows, it cannot be omitted without resulting in meaning change (i.e. polarity reversal) and it can be modified, reinforced and stressed: (14) a. Hy kom waarskynlik he come probably ‘He is probably coming’ b. Hy kom beslis nie1 he come definitely not ‘He is definitely not coming’



Chapter 5.  Jespersen off course 

c. Hy kom geensins nie2 he come not-remotely neg ‘He is not remotely coming’ d. Hy kom nie1 he come not ‘He is not coming.’ The question that now arises is whether examples such as those in (14) compromise Afrikaans’s status as a Stage III NC language? Biberauer (2007, 2008a) argues that they do not. If we consider the embedded clauses corresponding to the structures in (14), we see that these each contain the expected two nies: (15) a. Ek dink dat hy waarskynlik nie1 kom nie2 I think that he probably not come neg ‘I think that he probably isn’t coming.’ b. Ek dink dat sy hom nie1 ken nie2 I think that she him not know neg ‘I think that she doesn’t know him.’ c. Ek weet wie sy nie1 ken nie2 I know who she not know neg ‘I know who she doesn’t know.’ Against this background, Biberauer (op. cit.) follows den Besten (1986) and Robbers (1992) in proposing that (14)-type structures do not differ from their embedded counterparts in featuring two nies; they simply differ from the latter structures in involving a configuration which leads to deletion of the second nie (nie2). Den Besten (1986: 202) specifically postulates the filter in (16): (16) Doubly Filled Neg Filter * nie nie

(cf. den Besten 1986: 202; cf. also Robbers 1992):

For den Besten, then, the second in a sequence of two nies will necessarily be deleted. This is, however, too strong, as the following data show: (17) a. Hy kom nie1 nie1 nie2 he come not not neg ‘He’s not not coming’, i.e. he is coming b. Ek sien nie1 die tweede ‘nie’ nie2 I see not the second nie not ‘I don’t/can’t see the second nie.’

 Theresa Biberauer

In light of (17), Biberauer (2007, 2008a) proposes a haplology mechanism in terms of which the configuration in (18) is ruled out, while that in (19) is ruled in: (18) [φ… nie nie2] → nie nie2 (19) [φ… nie ] [nie2 …] In terms of Biberauer’s system, then, nie2 will undergo deletion at PF wherever it is spelled out adjacent to another nie located in the same phonological phrase, whereas it will survive deletion wherever adjacent nies are not assigned to the same phonological phrase. The generally accepted view that Afrikaans is a Stage III language is therefore not undermined by examples in which nie2 fails to surface: this element is always present during the syntactic derivation of the structures in question, its ultimate absence being a consequence of post-syntactic deletion at PF. With this much in place, we can now compare the distribution of nie2 in standard and colloquial Afrikaans (henceforth SA and CA respectively). As established above, clause-final nie2 is obligatory in SA clauses containing a negated element, regardless of whether this is the sentential negator nie1 or a specific negative quantifier. This is also true in CA, although native-speakers do occasionally “leave out” the final nie, a phenomenon which appears to be a consequence of performance factors as it tends to occur at the ends of particularly long sentences and/or where speakers are evidently rushed in their delivery. Final nie is, however, undoubtedly mostly present in CA and speakers are certainly not consistent about always leaving nie2 out in specific clausal contexts; therefore it seems justified to conclude that this variety does not differ from the standard in respect of the realisation of nie2 in clausal domains. In non-clausal domains, i.e. where negation has narrow scope (so-called constituent negation), the picture is strikingly different, however. There only appear to be two non-clausal contexts in which SA systematically requires two nies, and these are illustrated in (20) (see note 4 for an apparently idiosyncratic double-nie-containing form): (20) a. Nie1 die geld nie2, maar die tyd pla hom not the money neg but the time worry him ‘Not the money, but the time worries him.’ b. A: Wie het my boek gesien? who have my book seen ‘Who has seen my book?’ B. Niemand nie2 no-one neg ‘No-one.’ In (20a), a clearly non-clausal focused constituent which has been negated surfaces with two nies and in (20b), we see that negative quantifiers in fragment answers also prescriptively require nie2. Leaving aside the possibility that negative fragment answers in fact entail elided full clauses (pace Giannakidou 2000, 2005; cf. Watanabe



Chapter 5.  Jespersen off course 

2004 and Zeijlstra 2008 for discussion of the non-trivial problems associated with this type of analysis), it is clear that nie2 in SA is available in restricted constituent negation contexts. Generally, however, negated constituents are not independently associated with their own nie2. Thus the structures in (21) are both impossible in SA: (21) a. *Dit moet nie1 langer nie2 as 10 000 woorde wees nie2 it must not longer neg than 10 000 words be neg ‘It mustn’t be any longer than 10 000 words.’ [correct: Dit moet nie/nie langer as 10 000 woorde wees nie2] b. *Sy skryf nooit nie2 so baie nie2 she write never neg so much neg ‘She never writes so much.’ [correct: Sy skryf nóóit so baie nie2] As the prescriptively correct forms given beneath each example show, SA places additional emphasis on the required negation element rather than employing an “extra” nie2. In CA, by contrast, nie2 is productively used in these contexts, with both examples in (21) constituting well-formed CA structures.7 Ponelis (1993:  454) suggests that structures like (21b) only occur “very infrequently” and only in “highly informal style”, but this does not appear to be the case in CA today as all the informants I consulted 7. Native-speakers I consulted concur that clause-final nie2 may readily be dropped in structures of the type illustrated in (21a) and some (including myself) even feel that the structure lacking final nie2 sounds more natural than the alternative. The same informants did not have the same judgements about (21b), though: here final nie2 was universally regarded as the form that speakers would ordinarily use. This difference, the details of which we leave aside for future research, is interesting, particularly in view of the fact that there appears to be an idiosyncratic comparative-containing form that informants view as “perfectly standard” which requires nie2 to surface clause-internally: as (ii) shows, nie minder nie appears to be a fixed emphatic comparative which is somehow “sealed off ” from the rest of the clause, thereby precluding the need for the otherwise prescriptively obligatory clause-final nie2: (i) Moeder Natuur het vir nie1 minder nie2 as drie beskermende lae Mother Nature have for no less neg than three protective layers gesorg cared ‘Mother Nature provided no less than three protective layers’. (cf. Donaldson 1993: 410)

(ii) *Moeder Natuur het vir nie1 minder as drie beskermende lae gesorg nie2 As noted by an anonymous reviewer, the fact that a particular comparative structure ((i) above) should already be readily accepted by some native-speakers and that clause-internal nie2 is more generally a feature of negative comparatives in CA is also interesting if we look beyond Afrikaans since comparative contexts also seem to be a good host for so-called expletive negation (cf. Horn 1989, Espinal 1992, 2000, 2007 and Rooryck 2008). Worth noting in this connection is that positive comparatives do not permit “expletive” nie2 – cf. (iii) – although exceptive clauses and those containing approximative adverbs in CA, for example, appear to do so optionally– cf. (iv-v):

 Theresa Biberauer

readily accepted a wide range of examples featuring this “extra” nie2. Some of the available possibilities are illustrated in (22):8 (22) a. Ek sou nooit nie2 so iets doen nie2 I would never not such something do neg ‘I would never do that kind of thing.’ b. Sy het nêrens nie2 tuis gevoel nie2 she have nowhere neg home felt neg ‘There was nowhere she felt at home, i.e. she didn’t feel at home anywhere.’ (iii) Hy is langer as ek (*nie2) he is taller than I neg ‘He is taller than I am.’ (iv) Hy vertrek sonder dat ek agterkom (nie2) he leaves without that I realise neg ‘He leaves without me realising it.’ (v) Hy kon nouliks staan (nie2) he could barely stand neg ‘He could barely stand.’ Whether (iv-v), or, in fact, the comparative-containing examples cited above actually represent instances of expletive negation (assuming this to be a coherent notion) must, however, be open to doubt. This is because core expletive negation contexts identified by earlier researchers feature not only an expletive nie2, but also an extraneous nie1 in Afrikaans (cf. also Weiss 2002: 323ff on Bavarian ned). Consider (vi): (vi) Hoe gelukkig was ons nie1 gewees nie2! how lucky was us not been neg ‘How lucky we were!’ (vii) Totdat jy nie1 klaar is nie2, bly jy sit! until you not finished are neg stay you sit ‘You stay sitting until you’ve finished!’ (viii) Voordat jy nie1 ophou huil nie2, kry jy nie1 kos nie2 before you not stop cry neg get you not food neg ‘Until you stop crying, you won’t get any food.’  e leave to future research the question of how “extra” nie2-containing structures like (i) and (ivW v) relate to those, which, like (vi), feature both an “extra” concord element and an “extra” nie1. 8. Note that application of the differentiating criteria discussed in Section 2 make it clear in each case that we are dealing with an additional nie2 and not an extra nie1. Thus each of the nies under consideration cannot be modified, replaced by stronger forms or stressed, whereas they can – and standardly are – omitted. Also worth noting in connection with these structures is that they appear to be of the same type found in certain Dutch dialects, notably Aarschots, which has often been viewed as the source for Afrikaans nie2 (cf. Pauwels 1958, but see den Besten 1986 and Roberge 2000 for discussion of the problems associated with this hypothesis). See also Aelbrecht (2006) for recent discussion of what appears to be the same phenomenon in Asse Dutch.



Chapter 5.  Jespersen off course 

c. Hy het niks nie2 geld nie2 he have nothing not money not ‘He has absolutely no money.’ d. Hulle gee niemand nie2 ‘n kans nie2 they give no-one neg a chance neg ‘They don’t give anyone a chance.’ e. Geen mens nie2 kan dit verstaan nie2 no person neg can it understand neg ‘Not a soul can understand that.’ f. Hy het niks anders nie2 as ‘n Mercedes gery nie29 he have nothing other neg than a Mercedes driven neg ‘He drove nothing other than a Mercedes.’ As the examples show, both simple and complex expressions containing a negative quantifier can be rendered emphatic by the presence of an extra nie2. (22f) is particularly interesting as native-speakers agree that this example is considerably less informal than the other examples in (22), a matter which we leave aside here (although see also note 7 – it seems that comparative(-like) forms involving negatives are particularly compatible with “extra” nie2). At the opposite end of the spectrum, it is also worth noting that negative quantifier subjects – aside from geen mens nie2 as in (22e), which may be an idiosyncratic lexicalisation (cf. also nie minder nie, niks anders nie) – appear

9. Note that Donaldson (1993: 410) cites this example without clause-final nie2. Consultation with native-speakers, however, confirms my own judgement that nie2 is possible here. This judgement is further reinforced by a google search which produced structures such as the following, both, strikingly, drawn from academic texts: (i) Die vermeerdering van kennis is niks anders nie2 as die the increase of knowledge is nothing other neg than the natuurlike seleksie van hipoteses nie2 natural selection of hypotheses neg ‘Increasing knowledge is nothing other than the natural selection of hypotheses.’ [source: Snyman, B. 1995. Wetenskapbeelde in die Geesteswetenskappe. Pretoria: hsrc Press, p.42] (ii) Streng etimologies beteken ‘metodologie’ niks anders nie2 as die strictly etymologically mean ‘methodology’ nothing other neg than the logika onderliggend aan die implementering van wetenskaplike metodes in die logic underlying to the implementation of scientific methods in the bestudering van die werklikheid study of the reality ‘Strictly etymologically, ‘methodology’ means nothing other than the logic underlying the implementation of scientifc methods in the study of reality’ [source: Mouton, J. and H. Marais 1988. Basiese Begrippe. Pretoria: HSRC Press, p16]

 Theresa Biberauer

to be maximally incompatible with “extra” nie2, regardless of register. This fact, which we once again leave aside here10, is illustrated in (23): (23) *Niemand nie2 ken die antwoord nie2 no-one neg know the answer neg ≠ ‘No-one knows the answer.’ To summarise, then: CA makes systematic use of nie2 in a non-clausal context that is not generally available in SA, namely that in which specific negated constituents are emphasised. The use of nie2 in emphatic contexts contrasts with its absence in neutral contexts: wherever clause-internal nie2 is deleted in (22), a less emphatic interpretation results. Strikingly, nie2 is also generally dropped in negative fragment answers in CA (cf. also Donaldson 1993: 409–410), with the equivalent of (20b) in the spoken language being viewed by many speakers as an emphatic fragment answer. To generalise, then, whereas nie2 is obligatory in the clausal domain in both CA and SA, it is, to varying degrees optional in the non-clausal domain (cf. Biberauer 2008b for further discussion of the diachronic aspects of this difference). For both SA and CA, an acrossthe-board characterisation of the system in question in JC terms therefore seems impossible: while the clausal domain has ostensibly reached Stage III (at least insofar as this stage entails the obligatory presence of two negation elements delivering NC structures), the non-clausal domain is very evidently at Stage II (optional NC, with the optional negator therefore delivering an “extra” interpretive effect in accordance with our expectations regarding optionality). This hybrid state of affairs has not, to the best of my knowledge, been explicitly noted before. It is clearly not compatible with the idealised JC schema in (3), and even more articulated divisions of JC than that presented in (3) (cf. Jespersen 1917 and also recent discussion in Zeijlstra 2004 and Willis 2008) do not, also as far as I am aware, accommodate a language exhibiting the properties of either SA or CA. Partially, this is unsurprising as JC focuses on sentential/predicate negation and the developments discussed in this section are clearly affecting quantifier negation. Quantifier-negation de10. The disparity between pre- and postverbal negative quantifiers and, specifically, the fact that subjects – which, at least in subject-initial V2 clauses, differ from non-subjects in being preverbal – behave differently from negative quantifiers located postverbally immediately calls to mind the very well-known discrepancies between pre- and postverbal negative quantifiers in so-called non-strict NC languages (cf. Giannakidou 2000, 2005, Herburger 2001, and Zeijlstra 2004, 2008 for recent discussion). Linear notions are, however, clearly irrelevant in the Afrikaans case, as the following example shows: (i) Gister het niemand (*nie2) my gehelp nie2 yesterday have no-one neg me helped neg ‘Yesterday no-one helped me’ Regardless of whether a negatively quantified subject surfaces pre- or postverbally, then, it seems to be more resistant to reinforcement via an “extra” nie2 (cf. Biberauer 2008b for further discussion of the potentially misleading nature of linear considerations in discussions of JC).



Chapter 5.  Jespersen off course 

velopments are not, however, unconnected to those in the domain of predicate negation: firstly, negative quantifiers are a common source for reinforcing negators (cf. English noht and Dutch niet) and further, there is also a sense in which quantifier negation can undergo a “shortened JC” (cf. van der Auwera & Neuckermans 2004: 458ff and van der Auwera et al. 2006: 461; in (24), we abstract away from the precise meaning of the negation-related elements at Stage II/III ): (24) a. Sentential/predicate negation I. je ne dis → II/III. je ne dis pas → IV. je dis pas I not speak I not speak not I speak not b. Quantifier negation I. --- → II/III. je ne dis rien → IV. je dis rien I not speak nothing I speak nothing As (24) clearly shows, changes in the sentential negation system – specifically, the loss of NC that occurs when a language moves from Stage III to Stage IV – can correlate with changes in the quantifier negation system – here, reanalysis of a negative polarity item (NPI) requiring licensing by a concord element as an inherently negative quantifier. SA and CA, however, highlight what appear to be independent changes in sentential and quantifier negation, resulting in a “mix of stages” of the type that at the very least suggests that it might be useful to think not just of related JCs of the kind illustrated in (24), but of genuinely independent or multiple JCs which are simultaneously underway. Further, the SA and CA “mix of stages” also highlights another potentially very significant point that we have already touched on in the course of our discussion of the properties of neg1 and neg2 in Afrikaans (cf. Section 2 above), namely that all Stage III languages may not be formally identical; just as we have seen in generative investigations of other areas – e.g. word order, where, for example, distinguishing between the different types of SVO structures found in English, French and German facilitates an otherwise unavailable understanding of other, superficially unrelated and non-overlapping properties in these languages – it may prove productive to consider whether the superficially very similar properties observed in Stage III languages actually follow from underlyingly identical systems. The following section considers further evidence that suggests that SA and CA are rather different Stage III systems from those instantiated by the best-studied Western European languages.

4. Current developments II: Non-nie2-related reinforcement patterns Having considered nie2-related reinforcement developments in the previous section, the focus of this section will be on two major types of non-nie2-related reinforcement patterns, one relating to lexical substitution possibilities and the other to negative quantifiers.

 Theresa Biberauer

4.1

Negative quantifier-related developments

Before we consider the quantifier-related development, it is important to clarify the behaviour of negative quantifiers in SA. The standard expectation for NC languages is that negative quantifiers will differ from those in non-NC languages in exhibiting the behaviour standardly ascribed to so-called n-words (cf. Laka 1990, 1994). Giannakidou (2005) characterises an n-word as follows: (25) N-word: An expression α is an n-word iff: a. α can be used in structures containing sentential negation or another α-expression yielding a reading equivalent to one logical negation; and b. α can provide a negative fragment answer. We have already seen that negative quantifiers in SA do not satisfy (25b) (cf. (20b) above), although those in CA do, to the extent that the presence of nie2 necessarily brings with it an emphatic (i.e. additional negative) interpretation. Crucially, however, neither SA nor CA satisfy (25a). This is illustrated for sentential negation involving nie1 in (26) and for co-occurring negative quantifiers in (27): (26) a. Niemand is nie1 hier nie2 no-one is not here neg ‘No-one isn’t here’, i.e. everyone is here. b. Sy het nie1 niks gelees nie2 she have not nothing read neg ‘She didn’t read nothing’, i.e. she read something. c. Sy het niks nie1 gelees nie2 she have nothing not read neg ‘There is nothing she did not read.’ (27) a.

Niemand sien niks nie2 no-one see nothing neg ‘No-one sees nothing’, i.e. everyone sees something. versus [Niemand sien enige iets nie2 = lit: ‘No-one sees anything’]

b.

Hy sien nêrens geen probleme nie2 he see nowhere no problems neg ‘Nowhere does he see no problems’, i.e. he sees problems everywhere. versus [Hy sien nêrens enige probleme nie2 = lit. ‘He sees nowhere any problems neg’, i.e. he doesn’t see any problems anywhere]

c.

Hy gee nooit niks nie2 he give never nothing neg ‘He never gives nothing’, i.e. he always gives something. versus [Hy gee nooit iets nie2 = lit. ‘He gives never (some)thing neg’, i.e. he never gives anything]



Chapter 5.  Jespersen off course 

As the above examples show, then, negative quantifiers in SA do not exhibit NC relations either with sentential negators or with each other; each negative element word contributes its own negative semantics, just as negative quantifiers/indefinites do in standard Germanic more generally (cf. also van der Wouden 1994). This is in stark contrast to other NC varieties, where negative quantifiers do generally (with the exception of negative subjects in non-strict NCs – cf. note 10) behave in the manner outlined in (25): (28) a. I don’t owe no-one nothing = I don’t owe anyone anything

Variety of English

b. Nobody saw nothing = no-one saw anything (contrast Afrikaans (27a)) c. Personne (n’) a rien fait no-one neg has nothing done ‘No-one has done anything’ d. Sun na ka pô fe nadaxi fa He not asp can do nothing neg ‘He couldn’t do anything’

French

Santome (cf. Hagemeijer 2007: 192)

Given this SA behaviour, I will continue to avoid the label n-word (cf. note 2 above), referring to negative quantifiers instead. While negative quantifiers in SA seem to behave like genuinely negative elements rather than like NPIs, the way n-words do, the picture for CA is a little different: in addition to the DN structures in (27), this variety also permits negative quantifiers to surface in NC structures, a phenomenon known as negative spread in the literature (cf. Molnárfi 2004, van Gass 2007 for recent discussion of Afrikaans; cf. also Haegeman & Lohndal 2008 on West Flemish): (29) a.

Niemand sien niks nie2 no-one see nothing neg ‘No-one sees anything.’ [=sa ‘No-one sees nothing’, i.e. everyone sees something; cf. (27a) above]

b.

Jy vertel my ook nooit niks nie2 you tell me also never nothing neg ‘You also never tell me anything.’ [=sa ‘You also never tell me nothing’, i.e. you’re always telling me something; cf. (27c) above]

(30) a. Ons is nog nooit nêrens betrap nie2 us is still never nowhere trapped neg ‘We have never been trapped anywhere.’

(Volksblad 2002–01–19)

 Theresa Biberauer

b. En bowe-al futiel omdat dit niemand nêrens bring nie2 and above-all futile because it nobody nowhere brings neg ‘And above-all futile, because it gets nobody anywhere.’ (Beeld 1997–02–07) Ek het nog niks nêrens geteken nie2, maar my kontrak met I have still nothing nowhere signed neg but my contract with Natal verstryk begin Januarie Natal expire beginning January ‘I haven’t signed anything anywhere yet, but my contract with Natal ends at the beginning of January.’ (Rapport 1999–01–03) (all examples in (30) drawn from van Gass 2007) c.

Some speakers judge structures like (29–30) to have an emphatic flavour, although there is disagreement on this point and it is certainly clear that these structures do not require specific phonological marking (cf. Schwenter 2002, 2005, 2006 on the opacity surrounding what is meant by emphasis in cases of this type).11 CA does, however, feature double negative quantifier-containing structures that all speakers agree to be emphatic (in the sense of “stronger than the corresponding SA form”) – see (31):12 (31) a.

Hy wil (g’n) niks luister nie2 he want no nothing listen neg ‘He won’t listen at all.’ [cf. Colloquial English: He won’t do no nothing]

b. Jy gaan met hierdie ding (g’n) nêrens kom nie2 you go with this thing no nowhere come neg ‘You won’t get anywhere at all with this thing.’ c. “Teen aanstaande Mei gaan min groot Amerikaanse maatskappye against next May go few large American companies teenwoordig wees in Suid-Afrika en oor twee jaar gaan geen present be in South Africa and over two year go no niemand meer daar wees nie2,” het mnr. Patrick McVeigh nobody more there be nie has Mr Patrick McVeigh

11. Impressionistically, it seems that double negative quantifier expressions are so entrenched in Kaaps, a variety of Afrikaans spoken in the Cape (cf. McCormick 2002), that they are interpreted neutrally, with the prescriptively correct NPI-containing forms only being employed rather rarely. This is, however, a matter requiring more detailed investigation. 12. Crucially, these examples indicate that so-called Emphatic Negation – where one negative element reinforces another negative element, resulting in a negative interpretation that is stronger than it would have been in the absence of the second negative element (cf. Zeijlstra 2004: 58) – is in fact available in NC languages (pace Zeijlstra 2004, 2006).



Chapter 5.  Jespersen off course 

gister gesê yesterday said ‘“By the end of next May there will be very few large American companies present in South Africa, and in two years time there will be no-one at all there anymore,” said Mr Patrick McVeigh yesterday.’ (Die Burger 1986–10–24)

The above examples show clearly that g’n and, less commonly, geen can combine with a range of negative quantifiers, i.e. double negative-containing structures of this type are productive, with g’n niks being particularly so (cf. van Gass 2007 for corpus-based comparative figures).13 The origins of g’n can fairly clearly be related to usages of the type illustrated in (32), which are common in CA: (32) Ek het geen stokperdjies, geen meisies, geen niks nie2 I have no hobbies no girlfriends no nothing neg ‘I have no hobbies, no girlfriends, nothing.’ (Die Burger 1998–05–11) As the orthographic representations in (31a,b) and elsewhere indicate, g’n in double negative quantifier structures is in fact a phonologically reduced version of the basic negative quantifier geen (“no”). The indications that grammaticalisation has taken place are therefore there, a matter to which we return in Section 5.

4.2

Nie1-related developments: Lexical substitution

G’n is also a core ingredient in the nie1-related developments evident in CA. Firstly, as the examples in (31a,b) show, g’n is not obligatory; clauses containing just niks (“nothing”) can also successfully express negation. Importantly, however, such clauses are not interpreted neutrally: hy wil niks luister nie2 necessarily bears an emphatic interpretation, the nature of which we return to below. Secondly, g’n in (31a) may also surface independently of niks, i.e. it does not just serve a modifying function, being able to act on its own to express negation. As (33) indicates, structures containing just g’n resemble their niks-containing counterparts in expressing non-neutral negation: (33) a.

Ek is g’n jou vriend nie2! I is no your friend neg ‘I am not your friend.’ [sa = Ek is nie1 jou vriend nie2!]

13. As such, these structures seem very different to Zeijlstra’s (2006) so-called Emphatic Multiple Negative Expressions/emnes, which he argues to be non-productive lexicalised expressions.

 Theresa Biberauer

b.

Hy skryf g’n goeie stories nie2! he write no good stories neg ‘He does not write good stories!’ [sa = Hy skryf nie1 goeie stories nie2!]14

c. Dit is g’n/G’N so moeilik nie2 it is no so difficult neg ‘It’s not so difficult at all!’ An important aspect of the use of the structures in (33) is that they must be a response to a prior assertion, a state of affairs which calls to mind Schwenter’s (2005, 2006) information-structural considerations determining the use of non-standard negation structures (cf. Section 2). The examples in (34) illustrate the type of linguistic context that can felicitously precede the g’n-containing structures in (33): (34) a. Kom nou, my vriend! come now my friend ‘Come now, my friend!’ b. Jan skryf regtig goeie stories Jan write really good stories ‘Jan writes really good stories’ c. Ek sukkel, want dis moeilik I struggle because it’s difficult ‘I’m struggling because it’s difficult’ G’n also serves metalinguistically, as illustrated in (35): (35) Dit is g’n wonderlik nie2; dis fantasties! it is no wonderful neg; it’s fantastic ‘It’s not wonderful; it’s fantastic!’ Based on the above facts, we can conclude that g’n substitutes for nie1, reinforcing it, in what we may broadly think of as presuppositional negation contexts (but cf. Schwenter 2005 for more fine-grained discussion). In Horn’s (1989: 270) terms, g’n expresses contradictory (polar) negation. When nie1 is reinforced via substitution by niks (“nothing”), by contrast, it is not limited to presuppositional contexts; it does, however, always express what Horn (1989: 27) designates contrary (scalar) negation. This is illustrated in (36):

14. The fact that (33b) is not equivalent to Hy skryf geen goeie stories nie2, i.e. the full negative quantifier-containing form, which means “He writes no good stories (only bad ones)” once again underlines the synchronic distinction between this full negative quantifier form and grammaticalised g’n.



Chapter 5.  Jespersen off course 

(36) a. Sy is niks tevrede nie2 she is nothing satisfied neg She isn’t remotely satisfied’, i.e. she isn’t satisfied at all b.

Ons doen niks sulke dinge nie215 us do nothing such things neg ‘We don’t do any of those sorts of things’ [≠ Ons doen niks nie2 – niks isn’t an object here]

A final interesting aspect of the developments affecting predicate negation in CA is that many speakers permit both g’n and niks to be further reinforced by optional nie2 (cf. Section 3 above). This is shown in (37): (37) a. b.

4.3

Dit is g’n nie2 so moeilik nie2! it is no neg so difficult neg ‘It’s not so difficult at all.’ Sy is niks nie2 tevrede nie2 she is nothing neg satisfied neg ‘She isn’t remotely satisfied.’

Summary

What we see, then, is that the non-neutral “real negators” available in CA already appear to be so integrated into the system that they have reached what we might think of as Stage II (optional NC with concomitant interpretive effects) in the domain of quantifier negation. Furthermore, it is clear that quantifier-related developments in CA are affecting the manner in which negation can be realised, with CA in effect developing a system which facilitates a more transparent indication of the various types of negation usually expressed by nie1. Thus while SA does not distinguish between the neutral negator in predicate or wide-focus and “constituent” or narrow-focus negation – nie1 serving both functions – CA does: as consideration of the examples in the preceding section will show, niks and g’n niks serve exclusively as narrow-scope negation markers; g’n may mark either wide or narrow scope depending on the nature of the discourse-old information being negated. These developments therefore also indicate how important it is to consider how quantifier and sentential negation interact (cf. Section 3 above). Additionally, they suggest that “short-cycle” JC developments are not limited to Stage II/III – Stage IV developments of the type illustrated in (24); a Stage I – Stage II “short cycle” also seems to be possible. Finally, the fact that the quantifier negation-internal “short” JC development targets strong nie1 rather than weak nie2 15. The fact that (36b) permits niks to co-occur with an object clearly signals that it is compatible with direct objects and must therefore have lost its argument status. This is not a necessary development where the negative quantifier meaning “nothing” is harnessed as part of the negation system (cf. Poletto 2008 and Bayer 2009 for discussion).

 Theresa Biberauer

unambiguously indicates that phonological considerations are not conclusive in determining “next steps” in JC. In the Afrikaans case, the motivation for the introduction of new sentential negators is clearly semantico-pragmatic, with weak nie2 continuing to function as before. As also suggested by van der Auwera (this volume), it would therefore seem that there are at least two “next steps” for a language that has reached Stage III of JC: further weakening of a by now weakened original negator, i.e. progression to Stage IV, or the opposite development, namely further amplification of the existing Stage III structure via the introduction of new means of expressing the distinction between emphatic and neutral negation in a particular domain and/or facilitating a finer-grained distinction between the different functions filled by neg1. The following section will now attempt to offer a generative account of why Afrikaans has taken the particular route it has, focusing in particular on the question of why the at first sight very plausible “traditional” expectation for nie2 – cf. (38a) – is not borne out: (38) a. “Traditional” jc expectation for Afrikaans: Stage III → Stage IV, i.e. nie1 will become optional, ultimately to be replaced by nie2 Cf. French: ne … pas → (ne) … pas → pas Std French Colloquial French English: ne …(noht) → (ne) … no(h)t → not Old English Present-Day English *Afrikaans: nie1 … nie2 → (nie1) … nie2 → nie2 sa/ca Future Afrikaans Actual jc development in Afrikaans (cf. ca): nie1 → nie1 … (nie2) → nie1 … nie2 Dutch Early Afrikaans sa/ca i.e. stable Stage III, with a “short-cycle” development affecting nie1, i.e. the original negator being reinforced in different ways (lexical substitution/optional nie2)

5. Analysis As noted at the end of the previous section, this section will concentrate primarily on the fact that nie2 has not followed the same path of development as the concord elements in well-studied Western European languages. Section 5.2 will, however, also briefly consider the quantifier negation-related developments discussed in Section 4.2 above (cf. Biberauer 2008c for more detailed discussion of these developments). 5.1. Concord element-related developments At the core of our proposal is the idea that Afrikaans crucially differs from the languages which have undergone progression from Stage III to Stage IV in respect of the syntactic nature of its reinforcing element. If we consider the concord elements that



Chapter 5.  Jespersen off course 

were introduced into French, English, German and Dutch, for example, we observe that these elements (pas, no(h)t, niht, niet) all originated within the VP/vP domain, i.e. low on the clausal spine. Thus pas is a minimiser (cf. Kiparsky & Condoravdi 2006, Condoravdi 2008, Poletto 2008) with argumental (V-complement) origins, while the other forms are negative quantifiers which have frequently been argued to be V/vPrelated (cf. Weiss 1998, Roberts & Roussou 2003, Zeijlstra 2004, Jäger 2006, 2008 and Poletto 2008). In Afrikaans, by contrast, there is no indication that nie2 originated as a minimiser of any kind; instead, there is a strong argument that nie2 originated as a CPrelated discourse marker, which might originally have been a resumptive or emphatic tag negator (cf. Roberge 2000 for details of this proposal).16 If this is correct, (spokenlanguage) structures such as the following would have served as input for the obligatory NC pattern that ultimately became standardised in SA: (39) a. Het kan niet waar zijn, nee! it can not true be no ‘It can’t be true, no!’ b. Jij komt niet mee, ne? you come not with hey ‘You aren’t coming, hey/right?’

(Roberge 2000: 147)

[expectation = negative answer]

Diachronic considerations aside, there is a range of synchronic evidence that nie2, unlike its European counterparts, is a “high” concord element (pace Bell 2004a,b; cf. Biberauer 2007, 2008a for more detailed, and see also Oosthuizen 1998). Firstly, we observe a difference in the negation possibilities available to sentential complements of verbs that take both full and restructuring complements (cf. Wurmbrand 2001 for detailed discussion). Consider (40–41) in this connection: (40) a.

Ek probeer nie1 [om die boek te lees] nie2 (al lyk dit vir I try not inf-c the book to read neg even-if looks it for jou so) you so ‘I am not trying to read the book, even if it looks like it (i.e. as if I’m trying to) to you.’ (i.e. matrix negation)

b. Ek probeer [om nie1 die boek te lees nie2] (maar dis moeilik) I try inf-c not the book to read neg but it’s difficult ‘I am trying not to read the book, but it’s difficult (to resist doing so).’ (i.e. embedded negation)

16. Van Gelderen (2008) discusses a further potential source of concord elements, namely verbal heads (cf. also van der Auwera, this volume, 2009). Evidently, then, NEG2 can be drawn from a range of different stocks, which might lead us to expect that all concord elements will not behave identically, either synchronically or over time.

 Theresa Biberauer

(41) a. Ek probeer nie1 [die boek lees] nie2 (al lyk dit vir jou so) I try not the book read neg even-if look it for you so ‘I am not trying to read the book.’ (i.e. matrix negation) b. *Ek probeer [ nie1 die boek lees nie2 ] (maar dis moeilik) I try not the book read neg but it’s difficult (i.e. embedded negation unavailable) (40) features a full infinitival complement, introduced by the infinitival C-element om (cf. German um), whereas probeer in (41) takes a reduced complement, lacking not only an overtly realised C, but also the infinitival marker te, which might be analysable as a T-element in Afrikaans (cf. Biberauer & Roberts 2005).17 In the case of (40), both the matrix and the embedded clause may be negated. More specifically, it is possible for the matrix and the embedded clause to be negated as a whole (i.e. whole-clause negation) and, under the right discourse conditions (e.g. denials), it is also possible for sub-components of these clauses to be negated. This is illustrated in (42–43): (42) Matrix negation options: a. Ek probeer nie1 [om die boek te lees] nie2; net die tydskrif I try not inf-c the book to read neg just the magazine ‘I am not trying to read the book; just the magazine.’ b.

Ek probeer nie1 [om die boek te lees] nie2; ek probeer hom I try not inf-c the book to read neg I try him vertaal translate ‘I am not trying to read the book; I’m trying to translate it.’

(43) Embedded negation options a. Ek probeer [om nie1 die boek te lees nie2]; van tydskrifte I try inf-c not the book to read neg of magazines het ek niks gesê nie2 have I nothing said neg ‘It’s the book I’m trying not to read; I didn’t say anything about magazines (which I might consider going through).’ b.

Ek probeer [om nie1 die boek te lees nie2], maar ek sal hom I try inf-c not the book to read neg but I will him dalk bietjie deurblaai maybe a-bit through-page ‘I am trying not to read the book, but I might page through it a bit.’

In the case of (41), things are rather different. In this case, whole-clause negation is only possible in matrix clauses (cf. (41a)); the only negation possibility available in 17. Infinitives featuring just te are rare in Afrikaans (cf. Ponelis 1993: 292ff).



Chapter 5.  Jespersen off course 

embedded clauses is the focused constituent negation option highlighted in (42–43) above. Thus (41a) can be used as a translation equivalent of (40a), but (41b) cannot be used in this manner; it can only surface, with the right intonation, as a counterpart of either (42a) or (42b). These facts are illustrated in (44) and (45) respectively: (44) a. Ek probeer nie1 [die boek lees ] nie2; net die tydskrif I try not the book read neg just the magazine ‘I am not trying to read the book; just the magazine.’ b.

[cf. (42a)]

Ek probeer nie1 [die boek lees] nie2; ek probeer hom I try not the book read neg I try him vertaal [cf. (42b)] translate ‘I am not trying to read the book; I’m trying to translate it’

(45) a. * Ek probeer nie1 [die boek lees] nie2; van tydskrifte het ek * I try not the book read neg of magazines have I niks gesê nie2 [cf. (43a)] nothing said neg b. * Ek probeer nie1 [die boek lees] nie2; maar ek sal hom dalk I try inf-c not the book to read neg but I will him bietjie deurblaai maybe a-bit through-page [cf. (43b)] What we see, then, is that reduced infinitivals lacking both om and te do not appear to be able to license negation. Building on proposals by Wurmbrand (2001) and others, we propose that non-reduced (non-restructuring) structures like (40) and (42–43) feature a complete/non-defective clause structure, whereas restructuring complements like (41) and (44–45) are defective in featuring less clausal structure and, in particular, in lacking C-structure. If this is correct, we can understand the negation patterns discussed above as following from the fact that C-structure needs to be present in order to license nie2, without which negation more generally cannot be licensed. Further evidence that nie2 is a “high” element comes from ‘headlinese’. As (6b) illustrated, headlines obligatorily lack nie2. Numerous researchers (cf. Stowell 1991, Vinet 1993, Paesani 2006, Paul 2007) have previously proposed that headlinese features reduced structures. Once again, if nie2 is a “high” element, its obligatory absence in Afrikaans headlinese follows straightforwardly: the relevant left-peripheral structure is simply absent. Strikingly, however, the C-domain cannot be entirely lacking in Afrikaans headlinese as verb-containing structures seem to feature verb second (V2) structures. Consider (46): (46) a. Een uit elke drie loer lewensmaat se sms’e af one out every three peep life-partner gen smses down ‘One in three spy on partner’s text-messages.’ (Die Burger 2008–09–29)

 Theresa Biberauer

b. Wie anders het Lotz vermoor, wil regter weet who else have Lotz murdered want judge know ‘Who else murdered Lotz, judge asks.’ (Beeld 2007–11–02) The separated verb and particle in (46a) clearly indicate that verb movement has taken place; since Afrikaans lacks V-to-T movement (cf. Biberauer 2003, Vikner 2001), this must either be “short movement” to v18 or V-to-C movement. The fact that interrogatives like (46b) are also possible, however, suggests that the superficially V2 headline structures are best analysed as CPs, with the finite verb in C (cf. den Besten 1977/1983). Given that we already have evidence that nie2 is only licensed where CP-structure is present, the fact that nie2 ordinarily surfaces in negated V2 clauses (cf. (1b), (2b), etc.) suggests that it must be an extremely high element, occupying the higher reaches of the left periphery. In particular, if we assume an articulated CP of the type proposed in Rizzi (1997) and in much recent work, we might think of V2 clauses as either FinP, FocusP or TopicP structures, with the nature of the initial XP determining the type of CP; nie2 then occupies the highest head, Force, and attracts the V2-containing CP structure to its specifier (see below for a refinement of this proposal). This is outlined for different types of V2 clauses in (47):19 (47) a. [ForceP [FinP Subj Vf [TP … ]] nie2] -- neutral subject-initial V2 clause b. [ForceP [FocP Wh Vf [TP … ]] nie2] -- wh-initial V2 clause c. [ForceP [TopP Obj Vf [TP … ]] nie2] -- topical object-initial V2 clause 18. Alternatively, movement may be to a “low” position within the vP-domain, but we abstract away from the precise structure of the lower clausal domains, referring only to VP, vP and TP, i.e. the “cover terms” Chomsky has employed since 1995. For further discussion, see the following note. 19. As indicated in the outline diagrams, I do not assume the entire articulated CP-structure to be projected in the case of V2 clauses; the idea is, instead, that C represents a functional category whose internal feature structure reflects the hierarchical structure of the articulated CP, with only the features relevant to a given derivation being activated as probes. Thus, if we view features as [Attribute: Value] pairs, we can distinguish a probing/unvalued feature – [Person: __] – from a valued one – [Person: 3] – and attribute to C partial structures such as the following (outline font indicates non-activated features; * indicates a movement-triggering diacritic associated with the verbal “spine” [cf. Brody 1998] which C extends; and EF signals the “edge feature” ensuring XP-movement [cf. Chomsky 2008]): (i) C in a neutral V2 clause: *

: __ : __ : __ : __ , EF

(ii) C in a wh-question: : __ * : __ , EF : __ : __ See e.g. Poletto (2000), Roberts (2005), Mohr (2005) and van Craenenbroeck & Haegeman (2007) for a range of different accounts of V2 couched in terms of an articulated CP.



Chapter 5.  Jespersen off course 

Main-clause negated structures in Afrikaans therefore have the partial structure in (48): (48)

ForceP

Force’ Foc/Top/FinP Force

Foc/Top/FinP

nie2 ...

On this analysis, nie2 therefore behaves in the same manner as force-related clause-final particles that have been observed in a range of other languages (cf. Biberauer 2007, Poletto 2008 for references). To summarise, then, there seem to be good indications that concord elements in well-studied European languages are structurally “low”, whereas Afrikaans nie2, which occupies what stringwise appears to be “the same” position as these other reinforcers, seems to be structurally “high”. Another important difference between Afrikaans and the languages which have come full circle on JC is that the concord elements in the latter clearly “became negative” in the relevant sense20, with the result that they are generally analysed as Neg-elements in the literature (cf. Haegeman 1995, Zanuttini 1997, Rowlett 1998, van Kemenade 2000, Roberts & Roussou 2003, Wallage 2005; Jäger 2006, 2008; see, however, Rooryck 2008 for an opposing view). Thus Haegeman (1995) and Haegeman & Zanuttini (1996) propose the Neg Criterion in (49), in terms of which NC languages simply

20. For ease of exposition, we abstract away from the much-disputed question of the precise locus/representation of the negative semantics in negative structures; hence the vague formulation here. See Zeijlstra (2004, 2008), Penka (2007), Rooryck (2008) and Biberauer & Zeijlstra (2009) for recent discussion suggesting that matters may not be this straightforward.

 Theresa Biberauer

differ from their non-NC counterparts in consistently lexicalising both the Neg-head (usually via a concord element) and its specifier: (49) Neg-Criterion: a. A neg-operator must be in a Spec-head configuration with an X0 [NEG]; b. An X0 [NEG] must be in a Spec-head configuration with a NEG-operator

NegP 3 Spec Neg’ 3 Neg … [+Neg]

On this view, there may be parametric variation in respect of the height of NegP, with Germanic having a low, vP-related NegP and Romance a slightly higher, TP-related NegP (cf. references cited above). As first pointed out by Oosthuizen (1998), the Negelement analysis does not, however, seem satisfactory for Afrikaans as nie2’s distribution is not limited to negative contexts. Thus, as (50) illustrates, a currently ill-understood class of non-veridical operators (cf. Giannakidou 2005) seems to be able to license nie2 (cf. also note 7): (50) a. Ek kan tog (* onmoontlik) alleen die werk doen (nie2)! I can surely impossibly alone the work do neg ‘Surely I can’t possibly be expected to do the work on my own!’ (Oosthuizen 1998: 79) b. Jy sal my (*nouliks) daarvan oortuig kry (nie2)! you will me hardly there-from convinced get neg ‘You will hardly convince me of that!’ c. Ek weier om saam te kom (nie2) I refuse inf-c together to come neg ‘I refuse to come along.’ Given data such as (50), Oosthuizen (1998) suggests that nie2 is in fact a high polarity head, dominating CP, whose presence is not dependent on the projection of NegP (cf. also note 7). Biberauer (2007, 2008a) develops this idea further, pointing to nie2’s possible origins as a polarity-reinforcing element (cf. (39) above) and highlighting data that suggest that nie1 is best viewed as a member of the class of vP-related adverbials rather than as specifically the head of a positionally fixed NegP. In this paper, we have additionally seen that nie2 synchronically still seems to serve a very strongly polarityrelated function in CA: wherever it is optionally present – i.e. in non-clausal structures – its presence brings about a more emphatically negative interpretation. Where it is present in the absence of negation, it is likewise clear that its presence is determined by the presence of what we can imprecisely identify as a polarity-related element.



Chapter 5.  Jespersen off course 

Given these facts, then, it seems justified to view nie2 as the realisation of a CPrelated Pol-head (cf. Laka 1990, 1994 on parametric variation in respect of TP- vs CPrelatedness of Pol-heads21). At first sight it might seem plausible to suggest that CPrelated Pol is simply an instantiation of the highest clausal functional head, Force (cf. (48)). If Force, however, instantiates the highest projection of the “verbal spine” comprising the clause (cf. Brody 1998 and the discussion in note 19), this cannot be the case: as the discussion in Section 4.1 has clearly shown, nie2 cannot be analysed as a specifically clausal domain-related element since it is also able to surface finally in nonclausal structures. As such, I propose that nie2 in fact spells out a functional head that has undergone not only semantic bleaching, but also categorial bleaching. The fact that it is no longer restricted to negative contexts signals the former type of bleaching, while the fact that this element appears to have grammaticalised into a generalised polarity head, available both in clausal and non-clausal domains, points to additional categorial bleaching: unlike West Flemish en, which is restricted to the clausal domain and which Breitbarth & Haegeman (2008) consequently accord the value [+finite] as part of its feature specification (cf. note 21), Afrikaans nie2 cannot encode any specifically clause-related features. I therefore propose that it is a categorially deficient element which does not extend either “verbal” or “nominal” spines (cf. also Biberauer 2008d and Biberauer & Sheehan 2008) and, as such, lacks the [+V] or [+N] specification we would expect to find on non-defi21. Strikingly, en, the original neg1 in West Flemish, which is generally analysed as serving a concord function in the modern language (cf. Haegeman 1995 and much other work), also seems in fact to be a polarity head. Thus Breitbarth & Haegeman (2008) show that en in Flemish tussentaal (“in-between language”) and certain West Flemish dialects may in fact surface in both negative and positive contexts: (i)

Background: Salary scales are under discussion … Da kan toch niet? Da en kan toch niet?! That can mod.p. not that pol can mod.p not ‘That can’t be right? Surely that can’t be right?!’

(ii). A: Valère verkuopt da nie Valère sell that not ‘Valère doesn’t sell that.’ . B: J’ en doet Je verkoopt da wel. K’een der gisteren gekocht. He pol does He sells that indeed I have some yesterday bought ‘He does. He does sell that. I bought some yesterday.’ (Lapscheure dialect, cf. Haegeman 1995: 160; cf. also van Craenenbroeck 2004) For Breitbarth & Haegeman (2008: 14), then, en is the spellout of Pol [+affective, +emphatic], i.e. a feature-bundle which does not include a negative value. Its positioning suggests that it may, in Laka’s terms, be a CP-related Pol-head, like nie2 in Afrikaans. Cf. van Gelderen (2008) for further discussion of some languages that appear to resemble Afrikaans in featuring a CP-related Pol head. Biberauer (2008b, c) and Biberauer & Roberts (2008), in turn, argue that do–support in Modern English may instantiate a TP-Pol phenomenon.

 Theresa Biberauer

cient elements. As such, we might think of nie2 as an element which has been radically grammaticalised, having undergone not only the upward reanalysis that Roberts & Roussou (2003) and van Gelderen (2004) assume to underlie grammaticalisation processes more generally, but also, additionally, a categorial bleaching process which renders this grammaticalised element particularly “light” from a formal (featural) perspective. Afrikaans nie2, then, differs from the originally reinforcing concord (neg2) elements in better-studied European languages in a number of respects. Firstly, as we saw in Section 2 above, it is a weak element which clearly lacks the strengthening possibilities available to nie1 (neg1). Secondly, it lacks the negative specification clearly retained by the languages that proceeded to Stage IV and can, as such, be viewed as semantically deficient relative to these elements. Thirdly, it is evidently an element that occupies a much higher structural position than the other neg2 elements which eventually became “real” negators (neg1). Finally, it appears to be severely categorially deficient. This combination, I argue, makes nie2 unsuited as a possible candidate for “real” negator status, with the result that a Stage III → Stage IV development along the lines of what has been observed in the other languages considered here is not to be expected: nie2 has, as it were, “grammaticalised beyond the Cycle”. Cyclic developments are not, however, more generally ruled out, and various distinct JC developments, rather clearly motivated by semantico-pragmatic considerations, do indeed appear to be underway in Afrikaans.

5.2

Negative quantifier-related developments

Before concluding, we will briefly consider the quantifier negation-related developments discussed in Sections 4.1 and 4.2 above. We will firstly offer a few thoughts on the nie1 replacement strategies before doing the same for the multiple negative quantifier-containing structures. 5.2.1 N-words replacing nie1: Niks and g’n On the matter of the choice of the lexical items that serve as reinforcing negators in place of nie1, we note only that g’n (> geen – ‘no’) and niks (‘nothing’) correspond to the smallest elements on the quantification scale (cf. Horn’s 1989: 237 Boethian Square of Oppositions) and, as such, constitute relatively unsurprising reinforcement choices. Poletto (2008) furthermore observes that niente (‘nothing’) appears to be one of the four etymological sources from which all the sentential negators that have been identified in Italian dialects originate (the others being minimiser, polarity head of the type assumed by Laka and, finally, the anaphoric negator). The reason she proposes for the fact that nothing so commonly serves as a source for sentential negation is that it “is always the semantically (and syntactically) barest operator, in the sense that it has the smallest set of features because its lexical restrictor is virtually non-existent”. In other words, it can be viewed as the negation equivalent of what in the wh-domain and that in the complementiser-domain (cf. also Kayne 2008 for recent discussion along partially similar lines).



Chapter 5.  Jespersen off course 

An additional factor that appears to be relevant in the choice of niks in particular is that it is a negative quantifier that quantifies over entities, amounts, etc. In employing lexical substitutions to reinforce negation of different types, we might expect languages to draw on alternative negatively specified resources from the domain being negated, and Afrikaans certainly appears to do so in harnessing niks and g’n (which appears to have undergone reanalysis from a specifier element to a head – cf. van Gelderen’s 2004 Head Preference Principle – possibly because of the semantics of its restrictor (niks)). Similarly, various species of non-standard English also appear to draw on a negative quantifier quantifying over the same domain as the negative element they replace. Consider the examples in (51) in this connection: (51) a. I didn’t do that b. I never did that Here never, a temporal quantifier, replaces didn’t, an element that is generally agreed to be a T-element (cf. Biberauer & Roberts 2008 for recent discussion; cf. also Breitbarth, Lucas & Willis 2008 for recent discussion of English never; and Ramchand 2004 for more general discussion of the differences between vP- and TP-related negation elements). In certain respects, then, the developments involving niks and g’n in CA can readily be understood. More work is, however, required to understand, for instance, how the narrow-scope (“constituent negation”) use of these elements is similar and different to the wide-scope (“sentential negation”) use of the same element (i.e. niente-type negation elements) found in other languages, and whether and, if so, how the presuppositional associations of g’n might be captured formally. These are all matters which are left for future research. 5.2.2 The availability of NC and DN readings As we saw in Section 4.1, there are circumstances under which co-occurring negative quantifiers in CA can give rise to either an NC or a DN reading. From a theoretical perspective, this possibility is problematic, particularly for theories which assume the grammars of NC languages to differ in some fundamental (possibly, parametric) respect from non-NC languages (cf. Haegeman & Lohndal 2008 and Zeijlstra 2008 for recent discussions). The aim here is not to try to resolve this matter; instead, I simply wish to highlight a previously unnoted point about the availability of NC as opposed to DN readings in CA that may prove important in determining what the correct analysis of this phenomenon may be (cf. also Biberauer & Zeijlstra 2009 for further discussion). A striking fact about NC and DN interpretations in CA is that the ordering and prosodic properties of the negative quantifiers in the structure matter. Consider (52) to (53): (52) a. Jy vertel my ook nooit niks nie2 you tell me also never nothing neg ‘You also never tell me anything.’

[nc]

 Theresa Biberauer

b.

Jy vertel my ook nooit niks nie2 you tell me also never nothing neg ‘You also never tell me nothing’, i.e. you’re always telling me something [dn, cf. (27c) above]

c.

Jy vertel my ook niks nooit nie2 you tell me also nothing never neg ‘You also never tell me nothing’, i.e. you’re always telling me something ≠ ‘You also never tell me anything.’ [only dn with scrambled order]

(53) a.

Ek het nog niks nêrens geteken nie2, maar my kontrak met I have still nothing nowhere signed neg but my contract with Natal verstryk begin Januarie Natal end beginning January ‘I haven’t signed anything yet, but my contract with Natal ends at the beginning of January.’ (Rapport 1999–01–03) [nc]

b.

Ek het nêrens niks geteken nie2 … I have nowhere nothing signed neg ‘There is nowhere that I signed nothing’, i.e. for each place, I signed something ≠ ‘I haven’t signed anything …’ [dn]

In each case, we see that neutral intonation delivers an NC reading22 and that prosodic marking of a particular type produces DN (cf. also Falaus 2007 on Romanian, which appears to show similar behaviour). More specifically, the required prosodic marking strongly resembles that which is necessarily required in structures like (52c), i.e. in structures where an object (regardless of its negative/non-negative status) has clearly scrambled over an appropriate adverb; where the object and the adverb in question are negative, the only possible reading is a DN one, as shown above. This observation suggests that the generalisation about NC and DN readings may be that NC readings are only available in structures featuring an unraised negative quantifier. Further evidence that this generalisation may be important comes from a comparison of the intonation patterns associated with double negative quantifier-containing 22. This suggests that NC structures do not need to start off as emphatic, a view that is sometimes extrapolated from JC schematisations like (3). See also Giannakidou (2005) and Zeijlstra (2006), who propose that NC readings in non-NC languages require emphatic intonation, and van Gelderen (2008), who proposes that this may also extend to certain varieties of Mainland Scandinavian. For Oevdalian, a Scandinavian variety which today seems to exhibit NC quite systematically, though not obligatorily, Garbacz (2008), however, explicitly notes that NC is not connected to emphasis. It may therefore be the case that special intonation is required to signal negative readings that depart from the neutral “norm” associated with a given system – NC readings in non-NC systems and DN readings in NC-systems like Afrikaans, or it may be the case that non-NC (i.e. DN) readings are universally intonationally marked, regardless of the type of system involved. Evidently, further research is required to pin down the exact relation between negative interpretations and prosodic considerations.



Chapter 5.  Jespersen off course 

structures and those in which the so-called Diesing Effects (cf. Diesing 1992) are observable. As (54) illustrates, the prosodic properties associated with DN structures is the same as that associated with structures in which bare indefinites receive a specific reading, i.e. that associated with movement out of the domain of existential closure (for Diesing, the VP): (54) a. Niemand gee my niks nie2 no-one give me nothing neg ‘No-one gives me anything.’ – nc reading; no special intonation required b. Niemand gee my niks nie2 no-one give me nothing neg ‘No-one gives me nothing.’ – dn reading; special intonation required c.

Niemand gee my boeke nie2 no-one give me books neg ‘No-one gives me books (specifically).’ – specific (i.e. non-in situ) reading; special intonation required

Strikingly, Rooryck (2008:  9) makes what appears to be a parallel observation for French, noting that “negative interpretations of [n-words in NC languages like French– TB] behave like universal quantifiers, while the nonnegative interpretations behave as existential quantifiers”. Translated into the terms of Diesing’s Mapping Hypothesis, nwords in French receive negative interpretations where they are located outside VP, whereas they are interpreted as non-negatives whenever they are located within this domain. Furthermore, Rooryck (2008: 25) also cites French data from Corblin et al. (2004) which indicates that “stress takes the stressed n-word out of the scope of the scope domain of negation:23 (55) a. personne ne dit rien à personne nobody neg says nothing to nobody ‘Nobody says nothing to nobody.’

(dn reading)

b. Personne ne dit rien à personne nobody neg says nothing to nobody ‘Nobody say nothing to nobody.’

(dn reading)

23. This function of stress may also be relevant in completely unrelated contexts, such as structures like (i) where Condition C appears to be violated: (i) A: No-one likes John anyway! B: JOHN likes John!

 Theresa Biberauer

Further evidence that movement is a crucial ingredient in the understanding of the availability of DN vs NC readings comes from fronting: (56) a. Jy vertel my ook nooit niks nie2 you tell me also never nothing neg ‘You also never tell me anything.’

(cf. (52a)), dn/nc readings)

b. Nooit vertel jy my niks nie2 never tell you me nothing neg ‘You never tell me nothing’, i.e. obligatory dn reading. c. Niks vertel jy my nooit nie2 nothing tell you me never neg ‘There is nothing that you never tell me’, i.e. obligatory dn reading. As the examples show, NC readings are never possible, regardless of ordering, when one of the negative quantifiers has undergone fronting. What the CA data suggest, then, is that it is not just the featural make-up of negative quantifiers or of negation-related functional projections or the presence vs absence of negative operators of various kinds which determine how negation is interpreted in a particular language; in NC languages at least, movement and, consequently, syntax-PF mapping may also play a crucial role in determining the availability of NC vis-à-vis DN readings.24

6. Conclusion Consideration of negation-related developments in contemporary Afrikaans clearly show that this language is in a number of respects an atypical Stage III language. Although it superficially appears to have reached the same NC stage that English, German, Dutch and French previously reached, it is quite clear that the components corresponding to neg1 and neg2 in Afrikaans are rather different to those in its Indo-European relatives: whereas neg1 in the latter became a weak element allowing neg2 to take over its sentential-negation role, Afrikaans nie1 has remained strong, while nie2 appears to be both semantically and grammatically bleached. Given this, the fact that Colloquial Afrikaans (CA) and less conservative varieties of Afrikaans like Kaaps are not taking the “next step” in JC – loss of neg1 in favour of neg2 – emerges as readily understandable: the system underlying Afrikaans NC is in fact rather different to that which underlay the Western European languages which did progress to Stage IV of the Cycle. In particular, it seems that nie2 has “grammaticalised beyond the 24. Cf. Błaszczak & Gärtner (2005) for discussion of another context in which syntax-PF mapping and, consequently, prosodic considerations appear to determine the manner in which negation is interpreted.



Chapter 5.  Jespersen off course 

Cycle”, lacking the semantic and grammatical properties neg2 requires if it is to take over as the “real” negator. The case of contemporary Afrikaans thus highlights the importance of not approaching JC solely in string-wise terms; apparently, non-surface considerations such as the structural height of the concord element and the features it encodes are also crucial considerations in determining whether and if so, how, the Cycle will continue. Based on the facts considered here, then, the general hypothesis in (57) suggests itself: (57) Hypothesis concerning progression from Stage III to Stage IV Concord elements will only become “real” negators (i.e. undergo the change from Stage III to Stage IV) where these elements are (a) located suitably “low” within the clausal domain, and (b) bear appropriate substantive and categorial content. This hypothesis makes some clear, testable predictions. Firstly, we would in general expect concord elements drawn from the vP and TP domains to be available for reanalysis as potential “real” negators, whereas those drawn from the CP-domain would not be. Note that this statement does not compromise the availability of CP as the domain in which sentential negation is neutrally expressed in non-NC languages. As noted in the main text, languages differ in respect of the structural height of their sentential negators – Germanic having “low”, vP-related negation, Romance “higher” TPrelated negation and Celtic, for example, licensing negative particles in CP (cf. Duffield 1995, Roberts 2005 and Willis 2008 for discussion). (57), then, does not preclude the possibility of “real” negation being expressed in CP; only of a CP-related concord element being reanalysed as a “real” negator. The prediction, then, is that the “real” negators in languages with CP negation will not at some earlier stage of their development have been CP-related concord elements. Northern rural varieties of Brazilian Portuguese, which have been argued to feature a very high CP-related concord element (cf. Cavalcante 2007), represent a relevant test case here, as do Italian dialects featuring clause-final concord elements that appear to be drawn from the discourse domain (cf. Poletto 2008) and, to the extent that their final concord elements are likewise CP-related, the so-called Bipartite Final Negation languages discussed in Bell (2004a). See Biberauer (2008b) for consideration of these cases. Secondly, (57) also predicts that concord elements which derive from bleached “real” negators, but which have been retained, possibly optionally – cf. West Flemish en and possibly also Colloquial French ne – will not come full circle to reclaim the role of neg1: although they might be located “low” enough to be reanalysed, these elements lack the substantive and possibly also categorial content to serve as viable negators. For West Flemish, this seems correct (cf. Breitbarth & Haegeman 2008). (57), then, predicts at least two scenarios where concord elements will not displace an existing “real” negator. As the investigation of CA has shown, however, this does not necessarily mean that the “real” negator in such systems is immune to JC developments

 Theresa Biberauer

involving other elements. To the extent that JC is a wider phenomenon that extends beyond the domain of negation, the proposals and predictions made here should obviously also carry over to other domains in which Cycles have been observed.

References Aelbrecht, Lobke. 2006. Ellipsis in negative fragment answers. Ms, Katholieke Universiteit Brussel. Available on-line at: http://ling.auf.net/lingBuzz/000343. van der Auwera, Johan. 2009. On the diachrony of negation. To appear in The Expression of Negation, Laurence Horn (ed). Berlin: Mouton. van der Auwera, Johan & Neuckermans, Annemie. 2004. Jespersen’s cycle and the interaction of predicate and quantifier negation in Flemish. In Typology Meets Dialectology, Dialect Grammar from a Crosslinguistic Perspective, Bernd Kortmann (ed.), 454–478. Berlin: Mouton. van der Auwera, Johan, de Cuypere, Ludovic & Neuckermans, Annemie. 2006. Negative indefinites: A typological and diachronic perspective on a Brabantic construction. In Types of Variation. Diachronic, Dialectal and Typological Interfaces [Studies in Language Companion series 76], Terttu Nevalainen, Juhani Klemola & Mikko Laitinen (eds), 305–319. Amsterdam: John Benjamins. Bayer, Josef. 2009. Nominal negative quantifiers as adjuncts. Journal of Comparative Germanic Linguistics 12: 5–30. Bell, Arthur. 2004a. How N-words move: Bipartite negation and ‘Split-NegP’. In Triggers, Anne Breitbarth & Henk van Riemsdijk (eds), 77–114. Berlin: Mouton de Gruyter. Bell, Arthur. 2004b. Bipartite Negation and the Fine Structure of the Negative Phrase. PhD dissertation, Cornell University. Besten, Hans den. 1977/1983. On the interaction of root transformations and lexical deletive rules. In On the Formal Syntax of the West Germania [Linguistik Aktuell-Linguistics Today 3], Werner Abraham (ed.),47–131. Amsterdam: John Benjamins. Besten, Hans den. 1986. Double negation and the genesis of Afrikaans. In Substrata versus Universals in Creole Genesis [Creole Language Library 1], Pieter Muysken & Norval Smith (eds), 185–230. Amsterdam: John Benjamins. Biberauer, Theresa. 2003. Verb Second (V2) in Afrikaans: A Minimalist Approach to Word-order Variation. PhD dissertation, University of Cambridge Biberauer, Theresa. 2007. A closer look at negative concord in Afrikaans. Stellenbosch Papers in Linguistics Plus 35: 1–51. Biberauer, Theresa. 2008a. Doubling and omission: Insights from Afrikaans negation. In Microvariations in Syntactic Doubling, Sjef Barbiers, Margreet van der Ham, Olaf Koeneman & Marika Lekakou (eds), 103–140. Bingley: Emerald. Biberauer, Theresa. 2008b. Continuity and change in Afrikaans negation: Facts and implications. Paper presented at DiGS X (Cornell). Biberauer, Theresa. 2008c. Locating Afrikaans Afrikaans in Jespersen’s Cycle: A Comparative Investigation. Paper presented at CGSW 23 (Edinburgh). Biberauer, Theresa. 2008d. A third factor-based typology of disharmonic languages? Paper presented at the Cambridge-Delhi-Hyderabad-Nanzan Workshop (Hyderabad).



Chapter 5.  Jespersen off course  Biberauer, Theresa & Sheehan, Michelle. 2008. Disharmony, antisymmetry, and the Final-overFinal Constraint: Why clauses need nominals. Paper presented at the Ways of Structure Building conference. Biberauer, Theresa & Richards, Marc. 2006. True optionality: When the grammar doesn’t mind. In Minimalist Essays, Cedric Boeckx (ed.), 35–67. Amsterdam: John Benjamins. Biberauer, Theresa & Roberts, Ian. 2005. The loss of residual ‘head-final’ orders and remnant fronting in Late Middle English: Causes and consequences. In Comparative Studies in Germanic Syntax: From A(frikaans) to Z(ürich German) [Linguistik Aktuell-Linguistics Today 97], Jutta Hartmann & László Molnárfi (eds), 263–297. Amsterdam: John Benjamins. Biberauer, Theresa & Roberts, Ian. 2008. Subjects, tense and verb-movement in Germanic and Romance. Cambridge Occasional Papers in Linguistics 3: 24–43. Biberauer, Theresa & Zeijlstra, Hedde. 2009. Negative concord in Afrikaans: Filling the typological gap. To appear in Proceedings of NELS 39. Amherst MA: GLSA. Błaszczak, Johanna & Gärtner, Hans-Martin. 2005. Intonational phrasing, discontinuity, and the scope of negation. Syntax 8(1): 1–22. Breitbarth, Anne & Haegeman, Liliane. 2008. Not continuity but change: Stable Stage II in Jespersen’s Cycle. Paper presented at the Continuity and Change conference (Cambridge, UK). Breitbarth, Anne, Lucas, Chris & Willis, David. 2008. Never again: The regrammaticalization of NEVER as a marker of sentential negation in English. Paper presented at New Approaches to Grammaticalization 4 (Leuven). Brody, Michael. 1998. Projection and phrase structure. Linguistic Inquiry 29: 367–398. Cardinaletti, Anna & Starke, Michal. 1999. The typology of structural deficiency: A case study of the three classes of pronouns. In Clitics in the Languages of Europe, Henk van Riemsdijk (ed.), 145–233. Berlin: Mouton de Gruyter. Cavalcante, Rerrison. 2007. A negação pós-verbal no português brasileiro: Análise descritiva e teóretica de dialetos rurals de afro-descendentes. MA thesis, Universidade Federal da Bahia. Chomsky, Noam. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky, Noam. 2008. On phases. In Foundational issues in linguistic theory, Robert Freidin, Carlos Otero & Maria-Luisa. Zubizarreta (eds), 133–166. Cambridge MA: The MIT Press. Condoravdi, Cleo. 2007. Jespersen’s Cycle and the rise of NPIs. Paper given at the Workshop on Language Change (Amsterdam). Corblin, Francis, Déprez, Viviane, de Swart, Henriëtte & Tovena, Lucia. 2004. Negative concord. In Handbook of French Semantics, Francis Corblin & Henriëtte de Swart (eds), 417–452. Stanford CA: CSLI. Craenenbroeck, Jeroen van. 2004. Ellipsis in Dutch Dialects. PhD dissertation: Leiden. Craenenbroeck, Jeroen van & Haegeman, Liliane. 2007. The derivation of subject-initial V2. Linguistic Inquiry 38(1): 167–178. Dahl, Östen. 2001. Inflationary effects in language and elsewhere. In Frequency and the Emergence of Linguistic Structure [Typological Studies in Language 45], Joan Bybee & Paul Hopper (eds), 471–480. Amsterdam: John Benjamins. Diesing, Molly. 1992. Indefinites. Cambridge MA: The MIT Press. Donaldson, Bruce. 1993. A Grammar of Afrikaans. New York NY: Mouton de Gruyter. Duffield, Nigel. 1995. Particles and Projections in Irish Syntax. Dordrecht: Kluwer. Espinal, Maria Teresa. 1992. Expletive negation and logical absorption. The Linguistic Review 9: 333–358. Espinal, Maria Teresa. 2000. Expletive negation, negative concord, and feature checking. Catalan Working Papers in Linguistics 8: 47–69.

 Theresa Biberauer Espinal, Maria Teresa. 2007. Licensing expletive negation and negative concord in Romance languages. In La négation dans les langues romanes [Lingvisticæ Investigationes Supplementa 26], Franck Floricic (ed.), 49–74. Amsterdam: John Benjamins. Falaus, Anamaria. 2007. Negative concord and double negation: The Romanian puzzle. In Romance Linguistics 2006 [Current Issues in Linguistic Theory 287], José Camacho, Nydia Flores-Ferrán, Liliana Sánchez, Viviane Déprez and María José Cabrera (eds),135–148. Amsterdam: John Benjamins. Fox, Danny. 2000. Economy and Semantic Interpretation. Cambridge MA: The MIT Press. Garbacz, Piotr. 2008. Negative concord in Oevdalian. Paper presented at ScanDiaSyn Grand Meeting 2008 (Sandbjerg). Gass, Kate van. 2007. Multiple n-words in Afrikaans. Stellenbosch Papers in Linguistics Plus 35: 167–201. Gelderen, Elly van. 2004. Grammaticalization as Economy [Linguistik Aktuell-Linguistics Today 71]. Amsterdam: John Benjamins. Gelderen, Elly van. 2008. Negative Cycles. Linguistic Typology 12: 195–243. Giannakidou, Anastasia. 2000. Negative... Concord? Natural Language and Linguistic Theory 18: 457–523. Giannakidou, Anastasia. 2005. N-words and negative concord. In The Blackwell Companion to Syntax, Martin Everaert & Henk van Riemsdjik (eds), 327–391. Oxford: Blackwell. Haegeman, Liliane. 1995. The Syntax of Negation. Cambridge: CUP. Haegeman, Liliane & Lohndal, Terje. 2008. Negative concord and (multiple) agree: A case study of West Flemish. Ms, Universities of Lille III and Oslo/Maryland. Available on-line at: http://ling.auf.net/lingBuzz/000739. Haegeman, Liliane & Zanuttini, Raffaella. 1996. Negative concord in West Flemish. In Parameters and Functional Heads. Essays in Comparative Syntax, Adriana Belletti & Luigi Rizzi (eds), 117–179. Oxford: OUP. Hagemeijer, Tjerk. 2007. The Clause Structure of Santome. PhD dissertation, Lisbon University. Herburger, Elena. 2001. The negative concord puzzle revisited. Natural Language Semantics 9: 289–333. Horn, Laurence. 1989. A Natural History of Negation. Chicago IL: The University of Chicago Press. Jäger, Agnes. 2006. History of German Negation. PhD dissertation, Friedrich Schiller Universität Jena. Jäger, Agnes. 2008. History of German Negation. Amsterdam: John Benjamins. Jespersen, Otto. 1917. Negation in English and Other Languages. Copenhagen: A.F. Høst. Kayne, Richard. 2008. Why isn’t this a complementizer? Ms, NYU. Available on-line at: http:// ling.auf.net/lingBuzz/000726. Kemenade, Ans van. 2000 Jespersen’s cycle revisited: Formal properties of grammaticalisation. In Diachronic Syntax: models and mechanisms, Susan Pintzuk, George Tsoulas & Anthony Warner (eds), 51–75. Oxford: OUP. Kiparsky, Paul & Condoravdi, Cleo. 2006. Tracking Jespersen’s cycle. In Proceedings of the 2nd International Conference of Modern Greek Dialects and Linguistic Theory, Mark Janse, Brian Joseph & Angela Ralli (eds), 172–197. Mytilene: Doukas. Laka, Itziar. 1990. Negation in Syntax. PhD dissertation, MIT. Laka, Itziar. 1994. On the Syntax of Negation. New York NY: Garland. McCormick, Kay. 2002. Language in Cape Town’s District Six. Oxford: OUP. Miestamo, Matti. 2005. Standard Negation: The Negation of Declarative Verbal Main Clauses in a Typological Perspective. Berlin: Mouton de Gruyter.



Chapter 5.  Jespersen off course  Mohr, Sabine. 2005. Clausal Architecture and Subject Positions. Amsterdam: John Benjamins. Molnárfi, Laszlo. 2002. Die Negationsklammer im Afrikaans: Mehrfachnegation aus formaler und funktionaler Sicht. In Issues in Formal German(ic) Typology, Werner Abraham & JanWouter Zwart (eds), 223–261. Amsterdam: John Benjamins. Molnárfi, László. 2004. On the interpretation of multiple negation in spoken and written Afrikaans. In The Composition of Meaning, Alice ter Meulen & Werner Abraham (eds), 145– 167. Amsterdam: John Benjamins. Oosthuizen, Johan. 1998. The final nie in Afrikaans negative sentences. Stellenbosch Papers in Linguistics 31: 61–94. Paesani, Kate. 2006. The case of special registers. In The Syntax of Nonsententials. Multidisciplinary Perspectives [Linguistik Aktuell-Linguistics Today 93], Liljana Progovac, Kate Paesani, Eugenia Casielles & Ellen Barton (eds),147–182. Amsterdam: John Benjamins. Paul, Ileana. 2007. Missing topics in Malagasy headlines. Paper presented at the Canadian Linguistics Association (University of Saskatchewan). Pauwels, Jan. 1958. Het Dialect van Aarschot en Omstreken. Tongeren: Belgisch Interuniversitair Centrum voor Neerlandistiek. Penka, Doris. 2007. Negative Indefinites. PhD dissertation, Tübingen. Ponelis, Frederick. 1993. The Development of Afrikaans. Duisburg: Peter Lang. Poletto, Cecilia. 2000. The Higher Functional Field in the Northern Italian Dialects. Oxford: OUP. Poletto, Cecilia. 2008. On negation splitting and doubling. Paper presented at NORMS Workshop on Negation (Oslo, 1  March). Handout available at: http://www.hf.uio.no/tekstlab/ negasjon07/programme.html. Ramchand, Gillian. 2004. Two types of negation in Bengali. In Clause Structure in South Asian Languages, Anoop Mahajan & Veneeta Dayal (eds), 39–66. Dordrecht: Springer. Reinhart, Tanya. 2006. Interface Strategies: Optimal and Costly Computations. Cambridge MA: The MIT Press. Roberge, Paul. 2000. Etymological opacity, hybridization, and the Afrikaans Brace negation. American Journal of Germanic Linguistics and Literatures 12(1): 101–176. Roberge, Paul. 2002. Afrikaans: Considering origins. In Language in South Africa, Rajend Mesthrie (ed.), 79–103. Cambridge: CUP. Robbers, Karen. 1992. Properties of negation in Afrikaans and Italian. In Linguistics in the Netherlands, Reineke Bok-Bennema & Roeland van Hout (eds), 223–234. Amsterdam: John Benjamins. Robbers, Karen. 1997. Non-finite Verbal Complements in Afrikaans. A Comparative Approach. PhD dissertation, University of Amsterdam. Roberts, Ian. 2005. Principles and Parameters in a VSO Language: A Case Study in Welsh. Oxford: OUP. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: CUP. Rooryck, Johan. 2008. On the compositional nature of syntactic negation in French. Ms, Leiden University. Rowlett, Paul. 1998. Sentential Negation in French. Oxford: OUP. Schwenter, Scott. 2002. Pragmatic variation between negatives: Evidence from Romance. University of Pennsylvania Working Papers in Linguistics 8(3): 249–263. Schwenter, Scott. 2005. The pragmatics of non-canonical negation in Brazilian Portuguese. Lingua 115: 1427 – 1456.

 Theresa Biberauer Schwenter, Scott. 2006. Fine-tuning Jespersen’s cycle. In Drawing the boundaries of meaning: neo-Gricean studies in Pragmatics and Semantics in honor of Laurence R. Horn [Studies in Language Companion Series 80], Betty Birner & Gregory Ward (eds), 327–344. Amsterdam: John Benjamins. Stowell, Timothy. 1991. Empty heads in abbreviated English. Paper presented at GLOW. Vikner, Sten. 2001. Verb Movement Variation in Germanic and Optimality Theory. Habilitationschrift, Tübingen. Vinet, Marie-Thérèse. 1993. L’aspect et la copule vide dans la grammaire des titres. Langue française 100: 83–100. Waher, Hester. 1978. Die Probleem van die Bereik van die Ontkenning met Spesiale Verwysing na Afrikaans. PhD dissertation, University of Cape Town. Wallage, Phillip. 2005. Negation in Early English: Parametric Variation and Grammatical Competition. PhD dissertation, University of York. Watanabe, Akira. 2004. The genesis of negative concord: Syntax and morphology of negative doubling. Linguistic Inquiry 35: 559–612. Weiss, Helmut. 1998. Syntax des Bairischen. Studien zur Grammatik einer natürlichen Sprache. Tübingen: Niemeyer. Willis, David. 2008. A minimalist approach to Jespersen’s cycle in Welsh. Cambridge Occasional Papers in Linguistics (COPiL) 3: 1–23. Available on-line at: http://www.ling.cam.ac.uk/ COPIL/Vol3(Online)/1willis.pdf Wouden, Ton van der. 1994. Negative Contexts. PhD dissertation, University of Groningen. Wurmbrand, Susanne. 2001. Infinitives. Restructuring and Clause Structure. Berlin: Mouton. Zanuttini, Raffaella. 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford: OUP. Zeijlstra, Hedde. 2004. Sentential Negation and Negative Concord. PhD dissertation, Amsterdam University. Zeijlstra, Hedde. 2006. Emphatic multiple negative expressions in Dutch. In 77–103. ACLC Working Papers in Linguistics 2, Enoch Aboh & Miriam van Staden (eds). Amsterdam: University of Amsterdam Press. Zeijlstra, Hedde. 2008. Negative concord is syntactic agreement. Ms, Amsterdam University. Available on-line at: http://ling.auf.net/lingBuzz/000645.

part 2

Pronouns, agreement, and topic markers

chapter 6

Weak pronouns in Italian Instances of a broken cycle?*† Diana Vedovato

University of Padova In Modern Italian, a small number of personal pronouns show the typical characteristics of weak elements in the sense of Cardinaletti and Starke (1999): the third person subject pronouns egli, esso, essa, essi, esse and the third person plural dative loro. These elements belong to a formal language register, whereas in spoken Italian speakers tend to avoid them and prefer null or clitic forms instead. The presence of these elements in a pronominal paradigm like the Italian one seems to go against predictions made by certain Economy Principles. We will account for this peculiar situation following Vanelli’s (2003) analysis, i.e. we will look at weak subject pronouns as instances of a broken grammaticalization cycle. Prescriptive rules will be shown to be responsible for the persistence of weak pronouns in standard Italian.

Introduction Pronouns are often involved in grammaticalization processes, in which free forms weaken and then cliticize. The further stage of this cline is a change in functional * I had the opportunity to spend the spring semester of 2008 at Arizona State University. I would like to thank Elly van Gelderen and the participants of the Wednesday ASU reading group for enriching discussions. I am grateful to an anonymous reviewer and to Cecilia Poletto for helpful comments on a first version of this paper. All the remaining errors are obviously mine.

† Abbreviations:

AGR=agreement; C=complementizer; COLL=Colloquial register, DP=determiner phrase; LIP=Lessico di Frequenza dell’Italiano Parlato; SubjP=Subject of the Predication phrase; OVI=Opera del Vocabolario Italiano; TLIO=Tesoro della Lingua Italiana delle Origini; TP=Tense phrase; V=verb. Italian pronouns will be glossed reporting the Italian form and indicating the syntactic category (C=clitic; S=strong; W=weak); gender, number and case have been reported only on clitic forms. Italian null pronominal subject are glossed putting the corresponding English pronoun in round brackets. In all those cases, the person feature is distinctly realized by the verbal morphology, which I will not report.

 Diana Vedovato

category, i.e. clitics becoming agreement markers. When such a process starts, new elements often enter the paradigm to fill the gap left by the previous, now weakened, forms1. As has already been noticed, cycles do not take place in a random way, rather “there is a universal tendency for pronouns to be diachronically reanalyzed in a given order: strong>weak>clitic” (Egerland 2005: 1106). Studies on grammaticalization have pointed out that the mechanisms responsible for change are to be found in language acquisition2: children acquire the setting parameters of a language on the base of the linguistic input they are exposed to; whenever the input they receive is ambiguous, the interpretation will be driven by computational principles of Economy, such as the Head Preference principle: without clear counterevidence, full XPs will be reinterpreted as heads. As a consequence, new full phrases will replace the gap. In this “tug-of-war between economy (Spec to Head) and innovation (reintroduction of specifiers)” (van Gelderen 2004: 12), prescriptive rules, which are usually conservative, play a crucial role in modifying the changes. What I would like to show in this article, starting from Vanelli’s (2003) analysis, is that the few weak subject pronouns found in Standard Italian are hints of a broken grammaticalization cycle which involved subject pronouns in Romance varieties that were losing V-to-C movement. Prescriptive forces had a great role in keeping these elements in the paradigm, along with the plural dative loro. The result is that weak pronouns are used only in written and formal varieties of Italian and speakers do not always handle them with ease. In Section 1, I will start with a brief presentation of the theoretical framework that is adopted. Section 2 provides a quick overview of the Italian pronominal system. In Section 3, I will introduce two Economy principles that are relevant to linguistic cycles and, in Sections 4 and 5, I will show that Italian weak pronouns are instances of a broken cycle, emphasizing the role of prescriptive grammar in the matter. In Section 6, I will conclude with some data from an experiment carried out in a secondary school. These data show that children acquire weak pronouns quite late through education. The same data show that there is evidence of new cycles taking place in the language, but, once again, we do not know if and how these changes will be modified or stopped by standard rules.

1. Although very common, it is not always the case: sometimes the free form remains, while a homophonous one weakens. For example, in Friulan, a dialect spoken in the North-East of Italy, 1st and 2nd person singular subject clitic pronouns morphologically derive from the strong forms, which, in turn, have been maintained (indeed, both derive from the Latin nominative forms). 2. As Roberts & Roussou explain, “the goal of acquisition is to fix parameter values on the basis of experience – all parameters values must be fixed, but there is no requirement for convergence with the adult grammar (although this happens most of the time).” (Roberts & Roussou 2003: 13).



Chapter 6.  Weak pronouns in Italian 

1. The Italian pronominal system and the theoretical framework The category of personal pronouns includes syntactically different elements: some of them move in the structure as full DPs (although with some differences with respect to lexical DPs), others can surface only in few specific positions while they are banned from others (such as topic, focus, etc.). These elements are identified either as weak pronouns or clitics (Kayne 1975; Holmberg 1991). In the late 1990s, Cardinaletti and Starke (1999) proposed a hierarchical tripartite classification which could explain the different behavior of pronominal elements and capture the structural relation between weak and clitic elements. This hierarchical relation is sketched in (1): (1) a. Strong pronouns > deficient pronouns b. Strong (S) > weak (W) > clitic (C)

(= weak + clitic)

This classification is visible at all grammatical levels, across which strong, weak and clitic pronouns display different properties. Italian personal pronouns mainly fit in two of the three classes identified in (1), namely the strong class and the clitic one. They preserve case distinctions and oppose subject to complement forms, with clitics bearing case distinctions as well (accusative and dative). Null pro is commonly classified as a weak pronoun.

1.1

The Italian pronominal system

Italian is a consistent Null Subject language3. The realization of the null subject (pro) is regulated by pragmatic and syntactic factors: if its antecedent is ‘expected’ or prominent in the discourse4, and the pronoun surfaces in an unmarked syntactic position, it will receive a null spell-out. Otherwise a strong pronoun will be used. In sentence (2) with an unmarked reading, in order to have co-reference between the subject of the main clause [Marco] and the subject of the subordinate clause, the pronoun must be null, otherwise the overt pronoun can only refer to the direct object [Luca]: (2) Marcoi ha colpito Lucaj perché luij/Øi voleva rubargli i soldi. `Marcoi hit Lucaj, because hej/Øi wanted to steal his money.’ Being intrinsically thematic (see e.g. Sigurðsson 2008), first and second person pronouns have a slightly different distribution than third person with respect to strong 3. There is only one case in which the 2nd person singular is obligatory, namely in the present subjunctive, where there is no verbal morphology that differentiates the singular persons. Tu in that case displays the properties of weak pronouns. (see also Cardinaletti 2004: 126) 4. The notion of pragmatic relevance in the discourse has not been satisfactorily formalized, yet. Nonetheless, it is clear that the distribution of strong/deficient pronouns is the result of the interaction of pragmatic and syntactic factors. I direct the reader to Calabrese (1980, 1985) and subsequent works and Frascarelli (2007).

 Diana Vedovato

and deficient forms. This paper primarily deals with third person pronouns, hence I will not investigate this aspect of the question. In general, in all syntactic positions where deficient pronouns are not allowed, only the strong counterpart is allowed, and this is true for all persons: (3) a. Ho chiamato io la polizia, non Luca have called I the police, not Luca ‘It was me who called the police, not Luca.’

(contrastive focus)

b. *Ho chiamato la polizia, non Luca. have called the police, not Luca. ‘I called the police, not Luca.’ The ungrammaticality of (3b) only concerns the interpretation. If the pronoun is not phonetically realized, the only possible interpretation for (3b) is a contrastive focus on the object [la polizia]: ‘I called THE POLICE, not Luca’. Italian complement pronouns also fall into two distinct classes: strong pronouns and clitics. Strong pronouns syntactically behave as full DPs and can be used in PPs. Clitics are always adjacent to the verb and cannot occur in isolation, coordination or in syntactically marked structures, such as Cleft sentences, Focalizations, Clitic Left Dislocation (see Kayne (1975) for tests of clitichood). As in all other Romance languages, Italian complement clitics include an accusative and a dative series, both complete for all persons. Table 1 reports the forms of the paradigm: Table 1.  The Italian pronominal paradigm Subject

1st sing. 2nd sing. 3rd sing.

1st pl. 2nd pl. 3rd pl.

Complement

Strong

Strong

Object clitics

Dative clitics

io tu lui (m; + human) lei (f; +human) noi voi loro (m/f; + human)

me te lui (m; + human) lei (f; +human)

mi ti lo (m) la (f)

mi ti gli (m; coll.m/f) le (f)

noi voi loro (m/f; + human)

ci vi li (m.) le (f.)

ci vi gli (coll.; m/f)



Chapter 6.  Weak pronouns in Italian 

2. Out of the paradigm: weak pronouns As we have seen, the Italian pronominal paradigm is made up of two complete series of strong and deficient pronouns, both for subjects and complements. Nonetheless, there are some elements which do not belong to any of those classes, namely egli, which is used in subject position to refer to a third singular masculine [+ human] referent; essa, third singular feminine pronoun, which can refer to [+/– human] referents; esso, third singular masculine [– human]; esse third plural feminine [+/– human]; essi, third plural masculine [+/– human] and the third plural dative loro. These elements display the typical properties of weak elements. Even if disyllabic, weak pronouns cannot be used in all the syntactic contexts in which strong pronouns can be used. For example, they cannot be isolated (4a) or coordinated (4b). (4) a. Chi ha rotto il vaso? -*Egli vs -Lui `Who broke the jar?’ - *egliW vs -LuiS b. *Egli e suo fratello giocavano spesso con noi. vs Lui e suo *egliw and his brother played often with us. luiS and his fratello… brother… Moreover, they cannot be used in a marked position such as cleft sentences (5), nor can they be modified (6):5 (5) *Sono essi che mi hanno detto di venire vs Sono loro che mi hanno Are essiW that told me to come vs Are loroS that told detto di venire. me to come. ‘It was them that told me to come.’ (6) *Essi due non verranno vs Loro due non verranno. EssiW two will not come. vs LoroS will not come. `The two of them will not come.’ The weak dative loro, on the other side, can refer both to human and nonhuman antecedents, a typical semantic property of deficient elements, whereas its strong counterpart a loro is limited to a [+ human] reference. (7) a. Ho dato loro da bere have given loroW to drink ‘I gave them something to drink.’

(loro= to the children; to the flowers)

5. The only two elements that can modify a weak pronoun (but not the dative loro) are anch’ (also) and stesso (-self).

 Diana Vedovato

b. Ho dato da bere a loro (a loro= to the children; *to the flowers) have given to drink to themS ‘I gave them something to drink.’ Nevertheless, weak pronouns pattern in some respects with the strong series, for example they are not forced to be adjacent to the verb: (8) Essi, sapendo che il capitano li aveva traditi, lasciarono il paese. EssiW, knowing that their leader had betrayed them, left the country. ‘Knowing that their leader had betrayed them, they left the country.’ Weak pronouns do not form a complete series, rather they are isolated elements that create an asymmetry in the paradigm. We will provide data that can show two facts, among others: on the one hand, the use of these pronouns is limited to some formal language registers, on the other hand, in colloquial Italian, speakers reorganize the paradigm in a different way.

2.1

Written Italian

Egli/esso/essa/essi/esse and the dative loro are mainly used in written and formal Italian. Here are some examples from two national newspapers with the antecedents underlined: (9) Tra le tante dichiarazioni di queste settimane rilasciate dal presidente della Regione una ci è sembrata particolarmente interessante. Egli, parlando col governatore del Veneto, ha riconosciuto che nell’ Isola «non sempre è stato fatto un uso virtuoso dell’ autonomia. Among the several statements made in these weeks by the president of the Region, one seemed to us really interesting. He, talking with the governor of the Veneto region, recognized that… (from La Repubblica.it, 08/24/08)

(10) In un angolo sperduto al confine tra la Svezia e la Norvegia semi di pecci, una conifera sempreverde, attecchivano tra le rocce. Da essi sarebbero nati rigogliosi alberi che sono sopravissuti fino ai nostri giorni. In a remote corner on the border between Sweden and Norway seeds of pecci, an evergreen conifer, rooted among the rocks. From them, gorgeous trees would have sprouted… (from La Repubblica.it, 04/14/08) (11) ma certo ci induce a domandarci chi sono i nostri ragazzi. Li conosciamo veramente? [...] Non avremmo dovuto dare loro una forma di educazione per aiutarli almeno a vivere meglio certe situazioni? But certainly it leads ourselves to wonder who our guys are. Do we really know them? […] Shouldn’t we have given them a sort of education to help them to live some situations? (from Corriere della Sera.it, 04/13/08)



Chapter 6.  Weak pronouns in Italian 

Looking carefully at the examples above, another consideration is due: in both (9) and (10) the use of the weak pronoun is obligatory. In (9), egli could not be replaced by the null subject. In (9)’ the sentence is rewritten omitting the pronoun: (9)’ Tra le tante dichiarazioni di queste settimane rilasciate dal presidente della Regione una ci è sembrata particolarmente interessante. Parlando col governatore del Veneto, Ø ha riconosciuto che nell’Isola «non sempre è stato fatto un uso virtuoso dell’ autonomia. ‘Among the several statements made in these weeks by the president of the Region, one seemed to us really interesting. Taking with the governor of the Veneto region, Ø recognized that…’ In (9)’, since [il presidente della regione] is not the subject of the previous sentence, and consequently is not accessible by default as antecedent of the null pronoun, the realization of the subject pronoun egli favors the reading. Moreover, the presence of another possible antecedent, namely [il governatore del Veneto] requires egli to precede the parenthetic sentence, otherwise a misunderstanding would arise. Similarly, in (10), essi is the only personal pronoun allowed (otherwise a demonstrative would have to be used), since the zero form is excluded from PPs, while the strong third person plural pronoun loro cannot pronominalize the inanimate antecedent [seeds of pecci]6. As for the plural dative loro, its use is very frequent in controlled productions. Unlike examples (9) and (10), loro in (11) has a possible substitute in colloquial, informal Italian, namely the clitic gli (Non avremmo dovuto dargli una forma di educazione…): the variant does not create any change in the semantics of the sentence, rather it produces a stylistic drop.

2.2

Spoken Italian

Searching a corpus of spoken Italian, the LIP (Lessico di frequenza dell’Italiano Parlato) corpus, out of around 490.000 words, 15 occurrences of esso, 11 of essa, 28 of esse are found7. Only 40 occurrences of egli were found. Interestingly, 33 of them were produced in type D texts, i.e. “unidirectional exchange, with the addressee being present: school lessons (elementary and secondary school), university lectures, sermons, court pleadings”. Although belonging to a spoken language production, these texts are very formal.

6. In colloquial Italian, lui/lei/loro are very often used to pronominalize inanimate antecedents, too, but preferably if performing an agent theta-role (see § 6). 7.

There are no significant differences in the typology of the texts.

 Diana Vedovato

The same uneasiness with egli has been expressed by a quite significant number of informants who were asked to judge the following sentences: (12) Venne anch’egli alla festa. Came also egliW to-the party ‘He too came to the party.’ (13) Anch’egli ha partecipato alla preparazione della lotteria. Also egliW participated to the preparation of-the lottery ‘He too participated in the preparations for the lottery.’ Out of 17 interviewees, 9 of them claimed for (12), and 8 of them for (13), that the sentences are grammatical, but belong to the written register. The dative loro is used by speakers only in very formal contexts, while in colloquial Italian the clitic gli is preferred. Example (14) has been collected from a personal conversation: (14) (Talking about two musicians) Gli mando un messaggio con l’indirizzo. gli-3pl.DatC send a message with the address ‘I will send them a message containing the address.’ The data provided so far reveal that, in spoken Italian, speakers reorganize the paradigm as illustrated in Table 2. In both cases, a structurally more deficient pronoun is preferred to the weak one. In this paper, I will not consider the esso8 series referring to [-human] entities, because in this case a gap in the paradigm arises: since lui (strong) can refer only to 8. For a historical analysis of esso see Renzi (2000). As Cecilia Poletto points out to me, if we assume Cardinaletti & Starke’s generalizations, the semantic specialization of egli and esso could be a problem, because, being both weak, their distribution should not depend on the animate feature of the antecedent. A possible answer to this apparent puzzle lies (once again) in the history of Italian: Boström’s (1972) study on Italian and Florentine third personal pronouns provides ample evidence that, up to the 19th century, both egli and esso could refer to a non human antecedent. The first traces of a semantic specialization appear in Fornaciari (1881)’s grammar: “I PPS nella forma assoluta [=non clitica] non si debbono regolarmente riferire ad altro che a persona od a cosa personificata. Pure usano spesso gli scrittori antichi, e in qualche rara volta sarà lecito anch’oggi, riferire quelli di terza persona anche a cosa” [Pronouns in the absolutive form [i.e. not clitic] are to be referred to a person or personified elements. Ancient authors often used these pronouns to refer to things; nowadays, this usage might be only exceptionally admitted] (D.V.) (Fornaciari 1881 [1974: 59]). Esso could still be used to refer to [+ human] antecedents: “Esso serve a richiamare una persona o cosa poco prima nominata” [Esso is used to refer to a previously mentioned person or thing] (Fornaciari 1881 [1974: 84). Considering that in standard Italian essa, essi and esse are not sensitive to the semantic features of their antecedents, and that, except for egli, no other nominative outcome of the ille paradigm has survived in standard Italian (ella is very rare, elli, elle, eglino, elleno completely disappeared), the semantic specialization of the pair esso/egli is probably lexical, and does not lie in a structural difference.



Chapter 6.  Weak pronouns in Italian 

Table 2.  A comparison of standard and colloquial Italian

3rd singular subject (male) 3rd plural dative

standard Italian

colloquial Italian

ØW/egliW/luiS loroW/a loroS

ØW/luiS gliC/a loroS

[+human] antecedents, there is no personal pronoun bearing a [–human] feature, as (15a) shows, which can fill the syntactic positions reserved to strong elements: thus, the demonstrative pronouns are used in (15b). In (15c), I provide an example illustrating Left Dislocation, which is one of the contexts where only strong pronouns are allowed: (15) a. *Esso non l’ho mai visto! *ItW not lo-3s.m.Obj.C have never seen! `I’ve never seen this.’ b. Questo non l’ho mai visto! This not lo-3s.m.Obj.C have never seen! `I have never seen this.’

(esso=a movie)

(questo= a movie)

c. Lui non l’ho mai visto! (Lui≠a movie; lui=the boy with the blue shirt) HeS not 3s.m.Obj.C -have never seen! `Him, I’ve never seen.’

3. Economy principles In this section I present two principles that are helpful in understanding the grammaticalization cycle that affected personal pronouns and the resettlement of the paradigm as illustrated in Table 2. In the next section, instead, I will look at the history of Italian, and I will show which factors have influenced and modified the natural change of the paradigm.

3.1

Economy of Representation

The Economy of Representation principle, as presented in Cardinaletti & Starke (1999), is stated as follows: (16) Economy of Representations: Minimize Structure. Only if the smaller structure [= the pronoun with the smaller structure, D.V.] is independently ruled out, is the bigger alternative possible. (Cardinaletti & Starke 1999: 47) Being active in synchrony, ‘Minimize structure’ predicts that, wherever the syntactic context allows it, a deficient pronoun will be selected (i.e. a pronoun with ‘lighter’ structure). Otherwise, a strong pronoun will be spelled out.

 Diana Vedovato

3.2

Head Preference principle

Looking at diachrony, many changes seem to obey to another cognitive principle, which captures the tendency of speakers to reinterpret full or deficient XP as heads. The Head Preference principle, as van Gelderen (2004: 11) names it, is found in (17): (17) Head preference or Spec to Head Principle: Be a head, rather than a phrase.

(Van Gelderen 2004: 11)

It predicts that, whenever possible, a syntactic element will be analyzed as a head, rather than a full phrase. The application of this principle to personal pronouns is visualized in (18): (18) strong pronoun > weak pronoun > clitic (> agreement markers9) |_____Spec XP_______| |______  X°  _____| Pronouns can have the syntactic structure of full phrases (=strong pronouns) or of heads (clitics)10; when a process of weakening is triggered, pronouns lose part of their internal structure and change their syntactic category, most of the times showing a morphological reduction, as well (although this is not always the case). Being structurally related in a hierarchical way (Cardinaletti & Starke (1999)), it is plausible to expect pronouns to grammaticalize passing through each level of the hierarchy. The last possible stage, namely the change in functional category, has also been included in (18).

3.3

Consequences

There are at least three consequences of the interaction between (16) and (17): (19) a. In a paradigm with both weak and clitic pronouns, the latter are preferred; b. Weak pronouns could be reinterpreted as clitics; c. Since the shift of category implicates a different set of properties, restructuring the distribution of the pronouns implies that a new pronominal XP will enter the paradigm to fill the gap. A survey of linguistic changes allows us to collect some explicative examples for each point listed in (19). The best piece of evidence in favor of (19a) is the competition between the weak dative loro and the clitic gli in Modern Italian: the clitic, rather than the weak

9. For the cycle involving subject pronouns see van Gelderen 2008. 10. In Cardinaletti & Starke’s analysis, clitics are X°, weak are deficient XP while strong are full XP. In some recent literature, it has been proposed that clitics are not heads, rather remnant XPs. As for now, I will not take this analysis into account, even if it offers good analyses for some phenomena, such as Clitic Left Dislocation. Anyway, it does not invalidate our analysis, because in any case, in an “empty XP with just an X” (van Gelderen 2004: 11), the clitic would be the X°.



Chapter 6.  Weak pronouns in Italian 

form, fits in the paradigm of the spoken language. In the light of (19a), the persistence of sporadic weak elements in the Italian paradigm is not immediately comprehensible. The birth of Romance clitics is instead the result of the kind of process suggested in (19b), since Romance clitics derive from Latin weak pronouns: “weak forms are the source of Romance clitics and can explain most of their syntactic properties (e.g. their position)” (Salvi 1997: 1). The shift from a weak to a clitic category is the process that subject pronouns of old Northern Italian dialects underwent in the 16th century (see e.g. Vanelli 1987 and Renzi 1992), which in turn have been reanalyzed as AGR11 heads (“In the dialects the subject clitic is the strong AGR licensing pro in a Null Subject Language” (Rizzi 1986: 402)12). Analogously, from the weak 3rd plural dative loro in Old Senese a clitic lo’ originated (see Egerland 2005). Needless to say, weakening processes do not necessarily reach the weakest point of the path: there are languages that do not have clitics, rather they oppose a complete series of weak pronouns to a strong one (see Germanic languages in general, where the two series are sometimes homophonous, too). In this respect, the bizarre situation found in the Italian paradigm does not concern the presence of weak pronouns opposed to their strong counterparts, rather, it has to do with the presence of pronouns at different degrees of ‘structural deficiency’ which compete in the same context. The phenomenon under (19c) is a very common one in grammaticalization cycles; for instance, the Latin demonstrative ille has been the source of third person pronouns and of the demonstrative article: this change implied a modification in morphology and a shift in functional category. In order to fill the gap, a new reinforced demonstrative appeared: lat. eccum illum> ita. quello, that. In some Romance varieties the cline proceeded further, for example in Modern French the demonstrative has been newly reinforced by the deictic adverbs -ci, -la (‘here’, ‘there’). The use of strong complement forms as subject pronouns is to be understood as a process of that sort.

4. Subject cycle: Evidence from Old Romance The weak status of egli and the esso series reveals that they come from a strong counterpart. Indeed, in Old Italian, they were strong pronouns, i.e. they were allowed in all 11. Indeed in some dialects, a subject clitic and a quantifier can cooccur (i) Nessuno l’ha detto nulla Nobody 3s.m.Subj.C -has said anything ‘Nobody said anything.’ Further evidence in Rizzi (1986) and therein references.

(Rizzi 1986: 396): (Florentine)

12. Recall also the verbal endings of some dialects, which derive from agglutinated personal pronouns: for example, in the Lugano dialect (Swiss Lombard), the 2ndpl. pronoun is attached to the verb vegnu-f, where the f derives from vos `you.pl’ and the 2ndsing. t in cante-t `you sing’ derives from the pronoun tu ` (Poletto 2000).

 Diana Vedovato

the contexts from which they are nowadays banned (compare the following examples13 with (4) and (6) in Section 2): (20) a.

ed egli ed ella furono trattatori della […] morte del detto re and egliS and ellaS were cause of-the death of-the called king Andreas Andreas ‘And he and she have been the cause of the death of the so-called king Andreas.’ (G.Villani, Cronica, libro 13, cap.99, rr. 16–18 [1348])

comandarono che tutto il mondo in pace vivesse: ed egli due14 they ordered that the whole world in peace lived: and egliS two chiusero le porte del tempio […] closed the doors of the temple ‘They ordered the entire world to live in peace: and the two of them closed the doors of the temple.’ (Bono Giamboni, Orosio, libro 7, cap.7, p.449–450, rr. 20–22, 1) b.

Egli and the esso series had a weak (homophonous) counterpart, too, which was used when the null subject was not licensed, namely when the verb didn’t move to a higher position in the C domain: the subject could be dropped only from a position to the right of the verb15. This condition didn’t involve only third person pronouns, but the system of all persons. Egerland reports that “oltre agli usi evidentemente tonici dei pronomi soggetto, ci sono anche casi in cui i pronomi appaiono con valore di tema non marcato e sono presumibilmente atoni: è il cosiddetto uso debole” [besides the cases where subject pronouns are clearly used as tonic, there are cases in which pronouns display the value of unmarked theme and are presumably atonic] (D.V.) (Cardinaletti and Egeland forthcoming: 9). The pronouns lui/lei/loro were used only as oblique, but they were allowed (not obligatory, though) in some non-canonical subject positions, such as participial and gerundive sentences or when the subject was far away from the verb. A similar system is attested in most Old Romance languages, such as Old French, Old Northern Italian Dialects, and Florentine16. In Old Italian, indeed, “non mancano 13. From Cardinaletti & Egerland (forthcoming). 14. In Old Italian, egli was the form for the third person plural, too. 15. “Quando abbiamo un contesto di inversione, sia in frase principale che subordinata, il soggetto pronominale può essere espresso […] o anche rimanere inespresso […]; in questo caso il posto del soggetto non espresso sarebbe quello che segue immediatamente il verbo” [When we have an inversion context, in the main clause as well as in the subordinate clause, the pronominal subject can be spelled out […], or it can remain unpronounced […]; in the latter case the place of the unpronounced subject should be the one immediately following the verb] (D.V.) (Benincà (forthcoming: 13). 16. See Renzi (1983, 1992) for Florentine, Vanelli-Renzi-Benincà (1985), Vanelli (1987), Poletto (1995) for Northern Italian dialects, Roberts (1993) for Old French, and Benincà (2006) for all Romance varieties.



Chapter 6.  Weak pronouns in Italian 

i casi in cui anche in questo contesto il soggetto, il cui posto sarebbe immediatamente prima del verbo, non è espresso. In questo caso, abbiamo probabilmente a che fare con un registro di lingua diverso, in cui vige il principio del soggetto nullo” [There are also cases, in this context [=subordinate clauses with no V-to-C movement], in which the subject, whose place would be immediately before the verb, is not spelled out. In this case, we are probably dealing with a different language register, where the null subject principle is active] (D.V.) (Benincà forthcoming: 13). When V-to-C movement was lost, many Old Romance languages developed a series of subject clitics (passing through a ‘weak’ phase). In most cases, the oblique forms became the new strong subject pronouns. We can visualize the process in this way17: (Northern Italian Dialects and Florentine) (21) i. ii.

1st stage: CP wo

SubjP

wo

strong prns AgrSP* (Nominative forms) 3 pro Agr’ 3

TP

3

VP

2nd stage CP wo

SubjP

wo

strong prns AgrSP* (Oblique forms) wo weak prns Agr’ (Nominative forms) wo TP wo

VP

17. “AgrSP is the projection in which phi-features are checked on nominative DPs […] SubjP is the projection in which the “subject-of-predication” feature is checked. […] in turn, AgrSP can be split into discrete projections realizing different phi-features.” (Cardinaletti 2004: 121)

 Diana Vedovato

iii.

3rd stage CP wo

SubjP

wo

strong prns AgrSP* (Oblique forms) wo Agr’ wo

SubjCL (Nom forms)

TP

wo

VP

Modern Italian has not developed a similar system: it is a perfect null subject language. Nonetheless, as Vanelli (2003) points out, the fact that egli remained as a weak pronoun and that the oblique form lui is the only strong pronoun available is a clue that a weakening process started, but never ended. Italian shares the first two stages with Northern Italian Dialects and Florentine, but departs from them in the last stage: (22) i. 1st stage: = (21i) ii. 2nd stage: = (21ii) iii. 3rd stage: CP wo SubjP wo strong prns/egli AgrSP* (Oblique forms) 3 pro Agr’ 3 TP 3 VP

(Modern Italian)

The unsolved cycle illustrated in (22) is responsible for the spurious behaviour of egli. There is syntactic evidence for hypothesizing that in pre-subject position egli occupies the same syntactic position as lui, even though unlike lui, egli cannot further move up to other positions. Egli in Spec,SubjP would also account for the fact that egli and pro differ in some respects and are not completely interchangeable.



Chapter 6.  Weak pronouns in Italian 

5. A broken cycle: The influence of prescriptive grammar If the status of weak elements can be explained in terms of a broken cycle, their persistence in the paradigm is understandable in the light of the influence of grammarians in the development of standard Italian. Since the 16th century, grammarians have classified the use of lui/lei/loro in subject position as a ‘grammatical mistake’; the model they conformed to was based on the language of the three great authors of the 14th century, Dante, Petrarca and Boccaccio. As we have documented, in that period egli and the esso series belonged to a different category. But in the 16th century, when the standardization of the national language took place, the spoken language had already traces of the weakening process of egli and esso, and lui/lei/loro were commonly used as subjects18. The force of the authority did not change the situation in the spoken language, but the written register conformed to the norm, which is still active. Vanelli (2003: 1) reports two forceful quotations, one taken from a 17th century grammar and the other from a more recent one: (23) a. “Egli, ed Ella son sempre nel caso retto [= soggetto] […] E’ questo grav’errore a non pochi molto frequente dir, Lui ha fatto.” [Egli and ella are to be used in subject position. Saying Lui ha fatto is a frequent, serious mistake. (DV translation)] (Buommattei, Della lingua toscana, 1643 ) b. “Volendoci attenere alla norma grammaticale, sarebbe opportuno evitare l’impiego di lui, lei e loro al posto rispettivamente di egli (esso), ella (essa), essi ed esse.” [If we want to comply with the grammatical law, it would be necessary to avoid the use of lui, lei and loro instead of egli (esso), ella (essa), essi and esse. (DV translation)] (Battaglia / Pernicone, La grammatica italiana, 1951) All ancient and modern grammars include a section on this querelle, even if they deal with it simply from a stylistic point of view, without capturing the problematic status of weak pronouns and their peculiar syntactic properties, e.g. that they cannot surface in post-verbal position. In the last decades, the norm has relieved under the pressure of the spoken language, but the debate is still open, in particular in school teaching.

18. “La nuova forma lui (col femminile lei) si divulga nel quattrocento (Pulci, Poliziano). I grammatici del XVI secolo (per esempio il Varchi) condannavano un tal uso di lui. Ma nel XIX secolo la vittoria di lui appar definitiva; nel rifacimento del suo romanzo, il Manzoni cambia egli stesso in lui medesimo (Folli, 266). Nell’odierno toscano parlato egli appare ovunque sostituito da lui (cfr. AIS, 651)” [The new form lui (along with the feminine lei) spreads out in the 15th century (Pulci, Poliziano). The grammarians of the 16th century (for example Varchi) condemned such a use of lui. But in the 19th century lui definitely prevailed: in the review of his novel, Manzoni changes egli medesimo into lui medesimo (Folli, 266). In modern spoken Tuscan, egli is everywhere substituted by lui (see AIS, 651)] (D.V.) (Rohlfs 1967, II: 133). See also Boström (1972) and D’Achille (1990, ch. 6) for a survey over the centuries.

 Diana Vedovato

5.1

At which point of the cycle is egli?

In Modern Italian, egli appears as a ‘frozen’ element. Speakers learn it only through written and formal language (this means quite late, at least in elementary school) and the paradigm is split as we have seen in Table 2, with two micro-grammars depending on the linguistic register. Another consequence of this prescriptive policy is that egli is not as weak as it should be19. Although egli and pro are claimed to be both weak (Cardinaletti & Starke 1999: 47), some data suggest that the phonological reality of egli makes pro ‘the smallest’ one. A short questionnaire was submitted to some native speakers of Italian,20 asking them to identify the referent of the pronoun contained in the sentences. I report part of the results: (24) Gianni partirà quando Ø avrà finito il lavoro.

17: Ø = Gianni

(25) Gianni partirà quando lui avrà finito il lavoro. 15: lui = other referent 2: lui = Gianni (26) Gianni partirà quando egli avrà finito il lavoro 8: egli = Gianni 9: egli = other referent ‘Gianni will leave when Ø/egliW/luiS will have finished the work.’ In Cardinaletti & Starke (1999: 47), (27) is coindexed as follows: (27) a. Giannii partirà quando proi avrà finito il lavoro. (Italian) b. Giannii partirà quando eglii avrà finito il lavoro. `John will leave when he will have finished the work.’ Our results show that a significant number of interviewees do not accept this coindexation, and egli seems to pattern more with lui than with pro.

5.2

LoroDAT

In Modern Italian, 3rd plural dative pronominalization is expressed by the clitic gli rather than the weak loro. Prescriptive grammar has not accepted it as a standard feature, although in modern grammars it is quite frequent to find the clitic as a (colloquial) variant for loro.

19. Adopting a split-DP hypothesis à la Giusti (1996), thus considering the DP structure articulated in more that three layers, it would be possible to think that the weakening processes ‘peel’ small portions of structure each time, thus giving rise to structurally different deficient pronouns. This would explain why egli is weak, but fails to fall in some of the relevant predictions on its distribution. I thank Cecilia Poletto for suggesting this hypothesis to me. Unfortunately, I cannot develop it here, but I will take it into account for my future research. 20. All informants were aged 22–28.



Chapter 6.  Weak pronouns in Italian 

Unlike what observed for weak subject pronouns, loro is truly a weak pronoun in the sense of Cardinaletti & Starke (1999) (see Cardinaletti 1991), and its status has remained unchanged since the Old Italian period (Cardinaletti and Egerland forthcoming), apart from the possibility to surface in preverbal position (29), which is lost in Modern Italian (30)21: (29) deono dare noi quello che lloro piacie must give us that that loroW likes. ‘They must give us what they like.’ (Primo libricciolo di crediti di Bene Bencivenni, 7v, par 118) (30) *Io loro ho consigliato di venire I loroW have suggested to come. ‘I suggested for them to come.’ Looking at the Old Italian paradigm, two more weak dative pronouns are found, i.e. lei (to-her) and lui (to-him): (31) E io, rispondendo lei, dissi [… ] And I, answering lei.W, said… ‘And I, answering her, said…’ (Dante, Vita Nuova, cap.33, par.3)22 (32) né agrada lui d’intendere le nostre parole [...] nor likes luiW to-understand the our words ‘Nor does he want to understand our words.’ (Brunetto Latini, Rettorica, p.195, r.17) The interestingly point is that 3rd singular weak pronouns completely left the pronominal system, and only 3rd person singular clitics are used. As we already know loro never disappeared: it shares the paradigm with the clitic gli and their distribution is based on socio-linguistic factors23.

21. In Modern Italian loro can appear between the auxiliary and the past participle, a position not accessible to other elements (the construction belongs to a literary style, though) :

(i) Io ho loro consigliato di venire. I have loroW suggested to come

22. From Cardinaletti & Egerland (forthcoming). 23. From a semantic point of view, the dative loro is perfectly synonymous with gli (indeed, 35i=35ii), without any particular distribution depending on the coindexation with the antecedent as we observed for the subjects.

 Diana Vedovato

A 3rd plural dative clitic i/li is found in Old Italian, too, and it occurred mainly in clitic clusters (Cardinaletti and Egerland forthcoming 15): (33) e credon che la gente / lili ponga i·llarghezza [metta in conto di prodigalità, li consideri segni di prodigalità]... and believe that the people/liCliC put in generosity ‘and they believe that people consider them signs of generosity.’ (Brunetto Latini, Tesoretto, vv. 1478–1479) It must to be said that no weakening process seems to have operated on the weak loro, and therefore nothing similar to what happened to Old Senese lo’ is attested24 (Egerland 2005: 1120). The 3rd plural dative clitic form is the etymological outcome of the original Latin plural dative illis25. The first explicit rejection of the clitic gli denoting plural comes from Varchi (16th century), who forbids the form declaring that Dante did not use it (Varchi 1571: 271). Since then, the standard paradigm has been the one reported in the table below: The problem with Table 3 is that loro is structurally asymmetrical with respect to all other dative clitics. This asymmetry is particularly visible with clitic clustering: (34) a. gli-3s.m.DatC+lo-3s.m.ObjC → glielo V Ho comprato un libroi per Marioj e glielj+i’ho dato. (I) have bought a booki for Marioj and glielCj+i have given b.

le-3s.f.DatC +lo-3s.m.ObjC → glielo V Ho comprato un libroi per Luciaj e glielj+i’ho dato. (I) have bought a booki for Luciaj and glielCj+i have given ‘I bought a book for Lucia and I gave it to him/her.’

(35) a. loro-3plDatW+lo-3s.m.ObjC → lo V loro Ho comprato un libro per i bambini e l’ho dato loro (I) have bought a booki for the childrenj and lCi’ have given themWj Table 3.  Italian dative pronouns m. 3rd singular 3rd plural

f.

Gli

le loro

24. According to Egerland (2005), the trigger for the grammaticalization of lo’ has been morphophonological: the reduced form lo’ ( (pro-drop) > ja jesmĭ dalŭ … > (BE-loss) > jesmĭ dalŭ … The unavailability of (3c) in Old North Russian syntax (Zaliznjak 2004: 178 et passim; Zaliznjak 2008: 248) can be attributed to pragmatic or syntactic restrictions, namely prohibited probably due to a ban on pronoun doubling. Old North Russian (and Old Russian, as well, but to a more limited extent) possesses two sets of pronouns, differentiated by prosody and function (Zaliznjak 2004: 178). Table 1 shows contrasting behaviors between pronominal and verbal pronouns in this dialect. One may well question (i) whether such reanalysis proceeded in a monolithic way within the language, at the same speed and without any variation among dialects, and if not, (ii) what might have conditioned such disparity. As to question (i), as already suggested, I admit that Old Central Russian also realized the BE pronoun interpretations in a restricted manner whereas the northern dialect has completely reanalyzed BE as a pronoun. Evidence for these uneven developments comes from the way variants (3b) and (3c) are distributed. In Old Central Russian, a personal pronoun appears in emphatic cases, (3c). That is, the distinction is made between zero versus personal pronouns, whereas Old North Russian has a contrast between personal pronouns (weak) and verbal pronouns (strong), as outlined in Table 1.  Two kinds of pronouns in Old Russian (mainly in the North) Prosody Origin Function Examples

Full Pronominal Emphatic ja, ty

Clitic Verbal Neutral –jesmĭ, –jesi

 Kyongjoon Kwon

Table 2.  BE in Old Central Russian and Old North Russian Old Central Russian Most common type Ø   jesmĭ dalŭ Second common type ja   jesmĭ dalŭ Contrast Ø vs. pronoun Analysis pro-drop

Old North Russian

vs.

jesmĭ   Dalŭ ja   Dalŭ weak vs. strong pronoun BE=pronoun // no pro-drop for ½

Table 2 (cf. Cardinaletti and Starke 1999). What this table shows is that Old Central Russian still resorted to analytic (periphrastic) forms for past tense marking, while Old North Russian completely switched to a new system, in which l-participles are used as independent past tense marking, as is also evidenced by the rare use of aorist in birch bark letter documents. In other words, the lack of (3c) in Old North Russian and its higher frequency in (3b) in Old Central Russian combined suggest that the BE-pronominalization reanalysis was in full force in the north in contrast to its weak manifestation in the remaining areas. In what follows, I will disregard the marginal status of (3b) in Old Central Russian only with the purpose of emphasizing contrastive behaviors between these dialects for expository convenience. In the next section, it will be shown that the reanalysis in Old North Russian is truly the pronominalization of BE, not simply the case in which a null subject would be retrieved by verb morphology including BE. Language internal evidence is drawn first, and then typologically relevant examples are presented with a diachronic account.

2. Evidence 2.1

Language internal evidence

First, there are many cases where BE is used in an unexpected manner. In (4), for instance, a second person singular BE form jesi occurs even in the presence of another BE, i.e. a second person singular perfective budešĭ. The former BE jesi is a pronoun, functioning as the subject of a predicate. And budešĭ, being an auxiliary in future perfect budešĭ vidělŭ, agrees with it (cf. Zaliznjak 2004: 179). The discontinuity of verbal inflection follows from the fact that the clitic pronoun is placed immediately after the verb. A more evident example is given in (5), where the first person singular present form of a verb is preceded by BE. These examples strongly support the hypothesis that the erstwhile BE is grammaticalized as a full fledged pronoun for all intents and purposes. The pronominal use of BE is much more frequently observed in texts of northwest origin.



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

(4) A ženy ne vidělŭ jesi budešĭ vŭ sně and wife:Gen.Sg not see:pst.msc.sg [be:2Sg]pron will:2Sg in dream:Loc.Sg ‘And you will not have seen a woman in a dream.’ (Old North Russian, Kirik’s Queries, mid-12C) (5) kupilŭ jesmi na torgu, a togo žŭ jesmi ne buy:pst.msc.Sg be:1Sg at market but that:Gen.Sg part [be:1Sg]pron not znaju, u kogo kupilŭ know:prs.1Sg from whom:Gen buy:pst.msc.Sg ‘I bought (it) at the market, but from whom I bought, that I do not know.’ (Old North Russian, Pskov judicial charter, 1467) A second argument is based on chronology. As noted by Ševeleva (2002: 63), in the northern dialects BE was in use until the fifteenth to sixteenth century. Reanalyzing BE as a pronoun in full force, Old North Russian had a more stable system than Old Central Russian. In Old Central Russian, BE=pronoun analysis was minimal, thus very likely to undergo another change, i.e. its loss, when faced with the generalization of lparticiples for a preterit. Sentences (5) and (6) are from the fifteenth century, with BE used as a pronoun. (6) jesmĭ nyně na dorogu jexati xoščü [be:1sg]pron now on road:Acc.Sg go:inf want:prs.1sg ‘I will hit the road now.’ (Old North Russian, III Pskovian chronicle, 1473) Lastly, the negative particle ne as a rule comes after the verbal pronoun and before the l-participle, since the “verbal” subject would be negated instead if the particle came before the BE-verb *něsi (cf. Xaburgaev 1978: 46). The same goes for (5), in which a subject jesmi (read jesmĭ) scopes over the negated verb projection. The position of jesi in (7) and jesmi in (5) support their subject status, which is possible only through its (pro)nominal property. (7) jesi ne vzjalŭ [be: 2Sg]pron not take:pst.msc.Sg ‘you did not take….’ (Old North Russian, BBL 109, 115/5-121/5) Given the evidence presented above, it becomes clear that BE verbs in these contexts were clearly pronouns in Old Russian to varying degrees: the reanalysis of BE into a pronoun was observed in the north with much greater intensity.

2.2

Crosslinguistic evidence

In this section, I will present crosslinguistic data regarding the change from copulas to pronouns, though a discussion of the BE-copula lies beyond the scope of the examination. To the best of my knowledge, an exact parallel to the Old North Russian reanalysis of BE into pronoun is not represented in the literature.

 Kyongjoon Kwon

The first set of data concerning the change from copula to pronoun is the so-called pronominal copula, which is exactly a mirror image of pronominalized BE in its developmental path. This near-typological correspondence is well attested in several languages, such as Arabic, Hebrew, Polish and Scottish Gaelic (see references in Citko 2008). Polish, for instance, has three kinds of copula sentences without substantial semantic change. Note that the pronominal copula to is glossed as TO for the sake of convenience. (8) Polish: ‘Jan is my best friend’ a. Jan jest moim najlepszym przyjacielem Jan be.3Sg [my best friend]:Instr.Sg

verbal copula

b. Jan to mój najlepszy przyjaciel Jan to [my best friend]:Nom.Sg c. Jan to jest mój najlepszy przyjaciel Jan to be.3Sg [my best friend]:Nom.Sg

pronominal copula dual copula (Citko 2008: 262–263)

Based on typological data similar to Polish, Li and Thompson (1976) argued convincingly that copular elements may evolve from anaphoric pronouns through the reanalysis of topic-comment structure into a subject-predicate construction, (9). I assume that the Polish case followed this path, i.e. topicalization of the subject followed by a resumptive pronoun (Diessel 1999: 33–34, Rutkowski 2006; for a different proposal, Citko 2008: 263).7 (9) [topic NP1] [comment Pronoun NP2] → [subject NP1] [predicate Copula NP2] More direct parallels are found in Katz (1996). She reports on two cases of copula-topronoun innovation, one in Turkish and the other in Hebrew (pre-proto-Semitic to Biblical).8 To take an example, a Turkish word equivalent to BE is olmak, the semantics of which is roughly ‘to be, to become, to happen, to mature’. Through semantic attrition, its third person singular form ol came to function as a pronoun of the same number. To take her paradigmatic example, a question like ‘Who is Turkish?’ can be very conceivably answered as ol ‘he is’. Katz sees two possibilities of interpreting this surface form, from which a proposed change is derived. (10) Surface form: Turkish Ol a. [ol-Ø3 sg]v : intended original parsing b. [ol3 sg]pron + [Øbe:3sg]v : innovative parsing (Katz 1997: 130, with modification in representation) 7. It is worth noting that two copulas can co-occur as in (8c) and a verbal copula only dictates the instrumental case for complements as in (8a). But to account for them is beyond the scope of this paper. 8. Frajzyngier (1987) shows a similar change in Mupun, a West Chadic language, from a verb (‘to come’) into an anaphoric demonstrative (in this case, a proximate deictic).



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

This reanalysis was made possible, as Katz notes, by two factors. First of all, the null subject property for a third person singular form allowed a single word to make a sentence in the language, (10a). Second, another parsing could be possible due to the presence of a null copula, [Øbe:3sg] v. What results from this reanalysis is a change from a bi-morphemic word into a two-word sentence. What is implied in this reanalysis is that a change of any direction, from (10a) to (10b) or vice versa, would be possible, since (10b) can also potentially be parsed as (10a). The pronominalization of BE in Old North Russian and the like in Turkish challenge a widely-assumed but hotly disputed uni-directionality hypothesis of grammaticalization theory. To take an instance, Diessel (1999), who has extensively investigated how demonstrative pronouns develop into grammatical markers based on typological work, does not spare a sentence for the change in reverse order. The only hint that can be extracted from Diessel’s data pertaining to grammaticalization is that the pronominalization of BE may be seen as a mirror image phenomenon of exceptions to his grammaticalization data, which are variously called degrammaticalization (Norde 2008), degrammation (Andersen 2006a, 2008), anti-grammaticalization (Haspelmath 2004), and so on. Since pronouns and tense markers are essentially functional categories, it is not easy to determine to what kind of change the Old North Russian case belongs by a traditional notion of grammaticalization. Moreover, an auxiliary is no less grammatical (or functional) than a pronoun, nor is a pronoun less grammatical than an auxiliary. Therefore, instead of searching for a best fit in the massive realm of grammaticalization study, I will just mention a few contributions which may be interpolated to understand the Old North Russian example better. First, Andersen (2006a, 2008) projects grammaticalization into a broad context of language change and provides a series of reanalysis patterns, grammation, regrammation and degrammation, analogous to Roman Jakobson’s (1931) phonologization, rephonologization and dephonologization. And he specifically treated the BE=pronoun as involving two changes: while the reanalysis of present perfect as general past is considered as regrammation, the change from auxiliary to personal pronoun is seen as an instance of degrammation, that is, as a change by which an expression loses a part of its grammatical content through reanalysis (Andersen 2008). It is undoubtedly true that the auxiliaries “retain their content as person and number markers and become allomorphs of personal pronouns”, but it does not tell why such reanalysis could have happened. Katz’s (1996) proposal is interesting in this respect, in that she strives to find a cognitive ground common to pronouns and copulas. Capitalizing on the obvious fact that a third person pronoun is distinct from a first or a second person by requiring a greater level of abstraction and a presupposed existence of a participant, Katz claims that third person pronouns are cognitively related to predication of existence. She furthermore claims that copulas and third person pronouns code the same concept, ‘existence in time and space’, under different grammatical guises, one nominal and the other verbal. This might be essentially true but the purported cognitive contiguity of deixis and predication does not seem to be crucial to triggering the reanalysis in Old

 Kyongjoon Kwon

North Russian. In the following section, I will instead propose a formal analysis, in which the syntactic contiguity reflected in the linearization of elements and feature specifications of these two categories are the keys or triggers to condition the BE=pronoun analysis. In order to account for the earlier adduced Turkish and Hebrew examples, Katz proposed a term cyclical grammaticalization to account for the bi-directional nature of changes between pronouns and copulas. Taking only the first portion of this terminology seriously, I will give a full account of the cyclic nature of phenomena involving the BE=pronoun analysis in Section 3.2. I will not endorse any specific stand on the relationship of this specific analysis to the study of grammaticalization theory. Throughout the paper, I will adhere instead to a more general, not specific hypothesis-laden term reanalysis, which is assumed to be both fundamental cognitive ground and a major source of language innovations (Timberlake 1977; Andersen 2001, 2008).

3. Reanalysis in a syntactic framework 3.1

Pro Taraldsen’s generalization

In this section, I will provide a syntactic analysis of the mechanism by which Old North Russian and Old Central Russian diverged in the treatment of BE. In Late Common Slavic and Old Russian, verbal paradigms were rich enough to indicate what they are predicated of. In other words, the phi-features of verbs were sufficient to identify an unspecified pronoun pro, thus revealing a well-established correlation between rich verbal agreement and a null subject parameter. This correlation is known as Taraldsen’s generalization, the intuition of which has enjoyed large acceptance in scholarship (Taraldsen 1978). For instance, Rizzi (1986) reformulated it as “pro-module” by which pro must be licensed and identified by. Despite its descriptive elegance, there are reasons to question its validity as generalization. First of all, one may well wonder how rich is `rich enough’ to condition a null subject parameter setting.9 Secondly, there are a group of languages that allow pro-drop not on morphological grounds but by discourse (topic drop, or radical pro-drop by Huang 1984). Moreover, Holmberg (2005) recently pointed out that Rizzi’s “pro-module” is untenable in current Minimalist syntax, stating that by definition, uninterpretable phi-features in T cannot identify pro, which itself is an unspecified pronoun. My solution to this impasse is to assume that in null-subject languages the phifeatures of T are interpretable, thus pro Alexiadou and Anagnostoupoulou 1998, Barbosa 2007, but contra Holmberg 2005 and Roberts 2007. If pro is parameterized instead, so that in null subject languages pro has interpretable features, the null subject 9. Some proposals are advanced to measure the “richness” of verbal agreement to rescue Taraldsen’s generalization, for instance, Vikner (1997), Müller (2005), and Tamburelli (2006).



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

parameter will be reduced to the stipulation of pro, divorced from verbal inflection. The T-parameterization hypothesis adopted here is based on the following assumption: in consistent null subject languages morphological verb agreement is “pro(nominal)” to the extent that it has a D/N feature capable of checking the EPP via V raising to T (cf. Borer 1989, Alexiadou and Anagnostoupoulou 1998, Platzack 2003, and Barbosa 2007). In other words, the EPP is not checked by A-movement of subjects to [Spec, TP], as seen empty in (11). (11) Derivation of Old Central Russian perfect dalŭ jesmĭ TP T’ T dalŭ

Asp jesmĭ

Asp

vP

da-

-l

v’ v

…





The most common variant of the Old Central Russian perfect, (3a) dalŭ jesmĭ, is derived from the cyclic movement of a verb head (for a similar clausal structure for the Early English perfect, see McFadden and Alexiadou 2008). The first movement to AspP, , is motivated to obtain participial morphology –l (cf. Embick 2004). By the second movement, , the tense marker BE becomes cliticized to a verb head, yielding dalŭ jesmĭ. When a phonological word is present adjacently, in the [Spec, TP] or in C positions, however, an enclitic jesmĭ moves out of the phrase and cliticizes to its host, a prepositional phrase vŭ tomĭ in (12a) and a wh-noun cto (should read as čto) in (12b). (12) a. A vŭ tomĭ esmĭ ostalŭ but in that:Loc.Sg [be:1sg]pron remain:pst.msc.Sg ‘But therefore I remained.’ (Old North Russian, bbl 724, 124/5) b. upomni cto jesmĭ platilŭ … remember:imp what:Acc [be:1Sg]pron pay:pst.msc.Sg ‘Remember what (=how much) I paid …’ (Old North Russian, bbl 140, 142/5)

 Kyongjoon Kwon

My assumption that phi-features of T in null-subject languages are interpretable is based upon the rich information provided by the phrase merged with T position, i.e. dalŭ [i Number], [i Gender], jesmĭ [T], [i Number], [i Person]. In addition, I assume that number and gender features on l-form are checked upon the merger of AspP with T position, along with the person and number features on BE, which is evidently located in T. Turning to Old North Russian, the reanalysis of the perfect participle as past tense restructures the position of dalŭ into T, displacing jesmĭ. This reanalysis eliminates AspP from the structure. I submit that BE is clitic-adjoined to T, which is now occupied by an erstwhile l-participle, hence (13a) verbal clitic analysis. A crucial difference from the Old Central Russian case is the categorical status of BE, which I claim to be a pronoun in Old North Russian but an auxiliary in Old Central Russian. With the cliticization of BE=pronoun T’s uninterpretable phi-features are valued (for the idea of T serving as a host of clitics, see Terzi 1999). This assumption is also compatible with my efforts to preserve Taraldsen’s generalization, in that Old North Russian verb morphology becomes enriched to contain the required features for dispensing with the need of an overt subject. (13) Pronominalization of be in Old North Russian TP T’ T T dalŭ [T] [u Number] [u Gender]

vP CL jesmĭ [i Person] [i Number]

v’ v

…

b. pronominalization

What this cliticization ultimately implies is the emergence of a pronominal agreement pattern (cf. Jelinek 1984). In other words, verbal agreement in Old North Russian bears phi-features as [person] to dispense with the phonetic realization of subject in [Spec, TP]. This claim bears on Platzack’s (2003) account of why languages with rich verb morphology do not always have null subject parameter settings. According to him, languages with anaphoric agreement (e.g., German, Icelandic) have visible subjects, whereas those with pronominal agreement do not. In this respect, one may well claim that Old North Russian has pronominal agreement at least for some verbal



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

paradigms, such as past tense. This claim will be supported by the dialect specific verb formation in 3.2. As a dotted arrow indicates in (14), the reanalysis could go further to the extent that BE achieves the status of a full word, becoming a full fledged pronoun (14b) (recall that there are two kinds of pronouns in this dialect, see Table 1). This independent (non-clitic) use of BE was presented in examples like (4) and (5). (14) Feature transfer of be i. Old (Central) Russian

ii. Old North Russian

i number i gender Asp

i number i person T D

u number u gender T

i number i person D

Once the aorist and imperfect were removed from the inventory of the Old Russian tense system, T became more naturally associated with an l-participle than with BE. This innovation is economical and natural, since languages are not likely to have perfects without past tense. Generalization of the l-perfect as the general past tense marking had the direct consequence of the displacement of BE-auxiliary. One cannot free BE from l-participle, unless the latter achieves an independent status. The transfer of T from BE to the l-form, as indicated by an arrow in (14), renders BE subject to reanalysis. The Old Central Russian tense system is slow on the reception of this innovation, as is represented as still preserving the inherited structure (i.e. with AspP in (11)) from Late Common Slavic. This will become obvious in Section 4.1, where I discuss the dialectal differences in terms of dealing with BE. What is crucial now is that BE ended up carrying a set of phi-features without tense information. From the assumption that pronouns are no more than a Spell-Out of phi-features, there is good reason to conceive of reanalysis of BE into pronouns. This reanalysis is a result of collective operations; first, an uninterpretable number feature can be valued with an interpretable one on BE. Feature valuation in this case takes place under Spec-Head relationship, i.e. BE in [Spec,TP] and an l-form in T°. Second, the D feature in BE may have helped its reanalysis into pronoun. This will be given a detailed account toward the end of this section. Third, I propose that the presence of an interpretable phi-feature [person] makes jesmĭ eligible for occupying [Spec, TP] position under the assumption that [person] is the most prominent feature in reference to humans (cf. for a similar feature analysis on case, McFadden 2004, Sigurđsson 2008).

 Kyongjoon Kwon

Once freed from participles, the BE=pronoun can be used not in adjacency to lforms, but in combination with present tense. The second half of example (5), reproduced here as (15) has a verb in present tense, which specifies person and number in its affixal morphology. The occurrence of jesmi (read as jesmĭ) is nothing else than a subject of the predicate. (15) kupilŭ jesmi na torgu, a togo žŭ jesmi ne buy:pst.msc.Sg be:1Sg at market but that:Gen.Sg part [be:1Sg]pron not znaju, u kogo kupilŭ know:prs.1Sg from whom:Gen buy:pst.msc.Sg ‘I bought (it) at the market, but from whom I bought (it), that I do not know.’ Another crucial consequence of the BE=pronoun reanalysis is its relation to the null subject parameter setting. As pointed out earlier, Late Common Slavic and Early Old Russian had null subject property due to the presence of a D feature in verb morphology, following Alexiadou and Anagnostoupoulou’s (1998) argument. In this hypothesis, EPP feature on T is optionally valued by V-to-T raising, by which an uninterpretable D feature of verbs is checked at T position. I propose that this D feature resides in BE, since V-to-T raising as a way of realizing EPP is most obvious in formation of perfect tenses where BE is indispensable at the position of T. The presence of D feature in BE may have triggered its reanalysis as a pronoun, which is inherently a D-element. The innovation in the tense system, especially the generalization of the l-participle as a past tense, led to the change in the null subject parameter setting and subsequently to the loss of V-to-T raising.10 If my reasoning is right so far, then the lack of V-to-T raising in contemporary Russian receives a straightforward explanation. It is widely assumed that verbs raise only as high as an intermediate functional category, not high up to T position, as the position of an adverb indicates in (16) (cf. Pollock 1989; Bailyn 1995). This property in the contemporary language should be ascribed ultimately to the change in the pro-drop parameter. (16) Ja dumaju, čto Ivan často celuet (*často) Mašu Russian I think that Ivan:Nom often kisses Maša:Acc ‘I know that Ivan often kisses Mary.’ (Bailyn 1995: 58)

3.2

A solution to the so-called ‘l-less perfect’: The subject cycle

Old North Russian has a unique verb form, the so-called “l-less perfect (bezèlevyj perfekt).” On the surface it consists of two evident parts, a bare-stem verb plus a present form of BE. Since their semantics in context undoubtedly refers to past events, they are accordingly given “perfect” as their denomination. However, it has not been well 10. I assume that V-to-T raising did not instantaneously disappear, as is still reflected in the syntactic structure, (13).



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

accounted for why a bare stem was used to refer to past events and whether BE here is an exact analogue to the BE-auxiliary used in various types of perfect constructions. (17) l-less perfects in birch bark documents a. vzę jesme (should read jesmĭ) take:bare stem be:1Sg ‘I took.’ b. a ni posla jesi but not send:bare stem be:2Sg ‘But you did not send…’

bbl 482 (135/5)

bbl 99 (142/5)

c. jesi ododa (should read otŭda) be:2Sg give:bare stem ‘You gave …’ bbl 311 (151/5) (Old North Russian, Zaliznjak 2004: 144) With regard to the “abnormal” formation, Zaliznjak once conceived of this form as a combination of aorist and BE, but then withdrew his earlier position (Zaliznjak 1986: 146, followed by Nørgård-Sørensen 1997: 5) to assume the current view, which states that this is perfect tense/aspect but without –l. Zaliznjak’s initial view on this form relies on the surface shape of the verbal stem. Old Church Slavonic and its Russian successor had the aorist, whose second- and third-person singular forms have a zero ending, when a verb stem ends in a vowel, (18) (Borkovskij and Kuznecov 1965: 269–272; Lunt 2001:  104–107). Moreover, this claim has a great advantage of very naturally accounting for why l-less perfect occurs only with the first- and secondperson singular forms. (18) Aorist formation in Old Church Slavonic   1sg   2sg   3sg

vŭzę-s-ŭ vŭzę vŭzę

vŭzę-ti ‘to take’ 1pl 2pl 3pl

vŭzę-s-omŭ vŭzę-s-te vŭzę-s-ę

The term “l-less perfect”, however, does not tell us about anything more than its surface form, disregarding the mechanism of the verb formation. If it is truly a “perfect”, any explanatory account should address the immanent question of why the past (or perfect) tense morpheme –l is omitted in this specific case. My analysis of this verb formation relies on the main claim of this paper that BE is a pronoun in Old North Russian. In this context, I claim that the bare stem constitutes a verb stem and the following (though sometimes fleeting) BE is a verbal inflection. This claim fits nicely with a pronominal agreement pattern (as represented in (13b) and recast in (19) with an attested example).

 Kyongjoon Kwon

(19)

TP T’ T T posla– [iT]

vP CL jesi [i Person] [i Number]

v’ v

The bare stem for the past tense is well motivated by its approximation to that of infinitive stems. This claim requires some comment on the Russian verb conjugation. The presence of two stems for Russian verb conjugation has been convincingly presented from various perspectives, such as language acquisition, language development, dialectology and so on (cf. Flier 1978; Andersen 1980). According to Flier (1978), for instance, past and infinitive are grouped together, distinct from present and imperative, by the [participation] feature, the semantic corollary of which can be translated into the presence/absence of person marking in verbal forms in the current syntactic terms. In past and infinitive, the [person] feature can be expressed only by syntactic means, i.e. with an overt pronoun, whereas [person] can be easily recoverable from verbal inflection in present and imperative. More than anything else, these two types of verbal stems are visibly distinct in many cases. I will demonstrate this with some self-evident examples. (20) Verbs with prevocalic stem in Contemporary Russian (Andersen 1980: 287, (1a)) ‘to know’ ‘to become’ ‘to live’ Infinitive zna-t’ sta-t’ ži-t’ Past msc sg zna-l sta-l ži-l Present 3pl znaj-ut stan-ut živ-ut Imperative znaj-ø stan’-ø živ’-i In descriptive terms, instead of following the standard formation of perfect, Old North Russian used a default form of the past (with {Ø} instead of {l}) and a BE verb. In other words, the bare stem in l-less perfect is indexical of the past tense and followed



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

by a zero morpheme, which is comparable to the regular past morpheme, {-l}.11 The person/number maker BE is a verbal clitic, whose fleeting nature is similar to that of reflexive particle sja/s’, which later find its fixed place after verbal stems. It is here worth introducing van Gelderen’s Subject Cycle (or Subject Agreement Cycle), which I believe can well account for the cyclic nature of my data. Subsumed in a larger framework, Linguistic Cycles, the Subject Cycle is motivated by special reference to the principle of efficient computation which is crucial in language acquisition (cf. van Gelderen 2004, 2008, Chomsky 2005). At stage (a), nouns whether pronominal or nominal appear in [Spec, TP] position, as in English. In (b), the pronoun moves to the head position, namely T position. This is notably evidenced by colloquial French data. When verbs are coordinated, pronouns cannot be omitted, but should appear as dependent on the verb. In (21), the second conjunct écris must be preceded by a subject pronoun for its grammaticality. (21) a. Je lis et j’écris I read and I-write `I read and write.’

Colloquial French

b. *Je lis et écris (from van Gelderen 2008: 19) In the last stage (c), the pronoun is reanalyzed as agreement in T and this change subsequently requires [Spec,TP] position to be lexically filled. These three stages of changes are structurally represented as follows.

11. Michael Flier suggests another parsing possibility, namely {posla-Øpast-Øphi-feature}, considering l-perfect forms as morpho-phonemically novel but akin to the perfect forms of verbs with unsuffixed stems, i.e. Old Russian nesŭ ‘carried’, vezŭ ‘transported’, mogŭ ‘could’, tekŭ ‘flowed’, and so on. This seems not at odds with our analysis, i.e. (13a). However, the current analysis based on parsing {posla-Øpast-} can better account for an obligatory BE as a person marker without further stipulation. Moreover, Old North Russian, at least in birch bark documents, does not have a nesŭ-type perfect, but instead had a nesle-type for verbs with an unsuffixed stem: svezti ‘to take out’ (svezle 359); priteči ‘to hurry in’ (pritĭklĭ 752); reči ‘to say’ (rekle 748, reklŭ 3); izmęsti ‘to cause harm, mess up’ (izmakle Psk. 6); poběči ‘to run’ (poběglĭ 510); pogybnuti ‘to perish’ (pogiblo 445); ureči ‘to announce’ (uroklŭ 724).

 Kyongjoon Kwon

(22) Stages of the Subject Cycle a. TP DP pron

b.

T’ T

DP



T’ pron-T

VP 

VP



c.

TP T’

[DP] [pron]

TP

(ibid.,8, Figure 2.1.)

agr-T

VP ([…] indicates an optional phrase)

One may wonder whether the changes between subject and agreement make a real circle, i.e. whether a change from (c) to (a) is attested, as proposed by van Gelderen. Givón’s (1984) proposed cline, expanded by van Gelderen (2008) as in (23), does not seem to make a closed cycle. If changes are all geared toward a zero as in (23), there is nothing from which a new cycle can start. (23) a. demonstrative > third person pronoun > clitic > agreement > zero b. oblique > emphatic > first/second person pronoun > clitic > agreement > zero It is not easy to find a change from (c) to (a), at least in the realm of subject/agreement. In van Gelderen (2008) and Lohndal (in this volume) alike, there is no such change reported. This may be connected to some sort of economy principle. In this respect, it is presumably not a coincidence that van Gelderen’s well-known principles of efficient computation, Head Preference Principle and Late Merge Principle, can account for changes only from (a) to (b) and (b) to (c) respectively, leaving the change from (c) to (a) less motivated in terms of economy. I submit that the Old North Russian data can contribute to a better understanding of the Subject Cycle in two respects. First of all, I propose that the cycle can be repeated, thereby making a complete circle or spiral. Second, the direction of change can be in the opposite direction.



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

(24) Old North Russian in Subject Cycle



a. BE=pron, full pronoun [TP [DP jesmĭ] [T poslalŭ [VP …]]]

b. BE=clitic: pron-T [TPø [T poslalŭ jesmĭ [VP …]]]

b2. BE=clitic [TP ø [T posla-jesmĭ [VP …]]]

c1. BE=affix: agreement [TP ø [T poslalŭ-jesmĭ [VP …]]]

c2. BE=agreement [TP ø [T poslasmĭ [VP …]]] (Small letters correspond to van Gelderen’s stages, as in (22))

First, the start of change is c1, in which BE is both a Tense and an Agreement, i.e. having phi-feature checking property. After the innovation in the tense system that has the consequence of removing BE from the past tense inflection, BE became reanalyzed (or re-cycled) as a pronoun. The first stage is proposed to be like (b), in which BE is a clitic adjoined to T (‘verbal clitic’ analysis, (13a)). Further down the road of change, BE may achieve pronoun status (‘pronominalization’ analysis, (13b)).12 The order of reanalyzing BE, (c) > (b) > (a), is exactly the reversed one of the developmental stages in the Subject Cycle. This does not preclude the possibility of the originally proposed and typologically more common order for treating BE in Old North Russian. With respect to the formation of l-less perfects, the analogical base would be the stage at which BE is a pronoun, 12. Analogous developments of the Polish tense systems present similarities and differences which need to be considered in evaluating the Old Russian facts. Due to the lack of space, I will just mention some differences in passing. First of all, BE is a clitic adjoined to virtually any preverbal constituent, as seen below. a. b. c.

Nigdyśmy tego nie myśleli Nigdy tegośmy nie myśleli Nigdy tego nie myśleliśmy never that neg thought-be ‘We never thought that.’

(Andersen 1987: 31, (6))

S econd, our proposal of a bare stem for [past] distinguishes Old North Russian from the Polish case, where BE verbs are incorporated into the past forms but with a full l-participle (cf. Andersen 1987, Zaliznjak et al 2005). Polish developed a verbal clitic analysis, (13a) but BE did not develop into a pronoun. I further suggest that the dual nature – whether a clitic in (a) and (b) or an affix in (c) – of the Polish BE should be ascribed to the presence of two competing analyses, which can be readily represented as (b) and (c) in Subject Cycle, (23) (cf. Franks and Bański 1999).

 Kyongjoon Kwon

i.e. (a). Then, BE=pronoun is either cliticized to T position to render verbal inflection complete, hence (a) to (b), or BE=pronoun can be further reanalyzed as an Agreement, hence (b) to (c), possibly via (b). Whether BE in this case is an affix or a clitic is in fact hard to determine. This, however, does not undermine my analysis since the distinction between affixes and clitics is generally assumed to be scalar rather than discrete (Zwicky and Pullum 1983; Siewierska 2003: 340). As the cohesion between a verbal stem and BE becomes stronger, the latter can be fully reanalyzed as an agreement (cf. “pronominal incorporation’, Jelinek 1984), as evidenced by a morphologically fused but hitherto questionable form poslasmĭ ‘I sent’ (BBL Staraja Russa 39, 144/4). I assume that this is evidently an instance of (c), in which a pronoun becomes an Agreement. If this proposal is on the mark, the Subject Cycle should be either modified to accommodate the bi-directional nature of the changes or to be provided with an exceptional clauses.

4. The loss of null subject 4.1

Third person BE as a modal

In this section, I will first demonstrate how the third person form of BE behaves in comparison with other person forms. Second, I will suggest a hypothesis on why Old North Russian behaved so differently in treatments of BE. This leads to my final claim that this dialect is transitional in changing the null subject parameter, the setting of which changed from positive to negative in the history of the Russian language. First, with l-past forms, third person singular form of BE was evidently not used as an auxiliary, as seen from the two examples in (25a). In the right column, the only licit form is dalŭ without an auxiliary or a pronoun. This contrasts with how first and second person forms of past tense are formed in Old North Russian (see (3)). If third person behaved the same as the other persons, two predictions could be made; (a) dalŭ jestĭ and onŭ dalŭ should be two possible variants, with the former favored, (b) onŭ jestĭ dalŭ should be disallowed as is with *ja jesmĭ dalŭ. The second prediction is borne out as seen in (26.b.ii) but the first one (a) is not. (25) 3rd person singular in Old North Russian past tense a. Lexical noun i. Ivanŭ da-l-ŭ Ivan:Nom.Sg give-pst-msc.Sg ii. *Ivanŭ jestĭ dalŭ Ivan:Nom.Sg be:3Sg give-pst-msc.Sg b. Pronoun i. (*onŭ) da-l-ŭ he:Nom.Sg give-pst-msc.Sg



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

ii. *onŭ jestĭ da-l-ŭ he:Nom.Sg be:3Sg give-pst-msc.Sg iii. *da-l-ŭ jestĭ give-pst-msc.Sg be:3Sg If we assume that the third person BE jestĭ is also a pronoun, a straightforward explanation seems possible; obligatory pro-drop in the third person. We would have a very consistent picture in every person; Old North Russian has two sets of pronouns in every person but has obligatory pro-drop in third person. However, this is not true. First, the third person singular form jestĭ never surfaces in other tenses of verbal inflections, which denies it a pronoun status. Second, it has another function not observed in other dialects. That is, jestĭ in Old North Russian exhibited modal usage as early as the first half of the twelfth century (Zaliznjak 2004: 179–180). For instance, (26) is interpreted as “whether it is how the matter stands that / whether it is true that / he paid all the interest by himself ”. In other words, jestĭ is neither an agreement nor a pronoun, but a modal, which takes the whole proposition under its scope, challenging verifiability of the proposition. (26) ci jestĭ same vĭxu lixvu vŭdale whether:conj be.3Sg self.emphatic [all debt]:Acc.Sg give-pst-msc.Sg ‘… whether it is true that he himself paid all the interest.’ (Old North Russian, bbl 736, 121/2) Therefore it is safe to say that for third person, Old North Russian has only one option, which is a pronominal one, onŭ. This is of course the same as in the remaining parts of the Russian territory. Both in Old North Russian and Old Central Russian, a referential third person pronoun is dropped, thus the picture in (25) is applicable to Old Russian in general. There are two ways to relate in time the modal use of BE to the third person only pro-drop, i.e. to determine which one comes before the other. Though it is hard to prove one option over the other, I propose that this be considered as a case of exaptation, whereby the original distinction has been discarded but the morphological expression has been pressed into service to encode something else (Lass 1990; Smith 2006). That is, the early loss of BE in the third person is presumed to have conditioned the development of modal meaning. This leads to a more fundamental question why only the third person developed such idiosyncratic usage, while defying pronominalization reanalysis. It is a wellknown axiom that first and second persons are different from third person by being participants in speech acts, thus they are highly salient and easily recoverable compared to third person (cf. Vainikka and Levy 1999). For this reason, third person is often considered as not a true person referent (cf. Benveniste 1966). This contrast becomes manifest in various grammatical processes, some representative examples of which are Person-Case Constraint (Bonet 2007) and Romance auxiliary selection

 Kyongjoon Kwon

(Kayne 1993). Old North Russian demonstrates a similar effect, namely the lack of pronominalization in third person. In addition, I suggest that BE as an auxiliary was lost very early in Old North Russian, which may be due to language contact with non-Slavic languages or to the geographical isolation of the northwestern part of Russia from the remaining parts. Past passive participles, by being participles, require an auxiliary. In (26), however, the participle here is used as an independent predicate. In contrast, the use of BE is not totally excluded in Old Central Russian, (27). (26) Žiznobude pogublene u Syčevicĭ Žiznobud:Nom.Sg killed:ppp.msc.Sg at:prep Syčevicĭ:Gen.Sg ‘Žiznobud was killed at Syčevicĭ’s.’ (Old North Russian, BBL 607/562, 112/2) (27) a uže jestĭ všelŭ v našě vętičě and already be:3Sg come in:pst.msc.Sg in:prep our Vjatič ‘He already invaded our Vjatič tribe.’ (Old Central Russian, Zaliznjak 2004: 179)

4.2

Der dritte Weg: How to lose null subjects

Contemporary Russian is not a pro-drop language (see fn. 3). In this subsection I will consider how the pro-drop parameter setting can be altered in the history of the Russian language. Since pro-drop is assumed to be parameterized along the possibility of V-to-T raising, its change should be ascribed to changes in verbal morphology. This is exactly what is generally considered to be a trigger for the loss of null subjects for Brazilian Portuguese. The loss – more precisely restriction in distribution – of null subjects in Brazilian Portuguese has a morphological foundation. In the history of this language, there was impoverishment of the overt pronoun and agreement inventory, which ultimately altered the distribution of pro. As Roberts (2007) notes, the syncretism created between 2sg, 3sg and 1pl, and between 2pl and 3pl due to the loss of 2sg, 2pl, 1pl pronouns and associated verbs forms prevented null subjects from occurring. In (28), (copied from Roberts (2007: (23)), pronominal paradigms and corresponding inflections for the 1st-conjugation verb falar (‘to speak’) in European Portuguese and Brazilian Portuguese are listed. Considering that the European Portuguese paradigm reflects what Brazilian Portuguese used to be, it is not difficult to see the following facts, i) in European Portuguese stage, verbal inflections are distinct for each person and number, whereas Brazilian Portuguese has only three forms for expressing six possibilities, ii) Brazilian Portuguese has a reanalyzed pronominal system, such as você- for the second person and a gente for the first person plural ‘we’ (cf. Rodrigues 2004; Roberts 2007; Martins and Nunes, forthcoming).



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

(28) European Portuguese Brazilian Portuguese Sg Pl Sg Pl 1 (eu) falo (nos) falamos eu falo a gente fala 2 (tu) falas (vos) falais / (vocês) falam você fala vocês falam 3 (ele/ela) fala (eles) falam ele/ela fala eles falam In other words, merger of personal forms in Brazilian Portuguese led to the loss of a D-feature in T, consequently preventing the application of pro-drop. Morphological impoverishment, which has as its consequence a phi-feature neutralizing effect, is the most conceivable motivation for the loss of pro-drop property (cf. Müller 2005: 10). Another possibility is duly given more emphasis by Roberts (2007) who treats null subjects as weak pronouns. While demonstrating how languages can lose the null subject property, he introduced the French case to show that a change can occur in [Spec, TP] rather than in T position. In Old French, null subjects appeared only in verb-second (V2) contexts as in (29), thus a partial pro-drop language. (29) Tresqu’en la mer cunquist la tere altaigne [until the sea] conquered:3sg [the land high] ‘He conquered the high land all the way to the sea.’

Old French (Roberts 2007: 18a)

Briefly put, the V2 property was lost through the reanalysis of the CP-TP boundary and as its crucial consequence the licensing environment for null subjects was removed from Old French syntax. Then, the rise of overt weak pronoun of the je-series is claimed by Roberts to ultimately take over pro. Given the gradual nature of syntactic change, there must be a period of overlap, i.e. the coexistence of partial null subject and weak pronouns. Supposedly, this transitional period, which began in the Early Middle French period (ca. mid-fifteenth century), came to an end by the seventeenth century. Though I do not agree with Roberts in treating pro as a weak pronoun which undergoes syntactic deletion, it is worth noting that French developed a double series of subject pronouns and its weak (atonic) ones replaced pro. As Roberts (2007:  2) puts it, if Taraldsen’s generalization is valid, any language change that affects the “rich” agreement morphology may lead to the loss of consistent null subjects. This prediction is borne out by two exemplary cases; one whose verbal morphology underwent significant erosion as in Portuguese and the other which developed overt pronouns to take up [Spec, TP] position as in French. My data also support this prediction. Interestingly, Old North Russian exhibits both developments observed in the French and Brazilian Portuguese cases. On one hand, like Brazilian Portuguese, Old Russian verbal paradigms, especially with reference to past events, became impoverished to the extent that person distinctions ceased to be overtly expressed. This is the main trigger to the loss of null subjects in Old Russian in general. On the other hand, Old North Russian developed a weak series of pronouns from BE, making atonic pairs for tonic pronouns. I have shown that these verbal pronouns

 Kyongjoon Kwon

may contribute to maintaining the null subject property setting of the dialect. In other words, the Old North Russian verbal paradigms became rich, i.e. interpretable again in phi-feature specification, via cliticization of verbal pronouns to T, (12a). This reanalysis consequently eliminated the necessity of referential pronominal subjects. However, these atonic verbal pronouns could occupy [Spec, TP], as is the case with Middle French, through pronominalization as in (12b). This specific property places Old North Russian into a transitional stage in the development of null subject parameter, in that this dialect came to be a partial pro-drop, i.e. only in third person. Before concluding, let us tackle a fundamental question: why Old Central Russian and Old North Russian show different treatments of BE in every person. In terms of past tense marking related to l-participle forms, Old Central Russian exhibits two grammars; an old grammar composed of dalŭ jesmĭ and ja jesmĭ dalŭ with and without pro-drop and an innovative grammar with ja dalŭ. This situation does not come as a surprise, for syntactic change typically entails coexistence of competing grammars before complete transition (Kroch 1989). There is another reason for the competing grammars in Old Central Russian. As noted earlier, the loss of the BE-auxiliary, when coupled with pro-drop, would have jeopardized the syntactic structure, leaving only predicates in all three persons. A way to express persons had to be maintained and it was by adhering to an old way. In contrast, Old North Russian was quick to accept the innovation in tense system such that a defunct BE was put to a different use than as an auxiliary. And this recycling could not have been better, since it conveyed the person feature. As expected from the previous discussion, Old North Russian does not have two conflicting grammars, but instead a single new grammar, in which past tense is marked only with lforms and its subject can be expressed in two distinct ways.

5. Conclusion I have argued that different distributional facts pertaining to erstwhile l-participle forms result from different reanalysis patterns between Old Central Russian and Old North Russian. The pronominalization reanalysis, suggested by a few scholars in descriptive terms, has been recast in the current syntactic framework, in the hope that the specific reanalysis is provided with motivation within and without the linguistic system of the Old North Russian dialect. In doing so, I have suggested that the pronominalization of BE as such presents a change counter to the direction proposed in van Gelderen’s Subject Cycle. I have also looked at the null subject parameter setting in Old Russian for its developments in history among dialects and suggested that Old North Russian occupies a transitional stage between Old Russian and Modern Russian. The history of Old Russian pro-drop parameter should be investigated (i) in tandem with related issues, such as BE-loss in general – either as auxiliary or as copula –, clitic ordering, developments



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian 

of the past tense or the conditional form of be-verbs into modal usage, etc., and (ii) based on data considering dialectal differences, stylistic variations, discourse functions, etc. This work is just one step towards that larger project.

References Alexiadou, Artemis & Anagnostopoulou, Elena. 1998. Parametrizing AGR: Word order, Vmovement and EPP-checking. Natural Language and Linguistic Theory 16: 491–539. Andersen, Henning. 1980. Russian conjugation: Acquisition and evolutive change. In Papers from the 4th International Congress on Historical Linguistics [Current Issues in Linguistic Theory 14], Elizabeth Traugott, Rebecca Labrum & Susan C. Shepherd (eds), 285–301. Amsterdam: John Benjamins. Andersen, Henning. 1987. From auxiliary to desinence. In Historical Development of Auxiliaries, Martin Harris & Paolo Ramat (eds), 21–52. Berlin: Mouton de Gruyter. Andersen, Henning. 2001. Actualization and the (uni)directionality of change. In Actualization: Linguistic Change in Progress, Henning Andersen (ed.), 225–248. Amsterdam: John Benjamins. Andersen, Henning. 2006a. Grammation, regrammation, and degrammation. Tense loss in Russian. Diachronica 23(2): 231–258. Andersen, Henning. 2006b. Future and future perfect in the Old Novgorod dialect. Russian Linguistics 30(1): 71–88. Andersen, Henning. 2006c. Periphrastic futures in Slavic, divergence and convergence. In Change in verbal systems. Issues on explanation, Kerstin Eksell & Thora Vinther (eds), 9–45. Frankfurt: Peter Lang. Andersen, Henning. 2008. Grammaticalization in a speaker-oriented theory of change. In Grammatical Change and Linguistic Theory. The Rosendal papers [Linguistik Aktuell-Linguistics Today 113], Thórhallur Eythórsson (ed.), 11–44, Amsterdam: John Benjamins. Bailyn, John. 1995. A Configurational Approach to Russian ‘Free’ Word Order. PhD dissertation, Cornell University. Barbosa, Pilar P. 2007. Two kinds of subject pro. Ms. Benveniste, Émile. 1966. Problèmes de lingusitique générale. Paris: Gallimard. Bonet, Eulàlia. 2007. The person-case constraint and repair strategies. In Agreement Restrictions, Roberta D’Alessandro, Susan Fischer & Gunnar Hrafn Hrafnbjargarson (eds), 103–128, Berlin: Mouton de Gruyter. Borer, Hagit. 1989. Anaphoric Agr. In The null subject parameter, Osvaldo Jaeggli & Kenneth Safir (eds), 69–109. Dordrecht: Kluwer. Borkovskij, Ivanovič, Viktor & Savvič Kuznecov, Petr. 1965. Istoričeskaja grammatika russkogo jazyka, 2nd edn. Moscow: Nauka. Cardinaletti, Anna & Starke, Michal. 1999. The typology of structural dependency: A case study of the three classes of pronouns. In Clitics in the languages of Europe, Henk van Riemsdijk (ed.), 145–235. Berlin: Mouton de Gruyter. Chomsky, Noam. 2005. Three factors in language design. Linguistic Inquiry 36(1): 1–22. Choo, Sukhoon. 2003. The Decline of Null Pronominal Subjects in Old Russian. PhD dissertation, Indiana University. Citko, Barbara. 2008. Small clauses reconsidered: Not so small and not all alike. Lingua 118: 261–295.

 Kyongjoon Kwon Diessel, Holger. 1999. The morphosyntax of demonstratives in synchrony and diachrony. Linguistic Typology 3: 1–49. Embick, David. 2004. On the structure of resultative participles in English. Linguistic Inquiry 35: 355–392. Flier, Michael S. 1978. On the velar infinitive in East Slavic. In American contributions to the 8th International Congress of Slavists, Vol. I: Linguistics and poetics, Henrik Birnbaum (ed.), 269–306. Columbus OH: Slavica. Frajzyngier, Zygmunt. 1987. From verb to anaphora. Lingua 72: 155–168. Franks, Steven & Bański, Piotr. 1999. Approaches to `schizophrenic’ Polish person agreement. In Formal approaches to Slavic linguistics: The Seattle meeting, Katarzyna Dziwirek, Herbert Coats & Cynthia M. Vakareliyska (eds), 123–143. Ann Arbor MI: Michigan Slavic Publications. Gelderen, Elly van. 2004. Grammaticalization as Economy [Linguistik Aktuell-Linguistics Today 71]. Amsterdam: John Benjamins. Gelderen, Elly van. 2008. The subject agreement cycle. Ms, Arizona State University. Givón, Talmy. 1984. Syntax: A Functional-typological Introduction. Amsterdam: John Benjamins. Haspelmath, Martin. 2004. On directionality in language change with particular reference to grammaticalization. In Up and Down the Cline – The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Perridon (eds), 17–44. Amsterdam: John Benjamins. Holmberg, Anders. 2005. Is there a little pro? Evidence from Finnish. Linguistic Inquiry 36(4): 533–564. Huang, C.-T. James. 1984. On the distribution and reference of empty pronouns. Linguistic Inquiry 15(4): 531–574. Isačenko (Issatschenko) Alexander V. 1940. Tense and auxiliary verbs with special reference to Slavic languages. Language 16: 189–198. Jakobson, Roman. 1931. Principes de phonologie historique. Reprinted in his Selected Writings, 1(1962): 202–220. Jelinek, Eloise. 1984. Empty categories, case and configurationality. Natural Language and Linguistic Theory 2(1): 39–76. Katz, Aya. 1996. Cyclical Grammaticalization and the Cognitive Link between Pronoun and Copula. PhD dissertation, Rice University. Kayne, Richard. 1993. Toward a modular theory of auxiliary selection. Studia Linguistica 47(1): 3–31. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Lass, Roger. 1990. How to do things with junk: exaptation in language evolution. Journal of Linguistics 26: 79–102. Le Feuvre, Claire. 2004. Le développement de la phrase nominale dans les écorces de bouleau de Novgorod: Copule, auxiliaire et marque personnelle. Revues Études Slaves 75(3–4): 381–401. Li, Charles & Thompson, Sandra. 1976. Subject and topic: A new typology of language. In Subject and Topic, Charles Li (ed.), 457–489. New York NY: Academic Press. Lunt, Horace G. 2001. Old Church Slavonic Grammar, 7th revised edn. Berlin: Mouton de Gruyter. Lindseth, Martina & Franks, Steven. 1994. Licensing and identification of null subjects in Slavic. In Formal Approaches to Slavic Linguistics: The College Park Meeting, Jindřich Toman (ed.), 199–225. Ann Arbor MI: Slavica.



Chapter 7.  The subject cycle of pronominal auxiliaries in Old North Russian  Martins, Ana Maria & Nunes, Jairo. Forthcoming. Syntactic change as chain reaction: The emergence of hyper-raising in Brazilian Portuguese. In Historical syntax and linguistic theory, Paola Crisma & Giuseppe Longobardi (eds). Oxford: OUP. McFadden, Thomas. 2004. The Position of Morphological Case in the Derivation: A Study on the Syntax-morphology Interface. PhD dissertation, University of Pennsylvania. McFadden, Thomas & Alexiadou, Artemis. 2008. Perfects, resultatives and auxiliaries in Early English. Lingbuzz/000588. Müller, Gereon. 2005. Pro-drop and impoverishment. Ms, University of Leipzig. Norde, Muriel. 2008. Classifying degrammaticalization: Case studies. Paper read at New Reflections on grammaticalization 4, Leuven. Nørgård-Sørensen, Jens. 1997. Tense, aspect and verbal derivation in the language of the Novgorod birch bark letters. Russian Linguistics 21: 1–21. Petruxin, Pavel Vladimirovič. 2002. Semantičeskie klassy predikatov v razvitie vida v vostočnoslavjanskom. Russkij jazyk v naučnom osveščenii 3(1): 244–262. Platzack, Christer. 2003. Agreement and null subjects. Nordlyd 31(2): 326–355. Pollock, Jean-Yves. 1989. Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20(3): 365–424. Rizzi, Luigi. 1986. Null objects in Italian and the theory of pro. Linguistic Inquiry 17: 501–557. Roberts, Ian. 2007. Taraldsen’s generalisation and language change: Two ways to lose null subjects. Ms, University of Cambridge. Rodrigues, Cilene. 2004. Impoverished Morphology and A-movement out of Case Domains. PhD dissertation, University of Maryland at College Park. Rutkowski, Paweł. 2006. From demonstratives to copulas: A cross-linguistic perspective and the case of Polish. Journal of Universal Language 7(2): 147–175. Ševeleva, Marija Naumovna. 2002. Sud’ba form prezensa glagola byti po dannym drevnerusskix pamjatnikov. Vestnik Moskovskogo universiteta, Ser. 9. Filologija 6: 55–72. Siewierska, Anna. 2003. Person agreement and the determination of alignment. Transactions of the Philological Society 101(2): 339–370. Sigurđsson, Halldór Ármann. 2008. Externalization: The case of C/case. lingBuzz/000736. Smith, John Charles. 2006. How to do things without junk. The refunctionalization of a pronominal subsystem between Latin and Romance. In New perspectives on Romance Linguistics, Vol. 2: Phonetics, Phonology and Dialectology [Current Issues in Linguistic Theory 276], Jean-Pierre Y. Montreuil (ed.), 183–205. Amsterdam: John Benjamins. Tamburelli, Marco. 2006. Remarks on richness. University College London Working Papers in Linguistics 18: 439–455. Taraldsen, Knut Tarald. 1978. On the NIC, vacuous application and the that-trace filter. Distributed by the Indiana University Linguistics Club, Bloomington IN. Terzi, Arhonto. 1999. Clitic combinations, their hosts and their ordering. Natural Language and Linguistic Theory 17: 85–121. Timberlake, Alan. 1977. Reanalysis and actualization in syntactic change. In Mechanisms of Syntactic Change, Charles Li (ed.), 141–177. Austin TX: University of Texas Press. Vainikka, Anne & Levy, Yonata. 1999. Empty Subjects in Finnish and Hebrew. Natural Language and Linguistic Theory 17: 613–671. Vikner, Sten. 1997. Vo-to-Io movement and inflection for person in all tenses. In The New Comparative Syntax, Liliane Haegeman (ed.), 189–213. London: Longman.

 Kyongjoon Kwon Xaburgaev, Georgij Aleksandrovič. 1978. Sud’ba vspomagatel’nogo glagola drevnix slavjanskix analitičeskix form v russkom jazyke. Vestnik moskovskogo universiteta. Serija Filologija. 4: 42–53. Zaliznjak, Andrej Anatol’evič. 1986 Novgorodskie berestjanye gramoty s lingvističeskoj točki zrenija. In Novgorodskie gramoty na bereste (iz raskopok 1977–1983 gg.), Valentin Lavrent’evič Janin & Andrej Anatol’evič Zaliznjak (eds), 89–219. Moscow: Izdatel’stvo Akademii Nauk SSSR. Zaliznjak, Andrej Anatol’evič. 2004. Drevnenovgorodskij dialekt. 2nd edn. Moscow: Jazyki slavjanskoj kul’tury. Zaliznjak, Andrej Anatol’evič. 2008. Drevnerusskie ènklitiki. Moscow: Jazyki slavjanskoj kul’tury. Zwicky, Arnold & Pullum, Geoffrey K. 1983. Cliticization vs. inflection: English n’t. Language 59: 502–513.

chapter 8

Two instances of a broken cycle Sentential particles in Old Italian Cecilia Poletto

University of Venice Ca’ Foscari I analyze two cases of Old Italian sentential particles whose usage has changed in unexpected ways. I claim that this process has not occurred because the particles have undergone a reanalysis of their categorial properties but because a major change, the loss of V2, came about. Given that the CP is not obligatorily activated in Modern Italian, the two particles have restricted their usage to cases where the relevant CP projection is still available. If the analysis is correct, it shows that reanalysis of single functional items does not depend on the item itself, but can be the consequence of major syntactic changes to which the item readjusts. Once again, linguistic change is confirmed to derive from the complex interplay of various factors.

1. Introduction1 In this chapter, I intend to examine two examples of reduction of a diachronic cycle in Old Italian adopting the framework of diachronic cycles outlined in van Gelderen (2004). The first case is the particle e, which is a topic marker whose distribution has been radically reduced in Modern Italian (where it still plays the role of a topic marker only in very restricted contexts). The second case is the one of the marker sì, which, contrary to what several people have claimed (including myself), is not a real expletive for the SpecFocus position but an anaphor referring to the previous context which sets the sentence inside the context as a continuation of what has been just mentioned. In other words, sì marks the fact that there is no correction in the common assumptions 1. I thank the whole Paduan group for the common fruitful work during these last years, without Paola Benincà,Davide Bertocci, Federico Damonte, Jacopo Garzonio, Nicoletta Penello Diego Pescarini and Laura Vanelli my research life would have been much less interesting and exciting and my personal life much less rich. I also thank Lorenzo Renzi for pointing out to me the etymology the particle sì, which put me on the right track in the analysis of the modern particle. The abbreviations I use here are the standard ones used in the OVI project (see below footnote)

 Cecilia Poletto

between speaker and addressee, but simply that the sentence has to be added to the previous scenario. Modern Italian sì has lost this usage acquiring a new one, still related to Focus, but indicating contrast to the aforementioned context by correction. I claim that in both cases the reduction/change of the particle is due to a very major change in sentence structure, namely the loss of the Verb Second (henceforth, V2) property, i.e. movement of the inflected verb to a Focus projection in the CP layer. I adopt here the view on Old Italian proposed by Adams (1987), Vanelli et al. (1985), and Benincà (1984). The CP layer contains the following projections in the most recent version proposed by Benincà (2006): (1) [HT [SceneSett. [Leftdisl....[ListP [ [contr. cp1 adv/obj, [contr.cp2 circ.adv. [Inform. cp]]] frame theme focus In Old Italian, the verb is forced to move to one of the projections located in the Focus field (i.e. at least to Information Focus; if an XP is located higher, then the verb moves to the respective head as well). The projections above Focus occurring in the theme and frame fields can be filled and give rise to Verb third, Verb fourth orders. If only the Focus layer is realized, the linear order is V2. Therefore, although Old Italian does not respect the linear restriction of V2, it is nonetheless a `hidden’ V2 language, where the inflected verb moves to the CP domain (more precisely at least to the Information Focus projection). The two particles examined are located at the two extremes of the CP layer, e is a particle in the Frame field, while sì is located in the Focus field. It is well known that Italian lost the V2 property after the medieval period: this means that every main declarative clause does not necessarily have to be a whole CP; main clauses can be simple IPs. Furthermore, the inflected verb does not need to raise to the Information Focus head but can stay in the IP domain. Modern Italian even goes further in the loss of access to the CP domain, as the Information Focus position is blocked and only the Contrastive Focus projection remains available.2 Given this change, we could wonder what happens then of elements which originally marked the CP layer: they could either disappear or change into a marker of something else. The particle sì, which was originally an Information Focus marker, is reanalyzed as a Contrastive Focus marker, which is the closest projection still available in Modern Italian. Therefore, the particle has only undergone a minimal change, which is not due to its reanalysis as something new, but is forced onto the particle by the unavailability of the Information Focus projection. The development of the particle e is somewhat different: in Modern Italian it can still be used as a Continuation Topic marker, but only in those contexts in which the clause is an entire CP (for instance, interrogatives or exclamatives). In a way, the particle has not changed at all, it is the linguistic environment around the particle that has changed.

2. As I discuss in Section 4, not all languages that have lost V2 have also lost access to the Information Focus projection. This is a further development whose origin still remains obscure.



Chapter 8.  Two instances of a broken cycle 

More generally, the analysis of these two sentential particles I examine shows that elements which are already functional in the structure can cease to mark a given projection in the CP layer because of independent reasons. This means that at least some of the cases of an interruption of a grammaticalization cycle do not crucially depend on the properties of the element itself but on general properties of the language, namely the loss of the V2 property, as I will argue. This work can be read as a confirmation of the idea that it is the whole system that changes, not a single item or construction. The chapter is organized as follows: in Section 2, I analyze the particle e and propose an analysis which links the behavior of the particle to the fact that it is the coordination head and more specifically to the property that it takes a specifier and a complement of the same category. In Section 3, I show what the distribution of the second particle sì is, and analyze it as a marker linking the sentence to the previous context. In Section 4, I discuss the reason why Modern Italian has drastically reduced the usage of the two particles and derive it from the fact that the CP layer is “less active” in a sense which will be made more precise. Section 5 concludes the article and briefly hints at some research perspectives.

2. E as a topic marker As in Modern Italian, in Old Italian, the particle e is the conjunction head, which can conjoin sentences or phrases. However, it occurs in a number of contexts where it is clearly does not have this function:3 (2)

e quando avea forbiti i piedi ed elli tornava fuori e and when had cleaned the feet and he came.back outside and rinfangavalisi vie più e tornava a ricalpitare il letto. (F.F. 124) got.mudded more and came.back to step.on the bed `When he had cleaned his feet, he went back outside, covered them with mud, came back and went up onto the bed.’

A case like (3) cannot be interpreted as a conjunction, as the first sentence is an adverbial embedded clause, while the second is the main clause. The occurrence of e in these cases must be due to some other reason. If e were the conjunction particle, Old Italian

3. All examples are taken from the online OVI (Opera del Vocabolario Italiano) data base of the CNR (Consiglio Nazionale delle Ricerche ‘National Research Council’) available at the following website http://www.ovi.cnr.it/index.php?page=banchedati which contains all Old Italian texts from 1215 to 1350. Under Old Italian I mean the language of the Florentine texts from 1200 to 1315 approximately, following the standard usage. I use here the standard abbreviation of the Old Italian grammar (to appear): F.F. means Fiori di Filosafi and Nov. Novellino. The sources are indicated in the references.

 Cecilia Poletto

would have the peculiar property of conjoining embedded with main clauses, which is in general not an option in the most well studied Romance and Germanic languages: (3) Stando lo ‘mperadore Federigo e facea dare l’ acqua alle mani Being the emperor Federigo and let give the water to.the hands `While the emperor F. was standing there, he commanded to bring water for the people to clean their hands.’ (Nov. 177) Although punctuation is not a very reliable test in older stages of languages, it is interesting to note that e can occur at the very beginning of a sentence preceded by a full stop, as (4) shows: (4) a. Plauto fue uno grande savio, cortese in parlare. E scrisse queste P. was a great wise, kind in talking. And wrote these sentenze sayings ‘P. was a great wise man, who spoke very kindly. He wrote these sayings.’ (F.F. 104) b.

Scipio Africano fue consolo di Roma e fue tagliato di Scipio Africano was console of Rome and was cut of corpo a la madre e per ciò fue chiamato Cesare. body from the mother and therefore was called C. E dice uno filosafo che quelli che nascono in quel And says a philosopher that those that are.born in that modo son più aventurati way are more lucky `S. A. was console in Rome, he was born through a caesarean and for this reason he was named Caesar. Philosophers say that people who are born like this are luckier.’ (F.F. 140)

A further argument showing that in the cases above e is not a conjunction but a different sort of marker is its translation into Modern Italian: the first e in (5) is perfectly grammatical also in Modern Italian, while the second is completely excluded: (5) e poi, quando tutto ebbe dato, et elli si fece vendere, and then, when all had given, and he himself made sell `And then when he had given everything he let himself be sold.’ (Nov. 162) If e is not a conjunction here, what is it then? In Poletto (2006), I proposed that it is a topic marker. Here I would like to further refine this idea and claim that it is a marker located in the head of the Hanging Topic projection and licensing a null Hanging



Chapter 8.  Two instances of a broken cycle 

Topic, and more precisely the one referring back to the whole previous context.4 The effect of this null Hanging Topic is a sort of `continuation of the same discourse configuration’ (henceforth CDC). To put it bluntly, we can say that e is a continuity marker signalling the fact that the sentence has to be added to the established universe of discourse without further modifications of the scenario. This hypothesis explains why in these cases it always occurs sentence initially. Although Old Italian tolerates several Topic Phrases in front of FocusP, nonetheless, e is always the first element, which can be followed by Topics, but never be preceded by them. If it is a particle licensing a null Hanging Topic, this is exactly what we expect, as Hanging Topics are the highest type of Topic there can be (see Benincà and Poletto 2004 for arguments distinguishing Hanging Topics and Scene Setting elements which occupy a `Frame field’ on top of Left Dislocated Topics). The second phenomenon this analysis explains is the fact that e always triggers enclisis (if it is immediately followed by the verb). (6) a. e tenerlo and keep.it

(F.F. 135)

b e bevenne and drank.of.it

(F.F. 134)

In old and recent work, Benincà observes that that enclisis is not only found when the inflected verb is in first position (according to the well known Tobler-Mussafia law) but more generally when the Focus position is empty, even if there are one or more Topic positions. She analyzes enclisis as a result of verb movement to a Topic position, probably a position in the Theme field (crucially, notice that e is located further up in the Frame field). However enclisis is to be analyzed, the empirical observation (known as Tobler Mussafia law to traditional philologists) remains that the marker e behaves like other types of Topics. The idea that e can be a Hanging Topic marker of the CDC type also explains the fact noted above, namely that e can occur in front of both main and embedded clauses (giving the impression of a strange coordination structure between a main and an embedded infinitival or gerundive clause as in (7)): (7) quando entrò nella chiesa, et uno parlò e disse: when got into.the church, and one spoke and said: ‘when he entered the church, one of them spoke and said…’

(Nov. 189)

4. Notice that there are other languages where sentential particles mark exactly the continuation of a Topic. This is for instance the case of Chinese, where the particle ne indicates the continuation of the same discourse configuration (as in Li Boya 2006).

 Cecilia Poletto

Moreover, the CDC particle e can cooccur with the conjunction head e (thus giving the impression of “conjunction doubling”): (8) e, innebriato il pane dell’ odore che n’ uscia, del mangiare, and, putting the bread in.the smoke that of.it.came.out of.the food, e quelli lo mordea, e così, il consumò di mangiare, ricevendo il and he it bit and so, it finished of eating, getting the fumo e mordendolo. smoke and biting.it `He was putting the bread close to the smell which came out of the meat and then he ate the bread up biting it.’ (Nov. 177) In the example above, the first e is the conjunction element followed by an embedded clause, while the second is the Hanging Topic marker. We can conclude that e is not only the conjunction marker, but can also serve as a particle marking a Continuity (CDC) Topic. However, it is a striking fact that the element is one and the same for both cases. Obviously, I would not like to add to the complexity of the lexical entries of Old Italian and postulate that there are two homophonous e elements in the lexicon, but I will rather put forth the idea that there is a link between the two in the sense that there is only one lexical item e, which can be used either as a conjunction or as a CDC marker due to its categorial properties. The structure I assume for coordination is the one proposed by Kayne (1994) which has by now become standard:

(9) [CoordP XP [coord° e [XP]]

The conjunction head has two properties: a) its specifier and its complement must be of the same category and b) it can take whole CPs as specifier and complement. What I propose for the particle is the following structure in (10): (10) [TopicP ContextCP [Topic° e [[CP ]]] The topic marker e has a specifier and a complement of the same type: in the specifier there is the whole previous context, which is a CP, ad in the complement position there is also a whole CP. From the structural point of view, the only difference between (9) and (10) is in the labels. Thus, e maintains the property of taking a specifier and a complement which are of the same category in all its usages. From this point of view, the fact that the coordination head is used also as CDC Topic marker is not surprising at all. Some authors have already proposed that Hanging Topics can be a whole independent clause with partial deletion of the lexical material inside it (see Garzonio 2005 on this). According to this analysis, the fact that e can function as a topic marker is not a mere coincidence but depends on its original formal property, namely the necessity of taking a specifier and a complement which must have the same categorial status, not only on its semantics or on some other independent mechanism which “creates” new particles.



Chapter 8.  Two instances of a broken cycle 

The proposal crucially relies on the existence of null topics in Old Italian, and indeed we have evidence of other types of null topics in verb initial constructions as the following: (11)

Uno cavaliere pregava un giorno una donna d’ amore e diceale A knight was.praying one day a woman of.love and told.her intra l’intra l’altre parole com’ elli era gentile e ricco e bello among the other things how he was kind and rich and handsome a dismisura, “e ‘l vostro marito è così laido come voi sapete”; beyond measure “and the your husband is so ugly as you know” e quel cotal marito era dopo la parete della camera. ø Parlò e and that such husband was behind the wall of.the room Spoke and disse: “Eh, messer, per cortesia: acconciate li fatti vostri e non said: “Eh, sir, please, mind the facts yours and do.not isconciate li altrui”. (Nov. 231) spoil the others `One day a knight was begging a woman for love and told her among other things that he was kind, rich and very handsome, while her husband was ugly. The husband, who was behind the wall, spoke and said `please mind your own business and not that of others.”’

(12) “Iscrivi” disse quel re cortese “ch’ io obligo l’ anima mia a perpetua Write, said that king kind “that I oblige the soul mine to eternal pregione infino che voi pagati siate”. ø Morìo. Questi, dopo la morte, prison until that you paid are. Died. They, after the death, andaro al padre suo e domandaro la moneta. (Nov. 171) went to.the father his and asked the money ‘“Write”, said the kind king, “I oblige my soul to eternal prison until my debt is paid”. The king died. After his death they went to his father and asked for their money back.’ The whole context preceding the clause starting with the null topic has been added here to show that the V-initial clause indeed has a null element which must be recovered from the previous context. These sentences are not interpretable to contemporary Italian ears. Therefore, although both cases are subjects, the pro drop property is not enough to explain these cases. In cases like the above, there must be a lexically realized tonic pronoun in Modern Italian, which means that Modern Italian does not have null topics (at least not of this type). This fact can in turn explain why the topic marker e is not found in the Modern Italian counterpart of sentences as the ones illustrated in (2) to (8). However, it would be empirically false to assume that Modern Italian has no case of null topics of the CDC type at all. As a matter of fact, e can still be a topic marker in Modern Italian but only in interrogative elliptical clauses and in exclamative clauses. An interrogative is shown in (13). More generally, e acts as a topic marker in

 Cecilia Poletto

cases of special questions (in the sense of Obenauer (2004), where the presence of e can (but need not) licence wh-element in situ, as in (14): (13) a. E io? And I? ‘What about me?’ b. E adesso? And now? ‘Now what?’ (14) a. E cosa potrebbe fare in un frangente simile? rq interpretation And what could do in a case similar? ‘What the hell could he do in such a situation?’ b. E viene quando, allora? And come when, then? ‘When on earth is he coming then?’

Non echo wh in situ

An exclamative example is given in (15): (15) E che vestito che ti sei comprato! And what dress that yourself are bought `What a dress you bought!’ A special construction known as anaphoric anteposition also tolerates e as a topic marker (see Benincà 1988): (16) A: Gianni voleva comprarsi un castello Gianni wanted to.buy.himself a castle B: E un castello si è comprato! And a castle himself is bought ‘He bought a castle indeed!’ This distribution seems to indicate that the topic marker is parasitic on some sort of operator construction. However, this is not entirely correct, because the topic marker is excluded in contrastive focalizations, which are also analyzed as involving an Operator projection in the CP layer: (17)?? E IL vestito ha comprato, non il cappello And the dress has bought, not the hat Therefore, the usage of e as a Hanging Topic marker cannot directly depend on any element being located in the CP layer, nor on any sort of Operator in Focus, but must be restricted to those cases where the context must be relevant and present. The three constructions where e is still used all imply reference to the previous discourse, and at least in the cases of special questions and exclamatives occupy positions in the CP



Chapter 8.  Two instances of a broken cycle 

which are higher than Focus (see Portner and Zanuttini 2003 for exclamatives and Obenauer 2004 for special interrogatives). We can conclude that there are two conditions which must be met in order for e to function as a topic marker in Modern Italian: (a) its presence must be justified by reference to the immediate context, and (b) projections higher than Focus must be activated in the CP layer. In Section 4, I will come back on this second condition and try to make sense of it in terms of diachronic change.

3. The element sì The element sì meaning ‘so’ has several usages in Old Italian, some of which overlap with the adverbial form così, also meaning ‘so’.5 In what follows, I will describe the distribution of sì and compare this with the one of così, showing that they partially overlap, and then concentrate on the CP usage which I intend to analyze in detail. In a similar way to così ‘so’, sì can modify an adjective, an adverb, or a noun: (18) a. fue sì giusto e guardò sì le mani da.. was so right and looked so the hands that… `he was so right, and looked down at his hands so that...’ b. cominciò a tremare sì fortemente … began to tremble so strongly… `He began to tremble so strongly.’

(F.F. 110)

(Vita nuova 6)

(19) a. quando io vi dissi del cavallo cosa così maravigliosa, when I you told of.the horse thing so marvelous `when I told you such an incredible thing about the horse.’ (Nov. 129) b. onde picciolo guiderdone diedi a llui di così ricco insegnamento so small reward gave to him of so rich teaching `I gave him such a poor reward for such a precious advice.’ (Nov. 145) However, differently from sì, così can be a manner adverb, sì cannot: (20) Allora il lapidaro si rallegrò e prese l’ una pietra e miselasi in Then the stone-worker cheered up and took the one stone and put.it in mano e disse così: hand and said so `Then the stone worker cheered up, took a stone in his hand and said:’ (Nov. 124) 5. I do not know whether there is an etymological link between the two forms. Although sì looks like a short form of così, it is usually said to be derived from ‘sic’. Anyhow, sì has to be distinguished from the reflexive clitic si, which has a different etymological source. I will not pursue this question here.

 Cecilia Poletto

On the other hand, sì can occur in the CP before elements like come ‘as’ forming the sequence ‘so as’: (21) a. sì come appare a chi lo intende so how appears to whom it understands `So as it appears to whom can understand it.’

(Vita nuova 23)

b. Sì come elli parlava tra lloro di sì grande maraviglia so how he spoke to them of so great wonder `So as he spoke to them about such a wonder.’ It can also occur in front of the complementizer che: (22) a. sì che quasi dal principio del suo anno non apparve a me so that almost from.the beginning of the year not appeared to me `So that it appeared to me only at the beginning of the year.’ (Vita nuova 6) b. sì che li chiovi pareano 16 lettere so that the nails looked 16 letters `so that the nails looked like 16 letters.’

(Nov. 337)

When used as a CP operator, it can even climb into the main clause to a position located in front of the past participle, which is presumably an operator position: (23) a.

e ho sì saputo fare che li sudditi miei m’ hanno and have so been-able to.do that the subjects my me have cacciato chased.away `I have been so skilled that my subjects sent me away.’ (Nov. 143)

b. a chi mi sa sì pregare che io lo diparta to whom me knows so pray that I him take.away dagli altri from.the others `who can pray to me in such a way that I take him away from the others.’ (Nov. 167) The manner adverb così can also climb to the left of the inflected verb in V2 contexts, as any other adverb, but it never occurs in front of come or che: (24) E Guiglielmo, vedendo che così era sorpreso, parlò e disse And Guglielmo, seeing that so was surprised, spoke and said `And G., seeing him so surprised, spoke and told him...’ (Nov. 225)



Chapter 8.  Two instances of a broken cycle 

We can conclude that sì has access to the CP layer in various contexts, while così can only be contrastively focalized, as low adverbs. The structure which is relevant to the present work is the following one: (25) a. traendomi fuori de la veduta di queste donne sì mi domandò... taking.me out of the sight of these women so me asked `getting me out of those women’s sight, he asked me...’ (Vita nuova 58) b. E parlandomi così, sì mi cessò la forte fantasia And talking.me so, so me stopped the strong phantasy `(while he was) talking to me like that, I stopped dreaming.’ (Vita nuova 98) c. Poi che detta fue questa canzone, sì venne a me uno, … Then that said was this song, so came to me one, … `After this song was sung, a man came to me...’ (Vita nuova 133) d.

La volpe andando per un bosco sì trovò un mulo: e il The fox going through a wood so found a mule: and the mulo sì li Mostrò il piede dritto, mule so her showed the foot right `while the fox was going through the wood, she met a mule, who showed her his right foot.’

In all these cases, sì does not seem to mean ‘in this way’. In Poletto (2005), following Benincà’s (1995) intuition, I analyzed sì as an expletive located in SpecFocus. The structural arguments showing that sì is in SpecFocus are still valid (and will be presented below). However, I would like to propose the idea that sì is not an expletive, but has a meaning, though not exactly the one of ‘so’. It is an element indicating the relation between the clause and the context. Put it roughly, sì signals that the sentence is new information but has to be set against the preceding context. In a sense it is similar to e, though e licenses a Hanging topic which is linked to the preceding context though does not highlight the sentence as new information. Both elements are prosecutive, though sì adds something more: it adds the sentence to the context signaling its relevance. As e, sì is never found to the very beginning of a whole text, where there is no context yet to make reference to. This shows that the element is in itself not an expletive pronoun (contra Poletto 2005 and Ledgeway 2007). If it were a pure expletive, we would not expect this to be case, as expletive es in German can occur at the very beginning of a whole text. The second argument in favor of this idea is that other elements with a similar function (like for instance or ‘now’ in Old French in addition to sì) can be found, while there can only be one expletive per language. Therefore, it seems that the interpretation of sì as a marker which defines the relation between the proposition and the context is more adequate. Moreover, if sì were a real expletive, no V1 would be allowed in Old Italian, (except those cases in which there is a null operator in SpecFocus), as it is

 Cecilia Poletto

the case in Modern German, and the verb would never be in first position with enclisis. The last piece of evidence that sì is not a CP expletive is the fact that it can occur in embedded clauses: (26) a.

che l ferro, se l’ aopere, sì si logora, se no l’aopere that the iron, if it use, so it wears.out, if not it use la ruggine il consuma the rust it destroys `that the iron wears out if you use it, if you do not, it gets rusty.’ (F. F. 146)

b.

Leggesi del re Currado, del padre di Curradino, che, Reads-one of.the king Currado of.the father of Curradino, that, quando era garzone, sì avea in compagnia dodici garzoni di sua when was boy, so had in company twelve boys of his etade, age, `Here you can read about King Currado, Curradino’s father, who had twelve boys with him when he was a boy.’ (Nov. 232)

This is expected if we assume with Benincà (2006) that the CP layer can be activated in embedded clauses as well, but it is not if we consider sì as an expletive, as CP expletives like German es never occur in embedded contexts. Sì is clearly located in the Focus field in Old Italian, as it always occurs at the immediate left of the inflected verb: (27) a. sì s’ abacinò degli occhi so himself burnt his eyes `His eyes were dazzled.’ b.

(F. F. 105)

e, parlando spezialmente alli spiriti del viso, sì disse queste and, speaking particularly to.the spirits of.the face, so said these parole words ‘And talking to him in the face, he said no.’ (Vita nuova 8)

In the Old Italian data base of the OVI enterprise, there are no cases of sì separated by the inflected verb by elements other than clitics or negation, which shows that sì is very low in the structure of the CP. Moreover, sì very often occurs after an embedded temporal clause indicating either anteriority or simultaneity or after a TopicXP or the subject (which is probably a special type of topic). This is expected under this analysis, given that topic and embedded clauses are higher than Focus in the CP layer.



Chapter 8.  Two instances of a broken cycle 

It also often occurs in combination with e and the combination is always e preceding sì: (28) a.

E, che avrà cuore nobile et intelligenzia sottile, sì li And, that will.have heart noble and intelligence subtle, so him potrà simigliare per lo tempo che verrà can look.alike for the time that will.come `and who will have a noble heart and a sutble intelligence, and will look like him in the future.’ (Nov. 118)

b. in questo Pittagora sì cominciò... in this Pittagora so began `Precisely then, P. started...’

(F. F. 104)

The strongest piece of evidence that sì is located in the Focus field is the fact that it occurs only with proclisis. (29) e di ciò sì ne fue and of this so of.it was `and so it was of this.’

(F. F. 106)

(30) sì si ne diede questa penitenza so himself of.it gave this penance `He gave himself this penance.’

(F. F. 108)

Similarly to what has been proposed for e, which only occurs with enclisis, when it is immediately followed by the inflected verb, we can exploit the Tobler Mussafia law as a test to determine the position of sì, which behaves as foci, and not as higher topic elements. Summing up: the hypothesis that sì is a prosecutive operator setting the clause against the context explains: a. Why it can occur in embedded clauses b. Why it can be substituted by other similar elements c. Why it is never found at the beginning of a whole text The fact that sì is an Operator located in the Focus field explains: a. Its adjacency to the inflected verb, b. The fact that it occurs after topics, scene setting temporal elements, and hanging topics, and c. The fact that it only triggers proclisis. As a last point concerning sì, I would like to mention that the element has not disappeared from the language. Modern Italian has indeed cases of sì in the CP layer, as shown in (31), where (31abc) have the same meaning: (31) a. Gianni sì che lo fa G. sì that it does

 Cecilia Poletto

b. Sì che Gianni lo fa SÌ that Gianni it does c. Gianni lo fa sì Gianni it does sì ‘Gianni surely does it.’ Nevertheless this is not the type of sì found in Old Italian, as sì here is the assertive pro-sentence, not the adverb ‘so’. Though they are related, Modern Italian sì has evolved into a contrastive Focus marker which sets the sentence as a correction of the context, it is a marker which signals the denial of (part of) a previous utterance. In work related to the left periphery Benincà and Poletto (2004) show that the left periphery of Romance languages contains at least two distinct types of Foci, contrastive Focus and what is dubbed Information Focus. Some Romance languages or dialects, notably Modern Sicilian (see Cruschina 2006 for a detailed discussion on Information Focus in Sicilian), and crucially Old Italian make use of both types of Foci, so a sentence can either have contrastive or information focus located in the CP layer. In Modern Italian this is not so; the only possible Focus position independently available is the contrastive Focus position, for reasons which are not clear yet. In Modern Italian, Information Focus is only available in a parasitic way to Contrastive Focus namely when the Contrastive Focus position is already activated and contains a lexical XP. I will elaborate on this in the next section where I discuss the loss of V2. For the moment, let us take this observation as a fact and analyze how sì has changed from Old Italian to Modern Italian. If the position of Information Focus is not independently available in Modern Italian, then the change that occurred in the syntax of the particle sì is expected. A priori, there are two possibilities when Information Focus is blocked by the loss of V2: the particle could have disappeared from the language, or it could turn into something else. As seen above, the sentential marker sì has not disappeared from the language, but given that the Information position is blocked in Modern Italian, it has turned into a marker of Contrastive Focus, which is still freely available in Modern Italian. I propose that Modern Italian sì is a contrastive Focus marker signaling that the whole sentence is in contrast to the previous context (as shown in (31b)). One argument in favor of the idea that sì has turned from an information Focus to a contrastive Focus is provided by the fact that it has developed a negative counterpart, which did not existed in Old Italian, namely no, found in contexts like the following: (32) no che Gianni non lo fa no that Gianni not it does `Gianni won’t surely do it.’ Contrastive Focus with respect to the context can be either assertive or negative, information Focus cannot be negative. This explains the absence of the negative counterpart of sì, no, in Old Italian.



Chapter 8.  Two instances of a broken cycle 

In addition to this, Modern Italian also displays cases such as (33): (33) Gianni sì che lo fa Gianni sì that it does `Gianni surely does it.’ Here the sentence starts out with a topic located in front of the Focus marker (as in (31a)). In this case, the sentence is set as a contrast to the discourse only with respect to the Topic Phrase. Notice that, while there can be more than one topic in Modern Italian (as in all Romance languages), there can only be one topic of this type. I call this Ground and it restricts the domain of the contrast, as the ungrammaticality of the following sentence shows: (34) *Maria, Gianni sì che lo fa Maria Gianni sì that it does ‘As for Maria, Gianni surely does it.’ Other types of topics are not in principle excluded, showing that not all Topic Phrases are blocked, but that there can be only one per type: (35) Per Maria, Gianni sì che lo fa For Maria, Gianni sì that it does ‘Gianni surely does it to please Mary.’ Sì can also unexpectedly occur at the end of the clause, as in (31c). There are two possible ways to analyze the sentence final position found in Modern Italian: we either assume that the sentence initial and the sentence final positions are different, and sì can appear close to the VP (in fact lower than all VP elements, as it is sentence final) or we can hypothesize that the sentence final and sentence initial position are one and the same. The sentence final effect would then be due to movement of the whole clause to a GroundP (a type of Topic) position in front of the Focus marker. As I show in detail in Poletto (2008), there is empirical evidence that the second hypothesis is correct.6 The structure of the two sentences with sentence initial and sentence final sì are represented below: (36) a. [ GroundP [CPFocus SI [FinP [Fin° che …[IP Gianni lo fa]]] b. [SpecGroundP [IPGianni lo fa] [Ground° [CPFocus SI] [FinP [IP Gianni lo fa]]] [Fin° [IP Gianni lo fa]]]] In the first case, sì is sentence initial followed by the whole clause, with a low complementizer in Fin head.7 In the second, the whole IP has moved to the Spec of a GroundP 6. I briefly sum up here what I assume for Modern Italian sì. The reader is referred to Poletto (2008) for a detailed discussion of the empirical arguments and of the technical details of the analysis. 7. These structures are also an argument in favour of the idea that the complementizer in these constructions is in Fin. See also Belletti (2008).

 Cecilia Poletto

moving through the Spec of the FinP and thereby preventing the realization of the complementizer because of the doubly filled comp filter (see Poletto 2000 for a discussion on the notion of the doubly filled comp filter in a cartographic approach). I report here some of the arguments in favor of this hypothesis developed in Poletto (2008): the first is that NO occurs in sentence final position and only right dislocated items can occur after it. Those elements which cannot be right dislocated, as negative quantifiers or verbal particles are ungrammatical: (37) Ci sono andata si, al cinema there am gone yes, to.the cinema `I really went to the cinema.’ (38) *Ci sono andata si, da nessuna parte there am gone yes, to no place ‘I really did not go anywhere.’ (39) *Non mi ha detto no su Not me has told no off `He did not tell me off.’ According to the hypothesis illustrated above, this is due to the fact that the whole IP has to be moved, hence all IP-internal material has to occur before sì, and only elements which can be right dislocated (like definite PPs, but unlike Quantified PPs or verbal particles) can be found to the right of the negative marker. If sì were in IP, we would expect it to be followed by some non dislocated XPs. The second argument is that sì is incompatible with elements whose position is typically associated to the lower portion of the CP layer, like wh-items: (40) *Dove sei andato sì? Where are gone yes? ‘Where did you really go?’ (41) *Il ragazzo a cui ho telefonato sì, è Gianni The boy to whom have phoned yes, is John ‘The boy I really phoned is John.’ Sentence final sì is both incompatible with interrogative wh-items and with relative pronouns. This is expected if the two types of elements are banned by a minimality effect. On this basis I will assume that sì is always a contrastive Focus marker in Italian, so its position is still in the Focus layer, but has readjusted to the new grammar, which does not allow information focus. As for the more general reason why an element like sì has been selected for marking Focus (either Information or Contrastive) I propose that, as in the case of the particle e, it must be related to its intrinsic nature. In this case I think that it is its status as an operator which makes sì the ideal candidate to mark



Chapter 8.  Two instances of a broken cycle 

Focus. That sì is an operator is shown by the cases in Old Italian in which sì is in SpecCP followed by a complementizer: (42) a. sì che quasi dal principio del suo anno non apparve a me so that almost from-the beginning of his year not appeared to me `so that almost at the beginning of the year it appeared to me...’ (Vita nuova 6) b. sì che li chiovi pareano 16 lettere so that the nails looked-like 16 letters `so that the nails looked like 16 letters.’

(Nov. 337)

Here the meaning of sì is not “in this way” but “in such a way that”. In other words in this usage the adverb already connects the previous utterance to the following one. We can conclude that the particle sì in Old Italian and Modern Italian can be used as a Focus marker due to its operator nature. In Old Italian, it marks Information Focus, and it has been reanalyzed as a marker of contrastive Focus in Modern Italian. In the next section, I will try to provide an explanation for this change based on the loss of the general V2 property.

4. A broken cycle Both sentential particles examined here have undergone a change from Old Italian to Modern Italian: the case of the particle e has been reduced to contexts which have the common property of having an active CP (due to the presence of a Wh-item or an operator), sì has been reinterpreted as a contrastive focus marker instead of an information focus marker. It is well known that the syntax of Italian underwent a major change after the medieval period. It lost V2 and this had consequences for the licensing of null subjects and the distribution of pronouns (see Benincà 1984). It also was important for the availability of a low left periphery located on top of the low phase vP (see Poletto 2006) thereby banning DP objects located in between the auxiliary and the past particle. I would like to propose that the change in the particle distribution also depends on the loss of V2 though in two different ways. I adopt here the standard assumption that “loss of V2” means in technical terms that the inflected verb does not obligatorily raise to the CP domain in main declarative clauses. More precisely, the relevant projection inside the split CP domain to which the inflected verb moves in Old Italian is Information Focus, the lowest projection inside the Focus field. Therefore, after the medieval period the inflected verb can remain in IP in declarative main clauses, as the feature in Information Focus forcing verb movement is not there anymore. Notice however, that Modern Italian does not represent the immediate stage expected after this change. This is represented by languages like Modern Sicilian, where, though the verb is not forced

 Cecilia Poletto

to move to the Information Focus head, this projection is still available to an XP, as the grammaticality of sentences like the following in a question-answer pair show (see Benincà and Poletto 2004 and Cruschina 2006 on this): (43) A: Chi è? Who is? `Who is it?’ B: Montalbano sono Montalbano am ‘It is M.’ Modern Italian has evolved further from a stage where the Specifier of Information Focus is still available: this projection can only be occupied when the specifier of Contrastive Focus already contains an XP. There are other Romance languages which have evolved even further, losing movement to Information and Contrastive Focus altogether, for instance, standard French. At this point I do not know why Modern Italian is different both from Modern Sicilian and Modern French. One can speculate that the unavailability of the Information Focus position must be in line with some property of the low left periphery on top of vP, but I will not take a stand with respect to the motivation of the loss of V2, which has been under discussion for years in the diachronic literature. What can be observed from the texts is exactly what Roberts (1993) observes in the case of Old French: on the one hand subject inversion becomes rarer and rarer while cases of preverbal subjects increase, on the other there is a restriction on the elements which can be moved to Information Focus. I analyzed the first ten chapters and chapter twenty to twenty-three of Machiavelli’s “Il Principe”, written in the early part of the 16th century, and only found very few clear cases of subject inversion (i.e. a nominal subject located in between the auxiliary and the past participle). In the whole sample (over 100 pages) there are only three cases of subject inversion with an auxiliary verb (cf. (44)) and eight cases with modals (six with potere ‘can’ and two with dovere ‘must’). Cases with modal verbs are illustrated in (45): (44) a.

Spenti adunque questi capi, e ridotti i partigiani loro blown.off then these bosses and reduced the friends their amici suoi, aveva il duca gittato assai buoni fondamenti friends his had the duke thrown very good foundations alla potenza sua to.the power his ‘when he had killed those chiefs and captured their friends, the duke had laid solid foundations to his power.’ (p.221)

b. Aveva adunque Luigi fatto questi cinque errori had then Luigi made these five mistakes ‘So Luigi had made these five mistakes.’ (p.197)



Chapter 8.  Two instances of a broken cycle 

c.

Mentre che durò la memoria, sempre furono i Romani while that lasted.3sg the memory always were the Romans incerti di quella possessione unsure of that possession ‘as long as this story is remembered, the Romans never completely possessed that land.’ (p.203)

(45) a.

E deve soprattutto uno principe vivere con i suoi sudditi And must.3sg overall a prince live with the his subjects in modo che … in way that… ‘and above all a prince should live with his subjects so that...’ (p.237)

b. E con più facilità se le può un principe guadagnare and with more ease for-himself them can.3sg a prince gain `and a prince can gain them more easily.‘ (p.205) The second fact is that the class of elements which can be placed in front of the inflected verb without being left dislocations or contrastive focus is dramatically reduced: in Machiavelli’s text the vast majority of cases are represented by the adverbs sempre ‘always’ and its negative counterpart, mai ‘never’. This clearly shows that Information Focus is not as readily available (both to the inflected verb and to XPs) as it was in the medieval period. The fact that Information Focus is blocked evidently produces an effect on the particle sì, which is then reanalyzed as contained in the closest projection, namely Contrastive Focus. As a consequence, the particle takes over also the properties of the new projection where it is located, not only because it indicates a contrast with the previous discourse, but also because it has the same properties as other types of CP elements in Modern Italian in allowing remnant IP movement and thereby triggering sentence final position of the particle. The case of the particle e is more difficult to explain, as the particle has not changed its status, but only reduced the number of contexts where it can be used as such. In order to function as a topic marking the continuation of the context, the particle e exploits a structure which is the same as the one used when it represents a conjunction. Given that the context, which is located in SpecConjunctionP, is represented as a whole CP starting with a TopP, it is necessary that the following clause is also syntactically a whole CP starting with some Topic (hence higher than Focus). This is by no means a problem in Old Italian, where in all main clauses the inflected verb has to reach the Focus field and there are null topics available. However, in Modern Italian this is not the case. Therefore, using the conjunction as a topic marker becomes impossible in Modern Italian unless the sentence is already a whole CP starting with some projection higher than FocusP. Precisely in these cases (namely when also the second condition is met), it is indeed still possible to use the conjunction particle as a topic marker. This means that the particle itself has undergone no real reanalysis from Old Italian to Modern Italian, but that it is rather the structure of the language that has

 Cecilia Poletto

changed, preventing an extensive use of the conjunction marker as a topic marking prosecution of the context. More generally, we can conclude that these two cases show that what looks like the reanalysis of an isolated item is strictly related to the whole system. This might prove to be a more general pattern which applies to reanalysis throughout: no lexical item can be reanalyzed by itself and reanalysis must always be the effect of a more general restructuring of the whole system. If this is correct, any time we deal with some element changing (even just the development of a new modal auxiliary, as bisogna from Old Italian to Modern Italian, or of the class of adverbs changing from manner to speaker oriented adverbs), which seem unrelated to the general structure of the language, we should seek the general structural change which drives the smaller ones of single items.

5. Conclusion In this chapter, I have examined the distribution of two sentential markers from Old Italian to Modern Italian and have reached a number of conclusions: first of all, following Kaiser (2006ab), I concluded that neither of the two elements is an expletive in the left periphery. The second conclusion is that the change of the two items is driven by a very major change in the V2 system of the language and that the two elements have only “readjusted” to the general context they are in. Moreover, the usage as topic or focus markers of these items seems to be related to their invariant categorial properties. This ensures that, when the right context is provided, these elements can still behave as they did in the old system (e still is a topic marker when CP is active). The general conclusion we can draw from this case study is that we should not only consider the change of lexical elements per se, as specifiers becoming heads, or lower elements climbling to higher functional projections, but also the general picture of the language which requires the change to be activated. In some cases, we might find that the element per se has not changed at all, and that its categorical and/or semantic invariant features just adapt to the context it is in.

References Adams, Marianne. 1987. From Old French to the theory of pro-drop. Natural Language and Linguistic Theory 5:1–32. Belletti, Adriana. 2008. Answering strategies: New information subjects and the nature of clefts. Ms, University of Siena. To appear as Chapter 10 of Structures and Strategies. London: Routledge. Benincà, Paola. 1984. Un’ipotesi sulla sintassi delle lingue romanze medievali. Quaderni Patavini di Linguistica 4: 3–19.



Chapter 8.  Two instances of a broken cycle  Benincà, Paola. 1988. L’ordine degli elementi della frase e le costruzioni marcate. In Grande grammatica Italiana di consultazione, Lorenzo Renzi (ed.), 129–194. Bologna: Il Mulino. Benincà, Paola. 1994. La variazione sintattica. Bologna: Il Mulino. Benincà, Paola. 1995. Complement clitics in medieval Romance: The Tobler-Mussafia law. In Clause Structure and Language Change, Adrian Battye & Ian Roberts (eds), 296–325. Oxford: OUP. Benincà, Paola. 2006. A detailed map of the left periphery of medieval Romance. In Negation, Tense and Clausal Architecture: Cross-linguistics Investigations, Raffaella Zanuttini, Héctor Campos, Elena Herburger & Paul Portner (eds), 53–86. Washington DC: Georgetown University Press. Benincà, Paola & Poletto, Cecilia. 2004. Topic, focus and V2. Defining the CP sublayers. In The Structure of the CP, Luigi Rizzi (ed.), 52–75. Oxford: OUP. Boya, Li. 2006. Chinese Final Particles and the Syntax of the Periphery. Leiden: The Leiden University Centre for Linguistics (LUCL). Cruschina, Silvio. 2006. Informational focus in Sicilian and the left-periphery. In Phases of Interpretation, Mara Frascarelli (ed.), 363–385. Berlin: Mouton de Gruyter. Garzonio, Jacopo. 2005. Le frasi interrogative non-standard in fiorentino. Rivista Italiana di Dialettologia 28: 219–235. Gelderen, Elly van. 2004. Grammaticalization as Economy [Linguistik Aktuell-Linguistics Today 71] Amsterdam: John Benjamins. Kaiser, Georg. 2006a. Sobre a (alegada) perda do sujeito nulo no português brasileiro. In Para a história do português brasileiro, Vol. 6: Novos dados, novas análises, Tomo 1, Tânia Lobo, Ilza Ribeiro, Zenaide Carneiro & Norma Almeida (eds), 11–42. Salvador: Editora da Universidade Federal da Bahia. Kaiser, Georg. 2006b. Pronombres sujeto en construcciones impersonales de lenguas iberorrománicas. In Andolin gogoan. Essays in Honour of Professor Eguzkitza, Beatriz Fernández & Itziar Laka (eds), 513–530. Vitoria-Gasteiz: Universidad del País Vasco/Euskal Herriko Unibertsitatea. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Ledgeway, Adam. 2007. Old Neapolitan word order: Some initial observations. In Histories and Dictionaries of the Languages of Italy, Anna Laura Lepschy & Arturo Tosi (eds), 121–49. Ravenna: Longo. Obenauer, Hans-Georg. 2004. Non standard wh-questions and alternative checkers in Pagotto. In The Syntax and Semantics of the Left Periphery, Horst Lohnstein & Susanne Trissler (eds), 343–383. Berlin: Mouton de Gruyter. Poletto, Cecilia. 2000. The Higher Functional Field in the Northern Italian Dialects. Oxford: OUP. Poletto, Cecilia. 2005. Sì and e as CP expletives in Old Italian. In Grammaticalization and Parametric Variation, Maria Luisa Hernanz,Carme Picallo & Francesc Roca (eds), 206–235. Oxford: OUP. Poletto, Cecilia. 2006. Parallel phases: A study on the high and low left periphery of Old Italian. In Phases of Interpretation, Mara Frascarelli (ed.), 261–294. Berlin: Mouton de Gruyter. Poletto, Cecilia. 2008 The syntax of focus negation. Ms, University of Venice. To appear in Working Papers in Linguistics University of Siena. Portner, Paul & Zanuttini, Raffaella. 2003. Exclamative clauses: At the syntax-semantics interface. Language 79(1): 39–81 Roberts, Ian. 1993. Verbs and Diachronic Syntax: A Comparative History of English and French. Dordrecht: Kluwer.

 Cecilia Poletto Vanelli, Laura, Renzi, Lorenzo & Benincà, Paola. 1985. Tipologia dei pronomi soggetto nelle lingue romanze medievali. Quaderni Patavini di Linguistica 5: 49–66. Reprinted in Benincà, 1994.

Primary sources Alighieri Dante. Vita nuova, Michele Barbi (ed.), Firenze, Bemporad, 1932. Anonymous. Fiori e vita di filosafi e d’altri savi e d’imperadori, Alfonso D’Agostino, (ed.) Firenze, La Nuova Italia, 1979. Anonymous. Novellino EdizioneLibro di novelle et di bel parlar gentile: Nel qual si contengono cento nouelle altrauolta mandate fuori da messer Carlo Gualteruzzi da Fano, Guido Favati (ed.), Genova, Bozzi, 1970.

part 3

Copulas, auxiliaries, and adpositions

chapter 9

The copula cycle Terje Lohndal

University of Maryland It is well-known that copulas often emerged historically from demonstratives and pronouns. The present paper argues that copulas change cyclically. This claim is substantiated through a rich number of examples from various languages. Besides the change from demonstratives and pronouns to copulas, it is shown that copulas further may develop into affixes. Interestingly, there are also languages like Hebrew where the copula first disappears and then redevelops at a later stage. This paper describes the stages of the copula cycle, and argues that there also are changes that do not fit directly into the major stages. The cycle is understood through a formal theory of grammaticalization, where there are cognitive principles that help the child to acquire a language.

1. Introduction*1 It is a well-known fact that “demonstratives may develop […] into personal pronouns, which themselves may give rise to copulas.” (Heine and Kuteva 2002: 109). In the typological literature, this process is known as copularization: “the grammaticalization process which turns full verbs or other non-copular elements into copulas” (Hengeveld 1992:  237–256, Stassen 1997:  94–99; though see Pustet 2003 for a slightly different terminology). The aim of this paper is to argue that there is a systematic cycle where copulas emerge from demonstratives and pronouns or from verbs and might then develop further into auxiliaries and grammatical markers like affixes. Demonstratives and pronouns may even reappear from copulas. In addition to providing a typology of these patterns of change, I also aim at giving a theoretical analysis of these data within the Minimalist Program. Specifically I will approach the data from the perspective of a formalist understanding of grammaticalization as in Roberts and Roussou (2003)

* Parts of this paper have been presented at the Linguistic Cycles workshop at Arizona State University in April 2008. I am grateful to the audience for valuable comments. Thanks also to Werner Abraham, Brian Dillon, Jan Terje Faarlund, Elly van Gelderen, David Ingram, and David Lightfoot for their useful remarks.

 Terje Lohndal

and van Gelderen (2004ab). This framework, I claim, provides a useful way both to classify the changes and to understand the changes in question. The paper is organized as follows. Section 2 discusses an approach to the syntax of copulas. Section 3 discusses grammaticalization and economy, adopting the framework in van Gelderen (2004ab). The copula cycle is introduced in Section 4, and step by step data are introduced that motivate the copula cycle. Section 5 concludes the paper.

2. The syntax of copulas How to analyze copulas is a problematic and somewhat controversial question. I will not provide an overview here, but simply present and briefly motivate the framework I am adopting. A common assumption about copulas is articulated by Baker (2003: 40): This range of data implies that the copula in English is not involved primarily in the dynamics of theta-role assignment, but rather appears when the lexical head of the clause cannot bear finite tense and agreement morphology.

This has been the traditional approach. As has been argued elsewhere, there are many reasons to think that this view is wrong (Rothstein 1999, 2001, Lohndal 2006, Lohndal, Åfarli and Nygård 2008). However, the issue is not directly important for the present paper, and what I will say in the remainder is largely tangential to this issue. What is important, however, is the phrase structure we assume for copulas. In the next section, I will outline an analysis where copulas are heads of a predication phrase. In a seminal paper, Bowers (1993) proposes a predication phrase, PrP. Pr is a functional category where the external argument (the subject) sits in the specifier of Pr. The predicate is the complement, and can consist of either a VP, NP, PP or an AP. This can be illustrated by way of the following tree structure. (1) PrP wo subject Pr’ wo Pr predicate/xp This structure accommodates small clauses as well as main clauses. The structure for a small clause like (2) is provided in (3).

(2) Lisa considers Mary crazy.



(3) [IP Lisa considers [PrP Mary [Pr’ Pr [AP crazy]]]]



Chapter 9.  The copula cycle 

Above the PrP in the clause is the inflectional phrase, IP. To illustrate a more complete structure, (4) represents a typical sentence (I have omitted the higher layers).1 (4) PrP wo subject Pr’ wo Pr VP wo V’ wo V Eide and Åfarli (1999), based on Eide (1996), have argued, pace Bowers (1993), Baker (2003) and Mikkelsen (2005), that copulas can lexicalize the head of PrP (see also Adger and Ramchand 2003: 336). This has further been corroborated by Lohndal, Nygård and Åfarli (2008), although their overall conclusion is slightly different. It is not possible to repeat all the arguments in favor of this conclusion, but let us just look at one important one from Eide and Åfarli (1999). Eide and Åfarli (1999) note that there is an interesting symmetry between the particle som in Norwegian and the copula. First, they argue that the particle som can lexicalize the predication operator. A few examples supporting this are provided in (5)-(7). (5) Vi fant [Marit [(*som) naken/ *(som) nervevrak]]. we found Marit as naked / as nervous wreck ‘We found Marit nude/as a nervous wreck.’ (6) Vi så [Jon [(*som) rasende/ *(som) spøkelse]]. we saw John as furious / as ghost ‘We saw John being furious/as a ghost.’ (7) Vi returnerte [pakken [(*som) uåpnet / *(som) flypost]]. we returned parcel.def as unopened / as air mail ‘We returned the parcel unopened/as air mail.’ (Eide and Åfarli 1999: 160) An adjectival predicate does not permit the occurrence of the particle som, whereas a nominal small clause predicate forces the presence of som. The structures in (5)-(7) all have in common that you can always paraphrase them as a matrix sentence, in which case you evidently will have to replace som with the copula. This is illustrated in (8)-(9). 1. As Bowers (1993: 599–600) argues, a main verb moves from V to Pr. Later on many people have used vP as a notation for more or less the same thing as PrP (Chomsky 1995, Kratzer 1996), and some people use PredP instead of PrP (Adger and Ramchand 2003, Baker 2003). Some have generalized this approach even further such that a variety of heads can lexicalize what has been called a relator phrase (den Dikken 2006). There are differences between these two approaches, but I will set them aside in this paper as they are not directly relevant for what follows.

 Terje Lohndal

(8) Vi fant [Marit [som nervevrak]] → Vi fant Marit og [Marit we found Marit as a nervous wreck → we found Marit and Marit [var nervevrak]]. was a nervous wreck (9) Vi ser [dette [som faktum]] → [Dette [er et faktum]] we see this as a fact → this is a fact (Eide and Åfarli 1999: 165) As Eide and Åfarli (1999) note, the bracketed parts of the two structures in each example are very similar both semantically and syntactically, and the most striking difference seems to be the head of the PrP, viz. the copula or som. The difference in lexicalization is related to the selecting matrix element. Eide and Åfarli hypothesize that the selecting matrix element is a verb not selecting for a verbal category when som occurs and a tense element when the copula occurs. This predicts that som and the copula are in complementary distribution, except when the head optionally selects a verbal category or a non-verbal category. The latter is precisely the case with Norwegian perception verbs. Eide and Åfarli point to (10) and (11) as evidence that in such a case, there is no complementary distribution, which is a strong argument in favor of treating both som and the copula as lexicalizations of the predication operator. (10) Jeg så [naboen [som spøkelse]]. I saw neighbor.def som a ghost ‘I saw the neighbor as a ghost.’ (11) Jeg så [naboen [være spøkelse]]. I saw neighbor.def be a ghost ‘I saw the neighbor be a ghost.’

(Eide and Åfarli 1999: 167)

In sum, it seems to be valid to assume the copula to be the head of PrP. This will be important for what follows, thus I have taken some space to argue in favor of this position. In the next section, I will turn to the specific diachronic framework that I will be using throughout the paper.

3. Economy and grammaticalization In recent years, grammaticalization has become an important research topic within generative grammar, cf. Longobardi (2001), Roberts and Roussou (2003) and van Gelderen (2004ab, 2007, 2008ab). Within the functionalist paradigm, grammaticalization has been a crucial topic for decades, see e.g. Lehmann (1985), Heine, Claudi and Hünnemeyer (1993), Hopper and Traugott (2003) and Heine and Kuteva (2002, 2007). The present paper will mainly be using the specific proposal laid out by Elly van Gelderen. Although several aspects between her theory and Roberts and Roussou’s are similar, there are also important differences, which I cannot deal with here for reasons of space. However, van Gelderen has in recent work integrated grammaticalization and the



Chapter 9.  The copula cycle 

linguistic cycle, and since the topic of this paper bears on both issues, I think her framework is especially well-suited for analyzing the diachronic source and development of copulas. In the present subsection, I will present the main aspects of van Gelderen’s theory. First I will say something about economy and its role in our internal grammar, and then I will discuss van Gelderen’s specific implementation of grammaticalization as feature economy.

3.1

Economy, grammaticalization and features

Economy has played a pervasive role within the history of generative grammar. Here I am primarily interested in the economy relating to grammatical derivations, not in the kind of economy related to Ockham’s razor. With the introduction of the minimalist program in linguistic theory in the late 1980s (see Chomsky 1995 for a collection of the most central papers), economy became one of the core principles of the program. One crucial aspect of economy is that derivations have to be as cheap as possible, that is, the grammar always chooses the least costly derivation. Another aspect is that children’s acquisition of language is arguably driven by principles of economy (van Gelderen 2004ab, 2007, 2008ab, Faarlund 2008). Van Gelderen (2004ab) proposes two specific principles of grammaticalization, namely (12) and (13). (12) Head Preference Principle (hpp) Be a head, rather than a phrase (13) Late Merge Principle (lmp) Merge as late as possible These principles are argued to guide the child during acquisition, which also means that they rely on the input. That is, the principles can only work when the input is ambiguous such that there is more than one way to analyze the input data.2 This relies on the assumption that it is more economical to be a head than a phrase as heads are less complex, and the fact that late Merge means that movement is avoided. Movement, even if seen as another species of Merge (internal Merge, cf. Chomsky 2004), is more costly because it involves a process of copying and remerging a lexical item during the derivation. External Merge does not involve copying, which makes it cheaper (van Gelderen 2008b). Concretely, HPP means that a speaker will prefer to build structures such as (14) rather than (15). I have followed van Gelderen in labeling the projection FP and pro stands for a pronoun.

2. Obviously, this claim is in need of further substantiation which goes beyond the scope of this paper.

 Terje Lohndal

(14)

FP 3 pro …

(15)

FP 3 pro F’ 3 F ….

Importantly, the HPP is a very specific and non-variant principle, as opposed to e.g. Hawkins’ (2004) efficiency principle ‘Minimize Forms’ or ‘Stay’ within Optimality Theory (van Gelderen 2008b). The LMP implies that Internal Merge (‘Move’) is less economical than External Merge (pace Chomsky 2004), but I will refrain from discussing this further at this point. As van Gelderen convincingly has shown, there are real empirical LMP effects, which would be problematic to account for if Merge and Move were equally economical. The second principle of economy, the LMP, works most clearly in the case of heads. Some common examples of LMP at work are listed in (16) (van Gelderen 2008b). (16) Preposition to complementizer (e.g. for) (17) modals: v > Asp > T (18) preposition to aspect (e.g. on) (19) vP adverbials to TP/CP adverbials (20) Negative objects to negative markers One very common change that could be explained by reference to this principle is the well known change in lexical verbs becoming modals in English (e.g. Traugott 1972, Lightfoot 1974, 1979, 1999, 2006, Plank 1984, Roberts 1985, 1993, Kroch 1989, Warner 1993, Abraham 2002). In Middle English verbs like can, could, may, might, will, would, shall were full verbs. Lightfoot (2006: 90) summarizes the main differences by way of the following data in (21)-(30). (21) *He has could understand Chapter 4. (22)

He has understood Chapter 4.

(23) *Canning understand Chapter 4, … (24)

Understanding Chapter 4, …

(25) *He wanted to can understand. (26)

He wanted to understand.

(27) *He will can understand. (28)

He will try to understand.



Chapter 9.  The copula cycle 

(29) *He can music. (30)

He understands music.

(21)-(30) show that modal auxiliaries differ from ordinary verbs in their distribution. A modal does not occur with a perfective (21) or past participle (23), it does not occur in the infinitival complement of another verb (25), nor as the complement of another modal (27), and no modal may occur with a direct object (29). However, when modals were main verbs, these structures were generally possible. Under the LMP, this change is straightforward.3 (31) would be preferred compared to (32) because the modal auxiliary is based generated higher in the former than in the latter. (31)

TP 3 T vP might 3 v’ 3 v …

(32)

TP 3 T vP 3 v’ 3 v … might

Generally the LMP accounts for changes where a lexical head becomes a functional head, or where a functional head becomes a higher functional head, both frequently described in the literature on grammaticalization (see e.g. Heine & Kuteva 2002). However, the LMP can also account for changes in the position of lexical phrases whereby lexical phrases become base generated in the functional domain. Consider the examples in (33)-(34) (van Gelderen 2009). (33) Those who offend actually, are most grievously punished. (34) Actually, it is kind of an interesting problem. When actually was first introduced into the English language from French, it was an adjective. In the 15th century it is used as a vP adverb (33). It may have been topicalized 3. Notice, though, that the LMP does not say anything about when this change occurred. Obviously, the LMP is a principle that interacts with the external data, and unless the external data is such that the principle can kick in, it won’t.

 Terje Lohndal

frequently, leading to a change to a CP adverb (34). This change also shows that the LMP works for both phrases and heads. Recently, van Gelderen (2007, 2008b, 2009) has reanalyzed the HPP and LMP, collapsing them into one overall and more general principle, namely (35). (35)

Principle of Feature Economy Minimize the semantic/interpretable features in the derivation Adjunct Specifier Head > affix Semantic > [iF] > [uF] > --

This is among others based on Radford’s (2000) argument that interpretable features are acquired before uninterpretable features, and the latter ones trigger the grammatical system. Assuming that only heads are Probes (Chomsky 2008), (35) means that whenever the HPP is at work, we get a (new) Probe. Interestingly, (35) also provides us with a formal definition of what grammaticalization really is. It is a change whereby something becomes more economical, i.e. where semantic features are reduced (on the assumption that uninterpretable features are more economical than interpretable ones). This change in turn has syntactic consequences, and the remained of the paper will be dealing with syntactic consequences relating to copulas diachronically.

4. The copula cycle and features In this section I will present various stages of what I take to be a copula cycle. First I will present the cycle based on general theoretical considerations within the framework laid out in Section 3. This will function as a guiding heuristic when working our way through the examples in the following sections. In Section 4.2 I will describe changes where a specifier develops into a head, viz. mainly demonstratives or pronouns that become copulas. Section 4.3 deals with the stage where a full verb has developed to a copula, that is, a head-to-head change. This will be shown to be a prime example that the major cycle also has smaller sub-stages. Section 4.4 discusses changes where the copula develops into a grammatical marker, viz. a suffix. In Section 4.5 I sum up the data and argue that the cycle proposed in 4.1 accounts straightforwardly for the change.

4.1

The copula cycle

In the previous section I adopted a principle of feature economy that favors uninterpretable features over interpretable features. This predicts that functional elements will start from lexical elements, a prediction borne out by a wealth of attested examples in the literature. This is also true for copulas, and Pustet (2003: 54) says that “the main source for copulas seems to be verbs and pronouns”. Stassen (1997: 92–99) argues that



Chapter 9.  The copula cycle 

there are three major diachronic sources for (verbal) copulas. The first source is dynamic verbs “covering the whole or parts of the semantic domain which includes notions such as ‘do/make/build, ‘happen/occur’ ‘go/turn into/come/become’ and ‘act(like)’” (Stassen 1997: 92–93). Examples of languages mentioned by Stassen that employ this strategy are the Berber languages Tamazight and Shilha, various TibetoBurman languages, Lahu, Lisu and Cambodian. The second source is what Stassen calls copularization. This is a process where “one of the members of the set of locational support verbs becomes specialized as the support item for nominal predicates” (Stassen 1997: 94). Commonly it is the semantically most neutral verb that undergoes this change. One further step in this process is the gradual loss of locative meaning of the copular verb, thus the language shifts towards a lexical differentiation of nominal and locational support items. The Tanoan language Kiowa appears to be such a language, Stassen argues. The only examples in which the item do: can act as a locative support where are sentences where it is accompanied by the particle he ‘away, gone’ (37). Other locational contexts require the use of a set of different posture verbs (38)(39).4 (36) Te: Koy-gu ba-do: all Kiowa-und 2pl.und-cop ‘You are all Kiowas.’

Kiowa

(37) P’o: he: gya-do: moon away pl-cop ‘There was (temporarily) no moon.’ (38) Guy-te Ø-t’o: other 3sg-stay ‘There is someone else here.’ (39) E:go yi: ol e-cel kicoy-ka here two hair 3dual-cop in.soup-in ‘There are two hairs in the soup.’

(Stassen 1997: 94)

Other languages where copularization happens are the Yuman languages (Mojave, Cocopa, Yavapai) and in North Carolina Cherokee (Stassen 1997: 94). The third source is 4. The abbreviations used in the glosses are as follows: 1, 2, 3 = person; anim = animate; ant = anterior; aux = auxiliary; c = noun class; class = classifier; comp = comparative case; cop = copula; dcl = declarative; def = definite; dem = demonstrative; detr = detransitivizer; dist = distal; dual = dual; erg = ergative case; ess = essive case; ex = existential; fem = feminine; fin = finite; fm = focus-marker; focus = focus/topic particle; fut = future; gen = genitive case; hab = habitual; imp = imperative; inan = inanimate; instr = instrument; loc = locative case; masc = masculine; nm = nominalizer; nom = nominative; npl = nonplural marker; npst = non-past; pat = patient; perf = perfective; pl = plural; prog = progressive; prox = proximal; prs = present; pst = past; px = proximity marker; rel = relative marker; s = part of verb stem; sbj = subject; seq = sequential converb; sg = singular; super = superessive case; und = undergoer-marker.

 Terje Lohndal

the process of verbalization, which can be “viewed as the final stage of the grammaticalization process by which pronouns and discourse particles are reanalysed as abstract linking morphemes in predicate nominal sentences” (Stassen 1997:  95). This process is especially frequent in many African languages, e.g. Temne, Gola, Zande, Gbeya, Zulu and Shona, to mention but a few (see Stassen 1997: 96 for a more comprehensive overview). In the present paper, I will offer a different suggestion. I will propose a typology based on structural positions. On the basis of the framework presented in Section 3, I will try to explain the cyclic nature of the changes in terms of feature economy. Note, though, that my proposal is not incompatible with Stassen’s; it is another way of looking at the issues, and, hopefully also a way to contribute with some different and further insights. Basically, I will argue that there is a general copula cycle which looks like (40). Both the structural positions and the feature contents of the lexical items are present (cf. Chomsky 2000 et seq.). (40) demonstrative/pronoun specifier iF

> > >

copula head uF

> > >

grammatical marker affix --

Several interesting questions emerge from (40). What kind of specifier position do demonstratives occupy? How do prepositions that develop into copulas fit into (40)? Are there sub-changes internal to the cycle in (40)? Below I will address all of these questions. Section 4.2 discusses specifiers becoming heads, and I will argue that in most cases the relevant specifier is the specifier of a predication phrase. This specifier then turns into the head of the predication phrase (cf. the discussion in Section 2). In Section 4.3 I discuss the stage where the copula is a head, and argues that changes take place within this stage of the cycle. Section 4.4 discusses heads becoming affixes, that is, grammatical markers. Section 5 provides a short summary.

4.2

Specifier to head

This section will discuss a number of examples showing that specifiers may develop into copulas. The main part of the section will be devoted to demonstratives and pronouns becoming copulas, but I will also somewhat briefly discuss existentials and prepositions undergoing the same change. 4.2.1. Demonstratives and pronouns There are several cases where a demonstrative or a pronoun has developed into a copula.5 The most famous case is the one described in Li and Thompson (1977), and their paper will also play an important role in this section. However, towards the end I will 5. In some cases, it may be hard to determine whether a lexical item is a copula or a nominal element (cf. Déprez 2003). However, I will not discuss any such cases in this paper.



Chapter 9.  The copula cycle 

also show that there are several other attested cases of this change. I also think that the theoretical framework adopted in this paper is a good alternative compared to functional theories of grammaticalization when it comes to analyzing this change. As Devitt (1994: 143–144) points out, the change from demonstratives and pronouns to copulas “is somewhat problematic for grammaticization theory if semantic generalization is assumed to be the key process in the development of grammatical morphology”. As will hopefully become evident, I think an advantage of the formal theory of grammaticalization adopted here is that it is able to deal straightforwardly with this change. The most widely know case of a pronoun becoming a copula is the Mandarin shì, discussed in Li and Thompson’s (1977) seminal paper on the development of copula morphemes.6 In Archaic Chinese (11th -3rd B.C.), equational sentences did generally not have a copula. This is shown in (41). (41) Wáng-Tái wù zhě yě Wang-Tai outstanding person dcl ‘Wang-Tai is an outstanding person.’

Archaic Chinese (Li and Thompson 1977: 421)

In Modern Mandarin, the copula shì regularly occurs in equational sentences. (42) nèi-ge rén shì xuéshēng that-class man cop student ‘That man is a student.’

Modern Mandarin (Li and Thompson 1977: 422)

This modern copula was a demonstrative in Archaic Chinese, as shown in the following example. (43) fū-zĭ zhì yù shì bāng yě Confucius arrive at this nation dcl ‘Confucius arrived at this nation.’

Archaic Chinese (Li and Thompson 1977: 423)

However, shì also occurred in equational sentences in Archaic Chinese as an anaphoric demonstrative pronoun. (44) jì yù qí shēng yoù yù qí sĭ shí huò yě Archaic Chinese already wish him live also wish him die this indecision dcl ‘wishing him to live while wishing him to die, that is indecision.’ (Li and Thompson 1977: 424)

6. Diessel (1999: 143–147) argues against Li and Thompson’s (1977) analysis. He claims that the developments of nonverbal copulas from third person pronouns and from demonstratives follow two different pathways. However, I will argue that in both cases we are dealing with a case where a specifier becomes a head, which justifies subsuming the possible different pathways in the present paper. Possibly there might though be two separate sub-stages here.

 Terje Lohndal

Li and Thompson say that it seems fairly clear that the use of shì as a copula was productive by the late Han period (1st -2nd century A.D.). In the vernacular literature, one finds sentences such as (45) showing shì as a copula. (45) cĭ bái wù shì hé děng? this white thing cop what kind ‘What kind of stuff is this white thing?’

(Li and Thompson 1977: 425)

The important thing, however, is that it is the use of shì as an anaphoric demonstrative pronoun that gave rise to the modern copula construction: “These topic-comment constructions, then, set the stage for reanalysis: the topic-comment construction without a copula became a subject predicate construction with the anaphoric demonstrative pronoun shì being reanalyzed as a copula” (Li and Thompson 1977: 425). Importantly, Li and Thompson argue that there are some ambiguous data that favor this analysis. The data in (46)-(47) could in fact well be interpreted as subject-predicate equational sentences. They are given a topic-comment interpretation because there is no sentence in Archaic Chinese where shì functions solely as a copula verb. (46) zhī ér shĭ zhĭ, shì bù rèn yě Archaic Chinese know then use him, this not kind decl ‘To use him knowing (that he would rebel), that was unkind.’ (47) jì yù qí shēng yoù yù qí sĭ, shì huò yě already wish him live also wish him die, this indecision dcl ‘Wishing him to live while wishing him to die, that is indecision.’ (Li and Thompson 1977: 424) Since (46)-(47) are open to multiple analyses, this made it very likely that children analyzed these strings differently at a later stage, given that there is sufficient representation of the structures in the input (Lightfoot 1999). Ambiguity in the input (the primary linguistic data) is a common source for language change (and often very little input is required; see Westergaard 2008), thus it is not incidental that the demonstrative pronoun develops into a copula when we assume the principles of economy given in Section 2. Let me now show how this can be analyzed formally. We have seen that a demonstrative can develop into a copula in Chinese. I will suggest that the demonstrative is in SpecPrP, and that through the HPP it is analyzed as head of the PrP. Using the sentences in (48) and (49) (repeated from above for expository convenience), there is a change from (50) to (51). (48) Wáng-Tái wù zhě yě Wang-Tai outstanding person dcl ‘Wang-Tai is an outstanding person.’

(=(41))

(49) nèi-ge rén shì xuéshēng that-class man cop student ‘That man is a student.’

(=(42))



Chapter 9.  The copula cycle 

(50) PrP wo Wáng-Tái Pr’ wo Pr wù zhě (51) PrP wo nèige rén Pr’ wo Pr xuéshēng shì In order to see this change even clearer, consider the part in bold in (52) (=(46)). (52) zhī ér shĭ zhĭ, shì bù rèn yě Archaic Chinese know then use him, this not kind decl ‘To use him knowing (that he would rebel), that was unkind.’ The stage where shì was a demonstrative is given in (53), and the new stage where it is analyzed as a copula is given in (54). The structures are simplified by not showing any of the higher functional projections. (53) (54)

PrP wo

shì

Pr’

wo

Pr

bù rèn

PrP

wo 6 Pr’

zhī ér shĭ zhĭ ei 3 Pr bù rèn shì These examples show how a formal theory can be used to explain the changes in question. However, I would like to point out an important issue at this point. Van Gelderen (2004b: 73 ff.) has invoked exactly this specifier-to-head change to explain how demonstratives may develop into complementizers. I think the fact that demonstratives become both complementizers and copulas show a crucial point: it is of utmost importance in which specifier an element is. If an element is in SpecCP, it may become a complementizer; if it is in SpecPrP, it may become a copula. Consequently, an element in SpecCP is predicted not to develop into a copula. The latter change has to the best of my knowledge not been attested. Notice that this is not a novel proposal about the framework I am

 Terje Lohndal

adopting, but I think the data that have been discussed underline this fact even further. Let us now look at more cases of this change across various languages. Li and Thompson claim that analogous processes like the Chinese one have taken place in Hebrew, Palestinian Arabic and Wappo. What is particularly interesting for our purposes is that they argue that there are traits of full cycles in both Hebrew and Palestinian Arabic. I will use the former language to illustrate the process. In Hebrew, the triliteral verbal copula, h-y-y is not used in the present tense. Instead, one finds that equational sentences contain a demonstrative pronoun or a personal pronoun (if the subject is non-sentential). (55) [še nitnaged lo] ze mәguxax that we.will.oppose.to him “this” absurd ‘It would be absurd for us to oppose him.’

Hebrew

(56) moše (ze) student šeli Moshe that student my ‘Moshe is a student of mine’ (57) ata (hu) ha-ganav you “he” def-thief ‘You are the thief.’

(Li and Thompson 1977: 427)

For most speakers, according to Berman and Grosu (1976), the hu morpheme is optional in sentences containing pronominal subjects but obligatory in sentences with full DP subjects. (58) David hu ha-ganav David “he” def-thief ‘David is the thief.’ (59) *David ha-ganav David the-thief

Hebrew

(Li and Thompson 1977: 428)

There are several other restrictions on where hu appears (Li and Thompson 1977: 428– 429). These suggest, Li and Thompson argue, that pronouns are being reanalyzed as copulas in Modern Hebrew. Recent research has also supported this view: “Colloquial Hebrew uses the overt copula with ever increasing frequency” (Katz 1996: 87). Two arguments are advanced in favor of this claim. First, the morphemes ze and hu are also able to function as pronouns. (60) ze mešune ‘That’s strange.’ (61) hu ohev et-Rivka he loves acc-Rivka ‘He loves Rivka.’

Hebrew

(Li and Thompson 1977: 429)



Chapter 9.  The copula cycle 

Sentences such as (62) have completed the transition from topic-comment to subjectpredicate equational sentences according to Li and Thompson (1977: 429–431). (62) David hu ha-ganav David “he” def-thief ‘David is the thief.’

Hebrew (Li and Thompson 1977: 429)

I will not present the arguments in favor of this view. Instead I will direct attention to another phenomenon, discussed by Katz (1996). The following example shows that the copula does not agree in Person with its subject. The subject is second person, but the form of the copula is nevertheless the invariant hu. (63) ‘Ata hu ha-‘iš thou.masc cop def-man ‘You are the man.’

Hebrew (Katz 1996: 90)

Katz argues that this mismatch in persons shows that hu is a copula and not merely a deictic marker: “The loss of the category of person is one property which leads us to conclude that the erstwhile pronouns have grammaticalized in such utterances, and are for all intents and purposes full fledged copulas” (Katz 1996: 90; cf. Devitt 1994: 140). An interesting important point is nevertheless that we have seen a development where a copula disappears (h-y-y) and a new copula hu develops from a pronoun. As Li and Thompson (1977: 438) say, “Thus, it seems that the emergence and decline of the copula may be seen as a diachronic cycle. The cycle is clearly very complex”. This is further supported by data in Katz (1996) who presents evidence that copulas may become pronouns. Katz (1996: 118–133) argues that the Turkish third person singular pronoun o(n) developed from the third person singular present of the copula verb. In pre-nineteenth century texts, the usual form of the third person absolutive was ol, and not o as it is today. The ol root is also found in another instance, namely in the verb olmak ‘to exist’. Katz says that ol was also used as the equivalent of the copula in the early period. As evidence for the directionality ol > o, she points towards the fact that in earlier texts, both were used interchangeably. In modern texts, o is the only form. Regarding the verb olmak ‘to exist’, it is clear that this verb shares the common root with the demonstrative pronoun ol since -mak is the infinitival ending. Katz (1996:  122–123) also presents two main arguments for why the pronoun developed from the copula and not the other way around. First, it is only in the very earliest texts that the pronoun has a copula function. This seems to indicate that the copula function emerged earlier, and that the pronoun is the new form. Second, early Turkish oral texts used independent pronouns sparsely whereas Modern Turkish uses them amply. Here too, the pronoun form seems to be the innovative one. Based on this, it seems clear that the verbal form ol developed into the pronoun o(n) (see Katz 1996: 120 on the nasal). Going back to demonstratives becoming copulas, another interesting case is Panare. Panare has three different copulas, as shown in (64)-(66).

 Terje Lohndal

(64) maestro këj e’ñapa teacher cop.anim.prox Panare ‘The Panare is (prox) a.teacher.’

Panare

(65) maestro nëj e’ñapa teacher cop.anim.dist a.teacher ‘The Panare is (dist) a teacher.’ (66) e’chipen mën manko fruit inan mango ‘Mango is a fruit.’

(Gildea 1993: 55)

These examples show that the different copula depends on whether the third person subject is animate or inanimate and whether they are inside or outside the speaker’s sphere of perception. First and second person subjects in Panare have no copula. (67) maestro yu/amen teacher 1sg/2sg ‘I am/You are a teacher.’

Panare (Gildea 1993: 54)

Gildea (1993: 56) says that: “It is clear on morphological grounds that the copulas këj, nëj and mën are derived from nonverbal elements in that verbs take tense/aspect suffixes and prefixes which agree with the subject for person, while these copulas take no affixes and they vary suppletively to agree with the subject for animacy”. The nonverbal forms which most resemble these copulas are the pronouns kën ‘anim.invisible’, nëj ‘who?’ and mën ‘inan.invisible’. Gildea argues that the nonverbal copulas and these pronouns are derived from the same historical set of pronouns, viz. demonstratives. There are two reasons for this claim. First, demonstrative pronouns are not used with first or second persons, which fits well with the fact that nonverbal copulas are not used either with these persons. Second, the deictic opposition is common for demonstrative pronouns, cf. English this and that. As Gildea (1993: 58) points out, there is one difference between Panare and the examples discussed by Li and Thompson (1977). In the latter case, all word orders are subject-predicate, whereas in Panare the order is predicate-subject. He proposes that in Panare we have left dislocation (instead of right dislocation as in Chinese): A teacher [is] he, John. Thus the demonstrative was required as the subject, and a new analysis happened. It seems likely that left dislocation has worked as a kind of trigger in Panare, just as right dislocation has in Chinese. That is, both types of dislocations provided a natural domain for the emergence of the copula.7 What happens further, and which shows that the demonstrative has separated from the subject semantically is that the deixis expressed by the copula may differ from that expressed by the subject. An example is shown in (68).

7. It may be that prosody plays a role in these changes, but that question goes beyond the scope of this study.



Chapter 9.  The copula cycle 

(68)

maestro nëj mëj teacher anim.dist anim.visible ‘This (here) guy was a teacher.’ [he (prox) is (dist) a teacher]

Panare (Gildea 1993: 61)

The subject in (68) is mëj ‘anim.visible’, so the spatial dexis is proximal (i.e. the person is right here); the copula is nëj ‘anim.dist’, so the temporal deixis is distal (i.e. not right now). Distal temporal deixis is most commonly interpreted as past. Based on this, Gildea argues that the historical demonstrative pronouns have become synchronic tense markers. In the final step, these copulas develop into auxiliaries. The following examples show how këj and nëj are used as verbal auxiliaries for aspect-marked verbs (ë is a non-agreeing intransitive prefix). (69) ë’púmanëpëj këj Toman a.detr.hit.imperf.i anim.prox Toman ‘Tom is falling (and landing with an impact).’ (70) ?ë’púmanëpëj nëj Toman a.detr.hit.imperf.i anim.dist Toman ‘Tom is falling (dist).’

(Gildea 1993: 65)

The only difference between these clauses is the choice of auxiliary. In (69), the auxiliary is këj, and as the translation shows, the clause is in the present tense. (70) has the auxiliary nëj, and although the clause is translated into present tense, it is understood “to be occurring at a spatially distal location” (Gildea 1993: 65). Gildea (1993: 67–68) then shows that these auxiliaries share syntactic properties with prototypical tense markers in Panare; see his work for discussion. As Pustet (2003: 56) discusses, in the language Kenya Luo, the third person pronoun can optionally be used as a copula, as shown in (71). (71) dhákó ėn bé ˆr woman cop goodness ‘woman she is goodness.’

(Tucker 1993: 308)

(72) dhákó beˆr woman goodness ‘woman she is goodness.’

(Tucker 1993: 308)

Kenya Luo

In Lango, a language closely related to Kenya Luo, Pustet argues that the grammaticalization has moved further. The original pronominal function is not transparent any more, and the copula εn occurs not only with third person singular subjects. (73) án εn àdáktâl 1sg cop 1sg.doctor.hab ‘I am the doctor.’

Lango (Noonan 1992: 146)

 Terje Lohndal

Again we see a process where pronouns develop into copulas. This is also the case in Lakota. In this language the copula hécha is etymologically based on the demonstrative pronoun hé ‘this’ and the element cha ‘such’. Neither hé nor cha are compatible with verbal inflectional categories: hécha, however, can be inflected like a stative intransitive verb” (Pustet 2003: 56). (74) wašícu he-má-cha white.man s-1sg.pat-cop ‘I am a white man.’

(Pustet 2003: 68)

(75) Phaláni he-má-cha Pawnee.Indian s-1sg.pat.cop ‘I am a Pawnee Indian.’

(Pustet 2003: 69)

Lakota

Lastly, let us have a look at one more case study. McWhorter (1997) (see also McWhorter 1995) discusses whether copulas in English-based Atlantic creoles emerged due to influence from West African copula systems or whether copulas are the result of independent developments. Some data relevant to this issue are provided in (76)-(77). (76)-(77) show that Ewe uses one copula in an equative context and another copula in a locative context. (76) Ló é-nye tọmelã crocodile he-cop aquatic.animal ‘The crocodile is an animal that lives in the water.’ (77) É-lè xọ me he-cop house in ‘He is in the house.’

Ewe

(McWhorter 1997: 243)

The Sranan creole has a similar pattern, as shown in (78)-(79). (78) Mi na datra I cop doctor ‘I am a doctor.’ (79) A de na ini Sranan Kondre he cop loc inside Surinam ’He is in Surinam.’

Sranan

(McWhorter 1997: 243)

McWhorther argues against the hypothesis that the copula systems in creoles are the result of influence from the West African languages (though see Migge 2002 for a critical discussion). On the basis of data from Arends (1989), he shows that a zero copula is possible in the oldest Sranan documents (the example is from 1770). (80) Mi blibi joe wan bon mattie fo dem. I believe you a good friend for them ‘I believe you’re a good friend of theirs.’

Sranan (McWhorter 1997: 244)



Chapter 9.  The copula cycle 

Arends argues that the copula evolved from the demonstrative da ‘that’. (81) ‘Adjabre’, da Djutongo Sranan ‘adjabre’ that ‘Jews’ language’ ‘”Adjabre” is Saramaccan.’

(McWhorter 1997: 244)

The modern copula emerged when what McWhorter calls a resumptive to a preceding topic is analyzed as a copula. That is, the demonstrative da is analyzed as a copula at a later stage. McWhorter also discusses the development of the locative copula de. He discusses several ways of analyzing this issue (1997: 247–250), which I will not go into here. Instead, I will just give his suggestion as to how this copula emerged. In contemporary Saramaccan, the deictic adverb de is often optionally inserted into sentences in order to lend deictic emphasis. Thus using de shows that the utterance is expressive. (82) Nóiti fa mi de a Winikíi dẹ, nóiti mi jéi táa… never since I cop loc Winikii there never I hear talk ‘Never since I’ve been there at Winikii have I heard that …’

Saramaccan

(83) Dí Gaamá dí Kófi gó lúku dẹ dẹ ku suwáki dẹ the chief rel Kofi go see there cop with sickness there ‘The chief who Kofi went to look at is sick.’ (McWhorter 1997: 250) McWhorter argues that it is from this usage that de in Modern Saramaccan is derived. At first there was no expression of the locative copula, just as we have seen for the equative copula. (84) Dí wómi a wósu the man loc house ‘The man is at home.’

Saramaccan (McWhorter 1997: 250)

McWhorter says that “however, it would have been a common expressive strategy to insert an expressive deictic de between subject and the predicate, similar to today’s usage” (1997: 250). (85) Dí wómi de a wósu the man there loc house ‘The man is there at home.’

Saramaccan (McWhorter 1997: 250)

This deictic marker was now ripe to be analyzed differently by new speakers as these speakers already had heard a number of Ewe sentences where the copula was obligatory. (86) É lè xọ me he cop house in ‘He is in the house.’ Thus, de is no longer an adverbial but a copula.

Ewe (McWhorter 1997: 251)

 Terje Lohndal

(87) Dí wómi dé a wósu the man cop loc house ‘The man is at home.’

Saramaccan (McWhorter 1997: 251)

Hence, the transition to a copula is complete. In sum, we have seen a number of examples where demonstratives and/or pronouns develop into copulas. I have claimed that this is an instance of a preference in acquisition for analyzing elements as heads instead of specifiers, following van Gelderen (2004ab). In the next section I show another case of this change, namely copulas emerging from existentials. 4.2.2. Existentials Another pathway where copulas emerge is from existentials. This is not a widely attested change, but data from Chalcatongo Mixtec show that it exists: “Mixtec also allows a construction in which the existential is used with an adjective. This construction […] apparently has the same meaning as the normal copula + adjective construction” (Macaulay 1996: 87). To see this clearly, observe that the normal copula in Mixtec is ka. (88) xwã ka lúlí Juan cop small ‘Jus is small/short.’

Chalcatongo Mixtec (Macaulay 1996: 130)

The existential marker in the language is žoo ‘there is’. (89) inì kaxá waA žoo tenànà insides box def ex tomato ‘in the box there are tomatoes.’

Chalcatongo Mixtec (Macaulay 1996: 129)

The existentiale can also be used as a copula, which arguably is the new function. (90) kGsG žóó xáá Chalcatongo Mixtec pot cop new ‘the pot is new/there is a new pot.’ (literally ‘as for the pot, it exists new.’) (Macaulay 1996: 130) These two cases show that existentials may develop into copulas. I will argue that this change is just another specifier to head change. That is, simplified structures of (89) and (90) would look like (91) and (92). (91)

PrP wo žoo Pr’ wo Pr tenànà



Chapter 9.  The copula cycle 

(92)

PrP wo kGsG Pr’ wo Pr xáá žóó

This presupposes a certain view on existentials. It seems reasonable to analyze these existentials on a par with expletives like English there, which means that they are phrases. I will assume, following the argumentation in Richards (2004), Richards and Biberauer (2005) and Biberauer and Richards (2006) that expletives are merged in what corresponds to SpecPrP in my framework (cf. also Åfarli 2005 within a somewhat different perspective). Given that, the analysis should be straightforward. 4.2.3. Summary In this section I have discussed a number of examples of what I have analyzed as a change from an element being a specifier to an element becoming a head. The main cases have been demonstratives and pronouns that develop into copulas, but we have also seen that existentials may undergo a similar change. In the next section, I will discuss cases where a head develops into a different head.

4.3

Head-to-head

There are a number of cases where full verbs develop into copulas. However, there are also cases where copulas develop into auxiliaries. Both of these changes do not involve change of projection type – both are heads – but I will argue that the nature of the heads change. In this section I will provide examples of these changes and an analysis. 4.3.1 Full verbs It is well-attested in the literature that full verbs may develop into functional elements. A famous case is the development of English full verbs into modals (see the references in Section 3). A similar change is where full verbs become copulas. This has even happened in English; the past tense forms of be come from wes meaning ‘to stay, to remain’ (Devitt 1994: 128). In this section I will give some further examples and an analysis. In Alyawarra, copula constructions either use the verb anima ‘to sit’ or ayntima ‘to lie’. These verbs have still retained their basic postural sense. This is shown in (93)-(94). (93) an-il-ayant-a nhilantiya sit-lig-aux-imp here.still ‘Sit here for a while; stay here.’

Alyawarra

 Terje Lohndal

(94) ayinga alingkarr-a an-ina I.nom tired-nom sit-past.cont ‘I was tired; I sat feeling tired.’

(Devitt 1994: 129)

Similar cases are found in other languages. The simplex verb ci- in Chantyal has extended it meaning beyond the expression of a basic posture sense. The original use is shown in (95), and the new copula meaning in (96)-(97). (95) Ram kurci-phyaraŋ ci-I Ram chair-super sit-perf ‘Ram sat on the chair.’

Chantyal (Noonan and Grunow-Hårsta 2002: 82)

(96) cater nә kBi nә mastәr ci-wa pәri-m bBi-i then focus you focus teacher sit-nm happen-npst say-perf ‘”Your have to be the teacher,” they said from that time on.’ (97) mә˜ya kBi-ye Bәnuwar lu˜-i ci-si-m love you-gen appearance shine-ant sit-ant-npst ‘(My) love, your appearance seems to shine.’ (Noonan and Grunow-Hårsta 2002: 87) There is also a copula, mu (with the allomorphs –m and –mu), which derives historically from a verb meaning ‘sit/stay’. This copula has developed into an auxiliary verb and a non-past suffix on verbs (cf. Turunen 2006 for a similar change in Erzya). (98)(99) show the auxiliary and (100)-(101) show the affix. (98) na kam-ri tBo-wa-bBәnda pәyle, Ram kam la-wa mu Chantyal I work-loc arrive-nm-comp before Ram work do-nm cop.npst ‘Even before I get to work, Ram will be working.’ (99) cә din nә bәnnu nә bBәrә-ysi-wa mu-wa Bin-si-m that day focus gun focus fill-ant-nom cop-nm cop-ant-npst ‘That day, it turned out that the gun was loaded.’ (Noonan and Grunow-Hårsta 2002: 89) (100) na-sә sar-mu I-erg kill-npst ‘I’ll kill it.’ (101) әmrika-ne-ma-thõy la-si-rә ci-m America-person-pl-ess do-ant-seq sit-npst ‘He sits down like an American.’ (Noonan and Grunow-Hårsta 2002: 88) Interestingly, we see almost a full cycle here. First we have a full verb, then a copula, then an auxiliary and finally an affix. Below we will see further instances of these separate stages, but it is noticeable that we seem to find all of them gathered in one language. This also speaks in favor of the present theory as we would expect such cycles to emerge in individual languages.



Chapter 9.  The copula cycle 

Munro (1977) argues on the basis of data from Mojave that existential constructions also should be included in the possible sources for copulas (see also Hengeveld 1992: 254–255; and see e.g. Devitt 1994: 138 for a dissenting view on the direction of the change). The discussion centers around sentences like (102). (102) John kwab‘ide:-č Ø-ido-pč John doctor-sbj 3sg-cop-perf ‘John is a doctor.’

Mojave (Munro 1977: 445)

It is problematic to analyze a sentence like (102). The subject is John, although kwab‘de: ‘doctor’ is grammatically marked as the subject. Munro (1977: 450) proposes that the subject actually is John kwab‘ide: ‘John (is) doctor’, i.e a predication embedded by the verb ido. That is, the paraphrase is better given as (103). (103) There is [John (is) doctor] Under this analysis, the subject is the embedded non-verbal predication, and the existential verb may be expected to show agreement with this (inherently third person singular). The following examples show that this prediction is borne out. (104) Many ‘nye-č Ø-ido-pč 2sg 1sg-sbj 3sg-cop-pf ‘You’re me.’ (105) ‘inyep kwab?ide:-č Ø-ido-pč 1sg doctor-sbj 3sg-cop-pf ‘I’m a doctor.’

Mojave

(Munro 1977: 452–453)

However, if the verb in (102) were a regular copula, we expect that it would agree with the subject of the embedded non-verbal predication. This is indeed found. (106) ‘inyep kwab?ide:-č ?-ido-pč 1sg doctor-sbj 1sg-cop-pf ‘I’m a doctor.’

Mojave (Munro 1977: 452)

As Hengeveld (1992: 255) argues: “The fact that [(102)] and [(106)] exist side by side may be taken as a sign of copularization of the existential verb used in a reality construction”. Regarding the analysis of these cases, I claim that this change is fully compatible with the perspective adopted in this paper. Whereas full verbs project a VP and moves from V to Pr (copies indicated by strikethrough below), the copula is merged directly in Pr. In other words, instead of the structure in (107), we have the structure in (108).

 Terje Lohndal

(107) PrP wo subject Pr’ wo Pr VP verb wo V’ wo V verb (108) PrP wo subject Pr’ wo Pr predicate copula In (108) the copula is directly merged in the PrP, thus there is no VP below it. Alternatively, there might be a VP, but then this VP is lexicalized by a full verb (see Pustet 2003 for languages where this is the case), and there is no movement from V to Pr. In other words, this change is a consequence of the Late Merge Principle. The verb goes from a lexical element to a functional element. If argument structure is part of syntax through features (Hornstein 1999, 2001), these features will be lost when a full verb changes into a copula. In the next section, we will look at auxiliaries that develop from copulas. 4.3.2. Auxiliaries Copulas that have developed from full verbs can also develop further. As Kuteva (2001: 67) mentions, a copula derived from a bodily posture verb can develop further into an auxiliary.8 The Latin postural verb stare ‘stand’ developed into the copula estar in Spanish, meaning ‘to be (somewhere, or temporarily)’ (Comrie 1976: 102). In Modern Spanish, this verb has developed into a progressive auxiliary (109). We find the same structure in Italian (110). (109) estoy cantando I.stand singing ‘I am singing.’

Spanish

(110) sto cantando I.stand singing ‘I am singing.’

Italian (Comrie 1976: 102)

8. This shows that semantic restrictions are important to the cycle as well. However, I will not try to deal with the specific role of semantics in the present paper.



Chapter 9.  The copula cycle 

Similar developments are attested in Imonda (Kuteva 2001: 65): (111) agõ-ianèi sabla ed-ia ekuk lõh-ual-fna women-npl two px-loc distance stand-dual-prog ‘The two women were standing there in the distance.’ (112) pilin ed-ia fa-hõdõ-lõh-f. plate px-loc class-put.up-cop-prs ‘The plate is up there.’

Imonda

(Heine and Kuteva 2002: 282)

Though, as Kuteva points out, this is by no means a universal auxiliation process (2001: 67). However, we see traits of it in English, cf. (113). (113) The car is arriving. Devitt (1994) argues that the directionality in these cases is one where copulas develop into auxiliaries. He notes a strong tendency for forms that serve as auxiliaries for progressive or continuous aspect to also appear as copulas in locative complements. This is shown for two languages in (114)-(117). (114) amu e idu tani manji manamu 1p(in) that house in cop.prs.prt cop.1p(in) ‘We are staying in that house.’

Kui

(115) anu lakai mai 1s sacrifice.prs.prt cop.1s (116) nga di der al è 1s come part cop pres ’I am (in the act of) coming.’ (117) magma hpe chum hka ma al è chief masc house in neg cop pres ‘The chief is not in his house.’

Nung

(Devitt 1994: 174)

According to Devitt, in these cases the auxiliaries have developed from copulas. I will now show how these cases can be analyzed. An auxiliary is syntactically different compared to a copula in that the auxiliary is commonly taken to be directly merged in the middle field, the IP domain. Once again, this is a Late Merge effect. The change can be illustrated by comparing the somewhat simplified trees in (118) and (119).

 Terje Lohndal

(118)

IP wo I’ wo I PrP copula wo Pr’ wo P copula

(119)

IP wo I’ wo I PrP auxiliary wo Pr’ wo P VP verb 6 verb

In (118) the copula is merged in PrP and moved to IP (in order to get inflected for tense and other agreement properties). In (119), the auxiliary is directly merged in IP, and there is a different verb in the argument domain of the clause. The next section discusses a slightly different case where a head develops into a new head, namely adpositions becoming copulas. 4.3.3. Adpositions Adpositions may be taken as another possible source of copulas. In Gbeya, the copula nέ “resembles the preposition nέ in several respects” (Samarin 1966: 76). The copula nέ is peculiar because it is used in combination with other copular elements, and “seems to occur exclusively with the verbs f and ya, the singular and plural verbs ‘to be’” (Samarin 1966: 77). (120) wa yá nέ wéey 3pl cop.pl cop man ‘they are men.’ (121) téa K nε búu tree cop.sg. cop white ‘the tree is white.’

Gbeya

(Samarin 1966: 77)



Chapter 9.  The copula cycle 

Frajzyngier (1986) discusses Chadic languages and argues that a grammaticalization from preposition to copula has happened in several of these languages. He calls the copulas in these languages “locative copulas”. A few examples are provided in (122)-(124). (122) kòún à gá àmá buffalo cop in water ‘A buffalo is in the water.’

Bolewa

(123) kòún à gá àmá sà buffalo cop in water not ‘The buffalo is not in the water.’ (124) sùãá-nò à kò réwè shirt-1sg cop on tree ‘My shirt is on the tree.’

(Frajzyngier 1986: 372)

(122)-(124) are all ungrammatical if either à or the preposition following à is deleted. These examples can be categorized as belonging to the type (125). There is another type as well (126), which makes it possible to test whether we are dealing with a copula in (122)-(124), or whether we are dealing with two prepositions in a row. (125) X is in/at Y (126) X verb Z in/at Y If à is a preposition, we would expect it to occur in (125) as well as in (126). However, if à is a copula, we would not expect it to occur in (126). The relevant data are provided in (127)-(129), which show that the copula disappears in this configuration. (127) zéetì sùãá gà kó réwè put shirt in on tree ‘Put the shirt on a tree.’

Bolewa

(128) ísín zòu sùãá-nì gà gà ngírkì 3sg put shirt-3sg in in bag ‘He put his shirt into a bag.’

(Frajzyngier 1986: 372)

(129) mèmú ámà gà kólbà gà gà sàrá-nì person dem with bottle in in hand-3masc ‘That man has a bottle in his hand.’

(Frajzyngier 1986: 373)

Frajzyngier also discusses the direction of the change: Did locative copulas emerge from prepositions or did prepositions emerge from locative copulas? He discards the latter option based on the fact that prepositions are far more frequent than copulas, thus it makes more sense that in some languages a copula emerged than to say that such a copula disappeared from most of the languages. The question is now how we are to analyze this change within the current framework. On standard assumptions, both prepositions and copulas are heads. Adopting

 Terje Lohndal

the proposal in Pesetsky and Torrego (2004:  518), we can assume that prepositions have the following feature content (I have modified the notation slightly). (130) PP wo P DP uphi iphi iCase uCase In (130), uphi stands for unvalued phi-features whereas iphi stands for valued phifeatures. Pesetsky and Torrego (2001, 2004) argue that Case is best understood in terms of a “displaced Tense feature”, but I have just put “Case” instead of T for expository convenience. The unvalued phi-feature makes prepositions into a probe, which is necessary given the Agree framework in Chomsky (2000, 2001, 2004, 2007, 2008). The Case feature is a valued feature on P on this approach. Assuming the Economy of Features principle, we can argue that the learner will try to get rid of this feature if possible. When this Case feature is removed, the feature content is identical to copulas. A possible problem here is that the feature content also resembles complementizers, and a very similar story has been invoked in order to explain how prepositions may become complementizers (van Gelderen 2008b). At present it is not entirely clear what the main difference is. Judging from the data above, it seems that those prepositions that become copulas have a limited distribution. Specifically, they tend to occur in the middle field, which is not the case for many of those prepositions that become complementizers, at least not in English (van Gelderen 2008b). This might provide the learner with a cue, that is, there need to be a certain structural relationship in order for a preposition to become a copula in order for the Head Preference Principle to be involved in relation to the PredP. This is also related to the discussion regarding whether a demonstrative develops into a complementizer or a copula, cf. Section 4.2.1. At present I am not able to formulate very specific criteria for this relationship, but hopefully future research will elucidate this. Heine and Kuteva (2002: 100) also quote a few cases of a locative copula developing into a locative preposition. One such example is le ‘be at’ in Ewe with becomes a preposition meaning ‘at’. (131) agbalέá le kplKá dzí. Book.def be.at table.def on ‘The book is on the table.’ (132) me kpK lKri le mK dzí. 1sg see lorry at street top ‘I saw a lorry on the street.’

Ewe

(Heine and Kuteva 2002: 100)

Heine (1993) and Heine and Kuteva (2002: 101) view this as an instance of grammaticalization. In the present framework, I would view such examples as cases of renewal.



Chapter 9.  The copula cycle 

That is, a cycle starts over again, with the new preposition having a potential to redevelop into a copula, given the adequate external influence such that one of the economy principles is able to work. In this section we have seen a few examples of prepositions that change into copulas. In the next main section, I discuss cases where a head develops into a grammatical marker.

4.4

Head to affix

So far we have seen changes where a specifier becomes a head and where a head becomes a different head. In this section, I will look at cases where heads develop into affixes. Typologically, copulas as affixes are not very frequent: “In the majority of cases, copulas are free morphemes” (Pustet 2003: 39). The copula ni in Yoruba is homophonous with the general focus marker in the language (Stassen 1997: 96): (133) Oni‰òwo ni mi merchant cop 1sg ’I am a merchant.’ (134) Lálέ ni wKn dé in.the.evening fm they arrived ‘It was in the evening that they arrived.’

Yoruba (Ashiwaju 1968: 28)

(Bamgbọse 1966: 37)

Verbal copulas may evolve into affixes. Modern Turkish has two different copular affixes, though they occur in a very restricted set of contexts. One of them is the palatal glide y. Notice that it can only be used after non-consonants (135): (135) (ben) satici-y-im 1sg seller-cop-1sg ‘I am a seller.’

(Kornfilt 1997: 77)

(136) (ben) öğretmen-im 1sg teacher-1sg ‘I am a teacher.’

(Kornfilt 1997: 78)

Turkish

As Kornfilt argues, y is an insertion rule restricted to morpheme boundaries in copula sentences. If it were a general phonological rule of y-insertion between vowels, even restricted to morpheme boundaries, we would expect (137) to be grammatical. (137) *kitab-i oku-du-y-um book-acc read-past-y-1sg

Turkish (Kornfilt 1997: 78)

There is a diachronic fact supporting y as a copula. Until the early years of the Republic (that is, in the early 20th century), tense and agreement markers in copula sentences were attached to a free morpheme i in the past tenses (138)-(139).

 Terje Lohndal

(138) (ben) satici i-di-m 1sg seller cop-pst-1sg ‘I was a seller.’ (139) (ben) öğretmen i-di-m 1sg teacher cop-pst-1sg ‘I was a teacher.’

Turkish

(Kornfilt 1997: 80)

These inflected copular forms then developed to postclitics, and during this change, the copular i turned into the glide y. Still one does find the free inflected copula in written Turkish, according to Kornfilt, which nicely illustrates that there was such a previous stage and that the transition has not yet made it entirely into writing. Thus, the reduced forms are: (140) (ben) satici-y-di-m 1sg seller-cop-pst-1sg ‘I was a seller.’ (141) (ben) öğretmen-di-m 1sg teacher-pst-1sg ‘I was a teacher.’

Turkish

(Kornfilt 1997: 79)

There are other analyses of y as a copula; Jansky (1954: 36) argues that y is a linking consonant that appears in specific phonological contexts. However, I think Kornfilt has provided convincing evidence that this is not an adequate analysis and that y indeed is a true copula. The other copula element in Turkish is the suffix –DIr, a “suppletive form for the copula” (Kornfilt 1997: 81). The historical source of –DIr is the verb ‘to stand’ (Devitt 1994: 128, Pustet 2003: 58). It is used in the third person present only, and it is always optional, but this is the only place where it is optional (Kornfilt 1997: 82). (142) bu cadde-dir this road-cop.3prs ‘this is a road.’ (143) bu cadde this road ‘this is a road.’

Turkish

(Pustet 2003: 58)

Again, we see a copula emerging from a previous full verb.

4.5

Summary

In this section I have provided a number of examples in support of a copula cycle. The cycle, I have argued, looks like (144).



Chapter 9.  The copula cycle 

(144) demonstrative/pronoun > specifier > iF >

copula > head > uF >

grammatical marker affix --

I have argued that the major transitions are those where a specifier becomes a head and a head becomes an affix. However, I have also claimed that there are minor stages within each main stage in (144). In particular, a verb may develop into a copula. Since a verb is already a head, this is not a category change. Instead, I have argued that verbs and copulas are merged in different positions, viz. the VP and the PrP.

5. Conclusion The aim of this paper has been twofold. On the one hand, I have wanted to describe the major patterns in which copulas emerge and disappear, what I have called the copula cycle. On the other hand, I have tried to show that a formal theory of grammaticalization is a good framework in order to account for and explain these changes. When demonstratives and pronouns become heads, this is a change from the specifier of a predication phrase to a head of the same phrase. In terms of features, it means that interpretable features are replaced by uninterpretable features. I have also shown how full verbs become copulas, and subsequently how copulas may turn into auxiliaries or grammatical markers. In sum, I have argued in favor of a copula cycle involving three major stages, but also that these stages in many cases have a more fine-grained internal cartography.

References Åfarli, Tor. 2005. The basic nexus. Ms, Norwegian University of Science and Technology. Abraham, Werner. 2002. Modal verbs: Epistemics in German and English. In Modality and its Interaction with the Verbal System, Sjef Barbiers, Frits Beukema & Wim van der Wurff (eds.), 19–50. Amsterdam: John Benjamins. Adger, David & Ramchand, Gillian. 2003. Predication and equation. Linguistic Inquiry 34:325–359. Ashiwaju, Michael. 1968. Lehrbuch der Yoruba-Sprache. Leipzig: VEB Verlag. Bamgbose, Ayọ. 1966. A Grammar of Yoruba. Cambridge: CUP. Baker, Mark. 2003. Lexical Categories. Cambridge: CUP. Berman, Ruth & Grosu, Alexander. 1976. Aspects of the copula in Modern Hebrew. In Studies in Modern Hebrew Syntax and Semantics, Peter Cole (ed.), 265–285. Amsterdam: North Holland. Bhat, D.N.S. 1994. The Adjectival Category [Studies in Language Companion Series 24]. Amsterdam: John Benjamins. Biberauer, Theresa & Richards, Marc. 2006. True optionality: When the grammar doesn’t mind. In Minimalist Essays [Linguistik Aktuell-Linguistics Today 91], Cedric Boeckx (ed.), 35–67. Amsterdam: John Benjamins. Bowers, John. 1993. The syntax of predication. Linguistic Inquiry 24: 591–656. Chomsky, Noam. 1995. The Minimalist Program. Cambridge MA: The MIT Press.

 Terje Lohndal Chomsky, Noam. 2000. Minimalist inquiries. In Step by Step. Essays on Minimalist Syntax in Honor of Howard Lasnik, Roger Martin, David Michaels & Juan Uriagereka (eds), 89–155. Cambridge MA: The MIT Press. Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.), 1–52. Cambridge MA: The MIT Press. Chomsky, Noam. 2004. Beyond explanatory adequacy. In Structures and Beyond: The Cartography of Syntactic Structure, Vol. 3, Adriana Belletti (ed.), 104–131. Oxford: OUP. Chomsky, Noam. 2007. Approaching UG from below. In Interfaces + Recursion = Language? Chomsky’s Minimalism and the View from Syntax-Semantics, Hans M. Gärtner & Uli Sauerland (eds), 1–30. Berlin: Mouton de Gruyter. Chomsky, Noam. 2008. On phases. In Foundational Issues in Linguistic Theory, Robert Freidin, Carlos P. Otero & Maria-Luisa Zubizaretta (eds), 133–166. Cambridge MA: The MIT Press. Comrie, Bernard. 1976. Aspect: An Introduction to the Study of Verbal Aspect and Related Problems. Cambridge: CUP. Dikken, Marcel den. 2006. Relators and Linkers. The Syntax of Predication, Predicate Inversion, and Copulas. Cambridge MA: The MIT Press. Déprez, Viviane. 2003. Haitian Creole Se: A copula, a pronoun, both or neither? On the double life of a functional head. In Recent Development in Creole Studies, Dany Adone (ed.), 135– 173. Tübingen: Max Niemeyer. Devitt, Daniel. 1994. Copula Constructions in Crosslinguistic and Diachronic Perspective. PhD dissertation, SUNY Buffalo. Diessel, Holger. 1999. Demonstratives: Form, Function, and Grammaticalization [Typological studies in Language 42]. Amsterdam: John Benjamins. Eide, Kristin Melum. 1996. Som-predikativer. Cand.philol. thesis, Norwegian University of Science and Technology. Eide, Kristin M. & Åfarli, Tor A. 1999. The syntactic disguises of the predication operator. Studia Linguistica 53: 155–181. Faarlund, Jan Terje. 2008. A mentalist interpretation of grammaticalization theory. In Grammatical Change and Linguistic Theory: The Rosendal Papers [Linguistik Aktuell-Linguistics Today 113], Thorhallur Eythòrsson (ed.), 221–244. Amsterdam: John Benjamins. Frajzyngier, Zygmunt. 1986. From preposition to copula. In Proceedings of the Twelfth Annual Meeting of the Berkeley Linguistics Society, Vassiliki Nikiforidou, Mary VanClay, Mary Niepokui & Deborah Feder (eds.), 371–386. Berkeley CA: Berkeley Linguistics Society. Gelderen, Elly van. 2004a. Economy, innovation and prescriptivism: From Spec to Head and Head to Head. Journal of Comparative Germanic Linguistics 7: 59–98. Gelderen, Elly van. 2004b. Grammaticalization as Economy [Linguistik Aktuell-Linguistics Today 71]. Amsterdam: John Benjamins. Gelderen, Elly van. 2007. The definiteness cycle in Germanic. Journal of Germanic Linguistics 19: 275–308. Gelderen, Elly van. 2008a. Linguistic cycles and Economy Principle: The role of Universal Grammar in language change. In Grammatical Change and Linguistic Theory: The Rosendal Papers [Linguistik Aktuell-Linguistics Today 113], 245–264. Amsterdam: John Benjamins. Gelderen, Elly van. 2008b. Where did Late Merge go? Grammaticalization as feature economy. Studia Linguistica 62: 287–300. Gelderen, Elly van. 2009. The linguistic cycle. Ms. Gildea, Spike. 1993. The development of tense markers from demonstrative pronouns in Panare (Cariban). Studies in Language 17: 73–73.



Chapter 9.  The copula cycle  Hawkins, John. 2004. Efficiency and Complexity in Grammars. Oxford: OUP. Heine, Bernd. 1993. Auxiliaries: Cognitive Forces and Grammaticalization. Oxford: OUP. Heine, Bernd, Claudi, Ulrike & Hünnemeyer, Friederike. 1993. Grammaticalization: A Conceptual Framework. Chicago IL: University of Chicago Press. Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Heine, Bernd & Kuteva, Tania. 2007. The Genesis of Grammar: A Reconstruction. Oxford: OUP. Hengeveld, Kees. 1992. Non-verbal Predication: Theory, Typology, Diachrony. Berlin:Mouton de Gruyter. Hopper, Paul & Traugott, Elizabeth Closs. 2003. Grammaticalization. Cambridge: CUP. Hornstein, Norbert. 1999. Movement and Control. Linguistic Inquiry 30: 69–96. Hornstein, Norbert. 2001. Move! A theory of construal. Malden: Blackwell. Jansky, Herbert. 1954. Lehrbuch der Türkishen Sprache. Wiesbaden: Harrassowitz. Katz, Aya. 1996. Cyclical Grammaticalization and the Cognitive Link Between Pronoun and Copula. PhD dissertation, Rice University. Kornfilt, Jaklin. 1997. Turkish. London: Routledge. Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase Structure and the Lexicon, Johan Rooryck & Laurie Zaring (eds.), 109–137. Dordrecht: Kluwer. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Journal of Language Variation and Change 13: 199–244. Kuteva, Tania. 2001. Auxiliation: An Enquiry into the Nature of Grammaticalization. Oxford: OUP. Lehmann, Christian. 1985. Grammaticalization: Synchronic variation and diachronic change. Lingua e Stile 20: 303–318. Li, Charles & Thompson, Sandra A. 1977. A mechanism for the development of copula morphemes. In Mechanisms of Syntactic Change, Charles N. Li (ed.), 419–444. Austin TX: University of Texas Press. Lightfoot, David. 1974. The diachronic analysis of English modals. In Historical Linguistics, John M. Anderson & Charles M. Jones (eds), 219–249. Amsterdam: North Holland. Lightfoot, David. 1979. Principles of Diachronic Change. Cambridge: CUP. Lightfoot, David. 1999. The Development of Language. Acquisition, Change and Evolution. Malden MA: Blackwell. Lightfoot, David. 2006. How New Languages Emerge. Cambridge: CUP. Lohndal, Terje. 2006. The phrase structure of the copula. Working Papers in Scandinavian Syntax 79: 47–73. Lohndal, Terje, Nygård, Mari & Åfarli, Tor A. 2008. The structure of copular clauses in Norwegian. Working Papers in Scandinavian Syntax 82: 23–41. Longobardi, Guiseppe. 2001. Formal syntax, diachronic minimalism, and etymology: The history of French chez. Linguistic Inquiry 32: 275–302. Macaulay, Monica. 1996. A Grammar of Chalcatongo Mixtec. Berkeley CA: University of California Press. McWhorter, John. 1995. Looking into the void: Zero copula in the Creole mesolect. American Speech 70: 339–360. McWhorter, John. 1997. Lost in transmission: A case for the independent emergence of the copula in Atlantic creoles. In The Structure and Status of Pidgins and Creoles [Creole Language Library 19], Arthur K. Spears & Donald Winford (eds), 241–21. Amsterdam: John Benjamins. Migge, Bettina. 2002. The origin of the copulas (d/n)a and de in the Eastern Maroon Creole. Diachronica 19: 81–133.

 Terje Lohndal Mikkelsen, Line. 2005. Copular Clauses [Linguistik Aktuell-Linguistics Today 85]. Amsterdam: John Benjamins. Munro, Pamela. 1977. From existential to copula: The history of Yuman BE. In Mechanisms of Syntactic Change, Charles Li (ed.), 445–490. Austin TX: University of Texas Press. Noonan, Michael. 1992. A Grammar of Lango. Berlin: Mouton de Gruyter. Noonan, Michael & Grunow-Hårsta, Karen. 2002. Posture-verbs in two Tibeto-Burman languages of Nepal. In The Linguistics of Sitting, Standing and Lying [Typological Studies in Language 51], John Newman (ed.), 79–101. Amsterdam: John Benjamins. Pesetsky, David & Torrego, Esther. 2001. T-to-C movement: Causes and consequences. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.), 355–426. Cambridge MA: The MIT Press. Pesetsky, David & Torrego, Esther. 2004. Tense, case, and the nature of syntactic categories. In The Syntax of Time, Jacqueline Guéron & Jacqueline Lecarme (ed.), 495–537. Cambridge MA: The MIT Press. Plank, Frans. 1984. The modals story retold. Studies in Language 8: 305–364. Pustet, Regina. 2003. Copulas: Universals in the Categorization of the Lexicon. Oxford: OUP. Radford, Andrew. 2000. Children in search of perfection: Towards a minimalist model of acquisition. Essex Research Reports in Linguistics 34. Richards, Marc. 2004. Object Shift and Scrambling in North and West Germanic: A Case Study in Symmetrical Syntax. PhD dissertation, University of Cambridge. Richards, Marc & Biberauer, Theresa. 2005. Explaining Expl. In The Function of Function Words and Functional Categories [Linguistik Aktuell-Linguistics Today 78], Marcel den Dikken & Christina M. Tortora (eds.), 115–153. Amsterdam: John Benjamins. Roberts, Ian. 1985. Agreement patterns and the development of the English modal auxiliaries. Natural Language and Linguistic Theory 3: 21–58. Roberts, Ian. 1993. Verbs and Diachronic Syntax. Dordrecht: Kluwer. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: CUP. Rothstein, Susan. 1999. Fine-grained structure in the eventuality domain: The semantics of predicative adjective phrases and ‘be’. Natural Language Semantics 7: 347–420. Rothstein, Susan. 2001. Predicates and their Subjects. Dordrecht: Kluwer. Samarin, William J. 1966. The Gbeya Language. Grammar, Texts, and Vocabularies. Berkeley CA: University of California Press. Stassen, Leon. 1997. Intransitive Predication. Oxford: OUP. Traugott, Elizabeth Closs. 1972. A History of English Syntax: A Transformational Approach to the History of English Sentence Structure. New York NY: Holt, Rinehart & Winston. Tucker, Archibald N.1993. A Grammar of Kenya Luo (Dholuo). Köln: Köppe. Turunen, Rigina. 2006. Complex morphosyntactic features of nominal predicates in Erzya. SKY Journal of Linguistics 19: 173–187. Warner, Anthony. 1993. English Auxiliaries: Structure and History. Cambridge: CUP. Westergaard, Marit. 2008. Acquisition and change: On the robustness of the triggering experience for word order cues. Lingua 118: 1841–1863.

chapter 10

Rather On a modal cycle Remus Gergel

University of Tübingen The goal of this chapter is to contribute to the issues of the grammatical cycle and economy in the context of grammar change through (A) focus on an exponent of modalized meaning (which is distinct from the theoretically betterstudied modal auxiliaries); (B) the recognition of a corresponding type of change as an essentially cyclic development from an original meaning including a temporal component to its currently modalized semantics; (C) an initial analysis of such cyclic developments in relationship to grammatical theory and centered on logical forms.

1. Introduction12 The immediate aim of this chapter is to propose an account of the development of the word rather which gave rise to resulting meanings such as the one illustrated in (1).

1. This paper has profited from presentations of related material at Arizona State University, Cornell University, the University of Texas at Austin, and the University of Tübingen, the audiences and hosts of which are acknowledged. Special thanks are due to Sigrid Beck, Elly van Gelderen and Tony Kroch for their comments. I also wish to thank Nicholas Asher, Jacqueline Guéron, Mark Hale, Wayne Harbert, Jack Hoeksema, Sveta Krasikova, John Vanderelst and two anonymous reviewers for valuable feedback. The usual disclaimers apply. 2. The following abbreviations are used: BNC: British National Corpus; CGEL: Cambridge Grammar of the English Language (Huddleston & Pullum 2002); GGL: web-search based on the search engine Google; LF: logical form; QR: quantifier raising; RTS: rather-than structures. Examples retrieved from the Penn-Helsinki corpora of historical English are given with the standard token IDs used in the sources from which they have been extracted. Typically, Middle English examples are prefixed with the notation ‘CM’ in the sources; Old English examples are prefixed as ‘co’. (Cf. Kroch & Taylor 2000; Kroch, Santorini & Delfs 2004; Taylor, Warner, Pintzuk & Beths 2003 for full notational conventions and further philological information regarding the files included in the corpora).

 Remus Gergel



(1) She heard Spanish and Korean, Russian and Chinese, Arabic and Greek, Japanese, German, and French, but rather [q than feel intimidated …], [p she exulted in this variety of human sound]. (P. Auster, The Brooklyn Follies)

While a somewhat larger overview of the distribution of rather will be given in Section 2 below, the non-finite complement of rather in (1) is modalized and compared with an alternative on a scale, more specifically against a salient doxastic background. In view of the facts given (e.g. hearing a variety of languages around her), the natural expectation for the protagonist girl in Paul Auster’s piece of fiction cited above would be to feel intimidated. But instead, what holds true in the same situation is that she exulted in the context she was placed in. (We will discuss additional, frequently available modal backgrounds below). The major concern of this paper will be the key historical and grammatical developments at the syntax-semantics interface. By assumption, I will make crucial use of the level of logical form represented at all synchronic stages (LF; see, for example, Heim & Kratzer 1998 for motivating discussion of this component in a version including the notational variant of movement in the computation of meaning). The developments will be investigated by tracking down the semantically most relevant changes that led from a transparent form-meaning correspondence involving temporality and the comparative morpheme applied to a scale-sensitive item (for contrast, cf. the currently opaque -er obligatorily attached to rath-) towards a modalized meaning in which alternatives are compared. Synchronically, today, rather appears less transparent at least from a purely morphological point of view than in the original input structure to the cycle. But in fact it still compares, if in a less direct way, namely by establishing an ordering relationship between propositions. I will argue that the semantic development is a crucial ingredient in capturing this change. At the same time, as we will see, the linguistic change is only explained in grammar-theoretic terms if we let the semantics operate compositionally in a structure-sensitive manner, i.e. ultimately on tree structures – in our present case, those that are fit for interpretation. Expanding on these ideas, the immediate empirical focus of the chapter is the cyclic development of rather, which will be illustrated from a number of perspectives. Somewhat more specifically, the chapter is structured as follows. In Section 2, I sketch the inventory of relevant forms, meanings and changes. Section 3 gives the analysis and illustrates the developments from the perspective of the cycle. In Section 4, I discuss how this particular cycle may relate to economy and additional considerations on cycles. Finally, Section 5 provides a conclusion.

2. Rather in current and earlier English In this section, I introduce the main meanings and forms of rather in current English, that is, the present explanandum. I then consider the relevant facts that become visible



Chapter 10.  Rather 

from the diachronic trajectory of the word together with its category, meaning, and grammatical distribution.

2.1

Rather in current grammars

According to the Cambridge Grammar of the English Language (CGEL; Huddleston & Pullum 2002) rather can function as a less central governor in scalar inequality. There are four major types of meaning-structure correspondence in present-day English (PDE), cf. (2a-d). (2)

a. b. c. d.

The idiom would rather With bare infinitival and “in preference” meaning Contrastive link, meaning “not, instead of ” Pleonastic use, with rather than equivalent to than alone

The examples in (3) exemplify the four types given by the CGEL in correspondence with (2). (3)

a. b. c. d.

She would rather live in danger than die of loneliness and boredom. Many of them went to jail rather than pay the fine. Care rather than skill is all you need. These people are more likely to be referred to courts rather than to aid panels.

My main focus lies on the type given in (2b) and exemplified in (3b), i.e. the independent modal meaning, which plays a key role in understanding the grammaticalization process. Most typically, this meaning involved in bare-infinitives is indeed bouletic (just as in (3b)), that is, expressing a preference, but other more subtly modalized nuances can obtain as well (cf., e.g., (1) above). In terms of the form involved, we can extend the observation. Such modal meanings can be expressed not only by a bare infinitive, but also with further non-finite complementation; cf. the patterns with a present-participle form constructed with a null subject, shown in (4), as well as the less frequent distribution with an overt subject, as in (5).

(4) If you want to download any of these eBooks directly, rather than using the regular search system you may utilize the following addresses and just download by the etext year. (ggl)

(5) In a way rather than us reading the parables, the parables end up reading us. (ggl) While this type of distribution is language-specific, it is interesting nonetheless in the context of English syntax that when a subject appears at all in the nonfinite pattern, it gets a default case (and not nominative). This suggests that the complement of such rather than constructions is roughly speaking a reduced VP or vP. (But crucially not a TP; cf., e.g., Pesetsky & Torrego 2001 for an account of assignment of nominative case through T.)

 Remus Gergel

The type in (2a) is important both synchronically and diachronically, but in terms of its meaning it falls under the same rubric, namely of (bouletic) modality. It may in fact instantiate a modal harmony effect, from which the modal itself has later been removed. One possibility, then, would be to say that the originally reinforcing adverb rather has taken over the earlier modal function in the examples without the overt modal would (or another equivalent modal element; cf. below). The contrastive type mentioned in (3c) can be directly linked to metalinguistic comparatives. In fact, metalinguistic comparatives are arguably modalized as well (cf. also Giannakidou & Stavrou 2008 with a different focus). In terms of the distribution of the two types in English, an interesting test can be noted between the rather-than structures (RTSs) of primary interest here and the metalinguistic ones, drawing here on Dieterich & Napoli (1982). Consider first (6a) vs. (6b). (6) a. Harry walked to work rather than drive. b. Harry walked to work rather than drove. While a RTS such as (6a) is typically followed by a non-finite main-verb form (in English), the metalinguistic comparative in (6b) takes the finite form in the language.3 The preference or bouletic reading for the non-finite form can be made to fail by leaving the sentence without a logical subject, cf. the weather-verb expletive in (7a) and, by contrast, the lack of such an effect with the metalinguistic placed in the same context in (7b).

(7) a. #It snowed rather than rain. b. It snowed rather than rained.

Finally, the pleonastic type is marginally interesting from a descriptive take on grammaticalization since it seems to instantiate a bleached meaning. (I use this term descriptively; see, for example, von Fintel 1995, Traugott & König 1991, on some pitfalls and paradoxes related with the term ‘bleaching’ if taken to literally mean void of meaning in the general case). While the pleonastic type is fully outside of the concerns about modality addressed in this paper, we can offer an additional syntactic diagnostic that confirms the classification by the CGEL as a separate item. The diagnostic relies on comparative inversion, a relatively restricted possibility already by itself, optionally arising in certain clausal comparatives in English (cf. Culicover & Winkler 2008, Emonds 1970, Gergel 2008, Merchant 2003, among others). Comparative inversion (independently of rather) is illustrated in (8). (8) a. She hasn’t bought as many souvenirs as has her husband. b. Harvard undergrads generally give the impression of being far more supportive of their president than is the faculty. 3. An investigation of metalinguistic comparatives falls beyond the scope of this paper. Notice that the contrastive or metalinguistic comparatives can also appear in reduced structures, as visible in the examples given in the main text.



Chapter 10.  Rather 

Unlike other rather-constructions, the pleonastic type can also appear in conjunction with inversion of the finite element with the subject, as the attested example in (9) shows.

(9) Defined as monopolies that could deliver goods and services more efficiently rather than could a host of competing entities, natural monopoly utility companies would win state sanction to operate in an environment that allowed expansion and use of ever-larger generation technologies. (The Electric Utility Industry in 1965: At the Pinnacle of Success before the Blackout, By Richard F. Hirsh, Virginia Tech, online)

To summarize the subsection, we have seen some of the major types of rather in current English. I will focus on RTSs, the structures followed by a non-finite form, such as the bare infinitive. In the next section, I discuss the major uses (precursor forms of) rather could have in Old and Middle English.

2.2

Rather in earlier English

This subsection offers a sample of the relevant uses of rather at earlier stages of the language. Observing lexical change surrounding rather is certainly not new. The origin of the word is well known in the literature on English and it has been noted in many scholarly sources concerned with the history of the language; cf. Jespersen 1949; The Oxford English Dictionary; Mitchell 1985; Rissanen 1999; Stern 1931, among others. What is new to the best of my knowledge, however, is the investigation of the cyclic nature of this change and its implementation as a systematic type of development in language change related to the level of LF. The key meaning change revolves around an adverb indicating temporally marked properties at the beginning which developed systematically towards a modal element. We next turn to the crucial uses of rather. The origin of rather lies in the comparative form of the adverb (h)ræþe (and its numerous variants; cf. OED), derived from the adjective (h)ræþ, which could convey several properties, among which we find ‘soon’ ‘quick’, ‘swift’. Some Old English uses of (h)ræþe are illustrated in (10). (The sentences in (10), as most of the diachronic data here, have been extracted from the Penn-Helsinki-York corpora of historical of English; see in particular Kroch & Taylor (2000); Kroch, Santorini & Delfs (2004); Taylor et al. (2003). The data are given by way of reference to their standard corpus token identifications.) (10) a. On Sunnandæge mon sceal hraðor arisan to uhtsange. ‘On Sunday one shall earlier rise to morning song.’ (cobenrul, BenR:11.35.4.476) b. Quirinus þa eode to ðam cwarterne hraðe, ‘Quirinus then went to the prison quickly.’ (coaelhom, æHom 24:78.3806)

 Remus Gergel

Forþon hi ne besceawiaþ no hu late hi on þysne therefore they not consider not how late they on this middangeard acennede wurdon, & hu raþe hi him eft world born were and how soon they him afterwards of gewitan sceolan,... of depart shall ‘Therefore they didn’t consider how late they were born on this world and how soon they would depart from it.’ (coblick, HomS_17_[BlHom_5]:59.88.735) c.

Unsurprisingly, (h)ræþe could appear either in the positive or the comparative. Example (11) additionally illustrates the contrast of raþe as ‘soon’ with ‘late’ and the extraction of a degree argument through the implicit question; cf. hu late/hu-raþe, i.e. ‘how early/how late’. The wh-phrases orthogonally involve pied-piping, a language-specific process in degree questions (cf. Corver 1997, Grosu 1994). The temporal component is available in the entry of rathe in the Middle English period as well, as we can see in (11), and it is still available, if decreasingly frequent, in the comparative forms, as shown in (12) below. (11) a. Why ryse ye so rathe, ey, benedicite! ‘Why do you rise so early…?’ (Geoffrey Chaucer, The Miller’s Tale) b. and al so raþe he was iwarisd of his maladie. ‘and all so soon he was cured of his sickness.’ (cmkentse, 218.108) (12) a.

for þat Sonday was of þe raþer Ȝere, and nouȜt of þe newe ‘because that Sunday was of the earlier year and not of the new Ȝere þat... year that…’(cmpolych, VI, 101.709)

b. þe kyng blamede hym for he warned hym nought raþer; ‘the king blamed him because he warned him not earlier.’ (cmpolych, VI, 437.3207) The so-called contrastive link mentioned by the CGEL (cf. Section 2.1) is also observable in Middle English texts, clearly with antonymic contrasts as in (13a) and (13b). But the alternatives introduced can be more diverse as well, as illustrated in (13c). While this reading may not be crucial for the development from temporal to modal meanings, it illustrates the range of possibilities in terms of alternatives that rather could compare. (The possibility of the contrastive link is certainly continued towards the early modern period, as illustrated in (14) for EModE.) (13) a. I rede thee, certes, that thou, Lord, werke in swich wise with thy cherles that they rather love thee than drede. (cmctpars, 314.C1.1112) b. for he wold raþer gon bakward þan forward. (cmkempe, 10.179)



Chapter 10.  Rather 

c. for aȜenst an hondred of Egbert his knyȜtes, þat were pale men and lene, come a þowsand þat were rody and fat, and were raþer i-stuffed wiþ swoot þan with blood (cmpolych, VI, 289.2128) (14) a. ffor many Shippys and galyes towche ther rather thanne at Parence. (torkingt-E1-P1, 16.234) b. and rost him, basted often with Vinegar, or rather verjuice and butter, with good store of salt mixt with it. (walton-E3-P1,218.19) But turning to the Early Modern English period from the perspective of the intensional readings, namely the temporal and modal ones, we can easily observe that the overall availability of temporal interpretation of rath(er) virtually disappears. The other readings persist: the contrastive link, the modal readings joined by auxiliaries (for example, had is frequently found with a modal meaning from this time on; cf. (15a)), and also the bare infinitive with a distant selector (that is, not yet one that is necessarily adjacent to the than-clause), cf. (15b). Example (15c) with a temporal interpretation seems to be already a relic by the early ModE times. (15) a. he had rather be unknown and obscure (boethpr-E3-H,126.44) b. and chuseth rather to withdraw from himself many natural Pleasures, than run the hazard of losing that Money which he hath gathered. (boethpr-E3-H, 126.45) c. All the stocke thou cost of later or rather, From thy first fathers grandfathers fathers father, Nor all that shall come of thee to the worldes ende, Though to three score generations they descende, Can be able to make me a iust recompense, For this trespasse of thine and this one offense. (udall-E1-P2,L1209.382) Moving, then, from a temporally meaningful element to a characteristically modalized item, we need to answer the question what precisely happens in terms of the semantic representation involved. The most important facts to be explained in the remainder of this chapter are the following: (a) How did the change develop – Is there a systematic characterization in this connection? (b) What motivated the change? (c) What is the LF-role of the comparative morpheme –er?

3. The analysis In this section, I present the analysis of the main developmental stages of rather couched in terms of LF structure. In the first subsection, I discuss the formal tools required. In the second part, I illustrate how the change can be characterized in terms of logical forms. While the section requires some minimal semantic formalism, all of the tools introduced here are independently motivated.

 Remus Gergel

3.1

Introducing the semantics used

In this first subsection, I introduce the basic ingredients of the analysis that are necessary for a formal account of RTSs. To achieve that, I briefly discuss the issue of compositionality in language change; then I present the essentials in the semantics of quantifier raising (QR), comparatives and modality, respectively, that will be used further. To begin, there is a first sense of compositionality involved in language change which is usually formulated along the following lines. Developments in terms of meaning change can only be fully understood if we consider them at the propositional level. This first step is a departure from restrictions of the traditional research on semantic (alias lexical) change and is adopted here. While it is one word, the change of which is most conspicuously noticed when inspecting diachronic data, there are a series of other factors that change in relationship with the visible culprit in many interesting cases. Alongside potential morphological and phonological change, both the surrounding tree geometry of the word (including LF for the purposes of interpretation) and the way its lexical entry combines with the other nodes of the clause can thus typically change; cf. Eckardt (2007) for a perspicuous illustration of the latter based on going to. The idea that meaning change is more than lexical change or pragmatic conventionalization of single items is not new, but its more systematic exploitation is relatively recent (cf. Eckardt 2007 and Traugott & Dasher 2001, among others). What I would like to add to the picture is how a semantically motivated (and realized) movement such as the type observed in QR may effect a language change phenomenon. Overall, I would like to adopt a Fregean version of compositionality here and apply it to the diachronic case study at hand. This means in updated terms that a clear sense of the structure on which the interpretation principles can apply at every node in an LF tree needs to be addressed. To achieve that, we can next introduce a prerequisite, namely the standard version of QR based on movement (cf. May 1977; Heim & Kratzer 1998). Later in the chapter, we will see that the changes involved in rather will make use of the same types of mechanisms as QR transferred to degrees and times instead of individuals. But first things first: A classical topic in semantic theory is the issue of quantifiers in object position; cf. every park in (16). (16) Sue liked every park. Simply put, the issue arises through the following paradox. On the one hand, an object needs to saturate the first of the individual slots in the logical type of the transitive verb, namely <e, <e,t>>. So, the object must be of the type of the required individual for functional application to apply, that is <e>. On the other hand, however, there is a large body of evidence that quantifiers yield very distinct truth-conditional effects from those obtaining with individual-denoting DPs (in tautologies, contradiction scenarios etc.; cf. Heim & Kratzer 1998 for an overview). A way to solve the dilemma then, which we adopt here, is to move the quantifier phrase to a sister position of a



Chapter 10.  Rather 

truth-value denoting node (typically at the level of the IP/TP-adjunction), introduce a movement index via the process of predicate abstraction and give the moved quantifier phrase its rightful and independently expected logical type, namely t>, which can now moreover combine with the rest of the clause. In the low position, from which the quantifier started out the derivation, a trace of type <e> now saturates the object slot of the transitive verb and is bound by the movement index. This solves the paradox. The process is schematized in (17). (17) [every park [1 [Sue [likes t ]]]] It is possible to apply the same mechanics to other domains, in particular the comparative morpheme -er, which is the equivalent of a quantifier over degrees (see, for example, Beck 2008 and Heim 2000). In logical terms, this morpheme takes the thanphrase as an argument and is raised with it at LF, just like a quantifier with its first-argument sister NP. Notice that in its original (in outdated speech D-structure) position a gradable adjective requires saturation by a degree argument. This is parallel to the individual-type argument observed with a regular quantifier above. The comparative -er thus ends up, in this case also via movement, operating on two sets of degrees, paralleling the relation on sets of individuals that a run-of-the-mill generalized quantifier operates on. (18) Tempe is larger than Tübingen (is). (19) a. [ [-er [1 [than Tübingen is t large] ] ] [1 [Tempe is t large ] ] b. [[-er]] (λd ε Dd. Tübingen is d-large) (λd ε Dd.Tempe is d-large) The degree d to which Tempe is large exceeds/is greater than (e.g. on the population or surface scale) the degree d´ to which Tübingen is large. Closely related to the degree semantics illustrated above comes the notion of temporal comparison for which I draw on von Stechow’s (2006) approach to comparative adverbs of the sooner/later type (and some of their German equivalents e.g. früher/ später). For a relevant sentence such as (20), von Stechow proposes the LF in (21). (20) Alla came later than Caroline.

 Remus Gergel

(21) von Stechow’s Logical Form for temporal comparatives

While not all the details of this LF are relevant to the diachronic change, let me mention its basic features. First, a temporal adverb such as late or early is originally merged to a position adjoined to an AspP, which denotes a property of times, , (equivalently: a set of times, or a characteristic function of such a set). Following usual practice, the adverb then intersectively combines with the AspP yielding another AspP (of the same logical type, but now restricted by the additional condition that the set of times in question fulfill the condition imposed by late/early). Since the adverb comes as a comparative, it behaves as a quantifier, and it undergoes QR. In fact, it does so with its argument, the than-clause. The final ingredient required is modality. The appropriate interpretable structure for modality is closely tied to the research history of the conditional. While different bracketing options for the conditional have been proposed, Kratzer (1981, 1991, et seq.) has argued that a particularly suitable LF-structure is the one in which (semantically) the modal brackets with a restrictor first, and only then takes its nuclear scope, i.e. the “modalized” proposition. The restrictor can be either just the implicit contextual one providing the background (e.g. whether it is “deontic”, “bouletic”, “epistemic” etc.) or it can be enriched by a conditional clause. This yields the schema in (23) below, following von Fintel & Heim 2007, or in a more basic version the one in (22). I use the usual type-theoretic conventions, in which s stands for the type of possible worlds, t for truth values, v for events (and, largely equivalently, situations), i for times, e for



Chapter 10.  Rather 

individuals (or “entities” in the Montagovian tradition) and d for degrees. If a and b are types, then is also a type. In particular, it is useful to think of a denotation of type as a set of possible events/situations, as a set of times (or, equivalently, characteristic functions of such sets) etc. (22) Main Scopal Relationships for Modals [Modal [Restrictor]] [Proposition P] (23) Logical Form for Modality

(cf., e.g., Kratzer 1981 et seq.) (cf. von Fintel & Heim 2007)

wo , t> <s, t> 3 6 Modal <s, t> ‘P’ 3 , <s, t>> <s, t> 3 6 R w* ‘if Q’ Two notational amendments will be made to this. First, since I will not include contextual information in the logical trees, a simpler version will suffice for my purposes. But I will be explicit about the logical forms involved even in the simple versions since they are important for the current argument. The simplification will consist in having the restrictor (e.g. “if Q” in (23) above) directly as an argument of the modal, without R and w*. A second amendment has to do with the types involved. Following Kratzer (2007), among others, I will use possible situations/events to represent modality instead of the classic possible worlds. Plainly put, this will amount to representing the types of propositional sub-trees such as P and Q above as sets of situations/events rather than sets of possible worlds, i.e. of type instead of <s,t>, cf. the implementation in Section 3.2 below.

3.2

Change from temporal to modal meaning

By capitalizing on the research in semantics reviewed above, this subsection establishes the input and output grammars in the change of rather and offers an explanation of its basic developments in terms of logical forms. A key role will thus be played by the structural make-up of the interpretable clausal structure in which rather participates before and after the change. The starting point of the change is pragmatic. We make the general inertia assumption of diachronic syntax, namely that grammatical systems and in particular phrase-structures are mapped from their predecessors restrictively, if not perfectly (cf. e.g. Kroch et al. 2000, Roberts 2007). While the present claim is that the tree-geometry in terms of LF is significant in the dynamics of the change, notice that it can thus also hardly be expected to be an initiating factor of

 Remus Gergel

a semantic change. What is frequently the case, however, is that semantic change is pragmatically induced (cf. Eckardt 2006 and references cited there). We will adopt this motivation for the inception of the RTS change, too. But the question will be raised whether the pragmatic factor also fully explains the change. To place the discussion on a concrete footing, I next divide it into three parts tied to specific developments affecting logical forms. In terms of the LF structure involved then, the main stages were as follows: Main stage 1: Borrowing time scales for comparisons Earlier English rath(er) induced a temporally related scale, which could be exploited for degree constructions including comparatives, as seen in Section 3.1. This step involves a minimal adaptation from tense to a standard degree scale. It is one that can be accounted for synchronically (adopting von Stechow’s 2006 approach introduced above). Under such a view, the LF that has incorporated times as degrees and serves as the input to the change looks essentially as (24). (24) Pre-Reanalysis main comparative-temporal structure -er



‘rath (Q)’

λt

Past AspP () ‘rath (P)’

Main stage 2: Pragmatic Overload Cross-linguistically, modalized situations which are preferred or more likely, are frequently communicated through expressions originally meaning earlier, faster etc. To implement the observation, we can use, for instance, Eckardt’s (2006) notion of side-message. This is, in essence, nothing but an implicature at the beginning. The next point of the change is the one at which a side-message incorporates to the next-generation semantic entry. The net result here is that (the characteristic function of) the set of times in (24) denoted by the denotations (“Asp-phrases”) are not sets of times any longer, but sets of situations, now in a contextually given modal ordering (e.g. w.r.t. desires). Regarding the latter, we draw on e.g. Heim’s 1992 possibilistic account for desire predicates, but with two amendments: (i) the alternative to p is not necessarily non-p; cf. Villalta 2006 for scenarios independent of rather which make this extension plausible, and we note that RTSs are a cheap way to order distinct p and q; (ii) propositional subtrees denote sets of situations/events here.



Chapter 10.  Rather 

Main stage 3: Misguided interpretation function The change from sets of times to situations in the AspP becomes fatal for compositionally interpreting a temporal structure. Without the set of times, Past can serve no function, no interpretation principle can apply (a standard temporal reference, is introduced structurally higher-up, to the newly formed structure, but it is irrelevant for the node merging Past with AspP in post-change grammars). As a consequence, one is then stuck with two sets of situations and the rather predicate in what was a dependency originally created through movement. The latter is re-interpreted as being a firstmerged relation, rather than a moved generalized quantifier and, finally, the (syncategorematic) movement index has no application either (no compositional rule can use it) and is erased. The derivation of this output is schematized in (25) below. (25) Post-Reanalysis Modal Structure … Past rather ‘(Q)’



λt Past



AspP () ‘(P)’

The diachronic development is thus given a specific merge-over-move implementation that pertains to the semantic developments addressed here. This makes a good prediction in that it correlates with the cyclical character and the unidirectionality of the change. While changes of the rather-type are frequent cross-linguistically (perhaps precisely due to the easy availability of the starting implicature), the reverse does not seem to happen (even though an implicature that Jones does P faster because he utters that he prefers to do it would not be, per se, unimaginable).

4. More on cyclicity and economy In this section, I investigate how some additional considerations pertaining to linguistic cycles and economy carry over to the semantic cycle induced by rather. The first subsection gives some more details coming from diverse empirical areas (clause-type and interaction with modality in particular). The second part notes similar developments to rather. The final subsection explores to what extent “narrow-syntactic” and other economy constraints carry over to the LF-development.

 Remus Gergel

4.1

More towards modalizing rather

In this subsection, I address certain issues in the developmental stages of rather with particular focus on the Middle English and the Modern periods, which I take to be the crucial span for its grammaticalization. These issues include the transition towards independent modal meanings and the clausal patterns that chronologically preceded the bare-infinitve RTSs available in PDE. A central role is played by the developments towards modalized meanings. While this type of transition is observable at all attested stages of the language, it gains particular momentum in Middle English. (Only very few relics are left of the earlier temporal meaning in ModE.) Various modal constructions, including modal verbs or premodals, can be found in particular in the rather-clauses of Middle English.4 This is illustrated in (26a-c) with examples in which the modal is left-adjacent to rather, and in (26d) with the modal taking scope over both main and rather-clause. The rather-clause in the latter case is an infinitive, which is still selected by the modal, but due to the greater distance, it gives the effect of a quite modern complementation pattern, namely the “bare-infinitive” one appearing in the rather clause. (We return to the types of clauses involved in Middle English rather constructions below). Multiple modals used distributively over matrix and subordinate as in (26e) are also quite possible. (26) a. certes youre wyf oghte rather to be preised than yblamed. ‘Surely your wife ought rather to be praised than blamed.’ (cmctmeli,221.C1.153) b. For peraventure the nature of som man is so overthrowynge to for perhaps the nature of some man is so turned over to yvel, and so uncovenable, that the nedy poverte of his houshold evil and so inappropriate that the needy poverty of his household myghte rather egren hym to don felonyes might rather provoke him to commit misdeeds (cmboeth,453.C2.541) c. and wolde rather dye than lese þe right thereof. ‘and would rather die than lose the right thereof.’ (cmedmund,172.280) 4. Other means of introducing modal meanings and related LFs are possible, but are less clearly detectable from what I have seen. For example, an important factor in the distribution of non-temporal rather is the idiom þe raþer, available abundantly in Old English and still found in Middle English. This includes an original instrumental that also gave rise to the so-called correlative comparative of PDE (the more, the better). An analysis of such constructions views them as conditionals (Beck 1997). Recalling that a modal LF is in essence the LF of a conditional (the latter serving as restrictor), there is thus a possibility that the two developments have a common origin and a related LF.



Chapter 10.  Rather 

d.

And so sall gude dedis owtewarde noghte hyndire thi and so shall good deeds on the outside not hinder your deuocyone, bot raþer make it mare. devotion but rather make it greater (cmrolltr,33.699)

e.

but rather than I sholde be dishonoured, there wolde som good but rather than I should be dishonored there would some good man take my quarell. man take my side (cmmalory,36.1144)

An additional comment is in order regarding the modals. While the originally volitional would (together with variants) is particularly frequent, we can see from the above examples that other premodals such as ought, might and in particular shall cooccurred with rather as well. In terms of distribution, we have already noted that the rather structures of earlier English functioned distinctly from those of PDE, in that we do not find unselected bare infinitives but rather tensed full-fledged clauses. This is not to say that we do not find tensed clauses of various sorts in PDE as well (in particular the metalinguistic type comes to mind in this connection). We could claim that the cases in which the complement of rather is phrasal on the surface (and hence not clausal) is a full clause underlyingly, with mechanisms of ellipsis including comparative deletion at work (cf. Lechner 2004 for such derivations of comparatives in general). But the point is a different one. It seems that the grammaticalized patterns that are possible have changed, resulting in the new possibility of having tighter syntactic structure as well, specifically the bare forms. Infinitives at earlier stages were generally governed by an additional element that required them on independent grounds, as shown with the modals above. Without an independent governor (where the term is used in a descriptive sense as a selector), the most productive pattern is the one with full-fledged tensed clauses, and the possibility of having overt (and distinct) subjects in each of them. (27) Yet seye I nat that ye shul rather pursue to youre adversaries for pees than they shuln to yow. (cmctmeli, 235.C2.725) The possibility illustrated in (27) cannot be blamed entirely on parallelism requirements that would have imposed an overt subject in the embedded clause because of the overt subject in the matrix. It is possible to have non-overt subjects such as PRO in a potential superordinate and overt subjects in a subordinate clause, as in (28). (28)

Where is now so gret loue, zele, and fauour vn-to men of holy churche where is now so great love, zeal, and favor toward men of holy church and to þe pepull as had þe gret emperour Constantyne, þe wiche chose and to the people as had the great emperor Constantine who chose raþur [ pro all is liff tyme to be smytte with a leper] þan [he wold rather all his life time to be smitten with leprosy than he would

 Remus Gergel

suffure þe innocentes blod to be shed to saue hym]. allow the innocents blood to be shed to save him? (cmroyal, 253.225) Rather clauses have a clear propositional status. Their rich structure also becomes evident from an inspection of non-finite structures in ME. There are two immediate areas in which evidence can be gathered for this claim. First, for-to infinitives (that is, CP structures), are available in the complement position of rather, cf. (29). (29) and bade hym holde vppe the right of Holy Churche with alle hys myght and asked him hold up the right of holy churche with all his might and rather for to suffre dethe than lese the fredome of the Churche, …. and rather for to suffer death than lose the freedom of the church (cmedmund, 172.273) Second, another non-finite structure that appears larger than expected, at least from the chronologically backwards and grammaticalized perspective of the bare-infinitive RTSs, is generated by the appearance of to in examples of the type in (30). (30) & saide þat þai wolde neuer faile Kyng Arture, and raþere to ‘and said that they would never fail King Arthur and (would) rather bene dede: be dead.’ (cmbrut3, 82.2486) All in all, historically there are then two corroborating facts: clause tightening in ModE compared to the preceding period and the possibility of modalization which developed from overt modals joined by rather in an adverbial function towards the additional possibility of modalization through rather itself. Turning to current grammar, we can, of course, still witness rather with modals in the main clause, but it can appear as an independent modalizer as well, as mentioned. Further, modal iteration is also possible internally to the structure modalized by rather, another positive expectation for modality in general (cf. von Fintel & Heim 2007). The modal rather of RTSs itself, as expected, cannot iterate with actual core modals due to its non-finiteness requirement in English. But once we switch to periphrastic modal expressions, it becomes possible to find some corresponding examples generated, as the following attested examples with participles and bare-infinitives illustrate. (31) a. ‘Let’s try to be the first to create a network that opens a new window of distribution for us rather than having to go hat in hand to a USA or a Nick at Night or a TBS,’... (NY Times, Nov14, 2005) b. Hospitals will receive a flat fee of $350 for a series of seven clinic visits rather than being allowed to charge a fee for each visit. (NY Times, Dec 5, 1989) c. Most of the key members of the Department... are resigning. In fact, just about a week ago one resigned rather than come and have to testify under oath. (ggl)



Chapter 10.  Rather 

d. Contrast this with 104, where I sit at a table with the students, and where discussions are much more organic, and I facilitate rather than need to direct. (ggl) The interaction with the classical overt modal restrictors (recall that these were ifclauses) is harder to observe, but it is possible in some reduced cases: (32) Feedback is more useful if given soon after an event rather than if delayed. (UC Davis, Faculty Handbook, online) Further relics of the comparative nature of rather can be observed in examples such as (33) below; cf. modification by differentials of degree such as much and far. (33) a. I’d much rather be with the boys. (Rolling Stones) b. I’d far rather tell political jokes than be one. (The Independent 06/10, 1999)

4.2

Brief excursus into partially similar changes

A similar change in progress is also observable in English: cf. sooner and as soon in (34) and (35) below. (34) Anna would have cut off her hand sooner than have brought the girl to harm; (Mary Roberts Rinehart, The Street of Seven Stars, retrieved online) (35) Not what one expected of the wife of the senior partner, Tim observed, thinking smugly of his own immaculately turned-out Patrice, who would refuse to eat if she put on even an extra pound and who would as soon leave the house naked as without make-up. (bnc, AB9–2413) Besides the transparency of soon itself, the fact that this is a change less fully developed from the perspective of the cycle can be quickly seen from two distributional requirements. First, it appears to express preference (to the detriment of temporality) primarily only joined by a modal and, second, it has a wider (less fossilized) distribution in terms of the comparative form, including the equative, as illustrated in (35) above.5 Other languages give rise to similar constructions, and I only mention German here, because the modal flavor arising with eher (lit. ‘sooner’, ‘earlier’) comes close to an epistemic, likelihood reading. (36) a. [Context:] Die Perspektiven im Dienstleistungsbereich… werden in Berlin als gut eingeschätzt. (fhtw university memo, economic forecast) ‘The prospects in the service branch in Berlin are considered to be positive.’ b. Der Industriesektor wird eher nicht wachsen. the industry-branch will rather (lit.: sooner) not grow ‘For the industrial branch, it is more likely that it will not grow.’ 5.

Thanks to Jack Hoeksema for raising the issue of the equative at the workshop.

 Remus Gergel

Changes that led to elements similar to rather involving transformed comparatives and/or temporal elements are not hard to come by in other languages either. (Gergel in prep., for example, describes some of the apparently numerous cross-linguistic patterns.) But there is a caveat. The claim is not that everything that develops a semantics of modal ordering comes out of a movement or a comparative dependency. Such LFbased movement dependencies are rather one source of the construction. In some cases it may also be interesting to investigate changes that did not take place. Elly van Gelderen (p.c.) presents a particularly striking type of example (with a morphological twist): Alongside raþe, in Old English we find the form hrædlice, which however never took off in the sense of the cycle presented here. A relevant example (with a temporal meaning only) is (37): (37) & syððon hrædlice wendon westweard on Oxnafordscire. ‘and then soon turned westwards to Oxfordshire.’ (Peterborough Chronicle 1010.15) One possibility might perhaps be that this adverb is already born in the wrong place in the syntax (i.e. merged too high) to be able to undergo the LF-movement from a low position that is necessary for the change. There is a range of possibilities here for further research.6 Another (non-disjunctive) possibility, is that this adverb did not get into the stage of pragmatic overload (recall that this typically creates the potential to trigger the type of change) because there were not enough instances of the two meanings (in a comparative form of it, that is, which is infrequently attested) that would have had to compete.

4.3

Cycle theory

In this subsection, I discuss to what extent regularities uncovered in the research on cycles in general and in particular of the type conducted in narrow syntax carries over to meaning-structural developments. First, the development we have inspected in RTSs has the appearance of being cyclic in a simple intuitive sense which is reminiscent of other cycles: rather, the crucial ingredient in the change is added to certain intensional constructions and it comes to express the modal ordering by itself in the course of the change. Second, in a more theoretical sense, the change is characterized by upward movement in the LF structure and the loss of a movement dependency to the detriment of an option relying on (First)-Merge (or in essence synonymously: external Merge). It should have become clear from the analysis above that Move dies and Merge is preferred. To name but one recent work, this is reminiscent, for example, of Roberts & 6. A somewhat different type of restriction may also be worth noting for potential contrasting purposes: For one relatively well-known type of arguably high adverbs in German, namely a class built with the ending –weise, a comparative form is lacking.



Chapter 10.  Rather 

Roussou’s (2003) observation on the rise of certain functional categories from former movement dependencies. It in fact instantiates a general schema of a semantic counterpart (based on the core of QR) to such syntactic considerations. But there are also additional, specific considerations regarding the cycle to be addressed. We can investigate, for example, in which respects the change in RTSs turns out to conform to cycle and economy principles observed for syntactic change. I will illustrate this based on van Gelderen’s (2006) characteristics of cyclicity. One way in which RTSs show an economy effect that has been observed in the specialized literature concerned with linguistic cycle is by adhering to a version of Late Merge. Consider van Gelderen’s Late Merge Principle given in (38) below: (38) Late Merge Principle Merge as late as possible. A case in point to illustrate the syntactic development would be the history of an adverb such as actually developing out of an adjective, first towards a low VP-adjoined adverb that was synchronically perhaps moved and later, via (external) merge alone, towards a sentential adverb. There can be little doubt that the case of rather reproduces a somewhat similar trajectory. Since I focused on the semantic developments, let me recapitulate the major steps: namely temporal interpretation with a semantically bound trace low down and the step with a directly high-merged modal element after completion of the change, in which it was not possible to interpret the original element merged low down any longer. The similarity on an abstract level is striking and it perhaps raises the question of whether it would not be desirable to have just one explanation instead of two. But the real question is whether we are dealing with entirely the same phenomenon. Despite the abstract structural similarity, there are differences in the two classes. I don’t see how a merger-site effect alone could derive the meaning change witnessed in rather, which relies on interpretability (and other factors). Nor do I think, conversely, that an explanation in terms of compositional applicability of principles of interpretation alone can (or should) derive the large body of evidence gathered for example from the research on adverbs. While there remains (also) syntactic work in the area of rather and a better understanding of semantic effects might enrich our understanding in language change, including adverbs, a more crucial point emerges, namely that alongside other better-known factors (pragmatics, morphological triggers etc.), both the narrow-syntactic and the LF component indeed make reference to structure. Two other syntactic principles for which we can raise the question what kinds of correlates they yield in the realm of meaning change and in particular in the present casestudy of rather are given in (39) and (40) below, also drawn from van Gelderen’s work. (39) Head Preference Principle Be a head, rather than a phrase. (40) Specifier Incorporation Principle When possible, be a specifier (rather than an adjunct).

(hpp) (sip)

 Remus Gergel

The two principles reproduced above make direct reference to specific assumptions in the well-known X’-schema. Meaning per se is not sensitive to the particular shape of the schema – for example either the node T’ or its sister, the subject-denoting DP, can serve as a function taking the other constituent as an argument. The choice depends on whether the subject is individual-denoting or a quantifier. So the notation of the X´-schema does not affect meaning as such. But the computation of meaning is standardly calculated on the skeleton of a similar binary tree structure and I would like to argue that something similar to the syntactic tendencies can be detected in the present case study when we transfer the core insight of the observations to the LF context. How does this then specifically relate to the aspects of the change of rather? One observation to be culled is this: while the pre-change LF had the temporal adverb merged low and interpreted it intersectively as a modifier via its trace (recall von Stechow’s proposal on this), in the reanalyzed LF, rather participated in core argument-taking operations (functional application). The two arguments were the two propositions. Thus while the LF may not be able to talk about specifiers and heads, it nonetheless can talk about modifier vs. argument-structures, with rather developing towards the latter and thus paralleling the syntactic tendency observed above on the level of meaning. The fact that on the preference reading the than-clause must be right-adjacent to its selector (namely rather) in PDE offers additional evidence on the surface for this development toward argument-status from the perspective of functional application. Whether the latter type of tendency holds more generally of LF changes is not easy to predict and I leave the generalization for future work on structure-sensitive semantic change, when more cases are studied from this perspective. All in all, the emerging observation seems to be that the economy of derivation in the computation of LF follows directly from the way interpretation applies to the syntactic tree.

5. Conclusion This chapter has investigated the connection between structure and meaning change by focusing on the thus-far neglected level of logical form. I have argued that the changes that took place in the history of rather instantiate a cyclic development that led to a particular modal element expressing an ordering relationship between two propositions. While the research reported here is in its beginning, I hope to have offered an explanation at least to a part of the linguistic puzzle, namely why this particular type of change may be both frequent cross-linguistically and unidirectional. Clearly, this requires further investigation both in other languages and in English.



Chapter 10.  Rather 

References Beck, Sigrid. 1997. On the semantics of comparative conditionals. Linguistics and Philosophy 2: 229–271. Beck, Sigrid. 2008. Comparatives and superlatives. To appear in Semantics: An International Handbook of Natural Language Meaning, Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.). Berlin: Mouton de Gruyter. Corver, Norbert. 1997. Much-support as last resort. Linguistic Inquiry 28: 119–64. Culicover, Peter & Winkler, Susanne. 2008. English focus inversion constructions. Journal of Linguistics 44: 625–658. Dieterich, Thomas G. & Napoli, Donna. 1982. Comparative ‘rather’. Journal of Linguistics 18: 137–65. Eckardt, Regine. 2006. Meaning Change in Grammaticalization: An Enquiry into Semantic Reanalysis. Oxford: OUP. Eckardt, Regine. 2007. Grammaticalization and semantic reanalysis. http:// semanticsarchive.net. Emonds, Joseph. 1970. Root and structure preserving transformations. Bloomington IN: Indiana University Linguistics Club Fintel, Kai von. 1995. The formal semantics of grammaticalization. NELS 25: 175–189. Fintel, Kai von & Heim, Irene. 2007. Intensional semantics. Lecture notes. Ms, MIT. Gelderen, Elly van. 2004. Grammaticalization as Economy [Linguistik Aktuell-Linguistics Today 71]. Amsterdam: John Benjamins. Gelderen, Elly van. 2006. Linguistic cycles and economy principle: The role of universal grammar in language change. Ms, Arizona State University. (Also published as Gelderen, Elly van. 2008. In Grammatical Change and Linguistic Theory: The Rosendal Papers, [Linguistik Aktuell-Linguistics Today 113], Thórhallur Eythórsson (ed.), 245–264. Amsterdam: John Benjamins). Gergel, Remus. 2008. Comparative inversion: A diachronic study. The Journal of Comparative Germanic Linguistics 11: 191–211. Gergel, Remus. In preparation. Temporal comparisons and ways they change to modal expressions. Ms, Universität Tübingen. Giannakidou, Anastasia & Stavrou, Melita. 2008. On metalinguistic comparatives and negation in Greek. Ms, University of Chicago and Aristotle University of Thessaloniki. Grosu, Alexander. 1994. Three Studies in Locality and Case. London: Routledge. Heim, Irene. 1992. Presupposition projection and the semantics of attitude verbs. Journal of Semantics 9: 183–221. Heim, Irene. 2000. Degree operators and scope. SALT X: 40–64. Heim, Irene & Kratzer, Angelika. 1998. Semantics in Generative Grammar. Malden MA: Blackwell. Huddleston, Rodney & Pullum, Geoffrey. 2002. The Cambridge Grammar of the English Language. Cambridge: CUP. Jespersen, Otto. 1949[2006]. A Modern English Grammar on Historical Principles, Vol. 2: Syntax. London: Routledge. Kratzer, Angelika. 1981. The notional category of modality. In Words, Worlds, and Contexts, Hans-Jürgen Eikmeyer & Hannes Rieser (eds.), 38–74. Berlin: Mouton de Gruyter. Kratzer, Angelika. 1991. Modality. Semantik. Ein internationales Handbuch der zeitgenössischen Forschung, Armin von Stechow & Dieter Wunderlich (eds), 639–650. Berlin: Walter de Gruyter.

 Remus Gergel Kratzer, Angelika. 2007. Situations in natural language semantics. In Stanford Encyclopedia of Philosophy. Edward N. Zalta (ed.), http://plato.stanford.edu/entries/situations-semantics. Kroch, Anthony, Santorini, Beatrice & Delfs, Laurel. 2004. Penn-Helsinki Parsed Corpus of Early Modern English. Kroch, Anthony & Taylor, Ann. 2000. Penn-Helsinki Parsed Corpus of Middle English. Kroch, Anthony, Taylor, Ann & Ringe, Donald. 2000. The Middle English verb-second constraint: A case study in language contact and language change. In Textual Parameters in Old Language [Current Issue in Linguistic Theory 195], Susan Herring, Piet van Reenen & Lene Schoesler (eds.), 353–391. Amsterdam: John Benjamins. Lechner, Winfried. 2004. Ellipsis in Comparatives. Berlin: Mouton de Gruyter. May, Robert. 1977. The Grammar of Quantification. PhD dissertation, MIT. Merchant, Jason. 2003. Subject-auxiliary inversion in comparatives and PF output constraints. In The Interfaces: Deriving and Interpreting Omitted Structures [Linguistik Aktuell-Linguistics Today 61], Kerstin Schwabe & Susanne Winkler, 55–77. Amsterdam: John Benjamins. Mitchell, Bruce. 1985. Old English Syntax. Oxford: The Clarendon Press. Pesetsky, David & Torrego, Esther. 2001. T-to-C movement: Causes and consequences. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.), 355–426. Cambridge MA: The MIT Press. Rissanen, Matti. 1999 On the adverbialization of rather: Surfing for historical data. Language and computers 26: 49–59. Roberts, Ian 2007. Diachronic Syntax. Oxford: OUP. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: CUP. Stechow, Arnim von. 2006. Times as degrees. Ms, Universität Tübingen. Stern, Gustaf. 1931. Meaning and Change of Meaning. With special Reference to the English Language. Göteborg: Elander. Taylor, Ann, Warner, Anthony, Pintzuk, Susan & Beths, Frank. 2003. The York-Toronto-Helsinki Parsed Corpus of Old English Prose. Traugott, Elizabeth & Dasher, Richard. 2001. Regularity in Semantic Change. Cambridge: CUP. Traugott, Elizabeth & König, Ekkehard. 1991. The semantics-pragmatics of grammaticalization revisited. In Approaches to Grammaticalization [Typological Studies in Language 19], Elizabeth Traugott & Bernd Heine (eds.), 189–218. Amsterdam: John Benjamins. Villalta, Elisabeth. 2006. Context Dependence in the Interpretation of Questions and Subjunctives. PhD dissertation, Universität Tübingen.

chapter 11

Cycles of complementation in the Mayan languages*1 Clifton Pye

The University of Kansas This paper describes the cycles of complementation evidenced in many Mayan languages. The cycle introduces lexical items in a root clause to mark aspectual distinctions. Over time these aspectual markers become grammaticalized and eventually disappear, at which time new lexical items may be introduced to renew the cycle. This cycle has significant implications for our understanding of the Mayan verbal complex and the boundary between syntax and the lexicon. Comparative data from the Mayan language family illustrates the variability in time and direction of the Mayan complementation cycle.

Introduction In this paper I present a cycle of complementation that is evidenced in many Mayan languages. The Mayans are better known for their cycles of time rather than cycles of complementation, nevertheless there is sufficient comparative and historical evidence to support a cyclic introduction of clausal aspectual markers in Mayan languages. This cycle has significant implications for our understanding of the Mayan verbal complex and the boundary between syntax and the lexicon. Comparative data from the Mayan language family illustrates the variability in time and direction of the Mayan complementation cycle. Mayan languages have a verbal complex which contains inflectional elements for aspect, agreement and status. Agreement generally follows an ergative paradigm in which subjects of transitive verbs are cross-referenced by ergative prefixes while subjects * This research would not be possible without the generous support of the K’iche’ community and the assistance of Augustin Huix Huix and Pedro Quixtan Poz. I would also like to thank my colleagues on the Comparative Mayan Acquisition Project—Barbara Pfeiller and Pedro Mateo Pedro as well as the assistants in the Mayan language communities who keep the project running. I take responsibility for any errors or misinterpretations. This research is funded in part by grants from the National Science Foundation (BCS-0613120 and BCS-0515120) as well as a grant from the General Research Fund of the University of Kansas.

 Clifton Pye

of intransitive verbs and objects of transitive verbs are cross-referenced by absolutive clitics. Mayan languages use prefixes to distinguish between incompletive, completive, progressive and potential aspects. The aspectual distinctions coordinate with status suffixes, which also mark modality and transitivity (Kaufman 1990). I provide a generalized inflectional template for intransitive and transitive verbs in (1). The examples in (2) show how the verbal complex is realized in the Mayan language K’iche’1. (1) Inflectional Templates for the Mayan Verbal Complex a. Intransitive Template Aspect=Absolutive Stem-StatusIV b. Transitive Template Aspect=Absolutive Ergative-Stem-StatusTV (2) The Verbal Complex in K’iche’ Maya a. Intransitive k=at b’in-ik inc=B2 travel-indIV ‘You travel.’ b.

Transitive k=at inw-il-oh inc=B2 A1-see-indtv ‘I see you.’

1. K’iche’ is the official spelling adopted in Guatemala during the 1980s. The language name was generally spelled Quiché before this change. I use the new spellings for the names of the Mayan languages spoken in Guatemala. All Mayan words are shown in the practical orthography developed by the Proyecto Lingüístico Francisco Marroquín (Kaufman 1976) with a single exception: I use rather than for the glottal stop. The other orthographic symbols have their standard IPA values except: = /ts/, = /t∫/, = /'/, = /ts’/, =/t∫’/, <x> = /∫/, <j> = /x/, = /G/. I use the following abbreviations throughout the article: 1 2 3 4 6 A AP B CMP ENC DEP DIR DUB IND

first person singular second person singular third person singular first person plural third person plural ergative cross-reference antipassive suffix absolutive cross-reference completive aspect person enclitic dependent status directional dubitive marker indicative status

INC INCEP NOM PAS PREP PROG REC REM RN SUBJ TERM VI VT

incompletive aspect inceptive aspect nominalizing suffix passive suffix preposition progressive verb recent past remote past relational noun subjunctive marker terminative aspect intransitive verb transitive verb



Chapter 11.  Cycles of complementation in the Mayan languages 

In this paper I use evidence from a number of Mayan languages to argue that the verbal complex masks a structure of complementation. A complex clausal analysis accounts for the structure of the verbal complex better than a monoclausal analysis. Examples of the verbal complex like those in (2) appear as simple inflected verbs with a template like that in (1). I make the argument in this paper that a template which assumes a structure of complementation like those in (3) provides a better understanding of the synchronic and diachronic properties of the Mayan Verbal Complex. (3) Complement Structure of the Mayan Verbal Complex a. Intransitive Matrix Complement Aspect_Element=Absolutive [Stem-StatusIV] b. Transitive Matrix Complement Aspect_Element=Absolutive [Ergative-Stem-StatusTV] This complementation structure is the result of a historical cycle of aspectual reduction and renewal. The cycle of complementation begins when a lexical element is introduced to mark an aspectual distinction that is not marked overtly. The lexical item may be a verb or adverb. Both of these items select a complement clause that contains the main verb stem. The second stage of the cycle is marked by the gradual grammaticalization of the aspectual element. The meaning of the aspectual element becomes broader while its phonetic structure is reduced. Aspectual verbs lose their own aspectual and cross-referencing morphology, and gradually appear as bare roots. Eventually, in the third stage, the aspectual prefix disappears completely at which point the language contains an aspectual gap that may be filled by returning to Stage One with the introduction of a new aspectual element and the beginning of a new cycle (4). (4) Mayan Cycle of Complementation matrix Stage One Lexical Aspect Stage Two Grammatical Aspect Stage Three Zero Aspect

complement [Verb Complement] [Stative Argument] [Aspectless Predicate]

The cycle of complementation focuses on changes in the matrix element, but these changes have significant implications for the form of the complement. Complementation is a context of nominalization in Mayan languages, which can induce a mixed ergative cross-reference system on intransitive verbs and antipassive marking on transitive verbs. In the remainder of this paper I will explore the evidence which supports the cycle of complementation in Mayan languages. I will provide examples from various Mayan languages that illustrate each stage in the cycle. I then discuss historical evidence for the cycle within Yucatec.

 Clifton Pye



Proto-Mayan wo Huastecan Late Proto-Mayan wo Yucatecan Central Mayan wo Western Mayan Eastern Mayan 3 3 Greater Greater Greater Greater Tzeltalan Q’anjob’alan Mamean K’iche’an 3 Ch’olan Tzeltalan

Figure 1.  Genetic classification of Mayan languages (Kaufman 1976, 1990)

1. The Mayan language family The Mayan data used in this study are the result of more than three decades of modern linguistic description by native Mayan and non-native linguists alike. This outstanding database provides a fantastic resource for reconstructing the history of the Mayan languages. The Mayan language family contains some 30 separate languages with over seven million living speakers. The languages fall into four main historical subdivisions (Figure 1). Mayan languages have a largely agglutinative morphology with an ergative system of verb cross-referencing (Kaufman 1990). The ergative inflections typically crossreference the subjects of transitive verbs and nominal possessors. The absolutive inflections cross-reference subjects of intransitive verbs, direct objects of transitive verbs and subjects of non-verbal predicates. There are prevocalic and preconsonantal allomorphs of the ergative markers. Nominal arguments for subject, direct object and possessors are only used for emphasis or to disambiguate the reference of the pronominal cross-reference markers on verbs and nouns. Verbal utterances usually contain obligatory particles for aspect that coordinate with verbal status suffixes. The languages generally have a verb-initial underlying word order. Some languages have a verb, subject, object word order while others have a verb, object, subject word order. The underlying order varies with changes in definiteness and animacy (England 1994).

2. Stage one – lexical aspect In this section of the paper I provide examples of constructions that use lexical elements to mark aspectual distinctions. These elements range between fully inflected verbs, a special set of movement verbs, affect verbs that are not regularly inflected, and



Chapter 11.  Cycles of complementation in the Mayan languages 

adjectives and adverbs. These elements fall into two aspectual classes: stative and nonstative. Verbs inflected for aspect are non-stative predicates, while the other lexical classes are stative predicates. The stative predicates create an interesting ambiguity in that their complement could be considered either an argument or a complement. English uses an expletive pronoun and overt copula to distinguish between these two types of constructions (5). Mayan languages lack both expletive pronouns and copulas, so the equivalent structure is ambiguous (6). (5) English Stative Predicate Constructions a. Verb Argument Construction Travelling is hard. b. Verb Adjunct Construction It is hard to travel. (6)

K’iche’ Stative Predicate Construction k’ax k=0 b’in-ik hard inc=B3 travel-indiv ‘it is hard for one who travels.’

The complement verb b’in in (6) is inflected for aspect, agreement and status so neither of the English expressions in (5) provide an exact translation. Indicative status co-occurs with incompletive and completive aspect in K’iche’. The ambiguity between argument and complement has important consequences for the cycle of complementation that I described in the introduction. As the aspectual verbs become grammaticalized, they lose their own aspectual inflections and become stative predicates. The change to a stative predicate then brings a change from verb complement to stative argument, while promoting a change in the aspect marker from stative predicate to aspectual prefix. The change from non-stative to stative aspect clause provides an essential link between the first two stages of the cycle.

2.1

Verbal aspectual constructions

Mayan languages have two distinct classes of lexical verb constructions: aspectual verbs and movement verbs. Both of these verb classes are inflected for aspect and agreement, but they differ in the types of complements they select and argument expression. The movement verbs have been analyzed in the past as a type of verb incorporation, but a complement analysis provides a better account of their properties. I provide examples of these constructions in the next two sections of the paper. 2.1.1. Aspectual verbs Inceptive and Terminative aspects are commonly expressed by inflected matrix verbs in Mayan languages and so provide canonical examples of a full complementation

 Clifton Pye

structure that is used to mark aspectual contrasts. I provide examples of inceptive constructions in (7) and terminative constructions in (8). (7) Inceptive Constructions a. K’iche’ x=0 u-maji-j u-k’ayi-x-iik cmp=B3 A3-begin A3-sell-pas-nomiv ‘S/he began its selling.’

(Kaufman 1990: 85)

b.

Mam (England 1983: 300) n=chi ku7 teen xjaal belaara-l t-e jun weech inc=B6 dir be person watch-nomiv A3-rn one fox ‘The people began to watch the fox.’

c.

Tojolab’al och=on way-el begin=B1 sleep-nomiv ‘I began to sleep.’

d.

Tzeltal ya x-jajch-on ta way-el inc-begin-B1 prep sleep-nomiv ‘I begin to sleep.’

e.

Ch’ol mi j-kejel tyi wäy-el inc A1-begin prep sleep-nomiv ‘I begin to sleep.’

f.

Colonial Yucatec (1784; from Bricker 1981) k-u-hop’-ol k-meyah minan ven-el t-on inc-A3-begin-nomiv A4-work there.isn’t sleep-nomiv to-us ‘We begin working without having slept.’

(Kaufman 1990: 87 from Robertson 1980)

(8) Terminative Constructions a. K’iche’ x=0 u-tanab’a’ wa’-iim cmp=B3 A3-finish eat-nomiv ‘S/he finished eating.’ b.

Tzeltal ya j-lajin-0 s-jal-el inc A1-finish-B3 A3-weave-nomiv ‘I finish weaving.’

(Gilles Pollian, pc)

(Kaufman 1990: 85)

(Sántiz & Polian 2007)



Chapter 11.  Cycles of complementation in the Mayan languages 

c.

Tojolabal 0-ch’ak j-tsil-0 cmp-finish A1-tortilla.making-B3 ‘I finished making tortillas.’

(del Prado & Curiel 2007)

The examples in (7) and (8) have in common a matrix clause which contains a fully inflected verb that in turn selects a nominalized verb complement. Progressive constructions in Mayan languages illustrate the transition between Stage 1 and Stage 2 of the complementation cycle. Languages like K’iche’ preserve a fully inflected progressive verb that selects a complement. The progressive verbs in Poqomchi’ and Q’eqchi’ are inflected for person, but not aspect although the progressive in these two languages maintains an indicative status suffix. I provide examples of inflected progressive verb constructions in (9). (9) Inflected Progressive Verb Constructions a. K’iche’ k=at tajin k=0 a-tz’iib’a-aj inc=B2 prog inc=B3 A2-write-inddtv ‘You are writing it.’ b.

Western Poqomchi’ k’ahchi’-k=iin chi wir-ik prog-indiv=B1 prep sleep-nomiv ‘I am sleeping.’

c.

Q’eqchi’ yoh-k=in chi aa-sak’-b’-al prog-indiv=B1 prep A2-hit-PAS-nomiv ‘I am hitting you.’

(Kaufman 1990: 85)

(Kaufman 1990: 89)

(Dayley 1981: 21)

2.1.2. Movement verbs Movement verb constructions differ from the lexical verb constructions in their semantic field and the type of complement they select. As in many other languages, the movement verb ‘go’ in Mayan languages is often understood as denoting movement in time as well as in space although these constructions are more spatially oriented than their equivalents in English or Spanish. There are also significant differences in the optionality of the movement constructions across the Mayan languages. Movement constructions are purely optional in K’iche’ where their use is less frequent. They are more frequent in Q’anjob’al and their use is obligatory with transitive verbs in Mam (England 1994). A template for movement verb constructions is provided in (10). (10) Movement Verb Template Matrix Complement Aspect Move=Absolutive [Verb-depstatus]

 Clifton Pye

Movement verb constructions are distinguished from other complement constructions by the use of a verb of movement in the matrix clause. In K’iche’, Q’anjob’al and Yucatec, the movement verb selects a verb complement with a dependent status. The absolutive clitic appears with the movement verb in K’iche’ and Q’anjob’al, while it alters between the matrix clause and complement clause in Tzeltal and Yucatec. I provide examples of movement verb constructions in (11–15). I class the Mam terminative construction in (12b) under this heading since the Mam terminative verb behaves as a movement verb rather than a matrix verb. (11) K’iche’ Movement Verb Constructions a. x-uj-e: war-oq cmp-B4-go sleep-depiv ‘We went to sleep.’ b. x-0-e: qa-k’am-a’ cmp-B3-go A4-carry-depiv ‘We went to carry it.’ (12) Mam Movement Verb Constructions a. ma 0-tzaj t-tzyu-7n Cheep ch’it rec B3-go A3-grab-ap José bird ‘José grabbed the bird.’

(England 1983: 212)

b. ma chin b’aj aq’naa-n-a rec B1 finish work-ap-enc1 ‘I finished working.’

(England 1983: 303)

c. ma chin-x aaj-a b’eeta-l rec B1-go return-enc1 walk-nom ‘I went to walk.’

(England 1983: 299)

(13)

Q’anjob’al Movement Verb Construction ch=in-‘ul b’ey-oq inc=B1-come walk-depiv ‘I come to walk.’

(14) Tzeltal Movement Verb Constructions a. ya x-tal-on ta way-el inc-come-B1 prep sleep-nomiv ‘I come to sleep’ b. ya x-tal way-uk-on inc-come sleep-depiv-B1 ‘I come to sleep.’

(Pedro Mateo Pedro, pc)

(Gilles Pollian, pc)



Chapter 11.  Cycles of complementation in the Mayan languages 

(15)

2.2

Yucatec Movement Verb Construction bin lúub-uk-eč go fall-depiv-B2 ‘you are going to fall.’

(Barbara Pfeiler, pc)

Stative aspectual constructions

I analyze any predicate without overt aspect marking as stative. This analysis includes adjectives, adverbs, affect verbs, negation and perfect verb forms within the class of stative predicates. Such predicates frequently replace the overt aspect markers in the Mayan languages. Bricker (1981: 95) shows that adverbs sometimes replace the aspectual prefixes in Yucatec. A similar replacement occurs in Mam and Q’anjob’al. I provide examples of stative constructions which take verb complements in (16). (16) Adverbial Predicates a. K’iche’ k’ax k=at in-ch’ab’e-j difficult inc=B2 A1-speak.to-dtv ‘It’s hard for me to speak to you.’

(Kaufman 1990: 86)

b.

Mam jotx t-aaj-tz meeb’a boom! A3-return-dir orphan ‘Boom! the orphan returned.’

(England 1983: 233)

c.

Mam (Ana López Ramírez 2008, p.c.) xwit’l-al t-xi’-a jump-nom A2/3-go-enc1/2 ‘you go along jumping.’ (lit. ‘jumping is how you go’)

d.

Yucatec (Bricker 1981: 95) sáamal u-k’uč-ul tomorrow A3-arrive-nom ‘S/he arrives tomorrow.’ (lit. His/her arrival is tomorrow)

e.

Q’anjob’al (Mateo Pedro 2008, p.c.) amanq’an im-b’ey-i quickly A1-walk-indiv ‘I am walking quickly (lit. quick is my walking).’

It is interesting to observe that these adverbial contexts are also contexts of split ergativity/nominalization in Mam and Q’anjob’al. Their use as aspectual replacements could explain how the nominalized complements were extended to incompletive contexts in the Ch’olan and Yucatecan branches. Adverbs provide another source that Mayan languages can draw upon to mark aspectual contrasts.

 Clifton Pye

The literal translations of these sentences provide a better idea of their syntactic structure. The adverb serves as a stative predicate which takes a nominalized predicate as its subject. Q’anjob’al displays an interesting split between the ergative cross-reference marker and the indicative status suffix in these contexts. Q’anjob’al has extended its indicative suffix to nominalizing contexts. Bohnemeyer (1998: 338) makes a distinction between the bound and unbound aspect markers in Yucatec, and argues that all of the unbound markers (such as the inceptive and terminative markers discussed above) are stative predicates. He analyzed the completive prefix t- and the incompletive prefix k- in Yucatec as bound prefixes. Furbee-Losee (1976: 204) explicitly treats the aspect markers in Tojolab’al as stative verbs which take verbal complements. As evidence, she cites the use of oh, the future progressive tense, in affirmative replies to a question in the future tense. The reduplicated form oh-oh is a possible response to the question oh xa wahan (oh xa wah-an, FUT now go-FUT, “Are you going now?”). Bohnemeyer shows that the unbound aspectual markers in Yucatec can be used in the same way.

3. Stage two – grammatical aspect Once an aspectual distinction is marked, it becomes subject to grammaticalization. The indicative aspectual proclitics in K’iche’, Q’anjob’al and Ch’ol provide examples of this stage of the complementation cycle. (17) K’iche’ Indicative Proclitics a. Completive x=at inw-il-oh cmp-B2 A1-see-indtv ‘I saw you.’ b.

Incompletive k=at inw-il-oh inc=B2 A1-see-indtv ‘I see you.’

(18) Q’anjob’al Indicative Proclitics a. Completive max=ach hin-kol-o’ cmp=B2 A1-help-indtv ‘I helped you.’ b.

Incompletive chi=ach hin-kol-o’ cmp=B2 A1-help-indtv ‘I help you.’



Chapter 11.  Cycles of complementation in the Mayan languages 

(19) Ch’ol Indicative Proclitics a. Completive tyi a-mos-oy-oñ cmp A2-cover-indtv-B1 ‘you covered me.’ b.

(Vázquez Alvarez 2002: 162)

Incompletive mi a-mos-oñ inc A2-cover-B1 ‘you cover me.’

The K’iche’ dependent forms in (6b) demonstrate the attachment of the aspect prefixes to the verb is incomplete in that K’iche’ allows a small set of movement verbs (e: ‘go’, ul ‘arrive’ and opan ‘pass’) to appear between the aspect marker and the verb stem. Ch’ol expands the set of particles that intervene between the aspect marker and the verb stem. The examples in (20) illustrate the use of Ch’ol adverbial and modal particles. These examples provide evidence of the weak association between the aspect markers and the verb stem. The weakness is due to the origin of the aspect markers as a separate clause. (20) Ch’ol Adverbial and Modal Particles a. Already chonkol=ix k-wäy-el prog=already A1-sleep-nomiv ‘I am already sleeping.’

(Vázquez Alvarez 2002: 151)

b.

Dubitive muk’=ka a-wäy-el inc=dub A2-sleep-nomiv ‘Are you asleep?’

(Vázquez Alvarez 2002: 164)

c.

Subjunctive muk’=ik a-wäy-el inc=subj A2-sleep-nomiv ‘if you would sleep.’

(Vázquez Alvarez 2002: 162)

The progressive construction illustrates the incipient grammaticalization of aspect in a number of Mayan languages. The Mayan progressive construction originates as an independent verb which takes a complement clause as shown in the previous section. Overtime, the progressive verb loses its inflections and even part of its root. In some Mayan languages, the progressive may now be marked by a single consonant. The following examples show progressive constructions in different stages of reduction. (21)

K’iche’ Progressive tajin k=e: pet-ik prog inc=B6 come-indiv ‘they are coming.’

 Clifton Pye

(22) Ch’ol Progressive choñ a-wäy-el prog A2-sleep-nomiv ‘you are sleeping.’ (23)

Yucatec Progressive t in hóok’-ol prog A1 go_out-nomiv ‘I am going out.’

(Pye, field notes)

(Bricker et al. 1998: 400)

In this section I provided examples of grammaticalized aspect markers in a variety of Mayan languages. In these examples, aspect has the appearance of a verbal prefix, but clues to the clausal origin of aspect marking can still be found. Several languages allow movement verbs or adverbs to appear between the aspect marker and the verb stem. The progressive construction provides further evidence of the grammaticalization process in that the progressive originates as an inflected verb and is gradually reduced to a single syllable or consonant. The variety of progressive markers across the Mayan languages is the result of the cycle of complementation.

4. Stage three – null aspect The last stage of the Mayan complementation cycle is reached when an overt aspect marker is no longer required in certain contexts. The preceding sections provide evidence from various Mayan languages of the weak association between the aspect markers and the verb stem. The weakness is due to the origin of the aspect markers as a separate clause. Overtime the aspect clause loses its specific semantic contribution and becomes a more general marker of aspect. Since the status marker also expresses similar aspectual contrasts, the aspectual prefix can become redundant. In this case, the aspect marker is no longer required and some Mayan languages no longer require overt aspect prefixes. The incipient loss of aspect marking is seen in Mayan languages that display partial gaps in their aspectual paradigms. Dayley (1985: 80) notes that the completive aspect marker may optionally be omitted in Tzutujil before a consonant (24). (24)

Tzutujil Optional Aspect Omission (x)=0 war-i (cmp)=B3 sleep-indiv ‘S/he slept.’

(Dayley 1985: 80)

Malchic et al. (2000: 64) state that the second singular and plural incompletive prefixes are zero in Poqomchi’ and Poqomam. Caz Cho (2007: 63) states that a few dialects of Q’eqchi’ preserve the use of an oh- prefix that becomes zero in second singular transitive forms and first and second singular intransitive forms in remote past contexts (25).



Chapter 11.  Cycles of complementation in the Mayan languages 

(25)

Q’eqchi Intransitive Remote 0-in-war rem-B1-sleep ‘I slept (some time ago).’

(Caz Cho 2007: 64)

Tzeltal has a defective aspectual paradigm in that intransitive verbs do not have an overt completive marker and transitive verbs lack an overt incompletive marker (26). (26) Zero Completive Marking in Tzeltal a. Tzeltal Completive Intransitive Verbs 0-muy-on cmp-descend-B1 ’I descended.’ b.

Tzeltal Incompletive Transitive Verbs 0-k-il inc-A1-see ‘I see it.’

(Robertson 1992: 194)

(Robertson 1992: 194)

The Cholan languages demonstrate the complete loss of overt aspect markers. Chontal innovated a set of aspectual contrasts that rely on the status suffixes. Incompletive transitive verbs in Chontal have the suffixes –e’ or –Vn, while completive transitive verbs with third person objects have the suffix –i (27). (27) Chontal Aspect Loss on Transitive Verbs a. Incompletive u pul-e’-0 A3 burn-indtv-B3 ‘s/he burns it.’ b.

Completive u k’ux-i-0 A3 eat-indtv-B3 ‘s/he ate it.’

(Keller & Plácido 1997: 447)

(Keller & Plácido 1997: 438)

Intransitive verbs in Chontal indicate the aspectual contrast through the use of split ergative marking (28). Completive intransitive verbs only carry an absolutive person marker while incompletive intransitive verbs have an ergative prefix as well as a status suffix (-e, -o, -a, -an or -in). (28) Chontal Aspect Loss on Intransitive Verbs a. Incompletive u jom-e A3 climb-indtv ‘s/he climbs.’

(Keller & Plácido 1997: 458)

 Clifton Pye

b.

Completive wäy-on sleep-B1 ‘I slept.’

(Keller & Plácido 1997: 459)

Chontal shows one way in which Mayan languages may communicate without the use of overt aspect marking. The Chorti’ language provides a related, but distinct example of the loss of overt aspect marking. Robertson (1992: 174) states that transitive verbs in Chorti’ have lost all distinctions between completive and incompletive aspect. The intransitive verbs rely exclusively on the system of split ergative marking to mark the completive/incompletive contrast (29). (29) Chorti’ a. Transitive completive and incompletive in-xur-i-0 A1-cut-ind-B3 ‘I cut it.’ (past, present and future) b.

Intransitive incompletive in-wayan A1-sleep ‘I sleep.’

c.

Intransitive completive wayan-en sleep-B1 ‘I slept.’

(Robertson 1992: 175)

Huastec is another Mayan language that has lost its aspect marking. Robertson (1992) provides the examples of incompletive verbs (30). (30) Huastec a. Intransitive incompletive in-way-el A1-sleep-nom ‘I sleep.’ b.

(Robertson 1992: 213–214)

Transitive incompletive u-k’ap-al A1-eat-nom ‘I eat it.’

Since the Cholan, Yukatekan, Tzeltalan and possibly the Huastecan languages have been in contact over an extensive period, there is the possibility that the loss of aspect marking in these languages spread from one of these languages to the others. We have seen that aspect marking can be lost in the other Mayan languages. In this regard, it is



Chapter 11.  Cycles of complementation in the Mayan languages 

interesting to find that Hofling reconstructs Proto-Yukatekan without overt aspect marking (2006). He provides the transitive reconstructions shown in (31) and the intransitive reconstructions shown in (32). (31) Proto-Yukatekan Transitive Verbs a. Transitive incompletive 0 in-mach-ik-ech inc A1-grab-nomtv-B2 ‘I grab you.’ b.

Transitive completive 0 in-mach-aj-ech cmp A1-grab-indtv-B2 ‘I grabbed you.’

(32) Proto-Yukatekan Intransitive Verbs a. Intransitive incompletive 0 in-wen-el inc A1-sleep-nomiv ‘I sleep.’ b.

(Hofling 2006)

(Hofling 2006)

Intransitive completive 0 wen-ih-ech cmp sleep-indiv-B2 ‘you slept.’

In sum, many Mayan languages provide evidence for the partial or complete loss of aspect marking prefixes. These constructions illustrate the final stage of the cycle of complementation. One consequence of aspect loss is the promotion of the verb from a complement clause to a matrix clause. This promotion requires a radical reanalysis of the verb’s inflectional morphology. The ergative possessive markers on nominalized intransitive verbs become markers of person and aspect on the promoted intransitive verb. However, there is reason to think that the verbs remain in a complement clause even when aspect marking has been lost since this makes possible the reintroduction of a new matrix clause marking aspect and a renewal of the cycle of complementation.

5. The cycle of complementation – Yucatec My presentation to this point has relied upon evidence from different Mayan languages to illustrate the stages of aspect marking. Yucatec provides historical evidence for all three stages of the complementation cycle.

 Clifton Pye

Stage Three – Zero Marking As noted in the previous section, Hofling reconstructs Proto-Yukatekan without overt aspect markers on the basis that Mopan currently lacks such markers. (33) Proto-Yukatekan Incompletive Verbs a. Transitive 0 in-mach-ik-ech inc A1-grab-nomtv-B2 ‘I grab you.’ b.

(Hofling 2006)

Intransitive 0 in-wen-el inc A1-sleep-nomiv ‘I sleep.’

Stage One – Lexical Aspect Bricker (1981: 86) mentions that McQuown (1967: 243) thought that the source for the incompletive prefix k- in modern Yucatec (1b) was the verb lik ‘accustomed to’ found in Colonial Yucatec. Robertson (1992: 204) suggests the verb kah ‘to do, make’ was its source. The modern Ch’olan languages Ch’ol and Chontal currently make obligatory use of their ‘make, do’ light verb in complex expressions with a class of verbal nouns. Chontal further extends this construction to many transitive verb complements as a type of antipassive construction. The Ch’olan languages provide support for Robertson’s suggestion, but Yucatec would have had to take a further step of limiting the light verb’s use to incompletive contexts. (34) Colonial Yucatec a. Transitive kah in-mach-ik-ech do A1-grab-nomtv-B2 ‘I grab you.’ b.

(McQuown 1967; Robertson 1992)

Intransitive kah in-wen-el do A1-sleep-nomiv ‘I sleep.’

Stage Two – Grammaticalized Aspect The incompletive markers have become greatly reduced in Modern Yucatec, to the point where some are beginning to disappear. Modern Yucatec uses k- to mark incompletive aspect. The nominalizing suffixes on incompletive verbs in Yucatec betrays their origin in nominalized complement clauses.



Chapter 11.  Cycles of complementation in the Mayan languages 

(35) Modern Yucatec a. Transitive k in-mach-ik-ech inc A1-grab-nomtv-B2 ‘I grab you.’ b.

(Bricker 1981)

Intransitive k u wen-el inc A3 sleep-nomiv ‘S/he sleeps.’

Stage Three – Zero Marking Modern Yucatec uses the prefix h- as a completive marker on intransitive verbs. The h- is lightly aspirated and disappears in colloquial speech (36a). The Yucatec progressive marker is currently used as a full verb root táan, a prefix t- and is sometimes omitted (36b). Completive intransitive verbs and progressive verbs provide potential contexts for a new cycle of complementation in Yucatec. (36) Modern Yucatec a. Completive Intransitive Verbs (h)-wèen-en cmp-sleep-B1 ‘I slept.’ b.

Progressive (t) in hóok’-ol prog A1 go_out-nomiv ‘I am going out.’

(after Bricker et al. 1998: 400)

6. Conclusion Complementation induces a suite of morphological changes on both intransitive and transitive verbs in Mayan languages. These changes result in the use of ergative subject marking on intransitive complements (mixed ergativity), and the intransitivization of transitive complements (crazy antipassives). Mayanists usually discuss these phenomena separately without attempting a unified account of these changes (c.f. Larsen & Norman 1979), although Kaufman (1990) provides an integrated discussion. Fortunately, complementation is currently the focus of many studies in Mayan linguistics so much information is available for a wide range of Mayan languages (cf. Aissen 2007). The recognition of a cycle of complementation provides a new perspective on complementation in the Mayan languages and integrates the analysis of complementation with the agglutinative structure of the Mayan verbal complex.

 Clifton Pye

The Mayan cycle of complementation bears a certain similarity to the periphrastic origin of inflectional tense-aspect forms in the Romance languages (Hopper and Traugott 2003: 8–9). The Mayan languages provide several examples of this change in progress where different forms of the construction are still in competition. Evidence of the complementation origin of the aspect markers can be found in Mayan languages with advanced grammaticalization of aspect. The Mayan verbal complex thus stands halfway between a complex clause and a simple inflected verb. It demonstrates how far the grammaticalization process may extend without becoming fully actualized. The lack of actualization makes it possible in turn to renew the cycle of complementation. The Mayan verbal complex is unusual in that there is evidence that the verbal complex results from the grammaticalization of complement structures. The Cholan and Yucatecan languages preserve evidence of this process in their incompletive verb forms which have nominalized complements. The dependent construction may have originally derived from a complement to verbs of movement. It was later extended to irrealis contexts such as negation in Q’anjob’al and the subjunctive in Yucatec. The process of complementation results in distinct verb suffixes which register the different contexts of complementation. Adverb insertion and changes in absolutive placement hint at the origin of the indicative form as a verbal complement. The indicative forms are now grammaticalized as verbs rather than verbal complements. I made use of comparative data from different Mayan languages as well as historical data from Yucatec to argue that a complementation cycle is at work across the Mayan language family. Once the cycle has been described we can compare the rates at which the languages move through the cycle. Huastec, Yucatec and the Cholan languages have progressed relatively quickly to the point where some of the languages have lost overt aspect marking. The K’iche’an branch of the languages have preserved the aspect markers to a greater extent, but even in these languages aspect marking shows considerable flux (Robertson 1992). The comparison across the Mayan languages also reveals different pathways of complementation. The Yucatecan and Cholan languages have pursued a path of nominalization while Mam followed a path of dependent complements to movement verbs. K’iche’ has adhered closely to the indicative path. The cycle of complementation in Mayan languages has affected aspect marking disproportionately to the other parts of the Mayan verbal complex. The status and cross-referencing affixes have changed much less than the aspect markers. Complementation has also played a crucial role in creating and maintaining the characteristic agglutinative morphology in the Mayan verbal complex. The complementation structure of the verbal complex suggests that Mayan languages are unusual in their preference for complex structures over single clauses. Further evidence of this bias can be found in the light verb constructions of the Cholan languages, the serial verb constructions of the Tzeltalan languages and the nearly obligatory use of movement verbs in the Mamean languages. The complementation preference also has radical implications for the acquisition of Mayan languages. Children have to acquire Mayan languages from complex sentences only. Mayan children cannot



Chapter 11.  Cycles of complementation in the Mayan languages 

use the status suffixes correctly without understanding their distinctive contexts of complementation.

References Aissen, Judith. 2007. Handouts of the Workshop on Complementation. Taller sobre complementación en lenguas mesoamericanas. Antigua Guatemala, Guatemala: OKMA. Bohnemeyer, Jürgen. 1998. Time Relations in Discourse: Evidence from a Comparative Approach to Yukatek Maya. Wageningen: Ponsen & Looijen. Bricker, Victoria R. 1981. The source of the ergative split in Yucatec Maya. Journal of Mayan Linguistics 2(2): 83–127. Bricker, Victoria, Po’ot Yah, Eleuterio & Dzul de Po’ot, Ofelia. 1998. A Dictionary of The Maya Language As Spoken in Hocabá, Yucatán. Salt Lake City UT: University of Utah Press. Caz Cho, Sergio. 2007. Informe de Variación Dialectal en Q’eqchi’. Guatemala: Cholsamaj. Dayley, Jon P. 1981. Voice and ergativity in Mayan languages. Journal of Mayan Linguistics 2(2): 3–82. Dayley, Jon P. 1985. Tzutujil Grammar. Berkeley CA: University of California Press. England, Nora C. 1983. A Grammar of Mam, A Mayan Language. Austin TX: University of Texas Press. England, Nora C. 1994. Autonomia de los Idiomas Mayas: Historia e Identidad. Guatemala: Cholsamaj. Furbee-Losee, Louanna. 1976. The Correct Language: Tojolabal. A Grammar with Ethnographic Notes. New York NY: Garland. Hofling, Charles Andrew. 2006. A sketch of the history of the verbal complex in Yukatekan Mayan languages. International Journal of American Linguistics 72(3): 367–396. Hopper, Paul & Traugott, Elizabeth Closs. 2003. Grammaticalization. Cambridge: CUP. Kaufman, Terrence. 1976. Proyecto de Alfabetos y Ortografias para Escribir las Lenguas Mayances. Guatemala: Ministerio de Educacion. Kaufman, Terrence. 1990. Algunos rasgos estructurales de los idiomas Mayances con referencia especial al K’iche’. In Lecturas Sobre la Lingüistica Maya, Nora C. England & Stephen R. Elliott (eds), 59–114. Guatemala: CIRMA. Keller, Kathryn C. & Luciano G. Plácido. 1997. Diccionario Chontal de Tabasco. Tucson AZ: SIL. Larsen, Thomas W. & Norman, William M. 1979. Correlates of ergativity in Mayan Grammar. In Ergativity: Towards a Theory of Grammatical Relations, Frans Plank (ed), 347–370. New York NY: Academic Press. Malchic Nicolás, Bernardo, Manuel, Mó Isém, Romelia & Tul Rax, Augusto. 2000. Variación Dialectal en Poqom. Guatemala: Cholsamaj. Mateo Pedro, Pedro. 2008. Nominalization in Q’anjob’al. Ms, The University of Kansas. McQuown, Norman A. 1967. Classical Yucatec (Maya). In Handbook of Middle American Indians, Vol.5, Robert Wauchope & Norman A. McQuown (eds), 201–247. Austin TX: University of Texas Press. del Prado, Ramírez & Alejandro Curiel. 2007. Everything you wanted to know about complementation in Tojolabal and more! Taller sobre Complementación en Lenguas Mesoamericanas. Antigua Guatemala, Guatemala: OKMA. Robertson, John S. 1980. The Structure of Pronoun Incorporation in the Mayan Verbal Complex. New York NY: Garland Press.

 Clifton Pye Robertson, John S. 1992. The History of Tense/Aspect/Mood/Voice in the Mayan Verbal Complex. Austin TX: University of Texas Press. Sántiz, Roberto & Gilles Polian. 2007. Complementación en tseltal. Taller sobre complementación en lenguas mesoamericanas. Antigua Guatemala, Guatemala: OKMA. Vázquez Alvarez, Juan Jesús. 2002. Morfología del verbo de la lengua chol de Tila, Chiapas. MA thesis. Centro de Investigaciones y Estudios Superiores en Antropología Social, Instituto Nacional Indigenista.

chapter 12

The preposition cycle in English Cathleen Waters

University of Toronto Historical facts on prepositions show a cycle in which semantic bleaching of spatial meaning is followed by the addition of a new prepositional element. In this paper, I outline a syntactic process to complement this semantic process by focusing on a group of English prepositions that have similar semantic and syntactic properties: above, behind, below, beyond, inside, outside, and in front. I show that, over time, a structure with two adjacent heads, Loc and AxPart, is reanalyzed as the higher Loc head followed by a null AxPart. This reanalysis sets the stage for the renewal of the cycle with the addition of either a new AxPart element, or, the subsequent reanalysis of nested prepositions as Loc and AxPart heads.

1. Introduction1 The etymologies of some English prepositions demonstrate a cycle in which spatial meaning is reinforced by the periodic addition of a new spatial element. To complement this semantic process, I suggest a syntactic cycle involving the internal structure of prepositional phrases. I focus on a small group of English prepositions that have been posited by Svenonius (to appear) to form a class, and thus to have some commonality of internal syntactic structure. This group is comprised of above, behind, below, beyond, inside, outside, and in front. Focusing on only this group allows for a manageable scope, though the analysis is likely to apply to other spatial prepositions in English and to spatial prepositions cross-linguistically. In this first section, I outline the structure of the paper and I present some historical data for the group of prepositions under consideration. In addition, I review 1. Many thanks to the participants of the Linguistic Cycles workshop, especially Elly van Gelderen and the anonymous reviewer, and to Olena Tsurska, Derek Denis, and Annick Morin for their feedback and ideas on this paper. Thanks also to Michela Ippolito, Elizabeth Cowper and Diane Massam for comments on an earlier, related paper. Citation abbreviations are as follows: (Corpus/Work) where the corpora are OED = Oxford English Dictionary, PG = Project Gutenberg, G = Google search; a list of works and authors cited is given in the Appendix.

 Cathleen Waters

generally observed tendencies of the grammaticalization process for prepositions. I outline, in a pre-theoretical manner, the cycle suggested by the historical data, namely, the periodic addition of new spatial elements to existing prepositions. In Section 2, I summarize recent syntactic work by Svenonius on the classification of English prepositions and the internal syntactic structure of prepositional phrases. Using these findings, I demonstrate that variation in the behavior of different prepositions of the group is a result of a difference in whether one of the heads (specifically, the AxPart head) is null. In Section 3, I use this syntactic framework to describe a syntactic process that operates in tandem with the cycle of semantic bleaching. The cycle consists of three stages. In stage one, neither head is null. In stage two, the two non-null heads are reanalyzed as a single, higher head, though there may be synchronic variation. In the final stage, the variation disappears, and only the structure with the null lower head is possible. This reanalysis to a structure with an obligatorily null head creates an environment where a new, non-null head can be subsequently added in the lower position, renewing the cycle. To illustrate how the renewal process progresses, I introduce additional historical and present-day data which shows that while there is more than one mechanism to trigger the renewal process, the ultimate result is the same, renewed structure. To begin, I turn to previously observed tendencies in the grammaticalization of prepositions. Adpositions, often classed as function words, are considered inherently more grammaticalized than some other types of words, such as nouns (Hopper and Traugott 2003: 4). As different prepositions show different degrees of grammaticalization, a further distinction can be made between “functional” prepositions (e.g., for) and “non-functional” prepositions (e.g., through) (Roberts and Roussou 2003: 227). Prepositions are considered to be completely grammaticalized when they have lost all spatial content (Roberts and Roussou 2003: 222). Prepositional elements themselves frequently arise from grammaticalization of other forms, usually along one of two paths, though both paths are possible in the same language (Heine, Claudi and Hünnemeyer 1991:  235–6). First, they may arise from verbs, generally through an intermediate stage as co-verbs, and then, with the loss of verbal marking such as agreement, prepositions (Heine, Claudi and Hünnemeyer 1991: 2). Alternatively, prepositions may arise from nominals, frequently nouns referring to body parts (Heine, Claudi and Hünnemeyer 1991: 125) or nouns that refer to a part of an object (Hopper and Traugott 2003: 110), moving through a stage of adverbial use and then to a spatial prepositional use.2 Once established as spatial prepositions, other prepositional uses, such as expressing temporal, causal, or benefactive relationships may arise, the progression of which has been observed to involve increasing abstractness (e.g., agentive uses involving human participants develop before causal uses with non-humans) (Heine, Claudi and Hünnemeyer 1991: 157). However, even after more abstract meanings come into being, spatial meanings can persist, though the 2. Of course, prepositional use is not the obligatory next stage after adverbial use for nominals, see van Gelderen (2004: 82) for an example with negation.



Chapter 12.  The preposition cycle in English 

spatial meaning may be reinforced, as seen in “by the railway station > down by the railway station” (Hopper and Traugott 2003: 110). Presumably, this reinforcement is necessary as the spatial content has weakened, i.e. semantic bleaching has taken place. This process of reinforcing the prepositional content can be seen clearly in the etymologies of the prepositions seen in Figure 1; a new spatial element be- (and later a-) or side is affixed to an existing directional/prepositional element. The reinforcement seen in Figure 1 is now an inalienable element of each preposition. However, synchronically, we also see the optional addition of a second preposition-like element with a full prepositional phrase: (1) a. up above our heads b. down in the valley c. over by the bus stop This juxtaposition of two prepositional elements is crucial to the reanalysis process, and I return to it in Section 3. Before discussing the cycle, however, I discuss some syntactic proposals for the internal structure of prepositional phrases headed by the prepositions under study here. OED Etymologies be + uf + an by + up + case

a + bufan on + above

be + hind + ana by + behind + from

above

behind

be + 3eond + ana beyond by + the farther side + from be + lowe by + low with + out with + in

below

without + side within + side

Figure 1.  Preposition etymologies from the OED

outside inside

 Cathleen Waters

2. The internal structure of phrases with projective prepositions Using syntactic and semantic behaviors, Svenonius proposes a four-group classification schema for a subset of English spatial prepositions: Projective, Bounded, Extended, and Particle. Table 1 lists the current population for each type (Svenonius, to appear: 2). While some further semantic differences exist between members of a single group, the differences across groups distinguish the classes from each other and are reflected in his proposed differences in syntactic structure; the facts relevant to my subsequent description of the cycle are discussed here. The most fundamental distinction made by Svenonius is a distinction between a default reading expressing a location versus a path (to appear: 1). For instance, Svenonius argues that the Projective class (the focus of this paper) is more naturally read as expressing locative meanings; however, they can optionally have directional meanings, as shown in (2): (2)

a. b. c. d. e. f. g.

The plane flew behind the trees. The rabbit jumped inside the cage. The rabbit jumped outside the cage. The submarine sailed below the ice. The marathoners ran beyond the city limits. The revelers danced in front of the palace. The mountaineers climbed above the dam. (based on Svenonius, to appear: 19)

In contrast, the most natural reading of an Extended class preposition is to express motion along a path, as shown in (3): (3)

a. b. c. d.

The plane flew around the trees. The rabbit jumped through the cage. The rabbit jumped past the cage. The submarine sailed under the ice.

Table 1.  Preposition types (from Svenonius) Projective

Bounded

Extended

Particle

behind in front of

among between

around through

up down

inside

next to

across

on

outside

beside

along

off

above

upon

over

in

below

near

under

out

beyond

against

past

away



Chapter 12.  The preposition cycle in English 

e. The marathoners ran along the river. f. The revelers danced across the palace. g. The mountaineers climbed over the dam. (based on Svenonius, to appear: 19–20) In addition, Projective prepositions combine awkwardly with each other and other locatives (in non-list readings), but are perfectly acceptable in combination with Particles higher in the structure (e.g., up), as seen here in (4):3 (4) a. ?the boat drifted beyond above the dam. b. the boat drifted up above the dam. (based on Svenonius, to appear: 12) Svenonius argues that, in cases such as (4b), the use of up serves to indicate that the boat is higher than some salient viewpoint, such as the speaker (to appear: 17). In other words, (4b) would be paraphrased by something like up there above the dam. In forms such as (4b), the Particle occurs higher in the structure than the phrase containing the Projective preposition.4 Furthermore, the co-occurrence of Particle and Projective prepositions is more likely when both elements have similar meanings (Svenonius, to appear: 18). These two facts are key to the process of reanalysis discussed later. In addition to proposals about preposition classes, Svenonius has also suggested a universal hierarchy within the prepositional phrase in which “the contentful material of spatial adpositions is distributed over a series of functional heads” (to appear: 19). The structure that Svenonius gives for in front of the house is reproduced in (5):

(5) in front of the house (Svenonius, to appear: 5)



PlaceP wo Place AxPartP in 3 AxPart KP front 3 K DP of 6 the house

3. In some cases, Projective prepositions are acceptable in combination with lower Particles, but I note that these do not have the same interpretation: e.g., John walked to the store. His dog trotted behind on the sidewalk. can mean behind John, but not *behind there). I suggest that cases such as these are actually two prepositional phrases, the first of which has a null complement: [behind Ø (= John)] [on the sidewalk]. 4. My description below of the reanalysis process for prepositions holds regardless of whether up in (4b) is considered to be an adverb or a preposition, so I will use the Svenonius model that up is a preposition in these cases. I leave aside the question of the nature of particles occurring in phrasal verbs.

 Cathleen Waters

However, Svenonius subsequently states that Place can be “decomposed into at least the components Deg – Loc – Ax(ial)Part – K” (Svenonius, to appear: 7). As he does not indicate that this decomposition should be considered nesting (i.e. it does not appear that he is suggesting the Place head has four subparts above another AxPart), his statement suggests that prepositional phrases with Projection prepositions have the internal structure seen in (6):

(6) Deg – Loc – Ax(ial)Part – K – DP

Svenonius describes the Deg layer as home to measure expressions such as one meter in one meter above the bridge (to appear: 7). However, in cases such as in front of the house, which lack a measure phrase, Deg is presumably null. Therefore, although Svenonius does not explicitly state this conclusion, I will assume that elements such as the in element of in front of the house are Loc heads. Thus, I assume that the structure of cases such as (5) have the fuller internal structure seen in (7):

(7) in front of the house (decomposed)



PlaceP wo Deg LocP Ø 3 Loc AxPartP in 3 AxPart KP front 3 K DP of 6 the house

The head called Ax(ial)Part is proposed in Svenonius (2006) to explain cross-linguistic data. AxPart is a nominal-like element indicating a space which is determined in relation to a part of an object. For example, the front element of in front is an AxPart. Other AxParts in English include top and side (Svenonius 2006: 49). However, Svenonius clearly distinguishes AxPart from N; for instance, the use of an adjective with an AxPart is ungrammatical, whereas the use of the same adjective with the corresponding nominal is acceptable: (8) a. There was a kangaroo in the smashed-up front of the car. b. *There was a kangaroo in smashed-up front of the car. (Svenonius 2006: 50) It is also clear from (8) that the nominal and AxPart uses of front do not refer to the same space (Svenonius 2006: 50). Although AxParts commonly evolve from nouns, it is possible for AxParts to arise from something besides a nominal. For example,



Chapter 12.  The preposition cycle in English 

beneath arose from be (by) + niðan (down), where the second element was a direction, rather than part of an object (Svenonius 2006: 73). As I discuss below, this flexibility in what can become an AxPart is important to the cycle of preposition renewal. Furthermore, Svenonius argues that the of element in a phrase such as in front of the house is a case marker, rather than a (second) preposition (to appear: 4).5 However, of is not present in all prepositional phrases with Projective prepositions: (9) a. above/beyond/behind/below (*)of the bridge. b. in front/inside/outside of the house. As K is realized as of where one (non-null) nominal element, the DP (the house), is in complement position to another nominal-like element, the AxPart head (front), I assume that the presence of the case marker of occurs only when the KP has not directly been assigned structural case (e.g., by the Loc head), but rather has received inherent case, similar to the destruction of the city (Chomsky 1986: 192, see also Anderson 2006: 119).6 Having established the internal structure of prepositional phrases with in front, I will now discuss the structure of prepositional phrases with other Projective prepositions: above, below, beyond, behind, inside, and outside. It is immediately apparent that they do not all behave identically; the first four have some distinct characteristics from the latter two. The first four never appear with of, whereas the latter two can (see (9) above). Furthermore, the prepositions in (9b) are often described as being compositional in nature (e.g., Hopper and Traugott 2003: 110), whereas those in (9a) are usually treated as single morphemes (e.g., Svenonius, to appear: 9). Svenonius describes the structure of behind as consisting of an AxPart with a null Loc (to appear: 9). However, this analysis is problematic. First, if AxParts alone could perform prepositional functions, we would expect forms like (10) whereas these forms arise with non-null Loc heads, as seen in (11): (10) a. Ø front of the house b. Ø top of the table (11) a. in front of the house b. on top of the table In addition, we would expect that non-null Loc heads could combine with the AxPart heads in a way that could be paraphrased in relation to the space identified by the AxPart. However, as shown in (12), this is not currently the case for behind (as I describe in a later section, the availability of this interpretation appears to change over 5. I maintain the K head below it in order to be consistent with the hierarchy, though nothing in my analysis hinges on the open question of whether K is necessary or not, i.e. the analysis presented here is also viable if a DP or NP can take case directly. 6. The “rule of of-insertion is a ‘default case’, applying only when there is no preposition available that inherently assigns the theta-role” (Chomsky 1986: 194). This also implies that the K head in this structure in English is always null. For a more detailed discussion, see Waters (2008).

 Cathleen Waters

time). In case (12b), similar to the situation described above for up above the dam, the presence of an element such as in before a phrase with a Projective preposition has a meaning along the lines of in some salient place, behind the desk: (12) a. in front of the house = in the space to the front of the house b. in behind the desk ≠ in the space to the back of the desk Finally, if the assumption is correct that of­-insertion only takes place when a non-null AxPart and KP are in a sister relationship, then we would expect cases like behind to occur with of, which is not the case, as seen in (9a). However, an analysis in which these prepositions are Loc heads followed by null AxParts is more convincing. First, it correctly predicts the lack of of-insertion in (9a) because the null AxPart would not trigger insertion. In addition, it eliminates the predictions of AxParts without Loc heads in (10), as the Loc head is occupied. The only thing that must be explained, however, is why it is not currently possible for above, below, beyond, and behind to combine with other AxParts, if their AxPart is null. As I demonstrate in the next section, this constraint does not hold diachronically. At any point in time, however, there may be semantic, rather than syntactic constraints on combinations. Specifically, the inherent spatial meaning of cases such as above prevents their combination with overt AxParts, which is not true of Loc heads like in. While a broad definition of in would include a meaning of containment, the Loc element in of in front does not imply this containment (in front of the house can be in the open air, contained by nothing). Thus, it appears that a Loc head such as the in of in front has lost some of its inherent spatial reference. On the other hand, the Loc heads above, below, beyond, and behind do have reference to a spatial area as part of their meaning, e.g., above refers to a the space higher than some reference point. As AxParts also refer to spaces, we would expect them to be problematic in combination with any Loc head which refers to some area in space. Thus, the restriction on the Loc heads in (9a) and AxParts appears to be a semantic, rather than a syntactic issue. Given this evidence, the structure I assume for the prepositions in (9a) is given in (13): (13) above the house LocP wo Loc AxPartP above 3 AxPart KP Ø 3 K DP Ø 6 the house Finally, I turn to inside and outside. It is interesting to note that of­-insertion can apply optionally with inside and outside:



Chapter 12.  The preposition cycle in English 

(14) a. Mary is inside of the house b. Mary is inside Ø the house Furthermore, unlike above or beyond, for instance, these two prepositions are generally described as being compositional in nature with the form in + side (e.g., Hopper and Traugott 2003: 110; Svenonius, to appear: 4). This implies that the two prepositions in (14) have a Loc head (in and out) and an AxPart side. However, their variability with respect to of-insertion suggests that two structures are possible. Cases like inside of the house have a non-null AxPart and a structure parallel to (7). Cases like inside the house have a re-analyzed Loc head and a null AxPart, with a structure similar to (13). As I show in the next section, this variation is part of a cycle of reanalysis and renewal. In this section, I have demonstrated that the universal structure for prepositional phrases headed by Projective prepositions plays out in slightly different ways, depending on the preposition. There appear to be three different types of Projective prepositions: null AxPart (above, beyond, below, behind), optionally null AxPart (inside, outside), and overt AxPart (in front). I will show in the next section that this three-way division is three stages of a single grammaticalization process.

3. The preposition cycle: Structural and semantic reanalysis Returning to the historical data, we see some interesting facts about the members of the Projective group of prepositions. First, Table 2 shows the approximate date of first attestation as a preposition for each one. The prepositions which prohibit AxPart (the first four) are all attested much earlier than those with an optional AxPart. Furthermore, the item in the list with the most recent attestation (in front) is the one with an overt AxPart. Looking again at the data as seen in Figure 1, we see that, historically, above, below, beyond, and behind were all formed by the affixation of either a- or be- to various roots. In fact, Svenonius analyzes these cases as the “Place” heads (which I analyze as Loc heads) be- and a- combining Table 2.  Earliest attested date as a preposition for the Projective class Preposition beyond behind above below inside outside in front

First use as P (OED) 1000 1200 1340 1575 1790 1795 1847

 Cathleen Waters

historically with AxParts such as buven (2006:  74). Therefore, while a case such as above once had a non-null AxPart, it has subsequently been reanalyzed as a Loc head + null AxPart. Therefore, we see that, at different points in time, the non-null status of the AxPart varies. The data further suggest that this process is a cycle with three stages. In the first stage, a innovative prepositional element has both a Loc head and an overt AxPart (arising, for instance, from the grammaticalization and reanalysis of a relational noun, as discussed earlier). Later, the Loc + AxPart head is reanalyzed to Loc + null AxPart; this is the second stage, and, as we see with inside and outside, the system at this stage may have variability. Finally, the AxPart becomes prohibited, and the variability disappears. Figure 2 outlines this process, using prepositional phrases with the Projective class of prepositions as examples. Assuming that the process of inherent case marking between two nominals by of was in place by the end of the Early Modern English period, the model outlined in Figure 2 implies that forms with of should arise before those without of. Examining the earliest attestations for the forerunners of inside and outside (withinside and withoutside, respectively), we see this is indeed the case. (15a) shows the earliest data for the prepositional use of withoutside, which occurs with of, and which is fifty years earlier than the earliest attestation of the prepositional use of withoutside without a following case marker (15b). The data in (16a-b) show similar results for withinside. (15) a. 1638: Placing the Indians...without side of our soldiers in a ring battalia (oed/cmh) b. 1686: Why may not these imperfect Metalls…grow...withoutside the stalks (oed/rp) Stages of the reanalysis process Stage I: Separate Place head and AxPart [Place in [AxPart front [K of [DP the car]]]] Stage II: Variability [Place in [AxPart side [K of [DP the car]]]] [Place inside [AxPart [K [DP the car]]] Stage III: Non-null AxPart prohibited [Place above [AxPart [K [DP the car]]]

Figure 2.  Stages of the preposition reanalysis process (with examples)



Chapter 12.  The preposition cycle in English 

(16) a. 1712: What passes … within-side of those Vehicles b. 1760: The stake … they run up withinside the spinal bone

(oed/cmh) (oed/hb)7

This variability continued at the time of the loss of with to form inside and outside as seen in (17) and (18), yielding the variability that continues today. In fact, there is even documented intra-speaker variation, as seen in (19). (17) 1769: Wearing their shirts withoutside of their cloaths.

(oed/jb)

(18) 1774: Succeeding each other, one without side the other, like circles in disturbed water. (oed/og) (19) 1783: a. Withinside of the hedge there is a bamboo pagar or paling… (pg/wm) b. …a small kind of leech, which dropped on us from the leaves of the trees, and got withinside our clothes. (pg/wm) c. … these feasts are never suffered to take place withinside their own kampongs. (pg/wm) The process seen above is consistent with observed patterns of grammaticalization. In the first stage, the elements are independent and each has lexical meaning. For example, in and front are not combined in the orthography and both are polysemous with what would be recognized as free morphemes (in and the nominal front). However, their meaning together is already different than the sum of the preposition + nominal (as described above), suggesting they have moved along the cline towards grammaticalization. In the second stage, one element optionally, but not obligatorily, is null. Thus we see inside and outside appear as single words in the orthography, though the structures are variable. Finally, the AxPart becomes null as the previously independent elements are reanalyzed as a single, higher head, and their individual standing as morphemes may disappear. The reanalyzed element is interpreted as the higher Loc head for semantic and syntactic reasons. Semantically, the Loc head, rather than the AxPart head, provides directional information based on the region identified as the Ground by the DP (see Svenonius, to appear: 7 for a detailed account). Syntactically, this semantic distinction between the two heads suggests that the Loc head must have a feature such as [i-direction] which is not present in the AxPart. As spatial prepositional phrases convey directional meaning, we would expect that a reanalysis of Loc + AxPart would retain the [i-direction] feature and thus the reanalysis yields an element that is a Loc head. 7. The OED contains an entry which, at first glance, appears to be an earlier use of withinside followed directly by its object: both without and withinside other plants (1686: OED/RP). However, this instance of withinside occurs in conjunction with without, a preposition that, at that time, as today, directly assigned accusative case, as seen here: A little Chappell a mile without the Village (1632: OED/WL). As accusative case would be assigned by without, the process of ofinsertion would not apply.

 Cathleen Waters

In and of itself, this is a process of grammaticalization, but not a cycle. To complete the preposition cycle, a further step is necessary: a new AxPart must be added. Based on the historical data, there are two possible ways this can happen. In the more straightforward case, the cycle may be renewed simply by the addition of a new AxPart head. For example, Figure 3 (data from the OED) shows the evolution of outside. The adverb outen started as an adverbial element and then became an AxPart, in this case following with. However, by the early seventeenth century without had been reanalyzed as a Loc head, and thus it was possible to add a new AxPart (side). As demonstrated in the data above, the AxPart phrase began to be followed by of. Over time, with was lost, and the present day preposition outside came into use. A similar path was taken by inside (via withinside). The historical data also reveal a second means to renew the cycle. As seen in the data in Figure 1, an older element may be bolstered by the addition of a new prepositional element, such as by or on. For example, the prepositional element a (on) was added to bufan (above) in the early fourteenth century. However, as the Loc head is already (and always) occupied in the hierarchy, it would seem no position would be available for adding a new prepositional element. However, evidence from Present Day English suggests how an additional reanalysis allows the introduction of a new AxPart. The cycle for outside [Adverb outen] c.900

c.1640

[Loc with [AxPart outen]] [Loc without [AxPart Ø]]

reanalysis

[Loc without [AxPart side]]

new AxPart

[Loc without [AxPart side]]

reanalysis

c.1690

or [Loc withoutside[AxPart Ø]]

variability

[Loc out [AxPart side]] Present day

or [Loc outside[AxPart Ø]]

Figure 3.  The preposition cycle for outside

variability



Chapter 12.  The preposition cycle in English 

As noted earlier, it is possible to have what appears to be multiple adjacent prepositions, and these are more common when the two prepositions have similar meanings: (20) up above our heads The elements that can appear higher in the structure are Particles, including up, down and in; these Particles also occur as Loc heads. Thus, we have two versions of each Particle: e.g., up1, which occurs higher in the structure than the Loc head, and up2, which is a Loc head. A structure such as up1 above can be reanalyzed such that the second element (above) becomes an AxPart following the homophonous Place head (up2). This process does not involve the higher elements moving down the tree, but rather it is a reinterpretation of a whole structure. As noted earlier, AxPart can evolve from a variety of syntactic types, not only nominals, which allows some flexibility in what may be recruited for the AxPart role. Furthermore, as cases such as above appear to retain an inherent reference to a space, they are good candidates for reanalysis as AxPart. The presence of two preposition-like elements can plausibly lead to a reanalysis of the structure; this process is one of reinforcement, rather than a change in meaning. As the OED states, “The simple ufan originally expressed the whole idea of its successive expansions b(e)ufan, a-b(e)ufan.” Based on these observations, Figure 4 outlines how the cycle has taken place for above (all data from the OED). I have suggested in Figure 4 that we are approaching reanalysis because there is data from Present Day English sources which suggest that we are on the verge of the The cycle for above [Adverb ufan] c.1000

[Loc be [AxPart ufan]] [Loc be (e)ufan[AxPart Ø]]

reanalysis

[“P1” a [Place buven[AxPart Ø]]] c.1300 Present day

[Loc(“P2”) a [AxPart buven]] [“P1” up [Loc above[AxPart Ø]]

Figure 4.  The preposition cycle for above

reanalysis approaching variability

 Cathleen Waters

next renewal of the cycle. The data in (21) – (24) were collected from Google from what are plausibly native speakers: (21) This idea up above of Rahm Emanuel would be a waste of time and sorely needed funds. (G/G1) (22) So if I chose one of these other ones you would see a preview down below of what that line looked like. (G/G2) (23) I love the picture up above of your site (G/G3) (24) You’re on your way to down below of your own goddamned intention (G/G4) While the of phrases in (21) and (22) clearly relate to the nominal preceding up above and down below, they provide an environment for reanalysis. The data in (23) and (24), on the other hand, appear to be early indications of reanalysis, as it clear from the larger context in which of your site in (23) appears that it refers to the location (not the content) of the photo.8 Furthermore, the of phrase in (24) clearly does not apply to the preceding nominal (way). What determines which path is taken? Both cycles, the one seen with outside and the one with above, are reinforcements of the previous meaning, rather than introductions of a new meaning; the two paths achieve the same end. Therefore, there must be something outside of the cycle which determines which route a preposition follows. It is interesting to note that without contains an element (out) that also exists (polysemously) as a Particle. Thus, in cases where the Loc head resembles a Particle, the reanalysis may tend toward the addition of an AxPart, whereas Loc heads that do not resemble existing Particles may be analyzed in conjunction with a higher Particle.

4. Conclusion In this paper, I outlined the syntactic processes that contribute to the cycle of reinforcement observed in spatial prepositions in English by focusing on above, behind, below, beyond, inside, outside, and in front, prepositions with similar semantic and syntactic behavior. Using a recent proposal by Svenonius for a hierarchy of functional heads, and historical data, I demonstrated that the prepositions in this class are at different stages in a grammaticalization process, namely, the reanalysis of Loc and AxPart heads. The least grammaticalized of the group is in front, in which the AxPart is nonnull; the most grammaticalized are beyond, behind¸ below, and above, in which the AxPart is always null. Furthermore, I outlined how the presence of the null head ultimately creates a situation in which renewal of the cycle can occur. The renewal may take place with the addition of a new AxPart head, as seen in the evolution of outside. Alternatively, the Loc head and a higher Particle element may be reanalyzed as Loc + 8. As a native speaker of English, I find these last two slightly odd, but not nearly as bad as something like above of your site.



Chapter 12.  The preposition cycle in English 

AxPart, as has occurred in the history of above. These changes taken together form a grammaticalization process, the preposition cycle. In the future, it would be interesting to extend this model to other spatial and non-spatial prepositions, both in English and in other languages.

References Anderson, John M. 2006. Modern Grammars of Case. Oxford: OUP. Chomsky, Noam. 1986. Knowledge of Language. New York NY: Praeger. Gelderen, Elly van. 2004. Grammaticalization as Economy [Linguistik Aktuell-Linguistics Today 71]. Amsterdam: John Benjamins. Heine, Bernd, Claudi, Ulrike & Hünnemeyer, Friederike. 1991. Grammaticalization: A Conceptual Framework. Chicago IL: The University of Chicago Press. Hopper, Paul & Traugott, Elizabeth Closs. 2003. Grammaticalization. Cambridge: CUP. Project Gutenberg Online. http://www.gutenberg.org/wiki/Main_Page. Retrieved January 2009. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: CUP. Svenonius, Peter. To appear. Spatial P in English. In The Cartography of Syntactic Structures, Guglielmo Cinque & Luigi Rizzi (eds), Vol. 6. Oxford: OUP. Svenonius, Peter. 2006. The emergence of axial parts. Nordlyd: Tromsø Working Papers in Language and Linguistics 33: 49–77. The Oxford English Dictionary Online. Retrieved December 2007. Waters, Cathleen. 2008. The internal structure of place class prepositional phrases in English. Proceedings of the Canadian Linguistics Association annual meeting 2008.

Appendix – Abbreviations in data CMH: Collections of the Massachusetts Historical Society Ser. III. VI. 23. 1638. G1: http://www.examiner.com/blog/includes/loadComments.cfm?id=46655&blogID =536&commentcount=811&lastID=64636%2C62486&stRow=1%2C21%2C41. G2: http://movielibrary.lynda.com/transcripts/37048/WordPerfect_Office_X4_Essential_Training/Creating_a_table_of_contents/ G3: http://adifferentlight.wordpress.com/2007/04/03/do-you-really-want-to-hurtme-respecting-the-physical-boundaries-of-people-with-disabilities/ G4: http://wiki.eveonline.com/wiki/Under_the_Sea,_the_City_(Chronicle) HB: Brooke, Henry, 1703–1783. The fool of quality; or the history of Henry, Earl of Moreland 1760–72. JB: Bush, John. Hibernia curiosa. 1764.

 Cathleen Waters

OG: Goldsmith, Oliver, 1730–1774. A survey of experimental philosophy. 1774. RP: Plot, Robert, 1640–1696. The natural history of Staffordshire. 1686. WL: Lithgow, William, 1582–1645. The totall discourse of the rare adventures and painefull peregrinations of long nineteen yeares travayles III. 94. 1632. WM: Marsden, William, 1754–1836. The History of Sumatra: Containing An Account Of The Government, Laws, Customs And Manners Of The Native Inhabitants. 1783.

part iv

An experiment

chapter 13

The study of syntactic cycles as an experimental science Roeland Hancock and Thomas G. Bever University of Arizona

Linguistic cycles in the syntactic expressions of particular functions (e.g., negation) exemplify a dynamic competition between distinct representational architectural constraints. We study these constraints in an experimentally controllable synchronic paradigm, using situated artificial language learning to induce syntactic cycles. Adults and children learn to use different languages in production and comprehension. The languages appear with frequencycontrolled variability in alternate forms with a directionality that the representational architectural constraints predict. Each subject’s learned output is the training input of the next, to accelerate syntactic changes. Questions include: do learners pick up new usage patterns in the predicted cyclical direction; do children change their usage in the predicted direction faster or with weaker frequency bias than adults; what is the relation between emerging changes in language usage and grammaticality intuitions? This research program offers detailed information about cyclic transitions that can enrich an understanding of historical data. Most important, it can give direct experimental confirmation of the dynamic presence of the architectural constraints proposed as the cause of historical cycles.

Introduction The study of language histories has served the field of theoretical linguistics by taking historical changes as the effects of dynamic inner constraints on possible languages that create the changes. Of great interest for linguistic theory is the discovery that certain kinds of patterns of change are repetitive: in the simplest case, a particular language structure type 1, becomes type 2, and then type 1 again, and then type 2…etc. These are known as “historical cycles”, described in other chapters in this volume.1 1. Abbreviations include: HPP = Head Preference Principle; LMP = Late Merge Principle: CFP = Clitic Formation Principle; FRP = Functional Recovery Principle; ALL = Artificial Language Learning; sALL = Situated Artificial Language Learning.

 Roeland Hancock and Thomas G. Bever

Van Gelderen is a principal in creating a syntactic architectural theory of why the cycles exist – essentially a dynamic between two kinds of representational constraints in syntax. The important underlying theme is that at any given time, specific constructions in the overt syntactic structure of a language reflect a computational compromise between the different constraints on linguistic representations. Over time, one constraint dominates or replaces the other, and then is dominated or replaced by the other, resulting in the overt historical cycle. Typical overt cycles are seen in negation, agreement systems, pronominal systems, tense systems and other functional constructions. The child learner is parent to the adult language. Thus, we are interested in understanding the language constraints and learning dynamics in the individual language learner that result in the cycles of language change. The logic underlying our program is this: (1) a. the language learner is the dominant force in re-analyzing his/her language, b. historical cycles demonstrate the competition of different architectural principles of syntactic representations, c. the competing architectural principles must be available to the learner. Thus, the historical facts motivate a dynamic architectural theory, which in turn must claim that the dynamic is part of the language learner’s cognitive and linguistic repertoire. The problem we address here is how to study the learner’s access to the architectural principles in a more direct way than delineating the historical cycles. That is, we would like to find synchronic evidence that the child indeed applies the competing architectural constraints as s/he learns the language. Below, we outline how we are using a naturalistic, “situated” artificial language paradigm and transmission of language training across “generations” of learners to study experimentally the dynamics of language change. It is critical that the paradigm can be used with children who are still in the normal language learning age range. The kernel of the idea is first that we can structure the artificial language to be intermediate between different historically attested cycles, and examine which direction the language is changed by the learners as they make and transmit errors. Secondly, we can manipulate the dynamics of how the training is presented, presenting critical examples in different ways, to illuminate how the individual learner interacts with the data and imposes constraints on it. Finally, we can probe the relationship between learning to use a new grammar as knowledge shifts from one stage to another, with the emergence of grammaticality intuitions about the new structures. This affords a unique opportunity to study the relation between usage and intuitive categorical knowledge in a controlled setting. We first review salient aspects of van Gelderen’s proposals, current practice in artificial language learning, and recent artificial language studies of compressing the “evolution” of language. Then we introduce our paradigm and some sample studies. The reader should note that the application of sALL to the study of “synchronic cycles” was sparked by the recent conference that resulted in this volume. We have created the paradigm and are piloting studies but we do not yet have definitive data. Thus, the



Chapter 13.  The study of syntactic cycles as an experimental science 

initial goal of this paper is to outline the logic of the prediction that historical cycles make for the language learner, and to present a method for studying those predictions.

1. The economy of language change Van Gelderen (2004, 2008) proposes that two minimalist principles of computational efficiency and locality largely drive syntactic change in conjunction with a cyclic process of phonological weakening and strengthening. The reader will be aware that van Gelderen’s theory of reference is the minimalist program, although the principles she adduces may be statable in other frameworks. We do not dwell on the theoretical background here; the reader is again invited to consult the other papers in this volume. The principles we will use are given in (2) and (3):

(2) The Head Preference Principle (HPP): Material in head positions is more structurally accessible for feature checking compared to Spec-head checking.

The typical result of the HPP historically is that a double-marked process, e.g., for negation, which has to be analyzed as Spec+head, collapses to a single process, allowing analysis of the head without Spec.

(3) Late Merge Principle (LMP): Lexical insertion (external merge) is preferred in situ rather than requiring “move” (or “copying”, internal merge).

Typically, a lexical item in an embedded part of a phrase (e.g., in the VP) which moves (or is copied) outside of the phrase becomes directly inserted in its final position and takes on properties of a function word. For instance, nothing inside a VP becomes not outside the VP. The associated result is that the word that originally required a theta role, now becomes a pure “syntactic” word without a theta role. Two additional well-attested principles play a critical role in interacting with these principals in the dynamics of historical change. They are given as (4) and (5):

(4) Clitic Formation Principle (CFP): Functional elements with primarily formal content become incorporated phonologically as components of head phrases. (5) Functional Recoverability Principle (FRP): Functional elements must be sufficiently clear phonologically to trigger distinct computational processes.

While learnable and economically constrained by the first two principles, the usability of a grammar they motivate may suffer through reduced phonological clarity as the function word becomes a clitic, thereby reducing its recoverability (e.g., /not/ goes to / nt/). This can motivate the adoption of new elements to reinforce the increasingly phonologically weak marker to make sure there is functional recoverability. As additional elements (e.g. double negatives) appear, the cycle begins again. For example, consider

 Roeland Hancock and Thomas G. Bever

a canonical example of language change studied by Jespersen (1917) in Germanic languages. This is stated for English, following van Gelderen (2008) as Figure 1. Stages L1 to L3 depict possible NegP constituents at various stages of the negative cycle. At the L2 stage no thing becomes a grammaticalized negative marker, noght, and the phonologically weak negative marker ne gradually becomes a clitic, leaving an empty head under NegP. In L3, this position is filled with not, formerly in Spec, according to the HPP. Roughly at the same time, noghi (as a negative marker) weakens phonologically to not(n’t). The weakening eventually prompts the appearance of a new reinforcing element and the cycle begins again. This aspect of the cycle can be seen in those modern dialects of English that permit double negatives. Jespersen’s Cycle of historical changes in the negative construction is a classic and well-studied example of syntactic change. The evidence in general for syntactic cycles comes largely from incomplete historical texts, making difficult an empirical investigation of the underlying driving factors, such as van Gelderen’s hypothesized economy principles. Even the full negative cycle – portions of which are well-attested cross-linguistically – has not been completely documented in a single coherent historical model (Dahl 1979). NegP no thing

FRP-Phonological strenghtening and renewal of the cycle

L1 Neg’

Neg

VP

ne

no thing

LMP eliminates ‘movement’ from VP

L3 NegP

NegP Neg’

Neg not

Figure 1.  The negative cycle

no thing VP

Neg’ Neg

HPP fills phonologically weakened head with fromer Spec L2

ne

VP



Chapter 13.  The study of syntactic cycles as an experimental science 

The critical feature of van Gelderen’s ideas is the particular principles; the historical data serve as confirmation of them: thus, the incompleteness of historical data is not a crippling problem, since complete prediction of the historical changes is not intrinsically required by the claim that they result from the interaction of the particular structural constraints. However, it will be useful to be able to test the principles actively rather than being dependent on incomplete historical records. To that end, we are applying a naturalistic paradigm for “situated” artificial language learning that can then be used with adults and children to test the effect of the theoretical principles. Below we discuss some background involving artificial language learning. We then present our new paradigm and examples of how it may be used to test the synchronic principles that have been proposed to explain the diachronic cycles.

2. Artificial language learning and change In this section, we discuss artificial language learning in general, how it has been applied to language change, and some criticisms of the traditional paradigm.

2.1

Artificial language learning

In the more experimental areas of linguistic inquiry, artificial language learning (ALL) paradigms have been used to study conditions on language learnability and processing mechanisms. ALL paradigms have four salient features, given in (6): (6)

a. b. c. d.

the structure of the grammar to be learned; the way the evidence is presented to the subjects; the discriminative response expected from the subjects to show learning. the motive to learn the grammar.

In the most basic of these paradigms, a subject is passively exposed to strings generated by some form of a grammar (often finite state). After some training period, the subject’s ability to distinguish novel strings generated by the target grammar from ungrammatical strings is tested using a head-turn or listening time preference paradigm (in the case of infants; cf Kemler Nelson et al. 1995; Gómez & Gerken 1999). In the case of adults, a standard procedure is to present examples of grammatical strings and then to assess learning through explicit grammaticality or recognition judgments (Reber 1969). Procedures like these are limited to tests of simple pattern-recognition ability, usually of grammars with simple formal structure; attempts to teach passive subjects more complex grammars have been generally unsuccessful. Subjects can learn to make grammaticality judgments on more complex grammars, including simple context-free grammars, when explicit instruction is provided. In this version of the paradigm, a subject is presented with a series of grammatical and

 Roeland Hancock and Thomas G. Bever

ungrammatical strings and makes grammaticality judgment in response to each, receiving explicit feedback after each trial (cf. Friederici et al. 2006) In a very small number of cases, subjects are taught to use the language in some way, without being given explicit instruction in grammaticality – rather, their successful use is dependent on creating or understanding or extending sequences with correct grammatical knowledge (Bever and Hansen 1988). Perhaps the most natural procedure of training and learning an artificial language to date is that introduced by Friederici et al. (2002). Their ultimate goal involved examining brain activation patterns when making grammaticality judgments about the artificial language they have learned. Their language, “Brocanto”, had 14 words that characterize a small set of objects, verbs and locations in a matrix of locations on a screen. The language was interesting in that it included “moves” that objects could make from one location to another on the screen. Learning was inculcated by having subjects work in pairs, in which each subject had to tell the other the move s/he had just made in Brocanto, and the other subject had to make the corresponding move on his/her display. In this way, Friederici et al. captured many aspects of normal language learning: notably, subjects were motivated by the need to communicate, subjects received both production and comprehension experience, and the communication was entirely verbal (with automatic computer correction of errors). Subsequently, Morgan et al. (2009) have used a variant of Brocanto, but with subjects trained individually in separate training blocks with feedback for production and comprehension, as in Bever and Hansen.

2.2

All studies of language change

Over the past decade, a number of computational models have been developed in support of the idea that language ‘evolves’ as an adaptive means of transmitting cultural information through successive generations. These iterated learning models (ILMs), have recently been brought into the realm of behavioral experimentation using an ALL paradigm (Kirby, Cornish & Smith 2008). Kirby et al. draw ambitious conclusions from these experiments about mechanisms for the original evolution of language, which are not immediately relevant to our research. But their methodology may be useful when applied to a more grounded study of the dynamics involved in syntactic change. Kirby and colleagues use a recursive multi-generation paradigm, paralleling the prior simulations, in which subjects are instructed in a language as generated by a previous participant under the guise of learning alien descriptions for simple moving objects. Each subject passively views a number of picture-string pairings and is then called on to produce strings for a set of test pictures (half of which were previously seen by the subject). Starting from completely random picture-string pairs, the first few subjects face an impossible memory task and, as may be expected, produce a number of recall/transcription errors. Eventually, the strings given for novel pictures are the product of innovative systematization of various kinds that simplify learning and retention. The subject’s pairings, including any errors, are then passed to the next



Chapter 13.  The study of syntactic cycles as an experimental science 

“generation” of learners (i.e. the next subject in the experiment). The general result of these studies is to show that subjects tend to create paradigmatic “morphology” out of systematic errors in learning, memory and transmission to succeeding subjects. The final state of the iterative process generally looks like classic paradigms of a morphologically rich sub-language, in which individual sub-components of “words” are systematically related to particular objects or particular kinds of motions. In other words, over successive mis-learnings and mis-recalls, subjects converge on a componential relation between the words and what they refer to. Kirby et al. argue that this is a miniature replication of how language might have evolved. This is an ambitious interpretation of what is justified by their subjects “itemand-arrangement” solution to an otherwise impossible learning problem. But for our purposes, their method has pioneered a way to hasten the impact of formal constraints on possible languages on changes in time: as subjects learn what they can, and that becomes the model for the next set of subjects to learn from, we can trace dynamic changes in the linguistic structure itself. Hudson Kam & Newport (2005) study the acquisition aspect of language change, with particular focus on the regularization of creoles. In this paradigm, participants are instructed in artificial VSO languages containing various degrees of unpredictable variation in the use of determiners as part of N + Det structures. The degree of variation had no effect on the ability of participants to learn vocabulary or make forcedchoice grammaticality judgments. In a sentence completion task, however, children (mean age 6;4.10), but not adults, regularized determiner usage (or non-usage). It is hard to pin down how much of the training involved natural features, but the sentence completion task certainly recruited both comprehension and production processes.

2.3

Criticism of traditional ALL paradigms

Consider a broad outline of the circumstances in which children usually acquire knowledge of a language and grammaticality sense. The terms mentioned in (6) characterize and differentiate ALL paradigms. In (7), more detail is provided. (7) a. The structure of the grammar to be learned. This of course, is the subject of linguistic investigations and theories with many conflicting ideas and proposals. Yet, several features endure across many different linguistic theories: sentences are a natural unit of complete meanings, with structure and meaning compositionally derived from their constituent phrases; sentences have derivations, relating inner (aka ‘logical’) form to outer structures and phrasal sequences. b. The way the evidence is presented to the subjects. Children characteristically hear language in its natural use, not a formal or didactic setting with strictly grammatical sentences correctly applied. They hear fragmentary sentences, optional variation in particular constructions, and outright

 Roeland Hancock and Thomas G. Bever

ungrammatical sequences that are nonetheless communicatively effective. Most important, the language exposure is “open ended” in the sense that there is no formally circumscribed domain of discourse: at the same time, there is usually a set of focal topics with a good deal of repetition both of content and form. c. The discriminative response expected from the subjects. As they grow up, children are expected to pronounce their language correctly, to understand the sentences appropriately, and to speak in a natural way. They are not required to speak strictly grammatically, nor are they required to understand only grammatically correct and complete utterances. Rather, they are expected to speak and understand with the same variability, appropriateness and correctness as adults. Most important may be the fact that they are required not only to understand sentences but also to produce them as part of their natural language behavior. d. The motive to learn the grammar. This is the most vexed and controversial of matters in the relevant fields. The simplest functionalist answer is that learning the grammar is the best vehicle for achieving what is required in (c): that is, speaking and understanding fragmentary and sometimes ungrammatical sequences as well as grammatical sentences may be best facilitated via knowledge of the correct structures, and then behavioral habits that deviate from them in standardized ways. The simplest structuralist/ nativist answer is that the child is predisposed to learn particular kinds of grammatical structures, with great filtering power over the apparent variability in its linguistic experience. In a general sense, both approaches to the problem may be true, since they tend to complement each other. A different kind of model has been suggested by Bever, namely that the abstract grammar is acquired by individual children, one at a time, because it reconciles conflicting representations between what the child can say and what it can understand: in this sense, the language structure presents a canonical problem of the kind that humans like to solve: that is, learning the abstract grammar is intrinsically mental fun. (Bever 1975, 1992, 2008). In light of these characteristics of natural language learning, ALL paradigms necessarily place subjects in an unnatural learning situation. The most problematic aspect of the paradigm is the usual focus on grammaticality judgments, a rather sophisticated behavior that children are rarely explicitly taught. By explicitly teaching grammaticality judgments, these paradigms shortcut the mechanism through which the ability to make such judgments naturally appears, with unknown consequences for the internal grammatical representation. The lack of semantic context and production in many other ALL paradigms preclude their use for any serious study of language change (but note Friederici et al. 2002, which does capture many natural features).



Chapter 13.  The study of syntactic cycles as an experimental science 

Dulany, Carlson & Dewey (1984; also Perruchet & Pacteau 1990) provide evidence that the basic ALL paradigm does not produce (or at least does not measure) unified internal grammars. In these experiments, subjects provided explicit information about their grammaticality judgments, as by indicating where an ungrammaticality occurred in a string of letters. Explicit judgments were highly predictive of performance when the ungrammaticality was the result of a simple pattern violation, such as an illegal bigram. Subjects were unable to articulate more complex violations or violations of multiple simple rules. These failures of explicit knowledge were closely correlated with chance performance, suggesting that explicit knowledge often drives ALL performance. Natural language competence, in contrast, is to some extent disassociated from explicit knowledge and requires a more coherent knowledge (conscious or unconscious) of the grammar. A few of the paradigms we have reviewed mitigate the drawbacks of ALL in one way or another. The Kirby et al. paradigm, in which subjects basically tell and then retell simple utterances, involves creating “natural” errors and variability in what successive subjects experience. Yet the paradigm is relatively artificial in how the stimuli are presented, and the range of contents. The Hudson Kam et al. studies involve actual situations, experimenter-controlled variability in the critical data the child experiences and a relatively natural response (completing sentence fragments). It is also relatively unusual in using auditory and verbal activity rather than displayed or typed texts. Yet an important virtue of the paradigm also limits its use: subjects are in a real person-toperson context with natural responses possible and studying young children, still in the normal first-language-learning age range. This requires time-consuming video analysis of data instance by instance, with loss of much information and requiring inter-judge reliability measures. Friederici et al. (2002) come closest to a paradigm that is natural in relevant ways. We will now discuss an alternative which combines some virtues from the methodologies in Bever and Hansen, Friederici et al. and Kirby et al.

3. Situated artificial language learning Our goal is a paradigm that captures certain essential properties of ALL methods that will make it possible to study dynamics of learning and language change in children, with a relatively natural auditory/verbal learning situation. To this end, we have developed a novel ALL paradigm, “situated ALL” (sALL), that allows us to study links between language use and grammar acquisition. Our initial results show that moderately sophisticated grammars can be learned through usage in an experimental setting and demonstrate the predicted importance of integrating perception and production in learning. The main feature of the sALL paradigm is semantic context in the form of a simple 2-dimensional visual world. A sample display is shown in Figure 2, where in the real experiment some of the shapes have color.

 Roeland Hancock and Thomas G. Bever

Figure 2.  Sample computer interface presented to subjects

Using our custom software, subjects can freely create and move geometric shapes on the screen and change any of several visual attributes. A corresponding phrase structure grammar, as in Figure 3, can be used to provide a linguistic description of a given visual arrangement, specifying the appearance of each shape and their spatial relation to one another. For example, (8) describes the arrangement in Figure 4 using the grammar in Figure 3:

(8) red star solid green triangle dotted the yellow diamond left-of above the solid, red star is above the dotted green triangle that is left of the yellow diamond.

After a short practice session (creating arrangements from English instructions), subjects become proficient in the use of the software. In a typical experimental session, a subject is pseudo-randomly presented with either a sentence from the target grammar (displayed on screen or spoken through headphones) or a visual world. The subject is then prompted to produce a corresponding world or sentence. In the most natural case, linguistic stimuli are spoken via the MaryTTS text-to-speech system (Schröder & Trouvain 2003) and responses are verbal, processed by speech-recognition software with a limited vocabulary (Dragon Naturally Speaking). The visual world is manipulated using a touch screen display, allowing for a minimal technology barrier.



Chapter 13.  The study of syntactic cycles as an experimental science  S NP Pred NP (ADJ) Noun (Size)(Pattern)(Clause) Clause te Pred Pred ADJ Pred NP Rel Pred NP NegP Noun {square, rectangle, star, pentagon, diamond, triangle} Size {small, medium, large} Pattern {striped, dotted, solid} ADJ {red, yellow, green, blue, purple, thick, thin,whole, cutout} Rel {above, below, left-of, right-of, behind, infront-of} L1 NegP ne Rel (NOT) Noun {..., nothing, anything} L2 NegP (ne) Rel (Adv) NOT L3 NegP Rel NOT

Figure 3.  Rules for three simple artificial languages. Each language contains the basic rules and one of the last three rules for negation

Figure 4.  A production trial with the target “red star solid green triangle dotted the yellow diamond left-of above” (the solid, red star is above the dotted green triangle that is left of the yellow diamond)

 Roeland Hancock and Thomas G. Bever

After responding, a correct response is presented. These mapping trials constitute the training portion of the experiment and provide a continual measure of production and comprehension performance. Competence is periodically probed through grammaticality judgment trials. On these trials, a novel sentence is presented in isolation and the subject is prompted to judge whether or not the sentence is possible in the learned language. No feedback is given. Bever & Hansen (1988) used a very simple version of a similar paradigm to examine the role of bidirectional language usage in grammatical induction. Subjects participated in a production-only (picture>sentence), comprehension-only (sentence>picture) or mixed condition. Subjects in the mixed condition show markedly higher performance on production trials, despite receiving half the production training of those subjects in the production-only condition, and in making grammaticality judgments. These results support the hypothesis that bidirectional language usage facilitates the development of structural representations. The change in mapping performance also suggests a transfer effect between language comprehension and production which is modulated by the emerging grammatical representations (see Bever, 1975, 1992, 2008). Earlier studies and our streamlined paradigm open up the possibility that we can adapt it to study the basic principles that have been supported by diachronic syntactic cycles. The new version of the paradigm is one that children can master, indeed may do much better than adults, using the verbal input and output facilities. There are many uses for this paradigm, including sophisticated studies of neurological organization of different kinds of grammars with different kinds of training regimes. The new questions for the concerns of this paper are: (9) a. Can we replicate diachronic cycles synchronically? b. Can we elucidate with more detail than allowed from the historical data, the basic principles that constrain the cycles? c. Will the progressive direction of the cycles depend in part on both comprehension and production experience? d. How will the change in grammatical structure use be reflected in elicited grammaticality judgments. e. Are children under ten more likely than adults to respond categorically and rapidly go through cycles? In considering how the sALL paradigm might be used to study syntactic cycles and grammaticalization, we adopt van Gelderen’s (2008) theoretical framework of syntactic cycles as a dynamic process stemming from conflicting constraints. One possible methodology for studying a cycle like the classic negative one follows Hudson-Kam and Newport (2005), presenting variable mixtures of negative forms on comprehension trials and examining the corresponding frequencies on the production side. As discussed above, our paradigm is suitable for use with children, where any cyclic shift should be most apparent. This method can be extended and



Chapter 13.  The study of syntactic cycles as an experimental science 

possibly sharpened by applying a paradigm like that of Kirby et al., in which the input data to each subject is the final stage of output from a prior subject. The introduction of variability drawn from different stages of a cycle is critical to our synchronic studies. First, we cannot expect subjects to spontaneously develop “errors” in the direction of or against a cycle in such a short time – historical cycles develop over many generations of learning and adult use. Second, we can calibrate the effect of directionality by manipulating the relative frequencies of the alternate variants of a cycle – we can expect that the frequency of alternate forms that move in the predicted direction can be lower and have an effect than the frequency of alternate forms that move against the predicted direction. That gives us a quantitative tool to compare adults and children. All subjects may shift in the predicted direction to some extent, but we predict that the balance of alternate forward and backward forms that shift children in the predicted direction will be less extreme than for adults. A separate feature of our paradigm is that we can contrast changes in language behavior patterns governed by cycles against corresponding changes in grammaticality intuitions. This will allow us to investigate the dynamic relation between gradual shifts in artificially induced ‘dialects’ against shifts in grammatical representations. The proposed experiment illustrates an empirical approach to historical cycles, using the negative cycle as a convenient and well-understood example, but the paradigm has a straightforward application to virtually any other cycle. The basic principle comes from the Hudson Kam & Newport study, namely that learners (at least children) will tend to reduce variation during production. In the Hudson, Kam & Newport study, this behavior is likely the result of general learning mechanisms. Here, we introduce economy factors that should amplify (or dampen) the regularization effect. The training set (presented as a mixture of production/comprehension trials) conceptually consists of three languages, corresponding to the three stages of the negative cycle. Twothirds of the relevant training sentences use negative constructions; two-thirds show one of three non-cyclic variations (control languages). The overlapping sentences are fully crossed control markings and negative constructions with equal frequency for each combination. The frequency of each construction for each experimental condition is given in Table 1. The Forward condition (and Forward' which reverses the control mixtures) tests the tendency for the negative cycle to progress in the canonical direction. The Equal condition serves as a control with respect to the control constructions (presented with equal frequency) and measures any tendency for the cycle to progress in the absence of frequency differences. The Reverse (and Reverse', again reversing the control mixtures) condition allows the unidirectionality of the cycle to be tested. Subjects will be trained to an acceptable level of performance (note that any of the possible negative and control constructions will be acceptable on production trials). Following criterion, critical test production trials eliciting negative and marked constructions will be given.

 Roeland Hancock and Thomas G. Bever

Table 1.  Experimental language mixtures Basic Word Order

L1 L2 L3 Control 1 Control 2 Control 3 Control1,2,3 × L1,2,3

N N Neg V (Neg) N N (Neg) V [(Adv) Neg] N N V Neg

Mixture (%) Forward (Forward')

Equal

Reverse (Reverse')

4.7 9.5 19 4.7 (19) 9.5 19 (4.7) 3.7

11 11 11 11 11 11 3.7

19 9.5 4.7 19 (4.7) 9.5 4.7 (19) 3.7

The variability in case marking provides a control and baseline of regularization. Based on the results of Hudson Kam & Newport, adult participants should show a pattern of probability matching, rather than regularization, for case marking. The same behavior may be seen for negation, although a change in negative construction probabilities would be suggestive. Children should reduce variability in both the negation and case marking aspects of the language. A change in case marking probabilities should be purely the result of regularization; a differential change in negation probabilities would suggest that nonstatistical factors influence their distribution. Support for this can also be obtained by comparing the regularization towards L3 in the forward condition to L1 in the reverse condition. According to the economy explanation, L1>L2>L3 is the logical (and attested) ordering of the cycle. The L3>L2>L1 order is uneconomical since L3>L2 goes against the HPP and L2>L1 against the LMP. The production frequency shift is modeled as a function of regularization, grammaticalization and their interaction. From a regression perspective, this can be stated as (10), where R and G are regularization and grammaticalization functions, respectively: (10) f (Li) ≈ β1R(f(L)) + β2G(f(L)) + β3R(f(L))G(f(L)) + e The use of the control constructions provides an important baseline against which the effect of grammaticalization beyond the tendency for regularization can be determined. We can also deploy this paradigm in a version with successive subjects, following Kirby et al., to track the evolution of the different ways to express negation. Given the auditory/verbal manipulation we may be able even to enhance the cliticization of certain forms by pressuring subjects to speak quickly. Much of this is conceptual at the moment, and, of course, this particular experiment involves negation, which English speaking subjects have already learned in a form related to the historical variations. We can also use the paradigm with colleagues in countries with non Indo-European native languages (e.g., Chinese) to neutralize the impact



Chapter 13.  The study of syntactic cycles as an experimental science 

of directly related prior language experience. We will now give an outline of an experiment with a less attested cycle involving features not indigenous to English syntax. A different way to test the principles with English speaking subjects is to construct a cycle that could exist following the architectural principles, using a syntactic/semantic feature that does not play a role in English syntax. Consider for example, the relation between telicity, unaccusativity and reflexives, which does play a role in many other languages (e.g., Spanish: Tenny 1994, Sanz 2000). The sALL paradigm does allow for presentations and productions of different kinds of motions. For example, it can represent “explode” as a telic mono-argument predicate (also known as “unaccusative”) with a specific end point in time by showing an object that actually explodes visually: and “bounce” can be depicted as an atelic mono-argument without specific end points by having an object bouncing for the duration of the display (in sALL the depictions are actually dynamic, but Figures 5 and 6 represent them statically for this paper). We can construct a cycle, regardless of actual historical facts of English as to how telicity is marked, although there may be some evidence of actual cycles of this sort historically in the background of languages other than English. The different singleargument telic constructions would look like the sequence in (11), in order of change (the examples use actual English-like words for exposition, but in an sALL paradigm will be nonsense words): (11) the circle s[+telic]’exploded itself[+telic] The kinds of cases that we can present to subjects with differing probabilities are based on the following fictitious but theoretically motivated cycle, presented as Figure 7.

Figure 5.  A telic event (‘explode’) seen as a series of frames

 Roeland Hancock and Thomas G. Bever

Figure 6.  An atelic event (‘bounce’) seen as a series of frames

InnerAspP itself

L1 InnerAsp’

InnerAsp

VP

(se)

itself

LMP

FRP

L3 InnerAspP

InnerAspP InnerAsp’

InnerAsp

itself VP

InnerAsp

HPP

(it)se(lf)

Figure 7.  A theoretically motivated telicity cycle

InnerAsp’

L2

(se)

VP



Chapter 13.  The study of syntactic cycles as an experimental science 

Figure 7 depicts a cycle analogous to the negative cycle. In L1 an additional supportive word merges out of the VP to support the weakened clitic marker. Weakening continues and the support word becomes incorporated under Spec in L2. In L3, the clitic has disappeared and the support word appears in head position: (12) the circle exploded (it)self[+telic]. The cycle of cliticization and recoverability continues back to L1. The language also includes evidence for the transitive construction, namely (13). (13) the square exploded the circle The actual experimental paradigms based on this cycle can be organized in the same way as for the negative cycle example: subjects are presented with mixed cases that vary in the relative frequency of the adjacent construction types. The same experimental paradigm variants can be used to study the relative learning and directionality of mistakes, as well as paradigms that involve taking the learned output of one subject and using it as the training input for the next. We should note that while this cycle follows the architectural principles, it does not enjoy as much historical justification as many others. We find that to be a virtue – testing a cycle that should exist, but may be rare. Other, well attested cycles that are novel for English include subject/object agreement, and the copula cycle.

4. Conclusion This paper demonstrates how we can expand the empirical support for the structural principles proposed by van Gelderen. Our approach promises to supplement scattered historical data with synchronic experimentally replicable research. The outcome will be to verify the presence and dynamic competition of the structural principles. It will be of considerable significance if children impose the cycles relative to non-cyclic control structures and adults do not: this will lend further confirmation of the idea that the underlying architectural principles that result in cycles are part of the language learning child’s structural repertoire. An additional benefit may be that the paradigm allows for detailed analysis of “micro-steps” in the experimentally-induced miniature evolution of the cycles: this can sharpen the search for details in actual historical cycles that have hitherto been unnoticed. Finally, the data for actual cycles is based largely on documented usage, not grammaticality intuitions: we will be able to compare shifts in usage against shifts in grammaticality. This may give detailed insight into how shifts in usage emerge as shifts in grammatical representations. Since linguistic science at its best rests on accounting for details, this research may make some unexpected contributions to the field of historical linguistics and linguistic theory in general.

 Roeland Hancock and Thomas G. Bever

References Bever, Thomas G. 1975. Psychologically real grammar emerges because of its role in language acquisition. In Developmental Psycholinguistics: Theory and Applications, Daniel P. Dato (ed.), 63‑75. Washington DC: Georgetown University Round Table on Languages and Linguistics. Bever, Thomas G. 1992. The logical and extrinsic sources of modularity. In Modularity and Constraints in Language and Cognition [Minnesota Symposia on Child Psychology 25], Megan Gunnar & Michael Maratsos (eds.). Hillsdale NJ: Lawrence Erlbaum Associates. Bever, Thomas G. 2008. Remarks on the individual basis for language universals. In Of Minds and Language: The Basque Country Encounter with Noam Chomsky, Massimo Piatelli-Palmarini, Pello Salaburu & Juan Uriagereka (eds), 275–296. Oxford: OUP. Bever, Thomas G. & Hansen, Ralph E. 1988. The induction of mental structures while learning to use symbolic systems. Proceedings of the 10th annual meeting of the Cognitive Science Society: 132–139. Hillsdale NJ: Lawrence Erlbaum Associates. Dahl, Östen. 1979. Typology of sentence negation. Linguistics 17: 79–106. Dulany, Don, Carlson, Richard & Dewey, Gerald. 1984. A case of syntactical learning and judgment: How conscious and how abstract? Journal of Experimental Psychology 113: 541–555. Friederici, Angela, Steinhauer, Karsten & Pfeifer, Edmund. 2002. Brain signatures of artificial language processing: Evidence challenging the critical period hypothesis. Proceedings of the National Academy of Sciences 99(1): 529–534. Friederici, Angela D., Bahlmann, Jörg, Heim, Stefan, Schubotz, Ricard I. & Anwander, Alfred. 2006. The brain differentiates human and non-human grammars: Functional localization and structural connectivity. Proceedings of the National Academy of Sciences 103: 2458–63. Gelderen, Elly van. 2004. Grammaticalization as Economy [Linguistik Aktuell-Linguistics Today 71]. Amsterdam: John Benjamins. Gelderen, Elly van. 2008. Linguistic cycles and economy principles: The role of Universal Grammar in language change. In Grammatical Change and Linguistic Theory: The Rosendal Papers [Linguistik Aktuell-Linguistics Today 113], Thórhallur Eythórsson (ed.), 245–264. Amsterdam: John Benjamins. Gómez, Rebecca & Gerken, LouAnn. 1999. Artificial grammar learning by 1-year-olds leads to specific and abstract knowledge. Cognition 70: 109–135. Hudson Kam, Carla L. & Newport, Elissa. 2005. Regularizing unpredictable variation: The roles of adult and child learners in language formation and change. Language Learning and Development 1: 151–195. Jespersen, Otto. 1917. Negation in English and Other Languages. København: A. F. Høst & Søn. Kemler Nelson, Deborah G., Jusczyk, Peter W., Mandel, Denise, Myers, James, Turk, Alice & Gerken, LouAnn. 1995. The Head-turn preference procedure for testing auditory perception. Infant Behavior and Development 18: 111–116. Kirby, Simon, Cornish, Hannah & Smith, Kenny. 2008. Cumulative cultural evolution in the laboratory: An experimental approach to the origins of structure in human language. Proceedings of the National Academy of Sciences 105: 10681–10686. Morgan, Kara, Sanz, Christina, Steinhauer, Karsten & Ullman, Michael. 2009. Acquisition of gender agreement in second language learners: An event-related potential study. Ms. Perruchet, Pierre, & Pacteau, Chantal. 1990. Synthetic grammar learning: Implicit rule abstraction or explicit fragmentary knowledge. Journal of Experimental Psychology: General 119: 264–275.



Chapter 13.  The study of syntactic cycles as an experimental science  Reber, Arthur. 1969. Transfer of syntactic structure in synthetic languages. Journal of Experimental Psychology 81: 115–119. Sanz, Montsarrat. 2000. Events and predication. A New Approach to Syntactic Processing in English and Spanish [Current Issues in Linguistic Theory 207]. Amsterdam: John Benjamins. Schröder, Marc & Trouvain, Jürgen. 2003. The German text-to-speech synthesis system MARY: A tool for research, development and teaching. International Journal of Speech Technology 6: 365–377. Tenny, Carol. 1994. Aspectual Roles and the Syntax-Semantics Interface. Dordrecht: Kluwer.

Author index A Adams, Marianne  186, 204 Aelbrecht, Lobke  102, 126 Åfarli, Tor  210-2, 229, 239-41 Aissen, Judith  281, 283 Alexiadou, Artemis  166-7, 170, 181, 183 Alighieri, Dante  206 Anagnostopoulou, Elena  181 Andersen, Henning  10, 157-8, 165-6, 172, 181 Anderson, John M.  71, 241, 291, 299 Anderwald, Lieselotte  37, 66 B Bailyn, John F.  170, 181 Baker, Mark  4, 10, 210-1, 239 Bański, Pjotr  175, 182 Barbiers, Sjef  12, 34, 36, 49, 62, 66-7, 126, 239 Barbosa, Pilar P.  166-7, 181 Beck, Sigrid  251, 263 Bell, Arthur  93, 113, 125-6 Belletti, Adriana  89, 128, 152, 154, 204, 240 Benincà, Paola  145, 154, 156, 186, 189, 192, 195-6, 198, 201-2, 204-6 Benveniste, Émile  177, 181 Bernini, Giuliano  16, 32, 37, 50, 51, 54, 66 Besten, Hans den  76, 99, 116 Beths, Frank  241, 264 Beukema, Frits  66, 239 Biberauer, Theresa  4, 6, 49, 51-2, 67, 91, 93-4, 99-100, 104, 112-4, 116-9, 121, 125-7, 229, 239, 242 Blancquaert, E.  50, 62, 67 Błaszczak, Johanna  124, 127 Bohnemeyer, Jürgen  274, 283 Bonet, Eulàlia  177, 181 Bopp, Franz  2, 10 Borer, Hagit  167, 181

Borkovskij, Viktor Ivanovič  81, 87, 89, 157, 171, 181 Boström, Ingmar  140, 149, 155 Bowers, John  210, 211, 239 Boya, Li  189, 205 Breitbarth, Anne  16, 32, 41, 46, 59, 61, 65, 67, 119, 121, 125-7 Bricker, Victoria R.  270, 273, 276, 280-1, 283 Brinton, Laurel  7, 10 Brody, Michael  119, 127 Brooke, Henry  299 Brown, Sue  82, 85, 86, 89 Buridant, Claude  40, 67 Burridge, Kate  38, 67 Bush, John  299 C Calabrese, Andrea  135, 155 Cardinaletti, Anna  96, 127, 133, 135, 141-2, 144, 148-50, 155, 162, 181 Caubet, Dominique  43, 67 Cavalcante, Rerrison  125, 127 Caz Cho, Sergio  276, 277, 283 Chaker, Salem  43, 67 Chomsky, Noam  1, 8, 11, 75, 89, 127, 173, 181, 213, 214, 216, 218, 236, 239-40, 291, 299, 320 Choo, Sukhoon  159, 181 Church, Kenneth  26, 32, 80, 85, 89, 171, 182 Citko, Barbara  164, 181 Claudi, Ulrike  4, 6, 11, 212, 241, 286, 299 Condillac, Etienne Bonnot de  2, 11 Condoravdi, Cleo  16, 19, 33, 42, 69, 96, 97, 113, 127, 128 Corver, Norbert  248, 263 Croft, William  36, 53, 54, 67 Crowley, Terry  56, 67 Cruschina, Silvio  198, 202, 205 Culicover, Peter W.  246, 263 Curiel, Alejandro  271, 283

Curnow, Timothy J.  56, 67 D Dahl, Östen  37, 57, 67, 97, 127, 306, 320 Dasher, Richard  3, 7, 12, 250, 264 Dayley, Jon P.  271, 276, 283 De Cuypere, Ludovic  54, 55, 67 De Vogelaer,Gunther  12, 34 Delfs, Laurel  247, 264 Denison, David  9, 11 Deprez, Viviane  16, 32, 33 Detges, Ulrich  64, 67 Devos, Maud  66, 67 Diesing, Molly  123, 127 Diessel, Holger  164, 165, 182, 219, 240 Dieterich, Thomas J.  246, 263 Donaldson, Bruce  93, 104, 127 Donhauser, Karin  38, 67 Duffield, Nigel  125, 127 E Early, Robert  4, 11, 36, 55, 57, 63, 67 Eckardt, Regine  46, 67, 68, 250, 254, 263 Egerland, Verner  134, 143-4, 149, 150, 155 Eide, Kristin M.  211-2, 240 Embick, David  167, 182 Emonds, Joseph  246, 263 England Nora C.  268, 270-3, 283 Espinal, Maria Teresa  127-8 F Falaus, Anamaria  122, 128 Fernández, Soriano  155, 205 Fintel, Kai von  246, 252, 253, 258 Fischer, Olga  67-8, 181-2 Fleischer, Jürg  10-11 Flier, Michael S.  172, 182 Floricic, Franck  49, 68, 128 Fornaciari, Raffaello  151, 155 Fox, Danny  97, 128

 Cyclical Change Frajzyngier, Zygmunt  182, 235, 240 François, Alexandre  54, 55, 56, 68 Franks, Steven  159, 175, 182 Frascarelli, Mara  155, 205 Furbee-Losee, Louanna  274, 283 G Garbacz, Georg von der  122, 128 Gardiner, Alan H.  2, 16, 33, 35, 42, 68 Gärtner, Hans-Martin  89, 124, 127, 240 Garzonio, Jacopo  190, 205 Gass, Kate van  107, 108, 109 Gelderen, Elly van  6, 8-9, 16, 29-30, 32, 36, 66, 73, 79, 120-1, 134, 142, 157, 159, 173-5, 180, 185, 210-6, 228, 236, 260-1, 304-7, 314, 319 Gergel, Remus  5, 6, 243, 246, 260, 263 Giannakidou, Anastasia  33, 76, 90, 100, 106, 118, 128, 246, 263 Gildea, Spike  224-5, 240 Giusti, Giuliana  148, 155 Givón, Talmy  2, 7, 11, 73, 90, 174, 182 Goldsmith, Oliver  300 Gorškova, Ol’ga V.  84, 90 Greenberg, Jospeh  2, 11 Grosu, Alexander  222, 239, 248, 263 H Haegeman, Liliane  29, 33, 46, 59, 65, 67-8, 90, 107, 117, 119, 121, 125, 127-8, 183 Hagège, Claude  9, 11 Hanks, Patrick  26, 32 Hansen, Britt M.  47, 68 Hansen, Ralph  308, 311, 314, 320 Harris, Alice C.  40, 68, 181 Haspelmath, Martin  68, 165, 182 Heim, Irene  244, 250, 251, 252, 253, 254, 258, 263, 320 Heine, Bernd  2, 4, 6, 7, 9, 11, 12, 209, 212, 215, 233, 236, 241, 264, 286, 299 Herburger, Elena  128, 154, 205 Hicks, Glyn  8, 11 Hiraiwa, Ken  75, 90 Hock, Hans Henrich  43, 68

Hodge, Carleton  2, 4, 11, 73, 90 Hoeksema, Jack  4, 5, 15, 16, 20, 27, 28, 30, 33, 38, 48, 56, 68, 76, 90 Hofling, Charles Andrew  279, 280, 283 Holmberg, Anders  135, 155, 166, 182 Honda, Isao  54, 68 Hopper, Paul  3, 8, 11, 67, 68, 127, 212, 241, 282, 283, 286, 287, 291, 293, 299 Horn, Lawrence R.  20, 33, 34, 48, 59, 68, 70, 87, 90, 110, 120, 126, 128, 130 Huang, C.-T. James  166, 182 Huddleston, Rodney  245, 263 Humboldt, Wilhelm von  2, 11 Hünnemeyer, Friederike  4, 6, 11, 212, 241, 286, 299 I Isačenko, Alexander V.  157, 182 J Jäger, Agnes  16, 22, 27, 33, 68, 113, 117, 128 Jakobson, Roman  165, 182 Jelinek, Eloise  168, 176, 182 Jespersen, Otto  2, 6, 9, 11, 15, 16, 18-21, 26-7, 29, 31-7, 39-42, 467, 49-50, 53-4, 56, 60, 64-71, 74, 90-2, 96, 104, 126-8, 130, 247, 263, 306, 320 Joly, André  45, 68 K Kaiser, Georg  204, 205 Kari, James  6, 11 Kasombo Tshibandi, Michael  67 Katz, Aya  5, 11, 164-6, 182, 222-3, 241 Kaufman, Terrence  266, 268, 270-1, 273, 281, 283 Kawaguchi, Yuji  44, 68 Kayne, Richard  120, 128, 135-6, 155, 178, 182, 190, 205 Keller, Kathryn C.  277, 278, 283 Kemenade, Ans van  16, 68, 117 Kiparsky, Paul  16, 19, 33, 42, 69, 96-7, 113, 128 König, Ekkehard  246, 264 Kornfilt, Jaklin  237-8, 241 Kouwenberg, Sylvia  51, 69

Kratzer, Angelika  241, 244, 250, 252-3, 263-4 Kroch, Anthony  180, 182, 214, 241, 247, 253, 264 Kroskrity, Paul V.  54, 69 Kupść, Anna  77, 90 Kuteva, Tania  7, 9, 11, 209, 212, 215, 232-3, 236, 241 Kuznecov, Petr Savvič  157, 171, 181 L Labov, William  15, 27, 33 Ladusaw, William A.  15, 33, 34, 76, 90 Laka, Itziar  76, 90, 106, 119-20, 128, 205 Lambrecht, Knud  4, 11 Larrivée, Pierre  69 Larsen, Thomas W.  281, 283 Lass, Roger  177, 182 Le Feuvre, Claire  161, 182 Ledgeway, Adam  154, 195, 205 Lehmann, Christian  2, 11, 212, 241 Lenz, Barbara  37, 69 Levy, Yonata  177, 183 Li, Charles N.  164, 218, 219, 220, 222, 223, 224, 241 Lightfoot, David W.  3, 11, 214, 220, 241 Lindseth, Martina  159, 182 Lithgow, William  300 Lohndal, Terje  5, 6, 10, 90, 107, 121, 128, 174, 209, 210, 211, 241 López Ramírez, Ana  273 Lucas, Christopher  16, 22, 27, 34, 37, 54, 69, 121, 127 Lunt, Horace G.  171, 182 Lyons, Christopher  3, 11 M Malchic, Nicolás  276, 283 Manning, Christopher D.  26, 34 Marchello-Nizia, Christiane  69 Marsden, William  300 Martineau, France  33, 39, 44, 46, 68, 69, 90 Martins, Ana Maria  178, 183 Mateo, Pedro  272, 273, 283 May, Robert  2, 108, 109, 250, 264 McCormick, Kay  94, 108, 128 McFadden, Thomas  167, 169, 183 McQuown, Norman A.  280, 283



Author index  Meillet, Antoine  16, 34-5, 42, 69 Mel’ničenko, G.G.  83, 90 Merchant, Jason  246, 264 Miestamo, Matti  54-7, 65, 69, 97, 128 Mitchell, Bruce  247, 264 Mó Isém, Romelia  283 Mohr, Sabine  116, 129 Möhren, Frankwalt  44, 69 Molinu, Ludia  49, 68 Molnárfi, Laszlo  93, 107, 127, 129 Morgan, William  7, 12, 308, 320 Mougeon, Raymond  39, 44, 46, 69 Muller, Claude  37, 44-6, 69 Müller, Gereon  166, 179, 183 Munro, Pamela  231, 242 N Napoli, Donna J.  246, 263 Nater, H.F.  55, 69 Neuckermans, Annemie  36, 4950, 59-60, 62, 69, 70-1, 105, 126 Newmeyer, Frederick  3, 12 Norde, Muriel  165, 182-3 Nørgård-Sørensen, Jens  171, 183 Norman, William M.  281, 283 Nunes, Jairo  178, 183 O Obenauer, Hans-Georg  192, 193, 205 Oosthuizen, Johan  93, 113, 118, 129 Ouhalla, Jamal  29, 34 Overdiep, G.S.  60, 61, 62, 65, 69 P Paesani, Kate  115, 129 Parry, M. Mair  66, 69 Paul, Ileana  115, 129 Pauwels, J.L.  49, 50, 51, 52, 62, 70, 129 Penka, Doris  117, 129 Perle, Friedrich  42, 70 Pernicone, Vincenzo  147 Pesetsky, David  236, 242, 245, 264 Petrarca, Francesco  147 Petruxin, Pavel Vladimirovič  183 Pintzuk, Susan  33, 128, 264 Plácido, Luiciano G.  283 Platzack, Christer  167, 168, 183 Plot, Robert  300

Pohl, Jacques  70 Poletto, Cecilia  5, 113, 117, 120, 125, 129, 155, 185, 188-9, 195, 198-02, 205 Polian, Gilles  270, 284 Pollock, Jean-Yves  29, 34, 170, 183 Ponelis, Frederick  93, 101, 129 Portner, Paul  154, 193, 205, 263 Postal, Paul M.  20, 34, 48, 70 Postma, Gertjan  16, 34, 70 Press, Ian  80, 90 Przepiórkowski, Adam  77, 90 Pullum, Geoffrey K.  176, 184, 245, 263 Pustet, Regina  209, 216, 225-6, 232, 237, 238, 242 Pye, Clifton  5, 265, 276 R Radford, Andrew  8, 12, 216, 242 Ramat, Paolo  16, 32, 37, 49-51, 54, 66, 70, 181 Ramchand, Gillian  121, 129, 211, 239 Rechzieglova, Adela  77, 90 Reinhart, Tanya  97, 129 Renzi, Lorenzo  143, 154, 155, 156, 205, 206 Ringe, Donald  264 Rissanen, Matti  247, 264 Rizzi, Luigi  75, 90, 116, 128, 143, 155, 156, 166, 183, 205, 299 Robbers, Karen  93, 99, 129 Roberge, Paul T.  49, 51, 70, 113, 129 Roberts, Ian  16, 34, 54, 70, 113-4, 117, 120-1, 125, 127, 129, 155-6, 166, 178-9, 183, 202, 205, 209, 212, 214, 242, 253, 259-60, 264, 286, 299 Robertson, John S.  270, 277-8, 280, 282-4 Rodrigues, Cilene  178, 183 Rohlfs, Gerhard  156 Rooryck, Johan  117, 123, 129, 241 Roussou, Anna  16, 34, 54, 70, 113, 117, 120, 129, 156, 209, 212, 242, 261, 264, 286, 299 Rowlett, Paul  16, 34, 117, 129 Rutkowski, Pawel  164, 183 S Salvi, Ivan A.  143, 154-6

Sántiz, Roberto  270, 284 Santorini, Beatrice  247, 264 Schøsler, Lene  70 Schwegler, Armin  40, 49-50, 52, 70 Schwenter, Scott A.  17, 19, 34, 49, 70, 97, 108, 110, 129, 130 Ševelëva, Marija Naumovna  159, 161, 163, 183 Siewierska, Anna  176, 183 Sigurđsson, Halldór Ármann  169, 183 Sims, Lynn  7, 12 Smith, John Charles  89, 126, 177, 183, 308, 320 Starke, Michal  96, 127, 133, 135, 141-2, 148-9, 155, 162, 181 Stassen, Leon  209, 216-8, 237, 242 Stavrou, Melita  246, 263 Stechow, Arnim von  251, 252, 254, 262, 263 Stender-Petersen, Adolf  80, 90 Stern, Gustaf  247, 264 Stowell, Timothy  115, 130 Svenonius, Peter  285, 286, 288, 289, 290, 291, 293, 295, 298, 299 T Tamburelli, Marco  166, 183 Taraldsen, Knut Tarald  156, 166, 168, 179, 183 Tauli, Valter  2, 12 Taylor, Ann  247, 264 Terzi, Arhonto  168, 183 Thompson, Sandra A.  164, 182, 218-20, 222-4, 241 Timberlake, Alan  166, 183 Tooke, John Horne  2, 12 Torrego, Esther  236, 242, 245, 264 Traugott, Elizabeth Closs  2, 3, 7, 8, 11-2, 68, 181, 212, 214, 241-2, 246, 250, 264, 282-3, 286-7, 291, 293, 299 Travis, Lisa  30, 34 Tul Rax, Augusto  283 U Ura, Hiroyuki  75, 90 V Vainikka, Anne  177, 183 Vallduví, Enric  76, 77, 87, 90

 Cyclical Change van der Auwera, Johan  2, 4, 6, 10, 12, 16, 34, 36-8, 63, 66-7, 70-1, 97, 105, 112, 126 Van der Horst, Joop  30, 33, 40, 61, 71 Van der Horst, Kees  30, 33 Vanelli, Laura  133, 134, 143, 146, 147, 156, 186, 206 Varchi, Benedetto  150, 156 Vázquez Alvarez, Juan Jesús  275, 284 Vedovato, Diana  5, 9, 133, 156 Vennemann, Theo  40, 71 Vikner, Sten  116, 130, 183 Villalta, Elisabeth  254, 264 Vinet, Marie-Thérèse  44, 69, 115, 130 Visconti, Jacqueline  68 W Wackernagel, Jacob  71

Waher, Hester  93, 130 Wal, Marijke van der  40, 71 Wallage, Phillip  46, 61, 71, 117, 130 Waltereit, Richard  64, 67 Watanabe, Akira  100, 130 Wauchope, Robert  283 Weiss, Helmuth  113, 130 Willis, David  16, 22, 34, 54, 71, 104, 121, 125, 127, 130 Winand, Jean  71 Winkler, Susanne  246, 263, 264 Wurff, Wim van der  45, 66, 68, 71, 239 Wurmbrand, Susanne  113, 115, 130 X Xaburgaev, Georgij Aleksandrovič  159, 161, 163, 184

Xmelevskaja, Tatiana Aleksandrovna  84, 90 Y Yáñez-Bouza, Nuria  10, 12 Young, Robert  7, 12 Z Zaliznjak, Andrej Anatol’evič  159, 160, 161, 162, 171, 177, 178, 184 Zanuttini, Raffaella  29, 33, 34, 76, 77, 90, 117, 128, 130, 154, 193, 205 Zeijlstra, Hedde  16, 34, 37, 71, 74, 75, 78, 79, 85, 86, 89, 90, 101, 104, 113, 121, 127, 130 Zribi-Hertz, Anne  4, 5, 12 Zwarts, Frans  34 Zwicky, Arnold  176, 184

Subject index A absolutive  140, 223, 266-8, 271-2, 277, 282 actualization  181, 183, 282 adverb  3, 5-7, 27, 29-31, 36-9, 42, 63, 74, 96, 101, 122, 143, 170, 193-8, 201, 203-4, 215-6, 227, 246-7, 252, 260-2, 267, 274, 296 adverbial  4, 5, 15, 23, 27-32, 42, 74, 118, 187, 193, 214, 227, 258, 273, 275, 286, 296 Afrikaans  4, 49, 51, 67, 70, 89, 91-130 Agree  75, 78, 86, 236 agreement  1, 3, 4, 5, 7, 8, 31, 90, 78-9131, 134, 142, 166-79, 210, 231, 234, 237, 224, 265, 269, 286, 304, 319 Alyawarra  229 analytic  4, 73, 162 anti-grammaticalization  165 aorist  157-8, 160, 162, 169, 171 Arabic  16, 22, 43, 54, 69, 164, 222, 244 Arizona Tewa  54, 69 artificial language learning  30311 aspect  2, 3, 5, 7, 56, 73, 157-8, 171, 183, 213, 214, 224, 225, 233, 265-71, 273-6, 274-310, 317-9 Athabaskan (Athabascan)  6, 16 auxiliary  3, 5, 6, 7, 37, 157, 160-84, 201-2, 204, 213-5, 225, 229-34, 243, 249 Avava  56 Awa Pit  56 B Bantu  66, BE  157-84 Bella Coola  55 Berber  16, 43, 217 Berbice (Creole) Dutch  51 bleaching  41, 43, 47, 50, 53, 54, 56, 66, 97, 119-20, 246, 285-286

Bolewa  235 bound  251, 261, 274 Brabantic  38, 49-52, 62, 64, 65, 70, 126 Brazilian Portuguese  47, 49-52, 70, 125, 129, 178-9, 183 broken cycle  5, 133- 4, 147, 154, 185, 201 C Cambodian  217 Catalan  17, 22, 47, 78, 85-7, 90, 127 Chalcatongo Mixtec  228, 241 Chantyal  230 Chinese  4, 6, 16, 189, 205, 21922, 224, 244, 316 Chontal  277- 8, 280, 283 cline  3, 8, 133, 143, 174, 295 clitic  4, 22, 30, 31, 41, 69, 74, 133-5, 139, 140, 142-3, 148-51, 153, 161, 162, 168-9, 173-6, 180, 193, 196, 266, 272, 303, 305, 306, 319 clitic cluster  150-1 Cocopa  217 Colloquial Afrikaans  94, 100, 124 Colloquial French  5, 93, 112, 135, 173 Colloquial Italian  138, 140, 141, 153 Colonial Yucatec  270, 280 Common Slavic/Slavonic (see also Late Common Slavic)  80, 157 comparatives  22, 25, 101-2, 246, 250-60, 264 complement  22, 24, 29, 32, 61, 63, 113-5, 135, 136, 143, 154, 164, 187, 190, 210, 215, 233, 244, 245, 257-8, 267-85, 289, 291, 310 complementation, 245, 256, 265-83

complementizer  7, 120, 128, 194, 199-201, 214, 221, 236 conjunction  85, 158, 187-8, 190, 203-4 Coptic  42-3, 69 copula  1, 2, 3, 5, 7, 35, 57, 158-9, 163-6, 180, 209-42, 269, 319 CP  113, 116-9, 125, 145-6, 155, 179, 185-7, 190-205, 214, 216, 221, 229, 258 Czech  76-9, 90 D definite(ness)  3, 7, 11, 22, 240, 268 degrammaticalization  165, 183 degrammation  165, 181 demonstrative  2, 3, 4, 5, 7, 139, 141, 143, 164-5, 174, 209, 21829, 236, 239, 240 dependent status  272-5, 282 double negation  15-6, 29, 31, 59, 65, 68, 74, 76-7, 92, 108-9, 128, 305-6 doubling  18, 36-7, 47-9, 51-3, 58-9, 60-3, 64-6, 129, 130, 152, 154, 161, 190 drecative  20 Dutch  4, 6, 10, 16-7, 18, 20-2, 23, 24, 28, 30, 31-3, 34, 36, 38, 44, 46, 49, 51-2, 60-2, 66-7, 92-3, 96, 105, 112-3, 124, 127, 130 E economy  3, 5, 8, 46-7, 68, 79, 89, 128, 133-4, 141, 174, 210, 212-4, 216, 218, 220, 236-7, 243, 244, 255, 261-2, 263, 305-6, 315-6, 320 emphasis  35, 42, 45-6, 50, 52-4, 56, 58-9, 63, 65, 96, 101, 104, 108, 122, 179, 227, 268 emphatic  3-5, 7-8, 17, 19, 22, 419, 50-60, 63-6, 87, 95-7, 103-4,

 Cyclical Change 106, 108-9, 112-3, 122, 160-1, 174, 177 emphatic negation  42-3, 49, 52-3, 59, 64, 87, 97 EPP  167, 170, 181 ergative  217, 265-8, 274, 277-9, 281, 283 Erzya  230 Ewe  226-7, 236 exaptation  177, 182 exclamative  186, 191-3 F feature  3, 5-8, 15, 19, 21, 42, 73, 75, 78-80, 84-7, 116, 119-20, 125, 133, 141, 157, 159, 161, 16672, 179-80, 213, 216, 218, 232, 236, 239, 295, 305 feature checking  75, 86, 127, 175, 305 Feature Economy  3, 5, 8, 79, 213, 216, 218 feature loss  87 Flemish  59, 64, 66, 68, 107, 119, 125 focus  5, 30-2, 100, 111, 115-6, 1356, 185-6, 189, 192-3, 195-205, 230, 237 focus marker  5, 186, 198-201, 204, 237 French  4-5, 7, 9, 10, 16, 19, 20, 22, 29-31, 36-7, 39-42, 44-8, 53, 56, 60, 64, 93, 96, 105, 107, 112-3, 123-5, 143, 144, 173, 179-80, 195, 202, 215, 244 future  3, 7, 157-8, 162, 274, 278 G Gbeya  218, 234 German  10, 16, 20, 22-4, 28, 302, 42, 61, 105, 113-4, 124, 143, 168, 195-6, 244, 251, 259 Germanic  7, 22, 31, 43, 49, 92, 107, 118, 125, 143, 188, 306 Goal  75 Gola  218 grammarians (role of)  9, 56, 147, 154 grammaticality intuitions  140, 303-304, 307-10, 314-5, 319 grammaticalization  2-3, 5-8, 22, 30, 42, 46, 53, 57, 68, 70, 74, 133, 134, 141, 143, 154, 165-6, 187, 209-10, 212-3, 215-6, 218-9,

225, 235-6, 239, 245-6, 256, 267, 274-6, 282, 286, 293-6, 298-9, 314, 316 H hanging topic  188-90, 192, 195, 197 Head Movement Constraint  30 Head Preference Principle  2932, 121, 134, 142, 174, 213, 236, 261, 305 Hebrew  5, 164, 166, 183, 209, 222-3, 239 Huastec  268, 278, 282 I imperfect  157-60, 169 Imonda  233 indicative  17, 269, 271, 274, 282 Inertia Principle  31 interpretable  7-8, 75, 78, 79, 85, 87, 166-9, 180, 191, 216, 239, 252-3 interrogative  9, 17, 28, 44, 76, 116, 186, 191, 193, 200 intransitivization  281 Italian  5, 17, 49, 66, 68, 69, 70, 76-7, 78, 79, 85, 86, 120, 125, 129, 133-156, 183, 185-206, 232 J Jespersen Cycle  6, 15-6, 18-21, 26-7, 29, 31-2, 35-7, 41, 43, 46, 50, 53-4, 56, 60, 64-6, 74, 91-2, 96, 306 K Kenya Luo  225 Kiowa  54, 217 Kui  233 L Lahu  217 Lango  225 Late Common Slavic  157-61, 166, 169-70 Late Merge Principle  174, 213, 232-3, 261, 305 Latin  16, 37, 42, 134, 143, 150, 232 Lewo  4, 36, 55-7, 63-4, 67 Lisu  217 l-less perfect (bezèlevyj perfekt)  170-5 Loc, 285  290-8

loss of V2 (Verb Second)  170, 185-7, 198, 201-2 Lower Tanana  6 M Mam  270-3, 282-3 Mayan  265-83 Merge  22, 168, 213-4, 229, 231-4, 239, 252-3, 260-2, 305, 319 metalinguistic comparatives  246 Middle Dutch  27, 31, 61, 96 Middle English  46, 61, 65, 214, 247-8, 256 minimizer  7, 20-7, 29, 44, 48, 53, 54-6, 74, 79, 84, 87-8 modal  1, 5-7, 23, 176-7, 181, 202, 204, 214-5, 229, 243-62, 266, 275 modal restrictor  259 Modern Yucatec  267-8, 280-1 Mojave  217, 231 Motlav  55-6 movement  30, 32, 40, 116, 123-4, 134, 145, 167, 186, 189, 199, 201-3, 213, 232, 244, 250-1, 255, 260-1, movement verb  268-73, 275-6, 282 Multiple Agree  75, 86, 90 N Navajo  7, 12 negative cycle (see also Jespersen Cycle)  1-2, 4, 6, 35, 37, 73-4, 79, 306, 315, 319 negative marker  4, 6, 17, 36, 55, 64, 73-89, 94, 200, 214, 306 negative operator  78-9, 86, 124 negative polarity  15, 17, 19, 27-8, 32, 36, 44-6, 54, 60-5, 76, 105 negative quantifier  4, 21, 24-9, 32, 36, 51, 53-5, 65, 76, 92-3, 98, 100, 103-9, 113, 120-4, 200 NegP  15, 29-32, 79, 85-7, 118, 126, 306 nominalized  271-4, 279-82 non-stative (aspect)  269 North Carolina Cherokee  217 Northern Italian Dialects  143-6 Norwegian  31, 211-2 Novgorod  82-3, 158-9 n-word  6, 22-4, 27-8, 31, 73-4, 76-89, 106-7, 120, 123



Subject index  O of-insertion  291-3 Old Church Slavonic  80, 85, 171 Old English  8, 71, 74, 93, 112, 247, 256, 260 Old Italian  143-4, 149-50, 185-91, 193, 196, 198, 201-4 Old Senese  143, 150 ordering  121, 124, 180, 244, 254, 260, 262, 316 P Paamese  56 Palestinian Arabic  222 Panare  223-5 parameter  35, 134, 159, 166-70, 176, 178, 180 parsing  164-5, 173 perfect  157-8, 160-2, 165-72, 175, 273 Person-Case Constraint  177 phi-features  7-8, 75, 145, 166-9, 175, 179-80, 236 Polish  77, 164, 175 Poqomam  276 Portuguese  9, 47, 49-52, 125, 178-9 pragmatic overload  254, 260 preposition  5-7, 9-10, 52, 158, 167, 214, 218, 234-7, 285-99 preposition cycle, 285, 296-9 preposition stranding  10 prescriptive grammar/rules  5, 9-10, 24, 25, 100-101, 133-4, 147-8, 154 present tense  170, 222, 225 probe  3, 7, 8, 75, 116, 216, 236 progressive  232-3, 266, 271-6, 281 prohibitive  4, 32, 38, 63 projective preposition  288-94 pronoun  1-5, 7-8, 22-3, 36, 48, 54, 63, 133-54, 157-81, 191, 195, 200-1, 209, 213, 218, 219, 2206, 239, 269 Q quantifier raising (QR)  243, 250

R reanalysis  5-6, 21, 27, 29-30, 32, 44, 55, 58-9, 64-5, 74, 88-9, 97, 105, 120-1, 125, 157-70, 177-80, 185-6, 203-4, 220, 263, 279, 285, 286-7, 289, 293-8 renewal  1, 3, 6-8, 15, 18, 35-6, 46, 236, 267, 279, 285-6, 291, 293, 298 resumptive pronoun  164, 227 Romance  7, 46, 49, 118, 125, 134, 136, 143-5, 160, 177, 188, 198-9, 202, 282 Russian, 4-5, 73-89, 157-81 S Saramaccan  227, 228 school teaching  147 Scottish Gaelic  164 semantic negation  76, 82 Shilha  217 Shona  218 Sicilian  198, 201, 202, 205 situations  252-5 Slavic  46, 74, 76, 80, 89, 157, 178 Spanish  49, 76-7, 152, 232, 271, 317 specifier  8-9, 29, 75, 116, 118, 121, 134, 187, 190, 202, 204, 210, 216, 218, 221, 228-9, 237, 239, 261-2 spiral  2, 42, 174 Sranan  226-7 standard Italian  5, 133, 147, 154 stative  226, 267-9, 273-4 subject  2-4, 6, 8, 25, 37, 57, 63, 73, 77, 85, 87, 103-4, 107, 116, 133-5, 137, 139, 143-7, 149, 152, 157-81, 191, 196, 201-2, 210-11, 220-5, 227, 231-2, 245-7, 257, 262, 265, 268, 274, 281 subject clitics  145, 155-6 subject cycle  8, 73, 143, 157, 159, 170-6 syntactic agreement  78-9, 86, 89 synthetic  4, 73 T taboo  20-2, 24-5, 27, 32, 36, 48

Tamazight  217 Tariana  9 Temne  218 tense  62, 157-8, 160-2, 165, 16772, 175-6,180, 210, 212, 222, 224-5, 229, 234, 236-7, 254, 274, 282, 304 than-clause  249, 252, 262 topic  5, 116, 131, 135, 154, 164, 166, 185-99, 203-4, 212-3, 220, 223, 227, 250 topic marker  131, 185, 187-8, 190-3, 203-4 tripling  57-8, 62-6 Turkish  5, 164-6, 223, 237-8 Tzeltal  270, 272, 277 Tzutujil  276 U unidirectional  3, 16, 139, 255, 262, 315 uninterpretable  3, 8, 75, 78, 85, 166, 168-70, 180, 216, 239 V verbal complex  265-7, 281-283 Verb-second (V2)  115, 179, 186 V-to-C movement  116, 134, 145 W Wappo  222 weak pronouns  5, 133-5, 137-8, 143, 147, 149, 179 weakening  39-42, 44, 48, 59, 74, 79, 96-97, 112, 142-3, 146-8, 150, 154, 305-6, 319 Welsh  16, 22, 54 Western Poqomchi’  271 word order  30, 40, 59, 71, 105, 224, 268, 316 Y Yavapai  217 Yoruba  237 Yucatec  267, 272-4, 276, 279-82 Z Zande  218 Zulu  218

Linguistik Aktuell/Linguistics Today A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 148 Hogeweg, Lotte, Helen de Hoop and Andrej Malchukov (eds.): Cross-linguistic Semantics of Tense, Aspect and Modality. Expected October 2009 147 Gomeshi, Jila, Ileana Paul and Martina Wiltschko (eds.): Determiners. Universals and variation. vii, 241 pp. + index. Expected October 2009 146 Gelderen, Elly van (ed.): Cyclical Change. 2009. viii, 329 pp. 145 Westergaard, Marit: The Acquisition of Word Order. Micro-cues, information structure, and economy. xii, 242 pp. + index. Expected August 2009 144 Putnam, Michael T. (ed.): Towards a Derivational Syntax. Survive-minimalism. 2009. x, 269 pp. 143 Rothmayr, Antonia: The Structure of Stative Verbs. 2009. xv, 216 pp. 142 Nunes, Jairo (ed.): Minimalist Essays on Brazilian Portuguese Syntax. 2009. vi, 243 pp. 141 Alexiadou, Artemis, Jorge Hankamer, Thomas McFadden, Justin Nuger and Florian Schäfer (eds.): Advances in Comparative Germanic Syntax. 2009. xv, 395 pp. 140 Roehrs, Dorian: Demonstratives and Definite Articles as Nominal Auxiliaries. 2009. xii, 196 pp. 139 Hicks, Glyn: The Derivation of Anaphoric Relations. 2009. xii, 309 pp. 138 Siddiqi, Daniel: Syntax within the Word. Economy, allomorphy, and argument selection in Distributed Morphology. 2009. xii, 138 pp. 137 Pfau, Roland: Grammar as Processor. A Distributed Morphology account of spontaneous speech errors. 2009. xiii, 372 pp. 136 Kandybowicz, Jason: The Grammar of Repetition. Nupe grammar at the syntax–phonology interface. 2008. xiii, 168 pp. 135 Lewis, William D., Simin Karimi, Heidi Harley and Scott O. Farrar (eds.): Time and Again. Theoretical perspectives on formal linguistics. In honor of D. Terence Langendoen. 2009. xiv, 265 pp. 134 Armon-Lotem, Sharon, Gabi Danon and Susan Rothstein (eds.): Current Issues in Generative Hebrew Linguistics. 2008. vii, 393 pp. 133 MacDonald, Jonathan E.: The Syntactic Nature of Inner Aspect. A minimalist perspective. 2008. xv, 241 pp. 132 Biberauer, Theresa (ed.): The Limits of Syntactic Variation. 2008. vii, 521 pp. 131 De Cat, Cécile and Katherine Demuth (eds.): The Bantu–Romance Connection. A comparative investigation of verbal agreement, DPs, and information structure. 2008. xix, 355 pp. 130 Kallulli, Dalina and Liliane Tasmowski (eds.): Clitic Doubling in the Balkan Languages. 2008. ix, 442 pp. 129 Sturgeon, Anne: The Left Periphery. The interaction of syntax, pragmatics and prosody in Czech. 2008. xi, 143 pp. 128 Taleghani, Azita H.: Modality, Aspect and Negation in Persian. 2008. ix, 183 pp. 127 Durrleman-Tame, Stephanie: The Syntax of Jamaican Creole. A cartographic perspective. 2008. xii, 190 pp. 126 Schäfer, Florian: The Syntax of (Anti-)Causatives. External arguments in change-of-state contexts. 2008. xi, 324 pp. 125 Rothstein, Björn: The Perfect Time Span. On the present perfect in German, Swedish and English. 2008. xi, 171 pp. 124 Ihsane, Tabea: The Layered DP. Form and meaning of French indefinites. 2008. ix, 260 pp. 123 Stoyanova, Marina: Unique Focus. Languages without multiple wh-questions. 2008. xi, 184 pp. 122 Oosterhof, Albert: The Semantics of Generics in Dutch and Related Languages. 2008. xviii, 286 pp. 121 Tungseth, Mai Ellin: Verbal Prepositions and Argument Structure. Path, place and possession in Norwegian. 2008. ix, 187 pp. 120 Asbury, Anna, Jakub Dotlačil, Berit Gehrke and Rick Nouwen (eds.): Syntax and Semantics of Spatial P. 2008. vi, 416 pp. 119 Fortuny, Jordi: The Emergence of Order in Syntax. 2008. viii, 211 pp. 118 Jäger, Agnes: History of German Negation. 2008. ix, 350 pp. 117 Haugen, Jason D.: Morphology at the Interfaces. Reduplication and Noun Incorporation in Uto-Aztecan. 2008. xv, 257 pp.

116 Endo, Yoshio: Locality and Information Structure. A cartographic approach to Japanese. 2007. x, 235 pp. 115 Putnam, Michael T.: Scrambling and the Survive Principle. 2007. x, 216 pp. 114 Lee-Schoenfeld, Vera: Beyond Coherence. The syntax of opacity in German. 2007. viii, 206 pp. 113 Eythórsson, Thórhallur (ed.): Grammatical Change and Linguistic Theory. The Rosendal papers. 2008. vi, 441 pp. 112 Axel, Katrin: Studies on Old High German Syntax. Left sentence periphery, verb placement and verbsecond. 2007. xii, 364 pp. 111 Eguren, Luis and Olga Fernández Soriano (eds.): Coreference, Modality, and Focus. Studies on the syntax–semantics interface. 2007. xii, 239 pp. 110 Rothstein, Susan (ed.): Theoretical and Crosslinguistic Approaches to the Semantics of Aspect. 2008. viii, 453 pp. 109 Chocano, Gema: Narrow Syntax and Phonological Form. Scrambling in the Germanic languages. 2007. x, 333 pp. 108 Reuland, Eric, Tanmoy Bhattacharya and Giorgos Spathas (eds.): Argument Structure. 2007. xviii, 243 pp. 107 Corver, Norbert and Jairo Nunes (eds.): The Copy Theory of Movement. 2007. vi, 388 pp. 106 Dehé, Nicole and Yordanka Kavalova (eds.): Parentheticals. 2007. xii, 314 pp. 105 Haumann, Dagmar: Adverb Licensing and Clause Structure in English. 2007. ix, 438 pp. 104 Jeong, Youngmi: Applicatives. Structure and interpretation from a minimalist perspective. 2007. vii, 144 pp. 103 Wurff, Wim van der (ed.): Imperative Clauses in Generative Grammar. Studies in honour of Frits Beukema. 2007. viii, 352 pp. 102 Bayer, Josef, Tanmoy Bhattacharya and M.T. Hany Babu (eds.): Linguistic Theory and South Asian Languages. Essays in honour of K. A. Jayaseelan. 2007. x, 282 pp. 101 Karimi, Simin, Vida Samiian and Wendy K. Wilkins (eds.): Phrasal and Clausal Architecture. Syntactic derivation and interpretation. In honor of Joseph E. Emonds. 2007. vi, 424 pp. 100 Schwabe, Kerstin and Susanne Winkler (eds.): On Information Structure, Meaning and Form. Generalizations across languages. 2007. vii, 570 pp. 99 Martínez-Gil, Fernando and Sonia Colina (eds.): Optimality-Theoretic Studies in Spanish Phonology. 2007. viii, 564 pp. 98 Pires, Acrisio: The Minimalist Syntax of Defective Domains. Gerunds and infinitives. 2006. xiv, 188 pp. 97 Hartmann, Jutta M. and László Molnárfi (eds.): Comparative Studies in Germanic Syntax. From Afrikaans to Zurich German. 2006. vi, 332 pp. 96 Lyngfelt, Benjamin and Torgrim Solstad (eds.): Demoting the Agent. Passive, middle and other voice phenomena. 2006. x, 333 pp. 95 Vogeleer, Svetlana and Liliane Tasmowski (eds.): Non-definiteness and Plurality. 2006. vi, 358 pp. 94 Arche, María J.: Individuals in Time. Tense, aspect and the individual/stage distinction. 2006. xiv, 281 pp. 93 Progovac, Ljiljana, Kate Paesani, Eugenia Casielles and Ellen Barton (eds.): The Syntax of Nonsententials. Multidisciplinary perspectives. 2006. x, 372 pp. 92 Boeckx, Cedric (ed.): Agreement Systems. 2006. ix, 346 pp. 91 Boeckx, Cedric (ed.): Minimalist Essays. 2006. xvi, 399 pp. 90 Dalmi, Gréte: The Role of Agreement in Non-Finite Predication. 2005. xvi, 222 pp. 89 Velde, John R. te: Deriving Coordinate Symmetries. A phase-based approach integrating Select, Merge, Copy and Match. 2006. x, 385 pp. 88 Mohr, Sabine: Clausal Architecture and Subject Positions. Impersonal constructions in the Germanic languages. 2005. viii, 207 pp. 87 Julien, Marit: Nominal Phrases from a Scandinavian Perspective. 2005. xvi, 348 pp. 86 Costa, João and Maria Cristina Figueiredo Silva (eds.): Studies on Agreement. 2006. vi, 285 pp. 85 Mikkelsen, Line: Copular Clauses. Specification, predication and equation. 2005. viii, 210 pp. 84 Pafel, Jürgen: Quantifier Scope in German. 2006. xvi, 312 pp. 83 Schweikert, Walter: The Order of Prepositional Phrases in the Structure of the Clause. 2005. xii, 338 pp. 82 Quinn, Heidi: The Distribution of Pronoun Case Forms in English. 2005. xii, 409 pp.

81 FuSS, Eric: The Rise of Agreement. A formal approach to the syntax and grammaticalization of verbal inflection. 2005. xii, 336 pp. 80 Burkhardt, Petra: The Syntax–Discourse Interface. Representing and interpreting dependency. 2005. xii, 259 pp. 79 Schmid, Tanja: Infinitival Syntax. Infinitivus Pro Participio as a repair strategy. 2005. xiv, 251 pp. 78 Dikken, Marcel den and Christina Tortora (eds.): The Function of Function Words and Functional Categories. 2005. vii, 292 pp. 77 Öztürk, Balkız: Case, Referentiality and Phrase Structure. 2005. x, 268 pp. 76 Stavrou, Melita and Arhonto Terzi (eds.): Advances in Greek Generative Syntax. In honor of Dimitra Theophanopoulou-Kontou. 2005. viii, 366 pp. 75 Di Sciullo, Anna Maria (ed.): UG and External Systems. Language, brain and computation. 2005. xviii, 398 pp. 74 Heggie, Lorie and Francisco Ordóñez (eds.): Clitic and Affix Combinations. Theoretical perspectives. 2005. viii, 390 pp. 73 Carnie, Andrew, Heidi Harley and Sheila Ann Dooley (eds.): Verb First. On the syntax of verbinitial languages. 2005. xiv, 434 pp. 72 FuSS, Eric and Carola Trips (eds.): Diachronic Clues to Synchronic Grammar. 2004. viii, 228 pp. 71 Gelderen, Elly van: Grammaticalization as Economy. 2004. xvi, 320 pp. 70 Austin, Jennifer R., Stefan Engelberg and Gisa Rauh (eds.): Adverbials. The interplay between meaning, context, and syntactic structure. 2004. x, 346 pp. 69 Kiss, Katalin É. and Henk van Riemsdijk (eds.): Verb Clusters. A study of Hungarian, German and Dutch. 2004. vi, 514 pp. 68 Breul, Carsten: Focus Structure in Generative Grammar. An integrated syntactic, semantic and intonational approach. 2004. x, 432 pp. 67 Mišeska Tomić, Olga (ed.): Balkan Syntax and Semantics. 2004. xvi, 499 pp. 66 Grohmann, Kleanthes K.: Prolific Domains. On the Anti-Locality of movement dependencies. 2003. xvi, 372 pp. 65 Manninen, Satu Helena: Small Phrase Layers. A study of Finnish Manner Adverbials. 2003. xii, 275 pp. 64 Boeckx, Cedric and Kleanthes K. Grohmann (eds.): Multiple Wh-Fronting. 2003. x, 292 pp. 63 Boeckx, Cedric: Islands and Chains. Resumption as stranding. 2003. xii, 224 pp. 62 Carnie, Andrew, Heidi Harley and MaryAnn Willie (eds.): Formal Approaches to Function in Grammar. In honor of Eloise Jelinek. 2003. xii, 378 pp. 61 Schwabe, Kerstin and Susanne Winkler (eds.): The Interfaces. Deriving and interpreting omitted structures. 2003. vi, 403 pp. 60 Trips, Carola: From OV to VO in Early Middle English. 2002. xiv, 359 pp. 59 Dehé, Nicole: Particle Verbs in English. Syntax, information structure and intonation. 2002. xii, 305 pp. 58 Di Sciullo, Anna Maria (ed.): Asymmetry in Grammar. Volume 2: Morphology, phonology, acquisition. 2003. vi, 309 pp. 57 Di Sciullo, Anna Maria (ed.): Asymmetry in Grammar. Volume 1: Syntax and semantics. 2003. vi, 405 pp. 56 Coene, Martine and Yves D’hulst (eds.): From NP to DP. Volume 2: The expression of possession in noun phrases. 2003. x, 295 pp. 55 Coene, Martine and Yves D’hulst (eds.): From NP to DP. Volume 1: The syntax and semantics of noun phrases. 2003. vi, 362 pp. 54 Baptista, Marlyse: The Syntax of Cape Verdean Creole. The Sotavento varieties. 2003. xxii, 294 pp.  (incl. CD-rom). 53 Zwart, Jan-Wouter and Werner Abraham (eds.): Studies in Comparative Germanic Syntax. Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26–27, 2000). 2002. xiv, 407 pp. 52 Simon, Horst J. and Heike Wiese (eds.): Pronouns – Grammar and Representation. 2002. xii, 294 pp. 51 Gerlach, Birgit: Clitics between Syntax and Lexicon. 2002. xii, 282 pp. 50 Steinbach, Markus: Middle Voice. A comparative study in the syntax-semantics interface of German. 2002. xii, 340 pp. 49 Alexiadou, Artemis (ed.): Theoretical Approaches to Universals. 2002. viii, 319 pp.

48 Alexiadou, Artemis, Elena Anagnostopoulou, Sjef Barbiers and Hans-Martin Gärtner (eds.): Dimensions of Movement. From features to remnants. 2002. vi, 345 pp. 47 Barbiers, Sjef, Frits Beukema and Wim van der Wurff (eds.): Modality and its Interaction with the Verbal System. 2002. x, 290 pp. 46 Panagiotidis, E. Phoevos: Pronouns, Clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. 2002. x, 214 pp. 45 Abraham, Werner and Jan-Wouter Zwart (eds.): Issues in Formal German(ic) Typology. 2002. xviii, 336 pp. 44 Taylan, Eser Erguvanlı (ed.): The Verb in Turkish. 2002. xviii, 267 pp. 43 Featherston, Sam: Empty Categories in Sentence Processing. 2001. xvi, 279 pp. 42 Alexiadou, Artemis: Functional Structure in Nominals. Nominalization and ergativity. 2001. x, 233 pp. 41 Zeller, Jochen: Particle Verbs and Local Domains. 2001. xii, 325 pp. 40 Hoeksema, Jack, Hotze Rullmann, Víctor Sánchez-Valencia and Ton van der Wouden (eds.): Perspectives on Negation and Polarity Items. 2001. xii, 368 pp. 39 Gelderen, Elly van: A History of English Reflexive Pronouns. Person, Self, and Interpretability. 2000. xiv, 279 pp. 38 Meinunger, André: Syntactic Aspects of Topic and Comment. 2000. xii, 247 pp. 37 Lutz, Uli, Gereon Müller and Arnim von Stechow (eds.): Wh-Scope Marking. 2000. vi, 483 pp. 36 Gerlach, Birgit and Janet Grijzenhout (eds.): Clitics in Phonology, Morphology and Syntax. 2001. xii, 441 pp. 35 Hróarsdóttir, Thorbjörg: Word Order Change in Icelandic. From OV to VO. 2001. xiv, 385 pp. 34 Reuland, Eric (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. xii, 255 pp. 33 Puskás, Genoveva: Word Order in Hungarian. The syntax of Ā-positions. 2000. xvi, 398 pp. 32 Alexiadou, Artemis, Paul Law, André Meinunger and Chris Wilder (eds.): The Syntax of Relative Clauses. 2000. vi, 397 pp. 31 Svenonius, Peter (ed.): The Derivation of VO and OV. 2000. vi, 372 pp. 30 Beukema, Frits and Marcel den Dikken (eds.): Clitic Phenomena in European Languages. 2000. x, 324 pp. 29 Miyamoto, Tadao: The Light Verb Construction in Japanese. The role of the verbal noun. 2000. xiv, 232 pp. 28 Hermans, Ben and Marc van Oostendorp (eds.): The Derivational Residue in Phonological Optimality Theory. 2000. viii, 322 pp. 27 Růžička, Rudolf: Control in Grammar and Pragmatics. A cross-linguistic study. 1999. x, 206 pp. 26 Ackema, Peter: Issues in Morphosyntax. 1999. viii, 310 pp. 25 Felser, Claudia: Verbal Complement Clauses. A minimalist study of direct perception constructions. 1999. xiv, 278 pp. 24 Rebuschi, Georges and Laurice Tuller (eds.): The Grammar of Focus. 1999. vi, 366 pp. 23 Giannakidou, Anastasia: Polarity Sensitivity as (Non)Veridical Dependency. 1998. xvi, 282 pp. 22 Alexiadou, Artemis and Chris Wilder (eds.): Possessors, Predicates and Movement in the Determiner Phrase. 1998. vi, 388 pp. 21 Klein, Henny: Adverbs of Degree in Dutch and Related Languages. 1998. x, 232 pp. 20 Laenzlinger, Christopher: Comparative Studies in Word Order Variation. Adverbs, pronouns, and clause structure in Romance and Germanic. 1998. x, 371 pp. 19 Josefsson, Gunlög: Minimal Words in a Minimal Syntax. Word formation in Swedish. 1998. ix, 199 pp. 18 Alexiadou, Artemis: Adverb Placement. A case study in antisymmetric syntax. 1997. x, 256 pp. 17 Beermann, Dorothee, David LeBlanc and Henk van Riemsdijk (eds.): Rightward Movement. 1997. vi, 410 pp. 16 Liu, Feng-hsi: Scope and Specificity. 1997. viii, 187 pp. 15 Rohrbacher, Bernhard Wolfgang: Morphology-Driven Syntax. A theory of V to I raising and prodrop. 1999. viii, 296 pp. 14 Anagnostopoulou, Elena, Henk van Riemsdijk and Frans Zwarts (eds.): Materials on Left Dislocation. 1997. viii, 349 pp. 13 Alexiadou, Artemis and T. Alan Hall (eds.): Studies on Universal Grammar and Typological Variation. 1997. viii, 252 pp.

12 Abraham, Werner, Samuel David Epstein, Höskuldur Thráinsson and Jan-Wouter Zwart (eds.): Minimal Ideas. Syntactic studies in the minimalist framework. 1996. xii, 364 pp. 11 Lutz, Uli and Jürgen Pafel (eds.): On Extraction and Extraposition in German. 1996. xii, 315 pp. 10 Cinque, Guglielmo and Giuliana Giusti (eds.): Advances in Roumanian Linguistics. 1995. xi, 172 pp. 9 Gelderen, Elly van: The Rise of Functional Categories. 1993. x, 224 pp. 8 Fanselow, Gisbert (ed.): The Parametrization of Universal Grammar. 1993. xvii, 232 pp. 7 Åfarlí, Tor A.: The Syntax of Norwegian Passive Constructions. 1992. xii, 177 pp. 6 Bhatt, Christa, Elisabeth Löbel and Claudia Maria Schmidt (eds.): Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences. 1989. ix, 187 pp. 5 Grewendorf, Günther and Wolfgang Sternefeld (eds.): Scrambling and Barriers. 1990. vi, 442 pp. 4 Abraham, Werner and Sjaak De Meij (eds.): Topic, Focus and Configurationality. Papers from the 6th Groningen Grammar Talks, Groningen, 1984. 1986. v, 349 pp. 3 Abraham, Werner (ed.): On the Formal Syntax of the Westgermania. Papers from the 3rd Groningen Grammar Talks (3e Groninger Grammatikgespräche), Groningen, January 1981. 1983. vi, 242 pp. 2 Ehlich, Konrad and Jürgen Rehbein: Augenkommunikation. Methodenreflexion und Beispielanalyse. 1982. viii, 150 pp. With many photographic ills. 1 Klappenbach, Ruth (1911–1977): Studien zur Modernen Deutschen Lexikographie. Auswahl aus den Lexikographischen Arbeiten von Ruth Klappenbach, erweitert um drei Beiträge von Helene MaligeKlappenbach. (Written in German). 1980. xxiii, 313 pp.