The Evolution of Negation: Beyond the Jespersen Cycle 9783110238617, 9783110238600

Citation preview

The Evolution of Negation

Trends in Linguistics Studies and Monographs 235

Editor

Volker Gast Founding Editor

Werner Winter Editorial Board

Walter Bisang Hans Henrich Hock Heiko Narrog Matthias Schlesewsky Niina Ning Zhang Editor responsible for this volume

Volker Gast

De Gruyter Mouton

The Evolution of Negation Beyond the Jespersen Cycle

edited by

Pierre Larrive´e Richard P. Ingham

De Gruyter Mouton

ISBN 978-3-11-023860-0 e-ISBN 978-3-11-023861-7 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data The evolution of negation : beyond the Jespersen Cycle / edited by Pierre Larrive´e, Richard Ingham. p. cm. ⫺ (Trends in linguistics: studies and monographs; 235) Includes bibliographical references and index. ISBN 978-3-11-023860-0 (alk. paper) 1. Grammar, Comparative and general ⫺ Negatives. 2. Polarity (Linguistics) 3. Grammar, Comparative and general ⫺ Syntax. 4. Semantics. I. Larrive´e, Pierre. II. Ingham, Richard. P299.N4E96 2011 415⫺dc23 2011035434

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” 2011 Walter de Gruyter GmbH & Co. KG, Berlin/Boston Typesetting: RoyalStandard, Hong Kong Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany. www.degruyter.com

Table of contents Is there a Jespersen cycle? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pierre Larrive´e Negative words and related expressions: A new perspective on some familiar puzzles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Theresa Biberauer & Ian Roberts Negative words and negation in French . . . . . . . . . . . . . . . . . . . . . Marie Labelle Secondary negation and information structure organisation in the history of English . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ans van Kemenade

1

23 61

77

Looking high and low for NegP in early English . . . . . . . . . . . . . . . Eric Haeberli

115

Ne-drop and indefinites in Anglo-Norman and Middle English . . . . Richard Ingham

145

Looking at Middle English through the mirror of Anglo-Norman . . . Jack Hoeksema

165

Ne-absence in declarative and yes/no interrogative contexts: Some patterns of change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . France Martineau The early absence of the French negative marker ne . . . . . . . . . . . . Paul Rowlett

179 209

Atoms of negation: An outside-in micro-parametric approach to negative concord. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Viviane De´prez

221

Viviane De´prez: ‘‘Atoms of negation: An outside-in microparametric approach to negative concord.’’ Discussion . . . . . . . . . . Maj-Britt Mosegaard Hansen

273

vi

Table of contents

Negative polarity and the quantifier cycle: Comparative diachronic perspectives from European languages . . . . . . . . . . . . . . . . . . . . . . David Willis

285

Indefinite pronouns, synchrony and diachrony: Comments on Willis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Johan van der Auwera & Lauren Van Alsenoy

325

Subject index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Language index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

347 349

Is there a Jespersen cycle? Pierre Larrive´e 1. Introduction The purpose of this paper is to present a reconceptualisation of the central grammatical change that is the Jespersen cycle. The evolution of the grammar of negation is proposed to be best understood as pathways of changes for (families of ) items rather than as a syntactically-driven process. The paper does three things. It first introduces the research project from which emanate the contributions to this volume that are subsequently presented. It shows how pathways of change account for the development of n-words and adverbial reinforcements. Defining diagnostics of negative items are proposed before the argument is summarised in the conclusion. 2. Investigating the Jespersen cycle: Summary of contributions Some grammatical changes have been argued to proceed in cycles. Probably the best known and earliest was put forward by Jespersen (1917; for precursors, see van der Auwera 2009: 6; for a full bibliography of work on the subject, see the web site of the project Cycles of grammaticalization). He suggested that there was a cycle of sentential negation such that the preverbal negative loses substance and is adjoined by a postverbal reinforcement that ends up as the sole expression of negation. This can be illustrated with the idealised French data provided by Jespersen himself, with each example corresponding to a new stage of the cycle (on di¤erent ways of counting stages, see van der Auwera 2009): (1) Stage 1 Jeo ne dis. 1SG NEG say-PR-1SG

(Early French)

‘I don’t say.’ (2) Stage 2 Je ne dis pas. (Middle French) 1SG NEG say-PR-1SG NEG

2

Pierre Larrive´e

(3) Stage 3 Je dis pas. 1SG say-PR-1SG NEG

(Contemporary informal French)

A new cycle can be started as in French-based Creoles: (4) New stage 1 Mo pa di. 1SG NEG say-PR-1SG

(Cont. Louisiana French)

The major question that this raises is whether this process of language change is inevitable. Such inevitability seems to be suggested by the notion of a cycle, which in itself implies a series of di¤erent stages that lead back to an initial position. This is implicit in an analysis such as that by van Gelderen (2011), who sees linguistic cycles as the result of a tension between principles of economy applying to individual items on the one hand and on the other to syntactic arrangements. That an orderly movement from preverbal to postverbal negative marking and back again is found across a variety of unrelated languages might point to a mechanism internal to language organisation – taking up the old debate as to whether the organisation of languages includes the mechanisms of their own change as proposed by August Schleicher and challenged by Ferdinand de Saussure. Should all languages be expected to follow the same trajectories? What external or internal factors might intervene in the history of particular languages in such a way as to a¤ect outcomes? The causes of the Jespersen cycle is the object of the research conducted under the International Network Cycles of Grammaticalization, generously funded by The Leverhulme Trust. The Network held six seminars from June 2008 to December 2009 in Birmingham. They reappraised an often described but seldom explained central grammatical change. The first particular objective was to propose semantic and syntactic criteria for the identification of negative items, as the emergence of new negative items is the condition of the cycle. As the emergence of new negatives impacts on their ability to be used together in concord configurations, the second particular objective was to establish whether the history of negation is best accounted for by negative phrases or more recent feature-based analyses (Brown 1999, Zeijlstra 2004). These helped two general objectives to be achieved: the advocacy of using new data closer to everyday usage for earlier periods of well-documented languages, and a dialogue between formal and functional approaches following an original format of contributions and their discussion.

Is there a Jespersen cycle?

3

The exchanges of the seminars culminate in the contributions to this volume that put forward the reconceptualisation of the Jespersen cycle. The main articles deal with the themes of sentential adverbial negation (van Kemenade, Ingham, Martineau), n-words (De´prez, Willis) and negative concord (Biberauer & Roberts). The papers are presented below before elements of convergence between them are highlighted. The Jespersen cycle is primarily about the move from preverbal to postverbal negative items. Going back to Indo-European roots, the preverbal marker is lost in most Germanic languages early on. The period when English still had a preverbal markers is what Ans van Kemenade analyses. She looks at new data in order to establish what the syntactic contribution of Old English na is. It can be found in two syntactic positions. One is a high position that is situated above the tense phrase in root clauses. This high position seems to correlate with discursive functions associated with presupposed propositions. The other is a low position below TP that is targeted in non-root clauses. As syntactic di¤erences between root and non-root clauses are lost with the decline of the second-position verb constraint, so is the higher position, and the preverbal marker is then restricted to one dependent syntactic position that may explain why it was lost. The loss of the preverbal marker in Middle English is what is considered by Richard Ingham. He raises the point that the crucial period between 1240 and 1380 lacks a body of prose, which makes it di‰cult to track the course of these changes reliably. This lacuna is caused by the sociolinguistic fact that prose would have been written in Latin or AngloNorman during that period. It is therefore the legal prose of that variety of French that Richard Ingham examines to find traces of influences from changing Middle English. One such change is the loss of the preverbal negative. Ne-drop is found in late thirteenth century, with indefinites such as onc ‘never’, rien ‘nothing’ and nul ‘none’ expressing negation by themselves. The first cases of post-verbal negation pas without ne are recorded in the 1340s. The loss of the preverbal negator in Anglo-Norman thus provides indirect evidence for the chronology of change for the English negative, and explains why in this respect Anglo-Norman appears more advanced than its continental counterpart, from which it di¤ers little, syntactically, up to the late 1360s. The progressive loss of ne in French since the Classical period is mapped out by France Martineau. She considers literary and non-literary data that reflect social stratification, using notably corpora of letters produced by partially literate writers. These sources present data both from the Old as

4

Pierre Larrive´e

well as from the New World. The regional and social variables are considered with respect to internal parameters of sentence types. Interrogative and assertive sentences witness a di¤erent rate of ne loss until the dominant interrogative structure patterns with the assertive one in the nineteenth century. The innovation of post-verbal marker without ne is promoted by lower social classes, and in New World French as compared to the French of the Old World. The chronology of change is shown to be earlier than proposed in Martineau & Mougeon (2003), and the decline of preverbal marker use is reversed slightly at the beginning at the nineteenth century. The results conclusively show that the emergence of post-verbal negatives cannot be instigated by the loss of the preverbal marker, as the former follows the later by at the very least half a millennium. The emergence of new negative items is also seen when n-words evolve out of indefinites. This is the object of the contributions by Viviane De´prez and by David Willis. David Willis raises the question of the directionality of evolution paths for n-words. He looks at a variety of European languages and provides detailed documentation of the situation in Welsh. This allows him to defend the assumption that n-words go from existentials to indefinites to negative items, which finds support in the rare studies of first-language acquisition. He looks at alleged cases of backward development in Celtic and Slavonic and finds them to be inconclusive. His work shows that n-words rarely possess an exclusive status as a negative polarity or negative item, and have dual status. The conjunction of status seems to evolve along a definite pathway of contexts, where n-words in contemporary Welsh and French maintain polarity uses with strong licensers. The question of the dual status of n-words is considered by Viviane De´prez. She extends her earlier proposals on the internal structure of nwords. N-word status is acquired by items that turn into pronouns when determiners become obligatory with nouns in Classical French. That is because determiners fill the head position of nouns that must remain free for pronouns to move to intermediate and high positions within the determiner phrase, calling for an external licensing negative in intermediate position, its negative trait being visible at the edge of the phrase in high position. This structural flexibility accounts for the variability observed in contemporary Continental French, Quebec French, as well as Haitian and Martinique Creoles. Creoles provide critical evidence that not all items change at the same chronological rate, and that a micro-parametric examination of the structure of individual items provides a better understanding of pathways of change.

Is there a Jespersen cycle?

5

The evolution of negatives a¤ects the issue of negative concord, and whether a language can use several markers to express a single negative proposition. A general framework linking the emergence of negative words and negative concord is proposed by Theresa Biberauer and Ian Roberts. Elaborating on previously developed models, they seek to encompass both the behaviour of sentential negative and that of n-words. A negative feature is generated in the complementiser phrase, where it is either kept or donated to a dominated projection, which explains the variety of sentential markers found across languages. That feature acts as a Probe licensing other negative elements that are dependent Goals. The ability to enter in an Agreement relation defines the distribution of individual markers. Such a relation can be established between n-words in French but not in (standard) English, which allows only one such item, any further n-word having to be introduced in a focus phrase that stands in the way of Agreement. The intervention of focus should account for the pragmatic factors associated with the emergence of new negative items and configurations. A critical assessment of these proposals will not be attempted here, as it is provided by the extended discussion pieces following each paper that promise to take the research agenda forward. One major way in which this is achieved by this volume is by calling into question the idea that the evolution of negation is driven by a cycle. This challenge bears on the automatic nature of change and on the dependence between stages. If the change were automatic, it would be found in all languages, which it is not: Slavic languages do not go beyond stage 1. If where it happens, the change were structured by stages, it would a¤ect all the items of a category equally following the same chronology, and would lead a language to have only one stage at any one time; two predictions that are unsupported. To restrict ourselves to observations established in this volume: continental French ne´ant is fully negative well before rien is (Labelle); all AngloNorman negatives are not used without ne (Ingham; and Hoeksema for a similar situation in Dutch); the French Creole n-words equivalent to never and no one behave quite di¤erently with respect to the post-verbal negative adverb (De´prez); and the emergence of postverbal items in French does not bring about a synchronised decline of the preverbal marker (Martineau; the same argument is made by Breitbarth 2009: 87). The alleged causes of the Jespersen cycle are also called into question. The role of phonetic weakening of the preverbal item – already questioned by the case of Greek, where the phonetic strength of the preverbal marker does not prevent a Jespersen cycle (Kiparsky & Condoravdi 2006) – is shown by

6

Pierre Larrive´e

Ingham to be unlikely since the early loss of the preverbal particle is not found with all negatives. Furthermore, the Jespersen cycle only covers a portion of the change of grammatical negation, having little to say about n-words, and nothing at all about negative coordination and constituent negation for instance. An approach spelling out the possible pathways of change for (families of ) items allows one to account for the behaviour of individual elements without sacrificing them to expectations of automatic change between dependent stages. This permits an exploration of concrete conditions of change that help define the bridging contexts that lead new generations of learners to attribute a new status to markers. How an approach in terms of pathways of change would deal with the evolution of negation is what I want to do in the remainder of this paper, which is based principally on French data. I start with the obvious case of n-words and show how a semantic map may help make sense of the problematic duality of these items. I want to discuss what the change is, and when and why it happens. These questions are brought to bear on the evolution of post-verbal reinforcements. How a changing item can be said to be negative is considered in the third section. After envisaging whether a pathway approach can account for the impression of a cycle and discussing the possible counterexample of morphologically negative items, the conclusion summarises the argument.

3. Where do n-words come from? There is a recognised pattern of change by which negative pronouns such as rien emerge from indefinite nouns. The change involves three steps. The first is a noun that may have a determiner and that possesses an existential reading; it evokes a non-specific occurrence of an ontological category. (5) [. . .]

se prist a porpenser [. . .] REFL-3PS took to think

Quele rien li fust miez a faire [. . .]. what thing him-DAT be-PST better to do-INF ‘Then he started to think what thing was better to do.’ (Dial. Greg., circa 1212, 14vb; Anglo-Norman Dictionary) The second step is the noun without a determiner that has a negative polarity interpretation: It refers to an indefinite occurrence of an ontological

Is there a Jespersen cycle?

7

category in non-veridical contexts (Giannakidou 1998) such as negative, interrogative and conditional sentences. (6) [. . .] rien n’estoit espergne´ en tel cas [. . .]. anything ne be.IMP spared in such case ‘Not anything was spared in this case.’ (C.N.N., circa 1456–1467, 198; DMF2009) The third is what has apparently become a pronoun excluding a determiner that is interpreted as a negative. This is illustrated below by a contemporary French example where the negative value can be attributed to no other visible item than rien itself: (7) Rien arrive par hasard. ‘Nothing happens by chance.’

(http://roxylive.skyrock.com/, 2010)

The pattern of change is relatively well attested. The emergence of a negative value for n-words starts with the eighteenth century in the literary data analysed by Martineau & De´prez (2004; and Martineau this volume). The same type of data provides us with a not insignificant proportion of polarity uses in contemporary French (Gaatone 1971), including in conditional and interrogative contexts. These uses, however, are neither natural, nor are they attested in contemporary corpora of informal spoken French (Elicop, Corpus de la Parole, Corpus de Franc¸ais Parle´ Parisien des anne´es 2000), which provide examples only in contexts commanded by a comparative and by sans ‘without’ (as noted by Hansen this volume). The striking disparity between an extended set of negative polarity contexts in literary sources and the narrow range of such contexts in everyday exchanges would be reasons enough to revisit whether the change away from polarity uses might in fact have taken place far earlier than the seventeenth century. A look at the uses of rien in dialogical Anglo-Norman Year Books of Edward I and Edward III demonstrate a dramatic change from 1292 to 1346. Edward I provides 34 occurrences of rien with seven straight indefinite uses notably in interrogatives (2) and conditionals (2); such uses are not absent from Edward III, but they represent a smaller proportion (8 out of 64 occurrences), and new uses are attested with clear negative import (2 in contrast, and 2 others in non-verbal fragments, without ne). This calls for a reappraisal of the chronology of change, which can be accomplished on the basis of sources closer to dialogal exchanges (such as the electronic corpus Narrations and Dialogues put together by Richard Ingham and me, and freely available on request).

8

Pierre Larrive´e

Whatever the chronology that a detailed examination of better data might establish, the question arises as to the cause of these changes. They have been argued to relate to the morphosyntactic shape of the noun phrase. The observation that negative polarity and negative uses emerge in the absence of determiners lead to the analysis by De´prez that it is when the determiner becomes obligatory for nouns that n-words become pronouns. The absence of determiners with pronouns makes it possible for the item to figure in a high or a low position in the phrase. A lower position likens n-words to polarity items in that the unfilled head of the projection must be licensed by an external negative. A n-word in the head of the projection makes its intrinsic negative value visible. Thus, negative value corresponds to negative traits being at the edge of a phrase. This could well explain an otherwise puzzling contrast between polarity coordination that contents itself with one conjunction (Il n’a pas vu Richard ni Pierre ‘He ne has not seen Richard nor Pierre’, ‘He hasn’t seen Richard or Pierre’) and the fully negative coordination (Ni Richard ni Pierre n’ont (?? pas) e´te´ vu ‘Neither Richard nor Pierre ne have (?? not) been seen’) (Mouret 2004), where the head of the phrase is overtly filled by a negative item. For all its merits, this influential analysis does raise issues. One is chronology, as suggested by data from Anglo-Norman that well before determiners became compulsory had negative n-words, some of which without ne (Ingham this volume). Early French has a n-word ne´ant that clearly displays a majority of negative uses (Larrive´e & Ingham in press). Another issue is the role of grammatical categorisation. Not all n-words are pronouns: while aucun, nul, personne are, and while the nominal phrase nulle part used as an adverbial might be, adverbial jamais has none of the characteristics of pronouns, and neither was it ever a noun (cf. also the Welsh adverb dim, see Willis in press: 7). This raises the important question of what might be the common property of n-words. What is common to jamais and rien is their evocation of an ontological category (Haspelmath 1997; Floricic 2005). (They do, therefore, not represent a minimal value like other items with NPI uses do, on which among many others, see Eckard 2006 and Israel 2011 for functional perspectives, and Falaus 2010, Krifka 1998 for formal ones.) N-words are expressions that evoke inanimate referents, humans, time, place, and manner in the case of aucunement and nullement ‘in no way’. The relation to ontological categories might account for the central problem raised by n-words, i.e. that of their duality. N-words in French and many other languages maintain both polarity and negative uses through extended periods. Dual use of the n-word rien has

Is there a Jespersen cycle?

9

been the case for at least five centuries in Continental French. The association between the two readings extends to polarity readings being created out of negative ones. Judging by their morphology, the n-words nul ‘none, no one’ and nulluy ‘no one’ would have been negative in proto-French, before reverting in early literary French to polarity uses in the usual conditional and interrogative contexts, and becoming negative again early on in Anglo-Norman and later in Classical Continental French (other cases being discussed in this volume by van der Auwera & Van Alsenoy, as well as by Hoeksema). Such duality is a problem because neither use can easily be derived from the other. It, therefore, contradicts the assumption that semiological units should have a single underspecified meaning that is enriched by context. The assumption that the readings of items should be lost as a new one is gained, and that evolution should go in one direction, is called into question by presumed cases of back-formation. One way to capture the duality of n-words is through a revised semantic map approach as developed by Martin Haspelmath (1997). Semantic maps were designed to account for polyfunctional items that may take on several functions, and have the practical virtue of sidestepping the question of the semantic unity of items, and of leaving the study of particular interpretations to one side. A number of di¤erent functions can be taken on by the same synchronic paradigm of items as long as these functions are contiguous on the map. In concrete terms, an indefinite item that is found in conditionals could be used as a free choice item only if it is used in comparatives as well according to the proposed map (Haspelmath 1997: 68). Specific >Specific >Irrealis known unknow non-specific

>Questions

>Indirect negation

>Direct negation

>Conditionals >Comparative >Free choice

Apart from empirical issues raised in this volume by van der Auwera & Van Alsenoy, semantic maps beg the theoretical question of the conflation of syntactic and semantic categories. Free choice is a semantic label that concerns itself with a particular type of indefinite readings (see for instance Farkas 2002, Kratzer & Shimoya 2002 and Menendez-Beninto 2006 for recent approaches to free choice readings), whereas conditional is a syntactic environment, in which either a negative or an indefinite reading of a n-word is possible (highly literary si tu vois personne ‘if you see anyone’ vs. everyday si tu ne vois personne ‘if you see noone’). It might be desirable to separate the semantic interpretations of n-words and their

10

Pierre Larrive´e

distribution in a two-dimensional map that might look something like the following: Specific known Assertion # Attributive # ... #

>

Specific unknown Future # Modality # ... #

>

Non-specific Indefinite Questions # Conditionals # Comparatives # Without #

>

Negative Sentential # Constituent # ... #

This would help capture finer-grained generalisations about the pathways of change, as has been looked into by Hoeksema for Dutch (see also Vanderheyden 2010 for the distribution and competition between readings of aucun in the history of French). It seems that as in French, Welsh n-words end their polarity career in comparative and without environments. As with Haspelmath’s maps, the respective arrangements of functions are largely an empirical question. The theoretical question is the cause for this arrangement. Why should functions stand in the contiguous relation that they seem to? Why should n-words evolve forward and sometimes backward along contiguous functions, between negative polarity and negation rather than between negative and existential values? The answer is that contiguous functions share more semantic properties than those that are not (Larrive´e 2003). That is why they are more likely to be related by bridging contexts than other values may be, to lead to one being reanalysed as the other. Bridging contexts between negative and negative polarity are more likely than between negative and existential, and this clearly is an empirically testable claim that, should it be verified, would support the pathway approach proposed here. How common semantic properties are to be demonstrated for environments such as questions and conditionals as compared to comparatives and without phrases remains to be assessed. That such an assessment could be conducted is shown by the di¤erent subcategories of negatives proposed by Ton van der Wouden (1998). N-words are defined not by their grammatical category, but by the fact that they evoke ontological categories. Their evolution raises the central challenge of their duality: negative polarity and negative uses coexist for considerable periods of time. This duality constitutes a problem for the idea of cyclic change, which supposes a forward-looking series of clean breaks from one type of use to the next, rather than conjunctions of di¤erent uses that may allow back-formation. This conjunction need not be

Is there a Jespersen cycle?

11

random however, as suggested by the observation that the last polarity functions expressed by n-words both in Welsh and French are in comparative and without phrases. A semantic map is proposed that would define the contiguous functions and environments which can be taken on by n-words. Whether such an approach can be fruitfully applied to adverbial reinforcements is considered in the next section.

4. Where do adverbial reinforcements come from? The pattern of change that a¤ects adverbial reinforcements such as pas is similar to that of n-words, from measure phrase to polarity item to negative. The noun pas ‘step’ takes on a measure interpretation with predicates of movement. The Early French equivalent to Il ne marche pas would have meant ‘He doesn’t walk by so much as by one step’ (Meillet 1912). The lexically-circumscribed minimizer interpretation (Bolinger 1972, Israel 2001) is reanalysed as a general negative polarity item reading with any verb. The Old French equivalent to Il ne marche pas would have meant ‘He doesn’t walk at all’, with an emphatic value related to polarity status (Kadmon and Landman 1993). The negative polarity item is reanalysed as a negative adverb, and has a pragmatic value until it becomes the default negative (Larrive´e 2010 and references therein). Finally, the default negative becomes a light adverb in the terminology of Abeille´, in that it associates with the verb in a quasi-clitic way (Abeille´ and Godard 2001 among other publications), which can be captured by saying that it enters a functional negative phrase (Rowlett 1997, van Gelderen 2008, Haeberli this volume). The issue is that this pattern is largely speculative. Little evidence supports the hypothetical pathway from measure to polarity to negative. The attested readings in early French texts do not attest to an overwhelming use of pas, point or mie as measure phrases before they would become polarity items. The measure phrase interpretation largely appears as a reconstruction to make sense of how a polarity item like pas could come from the ordinary noun ‘step’. The polarity status is better established, although again there is no period where pas would have had a majority use as a polarity item before being attested as a negative. The fact that a post-verbal reinforcement like goutte ‘drop’, which distributes with sentence negation in a few phrases such as Je ne vois goutte ‘I can see not even a drop’, i.e. ‘I can’t see anything’, does not develop beyond polarity item status indicates that this status might have characterised other rein-

12

Pierre Larrive´e

forcement. There are some uses of pas in polarity contexts, although these are mainly in interrogatives (Martineau, and Rowlett, this volume, and references therein), and why other contexts do not host polarity pas is a question that can be raised by the parallel that is drawn here with nwords. It may be that these readings are actually a backward formation in Classical French literary registers. This is indicated by the fact that data closer to the vernacular provide no evidence of polarity uses for pas. A quick look at the uses of pas in Edward I Year Books show 50 uses, 43 straight sentential post-verbal negative with ne, 1 sentential negative where pas precedes ne immediately, 5 coordinated constructions with pas preceded by nent (4) or nun (1) ‘not’, and 1 constituent negation ( pas seysi ‘not seized’). Whatever the case may be, it remains that polarity uses never clearly precede negative ones. This supports a pathway perspective that allows the conjunction of functions, which is problematic for a cyclic understanding of change. The presumed passage from a negative polarity to a negative role modelled for n-words by De´prez is extended to negative adverbials by Roberts & Roussou (2003: 148–157). The general idea is that as nominals pas, point and mie become partitive quantifiers, they move up to a higher position in the phrase where they are reanalysed as negative adverbs. This upward movement does not depend on the emergence of determiners, or any other external trigger. It certainly does not depend on grammatical categorisation: gue`re ‘not much’ does not seem to have ever had a substantial number of nominal uses, and neither does plus ‘no more’. It is therefore not clear how the proposed reanalysis relates to either polarity status or general chronology. The pathway of evolution of a post-verbal reinforcement like pas could be placed on the map proposed for n-words. Specific known noun pas #

> Specific unknown minimizer # reading

> Non-specific > Negative Indefinite Questions # ... #

Sentential # Constituent #

What such a map suggests is that particular environments play a lesser role for reinforcements than for n-words. Other disparities include the lower extent to which functions are conjoined and to which back-formations are attested. These may be impressions due to the lack of detailed empirical studies of the functions of adverbial items in data reflecting everyday exchanges. These functions can nonetheless be placed on a semantic map to

Is there a Jespersen cycle?

13

account for the evolution of items. How items that have become negative can be identified as such is envisaged in the next and final section.

5. When is an item negative? As the pathways of change go from negative polarity to negative, a crucial question is when an item has become a negative. This question may appear so simple as to be barely worth asking: an item has become a negative when it introduces a negative value, however this is to be formalised in a particular theory. The source of that negative value is however not always easy to apportion, because the impact of negation over an utterance may make it unclear which item introduces it. This applies, in particular, to cases of multiple concording items. The fact that a n-word that follows another n-word (rien in Personne n’a rien fait ‘Nobody ne has done nothing’ can be replaced by a polarity item as in Personne n’a fait quoi que ce soit) is often taken to indicate that they must be polarity items themselves. Thus, no consensus exists as to whether Quebec French Il a pas rien fait ‘He has not nothing done’, that is ‘He hasn’t done anything’, hosts a NPI commanded by negative pas or a negative concording with the postverbal negator. Diagnostics concerning modification by adverbs such as absolutely and almost, types of complements, and locality are not always conclusive. The infelicitous modification by the adverbs almost and absolutely that is supposed to indicate polarity status (Hoeksema this volume), which is not often well-attested in medieval sources, may in fact demonstrate a di‰cult relation with the sentential negator that is found with polarity items (?*He didn’t do almost anything), as this transfers to the modification of other items (?*He didn’t do almost everything) (see also Rohrbaugh 2007). De´prez proposes that some types of complements may distinguish polarity and negative uses of the same n-word, which would indicate that rien in the Quebec French sequence just cited would have the same status whether it is used with pas or not (Il fait ( pas) rien d’autre ‘He does (not) do nothing of else’, ‘He does nothing else’), which contradicts her own predictions. Locality of relation between n-words might not indicate negative status, since negative n-words in Quebec French and Catalan are not always local (Larrive´e 2004: 131–132). Some interpretative criteria might well be decisive, but are not easy to apply on historical attestations. It is not clear that the following contrast from Greek can easily be carried over to early French.

14

Pierre Larrive´e

(8) Background: A: You were shopping all day. Did you buy anything? Clothes? Books? Records? B: a. #A, oxi. Dhen aghorasa KANENA vivlio. oh no not bought.1sg nbook #‘Oh, no. I bought no books.’ b. A, oxi. Dhen aghorasa kanena vivlio. oh no not bought.1sg nbook ‘Oh, no. I didn’t buy any books.’ (Giannakidou 2006) Likewise for the alleged tolerance of negatives to exceptions (Ja¨ger 2009: 12). (9) a.

I didn’t get anything for my birthday. (*/# Well a plantpot from my husband.) b. I got nothing for my birthday. (Well a plantpot from my husband.)

Conclusive diagnostics are found with n-words used as the sole item introducing a negative value. The case of n-words used as negative answers is hotly debated, as some suggest that it is the elided propositional material that may provide the negative operator (Boskovic in press; see de Swart 2010 for a challenge to that perspective). Such an elision is implausible in constituent and contrastive constructions. The negative value of nothing in a sequence such as He was looking at nothing cannot by definition be accounted for by an underlying propositional negative, precisely because no propositional scope is expressed by the constituent negation – He was looking at nothing pointedly does not means He wasn’t looking at anything, as made clear by the continuation since his eyes were closed, which can apply to the latter but not the former (see Larrive´e 2004: 70–76). The existence of constituent scope must therefore invalidate the claim (Laka 1990; Ladusaw 1992; Giannakidou 2000; Penka 2007) that n-words are never negative in themselves (see Mathieu 2001 for a general critical perspective; and Iatridou & Sichel 2011). Proposing an underlying negative operator with constituent scope (e.g. Penka 2007) is in essence a theoretical stipulation, and it is not clear what advantage this would o¤er over simply recognising the negative character of the n-word, assuming the proposal makes the right predictions. (In fact, some inaccurate predictions can be identified: The suggestion from a reviewer that in Max bought

Is there a Jespersen cycle?

15

a gramophone for no particular reason, the n-word concords with an intermediate negative operator – paraphrasable by It isn’t the case that there is some particular reason such that Max bought a gramophone for that reason – does not seem to o¤er an adequate rendition of the original – as apart from anything else the paraphrase should allow an embeded NPI as in It isn’t the case that there is some reason such that anybody bought a gramophone for that reason versus the infelicitous ?* Anybody bought a gramophone for nothing.) Likewise for contrastive constructions such as He voted Labour, not Lib Dems, which must indicate the negative value of the item. The possibility of double negative readings has been noted since Larrive´e (1995) as a criterion of the negative status of individual items: Rien must have negative potential in Quebec French since the sequences Il fait pas rien and Personne fait rien can also mean that something was done. The fact that contrastive constructions have been noted for pas and rien in the Anglo-Norman data provided above, and that topicalisation, constituent negation and double negation is attested for nul confirms that these items had negative uses well before Classical French. (11) Nule riens je n’i donroie. none thing I ne-him-DAT give. ‘Not a thing would I give him.’ (Colin Muset, thirteenth century, VII, 14. Rouquier 2004) (12) Saichent tuit que nous, [. . .] meismes et metons know-3PP everyone that we ourselves and put au neant et pronunc¸ons pour de nulle to nothingness and declare for of no value l’opposicion [. . .]. value the-opposition ‘That everyone knows that we did reduce and are reducing to nothingness and prononced as of no value the opposition [. . .].’ (Chartes et documents de l’abbaye de Saint-Magloire, t.3, 1330–1436, 70) (13) [. . .] mes je nes dirai mies por noiant. but I ne-it-PL-ACC say-FUT mie for nothing ‘But I won’t say them for nothing.’ (Lancelot du Lac, II 502. Richard Ingham, personal communication)

16

Pierre Larrive´e

(It is debatable whether (11) is a topicalisation, as one might expect subjectverb inversion in this case. The important observation is that sentence initial is not a position where negative polarity items are expected: One finds in informal French Rien du tout ils font ‘Nothing at all they do’, but not *Quoi que ce soit du tout ils ne font ( pas). The lack of parallelism with NPIs in each case is found in the other diagnostic contexts.) The same arguments have been shown to demonstrate the negative status of ne´ant (Larrive´e & Ingham in press). The diagnostics of constituent uses, contrastive ones and double negation are therefore put forward to establish when items have acquired a new function along the proposed pathway.

6. Conclusions Is there a Jespersen cycle? I have proposed that there is not. All the relevant items do not undergo the same syntactic evolution in successive mutually exclusive stages. Rather, each item slides along a pathway of ordered functions and environments. Following di¤erent chronologies as with ne´ant as compared to rien, several functions can be conjoined by individual items, and back-formation of negative polarity functions seem possible. Yet, surely, as illustrated by (1)–(4), a cyclic evolution may concern at least some sentential negatives? Post-verbal negators create the impression of a cycle as a result of their pattern of development. They first emerge as polarity items that generally follow their negative licensor. As they become negative themselves, the post-verbal item seems to have followed a cycle. Therefore, the impression of a cycle is a consequence of pathways of evolution of reinforcement going through a polarity phase. I want to make it clear that I am not denying the importance of syntax in the evolution of negation; I am asserting that this evolution is not caused by categorical syntactic processes, but by lexical changes. Another question that the pathway perspective raises is whether it is needed by those negatives relating to Indo-European *ne. The items prefixed in this way are largely assumed to have always been negative, and therefore imply no pathway at all. That would be why morphologically negative items have been observed to be used in post-verbal functions without undergoing a polarity step, as noted for Middle Dutch niets by van der Auwera (2009), and as is the case with early French ne´ant (Larrive´e & Ingham in press) and even apparently with non (as in the rare attestation Ne vous chaille non [. . .] (Cent Nouvelles Nouvelles, XXX), ‘NEG 2PP-ACC trouble NEG’, ‘Don’t trouble yourself ’).

Is there a Jespersen cycle?

17

The assumption that no evolution takes place with morphologically negative items – which is probably to blame for the reduced attention given to the evolution of morphologically negative items of English (but see Terasawa 2005) – might need to be reconsidered however. A reason to do so is provided by the literary Latin coordinator nedum, which seems to develop polarity uses before it disappears (Dupraz 2010), and pending detailed examinations, that seems to be what happens to French coordinator ne. That evolution is in line with the assumptions put forward by Breitbarth (2009), according to whom preverbal clitics must be reanalysed as some sort of polarity or a¤ective particles when other n-words become fully negative, although that hypothesis requires more empirical validation. While not all markers of this morphological family later develop polarity readings, the crucial fact that some do is indicative of a change that can be captured by pathways, but not by the idea of a cycle: No syntactic cycle is concerned by negative items becoming polarity items, and the semantic change from negation to polarity would in fact invert any semantic cycle that one might want to propose. This article has proposed a reconceptualisation of the Jespersen cycle on the basis of a converging theme in the contributions to this volume. The change in the grammar of negation is best thought of pathways of functions along which (families of ) items evolve. The achievement of negative status can be diagnosed by the criteria proposed, concerning double negative interpretations and uses as constituent negation and in contrastive constructions. These criteria and the proposed pathways invite future research to reconsider the function and distribution of individual negative items in texts that are likely to reflect vernacular language. Only those can provide a reliable access to the causes of grammatical change.

Acknowledgements I gratefully acknowledge the input to this article from Richard Ingham, Johan van der Auwera and the seminar participants. Richard’s tireless involvement made the book possible, as did the copy-editing talent of Robert Fox, the contribution of anonymous reviewers, and the guidance from the series editor Vo¨lker Gast and the team at Mouton de Gruyter, especially Birgit Sievert. Each made the journey an interesting one.

18

Pierre Larrive´e

References Sources Anglo-Norman Dictionary. http://www.anglo-norman.net/ Corpus de la Parole. http://www.corpusdelaparole.culture.fr/spip.php?article8 Corpus de Franc¸ais Parle´ Parisien des anne´es 2000. http://ed268.univ-paris3.fr/syled/ressources/Corpus-ParoleParis-PIII/Presentation.html DMF2009. Dictionnaire de Moyen Franc¸ais. http://www.atilf.fr/dmf/ Elicop. http://bach.arts.kuleuven.be/elicop/ Narrations et Dialogues en franc¸ais ancien: The Anglo-Norman Year Books Corpus. Available from the author on request. Studies Abeille´, Anne & Danie`le Godard 2001 A class of lite adverbs in French. In: Joaquim Camps & Caroline R. Wiltshire (eds.), Romance syntax, semantics and their L2 acquisition, 9–26. Amsterdam: Benjamins. Boskovic, Zeljko In press Licensing negative constituents and negative concord. MS, University of Connecticut. 14 pages. Available at http://web2.uconn. edu/boskovic/papers.html. To appear in Proceeding of NELS 38. Bolinger, Dwight 1972 Degree words. The Hague: Mouton. Breitbarth, Anne 2009 A hybrid approach to Jespersen’s cycle. Journal of Comparative Germanic Linguistics 12: 81–114. Brown, Sue 1999 The syntax of negation in Russian: A minimalist approach. Stanford: CSLI. Chierchia, Gennaro 2008 Broaden your views. Implicatures of domain widening, and the logicality of language. Linguistic inquiry 37 (4): 535–590. Cycles of Grammaticalization: Bibliography of work on the Jespersen cycle. Available at http://www1.aston.ac.uk/lss/research/researchprojects/cycles-of-grammaticalization/cycles-ofgrammaticalization-bibliography/ De´prez, Viviane 2000 Parallel (a)symmetries and the structure of negative expressions. Natural Language and Linguistic Theory 18 (2): 253–342.

Is there a Jespersen cycle?

19

Dupraz, Emmanuel 2010 Le gramme`me latin nedum: syntaxe et pragmatique. Paper read at the conference A Contrario, Universite´ de Caen, March 26, 2010. Hand-out, 7 pages. Eckard, Regine 2006 Meaning change in grammaticalization. Oxford: Oxford University Press. Falaus, Anamaria 2010 The importance of being small. An implicature-based approach to epistemic indefinites. Paper read at SALT 20. Farkas, Donka 2002 Varieties of indefinites. Proceedings of SALT XII, 1–25. Ithaca: CLC Publication. Floricic, Franck 2005 Aspects de la ne´gation dans les langues romanes. Lalies 25: 163– 194. Gaatone, David 1971 E´tude descriptive du syste`me de la ne´gation en franc¸ais contemporain. Geneva: Droz. Giannakidou, Anastasia 1998 Polarity sensitivity as (non)veridical dependency. Amsterdam/ Philadelphia: Benjamins. Giannakidou, Anastasia 2000 Negative. . . concord? Natural Language and Linguistic Theory 18: 457–523. Giannakidou, Anastasia 2006 N-words and negative concord. In: Martin Everaert & Henk van Riemsdijk (eds.), The Blackwell companion to syntax, 3: 327– 391. Oxford: Blackwell. Haspelmath, Martin 1997 Indefinite pronouns. Oxford: Oxford University Press. Horn, Laurence R. 2001 A natural history of negation. Stanford: CSLI. Iatridou, Sabine & Ivy Sichel 2011 Negative DPs, A-movement, and scope diminishment. Linguistic Inquiry 42 (4): 595–629. Israel, Michael 2001 Minimizers, maximizers and the rhetoric of scalar reasoning. Journal of Semantics 18: 297–331. Israel, Michael 2011 The grammar of polarity: pragmatics, sensitivity and the logic of scales. Cambridge: Cambridge University Press. Ja¨ger, Agnes 2009 History of German negation. Paper read at the seminar of the project Cycles of Grammaticalization, Birmingham City Univer-

20

Pierre Larrive´e sity, July 25, 2009. Hand-out, 28 pages. Available at http://www1.aston.ac.uk/lss/research/research-projects/ summary-of-fifth-seminar/

Jespersen, Otto 1917 Negation in English and other languages. Copenhagen: A.F. Høst. Kadmon, Nirit & Fred Landman 1993 Any. Linguistics and Philosophy 16: 353–422. Kiparsky, Paul & Cleo Condoravdi 2006 Tracking Jespersen’s cycle. In: Mark Janse, Brian D. Joseph & Angela Ralli (eds.), Proceedings of the 2nd International Conference of Modern Greek Dialects and Linguistic Theory, 172– 197. Mytilene: Doukas. Kratzer, Angelika & Junko Shimoya 2002 Indefinite pronouns. The view from Japanese. In: Yukiko Otsu (ed.), Proceedings of the 3rd Tokyo conference on psycholinguistics, 1–25. Tokyo: Hituzi Syobo. Krifka, Manfred 1998 The semantics and pragmatics of polarity items. Linguistic Analysis 25: 1–49. Ladusaw, William A. 1992 Expressing negation. In Chris Baker & David Dowty (eds.), Proceedings from Semantics and Linguistic Theory, 2: 236–259. Ohio State University. Lagorgette, Dominique & Pierre Larrive´e 2007 Une re´interpre´tation contemporaine du marqueur diachronique onc. In: Denis Bouchard, Ivan Evrard & Etleva Vocaj (eds.), Repre´sentation du sens linguistique: actes du colloque international de Montre´al, 293–311. Brussels: De Boeck. Laka, Itziar 1990 Negation in syntax. On the nature of functional categories and projections. Ph.D. dissertation, MIT. Larrive´e, Pierre 1995 Quantifieurs ne´gatifs et double ne´gation. Actes des neuvie`mes Journe´es de linguistique, 207–212. Que´bec: CIRAL. Larrive´e, Pierre 2033 Variation diachronique, variation synchronique et re´seaux de polyse´mie. Verbum XXV (4): 431–442. Larrive´e, Pierre 2004 L’association ne´gative: Depuis la syntaxe jusqu’a` l’interpre´tation. Geneva: Droz. Larrive´e, Pierre 2010 The pragmatic motifs of the Jespersen cycle: Default, activation, and the history of negation in French. Lingua 120 (9): 2240– 2258.

Is there a Jespersen cycle?

21

Larrive´e, Pierre & Richard Ingham In press Variation, change and the status of negatives in peripheral varieties of Old French: The case of ne´ant. MS, Aston University and Bimingham City University. 9 pages. Martins, Ana Maria 2000 Polarity items in Romance: Underspecification and lexical change. In: Susan Pintzuk, George Tsoulas & Anthony Warner (eds.), Diachronic syntax: Models and mechanisms, 191–219. Oxford: Oxford University Press. Martineau, France & Viviane De´prez 2004 Microparametric variation and negative concord. In: Julie Auger, Clancy Clements & Barbara Vance (eds.), Contemporary approaches to Romance linguistics. Selected papers from the 33rd linguistic symposium on Romance languages, 139–158. Amsterdam: Benjamins. Martineau, France & Raymond Mongeon 2003 Sociolinguistic research on the origins of ne deletion in European and Quebec French. Language 79 (1): 118–152. Mathieu, E´ric 2001 On the nature of French n-words. In: Corinne Iten & Ad Neeleman (eds.), UCL Working Papers in Linguistics 13: 319–352. London: University College London. Meillet, Antoine 1912 L’e´volution des formes grammaticales. Scientia 12 (6) : 384–400. Menendez-Benito, Paula 2006 Exclusive choices. In: Christopher Davies, Amy Rose Deal & Youri Zabbal (eds.), Proceedings of the 36th meeting of the North-East Linguistic Society, 467–480. Amherst: GLSA. Mouret, Franc¸ois 2004 The syntax of French conjunction doubling. Proceedings from the Annual Meeting of the Chicago Linguistic Society 40 (2): 193– 207. Chicago: Chicago Linguistic Society. Penka, Doris 2007 Negative indefinites. Ph.D. Thesis, Universita¨t Tu¨bingen. Pollock, Jean-Yves 1989 Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20 (3): 365–424. Roberts, Ian & Anna Roussou 2003 Syntactic change: A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. Rohrbaugh, Eugene 2007 The role of focus in the licensing and interpretation of Negative Polarity Items. In: Danielle Forget, Paul Hirschbu¨hler, France Martineau & Maria-Luisa Rivero (eds.), Negation and polarity: Syntax and semantics, 311–322. Amsterdam/Philadelphia: Benjamins.

22

Pierre Larrive´e

Rouquier, Magali 2004 Re´flexions sur l’analyse en zones de la proposition et l’approche pronominale. Travaux de Linguistique 48 (1): 53–75. Rowlett, Paul 1997 Jespersen, negative concord and A’-binding. In: Danielle Forget, Paul Hirschbu¨hler, France Martineau & Maria-Luisa Rivero (eds.), Negation and polarity: Syntax and semantics, 323–340. Amsterdam/Philadelphia: Benjamins. de Swart, Henrie¨tte 2010 Expression and interpretation of negation. London: Springer. Terasawa, Jun 2005 Negative constructions in Old English: The question of Cynewulf ’s authorship. In: Yoko Iyeiri (ed.), Aspects of English negation, 15– 25. Amsterdam/Philadelphia: Benjamins. van der Auwera, Johan 2009 The Jespersen cycles. In: Elly van Gelderen (ed.), Cyclical changes, 35–71. Amsterdam/Philadelphia: Benjamins. Vanderheyden, Anne 2010 Dire la ‘‘singularite´ inde´termine´e’’ dans l’histoire du franc¸ais. Paper read at the conference Diachro V, ENS Lyon, October 22 2010. Hand-out, 8 pages. van der Wouden, Ton 1998 Negative contexts: Collocation, polarity and multiple negation. London: Roudedge. van Gelderen, Elly 2008 The negative cycle. Linguistic Typology 12 (2): 195–243. Willis, David In press A minimalist approach to Jespersen’s cycle in Welsh. In: Dianne Jonas, John Whitman & Andrew Garrett (eds.), Grammatical change: Origins, nature, outcomes. Oxford: Oxford University Press. Zeijlstra, Hedde 2004 Sentential negation and negative concord. Ph.D. dissertation, University of Amsterdam.

Negative words and related expressions: A new perspective on some familiar puzzles Theresa Biberauer & Ian Roberts 1. Introduction The focus of this paper is some well-known data concerning negation in (Standard) English, a non-negative concord (NC) language, Italian, a nonstrict NC language, and French. Our aim is to develop an Agree-based minimalist analysis of the relationship between indefinites and negative expressions of various kinds. The principal merit of this approach is that it integrates the analysis of negation and NC with that of indefinites, as well as clarifying the role of the Agree relation in these phenomena. In conclusion, we will also consider the potential application of this type of approach in the diachronic domain.

2. Agree, negation and indefinites In this section, we introduce our main assumptions regarding negation and related matters. Aside from the Agree relation itself, these assumptions concern the structure and interpretation of negative clauses, the nature of C and v as Probes, and the general nature of indefinites. Many recent accounts of NC rely on the Agree relation introduced in Chomsky (2000, 2001); see for example Ingham (2003, 2007), Roberts & Roussou (2003), Watanabe (2004), Zeijlstra (2004 et seq.), Haegeman & Lohndal (2010). Agree can be defined as in (1): (1) a Agrees with b where: a. a and b have non-distinct formal features; b. a asymmetrically c-commands b; c. there is no g non-distinct in formal features from a such that g c-commands b and a c-commands g. Where these conditions are met, a is referred to as the Probe and b the Goal of the Agree relation. A Negative Agree-relation is thus an instance

24

Theresa Biberauer & Ian Roberts

of Agree in the sense of (1) where the Probe and Goal share the feature Negative. Thus for g to break the Agree relation between a and b, g must be specified as Negative, i.e. g must be a negative element independent of a and b (this can be thought of as a specific instance of Rizzi’s 1990, 2001 Relativized Minimality). A further condition applies to Agree: a and b must be active, i.e. they must bear unvalued features which need to be valued before the derivation reaches the LF interface.1 Some general assumptions about the structure and interpretation of clauses, and regarding the nature of indefinites, must be articulated before we can enter into more detailed discussion of negation. First, following Acquaviva (1996: 298), we make the following assumption about the semantic nature of clausal negation: (2) Sentential negation ¼ closure of the temporal variable by a negated existential. In order to fully understand (2), we need to understand at least the rudiments of how non-negative clauses are interpreted in relation to tense and events (cf. also Ramchand 2008 for recent discussion). Consider, then, a very simple English sentence like (3): (3) John left. Following a tradition going back to Davidson (1967), we take the relevant parts of the logical form of (3) to be as in (4): (4) pept[t < to ] (AT (e,t) & leave (e, j)) Here, (4) states that ‘‘There is a time t, which precedes the speech time t o, and the event of John leaving took place at t’’. In neo-Reichenbachian terms (adopted by Ramchand (2008) among many others) we can take the clausal functional head T to provide the Reference Time which gives the restriction on the temporal quantifier, i.e. [t < t o]; we can thus, perhaps rather simplistically, take [t < t o] as the denotation of the feature [Past] (cf. Stowell 1996; see also Cinque 1999 for similar assumptions in the context of a more elaborate approach to clause structure). The Speech Time is contextually given, presumably via C, as in Enc¸’s (1987) theory of T-anchoring. The temporal quantifier is therefore associated with C (a fact which necessarily has important consequences for the interpretation 1. The Agree relation involving the Negative feature is quite similar to the notion of negative chain as adopted in Haegeman (1995) and elsewhere.

Negative words and related expressions

25

of C-less structures; cf. Martin 2001). AT may correspond to an aspectual head in the clause structure (giving a punctual reading here, as opposed to other aspectual operators one may imagine). AT relates the Event Time to the Speech Time and therefore the Reference Time; we take the Event Time to be an argument of the predicate, as in Higginbotham (1985). We can thus repeat (4) along with a specification of which clausal positions contribute which aspects of the logical representation, as follows: (5) pept [t > t o] (AT (e, t) & leave (e, j)) C T Asp/v V In these terms, simple clausal negation involves negation of the existential. This is how we understand Acquaviva’s assumption in (2).2 Now, clausal negation is not morphosyntactically realised in the Csystem in all languages, although it is in some, e.g. Irish, as (6) shows: (6) Nı´ dhearna Sea´n aon dearmad ar bith. Neg made Sean a mistake any ‘Sean didn’t make any mistakes.’ Following McCloskey (1996, 2001), the finite verb dhearna is in T in (6), and the negator nı´ is in C, so the relevant parts of the structure are as in (7):3 (7) [CP Nı´ [TP dhearna [ Sea´n aon dearmad ar bith ]]] Moreover, the phenomenon of ‘‘scope-splitting’’ discussed in von Stechow (1993), Penka & von Stechow (2001), and Penka (2002, 2007, 2010) provides evidence for an abstract negation operator in C in English and other languages. This phenomenon is observed in examples like (8): (8) Not every pearl can be above average size.

2. In terms of a split CP of the kind proposed by Rizzi (1997), the existential in (5) would be best associated with Fin (in fact, to the extent that non-finite contexts are modalised and/or irrealis, the operator in that position may be a di¤erent one; see Roussou 2000 on this). Evidence that negation must be low in the C-domain comes from the fact that negation never outscopes clausal force operators (Han 2001), which are presumably located in Force. 3. This analysis raises well-known problems concerning the position of the subject. See McCloskey (1996) for discussion in relation to Irish, and Roberts (2005, Chapter 1) in relation to Welsh.

26

Theresa Biberauer & Ian Roberts

Examples of this kind have an interpretation in which the relative scopes are as shown in (9): (9) [ NOT can [ [every pearl ] . . . . .]] With this order of scopal relations, (8) expresses the analytic truth that it is not possible for every pearl to be above average size. This reading involves scope-splitting in the sense that it appears that the modal is scopally interpreted ‘‘in between’’ not and every, despite the fact that the latter elements are generally (but not universally: see Sportiche 2005) thought to belong to a constituent which excludes the modal. Since the modal has scope over the universal, the universal has a de dicto interpretation. Following Lechner (2005), who follows Penka & von Stechow (2001), we take it that in examples like (8), not every pearl bears a feature [þneg] which requires it to appear in the local scope of an abstract operator NOT, and an interpretive convention regulates the transition from the morphologically negative QP to its contradictory. For our purposes, the crucial point is that the abstract NOT element corresponds to the locus of clausal negation in C, i.e. NOT corresponds to the negator of the clausal existential of Acquaviva’s analysis in (2). A further general assumption, following Chomsky (2008), is that phase heads (C and v in the clause) are the only true (inherent) Probes. Following Ouali (2008), we propose that C’s features may relate to those of T by means of three parametrised operations: KEEP (C retains its features), DONATE (C’s features transfer to T) and SHARE (C’s features are shared between C and T). Given this, we expect that negation can in principle be morphologically realised in any one or both of two positions in the C-phase, C or T (where C and T may be thought of as cover-terms for two more articulated domains of functional structure); this expectation is borne out (see below). Additionally, we see that negation can be morphologically realised within the vP-domain (cf. Germanic). We take this to mean that phase heads within the same domain (clausal, nominal, etc.) may – possibly, must – also SHARE features (although the other feasible parametric option, DONATE, would not be available in this case, given the semantic representation underlying sentential negation discussed above). In some languages, then, only the SHAREd features within the vP-domain are overtly realised. As just mentioned, Irish is a case of negation being directly realised in C, and the same is true for Modern Greek (cf. Roussou 2000). English, in turn, is a typical case of negation in the T domain, while German is

Negative words and related expressions

27

an instance of negation being realised in the lower part of the clause. The three cases are illustrated in (10a), (10b), and (10c) respectively: (10) a.

b.

Dhen tha dhiavaso to vivlio. not prt read-1s the book ‘I will not read the book.’

[Greek: Neg in CP]

I will not read the book.

[English: Neg in TP]

c. Ich werde das Buch nicht lesen. [German: Neg in vP] I will the book not read ‘I will not read the book.’ This shows a simple case of morphosyntactic variation in the realisation of a constant logico-semantic property. In Greek, negation is realised directly in the position of the operator, C (the features are KEPT in C); in English (and Romance) the features are DONATEd to T (we will say more about this in the next section); and in German, the features are SHAREd by C and v.4 We now turn to the question of the representation of indefinites, which will lead to a clearer picture of how the Agree operation may be involved in licensing NPIs and determining the interpretation of indefinites in certain contexts. Following Kamp (1981) and Heim (1982), we assume that indefinite DPs should be treated as intrinsically lacking in quantificational force, i.e. as denoting free variables. Simple indefinites, such as those marked with the indefinite article, can be unselectively bound in various ways, including by existential closure in the sense of Heim (1982). NPIs, on the other hand, must be bound by operators of a particular semantic class (for our purposes, this class can be defined as either downward-entailing or non-veridical). We tentatively take negative indefinite expressions such as nobody, nothing, etc., to have intrinsic negation, in a sense to be clarified 4. There is evidence that in some West Germanic languages (e.g. German, Dutch, West Flemish and Afrikaans) the overt negation is in fact lower than the left edge of vP (inter alia Bayer 1990; Haegeman 1995; Weiss 2002; Barbiers 2002; Biberauer 2008). In that case, given that we take it that C-level negation is not in the highest C-position (Force), but rather a lower one (Fin; see Note 3), there is an interesting parallel regarding the position of negation in the two clausal phases.

28

Theresa Biberauer & Ian Roberts

below, although we will ultimately see reasons to assimilate them to the Romance n-words in certain respects. Let us now consider how the Agree relation may be relevant to understanding the ways in which indefinites interact syntactically and semantically with clausal operators. As noted above, NC and perhaps NPI licensing too appear to lend themselves to Agree-based analyses (see Zeijlstra 2008 for recent discussion). In the case of NC, this is simply due to the fact that it is a concord relation, in which a morphosyntactic negative feature appears to be ‘‘spread’’, in the manner of agreement marking more generally, to other elements which are arguably not bearers of true, semantic negation. In the case of NPI-licensing, there is a clear sense in which the NPI is a dependent element which, without entering into the appropriate relation with something else, cannot be interpreted. Furthermore, both NC and NPI licensing are local relations. Attractive though the Agree approach is, there are three serious problems with it. First, the putative Agree relation appears to hold in an ‘‘inverse’’ fashion. This can be seen if we compare a ‘‘standard’’ case of Agree of the kind discussed in Chomsky (2001) with the case of NPIlicensing. Consider (11) in this connection: (11) a. b.

v*[uj]

. . . [VP V DP[ij, uCase] ] not [iNeg] . . . [VP V anyone[uNeg]]

In the standard case of Agree between v* and the direct object in (11a), the Probe bears unvalued agreement-/j-features (i.e. person, number and gender features) which make it active, initiating the Agree relation. The Probe consequently searches its c-command domain for an active element (i.e. one associated with one or more unvalued features) bearing the valued counterparts of its unvalued features. In the case of (11a), the jProbe on v* will encounter the direct object which bears the requisite unvalued feature to render it an active Goal, namely an unvalued Case. The Agree relation then values v*’s j-features and the object’s Case feature. In (11b), on the other hand, the putative Probe not has no unvalued feature and therefore, strictly speaking, cannot initiate the Agree relation. The NPI, on the other hand, has an unvalued feature, but is c-commanded by the element with the corresponding valued one; in this sense, a putative Agree relation involving these two elements would have to be an ‘‘inverse’’ one. Consequently, it is ‘‘inverse’’ rather than standard (1)-type Agree that features in the Agree-based analyses of negation-related phenomena in Zeijlstra (2004 et seq.), Penka (2007, 2010), Roberts (2007), Ja¨ger (2011),

Negative words and related expressions

29

and, in part, Haegeman & Lohndal (2010). Second, the features postulated in Agree-based accounts of NC and NPI licensing should clearly have a certain semantic plausibility and it should therefore be possible to distinguish them as valued and unvalued in a plausible way. Consider, for example, the j-features associated with v* and the direct object in (11a): that the nominal phrase should be associated with valued j-features, while v* bears corresponding unvalued features seems semantically plausible as it is usual to think of verb-object agreement relations as being directional, with the properties of the nominal determining the marking that appears on the verb (a marking which is consequently also referred to as agreement marking). It is, however, not clear how this applies in the case of a feature such as [Neg]: what could plausible values of such a feature be? Third, if phenomena such as NPI-licensing and NC are to be related to one another, there should be some relationship between the Probe and Goal which parallels that between j-features and Case (cf. George & Kornfilt 1981 on the naturalness of postulating a relation between agreement features and Case). Here we will sketch an approach that meets the requirements just noted, while avoiding the problems usually associated with Agree-based analyses. In order to show how this is possible, however, the internal phasal structure of DP first needs to be clarified. We take the nominal to have at least the following structure:5 (12) a. [DP D [ NumP Num [nP n [NP N .. ]]]] b. [DP [[ D the] [NumP three Num[þpl] [nP [n [N boys]- n ] [NP (N) .. ]]]] This clearly parallels the simplest version of clause structure, as assumed in Chomsky (2000, 2001) and elsewhere. Both structures divide into two phases: a ‘‘lexical’’ phase, nP/vP, and a ‘‘discourse-related’’ phase, CP/DP (cf. Chomsky’s ‘‘duality of semantics’’). As observed in Note 6, the n/v head controls the realisation of the lexical argument structure of the lexical head (which may have no intrinsic category; cf. Marantz 1997), while D/C controls the interaction between the whole category and external forces (Case, agreement, and ‘‘external’’ selection properties, along with discourse (speech act, definiteness) properties).

5. We assume that the lexical head N moves to the semi-functional category n here, by analogy with V-to-v movement in clauses. Following Marantz (1997), the categories n and v can be thought of as nominalising and verbalising categories respectively.

30

Theresa Biberauer & Ian Roberts

Our approach to the licensing of di¤erent kinds of indefinites is based on two central ideas. The first is that the DP phase contains an internal Probe-Goal pair (Num, located within the DP phase itself, and n, located at the edge of the nP phase)6 and additionally also contains a further Goal (D) which interfaces with an external Probe (cf. Haegeman & Lohndal 2010 for an independently arrived at, but in part similar analysis). In the case of j-feature Agree, for example, this can clearly be seen in languages with more morphological marking than English, e.g. French: (13) Jean voit [ DP la vieille femme] ‘John sees the old woman.’ Here we have the usual Agree relation between v* and the object DP, or more precisely its D head (see (11a)). The features of the Goal D value the j-features of v*. At the same time, Num probes for the number and gender features of the nP, vieille femme (if only phase heads are true Probes, then we must add that Num inherits its probing features from D). Our second idea is that the Agree relations, both internal and external to DP, which are related to licensing features whose content involves negation and quantification, create relations which can be construed as variable-binding. So, we propose that, inside DP, D binds a variable associated with the nP (the lower phase).7 This relation corresponds to the semantic relation which provides the restriction on the quantifier, conventionally written as Q(x) . . . N(x), as in the usual denotation for an expression such as Every boy thinks as every(x)boy(x) [think(x)], where every ¼ Q and boy ¼ N. For strong quantifiers, there may be no more to say, the nuclear scope being provided in some way independent of Agree. 6. According to the so-called Phase Impenetrability Condition (PIC) of Chomskyan phase theory, a Probe is only able to ‘‘see’’ (and thus probe) elements located within its own phasal domain (thus NumP in our case) and elements located at the edge of the next phasal domain down (thus nP in our case); the complement of n (NP) is opaque to probing by D and Num, but N-to-n raising renders N accessible to Probes from the higher DP phase. 7. Note that we are not assuming a coindexing relation between D and N here. The variable-binding interpretation arises as a result of the combination of the intrinsic quantificational content of D and the Agree relation between D and N (which consists in partial feature-identity). Beyond the standard assumptions that D may be quantificational and that Nouns may contain an open variable, we are making no claims here about the denotations of the heads that make up the nominal system, as this would go well beyond the scope of this paper.

Negative words and related expressions

31

Strong D has su‰cient intrinsic content to determine its quantificational force and therefore to make the Q(x) . . . N(x) relation fully interpretable at the semantic interface. On the other hand, we assume that this is not the case at least for certain kinds of indefinites, and perhaps for weak quantifiers generally. In these cases, D lacks su‰cient intrinsic content to determine the interpretation of Q(x) . . . N(x) on its own. Therefore, these Ds must instead function as a Goal for a DP-external Agree relation. Thus, some element outside DP has to Agree with the D in order for the internal structure of DP to be interpretable at all (one could think of this as a relation binding a second-order variable). Given established constraints on binding, it clearly has to be the case that the element in question must be part of the ‘‘A’-system’’, the system of sentential operators.8 In Agree terms, we can designate the relevant feature of D as an unvalued Operator feature, or [Op:__], in attribute-value terms. This feature requires some element of the A’-system to license it. So, schematically, we have an Agree relation of the following type: (14) C/T[Op:Val] . . . D[Op :__] A rough, non-exhaustive characterisation of the possible values of Op in (14) includes the interrogative feature Q of yes/no questions, conditional features, perhaps certain irrealis mood features and negation (we will briefly comment on wh-questions below). What does it mean to say that these features Agree with D? Intuitively, it does not seem to make sense to postulate, for example, that a weak D in the scope of an interrogative operator is itself interrogative. This is for the obvious reason that DPs cannot be interrogative (leaving aside ‘‘hidden questions’’ as in John asked the time, which probably have a more complex structure; and, for selfevident reasons, wh-elements). However, sentential operators can be classified according to their semantic properties. We propose that the most important of such properties is veridicality (as argued extensively by 8. In spirit, this proposal is similar to that which Szabolcsi (2004) proposes for PPIs. Like Szabolcsi, we assume that the licensing of ‘‘weak’’ DPs requires more than one licensor: j to bind n DP-internally (an operation which may give rise to an existential interpretation as default) and, additionally, a DPexternal licensor. Where PPIs surface as part of DPs, then, we expect these DPs to have the (dependent) structure of non-islands. This seems to be correct: (i) (ii)

He bought some books about Darwin. Who did he buy some books about __?

32

Theresa Biberauer & Ian Roberts

Giannakidou 1998). Giannakidou’s (1998: 106) definition of veridicality is given in (15): (15) Let Op be a monadic propositional operator. The following statements hold: a. Op is veridical just in case Op p ! p is logically valid. Otherwise Op is non-veridical. b. A non-veridical operator Op is antiveridical just in case Op p ! Bp is logically valid. So, for example, the interrogative operator Q is non-veridical, in that Is it raining? entails neither that it is nor that it is not raining. By contrast, alethic modality is veridical, while negation is clearly anti-veridical, and deontic modality, for example, is, again, non-veridical. We suggest that the ‘‘A’-Agree’’ relations between operators in the C/T system and weak Ds concern the syntactic reflex of a veridicality feature. For D, the value of the veridicality feature is interpreted in relation to the existential entailments of the DP, i.e. D’s content, determined by Agree, influences the interpretation of the default existential quantification inside DP. The following examples illustrate: (16) a. b.

John saw someone. Did John see anyone?

In (16a), there is a positive declarative Operator, which is semantically veridical, and which binds a variable in the DP someone. More precisely, assume that C contains an Operator that has a [declarative] feature. This feature is morphologically unrealised in English, but is realised as an a‰rmative marker in many languages, e.g. Welsh, see Roberts (2004). The declarative feature is inherently veridical. This Operator, OpDecl, then Agrees with the D of someone, i.e. some. This lexical item is inherently specified as a PPI, i.e. as requiring binding by a veridical operator (see Szabolcsi 2004 and Note 9).9 The required binding is instantiated by the Agree relation between OpDecl and D; this gives a particular value to 9. In fact, Szabolsci argues that PPIs need to be bound by two NPI-licensing operators. If these are anti-veridical operators, then the combined e¤ect of these operators on p would be BBp, equivalent to p and so to a veridical operator. We leave aside here the question of whether this is true in every case of PPI licensing. For our purposes, the crucial point is simply that PPIs are licensed, at least in part, by a relation with what we construe as an operator in C.

Negative words and related expressions

33

D, i.e. the value of existential import, which allows this D to be interpreted as an existential quantifier. D is then able to bind the variable in N. We can represent these Agree relations as follows: (17) [C OpDecl] . . . [DP D[p](x)one(x) ] Agree1

Agree2

Agree1 is the clausal A’-relation between the declarative OP (OpDecl) and D, brought about by the inherently active nature of the Operator and the unvalued feature on D.10 Agree2 is the binding of the argument position in N, brought about by D’s intrinsic property as an operator combined with the fact that N contains an open variable. The combined result of the two Agree relations is that the sentence is interpreted as veridical and the D is interpreted as existentially quantifying over a set of existing individuals. An obvious question is what happens in examples like (18), where there is no weakly quantified DP for the operator in C to probe: (18) John saw Mary. Here we clearly want to maintain that there is a declarative, veridical operator in C. But what does the operator Agree with here? The natural answer appears to be the event variable, in terms of the logical forms containing event variables of the type originally proposed by Davidson (1967) and given in (4) and (5) above. This can be captured by assuming that the

10. The idea that Probes bearing Operator features are inherently active may seem stipulative, but there is a sound conceptual justification for it. If, as mentioned above, Agree relations can create operator-variable relations, and if, as is often assumed, vacuous quantification is not allowed in natural language, then probes bearing an Operator feature must be inherently active since they are required to bind a variable, i.e. Agree with a Goal. This idea has an interesting implication for the operation of Agree in the A-system. It is standardly assumed that bearing (unvalued) structural Case makes a DP an active Goal (cf. Chomsky 2001). Extending the operator-variable idea to the A-system, we can think of the Case feature as indicating that a DP has an open variable position, i.e. it is an argument. We can then relate the presence of the Case feature to a general Visibility Condition on arguments, ultimately to the y-criterion. (Of course, this line of reasoning founders in the case of expletives, but this shows that the system is driven by formal, not semantic, considerations, in that we have to allow for variables which have no denotations here. This does not, however, mean that the formal relations do not generally have semantic counterparts.)

34

Theresa Biberauer & Ian Roberts

operator in C Agrees with the event argument, which we take to be located in v. Of course, this must also happen in examples like (16a), suggesting that there are in fact three Agree relations here: between Op in C and v, between v and D, and between D and N. Things are still more complex if we take Tense into consideration, as in (4) and (5). An example like Someone left would, in these terms, have the logical form (19): (19) ptpepx [t > t o] (AT (e,t) & leave (e,x)) The three existential quantifiers are the instantiation of the Agree relation which originates with the operator in C. In a sense, then, we have a phenomenon of ‘‘existential spreading’’, caused by iterated Agree. We will leave these further complexities aside here, as what we are interested in is the relation between the A’-operator and the weak quantifiers in D, although we can see that this relation is somewhat indirect. Let us now turn to (16b). Here there is a Q-operator in C (which in English root clauses has the property of triggering subject-auxiliary inversion, instantiated here by did-raising; again, in many languages a special morpheme appears in this context). This operator is semantically nonveridical, as we have seen. It probes the D any, which has the lexical property of being an NPI (leaving free choice any aside). In Agree terms, this means, following Giannakidou’s essential insight, that any must be in the scope of a non-veridical operator, which in this example it is. With the single di¤erence of the feature of the operator, Q rather than Decl, we arrive at an LF just like (17) (bearing in mind that ‘‘Agree1’’ is a cover term for up to three Agree relations). Again, D is interpreted as an existential, as in (16a). The existential is inside the scope of the Q-operator, and therefore, since the Q-operator is non-veridical, there is no existential entailment. This contrasts with the wh-question in (20): (20) Who did you see? Here the existential (denoted by wh; see Beck 1996a,b) is outside the scope of Q and so the entailment holds. In some languages, e.g. Mandarin, the di¤erence between (16b) and (20) is conveyed purely by di¤erent sentential particles, suggesting that these are associated with di¤erent features and perhaps therefore determine di¤erent scope for the existential. We can go a step further in our analysis of any-DPs, and thereby capture the nature of their dependency on their licensers. We have said that any is the Goal of an Agree relation with the operator in C. Roberts (2010a) introduces the notion ‘defective Goal’, which can be defined as follows:

Negative words and related expressions

35

(21) A Goal G is defective if G’s formal features are a proper subset of those of G’s Probe P. Owing to the fact that the features of defective Goals are subsumed in those of their Probes, where such a Goal values features of its Probe, the resulting structure is one where the Goal appears as a copy of (a subset of the) features of the Probe. In this case, then, the copy-spell-out mechanism which in other cases spells out just the highest copy in a movement dependency and deletes all others treats the features of the Goal valued on the Probe as an element incorporated in the Goal. In Roberts (2010a) this idea is employed in an account of cliticisation (and head-movement more generally); taking Romance complement pronouns to be js rather than Ds, due to their generally defective nature (see Cardinaletti & Starke 1999), these elements are defective Goals in relation to transitive v*, since their feature content is simply (valued) j-features, which are also present on v*. After valuing the j-features of v*, these features are spelled out as a clitic incorporated in v* and the copy is deleted at PF. With this in mind, it may be desirable to treat NPI any as a defective Goal. Suppose that any is not itself the D-element of an any-DP, but rather occupies Num in terms of the structure in (12), with D being phonologically unrealised. This D element has an Op feature, the Goal, which is shared with the Probe (the negative operator in T). It has no other feature; hence it counts as a defective Goal, and incorporates with the Probe. Although this case of incorporation is not directly morphologically visible as D-to-Neg movement as, by assumption, the D is phonologically null, it has certain indirect e¤ects. The relevant parts of the structure are: . . . [T didn’t-Op1] . . [VP see [DP Op2 anything ]] . . , where Op2 undergoes head-movement to Op1. The incorporation operation has consequences for the distribution of NPI any, given the following corollary of the account of incorporation and defective Goals put forward by Roberts (2010a): (22) A Probe P can act as an incorporation host for a Goal G only if P lacks an EPP-feature capable of attracting G. The reason for this is that an EPP-feature on the Probe causes the Goal to move to the specifier of the Probe, but if the Goal is defective, it will have to incorporate. Assuming these two options to be incompatible,11 then, we can see why NPI any cannot appear in subject position with a negative

11. See Roberts (2010b) for a di¤erent view.

36

Theresa Biberauer & Ian Roberts

auxiliary in T (as in *Anybody didn’t see Mary): negative T is the incorporation host for the D of the any DP, and, as such, that DP cannot move to satisfy T’s EPP-feature. If an expletive satisfies the EPP-feature, by contrast, the sentence is grammatical:12 (23) a. b.

*Any students weren’t arrested. There weren’t any students arrested.

More generally, to the extent that incorporation is an e¤ect of the copyspell-out operation, NPI-licensing will not be amenable to reconstruction: (24) a. b.

*Pictures of any supermodel, we didn’t sell. We didn’t sell pictures of any supermodel.

Finally, consider a sentence containing a negatively quantified DP: (25) John saw no-one. Here, the D clearly contains a semantic negation. But aside from this, it has no inherent quantificational force; it is a weak quantifier. Hence the usual Agree relations hold. The Agree relation with the Decl operator gives existential force to the quantifier. The result is a veridical sentence which asserts a negative: hence the entailment of non-existence of the object of see. The foregoing is just a sketch of some aspects of the syntax and interpretation of indefinites within an Agree-based theory. Clearly many questions are left open, but the essential points are covered here. The key ideas for our purposes are: (a) that weak quantifiers act as Goals for an Agree relation with a higher operator, and (b) that sentential operators in C vary in veridicality, something to which the bound Ds are sensitive. Having seen the basic assumptions about the realisation of negation, let us apply these ideas more concretely to English. 12. In (23b) it is likely that any students undergoes A-movement from a position following the participle ( pace Chomsky 2001; see Collins 2005). Note that, in this case, the Probe (v) lacks the negative operator and therefore the operator D of the any-DP is not a defective Goal in relation to it; no incorporation can therefore take place here. It is important to see that the notion of defective Goal is a relative one, since it depends on the set-theoretic relations between the feature sets of the Probe and the Goal.

Negative words and related expressions

37

3. Negated arguments in English Negated arguments are realised by an Agree relation between CNeg and D, interpreted as ‘‘for no x . . . x’’. Suppose that in order to negate a clause, a negative feature must be introduced and that that negative feature must have at least one morphological realisation, which we mark as X* (X the position where the feature is realised, * the diacritic for PF-realisation; see Roberts & Roussou 2003). Then we have the following possibilities:13 (26) a. (Neg, D*) b. (Neg*, D) c. (Neg*, D*) d. *(Neg, D) We can consider these options in relation to the following English sentences: (27) a. I saw nothing. b. I didn’t see anything. c. I didn’t see nothing. d. *I saw anything. Let us now consider these one by one. In (27a), negation is not realised as a separate morpheme in the clausal system, but on D – this is thus a case of (26a). In other words, C is a declarative operator and D (no) is interpreted as Bpx. The Agree relations in (27a) are as follows: (28) [C OpDecl] . . .

[vP v/V(e)

[DP D[Bp](x)

thing(x) ]

(27a) receives the interpretation: (29) B[pxptpe[PAST(t)](AT (e,t) &[ see (e,I,x) & [thing(x)]))]] C T Asp/v v/V NP The existential binding of the temporal variable t is introduced in C as part of the tense/mood specification of the clause; the existential binding of the event variable is related to the aspectual specification of the clause (although in the case of punctual past tense as in (27a), both of these 13. Note that the asterisk in (26d) is the usual asterisk representing ungrammaticality of the sentence to its right, unlike the occurrences in (26a–c), which represent Roberts & Roussou’s notation for overt morphophonological realisation.

38

Theresa Biberauer & Ian Roberts

might result from existential closure). The existential binding of the variable introduced by the noun thing results from the Agree relation. We indicate the negated existential associated with no as having wide scope; in (29), this has no consequences since the only other operators are existentials (note that there is no need for a QR-type operation here). Below we will suggest that no-DPs in Standard English may be associated with a null focus operator which can have the e¤ect of giving wide-scope interpretations to no. In (27b), clausal negation is overtly realised as n’t (on the nature of negative auxiliaries in English, see Biberauer & Roberts 2010). This element combines with anything to form a single negation. So this is an example of (26b): (Neg*, D). Any, on the other hand, is not intrinsically negative, as its appearance in non-negative environments shows (e.g. Did you see anyone? etc.). But the any-DP is a weak DP, and in fact one which imposes a non-veridicality requirement on the operator which binds it. This requirement is met since C is negative, although in English, as we saw above, the morphological realisation of negation is DONATEd to a negative auxiliary in T. The Agree relations minimally contrast with those in (24), viz.: (30) [C OpNeg] . . . [vP v/V(e)

[DP D (x)

thing(x) ]

Clearly, (30) can give the same logical form as (28), namely (29). (27c) can only be interpreted as double negation in Standard English. Superficially, a natural account of this fact would appear to be that this is because no realises a negative D and not/n’t a negative T: where not/n’t is realised in a higher structural position, one might then expect that the no-DP cannot be interpreted as part of the same Agree relation (by the non-intervention clause in the definition of Agree in (1iii)); consequently, no should be interpreted as introducing an autonomous negation, giving the observed double negation interpretation. Furthermore, the fact that there are varieties of English where (27c) is grammatical as single negation may suggest that these varieties di¤er from Standard English in featuring no-DPs which, despite their overt morphology, do not realise a negative D. Consequently, no-DPs in these varieties can Agree with CNeg in a way not dissimilar to any-DPs, delivering a single-negation interpretation. We would, however, like to suggest a more appealing and, we believe, more elegant alternative, which allows us to assimilate the analysis of no-DPs to that of Italian n-words to be given below and which also has important diachronic implications. This builds on an observation first made in

Negative words and related expressions

39

Haegeman & Zanuttini (1996) about the role of focus in NC systems (cf. also inter alii Haegeman & Zanuttini 1996; Corblin 1996; Sura´nyi 2006; Falaus 2007; Rooryck 2008; Biberauer 2009; Biberauer & Zeijlstra 2009, 2010, 2011, for recent discussion). The proposal is that two negative elements in Standard English necessarily result in double negation as a consequence of the fact that one of these elements is obligatorily focused, with the result that it is e¤ectively ‘‘sealed o¤’’ from the rest of the clause. The obligatory focus requirement, which also holds for NC languages where speakers wish to produce doublenegation structures (cf. the references cited at the end of the previous paragraph), can plausibly be thought of as a phonological reflex of the fact that double negation structures are necessarily heavily discourse-marked (cf. Horn 1989; Zeiljstra 2004; Biberauer & Zeijlstra 2010b): in the ordinary course of events, multiple negation-containing structures where each negative contributes its own negation fall foul of pragmatic principles like Grice’s (1975) Maxim of Manner, which entails the avoidance of obscure, ambiguous, unduly prolix and poorly organised expressions; consequently, they are mostly avoided, with the exceptions being circumstances which warrant the additional processing e¤ort (cf. inter alia Sperber & Wilson 1986 et seq. for detailed discussion). In the case of double negation structures, we can think of the need for additional processing e¤ort being clearly signalled by focus intonation (cf. Liberman & Sag 1974 on the so-called contradiction contour associated with double negation structures, and also Corblin 1996). Following Aboh (2010), we take it that double negation structures, just like any other discourse-marked structure, derive from a Numeration which contains appropriately specified lexical items, including information about their focus, topic, etc. status. For a well-formed double negation sentence, this implies that one of the occurrences of negation in a double negation Numeration will be associated with focus.14 14. Double negation-containing structures in which something other than a negative element is focused are possible, as is illustrated in (i): (i)

A: John really is the limit: he greeted NO-ONE last night! B: He didn’t GREET no-one; he NOTICED no-one.

Crucially, B’s response is only well-formed in response to a previous utterance in which no-one is explicitly mentioned and that mention, moreover, is focused. Contrast the felicity of A’ as the prelude to B: A’: John really is the limit: he didn’t greet ANYONE last night! This parallelism requirement clearly directs the hearer to an independently negative interpretation of no-one in B, giving the observed double negation

40

Theresa Biberauer & Ian Roberts

Unlike in languages with lexicalised focus and topic markers (e.g. Gungbe or Japanese), ‘association with focus’ in English can be thought of as entailing the addition of a focus feature or focus ‘‘shell’’ to the negative quantifier that derives from the Lexicon (in general terms, we can think of this as just another form of the ‘‘full inflection’’ discussed in Chomsky 1995). (31) illustrates the case of a double negation structure in which the no-DP is focused: (31) a. b.

He didn’t greet NO-ONE. [C OpNeg] . . . not/n’t [vP v/V(e)[Foc OpFoc [DP DNeg (x) thing(x)]

Here no-one is associated with a (prosodically realised) local focus operator, which gives rise to the absence of negative concord in Standard English. We treat this as an extra projection above the DP, giving the structure [ FocP [ DP . . . ]] (cf. inter alia Ihsane 2008, Troseth 2009, and Biberauer 2009). The focus operator in Foc, being part of the system of A’-operators and thus counting as an intervener, blocks the Agree relation between the no-DP and the clausal negation in T, or indeed Agree with any negation external to FocP. This means that the no-DP is associated with a separate Agree relation from the clausal negation, and hence a double negative interpretation necessarily results wherever focused no-DPs co-occur with other negative elements (sentential negators or other no-DPs). The same is true wherever the sentential negator is the focused element (cf. He DIDN’T greet no-one). structure. From a formal perspective, one way of understanding this would be to view it in a way similar to question-answer pairs in languages that make use of verbal forms to express positive answers, another polarity-related context (cf. Holmberg 2001, 2003 and also Aboh 2010 for discussion). In these cases, answers necessarily require identical replication of the declarative counterpart of the associated question, part of which then undergoes elision. In the present case, this identical replication requirement would clearly entail that no-one is in fact associated with focus in B just as it evidently is in A and in doublenegation structures more generally; it is only the presence of a further (contrastive) focus on the predicate that leads to the already-mentioned focus being ‘‘downgraded’’ in the context of the response. Again, then, we see that double negation results from the presence of focus structure. This structural identity approach also allows one to draw a parallel to denial contexts, another one in which double negation structures surface (cf. Horn 1989 for discussion).

Negative words and related expressions

41

In NC varieties, by contrast, there is no absolute requirement for co-occurring negative elements to be focused, with the result that all the negative elements in a given structure may Agree with CNeg , as illustrated in (32):15 (32) a. b.

He ain’t greeted no-one. [C OpNeg ] . . . ain’t [vP v/V(e)

[DP DNeg (x)thing(x) ]

Here, then, we have a structure in which the clausal phase-heads can successfully probe the negative elements contained in the clause in the manner outlined in Section 2. This results in a single Negation-oriented Agree relation (a single Negation Chain, in the terminology of Haegeman 1995). Worth noting here is that this perspective on no-DPs crucially entails that these elements resemble any-DPs in requiring binding by a clause-related operator, although this does not have to be a non-veridical operator; they, however, di¤er, from any-DPs in that D contains a negative existential. This means that they are not defective Goals, and are therefore unable to incorporate with the Probe. The negative existential imposes a negative interpretation on the operator-variable relation when no other negation is present, along the lines described in relation to (27a) above: here, then, we have an Agree relation in which the clausal (operatorbearing) Probes (C, T) probe a single overtly realised negative element. If one or more further negation elements are present, they can also be probed by the clausal Probes, with the result that there is once again a single negation-oriented Agree relation, yielding a single Negation reading. In our terms, then, the di¤erence between Standard and NC varieties of English lies in the obligatory Agree-compromising focus requirement associated with multiple-negation-containing structures in the former (in the latter, it is merely an option). We briefly return to the diachronic implications of this proposal in Section 5. Returning to Standard English, (27d) is an example where the requirement for binding of an any-DP by a non-veridical operator in C is not met, since unspecified C is interpreted as declarative (and hence veridical) 15. Recall that structures featuring two or more negative elements which each contribute their own semantic negation are pragmatically marked. By contrast, the same is not true of structures containing two or more negative elements which do not each contribute their own semantic negation, i.e. NC structures. The absence of any kind of marked intonation pattern, which characterises these structures, is therefore to be expected.

42

Theresa Biberauer & Ian Roberts

by default. Since any cannot be appropriately bound as the consequence of an Agree relationship, the sentence is ungrammatical. To summarise, we make the following claims about Standard English and non-NC varieties of English more generally: – C DONATEs its interpretable Neg-feature to T; not/n’t is the spellout of this DONATE operation. – no-DPs (nothing, no-one, etc.) have a negative existential in D, which is, in the absence of a Focus ‘‘shell’’, bound by the Op in C as a reflex of Agree. – Multiple no-DPs or no-DPs and not/n’t do not enter into NC relations because these structures necessarily involve Focus in Standard English. Focus is realised as an Operator in a left-peripheral DP Focus shell, and hence, by the non-intervention clause in the definition of Agree, e¤ectively ‘‘seals o¤’’ the DP from the rest of the clause, making it unable to enter into Agree relations with the operator in C, and thus resulting in multiple negation/the absence of NC. Having seen how the basic ideas work in English, let us now turn to Romance languages.

4. Negated arguments in Romance languages Before turning to French, let us look at Standard Italian. The reason for doing this is that Standard Italian is a good example of a (non-strict) NC language. This is shown in (33): (33) a.

b.

Non ho visto nessuno. not I-have seen no-one ‘I haven’t seen anyone.’ Nessuno (*non) mi ha visto. no-one not me has seen ‘No one has seen me.’

c. Non mi ha visto nessuno. not me has seen no-one ‘No one has seen me.’ d.

*Ho visto nessuno. I-have seen no-one

Negative words and related expressions

43

As in English, CNeg DONATEs its negative property to T in standard Italian, where it is realised as non. This entails that non lexicalises a valued (and thus interpretable) negative feature, as also proposed in Zeijlstra (2008), and Biberauer & Zeijlstra (2009, 2010, 2011). The main point of interest here is, however, the characterisation of the elements in the class exemplified by nessuno, the n-words (Laka 1990, 1994). We suggest that n-words are like no-DPs in NC varieties of English – and, thus, like unfocussed no-DPs in Standard English. This means that the Probe in C can enter into a Multiple Agree relation with the various n-words in Italian, when more than one is present, and this gives rise to NC. LF interprets Multiply Agreed-with n-words as involving a single negation (see the discussion of quantifier absorption in relation to Agree below), just as it would interpret the various Agreed-with elements in a DP as involving only one interpretable entity. We believe that there is therefore a very good case for Multiple Agree playing a role in NC languages, whereas it is not able to do so in non-NC systems like Standard English because of the intervening FocusPs. For us, then, Multiple Agree in the domain of Negation is what gives rise to Negative Concord (cf. also Zeijlstra 2004, 2008 for a di¤erent implementation of this idea; pace Haegeman & Lohndal 2010). If Italian n-words are like English no-DPs in that their D contains a negative existential, we expect them to be non-defective Goals and therefore unable to incorporate with their Probe. Like other indefinites, Italian n-words have an unvalued Operator feature, which makes them active Goals. The negative feature, which is independent of the operator feature, has various e¤ects. First, it imposes a negative interpretation on the operator-variable relation where no other negation is present.16 Second, it can act as an intervener in certain cases, as we will see in the discussion of French below.

16. Actually nessuno, like any, can appear with a non-negative interpretation in other polarity-licensing contexts (Rizzi 1982: 122): (i)

Mi chiedo se Gianni abbia contattato nessuno. Myself I-ask if Gianni have(subjunctive) contacted anyone ‘I wonder whether Gianni has contacted anyone.’

(ii)

Mi chiedo se nessuno abbia contattato Gianni. Myself I-ask if anyone have(subjunctive) contacted Gianni ‘I wonder whether anyone has contacted John.’

44

Theresa Biberauer & Ian Roberts

In (33a), then, there is a non-veridical operator in C and this element Agrees with the D of nessuno. This, like Standard English examples (27a,b) and non-standard English (27c), gives rise to the logical form in (29). In (32b) we clearly see the contrast with English NPIs. Nessuno is able to appear in subject position, since it is not a defective Goal. So it can satisfy T’s EPP-feature, like any other subject (although this feature is optional in Italian, see Note 18). Here nessuno Agrees with the nonveridical operator in C. However, as (33b) shows, nessuno is only allowed in subject position provided there is no independent expression of clausal negation. We can account for this if we assume that the feature-inheritance relation between C and T which DONATEs C’s negation features to T is subject to a non-intervention condition, like Agree; this seems natural to the extent that inheritance is related to selection (see Chomsky 2008) and selection is seen as a species of Agree (cf. inter alia Pesetsky & Torrego 2007; Rizzi 2008; Biberauer, Holmberg & Roberts 2009; Cecchetto & Donati 2010; in fact, selection can be thought of as vacuously subject to a non-intervention condition, being intrinsically too local an operation for intervention to ever make a di¤erence). In that case, since nessuno occupies the subject position (SpecTP), it asymmetrically c-commands the normal position of non, which is T. Therefore nessuno blocks the inheritance of C’s negative features by T; therefore they are not realised by non. The fact that nessuno has a negative feature is enough to satisfy the above-mentioned constraint that some element in a negative clause must overtly realise negation, with the fact that nessuno is located in SpecTP satisfying the Italian-specific constraint that negation must be realised in the T-domain in (standard) Italian.17,18 Italian non can be seen as a form that is spelled out at PF if and only if there is no accessible n-word subject As these examples show, there is no subject-object asymmetry here. It is clear that þWh se licenses nessuno. The same is true of at least personne and aucun in French (Martins 2000: 200, 218–219). For speakers who accept examples of this kind, then, we have to treat nessuno’s negative feature as optional, or more plausibly, we have to assume that they have two nessunos in their mental lexicon, one of which is an n-word and the other of which has in fact become an NPI (cf. Ja¨ger (2011) for discussion of cases which have undergone this change towards ‘‘decreasing negativity’’). 17. This account of (33b) raises a number of technical questions concerning how inheritance interacts with movement. Chomsky (2008) assumes that C is merged before movement to SpecTP is triggered, the order of operations being Merge C > inheritance > Move. This derivation violates the Extension Condition, since movement to SpecTP does not extend the root if C has already been merged. Moreover, if move follows inheritance, then the raised subject

Negative words and related expressions

45

which can inherit C’s Neg-features. Italian non thus di¤ers from English not/n’t in being a PF spellout element, rather than one which is always present in the Numeration so that it can be merged at the appropriate point in the derivation of relevant negative sentences.19 (33c) is very similar to (33a). There is a non-veridical operator in C, morphosyntactically realised by non in T, and this element Agrees with the D of nessuno in the postverbal subject position. Since nessuno is in does not intervene in the inheritance operation. However, we could think that T’s EPP-feature is intrinsic to T, not inherited (cf. Biberauer 2010), and that T thereby triggers movement to its Specifier before inheritance takes place. This would allow us to retain the Extension Condition. It might be thought that T’s raising of the external argument constitutes look-ahead, but the most local elements to T (in fact, the only ones not blocked by an intervener) are SpecvP, i.e. the external argument, and vP. According to Biberauer & Richards (2006), exactly these options are available for T. Another problem concerns the intervention role of the subject: should not the subject, as a DP bearing j-features, intervene for inheritance of j-features from C to T? This suggests that in fact C directly Agrees with the subject in languages where the subject is in SpecTP, with T simply providing a movement trigger. This is not the case in languages with medial negation, since we see the morphosyntactic realisation of the negative features in T. So we conclude that where there is a morphological realisation of features in T, inheritance takes place. This analysis has obvious implications for the position of the subject in languages with subject clitics. If, as Roberts (2005) argues, the agreement morphemes of the Celtic languages are subject clitics, then the fact that subjects do not raise to SpecTP in these languages can be accounted for. 18. In strict NC languages (e.g. many Slavic languages, Romanian and Hungarian), the variant of (33b) with the negation realised is grammatical. Here C arguably KEEPs its negation feature and can therefore Agree with all the potential probes it c-commands. A clausal negator that appears here can be treated as comparable to French ne, i.e. as something other than a purely negative element (see below). (33b) is grammatical for many Italian speakers in denial contexts with a double negation interpretation. Here nessuno is enclosed in a Focus projection, which has the e¤ect of licensing the Operator feature internally to FocP. In this case, we take it that nessuno is raised to SpecFocusP. In this position it has scope over non and is interpreted as negative, due to the focus operator; hence the double negation interpretation. There is no intervention e¤ect, since nessuno does not move through SpecTP (we assume that Italian T does not have an obligatory EPP-feature; cf. Biberauer & Richards 2006 and Biberauer 2010 on the interpretive e¤ects typically associated with optional EPP-features). 19. This is only true for neutral/unfocused non, i.e. the spell-out of C’s DONATEd features; stressed NON realises a distinct feature bundle, i.e. [Foc [Neg]], and thus has to be present in the Numeration of a structure in which it occurs.

46

Theresa Biberauer & Ian Roberts

this postverbal position (which is probably to be identified with SpecvP), it does not intervene between C and T, allowing non to be spelled out. (33d) is not ruled out due to ill-formed Agree relations, as these could be like those in (33b), with nessuno Agreeing with the null negative operator in C. Instead, what is at work here is the parametric property of Italian that requires negation to be morphosyntactically realised in the C/T-system, since C DONATEs its negation feature to T, but does not SHARE it with v. If the postverbal subject is in SpecvP, that requirement is not satisfied here. To summarise, in Italian: – C donates its Neg-feature to T, as in English. However, this feature is not always realised as non, owing to the possible intervention of an n-word subject (e.g. nessuno);20 – N-words are like English no-DPs, except that they are less systematically associated with a Focus shell, and hence typically enter into the (possibly Multiple) Agree relations which give rise to NC; – There is no counterpart to NPI any. Let us now look at Modern French. The relevant examples are given in (34): (34) a.

Je n’ai vu personne. I neg-have seen no-one ‘I have seen no-one.’

b.

Personne ne m’a vu. no-one neg me-has seen ‘No-one has seen me.’

c. Je n’ai pas vu Marie. I neg-have not seen Marie ‘I haven’t seen Marie.’ d.

*Je n’ai pas vu personne. I neg-have not seen no-one

20. Note that the English counterpart to (33b), with overt clausal negation, is very marginal and only admits a double negation interpretation: (i)

?? No-one/Z NO-ONE hasn’t done it.

This is predicted by our analysis: the only way in which the no-word fails to act as an intervener for inheritance is by being associated with a Focus shell, which necessarily gives rise to double negation here.

Negative words and related expressions

47

(34a) looks exactly like (33a), if we assimilate ne to non. However, (34b) shows that this is incorrect. If ne were the equivalent of non, we could say that (34b) di¤ers minimally from (33b) in that clausal negation ne is required. Here, rather than treat ne as a realisation of negation features inherited from C, we follow the traditional view and take it to be either a form of expletive negation, containing a semantically inert negative feature or a polarity (Pol) head (cf. recent proposals in Rooryck 2008; Biberauer 2008, 2009; Breitbarth & Haegeman 2009, all of whom build on Laka 1990; cf. Zeijlstra 2009 for an alternative perspective in terms of which ne is an NPI. Crucially, all these approaches have in common that ne would be probed by a non-veridical operator in C). Regardless of which perspective is the most feasible one, the core point for us is that ne is not a negative element under any of the proposals mentioned here. There is, then, no sense in which any unvalued feature on ne can be valued negative through an Agree relation with something else (if it is a Pol head, it may, however, receive the veridicality value specified on C, with C once again entering multiple Agree relations with the various polarity-related elements in the structure). Ne therefore does not act as an intervener for personne in (34a), and the same is true for other structures featuring n-words following ne. In (34b), personne again acts as the Goal for the null operator in C. There is no feature inheritance, and so no interaction between personne and ne of the kind described for nessuno . . . non above. We can therefore maintain that personne has a negative feature, like nessuno. In (34c) we see the characteristic Modern French double exponence of clausal negation in the form of ne . . . pas. Again, rather than treating these elements as reflexes of the operator in C, we propose that pas functions essentially as an n-word (see also De´prez 1999, 2000 and, more recently, Zeijlstra 2010). In terms of the characterisation of Italian n-words given above, then, pas has a negative feature and an unvalued operator feature. It is adjoined to vP and Agrees with the c-commanding negative operator in C. Since ne is essentially inert, this Agree relation between pas and the negative operator hosted in C establishes the clausal negation (the variable introduced by the unvalued feature on pas presumably corresponds to the temporal variable; we return to this point below). In these terms, we can see why (34d) is ungrammatical.21 This is because personne’s Operator 21. Some speakers marginally accept (34d) with a double negation reading. As with double negation readings more generally, this possibility is subject to the presence of appropriate intonation.

48

Theresa Biberauer & Ian Roberts

feature cannot be valued owing to the presence of pas intervening between it and the negation in C; we will suggest below that pas prevents the formation of a single Agree relation (C, pas, personne) because it is of the wrong semantic type. We can also observe that French di¤ers parametrically from Italian in not requiring morphosyntactic realisation of negation in the C/T system; negation features are located in v, and presumably do not spread at all. Other kinds of examples with multiple n-words are grammatical, of course: (35) Personne n’a rien dit. no-one neg-has nothing said ‘No one said anything.’ We take it that the negative operators are able to combine to form a single complex operator thanks to Multiple Agree (see Haegeman 1995; pace Haegeman & Lohndal 2010, who specifically assume absorption to fall out from single Agree relations which maximise Match). In these cases, as well as creating an operator-variable configuration, Agree creates a complex operator; e¤ectively it gives rise to ‘‘absorption’’. Absorption of wh-quantifiers was first proposed by Higginbotham & May (1981) and Aoun, Hornstein & Sportiche (1982) as a way of accounting for the fact that several wh-operators appear to act as a single complex quantifier in examples like: (36) Who bought what? ‘‘Covert’’ wh-movement (which may simply be a wh-Agree relation with no movement, see Chomsky 2000 et seq.; or movement which takes place in Narrow Syntax, but is not reflected at PF, owing to low-copy spellout, see Bobaljik 2002) brings two sets of wh-features together to form a single complex operator binding two variables. This facilitates the pair-list interpretation of (36) (i.e. where the question is answered as John bought a house, Mary bought a car, etc.). The operation essentially combines features of the same type (wh) occupying specifiers of the same category, C (see also McDaniel 1988; Barss 2000; Sura´nyi 2006 for discussion). In addition to wh-elements, absorption also applies to covertly raised quantified phrases. It is thus quite plausible that it can apply to several negative elements in a single domain, as in (35). Importantly, absorption can combine personne and rien and (35) and who and what in (36), but it cannot combine personne and pas in (34d). This is because the latter are distinct operators, binding distinct kinds of

Negative words and related expressions

49

variables: personne is associated with an individual variable, as is rien, while pas, as mentioned above, is associated with a temporal variable. Similarly, who and what are able to absorb because they bind individual variables, while whether, although a wh-expression, cannot combine with either: (37) *John asked whether Mary saw who. (37) cannot yield a pair-list interpretation of the kind ‘‘John asked if it was true that Mary saw Bill and false that Mary saw Dave’’. This is because whether is associated with a propositional variable rather than an individual variable (cf. Haegeman & Lohndal 2010 for more detailed discussion of an independently arrived at formal, Agree-based implementation of this idea). This concludes our brief discussion of French and Italian negation and n-words. In our terms, then, French has the following properties: – Unlike both English and Italian, French has no requirement for the morphophonological realisation of negation in the T-system; ne realises a non-negative feature of some kind (possibly Pol), while pas, situated within the edge of vP, realises clausal negation; – French n-words, on the other hand, are like their Italian counterparts, in particular in that they participate in Multiple Agree, although the di¤erent properties of both ne and pas as compared to Italian non give rise to di¤erent interactions between clausal negation and n-words; – Again, there is no counterpart to English any. The central point is that these systems di¤er from one another only very slightly, once the mechanisms licensing negation and related expressions are properly understood. Minimally di¤ering underlying systems, then, can give rise to superficially quite di¤erent clusters of phenomena.

5. Conclusions We have proposed an account of clausal negation, negative indefinites, n-words and NPIs using the basic mechanism of Agree, which we think avoids the pitfalls of other Agree-based analyses discussed in Section 2. Our main proposals are that negative indefinites and n-words both have interpretable negative features (which yield a semantic negation once part of an Agree relation) and uninterpretable Operator features. In both cases, the uninterpretable Operator features may either Agree with any clausal operator, or, where they are focused, they may be bound by a (null) local

50

Theresa Biberauer & Ian Roberts

focus operator. Formally, then, no-DPs and n-words are identical (cf. also Haegeman 1995 and de Swart & Sag 2002); the di¤erence in their interpretive possibilities follows to a large extent from the independent fact that the former are part of systems featuring any-DPs, whereas the latter are not. More specifically, systems lacking any-DPs will employ n-words (formally: unfocussed no-DPs) both in contexts in which the n-word is the only negation element and in contexts where the n-word co-occurs with other negative elements (sentential negators and/or further n-words). In the latter case, the clausal operator will Agree with all of the negative elements, resulting in a Multiple Agree configuration involving multiple negative elements, which is interpreted at LF as NC. The same possibility is also available in systems featuring any-DPs, as NC varieties of English show (cf. also NC varieties of German, Dutch and Afrikaans, all of which have any-DPs), but it is worth noting that the standard varieties in each case do not permit no-DPs to enter into Multiple Agree relations delivering NC interpretations; wherever multiple negative elements cooccur in these varieties, at least one of them is required to be focused, thereby ‘‘sealing o¤ ’’ the relevant element from the clausally mediated negative Agree relation. This (prescriptively imposed) convention, then, ensures that no- and any-DPs do not serve as doublets in these systems (cf. inter alia Kiparsky 1973; Anderson 1992; Kroch 1994; Wunderlich 1996; Noyer 1997; Stump 2001; Embick 2008 on the avoidance of doublets and blocking). Worth noting here is that the Focus shell which precludes NC readings in Standard English also plausibly plays a role in the context of Jespersen’s Cycle: if the elements that initially reinforce the sentential negator are enclosed in a Focus shell, they will not be able to enter into Agree relations with a clausal operator, an unproblematic state of a¤airs while these elements are simple bare nouns, lacking unvalued features; once they become more established, we can think of one of the formal e¤ects of bleaching as being the loss of the previously obligatory Focus shell, with the result that the reinforcer is in principle available for reanalysis as an unvalued feature-bearing element, i.e. as an element which may form part of the Agree chain initiated by a clausal operator (an NPI). In this way, initially non-negative elements may become integrated into an NC system. Returning to our discussion of indefinites and their formal structure: in our terms, NPIs di¤er from both no-DPs and n-words in that they lack a negation feature. In fact, we suggested that any-DPs have a null D which incorporates with the Probe; hence the strong, non-reconstructible c-command-based licensing requirement on these elements. The basic

Negative words and related expressions

51

clausal and DP-internal Agree relations at work here are also at work with indefinites more generally (although we did not enter into detail regarding the analysis of PPIs, for which see Szabolcsi 2004), and some of them hold for definite (strongly quantified) DPs. In this way, we have integrated a good part of the analysis of negative DPs of various kinds with the general analysis of DPs. Here we have shown how this analysis might be applied to English, Italian and French. Obviously, our approach should be tested against data from a wider range of languages, and also, crucially, in the diachronic domain.

Acknowlegement This paper derives from the presentation given at the first Cycles of Grammaticalization Workshop (Aston University, 7 June 2008). Ian Roberts would like to thank the participants at that seminar for their comments, and Richard Ingham and Pierre Larrive´e for organising it. The content of the present chapter di¤ers somewhat from the presentation, in that it reflects subsequent joint work, some of which is reported in Biberauer & Roberts (2011). The usual disclaimers apply.

References Aboh, Enoch 2010

Information structuring begins with the Numeration. Iberia 2 (1): 12–42. Alexiadou, Artemis, Liliane Haegeman & Melita Stavrou 2007 The Noun Phrase in the generative perspective. Berlin/New York: Mouton de Gruyter. Anderson, Stephen 1992 A-Morphous morphology. Cambridge: Cambridge University Press. Aoun, Joseph, Norbert Hornstein & Dominique Sportiche 1982 Some aspects of wide scope quantification. Journal of Linguistic Research 1: 69–95. Aquaviva, Paolo 1996 Negation in Irish and the representation of monotone decreasing quantifiers. In: Robert Borsley & Ian Roberts (eds.), The syntax of the Celtic languages, 284–313. Cambridge: Cambridge University Press.

52

Theresa Biberauer & Ian Roberts

Barss, Andrew 2000

Bayer, Josef 1990

Barbiers, Sjef 2002

Beck, Sigrid 1996a Beck, Sigrid 1996b Biberauer, 2008

Biberauer, 2009

Biberauer, 2010

Biberauer, 2009

Biberauer, 2010 Biberauer, 2006

Minimalism and asymmetric wh-interpretation. In: Roger Martin, David Michaels & Juan Uriagereka (eds.), Step by step: Essays in honor of Howard Lasnik, 31–52. Cambridge: MIT Press. What Bavarian negative concord reveals about the syntactic structure of German. In: Joan Mascaro´ & Marina Nespor (eds.), Grammar in progress, 13–24. Dordrecht: Foris. Microvariation in negation in varieties of Dutch syntactic microvariation. In: Sjef Barbiers, Leonie Cornips & Susanne Van der Kleij (eds.), Syntactic microvariation, 13–40. Amsterdam: Meertens Institute Electronic Publications in Linguistics. Wh-constructions and transparent Logical Form. Ph.D. dissertation, Universita¨t Tu¨bingen.

Quantified Structures as barriers for LF-movement. Natural Language Semantics 4: 1–56. Theresa Doubling and omission: insights from Afrikaans negation. In: Sjef Barbiers, Margreet van der Ham, Olaf Koeneman & Marika Lekakou (eds.), Microvariations in syntactic doubling, 103–140. Bingley: Emerald. Theresa Jespersen o¤ course? The case of contemporary Afrikaans negation. In: Elly van Gelderen (ed.), Cyclical change, 91–130. Amsterdam: Benjamins. Theresa Semi pro-drop languages, expletives and expletive pro reconsidered. In: Theresa Biberauer, Anders Holmberg, Ian Roberts & Michelle Sheehan (eds.), Parametric variation: Null subjects in minimalist theory, 153–199. Cambridge: Cambridge University Press. Theresa, Anders Holmberg & Ian Roberts Linearization and the architecture of grammar: A view from the Final-over-Final constraint. In: Vicenzo Moscati & Ermilio Servidio (eds.), StiL – Studies in Linguistics (Proceedings of XXXV Incontro di Grammatica Generativa), 78–91. Theresa, Anders Holmberg & Ian Roberts A hierarchical universal and its consequences. MS, Universities of Cambridge and Newcastle. Linguistic Inquiry. Theresa & Marc Richards True optionality: When the grammar doesn’t mind. In: Cedric Boeckx (ed.), Minimalist essays, 35–67. Amsterdam: Benjamins.

Negative words and related expressions

53

Biberauer, Theresa & Ian Roberts 2010 Subjects, tense and verb-movement. In: Theresa Biberauer, Anders Holmberg, Ian Roberts & Michelle Sheehan (eds.), Parametric variation: Null subjects in minimalist theory, 263–303. Cambridge: Cambridge University Press. Biberauer, Theresa & Ian Roberts 2011 Comments on Ja¨ger ‘‘Anything is nothing is something’’: On the diachrony of polarity types of indefinites. Natural Language and Linguistic Theory 28: 823–836. Biberauer, Theresa & Hedde Zeijlstra 2009 Negative concord in Afrikaans: Filling the typological gap. To appear in S. J. Mullin Lima & B. Smith (eds.), Proceedings of NELS 39. [A slightly longer version is also available on LingBuzz: http://ling.auf.net/lingBuzz/000830] Biberauer, Theresa & Hedde Zeijlstra 2010 Negative changes: Three factors and the diachrony of Afrikaans negation. In: Juanito Avelar, Sonia Cyrino, Charlotte Galves & Ruth Lopez (eds.), Diachronic syntax: Parameter theory and dynamics of change, Oxford: Oxford University Press. Biberauer, Theresa & Hedde Zeijlstra 2011 Afrikaans negation: Filling a typological gap. To appear in Journal of Semantics. Bobaljik, Jonathan 2002 A chains at the PF-interface: Copies and ‘covert’ movement. Natural Language and Linguistic Theory 20: 197–267. Breitbarth, Anne & Liliane Haegeman 2009 Not continuity but change: stable Stage II in Jespersen’s Cycle. MS, University of Cambridge and University of Ghent. Cardinaletti, Anna & Michal Starke 1999 The typology of structural deficiency: A case study of the three classes of pronouns. In: Henk van Riemsdijk (ed.), Clitics in the languages of Europe, 145–235. Berlin: Mouton de Gruyter. Cecchetto, Carlo & Caterina Donati 2010 On Labeling: Principle C and Head Movement. Syntax 13 (3): 241–278. Chomsky, Noam 2000 Minimalist inquiries: the framework. In: Roger Martin, David Michaels & Juan Uriagereka (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–156. Cambridge: MIT Press. Chomsky, Noam 2001 Derivation by phase. In: Michael Kenstowicz (ed.), Ken Hale: A life in Language, 1–53. Cambridge: MIT Press. Chomsky, Noam 2008 On phases. In: Robert Freidin, Carlos Otero & Maria-Luisa Zubizarreta (eds.), Foundational issues in linguistic theory, 133– 166. Cambridge: MIT Press.

54

Theresa Biberauer & Ian Roberts

Cinque, Guglielmo 1999 Adverbs and functional heads: A crosslinguistic perspective. New York: Oxford University Press. Collins, Christopher 2005 A smuggling approach to the passive in English. Syntax 8: 81– 120. Corblin, Francis 1996 Multiple negation processing in natural language. Theoria 17: 214–259. Corblin, Francis, Viviane De´prez, Henriette de Swart & Lucia Tovena 2004 Negative concord. In: Francis Corblin & Henriette de Swart (eds.), Handbook of French semantics, 417–452. Stanford: CSLI. Davidson, Donald 1967 The logical form of action sentences. In: Nicholas Rescher (ed.), The logic of decision and action, 81–95. Pittsburgh: University of Pittsburgh Press. De´prez, Viviane 1999 The roots of negative concord in French and French-lexicon Creoles. In: Michel DeGra¤ (ed.), Language creation and language change: Creolization, diachrony, and development, 375– 428. Cambridge: MIT Press. De´prez, Viviane 2000 Parallel (a)symmetries and the internal structure of negative expressions. Natural Language and Linguistic Theory 18: 253–342. Embick, David 2008 Variation and morphosyntactic theory: Competition fractionated. Language and Linguistics Compass 2 (1): 59–78. Enc¸, Murvet 1987 Anchoring conditions for tense. Linguistic Inquiry 18: 633–657. Falaus, Anamaria 2007 Negative concord and double negation: The Romanian puzzle. In: Jose´ Camacho, Nydia Flores-Ferra´n, Liliana Sa´nchez, Viviane De´prez & Marı´a Jose´ Cabrera (eds.), Romance Linguistics 2006. Selected Papers from the 36th Linguistic Symposium on Romance Languages (LSRL), New Brunswick, March–April 2006, 135– 148. Amsterdam: Benjamins. George, Leland & Jaklin Kornfilt 1981 Finiteness and boundedness in Turkish. In: Frank Henry (ed.), Binding and filtering, 105–127. Cambridge: MIT Press. Giannakidou, Anastasia 1998 Polarity sensitivity as (non)veridical dependency. Benjamins: Amsterdam. Grice, H. Paul 1975 Logic and conversation. In: Peter Cole & Jerry Morgan (eds.), Syntax and semantics, volume 3, 41–58. New York: Academic Press.

Negative words and related expressions

55

Haegeman, Liliane 1995 The syntax of negation. Cambridge: Cambridge University Press. Haegeman, Liliane & Ra¤aella Zanuttini 1996 Negative concord in West Flemish. In: Adriana Belletti & Luigi Rizzi (eds.), Parameters and functional heads: Essays in comparative syntax, 117–179. Oxford: Oxford University Press. Haegeman, Liliane & Terje Lohndal 2010 Negative concord and (multiple) Agree: a case study of West Flemish. Linguistic Inquiry 41: 181–211. Han, Chung-Hye 2001 Force, negation and imperatives. The Linguistic Review 18: 289– 325. Heim, Irene 1982 The semantics of definite and indefinite Noun Phrases. Ph.D. dissertation, University of Masschusetts Amherst. Higginbotham, James 1985 On semantics. Linguistic Inquiry 16: 547–593. Higginbotham, James & Robert May 1981 Questions, quantifers and crossing. Linguistic Review 1: 41–79. Holmberg, Anders 2001 The syntax of yes and no in Finnish. Studia Linguistica 55: 141– 174. Holmberg, Anders 2003 Yes/no questions and the relation between tense and polarity in English and Finnish. In: Pierre Pica (ed.), Linguistic Variation Yearbook 3, 43–68. Amsterdam: Benjamins. Horn, Laurence 1989 A natural history of negation. Chicago: University of Chicago Press. Ihsane, Tabea 2008 The layered DP: Form and meaning of French indefinites. Amsterdam: Benjamins. Ingham, Richard 2003 The development of the expletive negative construction in Middle English. Transactions of the Philological Society 101 (3): 411–452. Ingham, Richard 2007 NegP and negated constituent movement in the history of English. Transactions of the Philological Society 105 (3): 1–33. Ja¨ger, Agnes 2008 History of German negation. Amsterdam: Benjamins. Ja¨ger, Agnes 2011 Anything is nothing is something. Natural Language and Linguistic Theory 28: 787–822.

56

Theresa Biberauer & Ian Roberts

Kamp, Hans 1981

A theory of truth and semantic representation. In: Jeroen Groenendijk, Theo Janssen & Martin Stokhof (eds.), Formal Methods in the Study of Language, 277–322. Amsterdam: Mathematical Centre. Reprinted in Jeroen Groenendijk, Theo Janssen & Martin Stokhof (eds.), 1984, Truth, interpretation, information, 1–41. Dordrecht: Foris. Reprinted in Paul Portner & Barbara Partee (eds.), 2002 Formal semantics: The essential readings, 189–222. Oxford: Blackwell.

Kayne, Richard 1975 French syntax. Cambridge: MIT Press. Kayne, Richard 1984 Connectedness and binary branching. Dordrecht: Foris. Kiparsky, Paul 1973 ‘Elsewhere’ in phonology. In: Stephen Anderson & Paul Kiparsky (eds.), A Festschrift for Morris Halle, 93–106. New York: Holt, Reinhart, & Winston. Kroch, Anthony 1994 Morphosyntactic variation. In: Kathryn Beals (ed.), Papers from the 30th Regional Meeting of the Chicago Linguistics Society: Parasession on Variation and Linguistic Theory, 180–201. Chicago: Chicago Linguistic Society. Laka, Itziar 1990 On the syntax of negation. Ph.D. dissertation, MIT. Laka, Itziar 1994 On the syntax of negation. New York: Garland. Liberman, Mark & Ivan Sag 1974 Prosodic form and discourse function. Proceedings of the Chicago Linguistics Society 10: 416–427. Marantz, Alex 1997 No escape from syntax. University of Pennsylvania working papers in linguistics 4: 201–225. Martin, Roger 2001 Null case and the distribution of PRO. Linguistic Inquiry 32 (1): 141–166. Martins, Anna-Maria 2000 Polarity items in Romance: Underspecification and lexical change. In: Susan Pintzuk, George Tsoulas & Anthony Warner (eds.), Diachronic syntax: Models and mechanisms, 191–219. Oxford: Oxford University Press. McCloskey, James 1996 The scope of verb-movement in Irish. Natural Language and Linguistic Theory 14: 47–104. McCloskey, James 2001 On the morphosyntax of wh-extraction in Irish. Journal of Linguistics 37: 67–100.

Negative words and related expressions

57

McDaniel, Dana 1989 Partial and multiple wh-movement. Natural Language & Linguistic Theory 7 (4): 565–604. Noyer, Rolf 1997 Features, positions, and a‰xes in autonomous morphological structure. New York: Garland. Ouali, Hamid 2008 On C-to-T j-feature transfer: the nature of agreement and antiagreement in Berber. In: Roberta D’Alessandro, Gunnar Hrafnbjargarson & Susann Fischer (eds.), Agreement restrictions, 159–180. Berlin: Mouton de Gruyter. Penka, Doris 2002 Zur Semantik der negativen Indefinita im Deutschen. Universita¨t Tu¨bingen: Tu¨bingen-Linguistik-Report Nr. 1. Penka, Doris 2007 Negative indefinites. Ph.D. dissertation, Universita¨t Tu¨bingen. Penka, Doris 2010 Negative indefinites. Oxford: Oxford University Press. Penka, Doris & Arnim von Stechow 2001 Negative Indefinita unter Modalverben. In: Reimar Mu¨ller & Marga Reis (eds.), Modalita¨t und Modalverben im Deutschen, 263–286. Hamburg: Helmut Buske Verlag. Pesetsky, David & Esther Torrego 2007 The syntax of valuation and the interpretability of features. In: Simin Karimi, Vida Samiian & Wendy Wilkins (eds.), Phrasal and clausal architecture: Syntactic derivation and interpretation, 193–220. Cambridge: MIT Press. Ramchand, Gillian 2008 Verb meaning and the lexicon. A first-phase syntax. Cambridge: Cambridge University Press. Rizzi, Luigi 1982 Topics in Italian syntax. Dordrecht: Foris. Rizzi, Luigi 1990 Relativized minimality. Cambridge: MIT Press. Rizzi, Luigi 1997 On the fine structure of the left periphery. In: Liliane Haegeman (ed.), Elements of grammar, 281–337. Dordrecht: Kluwer. Rizzi, Luigi 2001 Relativized minimality e¤ects. In: Mark Baltin & Chris Collins (eds.), The handbook of contemporary syntactic theory, 89–110. Oxford: Blackwell. Rizzi, Luigi 2008 On delimiting movement. Paper presented at GLOW 31, Newcastle.

58

Theresa Biberauer & Ian Roberts

Roberts, Ian 2004

Roberts, Ian 2005 Roberts, Ian 2007 Roberts, Ian 2010a Roberts, Ian 2010b

The C-system in Brythonic Celtic languages, V2 and the EPP. In: Luigi Rizzi (ed.), The cartography of CP and IP, 297–328. Oxford: Oxford University Press. Principles and parameters in a VSO Language: A case study in Welsh. Oxford: Oxford University Press. Diachronic syntax. Oxford: Oxford University Press. Clitics, incorporation and head movement. Cambridge: MIT Press.

A deletion analysis of null subjects. In: Theresa Biberauer, Anders Holmberg, Ian Roberts & Michelle Sheehan (eds.), Parametric variation: Null subjects and minimalist theory, 58–87. Cambridge: Cambridge University Press. Roberts, Ian & Anna Roussou 2003 Syntactic change. A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. Rooryck, Johan 2008 On the compositional nature of syntactic negation in French. MS, Leiden University. Roussou, Anna 2000 On the left periphery: modal particles and complementisers. Journal of Greek Linguistics: 65–94. Sperber, Daniel & Deirdre Wilson 1986/1995 Relevance: Communication and cognition. Oxford: Blackwell. Sportiche, Dominique 2005 Cyclic NP structure and the interpretation of traces. In: Hans Broekhuis, Norbert Corver, Jan Koster, Riny Huybregts & Ursula Kleinhenz (eds.), Organizing grammar: Linguistic studies in honor of Henk van Riemsdijk, 599–607. Berlin: Mouton de Gruyter. von Stechow, Arnim 1993 Die Aufgaben der Syntax. In: Joachim Jacobs, Wolfgang Sternefeld, Arnim Von Stechow & Theo Vennemann (eds.), Ein Internationales Handbuch zeitgenossischer Forschung, 1–88. Berlin: Mouton de Gruyter. Stowell, Timothy 1996 The phrase structure of tense. In: Johan Rooryck & Annie Zaenen (eds.), Phrase Structure and the Lexicon, 277–292. Dordrecht: Kluwer. Stump, Gregory 2001 Inflectional morphology: A theory of paradigm structure. Cambridge: Cambridge University Press.

Negative words and related expressions

59

Sura´nyi, Balaczs 2006 Quantification and focus in negative concord. Lingua 116: 272– 313. de Swart, Herniette & Ivan Sag 2002 Negative concord in Romance. Linguistics and Philosophy 25: 373–417. Szabolcsi, Anna 2004 Positive polarity – negative polarity. Natural Language and Linguistic Theory 22 (2): 409–452. Troseth, Erika 2009 Degree inversion and negative intensifier inversion in the English DP. The Linguistic Review 26 (1): 37–65. Watanabe, Akira 2004 The genesis of negative concord. Linguistic Inquiry 35: 559–612. Weiss, Helmut 1998 Syntax des Bairischen. Studien zur Grammatik einer natu¨rlichen Sprache. Tu¨bingen: Niemeyer. Weiss, Helmut 2002 Three types of negation: A case study in Bavarian. In: Sjef Barbiers, Leonie Cornips & Susanne Van der Kleij (eds.), Syntactic microvariation. Electronic Publication: Meertens Instituut. Wunderlich, Dieter 1996 Minimalist morphology: the role of paradigms. In: Geert Booij & Jaap van Marle (eds.), Yearbook of Morphology 1995, 93– 114. Dordrecht: Kluwer. Zanuttini, Ra¤aella 1997 Clause structure and negation. Oxford: Oxford University Press. Zeijlstra, Hedde 2004 Sentential negation and negative concord. Ph.D. dissertation, University of Amsterdam. Zeijlstra, Hedde 2008 Negative concord is syntactic agreement. MS, Amsterdam University. [Available at: http://ling.auf.net/lingBuzz/000645]

Zeijlstra, Hedde 2009

On French negation. Proceedings of the Berkeley Linguistics Society 35: 447–458.

Negative words and negation in French Marie Labelle 1. Introduction The complex cross-linguistic patterns of negation and negative concord in the world’s languages pose a challenge for linguistic theory. Biberauer & Roberts (this volume) approach the problem by postulating an Agree relation between a negative clausal operator functioning as a Probe and another element functioning as the Goal. The model they develop is based on cross-linguistic di¤erences in the location of the Probe and in the features of the Goal. Our aim in the present commentary is to consider how this framework accounts for the French negation cycle and for phenomena related to negative concord and negative polarity in French. It will be shown that, while the framework has some attractive features, its empirical coverage is limited. We will first examine the case of Modern French, and then turn to Old French and historical change.

2. On the position of the negative operator in Modern French Biberauer & Roberts propose that, in Modern French, the clausal negative feature is in C (C KEEPS it feature; it does not DONATE it to T). Ne is not a negative element; it does not participate in the Agree relation between the negative operator in C and a term within the sentence. It is either a form of expletive negation, containing a semantically inert negative feature, or an element that realises a non-negative feature of some kind, possibly a polarity feature. The authors seem to favour the second option, where ne is probed by a non-veridical operator in C. There are problems with this option, however. In this framework, all clauses have an operator in C, which may be veridical or non-veridical, the negative operator being a type of nonveridical operator, namely an anti-veridical operator. Non-veridical operators Agree with negative polarity items. Biberauer and Roberts propose that ne is ‘‘either a form of expletive negation, containing a semantically inert negative feature or a polarity (Pol) head’’ (this volume). There are

62

Marie Labelle

arguments for preferring the first option over the second. Notice that, if ne does not belong to the same agreement chain as negative words (n-words) e.g. if it has a polarity feature bound by a polarity head structurally distinct from the negative head binding n-words, there is no way of expressing the dependency relation between ne and n-words described by Godard (2004). Godard argues that the domain of this relation is the same as the domain in which negative concord (NC) between two negative quantifiers is possible: the two negative elements must normally belong to the same clause, but, with a specific set of non-factive verbs the dependency may be extended to an embedded infinitival VP complement (1). However, this is not possible when one negative expression is an adverb (2) or when the negative expression belongs to an infinitive VP subject (3). (Examples (1a), (2a) and (3a) are from Godard 2004.) (1) a.

b.

(2) a.

b.

(3) a.

b.

Cet inculpe´ n’essaie de convaincre personne. the accused ne-tries to convince nobody (Godard 2004: 360, example (16a)) Personne n’essaie de convaincre personne. (NC) nobody ne-tries to convince nobody ‘Nobody tries to convince nobody.’ *Paul n’essaie de jamais travailler. Paul ne-tries to never work (Godard 2004: 360, example (16b)) *Personne n’essaie de jamais travailler. nobody ne-tries to never work ‘Nobody tries to never work.’ *Aujourd’hui, voir personne n’est souhaitable today, to-see nobody ne-is desirable (Godard 2004: 360, example (17a)) Voir personne ne plaıˆt a` personne. (DN only) to-see nobody ne pleases to nobody ‘To see nobody pleases nobody.’

This parallelism between the domains for NC and for ne . . . n-word follows from an operator analysis if the negative operator Agrees with all n-words in its domain as well as with ne. Thus, even if ne does not carry the clausal negative feature, it must be part of the negative agreement chain.

Negative words and negation in French

63

In addition, ne is perhaps semantically inert in not carrying the feature for clausal negation, but it is relevant for semantic interpretation because it indicates the scope of negation, the clausal domain in which the negative dependency is observed (Kayne 1984; Godard 2004: 358, examples (14a–b)): (4) a.

Paul accepte de ne recevoir personne. Paul agrees to ne see nobody ‘Paul agrees not to see anyone.’

b.

Paul n’ accepte de recevoir personne. Paul ne agrees to see nobody ‘Paul does not agree to see anyone.’

The contrast in (4) follows if ne is a morphological reflex of the presence of a negative operator in the c-commanding C. Therefore, ne must agree with the negative operator in C, and it seems preferable to treat it as carrying a formal (but semantically inert) negative feature checked by the negative operator (this would be like gender agreement between an adjective and a noun) than to treat it as carrying a non-negative (polarity) feature. This would explain the apparent negative dependency between ne and a negative word: both are Probed by the same operator.

3. Explaining Negative Concord Biberauer & Roberts account for Negative Concord by quantifier absorption as a result of Multiple Agree. When many n-words Agree with a negative operator in C, they are able to combine to form a single complex operator. A system that explains NC via Agree must account for the ambiguity of (5), which allows a NC reading and a double negation (DN) reading. (5) Personne ne croit personne. nobody ne believes nobody (Corblin, De´prez, de Swart & Tovena 2004: 435) (i)

Bpx py Believe (x,y) [NC] ‘Nobody believes anyone’

(ii)

Bpx Bpy Believe (x,y) [DN] ‘Everybody believes someone’

64

Marie Labelle

It has often been observed that in French the NC reading is the natural reading. Here, the idea that DN obtains when one item is focused, proposed by Biberauer & Roberts for standard English and by De´prez (2003) for French, is coherent with the observation that stress makes it easier to obtain the DN reading (Corblin, De´prez, de Swart & Tovena 2004: 425). It could also explain the observation that tri-negative readings do not obtain (Corblin, De´prez, de Swart & Tovena 2004: 432). Since a tri-negative reading would require two negative words to be contrastively focused, it is not expected to occur. A question arises in this perspective: how to analyse strict NC languages where clausal negation co-occurs with negative quantifiers, like some varieties of English, or Que´be´cois French, where J’ai pas vu personne ‘I didn’t see nobody’ is grammatical with a single negation reading. Here two distinct explanations are available. To distinguish the varieties of English, Biberauer & Roberts simply assume a dialectal di¤erence: in NC varieties, ‘‘[t]here is no absolute requirement for co-occurring negative elements to be focused, with the result that all the negative elements in a given structure may Agree with CNeg.’’ (this volume). De Swart & Sag (2002) similarly rely on a parametric di¤erence between languages: languages are free to include or not sentential negation in the concord system; Quebec French allows it, but not Standard French. It would be interesting if this followed from other properties of the system. In a second line of explanation, Biberauer & Roberts propose to treat pas, not as a reflex of the negative operator in C, but as a word that ‘‘functions essentially as an n-word’’ Agreeing with the operator (this volume). They account for why *Je n’ai pas vu personne ‘I didn’t see nobody’ is rejected in Standard French by claiming that quantifier absorption is blocked because pas is of the wrong semantic type: personne is associated with an individual variable whereas pas is associated with a temporal variable. In this perspective, Quebec French remains unaccounted for. In addition, this immediately raises the question of why jamais ‘never’ can enter NC with personne: Je n’ai jamais vu personne ici ‘I never saw anyone here’. Arguably, never is associated with a temporal variable. De´prez (1999, 2000, 2003), who similarly assumes that the formation of a polyadic quantifier is blocked because pas is of the wrong semantic type, characterises pas as a propositional operator, which avoids this problem. All things considered, however, there may be no gain in treating pas as a type of n-word if, as suggested by De´prez (1999: footnote 31), there are some polarity items that are licensed by pas but not by n-words. This would be the case for example in (6).

Negative words and negation in French

(6) a.

b.

65

Cette a¤aire-la` n’ a pas fait un pli. this thing-there ne has not made a wrinkle ‘This thing did not make a wrinkle; it was a sure thing.’ *Rien n’ a fait un pli. nothing ne has made a wrinkle ‘Nothing made a wrinkle.’

In a framework in which personne, pas, and polarity items are licensed by the same operator, the contrast in (6) remains a mystery. In fairness, however, many, perhaps most, NPI’s are licensed by the negation as well as by n-words (see (7)), and the oddness of (6b) may find some other explanation. (7) a. Le gouvernement n’a pas leve´ le petit doigt pour les aider. the government ne-has not raised the little finger to them help ‘The government did not raise a finger to help them.’ b. Personne n’a leve´ le petit doigt pour les aider. nobody ne-has raised the little finger to them help ‘Nobody raised a finger to help them.’ 4. On the di¤erence between negative concord and negative polarity Within the framework developed by Biberauer & Roberts, both n-words and negative polarity items (NPIs) are Goals for a negative operator in C. If Agree with a Probe in C is crucially responsible for licensing n-words and NPIs, the two types of elements should obey the same syntactic constraints. However, this prediction is not verified. In French as well in other languages, NPIs are licensed across islands, but n-words are not (Corblin, De´prez, de Swart & Tovena 2004; De´prez 1999). (8) Je n’engagerais pas quelqu’un I ne-would-hire not someone qui ait fait quoi que ce soit/*rien contre moi. who had done anything/nothing against me ‘I would not hire someone who had done anything/nothing against me.’ Notice also that, in French, NPIs must be c-commanded by the marker of clausal negation pas (or by an n-word), and they cannot occur in subject

66

Marie Labelle

position of a negative clause, whereas n-words are largely incompatible with pas, and are possible in subject position (Corblin, De´prez, de Swart & Tovena 2004: 425; De´prez 1999: 381, example 18). (9) Je n’ai pas vu qui que ce soit/*personne aujourd’hui. I ne-have not seen anyone/nobody today ‘I have not seen anyone/nobody today.’ (10) a.

b.

*Qui que ce soit n’est ( pas) venu aujourd’hui. anyone ne-is (not) come today ‘Anyone did not come today.’ Personne n’est venu aujourd’hui. nobody ne-is come today ‘Nobody came today.’

It is unclear how an Agree approach can account for the larger domain of NPI relative to n-words or for the di¤erences in their licensing environments.1

5. Accounting for linguistic change In a volume on cycles of grammaticalization, it is worth exploring how a cross-linguistic approach of negation can account for the French negation cycle. Ne, which had negative force in Old French, lost this property in favour of pas, initially a negation reinforcer. 5.1. Ne in Old French In Old French, ne can be the sole negative element in the clause. (11) Karles li Magnes de nos n’avrat aı¨e. Charles the Great from us ne will-have help ‘Charlemagne will not have help from us.’ (Roland, v. 1732)

1. The problem is compounded if one assumes for NPIs Biberauer & Roberts’ analysis of English any-DPs, treated as an involving head adjunction to the head containing the negative operator. If NPI involves head adjunction, NPI licensing should be more local than NC.

Negative words and negation in French

67

It is tempting to think that ne in (11) is like modern Italian non. But it becomes immediately clear that this is not the case. In Italian, preverbal subject n-words do not co-occur with non (12a), whereas in Old French, the same elements trigger the presence of ne on the verb (12b–d): (12) a.

b.

*Nessuno non mi ha visto. nobody NEG me has seen. ‘Nobody saw me’ Nus ne le puet conforter. nobody ne him can comfort ‘Nobody could comfort him.’ (Aucassin et Nicolete, VII, 4th l.)

c. Geres nient ne deut estra fruissiet ico´ que nient scarcely nothing ne must be destroyed so that nothing ne parmaint ad aurier an eglises. ne remain to adore in churches ‘Scarcely anything should be destroyed so that nothing remains to be worshipped in churches.’ (St Albans Psalter, p. 68)2 d.

[Ne vos ne il ] n’ i porterez les piez, [. . .]. neither you nor he ne LOC will-carry the feet ‘Neither you nor he will go there.’ (Roland, v. 260)

Instead of being like modern Italian (where it is proposed by Biberauer & Roberts that C DONATES it negative feature to T), Old French is like strict NC languages that allow the equivalent of Nessuno non mi ha visto, for example French Based Creoles (De´prez 1999, 2000). For these languages, Biberauer & Roberts (this volume, footnote 13), posit that ‘‘C KEEPS its negation feature and can therefore Agree with all the potential probes it c-commands. The clausal negator on T is treated as comparable to [Modern] French ne: something other than a purely negative element.’’ Evidently, this analysis is unsatisfactory as it does not explain the di¤erence between Old French, where ne could be the sole element expressing clausal negation, and contemporary French. But if, in Old French, C does not KEEP its feature but DONATES it to T, allowing ne to realise negative morphology, then (12b–d) should be as ungrammatical as (12a) because the system does not allow T to receive the negative feature from C 2. Source of the text and of the English translation: http://www.abdn.ac.uk/~lib399/ english/translation/trans068.shtml (c. 1130)

68

Marie Labelle

when a negative subject intervenes. The only way out of this problem would be to claim that the words in subject position in (12b–c) are not n-words. We come back to this issue in section 4.2. Whilst on the status of ne, let us consider Schapansky’s (2002) analysis of the di¤erence between ne and ne . . . pas in sentences like those in (13) (Schapansky 2002: 797, example 9): (13) a. b.

Je ne sais pas ou` aller (tout est re´serve´ ). ‘I don’t know where to go (everything is booked out).’ Je ne sais ou` aller (*tout est re´serve´). ‘I don’t (quite) know where to go (*everything is booked out)’

Schapansky argues that ne pas marks contradictory negation, while ne marks contrary negation. Contradictory negation is a relation between two propositions, p and Bp, such that p excludes Bp. Because it is quantifying, contradictory negation yields the no-option interpretation observed in (13a). With contrary negation, both p and Bp can be false at the same time, but they cannot both be true. Contrary negation is not quantifying, and the reading of (13b) is that there is a choice of options. In a nutshell, Schapansky proposes that the distinction stems from a feature di¤erence between ne and pas: ne is [þneg, –Q], pas is [þneg, þQ]. Because it is –Q, ne must form a negative association with a þQ element that may be an n-word, or a pseudo-modal verb; in the latter case, the negative interpretation of the sentence is construed, which results in a contrary reading. Arguably, the grammar described by Schapansky reflects a stage of the language typical of Classical French (seventeenth-eighteenth century), which survives as an archaism (Foulet 1968: §350). The analysis she proposes suggests a characterisation of the negation cycle in which ne gradually lost some features: from [þneg, þQ] in Old French to [þneg, –Q] in Classical French, to the absence of substantive semantic features in contemporary French; conversely, pas would have gained some features, going from [–neg, –Q] to [þneg, þQ] via a [–neg, þQ] stage. Can an operator analysis account for the facts? Contradictory negation may be associated with an anti-veridical operator, and contrary negation with a non-veridical operator. This being said it is unclear how Agree with these operators might account for the contrast in (13). 5.2. On the nature of the Goals In Old French, among the various elements that must be considered as potential Goals for an Agree relation with a negative operator are bare NP’s, words like rien, personne, aucun (that later became Modern French

Negative words and negation in French

69

n-words), Old French n-words like nient, nul, and negation reinforcers like pas, mie, point. For lack of space, we do not discuss the case of negation reinforcers. The other words all occupy an argument position and they should all Agree with the negative operator: they are bare indefinites or weak quantifiers. However, a quick survey for the purposes of the present paper shows that they all seem to have a distinct distribution. Nient ‘nothing’ is the Old French equivalent of modern French rien. As we saw, nient co-occurs with ne even from the subject position (12c). According to Foulet (1968: 278, §407), it is the most clearly negative word of Old French. It is found in contexts like those in (14), and it can be modified by almost (see 12c). This suggests that nient has quantificational force. (14) a. A vos que mont? fait cil. – Nient, fait Aucassins. to you, what amount? makes he. – Nothing, makes Aucassin ‘To you, what does it matter? said he. – Nothing, said Aucassin.’ (Aucassin et Nicolete, XXIV, 28–9. Foulet 1968: 279, §407) b. Se de nient poons vivre. if of nothing we-can live. ‘If we can live on nothing.’ (Le garc¸on et l’aveugle, 91–6. Foulet 1968: 279, §407) c. [. . .] vous les ave´s et jou nient. you them have and I nothing ‘You have them, and I nothing.’ (Le jeu de la feuille´e, 638–9. Foulet 1968: 278, §407) In our survey, the dependency between ne and nient is always local. The following example, where nient is in a position where NPIs are licensed, gives rise to a double negation reading. This suggests that nient is truly an n-word. Hence the problem noted for (12c) remains. (15) En tote la terre n’ot feme a haut baron ki nient vaille. in all the land ne-had wife of high baron who nothing be-worth ‘In all the land there was no wife of a high baron who was worth nothing.’ (Escoufle 1779. Englebert 1982: 29) The word nul also co-occurs with ne even from a subject position (12b), and it enters NC with another negative element, including nient (16). Notice that nul c-commands nient in (16a):

70

Marie Labelle

(16) a.

Ne nul des siens nient ne sout [. . .]. andneg none of his-people nothing ne knew ‘And nobody knew anything.’ (Marie de France, Lai du Bisclaveret, v. 28)

b.

Ne vos ne nus gentis hom que ge trover puisse [. . .]. neither you nor no gentle man that I find could ‘Neither you nor any gentleman that I could find’ (Lancelot, p. 149)

c. Nuls reis de France n’ out unkes si vaillant. no king of France ne had ever so valiant ‘No king of France was ever so valiant.’ (Roland, v. 1168) However, contrary to nient, nul appears in NPI contexts not accessible to n-words, in particular non-local contexts (17b, c): (17) a.

b.

Sire, conoissiez vos nule de noz serors qui laians sont? Sire, know you none of our sisters who here are ‘Sire, do you know any of our sisters who are here?’ (Lancelot, p. 170) [. . .] car vos estiez for you were que nus hom than no man

assez plus liez de doner much more happy to give ne fust del prendre [. . .]. ne was of-the take

‘for you were much more happy in giving than any man was of taking.’ (Lancelot, p. 222) c. [. . .] ge ne cuit que nus home mortel l’osast voir [. . .]. I ne thing that no man mortal him-would-dare see ‘I do not think that any mortal man would have dared see him.’ (Lancelot, p. 222) d.

[. . .] a ceste parole ne fu nus hom qui onques meı¨st at this word ne was no man that ever put nul contredit [. . .]. no refutation ‘Upon these words (there) wasn’t any man who ever o¤ered any refutation.’ (Lancelot, p. 110)

Negative words and negation in French

71

This distribution – concordance with negation even from the subject position and occurrence in NPI licensing contexts – is similar to that of the French Based Creole n-words, analysed by De´prez as bare N’s. But, nul is not an N0; in many of the above examples, e.g. (16c) and (17c), it occupies a position within the left periphery of the DP: D0, or perhaps a Num0. If nul occupies Num0, it might be tempting to try to extend to it Biberauer & Roberts’s analysis of any in terms of incorporation of a null D0 with the Probe. However, this would require the Probe to be in C despite the fact that ne realises clausal negation, and it would imply head movement across islands to account for non-local NPI contexts. Let us now turn to bare NPs. In Old French, bare NPs are licit even in subject position (18a). In (18b), ne scopes over the bare NP indefinite in subject position. However in (18c), there is a clear feeling that the bare NP is outside of the scope of negation. This is unexpected in a context in which a negative operator in C Probes the structure and binds the free variables under its scope. (18) a.

Paien chevalchent par cez greignurs valees. pagans ride by these wide valleys ‘(The) pagans ride through those wide valleys.’ (Roland, v. 710)

b.

Hume nel veit ki mult ne s’espoant. man ne-it-ACC see who much ne REFL-terrify ‘No man sees it without being frightened.’ (Roland, v. 1433)

c. Hom ki la vait repairer ne s’en poet. man who there goes come-back ne REFL-GEN can ‘A man who goes there can not come back’ (Roland, v. 311) Finally, the word rien is a noun meaning ‘thing’ in Old French. It is found in positive and negative sentences, and it can be modified, in particular by nul ‘none’ (19a–b). (19) a.

b.

Je sui la plus desconseilliee riens del monde. I am the most desparate thing of-the world ‘I am the most desperate thing in the world.’ (Lancelot, p. 76) ge ne desir tant nule rien. I ne desire as-much no thing ‘I do not desire as much any thing.’

(Lancelot, p. 78)

72

Marie Labelle

Like other nouns, rien is licit as a bare nominal. However, when it appears in subject position of a negative sentence (20), it seems to always be under the scope of ne, contrary to what we saw for other bare NPs (see (18)). (20) Riens ne me porroit retenir. (no)thing ne me-ACC could stop. ‘Nothing could stop me.’ (NOT: ‘Something could not stop me.’) (Erec et Enide, v. 5472–3) This rapid survey shows that the various elements discussed do not have the same distribution. We could add that, to our knowledge, polarity items with no intrinsic negative content are not licensed in subject position; however, they are compatible with pas (which Hansen 2009 interprets as showing that pas must very quickly have grammaticalized from an NPI to a negative particle proper): (21) Tuit vo Franceis ne valent pas meaille. all your Frenchmen ne are-worth not dime. ‘All your Frenchmen are not worth a dime.’ (Cour. Louis, v. 2433, in Hansen, 2009: 229) The di‰culty with the Agree approach to negation developed by Biberauer & Roberts is to account for the di¤erences in licensing conditions of these words, which should all be Goals for a negative Probe. Once a solution to this problem is found, a further challenge will be to account for how the system evolved between Old French and contemporary French. De´prez (2003) establishes a relation between the licensing of bare nominals and the possibility for an n-word to concord with negation. She accounts for the evolution of rien by proposing that when bare nominals were no longer grammatical, rien survived by raising to the D position and acquiring a (negative) quantificational meaning (p. 282; see also De´prez & Martineau 2004 on aucun). This type of approach does not seem to be readily available in Biberauer & Roberts’ framework, however. 5.2.1. Negative concord at the clausal level In Old French, as in contemporary French, a non-negative ne is found in an embedded clause when the main clause contains an explicit or implicit negation.

Negative words and negation in French

73

(22) Nes poet guarder [que mals ne l’ i ateignet.] (he) ne-them-ACC can prevent that misfortune ne him there reach ‘He cannot prevent misfortune from reaching him there.’ (Roland, v. 9) Espinal (2000) argues that expletive negation is found in clauses that are in the semantic scope of a non-veridical expression, and she develops an analysis in terms of negative concord of the negative marker with a non-veridical operator. This would be a natural extension of the Agree approach to NC if a principled account can be provided for when ne Agrees with or realises the antiveridical, negative clausal feature, and when it simply concords with a non-veridical operator, in which case it has no negative value. As a last example, let us mention that, in Old French, negative coordinated clauses are regularly introduced by a negative coordinator, ne. This could be analysed as having the coordinator concord with a negative feature in the C of the coordinated clause. (23) [. . .] car do pere fu il nule seu¨e, for of-the father ne was it no known-thing, ne elle nel vost dire a nelui. and-neg she ne-it wanted say to nobody ‘for nothing was known of the father, and she did not want to tell it to anybody.’ (Lancelot, p. 94) Should it be possible to extend Biberauer & Roberts’s framework to account for these facts, it would provide strong support for an approach in terms of Agreement with a negative operator in C.

6. Conclusions In sum, assuming a negative operator in C may provide an interesting account of some of the facts discussed here, for example the uniformity of the domains for the relation between ne and an n-word and for NC between two n-words, or the concord of a clause with a negative element in its c-commanding domain. Thus, the ideas developed by Biberauer & Roberts are worth exploring further. However, the empirical coverage of the framework in its present state seems to be restricted. The relation Biberauer & Roberts try to establish between indefinites, weak quantifiers

74

Marie Labelle

and n-words appears to be a handicap more than a positive outcome of their approach, as it seems too restricted to provide an explanatory account of the domains within which various types of Goals are licensed, or of the evolution of these domains with time. Further developments of this approach would need to address these issues and to provide answers for the questions that are the focus of this collection, namely, the definition of clear criteria for identifying when a polarity item becomes a negative element, the definition of means to recognise the presence of a negative phrase, and, ultimately, a proper treatment of the negation cycle.

References Sources Aucassin et Nicolete. 1973 [end 12th c.] Erec et Enide. 1890. [c. 1170]

Aucassin et nicolete. Edited by Jean Dufournet. Paris: Garnier-Flammarion. Christian von Troyes, Sa¨mtliche Werke nach allen bekannten Handschriften herausgegeben von Wendelin Foerster, t. 3. Niemyer: Halle.

Lancelot. 1991 [c. 1215–1235] Lancelot du lac. Paris: Livre de Poche. Lai du Bisclavet. 1820 [c. 1160–1175] Poe´sies de Marie de France Edited by Jean-BaptisteBonaventure de Roquefort, p. 180. Paris. Roland. 1985 [end.11th c.] La chanson de Roland. Edited by Ge´rard Moignet. Paris: Bordas. Studies Corblin, Francis, Viviane De´prez, Henrie¨tte de Swart & Lucia Tovena 2004 Negative concord. In: Francis Corblin & Henrie¨tte de Swart (eds.), Handbook of French semantics, 417–452. Stanford: CSLI. De´prez, Viviane 1999 The roots of negative concord in French and French based Creoles. In: M. Degra¤ (ed.), Language creation and language change: Creole, diachrony and development, 375–428. Cambridge: MIT Press. De´prez, Viviane 2000 Parallel (a)symmetries and the structure of negative expressions. Natural Language and Linguistic Theory 18 (2): 253–342.

Negative words and negation in French

75

De´prez, Viviane 2003 Concordance ne´gative, syntaxe des mots-N et variation dialectale. Cahiers de linguistique franc¸aise 25: 97–118. De´prez, Viviane & France Martineau 2004 Micro-parametric variation and negative concord. In: Julie Auger, J. C, Clements & Barbara Vance (eds.), Contemporary approaches to Romance linguistics, 139–158. Amsterdam/Philadelphia: John Benjamin. Englebert, Annick 1982 L’opposition ne/ne pas en ancien franc¸ais: une question d’incidence? B.A. Thesis, Universite´ libre de Bruxelles. Espinal, M. Teresa 2000 Expletive negation: Negative concord and feature checking. Catalan working papers in linguistics 8: 47–69. Foulet, Lucien 1968 Petite synaxe de l’ancien franc¸ais. Paris: Champion. Godard, Danie`le 2004 French negative dependency. In: Francis Corblin & Henrie¨tte de Swart (eds.), Handbook of French semantics, 351–389. Stanford: CSLI. Hansen, Maj-Britt Mosegaard 2009 The grammaticalization of negative reinforcers in Old and Middle French: a Discourse-Functional approach. In: Maj-Britt Mosegaard Hansen & Jacqueline Visconti (eds.), Current trends in diachronic semantics and pragmatics, 227–251. Oxford: Emerald. Kayne, Richard 1984 Connectedness and binary branching. Dordrecht: Foris. Schapansky, Nathalie 2002 The syntax of negation in French: Contrariety versus contradiction. Lingua 112: 793–826. De Swart, Henrie¨tte, & Ivan A. Sag 2002 Negation and negative concord in Romance. Linguistics and Philosophy 25: 373–417.

Secondary negation and information structure organisation in the history of English Ans van Kemenade 1. Introduction In this paper, I re-examine change in secondary clause negation at various stages in the history of English. Much has been written about this topic, e.g. Frisch (1997), Hulk & van Kemenade (1997), van Kemenade (1999), Haeberli (2000), van Kemenade (2000), van Gelderen (2008), Haeberli & Ingham (2007), Wallage (2005), (2008). Most of the work cited above adopts the concept of a NegP (a notable exception is Frisch 1997), which accommodates a negative head, in early English the preverbal clitic negator ne, and a secondary negator, the precursor of present-day English not. The concept of NegP provides an appealing and insightful framework for understanding those stages of the history of English where clause negation features two negative markers: ne and some form of secondary negation, and for the shifting distribution of these two negative markers over the history of English. A pivotal matter in this discussion is the syntactic status and position of the secondary negation marker. In Old English, secondary clause negation is a minor pattern and as such is expressed by na (and spelling variants thereof, primarily no; I will refer to this as na), which is also employed as a (negative) marker in contexts of comparison and contrast (with a reading: ‘not this, but that’). The Old English cognate of not is nawiht, literally ‘no creature’, which in Old English is employed as an emphatic negator, meaning something like ‘not at all’. It is fair to say, then, that there is no direct formal continuity between the Old English secondary clause negator na and not as it appears from early Middle English onwards. The aim of this paper is twofold: the first and foremost is to re-examine secondary clause negation in Old English, in particular the status of the marker na in Old English. In van Kemenade (2000), this is treated on a par with not as it appears in Middle English and later stages, and is analysed as a SpecNegP element, where NegP is located in a high position in the clausal architecture throughout the Old and Middle English periods, dominantly following pronominal subjects and preceding nominal sub-

78

Ans van Kemenade

jects. Here, the status of the marker na is reconsidered in the light of its similarity in behaviour to discourse particles such as þa and þonne ‘then’, as analysed in van Kemenade & Los (2006), van Kemenade (2009), van Kemenade & Milicev (2010). It will be argued that the secondary negator na in Old English has two basic uses: one in which it is a secondary clause negator low in the clause, and one in which it is used as a discourse particle that is relatively high in the clause, often in contexts of comparison and contrast. The first use is attested in main clauses without inversion of subject and finite verb, and is almost categorical in sub-clauses. The second use is restricted to root clauses with V to C movement. The transition to Middle English is also considered in this light, as it witnesses substantial loss of discourse particles: not, the grammaticalised version of the Old English emphatic negator nawiht, takes the place of Old English na as a secondary clause negator. At that stage, we begin to see evidence that not in V to C movement contexts is part of a NegP relatively high in the structure. We will consider this development as an instance of grammaticalization. The second aim of the paper is to re-examine the status and position of the secondary negator in the clausal architecture, following up van Kemenade (2000) and responses to this, in particular Haeberli & Ingham (2007) who argue, mostly on the basis of evidence from sub-clauses in the earliest period of Middle English, that the position of NegP is considerably lower than is claimed by van Kemenade (2000). The Middle English evidence shows a continuation of the Old English pattern: the position of not is high in root clauses with V to C movement, now also including negative questions, and it is low in other contexts. Both of these patterns are continued well into the early Modern period, until negative contraction progressively obliterates the erstwhile positional contrasts. We will analyse the rise of negative contraction as a progressive (grammaticalising) reanalysis of not as a negative head. The paper is organised as follows: Section 1 provides the backdrop for the treatment in this article. Section 2 addresses negation in Old English, including secondary clause negation. Section 3 discusses the grammaticalization of not in the subsequent history of English; section 4 concludes.

2. Background In this section, I briefly sketch the discussion on clause negation in the history of English as in the literature cited above, by way of background to

Secondary negation and information structure organisation

79

the argument presented here. Clause negation in Old English is predominantly marked by the preverbal clitic negator ne. Van Kemenade (2000, and see also Haeberli 2000), mainly studying root clauses, argues that the pattern of multiple sentential negation with ne and not in Middle English is foreshadowed as a minority pattern in Old English in the form of a combination of ne and na. We briefly consider two Old English examples in (1) in the clausal architecture as in (2): (1) a. þonne ne miht þu na þæt mot ut ateon of ðæs mannes eagan then not could you not the speck out draw of man’s eye ‘then you could not draw the speck out of man’s eye’ (ÆHomP.XIII.153) b. Ne sæde na ure Drihten þæt he mid cynehelme oððe not said not our Lord that he with diadem or mid purpuran gescryd, cuman wolde to us with purple clothed, come wanted to us ‘Our Lord said not that He would come to us with a diadem or clothed in purple.’ (ÆLS.XXXI.762) (2)

80

Ans van Kemenade

The head of NegP is the negative marker ne, an inflectional prefix almost never separated from the finite verb, which is checked against the negative head, overtly in clauses with syntactic V-movement, covertly in others. (2) has syntactic V-movement: ne miht is base-generated as the head of the VP, and is moved overtly, to T, Neg, F, and C (negative-initial clauses have V to C movement). Personal pronouns in this analysis are in SpecFP, where F should be taken for the time being as a functional slot. The element na is the secondary negator occupying the specifier of NegP; its position is therefore fixed. In this structure, (1b) is derived as follows: Ne is treated as a syntactic constituent, the first constituent in SpecCP (for arguments, see van Kemenade 2000); the finite verb sæde has been moved to C, as typical of questions, negative-initial clauses and clauses introduced by þa/þonne in Old English; na is the secondary negator; ure Drihten is the DP-subject in SpecTP. Pronominal subjects (nearcategorically) occur on the left of the secondary negator – this position is dubbed SpecFP here (we will return to this below). This structure is essentially maintained in Middle English, as in (3): (3) a.

b.

ne þarf þu naut dreden þt attri neddre of helle not need you not fear the venomous snake from hell ‘You need not fear the venomous snake from hell.’ (CMANCRIW-1, II.108.1354) nule nawt þi leofmon þoli na leas not-will not your beloved tolerate no false lihe þe longe deceive you long ‘Your beloved will not allow any false thing to you for long’

þing ta thing to

deceive (Juliana.33.332)

In these early Middle English examples, the position of noht/nawt (henceforth referred to as not) is identical to that of na in Old English: the pronominal subject in (3a) precedes the secondary negator naut, whereas the nominal subject þi leofmon in (3b) follows nawt. Van Kemenade (2000) furthermore quotes evidence for a lower negation marker, especially in Middle English, following an object pronoun. Examples are given in (4): (4) a. b.

And freten hym, for that they knewe hym naught. (Chaucer, Knight’s Tale 2068) I woot right wel, thou darst it nat withseyen (Chaucer, Knight’s Tale 1140)

Secondary negation and information structure organisation

81

Such examples are analysed as involving a low negative adverb, perhaps in the specifier of a second NegP lower in the clause, immediately above VP (I will remain agnostic on its status here). Patterns such as those in (4) then involve scrambling of the pronominal object across the negator, in the case of (4a) also involving V to T movement.1 Haeberli & Ingham (2007), following up Haeberli (2000) and van Kemenade (2000) note, for the earliest period of Middle English (AD 1150–1250), that these analyses make predictions that turn out to be inappropriate for sub-clauses in that period: pronominal objects, though not nominal objects, may occur in a position preceding not as well. Haeberli & Ingham analyse this in terms of a NegP below T. In this structure, both subject positions precede the secondary negator, as in (5): (5) a.

b.

Gif ðat hali writ ne wiðseið ðe naht if that holy text NEG prevents you not ‘if that holy text does not prevent you’ (CMVICES1, 101.1223) ac it ne openede hem noht þe blisse of heuene but it NEG opened them not the bliss of heaven ‘But it did not open the bliss of heaven to them’ (CMTRINIT, 87.1165)

Haeberli & Ingham extend this analysis to main clauses, including those where the nominal subject follows not, and thus account for secondary negation in terms of a low NegP throughout, arguing that the pattern with low negation noted by van Kemenade (2000) for late Middle English is already the most typical one in early Middle English. We will come back to this issue in section 3.

3. Secondary negation in Old English I now first examine the status of secondary clause negation by na in Old English. Clause negation in Old English is expressed predominantly by the preverbal clitic negator ne alone, as in (6) (from van Kemenade (2000):

1. This shows that in the M1 period, the object shift pattern that is contingent on V-movement, as we will discuss in section 4, was not yet in place.

82

Ans van Kemenade

(6) a.

ne sende se deofol ða fyr of heofenum, þeah þe not sent the devil then fire from heaven, though that hit ufan come it from-above came ‘the devil sent not fire from heaven, though it came from above’ (ÆCHom.i.6.13)

b.

Nolde se Hælend for his bene swaþeah hym not-wanted the Lord for his prayer however him fram gewitan from depart ‘The Lord, however, did not want to honour his prayer to go from him.’ (ÆHom.Pope.XIV.199)

Multiple clause negation is a minority pattern in Old English. The Old English cognate of not, the negative noun nawiht ‘no creature’, is used as a negated noun as in (7), or an emphatic negative adverb as in (8). (7) Eala leof lareow, ealle niht we swuncon, on idel wacigende, lo, dear teacher, all night we toiled, in vain watching, and we naht ne gefengon. and we naught not caught ‘Lo, dear teacher, we worked all night, keeping vigil in vain, and we caught nothing.’ (coaelhom, ÆHom_15:19.2145) (8)

þæt he þone manfullan Bretta cyning mid his unmætum weorode, that he the sinful of-Brits king with his monstrous host, þæm he gealp þæt him nowiht wiðstandan meahte, which he boasted that him not-at-all resist could, ofslog and acwealde in þære stowe, defeated and killed in that place ‘that he defeated and slew the sinful king of the Britons there, along with a monstrous host which he boasted was invincible.’ (cobede, Bede_3:1.154.13.1477)

The origin of not in no-wiht should not be taken to suggest that no is a just a constituent negator: although na/no is also employed in contexts of contrast and comparison (as discussed below), negation of nouns in Old English is largely expressed by negative quantifiers inflected as adjectives.

Secondary negation and information structure organisation

83

Clause negation with another negative marker beside ne thus seems to be restricted to cases with na as illustrated in (1). Recall that the position of na in Old English root clauses was taken by van Kemenade (2000) to indicate the position of NegP in the clausal architecture. More recent work shows that the positional di¤erences between pronominal and nominal subjects and objects are evident in a wider range of constructions in Old English. In particular, van Kemenade & Milicev ([2005] 2011), van Kemenade & Los (2006), and van Kemenade (2009) show that there is a range of clause-typing and modal uses of adverbs in Old English that in terms of word order serve to separate discourse-given and discourse-new subjects and objects of various types. With respect to their word order evidence, two such adverbs have been studied in particular: þa and þonne ‘then’. We will here take the results of these case studies as a model, to see if further insight into the properties of Old English na can be gained by comparing them with those of the adverbs þa and þonne. We will therefore first summarise the argument with respect to these adverbs, before we move on to a closer examination of clause negation with na in Old English. 3.1. Excursus on þa/þonne as discourse particles in Old English Van Kemenade & Milicev ([2005] 2011) identify di¤erential subject positions in Old English, following up Haeberli (2000) and van Kemenade (2000): a high position where pronominal subjects normally appear (cf. SpecF in (2)), and a low one mostly for nominal subjects (cf. SpecT in (2)). These positions are distinguished in several contexts in Old English: one is that of root clause questions, illustrated in (9) and (10), where the finite verb has been moved to the highest functional head C, and a pronominal subject (and optionally a pronominal object) normally precedes the discourse markers þa/þonne ‘then’, while a nominal subject most typically follows them. As discussed in van Kemenade & Los (2006), the two subject positions are also found in sub-clauses, illustrated in (11) and (12). (9) Hu mæg he ðonne ðæt lof & ðone gilp fleon. how may he then praise and vainglory avoid ‘How can he avoid praise and vainglory?’

(CP. ‘9.57.18’)

(10) Hu gerades mæg ðonne se biscep brucan ðære hirdelican are. how properly may then the bishop enjoy of-the pastoral dignity ‘How, then, can the bishop properly enjoy the pastoral dignity?’ (CP. ‘18.133.3’)

84

Ans van Kemenade

(11) He ne mihte swaþeah æfre libban, þeah ðe he hine þa ut alysde, he not-could nevertheless ever live, though they him then released ‘Nevertheless, he could not live forever, though they then released him.’ (coælive, ÆLS[Ash_Wed]:119.2763) (12) Gif him þonne God ryhtlice & stræclice deman wile. if him then God justly and strictly judge will ‘If God will then justly and strictly judge him.’ (CP. ‘5.45.20.257’) Table 1 illustrates the distribution of subject types over the two positions in root questions in the York Corpus of Old English (YCOE, Taylor et al 2003). Pronominal subjects occur almost exclusively in the position preceding þa/þonne (98.9%), while nominal subjects may appear in either position, with a preference for the lower one (82%). Table 2 (based on van Kemenade & Los 2006: 2312) shows a similar distribution in sub-clauses. Table 1. Order of subject and þa/þonne in root clause questions in OE DP subject

Pro subject

Subject-þa/þonne

11 (18%)

90 (98.9%)

þa/þonne-subject

50 (82%)

1 (1.1%)

61 (100%)

91 (100%)

Total

Table 2. Order of subject and þa/þonne in sub-clauses in OE DP subject

Pro subject

Subject-þa/þonne

129 (36%)

1116 (99.6%)

þa/þonne-subject

229 (64%)

Total

358 (100%)

5 (0.4%) 1121 (100%)

Van Kemenade & Los (2006), following up van Kemenade & Milicev ([2005] 2011) argue that this reflects a discourse structuring strategy: constituents left of þa/þonne are interpreted as discourse-given. The following types of arguments may appear on the left of þa/þonne: 2. The numbers are lower than those in van Kemenade & Los (2006): at the time, we were not aware that the IP-level in root clause questions is coded as a sub-clause in YCOE. The numbers here only include sub-clauses.

Secondary negation and information structure organisation

85

– – – –

Personal pronouns (subject and, optionally, object) Indefinite pronouns Impersonal pronouns (the subject man) Demonstrative pronouns (independently used subjects and, optionally, objects) – Some definite DPs

We first consider what we mean by ‘discourse-given’ (Given) vs. ‘discoursenew’ (New), amidst the proliferation of terms in the literature. They refer broadly to information that is known or presupposed (Given) and pragmatically unrecoverable (New). The notion New in particular is not a primitive, but allows further breakdown into whether the information is discourse-new or addressee-new, which is why Lambrecht (1994) prefers the term ‘‘pragmatically unrecoverable’’. Lambrecht points out that entities previously mentioned in the discourse can still be New in the sense that their association with a particular topic is new. There is a good deal of overlap in the literature between Givenness and topichood. The term ‘topic’ is here reserved for ‘aboutness topics’. These are defined by Gundel (1988: 210) as follows: ‘‘An entity E. is the topic of a sentence, S, i¤ in using S the speaker intends to increase the addressee’s knowledge about, request information about, or otherwise get the addressee to act with respect to E’’ (cf. also Reinhart 1982). Topics tend to be Given, but they do not have to be. Clauses may introduce a new entity while at the same time using it to denote a topic as in (13), which introduces the topic in the Common Ground content (Krifka 2007: 42): (13) [A good friend of mine] Topic [married Britney Spears last year]Comment . New information is most typically presented as comment, saying something about the topic. We use the term ‘‘focus’’ for constituents that are highlighted in some way, which is accomplished in present-day English by prosodic and/or syntactic marking (clefts). The e¤ect of such marking includes meanings of contrast and exhaustive identification. Krifka (2007: 44) characterises the various types of focus as indicating the presence of alternatives, as in (14): (14) A: B:

What do your siblings do? [My [SISter]Focus ] Topic [studies MEDicine]Focus , and [my [BROther]Focus] Topic is [working on a FREIGHT ship]Focus

86

Ans van Kemenade

In the first clause of B’s response in (14), focus on sister indicates an alternative to the topic ‘‘my sister’’, namely ‘‘my brother’’, and this prosodic marking is used by the speaker as a signal to the hearer that the answer is not finished with the first topic (the sister) but will also include information on another topic (the brother) (Krifka 2007: 44). This definition usefully includes focus-marked constituents that are also topics, like the topicalized object Baseball in (15): (15) G: Do you watch football? E: Yeah. Baseball I like a lot better.

(Birner & Ward 1998: 38)

Birner and Ward (1998) note that the contrast in (15) evokes partiallyordered sets: the earlier mention of football evokes the full category of sports. Krifka’s label ‘‘focused topics’’ fruitfully combines these aspects. A central concern of information structure is partitioning Given from New information. The general consensus seems to be that information structure is utterance-internal and its locus is the clause, whereas discourse structure concerns patterns of cohesion beyond the clause level. The two are closely linked because activation status and accessibility of referents require information from the previous discourse. Returning to our adverbs þa/þonne, we claim that they serve to partition Given and New information. As for the arguments that may precede or follow them, the first four types of argument are Given.3 The position of definite DPs is variable with respect to the diagnostic adverb.4 Here, van Kemenade & Milicev note that the definiteness marker in the Old English DP belongs to the paradigm of the weak demonstratives (the se paradigm), which is marked for case, number and gender; weak demonstratives can also be used as independent pronouns and as relative pronouns. This versatile use of the paradigm shows that it represents an independent strategy of pronominal reference: when used as a definiteness marker, it allows (but does not force) a definite DP to be Given in the

3. This is less than clear for indefinite pronouns. Van Kemenade & Milicev (2010) show that the cases of indefinite pronouns preceding the diagnostic adverb reflect a free choice from a fixed set of referents given in the discourse. Man is analysed as an impersonal pronoun, whose reference can always be inferred from the context. 4. Independently used demonstrative pronouns include cases where the pronoun is the antecedent of a relative clause. An example of this is (26) below. It is the relative clause that grounds the demonstrative antecedent in the discourse.

Secondary negation and information structure organisation

87

sense that it has specific reference to a discourse antecedent. Given definite DPs occur on the left of þa/þonne, whereas New or focussed definite DPs occur on their right. Examples (16) and (17) illustrate this: (16) Gif ðonne se sacred bið ungerad ðæs lareowdomes, if then the priest is unskilled in-the instruction, hwæt forstent ðonne his gehlyd? what avails then his cry? ‘If the priest is unskilled in instruction, what will his cry avail?’ (cocura, CP, 15,91,25) (17) Þa se biscop þæt when the bishop that þa licode him seo then pleased him the

þa geseah, þe him big sæt, then saw, who him by sat, arfæste dæd þæs cyninges; virtuous deeds of-the king (cobede, Bede_3:4.166.8.1593)

The definite subject DPs in (16) and (17) have di¤erent readings: (16) is about priests in a generic sense, ‘the priest in his o‰ce as priest’; this clause is not about a specific priest. (17), on the other hand, is about one specific bishop mentioned before in the discourse (the one who is sitting next to him). Corpus analysis reported in van Kemenade, Milicev and Baayen (2008) yields strong statistical support for this correlation: definite DPs preceding the adverb þa/þonne are Given by having specific reference to a discourse antecedent; definite DPs following þa/þonne have a generic reading and are thus New. Old English thus has an inflectional paradigm that provides an additional strategy of pronominal reference, and it has a set of adverbs that serve as discourse partitioners. Van Kemenade & Milicev ([2005] 2011) propose that adverbs like þa and þonne structure parts of the utterance/proposition itself with respect to the preceding discourse, as in (18). (18) [previous discourse] [utterance presupposition þa/þonne focus] We will now consider the analysis of such discourse-partitioning adverbs within the clausal architecture. Here, we point to their similarity to modal particles in German. The formal modelling of modal particles is surrounded by a variety of complex issues. We briefly discuss these here, leaving more detailed consideration for further study of a wider range of particles.

88

Ans van Kemenade

The Old English facts concerning þa and þonne support the following observations: semantically they can be used in their literal sense as temporal adverbs, but in their use as discussed here, they also have modal semantics. Bayer & Obenauer (in press) and work cited there, shows for German modal particles that they express the speaker’s attitude about him/herself or about the hearer with respect to the propositional content of the utterance. Thus, German questions with denn are not necessarily requests for information; they signal reference to some common ground with the hearer. Their semantics is closely related to illocutionary force. This may be fruitfully compared with the Old Englis questions in (9) and (10) above, which are rhetorical questions and thus assume a common ground between speaker and hearer. In their use as modal particles, þa/þonne serve a discourse-partitioning use as motivated in the references cited above. In this use, they lose a great deal of their lexical adverbial meaning and acquire modal meanings: they are semantically bleached, which suggests that they are grammaticalised. One current way of conceptualising the special syntax of modal particles in a formal syntactic framework is to associate adverbs with functional slots in the clausal architecture (cf. the clausal hierarchy argued for in Cinque 1999). Such an approach to discourse particles is adopted in Aboh (2004, 2006) for Gungbe, and in Bayer (in press) for modal particles in German. Bayer (in press) analyses the German modal particle denn in present-day German questions: in a sequence of functional heads as in (19), denn is the head of a PrtP. It is not an incorporating head – it cannot incorporate the finite verb under V to C movement.5 The particle must be c-commanded by a Question-feature. Denn as a functional head is preceded by an aboutness topic, which must move to the topic field preceding the particle, and is followed by the extended VP or another adverb or particle, under observance of Cinque’s hierarchy. (19) [FinP Fin [TopP topic* [PrtP [Prt denn] [VP (ext) . . . topic* . . . ]]]] In this approach, word orders are derived in a one to one association with functional slots in a hierarchy of functional features: the linear order 5. Bayer & Obenauer (in press) extend this analysis to what they call ‘special questions’, questions which are not interpreted as simple requests for information. In these question types, modal particles in German may form a constituent with a wh-phrase, which is subsequently moved to SpecCP. This is a pattern that also occurs in Northern Middle English, as discussed in de Haas (forthcoming).

Secondary negation and information structure organisation

89

of topic and particle is derived by their respective positions in the functional hierarchy. I will follow up this analysis for the purposes of my treatment here, with one proviso: I will assume that the particle is in a specifier position. This allows an account of the broader range of uses of þa/þonne in terms of semantic bleaching as the result of movement to a functional specifier, whilst their basic meaning is preserved, and it leaves open the possibility that over time, modal particles may grammaticalise to functional head status, as seems to be the case in German.6 With respect to Old English na in its high position in root V to C contexts, I will implement this as follows: na may move to the specifier of a mood-related position associated with contrast, which takes contrastively focussed entities in its scope. This will be further motivated in the following section. Na has an unvalued feature that is valued by a negative operator in C. Note that this implies that high na only occurs in negative-initial clauses with V to C movement. For Old English, this is correct: na is rare in questions and other types of V to C contexts. The total of five examples found are compatible with a high position, but not conclusive, and they have a secondary negation reading, not a contrastive reading: the contrastive meaning contribution apparently precludes this.7 The situation for Middle English is di¤erent, and will be discussed in section 4. (20) [CP Op Vf ] [ topic area [ PrtP na [ Prt0] [. . . . Vf ]]] We now turn to discussion of na in Old English. 6. Detailed consideration of this issue is a separate matter beyond the scope of this article. The underlying idea is that the onset of semantic bleaching is the result of movement to a functional specifier, a step in a grammaticalization process (cf. van Kemenade 2000). While the basic and literal use of þa/þonne is that of a temporal adverb, movement to the specifier of a mood-related position involves bleaching to modal semantics; movement to SpecCP yields a bleached meaning of signalling discourse continuation, as in the very frequent pattern with þa/þonne introducing a V to C clause (van Kemenade & Los 2006). 7. Note however, that this includes one example of a rhetorical negative question, as in root questions with þa/þonne as in (9)–(10). (i)

Ac ne geseoþ ge na þone eadigan Petrum & Paulum þa ealdormen þara apostola? but not see you not the blessed Peter and Paul the princes of the apostles ‘but don’t you see the noble Peter and Paul, the principal among the apostles?’ (cogregdC, GDPref_and_4_[C]:12.277.4.4049)

90

Ans van Kemenade

3.2. Secondary negation by na In the evidence discussed above, the positional similarity between na and discourse adverbs is at first glance striking. I will first explore this further. The data here are based on exhaustive searches for na in the York Corpus of Old English (YCOE, Taylor et. al 2003). The text referencing is from this corpus. Na is marked in YCOE as an adverb phrase ADVP dominating Neg þ ADV, which appropriately distinguishes it from negative quantifiers. Na hardly ever combines with þa/þonne in one clause and, in the handful of examples, no clear relative order can be established. We focus first on negative root clauses with V to C movement, since it is there that na primarily serves a diagnostic function for the position of the subject (na is rare in root questions, as noted above). Although na nearly always combines with the preverbal clitic negator ne, there are cases where it does not seem to have clausal scope. Two contexts stand out here: na may precede another adverb as in (21a), or a quantifier, as in (21b). (21) a.

b.

Þa ferde Martinus na swyðe feor þanon then departed Martin not very far from-thence ‘Then Martin departed not very far from thence’ (coaelive, ÆLS_[Martin]:444.6248) þonne ne bið þara fæstendaga na ma þonne syx & þritig. then not are of-the fast-days no more than six and thirty ‘then there remain no more than thirty-six of the fast-days’ (coblick, HomS_10_[BlHom_3]:35.159.472)

This context forces a reading in which na modifies the adverb or quantifier only. These cases were therefore excluded from the search. In the second context, na seems to act as a conjunction contrasting two constituents or clauses. Two examples of this are given in (22). (22) a.

Ne fæsð se no Gode ac him selfum, not fasts he not-for God but himself, se ðe ðæt nyle ðearfum sellan ðæt he ðonne on mæle læfð, who that not-will the poor give what he then of meal leaves, ac wile hit healdan eft to oðrum mæle, but will it keep then for another meal, ðæt he eft mæge his wambe mid gefyllan. that he afterwards may his belly with fill

Secondary negation and information structure organisation

91

‘He fasts not for God but for himself, who will not give the poor what he leaves of his meal, but wishes to keep it for another meal, to fill his belly with it afterwards.’ (cocura, CP:43.317.3.2120) b.

Ne sohte crist na ða modigan, not sought Christ not the proud, þa ðe mycele beoð on hyra geðance; those who great are in their imagination, ac þa þe beoð lytle and eaðmode on heora heortan: but those who are little and humble in their hearts: þa cumað to godes rice: these come to God’s kingdom; ‘Christ sought not the proud, those who are great in their own imagination, but those who are little and humble in their hearts, these shall come to God’s kingdom.’ (cocathom1, ÆCHom_I,_9:250.53.1617)

In (22a), two dative NP’s are contrasted: ‘not God, but himself ’. In (22b), two accusative NP’s embedding relative clauses are contrasted. It seems reasonable to think of na here as a negative conjunction marking the first member of a contrastive coordination. Given this, it is of considerable interest for our present purposes that in quite a few of these contrastive contexts, na does not immediately precede the first of the two contrasted constituents. An interesting illustration of this are (23a) and (23b): in these two examples, which apart from the position of na are identical and occur in the same sermon, the contrast is between the two PP’s na be hlafe anum, ac be ðam wordum ðem. In (23a), na immediately precedes the PP, but if we take that reading, it is remarkable that na in (23b) is in a position on the left of the subject. The same holds of (23c) where na introduces a contrast between two PP’s: æt nanes iudeisces mannes with ac æt rihtgelyfedra manna byrgenum. Note however, that na is between the subject God and the direct object þas wundra.

92

Ans van Kemenade

(23) a.

b.

Ne leofað se man na be hlafe anum, not lives the man not by bread alone, ac lyfað be eallum ðam wordum þe gað of godes muðe but lives by all the words that go from God’s mouth ‘Man lives not by bread alone, but by all the words that go from the mouth of God.’ (cocathom1, ÆCHom_I,_11:266.13.1988) ne not ac but

lifað lives lifað lives

na se mann be hlafe anum, not the man by bread alone, be ðam wordum ðe gað of godes muðe by the words that go from God’s mouth (cocathom1, ÆCHom_I,_11:267.50.2028)

c. Ne wyrcð God na þas wundra æt nanes iudeisces mannes not works God not these miracles at no Jewish man’s byrgene ne æt nanes oðres gedwolan. ac æt rihtgelyfedra sepulchre, nor at no other heretic’s, but at orthodox manna byrgenum [. . .] man’s sepulchres, ‘God works these miracles not at any Jewish man’s sepulchre, nor at any other heretic’s, but at the sepulchres of orthodox men.’ (cocathom1, ÆCHom_I,_20:344.253.4081) I hypothesise that in cases like (23b–c), na is not in the position marking the first member of the contrastive coordination, but occupies a position more to the left: the high position for na. To further explore the nature of the position for secondary na in root V to C contexts, I have therefore excluded cases like (22) and (23a), where na as far as we can tell marks the constituent, and I have included cases like (23b–c), as these may provide evidence for the high position of secondary na.8 With these restrictions in mind, I consider the figures for inverted negative main clauses with na and for sub-clauses, as given in tables 3 and 4: 8. This detailed check of cases of contrastive coordination where na may occur on the left of nominal subjects, was done only in the search file for nominal subjects. This is because pronominal subjects are near categorically in the position on the left of na, and we thus do not expect to find any added diagnostic value.

Secondary negation and information structure organisation

93

Table 3. Old English negative-initial root clauses with na DP subject

Pro subject

Man

Subject-na

39 (40.2%)

342 (100%)

7 (100%)

Na-subject

58 (59.8%)

0

0

97

342

7

Total N

The results are quite clear for personal pronoun subjects and for the impersonal pronoun man: these always precede na in both clause types, and I will not consider these in further detail. Nominal subjects present a very di¤erent picture: a majority of 59.8% follows na. It may be noted that this also includes clauses with unaccusative verbs, a context in which especially DP subjects, and more especially when they are focussed, may be in a lower position (van Kemenade 1997). But the facts in Table 3 show that the nominative in an unaccusative context (whether nominal or pronominal) is clearly sensitive to IS considerations, and transitive verbs are attested here as well. An unaccusative context is given in (24), and (25) gives an example with a transitive verb: (24) Root clauses. Subject-na Nis þis na gesæd be manna sawlum, not-is this not said of men’s souls, ac be manna lichaman þe formolsniað to duste, but of men’s bodies that moulder to dust ‘This is not said of men’s souls, but of men’s bodies that moulder to dust.’ (coaelive, ÆLS[Ash_Wed]:27.2717) (25) Root clauses. Na-subject Ne cwæð na se Symeon þæt Cristes swurd sceolde þurhgan not said not the Simeon that Christ’s sword should pierce Marian lichoman: ac hyre saule. Mary’s body, but her soul ‘The old Simeon said not that Christ’s sword should pierce through Mary’s body, but her soul. (cocathom1, ÆCHom_I,_9:254.176.1725) Given the correlation between the information structure status of the subject and its position with respect to the particle, as established quanti-

94

Ans van Kemenade

tatively in van Kemenade, Milicev & Baayen (2008), I will now consider whether the positional variation of subjects with respect to na can be accounted for on the basis of the IS status of the constituents preceding na. We start by observing that the fact that personal pronouns and man always precede na suggests that discourse-referential properties play a key role in distinguishing between the positions on the left and right of na. With this in mind, we take a closer look at DP subjects, checking the context in the texts. I will discuss the results in two subsections. 3.2.1. Negative root clauses with V to C movement, DP subject-na order In this context, the findings are entirely parallel to those for þa/þonne. The 39 cases numbered in Table 3 include 13 definite DPs, 18 independently used demonstratives, 3 proper names9, 5 possessive DPs, all referring to a discourse antecedent, including forward reference. An example of the latter is (26), where the subject is an independently used demonstrative pronoun that is the antecedent of a relative clause (26) Ne bið se no gefylled ðæs Halgan Gæsðæs se ðe not is he not filled of-the Holy Ghost who that on ðære smyltnesse his monðwærnesse forlæt ðone in the tranquillity of-his gentleness gives-up the wielm ryhtwislices andan fervour of-righteous zeal ‘He is not inspired with the Holy Ghost, who in the tranquillity of his gentleness gives up the fervour of righteous zeal.’ (cocura, CP:40.291.9.1911) 3.2.2. Negative root clauses with V to C, DP subject following na In the position following na, numbering 58 in table 3, we would expect to find DPs with a generic reading, bare NP’s, indefinites and so on. This is true in 37 cases, including some unaccusatives with late subjects. Two examples are given in (27):

9. Proper names tend to straddle the two positions, as is shown quantitatively in van Kemenade, Milicev & Baayen (2008).

Secondary negation and information structure organisation

95

(27) a.

Ne wende na Ezechias Israhela kyning ðæt he gesyngade, not thought not Hezekia Israel’s king that he sinned ða he lædde ða ællðeodgan ærenddracan on his maðmhus, when he led the foreign treasures into his treasury, & him geiewde his goldhord. and them showed his treasurers ‘Hezekiah, king of Israel, did not think that he sinned when he led the foreign treasurers into his treasury and showed them his treasures.’ (cocura, CP:4.39.2.197)

b.

Ne do na se Godes þeowa Godes þenunge for sceattum, not do not the Lord’s servant God’s ministry for treasures ac to þy þæt he geearnige þæt ece wuldor þurh þæt. but to that that he earn the eternal glory through that ‘The Lord’s servant shall not do God’s ministry for material gain, but to earn eternal glory through it.’ (colwsigeXa, ÆLet_1_[Wulfsige_Xa]:72.88)

In (27a), King Hezekiah is newly introduced and is mentioned as an example of how a ruler is distracted when he has too many things to attend to. In (27b), the DP-subject has a generic reading: ‘(what is typical of ) the Lord’s servant’. A further 7 cases have a subject previously mentioned as another term in the discourse, but where a pronominal object preceding na apparently overrides the IS status of the nominal subject, as in (28). (28) Ne gedyde næfre se mildheorta Dryhten, ne an his mode not did never the merciful Lord, nor in his mind ne gebrohte swelce hreowsunga, gif he hit æfter ðæm not brought such repentance, if he it after that auht swiðe wrecan wolde. He gecyðde swiðe mildheortlice aught severely punish wanted. He proclaimed very mercifully ðæt he him deman nolde, ða he gedyde ðæt hi him selfe that he them judge not-would, when he made that they them self ær beforan demdan. Be ðæm is awriten . . . & eft hit wæs before judged. About that is written . . . and again it was gecweden ðurh Sanctus Paulus: ðær we us selfum demden, said through Saint Paul: where we us self judged, ðonne ne demde us no God. then not judged us not God

96

Ans van Kemenade

‘The merciful Lord would never have caused or brought into his mind such repentance, if he wished afterwards to punish it with any severity. He proclaimed, very mercifully, that he would not judge them, when he made them judge themselves before. Of which is written. And again, it was said through Saint Paul: ‘‘When we judged ourselves, God judged us not.’’ ’ (cocura, CP:53.415.2.2865) This leaves 12 cases of DP subjects which have been previously mentioned in the discourse. These include some generic NP’s (‘man’, expressed as se mann; ‘the mass-priest’ (in his o‰ce as priest), 4 cases where the subject refers to a unique entity (God, the holy Father, the holy Ghost, the soul), but also 4 cases where a definite DP-subject might have specific reference to an antecedent. Two of them are proper names (se cyning Totila ‘king Totila’, se Symeon ‘Simon’) for which the context does not force a specific reading, a third is the antecedent of an immediately following restrictive relative clause (29a), and a fourth is (29b), in which the subject seo racu is repeated at intervals through the text, without clear reference to it as an antecedent. (29) a.

Nis na seo orðung ðe we utblawað & inn ateo ure sawul, not-is not the breath that we out-blow and in draw our soul Ac is seo lyft þe we on lybbað on ðisum deadlicum life but is the air that we on live in this mortal life ‘It is not the breath which we blow out and draw into our soul, but it is the air that we live on in this mortal life.’ (cotempo, ÆTemp:10.7.317)

b.

ac us ne segð na seo racu to hwam he hine sette, but us not tells not this exposition to what he it wrote, buton þæt he sealde soðe gebysnunge eallum dædbetendum, except that he gave true example to-all penitents ‘[. . .] but the story does not tell us why he wrote it, except to give true example to all penitents.’ (colsigewZ, ÆLet_4_[SigeweardZ]:1147.562)

My conclusion is that in negative root clauses with inversion, the position of the nominal subject with respect to na is consistent with the findings for þa/þonne as quoted above. I therefore claim that DP subjects preceding na are in the higher position in the topic area, whereas nominal subjects following na are in the lower subject position. The position of

Secondary negation and information structure organisation

97

na in root clauses with V to C-movement is thus high: above TP in the analysis of van Kemenade (2000). We now turn to sub-clauses. 3.2.3. Sub-clauses The figures for sub-clauses are given in Table 4: Table 4. Old English sub-clauses with secondary negator na DP subject

Pro subject

Man

Subject-na

53 (88.3%)

181 (100%)

7 (100%)

Na-subject

7 (11.7%)

0

0

60

181

7

Total N

Table 4 shows that the figures for nominal subjects in sub-clauses contrast sharply with those in negative root clauses with V to C movement: the DP subject precedes na in 97% of the cases and there are indeed only 7 cases where the subject follows na. Let us consider DP-subjects preceding na first: here too, this dataset presents a very di¤erent picture from that in negative root clauses with inversion. Although the nominal subject in many cases has specific reference to an antecedent, this is not true in a substantial 19 cases: this includes definite DPs with a generic reading as in (30), but it also includes quantified NP’s as in (31) and bare plurals as in (32). (30) Be ðæm suiðe wel wæs gecueden to Ezechiele ðæm witgan about this very well was said to Ezekiel the prophet ðætte ða sacerdas ne scoldon no hiera heafdu scieran mid that the priests not should not their heads shave with scierseaxum, ne eft hi ne scoldon hira loccas razors, nor on-the-other-hand they not should their locks lætan weaxan, ac hie scoldon hie efsigean mid scearum. let grow but they should them clip with scissors ‘Concerning which it was well said to the prophet Ezekiel that the priests were not to shave their heads with razors, nor, on the other hand, let their locks grow, but clip them with scissors.’ (cocura, CP:18.139.11.945)

98

Ans van Kemenade

(31) Hu ne wast þu þæt manig þing ne bið no ongiten how not know you that many things not are not perceived swa swa hit bið, ac swa swa þæs andgites mæð so so it is, but so so the meaning’s proportion bið þe þæræfter spyreð? is which there-behind goes? ‘How is it that you don’t know that many things are not perceived as they are, but according to the meaning behind them?’ (coboeth, Bo:41.145.5.2889) (32) Nis eac nan wundor þeah us mislympe, forðam we witan not-is also no wonder though us went-bad, because we know ful georne þæt nu fela geara men na ne rohton foroft well that now many years men not not cared very-often hwæt hi worhton wordes oððe dæde what they wrought of-words or of-deeds ‘It is no wonder that things went bad for us, since we know full well that for many years, people often haven’t cared what they say or do.’ (cowulf, WHom_20.2:127.1724) These findings contrast sharply with those for negative root clauses with inversion where the nominal subject precedes na, and which have specific reference to an antecedent. Since the division of labour between the various subject positions in root clauses with V to C movement is clear-cut, this must mean that in sub-clauses, na is lower than in root clauses with V to C movement. Note in this respect that in (32), two temporal adverbs (nu and fela geara) precede the subject, which in turn precedes na, suggesting that the position of na is low. This conclusion is reinforced by a further independent characteristic of this dataset: it contains a number of SubjectVf-na word orders that is unusually high for sub-clauses in Old English: 40 out of 53 examples (see e.g. examples (30) and (31)). On the assumption that the subject is in the higher position, or the lower subject position SpecT, and the finite verb has been moved to T, the conclusion is that na is below T in sub-clauses. This leaves seven examples of na in sub-clauses where the nominal subject follows the secondary negator. The properties of the nominal subject in these cases are consistent with those for nominal subjects following na in root clauses: they have no discourse antecedent; they are bare NP’s or, when they are definite, tend towards a generic reading. On our analysis

Secondary negation and information structure organisation

99

here, this would mean that na in these contexts is high. We will come back to this below. Two examples are given in (33): (33) a.

b.

& þonne ðu þas wyrte mid hyre wyrttruman niman wylle, and when you the root with her root-stock take want, ðonne warna þu þæt hy na sunne bescine, ðy læs hyre hiw then take-care you that it not sun beshines, lest her colour & hyre miht sy awend þurh ðære sunnan beorhtnysse. and her power be changed through of-the sun brightness ‘When you want to uproot the root, then be careful that the sun doesn’t shine on it, so that its colour and power are not a¤ected by the brightness of the sun.’ (coherbar, Lch_I_[Herb]:182.1.2618) Forðæm ðe na se ðorn ðære gitsunga ne wyrð because that not the thorn of-the greed not becomes forsearod on ðæm helme, gif se wyrttruma ne bið færcorfen withered in the crown if the root not is cut-o¤ oððe forbærned æt ðæm stemne. or burnt at the stem ‘[. . .] because the thorn of greed does not wither in the stem, unless the root has been cut o¤ or burnt at the stem.’ (cocura, CP:45.341.9.2292)

In (33a), an object pronoun precedes na, whereas the bare NP sunne follows it. In (33b), the subject se dorn þære gitsunga is mentioned for the first time in the discourse as a simile for how avarice should be dealt with. These findings raise the question whether the divergent positioning for na in root clauses with V to C movement and in sub-clauses respectively is related to V to C movement. We therefore now briefly consider root clauses with na in which the finite verb has not been moved to C. The main findings are given in table 5: Table 5. Old English root clauses, no V to C movement, with na DP subject

Pro subject

Man

Subject-na

115

191

2

Na-subject

6

2

1

121

193

3

Total N

100

Ans van Kemenade

These findings are very much in line with those for sub-clauses. Furthermore, as in sub-clauses, the incidence of subject-Vf-na word order is very high (102 out of 115). Again, on the assumption that the finite verb has been moved at least to T, this implies that na is below T. I therefore conclude that the high position for na is a characteristic of clauses with V to C movement. (34) gives an example: (34) Se halga wer þa cwæð, wif ne sceal na faran to wera the holy man then said, woman not shall not go to men’s fyrdwicum, ac wunian æt ham; camps, but remain at home ‘The holy man then said: a woman should not go to men’s camps but remain at home. (coaelive, ÆLS_[Martin]:1095.6683) 3.2.4. Discussion The case study in this section presents evidence that the position of na di¤ers between clauses with V to C movement (negative root clauses with inversion) and other clause types: in root clauses with V to C movement, na is between the higher and the lower subject position. In main clauses without V to C movement and in sub-clauses, on the other hand, the secondary negator is dominantly below the lower subject position, except perhaps in a few cases where the subject follows na. We have also seen evidence that the high position for secondary negation in root V to C contexts may be a derived one: in these contexts, na is high, whether it is a clausal negator or a contrastive constituent negator. The facts here are best accounted for by distinguishing two uses of na, corresponding to two di¤erent positions. In the case of clause negation, this is that of an adverb on the left of the VP, as in (35a). In the case of contrastive negation of constituents or clauses, na is an adverb contrastively marking a constituent or clause, as in (35b): (35a)

(35b)

Secondary negation and information structure organisation

101

In the higher position for na, it has the status of a modal particle associated with contrast, that takes in its scope focussed material which may consist of two propositions that are being contrasted, with a reading: ‘not this, but that’. Any non-focussed material, such as Given material, must move to the topic area, in line with (20) above: (35) [CP Op [C0 Vf ] [ topic area [prtP na [Prt0 ] [. . . . Vf ]]] Na has an unvalued feature for negation, which is valued by the negative operator in CP. This accounts for the fact that the high position for na is found almost exclusively in V to C contexts. The cases where the subject follows na, as in (33b) in a sub-clause, show that the higher position, with particle use, is not excluded in other contexts. What these often have in common is contrast with the context: (33b) above is a good example of how na has scope over two propositions that are being contrasted: the thorn of greed cannot wither in the crown if the condition that the root is cut o¤ is not met. (cf. Milicev in preparation for more detail). This shows that the precise licensing of the high position for na is perhaps more complex than I have so far made out here. I leave for further research a more detailed consideration of how na and other modal particles feature in contexts of contrast.

4. The grammaticalization of not in Middle English In the previous section, we have seen that there are clear distributional di¤erences with respect to the position of na between root clauses with V to C movement on the one hand, and other main clause and sub-clauses on the other hand. We will show here that, while the nature of the secondary negator changes substantially in the transition to Middle English, a di¤erence between V to C movement clauses and other clauses in the distribution of the secondary negator is largely maintained. This di¤erence, I claim, is much in line with the distinction made above for Old English. Recall that in negative main clauses with inversion in Old English, the articulation of two di¤erentiated subject positions as separated by na when used as a modal particle is fairly clearcut. Non/V to C contexts, on the other hand (main clauses without V to C movement and sub-clauses), are near-categorically subject-initial. Negative clauses with na without V to C movement thus seem to be well on the way to simply becoming subject-initial in Old English, with a low negation marker. For contexts with other modal particles in sub-clauses, this is a situation that occurs

102

Ans van Kemenade

only later, in the transition to Middle English (as shown by van Kemenade & Los 2006, van Kemenade 2009) for contexts with þa/þonne). The modal particle use of na is lost, in line with the fate of other Old English discourse particles in the transition to Middle English (see the discussion in van Kemenade & Los 2006 for þa/þonne), and na is found in Middle English as a negative quantifier marking constituent negation (much like no in present/day English as in no man is an island ). Not as the grammaticalised version of Old English nawiht as discussed above, is established as the secondary negation marker and expands its domain as clause negator in the course of the Middle English period. During the early Middle English period, it increases in use (for details, see Jack 1978a–c, and Iyeiri 2001). The use in contrastive contexts characteristic of Old English na in V to C contexts seems to be largely lost. I will assume, therefore, that the high negation position in V to C contexts is best characterised as NegP in Middle English. What is maintained in Middle English is the distinction between V to C contexts and non-V to C contexts. I now turn to the evidence for the higher and the lower negation position in Middle English. For the position of negation in sub-clauses, we will first consider the proposal by Haeberli & Ingham (2007): they note correctly that van Kemenade’s (2000) analysis in terms of high negation is problematic for sub-clauses in early Middle English, and they regard the high negation position as an archaic structure. Looking at the distribution of secondary negation in the first period of Middle English (AD 1150–1250), they find much evidence for a low negator that they characterise as NegP. This analysis is based primarily on the fact that patterns such as those in (36) (from Haeberli & Ingham 2007: 16) with object pronouns are found from early Middle English onward: (36)

þt ich ne seo hire nawt heonne-forð mare that I not see her not henceforth anymore ‘[. . .] that I will not see her anymore.’ (CMJulia, 123.489)

Haeberli & Ingham’s argument for a low secondary negator in sub-clauses in the M1 period is convincing: whether the subject is nominal or pronominal, Vf-Obj-not order with pronominal objects is attested on a large scale throughout the Middle English period and considerably beyond (it does not seriously decline before the middle of the seventeenth century as we will see below). And indeed, the same is true for subject-initial main clauses. However, Haeberli & Ingham’s argument faces problems with

Secondary negation and information structure organisation

103

inverted main clauses with not: first, in inverted main clauses with not, nominal subjects are still routinely found in the lower subject position. This is generally evident from the relatively high incidence of inversion with nominal subjects in V2 constructions until the late Middle English period (see van Kemenade and Westergaard to appear); in inverted main clauses with not, New DP subjects are found on the right of not until the end of the Middle English period and beyond, as the figures in Table 6 show (table 6 combines all root V to C contexts with not, including root questions which also begin to be found with secondary negation in Middle English). Table 6. Inverted main clauses in PPCME2 with not in Middle English, by period DP-subject-not

not-DP-subject

Total

M1

23

11

34

M2

15

4

19

M3

17

11

28

M4

1

5

6

The lower position for DP subjects seems incompatible with the assumption of a negation marker below T.10 Some examples are given in (37): (37) a.

Nalde nawt godd leoten his Martyrs licomes liggen not-wanted not God su¤er his martyrs’ corpses lie to Forleosen to perish ‘God would not tolerate that his martyrs’ corpses lie to perish’ (M1, CMKATHE, 49.482)

10. Haeberli & Ingham argue that in many cases, the nominal subject following not combines with an unaccusative verb and may be in SpecVP, following up van Kemenade (1997). This leaves a number of examples where a nominal subject following not combines with a verb that has an external argument. In these, they argue, the nominal subject must be in SpecTP, in line with the analysis here. They regard these cases, with a high position for not, and the nominal subject in SpecTP, as residual in the M1 period. We will see below, however, that in root contexts with V to C movement, nominal subjects appear in this position until well into the early Modern period.

104

Ans van Kemenade

b.

Before er man synnid, mizt not wille be disceyuid in his before ever man sinned, might not Will11 be deceived in its chesyng, in its ouyng, ne in none of his werkes choosing, in its loving, or in any of its works? (M3, CMCLOUD, 117.596)

Secondly and again in main clauses, the assumption of a low negation marker would seem to predict that in inverted main clauses as much as in subject-initial ones, pronominal objects occur on the left of not. This prediction is not borne out: pronominal objects in inverted main clauses precede not in any numbers only during the M1 period, less so in the M2 period. At that stage, pronominal objects could still occur in the high position for Given elements as in Old English, but this was lost except in unaccusative contexts. The figures are given in table 6, with some examples in (38). Although such examples are not necessarily incompatible with a low position for not, they contrast in several ways with non-V to C contexts. Admittedly, the overall number of V to C movement contexts with a subject and an object pronoun is not very high. Several factors account for this: the first of these is that the very frequent negative-initial clause pattern introduced by ne þ finite verb is declining after the M2 period, which severely reduces the number of V to C contexts over the course of the Middle English period. Furthermore, secondary negation only very sporadically occurred in questions in Old English and only begins to do so from the early Middle English period onward (Jack 1978a-b-c). But if we consider the pattern itself, there is no reason why the object pronoun should not readily appear on the left of not if the position for not is low. What we see, however, is that object pronouns may occur there during the M1 and M2 periods, but the last example of this pattern is in the M3 period. We could ascribe this to the fact that object shift to a higher position comes to be restricted to contexts with movement of the lexical finite verb, to a higher functional position, e.g. V to C. But movement of lexical finite verbs was still alive and kicking in the M3 period and only begins to decline in the course of the early Modern period. In contrast, we will see below in Table 7 that in non-V to C contexts, object pronouns robustly precede not until well into the early Modern period and beyond. With this in mind, we turn to object pronouns on the left of not, where I claim that not is in the high position:

11. Will is spelt with a capital here because it is allegorically personified.

Secondary negation and information structure organisation

105

Table 7. Middle English inverted root clauses with object pronoun preceding not in PPCME, per period V-S-Opro-not

V-S-not-Opro

Total

M1 AD 1150–250

5

5

10

M2 AD 1250–1350

7

7

14

M3 AD 1350–1420

1

9

10

M4 AD 1420–1500

0

6

6

(38) why þan fyndes þou hym noght? Why then find you him not?

(M24, CMROLLEP, 76.214)

We now contrast this with non-V to C contexts, where object pronouns precede not in robust numbers until the end of the early Modern English period. The figures in table 6 therefore show that the high position for the object pronoun is in the topic area preceding the high secondary negator in root V to C contexts, and not that of an object pronoun preceding a low secondary negator. Object pronouns preceding not in non-V to C contexts persist robustly in pre-not position to the end of the early Modern period and beyond. This has been analysed on a par with the object shift pattern well-known from the Scandinavian languages (Holmberg 1986; for a treatment of early Modern English, see Roberts 1996). In these analyses, object shift is taken to be contingent on V-movement: the object occurs on the left of NOT only if the finite verb has been moved to T.12 The pattern dies out gradually as the result of progressive loss of V to T movement and its incompatibility with auxiliaries, including periphrastic DO, which is strongly on the rise during this period (see e.g. Kroch 1989); Table 7 includes cases

12. Note, however, that the alternative pattern with Subj-Vf-not-Opro, where Vf represents a lexical finite verb in Table 7, would have to be derived by V to T movement as well. This suggests that V to T movement allows object shift, but object shift is not contingent on V to T movement.

106

Ans van Kemenade

with lexical finite verbs only, and it shows that object pronouns dominantly precede not as long as the pattern is around, which is beyond 1700. Table 8. The order of finite verb, not and object pronoun in main clauses without V to C movement in late Middle English and early Modern English, per period Subj-Vf-Opro-not

Subj-Vf-not-Opro

Total

M4 AD 1420–1500

26 (66.7%)

13

39

E1 AD 1500–1569

59 (62.1%)

36

95

E2 AD 1570–1639

88 (72.7%)

33

121

E3 AD 1640–1710

11 (73.3%)

4

15

I conclude from these sets of facts that in Middle English, as in Old English, the position for secondary negation in root clauses with V to C movement is higher than that in other contexts. The analysis advanced for this discrepancy in Old English seems entirely compatible with the facts as discussed here for Middle English: the secondary clausal negator in questions is in the specifier of a high functional head that separates Given from New material; in the course of the Middle English period, the position on the left of high not becomes restricted to pronominal subjects. The negative feature of this high Neg is valued by an operator in C, as in (39), where we assume that an interrogative operator is compatible with this: (39) [CP Op [C0 Vf ] [ topic area [NegP not [Neg0 Vf ] [. . . . Vf ]]] 4.1. Not becomes a negative head What we have seen here is a contrast between root V to C contexts and other contexts that is pervasive throughout the history of English. This has become clear when we consider in detail the properties of na in Old English, contrasting V to C contexts with others. This highlights the fact that the contrast between these two types of environments has been remarkably stable over time. Table 8 gives the figures for the position of pronominal subjects with respect to not, and they show that this relic of

Secondary negation and information structure organisation

107

discourse-flexible grammar is still robustly attested at the end of the early Modern period.13 Table 9. The order of pronominal subject and not in root V to C contexts in late Middle English and early Modern English Vf-Spro-not

Vf-not-Spro

Total

M4 AD 1420–1500

41 (91.1%)

4

45

E1 AD 1500–1569

188 (79.3%)

49

237

E2 AD 1570–1639

185 (86.4%)

29

214

E3 AD 1640–1710

66 (53.7%)

57

123

(40) a. b.

‘‘Damsel, knowyst þu not me?’’ ‘Damsel, don’t you know me?’

(M4, CMKEMPE, 118.2727)

knowe not ye how ye mysdeled on the plays? ‘Don’t you know how you cheated on the plays?’ (M4, CMREYNAR, 9.82)

The positional flexibility of subjects which we here ascribe to their information status started being lost towards the end of the Middle English period, as noted by van Kemenade (2000), following up Rissanen (1994, 1999). The four examples for M4 in the second column of table 8 represent the first instances of the final step in the grammaticalization of not, that of becoming a negative head. The analysis for the development apparent in table 8 is the following: once ne as a negative head had been lost in the course of the Middle English period, not came to be incorporated with the finite verb on its way to C. The NegP format provides an insightful account of this: Once ne has been lost as a negative head,

13. The finite verbs in Table 8 comprise lexical finite verbs as well as auxiliaries. As the loss of movement of finite lexical verbs to a higher functional position is lost over the course of the early Modern period (Kroch 1989), the pattern becomes restricted to auxiliaries, as noted in particular for do in Warner (2005).

108

Ans van Kemenade

not begins to show signs of assuming negative head status. In a structure like (40), when not becomes a head, it incorporates with the finite verb: (41) [CP [C0 Vf-not] [ topic area [NegP [Neg0 Vf -not] [. . . . Vf ]]] This process, essentially the rise of negative contraction, progresses further over the late modern period (Rissanen 1999) and progressively obscures the last remnants of the syntactically articulated left periphery found in V to C contexts with secondary negation spanning the earlier history of English. One implication of my analysis here is that the rise of negative contraction in negative questions, for which the figures in Table 8 provide evidence, is a separate development from negative contraction in non-V to C movement contexts. In negative questions as in (40), high not becomes a head element and comes to be syntactically incorporated with the finite verb as it is undergoes V to C movement via intermediate heads including Neg0. This is a change that starts in late Middle English, and progresses over the early Modern period, becoming restricted to auxiliaries in tandem with the loss of movement of lexical finite verbs. In non-V to C contexts, such as negative declaratives with a low negator as in Table 7, negative contraction arises as a result of the loss of the object shift pattern, by which finite verb and not become linearly adjacent. Note here too, that the object shift pattern is increasingly outnumbered by auxiliaries in T (including periphrastic do), which always yield Aux-not order. The date of this change must be later: the loss of V to T movement and the rise of periphrastic do are to be dated well into the early Modern period (Kroch 1989; Warner 2005), and the evidence from Table 7 shows that both patterns involving V to T movement do not shown dwindling numbers before the E3 period (late seventeenth century). This provides support for my analysis in terms of two di¤erentiated positions for secondary negation.14

5. Conclusions The account in this paper illuminates various aspects of secondary negation in the history of English. First, it gives more depth to the status and position of na as a secondary negator and a marker of contrast in Old English, by identifying a high position for na which seems to be restricted 14. Some independent support for this can be found in Han (2000) and Han & Kroch (2000), who study the rise of periphrastic do in imperatives.

Secondary negation and information structure organisation

109

to root V to C contexts, and a handful of contexts where a clause with na is somehow contrasted with the discourse. A second result is that the account here reconciles two ends of the debate on the position of secondary negation: where van Kemenade (2000) makes a case that the position for secondary negation is high, and Haeberli & Ingham (2007) show that in early Middle English, the position for not is low in most contexts. I have shown that there is a systematic and pervasive distinction between the position of secondary negation in root V to C contexts and other contexts. In Old English, this higher position is associated with contrastive negation, but allowed for clause negation as well. In Middle English, this position becomes a high clause negation position that we here continue to dub NegP. Further research will have to show how the high position of contrastive negation in Old English compares with the behaviour of other discourse particles and their role in clause structure.

Acknowledgements This material has been presented in whole or in part to the Leverhulme network Cycles of Grammaticalization, to the 7th Symposium on the History of English Syntax at the University of York (June 2010), and to various audiences at the University of Groningen (April 2010) and the Eberhardt Karls Universita¨t at Tu¨bingen (June 2010). I thank the audiences on these occasions. Thanks also to Eric Haeberli for his comments on a pre-draft, Josef Bayer, Richard Ingham, Erwin Komen and Pierre Larrive´e for their comments on an earlier version, and Theresa Biberauer, Roland Hinterho¨lzl, Bettelou Los, and Marit Westergaard for ongoing insightful and stimulating discussion on syntax and information structure. They are not responsible for any remaining errors.

References Aboh, Enoch Olade´ 2004 The Morphosyntax of complement-head Sequences. Clause structure and word order pattern in Kwa. New York: Oxford University Press. Aboh, Enoch Olade´ 2006 Complementation in Saramaccan and Gungbe: the case of c-type modal particles. Natural Language and Linguistic Theory 24 (1): 1–55.

110

Ans van Kemenade

Aboh, Enoch Olade´ 2008 Information Structure begins with the numeration. lingBuzz/ 000712 Allen, Cynthia L. 1995 Case marking and reanalysis: Grammatical relations from Old to Early Modern English. Oxford: Oxford University Press. Bayer, Josef In press From modal particle to interrogative marker: A study of German denn. In: Laura Bruge`, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro & Cecilia Poletto (eds.), Functional heads. Oxford: Oxford University Press. Bayer, Josef & Hans-Georg Obenauer In press Discourse particles, clause structure, and question types. The Linguistic Review. Bergen, Linda van 2000 Pronouns and word order in Old English, with particular reference to the indefinite pronoun man. Ph.D. dissertation, University of Manchester. Birner, Betty & Gregory Ward 1998 Information status and canonical word order in English. Amsterdam/Philadelphia: Benjamins. Cinque, Guglielmo 1999 Adverbs and functional heads: A cross-linguistic perspective. Oxford: Oxford University Press. Frisch, Stefan 1997 The change in negation in Middle English: A NegP licensing account. Lingua 101, 21.64. Gelderen, Elly van 2008 The negative cycle. Linguistic Typology 12 (2): 195–243. Gundel, Jeanette K. 1988 Universals of topic-comment structure. In: M. Hammond, E. Moravcsik & J. Wirth (eds.), Studies in syntactic typology, 209– 39. Amsterdam: Benjamins. Haas, N. de forthcoming Morphosyntactic change in Northern Middle English. Ph.D. dissertation, Radboud University Nijmegen. Haeberli, Eric 2000 Adjuncts and the syntax of subjects in Old and Middle English. In: Susan Pintzuk, George Tsoulas, & Anthony Warner (eds.), Diachronic syntax: Models and mechanisms, 109–121. Oxford: Oxford University Press. Haeberli, Eric & Richard Ingham 2007 The Position of negation and adverbs in Early Middle English. Lingua 117, 1–25.

Secondary negation and information structure organisation

111

Han, Chung-hye 2001 The evolution of do-support in English imperatives. In: Susan Pintzuk, George Tsoulas, & Anthony Warner (eds.), Diachronic syntax: Models and mechanisms, 275–295. New York: Oxford University Press. Han, Chung-hye & Anthony Kroch 2000 The rise of do-support in English: Implications for clause structure. Proceedings of the 30th North East Linguistics Society, 311–325. Amherst: GLSA. Holmberg, Anders 1999 Remarks on Holmberg’s generalization. Studia Linguistica 53: 1–39. Hulk, Aafke & Ans van Kemenade 1997 Negation as a reflex of clause structure. In: Danielle Forget, Paul Hirschbu¨hler, France Martineau & Maria-Luisa Rivero (eds.), Negation and polarity: Syntax and semantics, 183–208. Amsterdam/Philadelphia: Benjamins. Iyeiry, Yoko 2001 Negative constructions in Middle English. Fukuoka: Kyushu University Press. Jack, George 1978a Negative adverbs in Early Middle English. English Studies 59: 295–309. Jack, George 1978b Negation in later Middle English prose. Archivum Linguisticum (n.s.) 9: 58–72. Jack, George 1978c Negative concord in Early Middle English. Studia Neophilologica 50: 29–39. Kato, Kozo 1995 The interjection la and subject pronouns in Old English. In: H. Nakano (ed.), Linguistics and Philology, 23–40. Kogaka Shuppan. Kemenade, Ans van 1997 V2 and embedded topicalization in Old and Middle English. In: Ans van Kemenade & Nigel Vincent (eds.), Parameters of morphosyntactic change, 326–52. Cambridge: Cambridge University Press. Kemenade, Ans van 1999 Sentential negation and clause structure in Old English. In: Ingrid Tieken-Boon van Ostade, Gunnel Tottie & Wim van der Wur¤ (eds.), Negation in the history of English, 147–165. Berlin/ New York: Mouton de Gruyter. Kemenade, Ans van 2000 Jespersen’s cycle revisited: Formal properties of grammaticalization. In: Susan Pintzuk, George Tsoulas, & Anthony Warner

112

Ans van Kemenade

(eds.), Diachronic syntax: Models and mechanisms, 51–74. Oxford/ New York: Oxford University Press. Kemenade, Ans van 2002 Word order in Old English prose and poetry: The position of finite verbs and adverbs. In: Donka Minkova & Robert Stockwell (eds.), Studies in the history of the English language: A millennial perspective, 355–373. Berlin: Mouton de Gruyter. Kemenade, Ans van, Tanja Milicev & R. Harald Baayen 2008 Balance between discourse and syntax in Old and Middle English. In: Marina Dossena & Maurizio Gotti (eds.), English historical linguistics 2006. Amsterdam: Benjamins. Kemenade, Ans van & Tanja Milicev 2010 Syntax and discourse in Old and Middle English word order. In: Stephen Anderson & Dianne Jonas (eds.), Articles from the 8th diachronic generative syntax conference. Oxford: Oxford University Press. Kemenade, Ans van & Marit Westergaard To appear Syntax and information structure: The diverse nature of V2 in ME. In: Maria´-Xose´ Lopez-Couso, Bettelou Los & Anneli Meurman-Solin (eds.), Information structure in the history of English. Oxford: Oxford University Press. Krifka, Manfred 2007 Basic notions of information structure. In: Caroline Fery, Gisbert Fanselow & Manfred Krifka (eds.), The notions of information structure, 13–55. Potsdam Universita¨tsverlag. Kroch, Anthony 1989 Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Kroch, Anthony & Ann Taylor 2000 The Penn-Helsinki parsed corpus of Middle English 2. University of Pennsylvania, Department of Linguistics. Lambrecht, Knud 1994 Information structure and sentence form: Topic, focus, and the mental representations of discourse referents. Cambridge: Cambridge University Press. Reinhart, Tanya 1982 Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27: 53–94. Rissanen, Matti 1994 The position of not in Early Modern English questions. In: D. Kastovsky (ed.), Studies in Early Modern English, 339–48. Berlin: Mouton de Gruyter. Rissanen, Matti 1999 Isn’t it? or Is it not? On the order of postverbal subject and negative particle in the history of English. In: Ingrid TiekenBoon van Ostade, Gunnel Tottie & Wim van der Wur¤ (eds.),

Secondary negation and information structure organisation

113

Negation in the history of English, 189–205. Berlin/New York: Mouton de Gruyter. Rizzi, Luigi 1996 Roberts, Ian 1995

Parameters and functional heads: Essays in comparative syntax. Oxford: Oxford University Press.

Object movement and verb movement in Early Modern English. In: Hubert Haider, Susan Olsen & Sten Vikner (eds.), Studies in comparative Germanic syntax, 269–84. Dordrecht: Kluwer. Taylor, Ann, Anthony Warner, Susan Pintzuk & Frank Beths 2003 The York-Toronto Helsinki parsed corpus of Old English. York: Department of Language and Linguistic Science. Available through the Oxford Text Archive. Wallage, Phillip 2005 Negation in Early English: Parametric variation and grammatical competition. Ph.D. dissertation, University of York. Warner, Anthony 2005. Why DO dove: Evidence for register variation in Early Modern English negatives. Language Variation and Change 17: 257–280.

Looking high and low for NegP in early English Eric Haeberli 1. Introduction The emergence of a secondary negator in the early history of English raises the question as to how such an element is integrated into the syntactic structure. In her analysis, van Kemenade proposes that from the earliest attested stages until the Early Modern period, secondary negation is found in two structural positions, a low one at the edge of the VP and higher one in which the secondary negator can potentially precede subjects. With respect to the high secondary negator, it is argued furthermore that its occurrence depends on verb movement to C and that its distribution with respect to subjects is determined by aspects of information structure. This paper will explore various issues that arise from van Kemenade’s discussion and a somewhat di¤erent analysis is proposed. In section 1, I will examine the proposal that there is both high and low negation in Old English and I will conclude that such a distinction indeed seems to be warranted for Old English. Pursuing one of the main themes of this volume in section 2, I will then consider whether phenomena related to secondary negation in early English provide evidence for postulating a specific structural projection for negative elements in the clause structure (NegP). Section 3 explores the factors that determine the placement of secondary negation (high vs. low). In section 4, an analysis of the restriction of high negation to V-to-C contexts is proposed. Finally, in section 5, diachronic consequences of the proposals made in the earlier sections are explored. 2. Secondary negation in Old English: high and/or low? Old English (OE) sees the emergence of a secondary negator (na) that is occasionally used in addition to the standard preverbal clitic negator ne. Van Kemenade proposes that this new secondary negator can occur in two structural positions, a high one close to the left periphery and a low one close to the VP. Evidence for the high negation position comes from clauses that have commonly been analyzed as involving verb movement to

116

Eric Haeberli

the C position, i.e. clauses with the negated finite verb in initial position, clauses with initial þa/þonne (‘then’) or questions (cf. e.g. Pintzuk 1999). In such clauses, nominal subjects frequently occur to the right of na whereas pronominal subjects systematically occur to its left (cf. van Kemenade’s Table 3). The former observation concerning nominal subjects suggests that na can occur above TP. Evidence for a low placement of secondary negation comes from clauses not involving V-to-C movement (cf. van Kemenade’s Tables 4 and 5). In such clauses, nominal subjects systematically precede na and the inverted order is very rare. This means that the secondary negator can also occur in a lower position, i.e. below TP. In this section, I will examine the evidence for postulating two positions for secondary negation in Old English. 2.1. High secondary negation? As already noted by Einenkel (1913: 204), the secondary negator na can both precede and follow nominal subjects in Old English clauses featuring subject-verb inversion whereas it always precedes pronominal subjects. Quantitative confirmation of this observation is provided by van Kemenade in Table 3. In negative-initial root clauses with na, nominal subjects follow na in nearly 60% of the cases, but subject pronouns never do so. In her counts, van Kemenade excludes clauses where na is immediately followed by another adverb or a quantifier and clauses in which na seems to be part of a contrastive coordination involving another constituent introduced by ac (‘but’). However, these restrictions do not substantially alter the picture concerning the distribution of nominal subjects and na. If we simply count all cases of what in The York-Toronto-Helsinki Parsed Corpus of Old English (YCOE; Taylor, Warner, Pintzuk & Beths 2003)1 is coded as a negative adverb, there is a slightly lower rate with the secondary negator occurring to the left of nominal subjects, but the frequency does not fall much below 50%. Thus, there is clearly a substantial number of cases where the subject occurs in a structural position below that occupied by the secondary negator. If we assume that the nominal subject occurs in the inflectional domain, say SpecTP, na therefore seems to be able to occur above TP. However, such a conclusion might be a bit premature. As is well known, nominal subjects of unaccusative verbs can occur in a low position in Old 1. The OE data discussed in this paper are all taken from this corpus. Cited examples follow the referencing conventions of that corpus.

Looking high and low for NegP in early English

117

English (cf. e.g. van Kemenade 1997). This is shown for example by the fact that such subjects can occur to the right of the non-finite main verb in a construction with an auxiliary. If we interpret this observation as meaning that subjects of unaccusative verbs can remain in their underlying complement position of VP, the occurrence of such a subject to the right of na could simply mean that the secondary negator occupies a VPperipheral position and an analysis in terms of a secondary negator above TP would not be required. In order to avoid this possibility, we should restrict the empirical evidence used for the analysis of the placement of secondary negation to clauses with verbs whose subject is not an underlying complement. If na regularly occurs before a subject with such verbs, the conclusion that there is a high position for secondary negation seems to be on safer grounds than if it were reached on the basis of all kinds of clauses. As, among verbs taking one argument, the distinction between unaccusative and unergative verbs is not always a straightforward one to make, I propose to consider only verbs which have at least two arguments and whose nominative argument is clearly the external argument.2 A restrictive search along these lines supports van Kemenade’s findings to some extent. 13 clauses have the order ‘ne V–na–subject’ whereas 14 clauses have the order ‘ne V–subject–na’. However, among the 13 clauses with ‘na-subject’ order, two could be argued to involve constituent negation (one of the type ‘not all N’, one with the conjunction ac (‘but’) providing an alternative to the subject).3 A further six clauses contain an object that has been moved to the left of the subject and na (four pronominal objects between the finite verb and na, and two topicalized objects). This observation may be relevant for the structural analysis of the subject given the claim made 2. Impersonal verbs and passives of ditranstive verbs are therefore excluded as the nominative argument may occur in an object position in these cases (cf. Allen 1995). Furthermore, I also exclude clauses with a finite modal verb as modals can be argued to lack an external argument (cf. van Kemenade 1997: 336). Clauses in which the thematic roles are assigned by an adjective are not considered, either. In order to avoid interference by the heaviness of clausal arguments, I will focus on clauses with a nominal nominative argument and at least one additional nominal or pronominal argument. Finally, I will include all clauses with a negated finite verb in a position before a nominal subject. The negated verb may therefore not be in absolute initial position. 3. But see below for some remarks on why it may be problematic to discard examples for the purposes of analyzing sentential negation simply because they show a constituent negation interpretation.

118

Eric Haeberli

by Alexiadou & Anagnostopoulou (2001) that the external argument of a transitive verb can remain VP-internal in some languages if the object has vacated the VP. Whether this is the case for Old English or not would have to be examined in more detail. But even if we were to discard the two potential instances of constituent negation and the six cases with a fronted object, we are left with five clear cases of transitive verbs where both the subject and the object follow na. Two illustrations are given in (1). (1) a.

Ne byrð na se cniht butan intingan his swurd. not carries not the knight without cause his sword ‘The knight does not carry his sword without a reason.’ (colsigewZ, +ALet_4_[SigeweardZ]:1212.596)

b.

witodlice, Petrus, ne beþearf na seo sawl swa geradre wege certainly, Peter, not needs not the soul so straight way & færinge. and journey ‘Certainly, Peter, the soul does not need such a straight way and journey.’ (cogregdC, GDPref_and_4_[C]:36.314.24.4715)

It may be interesting to point out that three of the five clear cases of the type shown in (1) are from Gregory’s Dialogues (C) and the remaining two from Ælfric’s letters. Thus, once we focus on examples for which an analysis of the placement of the subject in terms of a specifier in the inflectional domain seems inevitable, the evidence for a high position for na is by no means abundant and restricted to two authors. However, given the fact that the use of secondary negation is still a clear minority option in Old English and that the number of examples with the order ‘subject-na’ in the same context is small as well, I will conclude, in line with van Kemenade, that ‘na-subject’ order does not only arise due to the occurrence of the subject in a low position in Old English but indeed also due to placement of na in a high one. However, the evidence for this option does not seem to be very robust. 2.2. Low secondary negation? As van Kemenade’s Tables 4 and 5 show, the occurrence of na to the left of a nominal subject is extremely rare in clauses not involving V-to-C movement, regardless of verb type. In main clauses, this observation would not be immediately incompatible with a high position for secondary negation. Following Pintzuk (1999), many authors assume that, when the

Looking high and low for NegP in early English

119

finite verb does not move to C in main clauses, it moves to the inflectional head below CP, i.e. to F in the structure presented by van Kemenade in (2).4 The order ‘Subject–V–na’ could therefore be derived through movement of the subject to the left periphery of the clause and verb movement to F. Given that NegP with na in its specifier is situated below FP in (2), the rarity of ‘na-subject’ orders in main clauses is expected. However, to assume that na is consistently situated in a high position in Old English would be problematic. Van Kemenade mentions the occurrence of object pronouns to the left of secondary negation in main clauses as evidence for a low negation position. The postulation of a position for object pronouns between FP and NegP in van Kemenade’s structure in (2) might indeed not be desirable and a lower NegP seems to be preferable. A possibly even clearer case for a low position of secondary negation can be made on the basis of the syntax of subordinate clauses. As is well-known, Old English exhibits a clause type asymmetry between main and subordinate clauses in the sense that the finite verb tends to occur towards the beginning of the clause in main clauses but towards the end of the clause in subordinate clauses. Furthermore, subject-verb inversion phenomena as found in main clauses are much more restricted in subordinate clauses. These contrasts suggest that there is an asymmetry concerning the distribution of the finite main verb in the two types of clauses. In Haeberli (2005), it is therefore proposed that verbs can occur in C or F in main clauses but that the verb does not move beyond T in subordinate clauses. If this analysis is correct, secondary negation would be expected to always occur preverbally in Old English subordinate clauses if there were only the high negation position shown in van Kemenade’s (2). However, this expectation is not borne out. Among a bit more than 300 subordinate clauses in the YCOE which contain an overt subject, a finite main verb or auxiliary and the secondary negator na, nearly half have the secondary negator after the verb/auxiliary (cf. van Kemenade’s examples (30) and (31) for illustrations). In subordinate clauses with nominal subjects, the rate is even higher and the postverbal position of the secondary negator is the majority option.5 These facts are incompatible with a unique high 4. I will adopt van Kemenade’s label FP in my discussion here. But cf. Haeberli (2000) for some evidence linking this position to subject-verb agreement (AgrP). 5. If we assume that na cannot be extraposed, these data suggest that at least about half of all subordinate clauses with a secondary negator have a headinitial structure. This is in clear contrast to data obtained with other diagnostic

120

Eric Haeberli

position above TP for na and the postulation of a lower one between TP and VP is necessary. Interestingly, there is one example of a subordinate clause in the YCOE where both the high and the low na seem to be realized. This is shown in (2). (2)

for þan þe se deað hit na ne elcað na. because the dead it not not delay not ‘because the dead do not delay it.’ (coalcuin, Alc_[Warn_35]:406.301)

If we assume that the finite verb in subordinate clauses is in T, the first na could be analyzed as occupying the secondary negation position above TP whereas the second na would occur in a position below TP. A similar example involving the somewhat less frequent secondary negator nauht is given in (3). (3)

forðæm ic hit no self nauht ne ondræde therefore I it not self not not feared ‘Therefore I did not fear it myself.’ (coboeth, Bo:20.47.5.846)

Having the verb in final position, (3) can be analyzed as involving headfinal structure. The subject and object pronouns would then occupy positions in CP and/or in FP within the structure (2) given by van Kemenade. Self could be argued to be associated with the lower subject position in TP. No thus occupies the secondary negation position above TP, whereas nauht occupies the one below TP. (2) and (3) are isolated examples within the YCOE, and we may have to treat them with some caution. Nevertheless, all the observations above taken together seem to confirm van Kemenade’s proposal that na can occur in more than one position in Old English. In the next section, I will explore the question as to how na can be represented in the syntactic structure and, more specifically, whether it is hosted by a NegP.

elements in Pintzuk & Haeberli (2008). There head-initial structure reaches frequencies between 1% and 25%, depending on the diagnostic element used. This contrast suggests that secondary negation in OE is favoured in head-initial structures (cf. also Einenkel 1913:211 for a similar point), an observation which may at least partly explain why the rise of the secondary negator is situated in the Middle English period, i.e. in a period when head-initial structure becomes the predominant one.

Looking high and low for NegP in early English

121

3. NegP or no NegP? One of the central issues in the recent literature on the rise of secondary negation in the history of English is its status in the clause structure. As observed by van Kemenade, many authors place the secondary negator in the specfier of a NegP. An alternative but less commonly held view is that the secondary negator has the status of an adverbial. Van Kemenade’s position concerning this issue remains somewhat uncertain. For high na, two options are presented. In the structures shown in (2) and in (39), na is located in a NegP whereas in (20), na occupies the specifier of a PrtP, a position it might share with other elements like þa and þonne. As for low na, it is described as ‘‘an adverb on the left of the VP’’ and put under a projection XP above VP in the related structure in (35a). In this subsection, I will examine whether the data related to the secondary negator na can be used to make a case for NegP in the clause structure of Old English. As van Kemenade observes in her introductory paragraph, ‘‘[t]he concept of NegP provides an appealing and insightful framework for understanding those stages of the history of English where clause negation features two negative markers: ne and some form of secondary negation, and for the shifting distribution of these two negative markers over the history of English’’. According to this observation, the emergence of the secondary negator na in Old English would best be analyzed in terms of a NegP. Although the presence of a head (ne) and a non-head (na) related to negation could indeed be easily captured by saying that the head occupies the head position of a NegP and the nonhead occupies its specifier, this would be by no means the only possibility. Ne could be analyzed as the spell-out of a negative feature on some other functional head, whereas na could be analyzed as some kind of an adverb. For many empirical aspects of the syntax of negation and its development, the two options might perform equally well. Very often, the choice is therefore simply a theoretical one. Clause structures inspired by the Minimalist Program might do away with NegP in the same way that AgrP has been discarded, whereas cartographic approaches postulate very rich clause structures in which NegP would no doubt have its place. The task is then to identify phenomena for which the choice of analysis actually does matter. For example for present-day English, it can be argued that, if not simply had the status of an adverb, its presence should not trigger do-support because other negative adverbs like never do not require do-support, either. But what about earlier stages of English? Are there phenomena that would also allow us to identify the presence of a NegP?

122

Eric Haeberli

Haeberli & Ingham (2007) indeed argue that there is empirical evidence for the presence of a NegP in the clause structure of Early Middle English (EME) and that this evidence comes from the placement of the secondary negator. The basic idea is that, if the secondary negator had the status of an adverb, its distributional properties should correspond to those of adverbs. On the other hand, if the secondary negator had distinctive distributional properties, it would be plausible to attribute to it a specific position in the clause structure that we could label SpecNegP. The evidence that Haeberli & Ingham identify as relevant for the postulation of NegP in Early Middle English concerns the interaction of objects with adverbs and secondary negation. In clauses with a finite verb and a postverbal secondary negation, nominal objects systematically follow the secondary negator. In the Early Middle English data considered by Haeberli & Ingham, the order ‘V–not–O(DP)’ can be found 45 times in main and subordinate clauses, whereas there is not a single instance where the object would precede the secondary negator in this context (2007:14). A similar but slightly less robust observation can be made for clauses with a finite auxiliary. The order ‘Aux–not–O(DP)–V’ is found 8 times, but the order ‘Aux–O(DP)–not–V’ does not occur. The systematic absence of the ‘O(DP)–not’ order in these contexts suggests that the secondary negator is placed too high for the object to move across it. The situation is considerably di¤erent with adverbs. In 369 clauses with a finite verb and a postverbal adverb, nominal objects precede the adverb in 113 (30.6%) cases and follow it 256 times (69.4%) (2007:12). Pre-adverb position is even more common in auxiliated clauses with objects preceding the non-finite main verb. ‘Aux–O(DP)–adverb–V’ occurs 25 times (59.5%) whereas the order ‘Aux–adverb–O(DP)–V’ is found 17 times (40.5%) (2007: 13). The range of adverb types that follow the object is fairly varied. Given this sharp contrast between secondary negation and adverbs, Haeberli & Ingham conclude that the secondary negator is not simply an adverb in Early Middle English but that it occupies a position specifically related to negation (SpecNegP) which a nominal object cannot move across. The question that we may raise then is whether Old English na had a similarly distinctive status with respect to nominal objects as the Early Middle English secondary negator. The answer to this question seems to be negative. The relevant figures are given in Table 1. Although the evidence is not overwhelming, the figures in Table 1 suggest that nominal objects do not systematically occur to the right of the secondary negator in Old English. The 5 out of 53 examples found in Old English contrast with the 0 out of 45 in Early Middle English. As in other

Looking high and low for NegP in early English

123

Table 1. Secondary negation, tensed verb/auxiliary and nominal object in YCOE Root clauses

Subordinate clauses

S V O Neg

3

0

S V Neg O

19

13

S Aux O Neg V

2

0

S Aux Neg O V

11

5

Total O Neg

Total Neg O

3 32 2 16 5 (9.4%)

48 (90.6%)

cases involving na, questions as to whether na truly functions as an independent secondary negator rather than as a constituent negator may arise in these examples. However, a note of caution seems to be necessary here. Even if we get a constituent negation interpretation, it is not certain whether syntactically we do not have a case of sentential negation. In Old English, this is suggested by the presence of the preverbal negative clitic ne in these examples. Similarly, in a present-day English sentence like John didn’t buy newspapers but books, the presence of but suggests a constituent negation interpretation for the object DP in the first part of the sentence, but the occurrence of do-support nevertheless seems to make an analysis in terms of sentential negation necessary at a syntactic level. Thus, it might be delicate to simply eliminate clauses like (4a) which might have a constituent negation interpretation from our consideration. However, even if we did so, we are still left with two clauses with ‘O-Neg’ order for which a constituent negation analysis seems unlikely (4b, c).6 6. Both clauses are followed by a clause introduced by ac, but what follows does not seem to correspond simply to a constituent that could be negated by na in (4b) and (4c). The continuations of (4b) and (4c) are given below: (i) ac he nolde his cwyde awendan ðurh ðone þe he But he not-wanted his speech change through that who he bead his leorningcnihtum: summoned his disciple ær his ðrowunge þus cweðende before his su¤ering thus speaking: ‘But he did not want to change his speech through him who he summoned as his disciple before his su¤ering, thus speaking:’ (cocathom2, +ACHom_II,_8:68.45.1377)

124

Eric Haeberli

(4) a.

ac þæt ne dereð elles þam na þe swyðor þe þa but that not hurts otherwise that not the more who the ðenunga underfoð [. . .] service receives ‘But that does not hurt that one more who receives the service.’ (cowulf, WHom_8c:49.612)

b.

He ne andwyrde ðam wife æt fruman na for modignysse. he not answered the woman at first not for pride ‘At first, he did not answer the woman out of pride.’ (cocathom2, +ACHom_II,_8:68.45.1376)

c. Þa ða awyrgdan gastas, þe ðær stodon . . . ne mihton then the cursed souls, who there stood . . . not could ða oðre men na geseon, the other men not see ‘Then the cursed souls who stood there . . . could not see the other men, . . .’ (cogregdC, GDPref_and_4_[C]:40.326.3.4904) That nominal objects can precede secondary negation is also confirmed by clauses where both the secondary negator and the full DP object precede the finite verb, i.e. in clauses that would traditionally be analyzed as involving head-final structure. Among 28 main and subordinate clauses with na and a nominal object preceding the finite verb, 8 (28.5%) exhibit the order ‘O-Neg’. Most of these cases do not seem to involve constituent negation. Two illustrations are given in (5). (5) a.

ðæt he ðæt good na ne dyde ðær he hit for ðæm ege that he that good not not did where he it for the fear dorste forlætan. dare abandon ‘that he did not do that good deed when he dared to abandon it for fear.’ (cocura, CP:37.265.10.1724)

(ii) ac hi hwæðre ongeaton heora andweardnesse in ondetnessum. . . but they nevertheless recognized their presence in confessions. . . ‘but they nevertheless recognized their presence in confessions. . .’ (cogregdC, GDPref_and_4_[C]:40.326.3.4905)

Looking high and low for NegP in early English

b.

125

he þæt no mid weorce ne gefremme; he that not with work not accomplish ‘he shall not accomplish this with work;’ (cobenrul, BenR:2.11.17.178)

Finally, in clauses involving V-to-C movement with a finite verb or auxiliary, nominal objects also occur to the left of na in six cases. However, all of these examples raise issues of constituent negation, as the two cases in (6) show (lang in 6a, to medsceatte in 6b). (6) a.

b.

Ne forlet ure Drihten þysne middangeard na leng buton not leaves our Lord this world not longer without lareowum þonne twa hund wintra, teachers than two hundred winters ‘Our Lord does not leave this world without teachers more than 200 years,’ (coblick, HomS_21_[BlHom_6]:71.103.897) Ne not ac but

onfeng he ðæt na to medsceatte, accepted he that not as payment, forðon þe he wæs þyr me gehalsod. because he was through me entreated

‘He did not accept that as payment, but because he was entreated in my name.’ (comart3, Mart_5_[Kotzor]:Se27, A.16.1877) In conclusion, the evidence for ‘O-Neg’ order with nominal objects in Old English is not abundant, but it nevertheless suggests that this order is not entirely excluded. Given that various types of adverbs can also occur to the left and to the right of nominal objects in Old English, the object data do not allow us to argue for the presence of a dedicated position for negation in the clause structure of Old English in the same way that this is possible for Early Middle English. Instead, the secondary negator na may simply have the status of a regular adverb. Phenomena related to object pronouns also suggest that the syntax of secondary negation in Old English is not regulated as strictly as in Early Middle English. In clauses with a finite main verb followed by a pronominal object and a secondary negator, the pronominal object always precedes the secondary negator in Early Middle English (0 instances of the opposite order out of 38; Haeberli & Ingham 2007: 16). In Old English,

126

Eric Haeberli

however, we can find orders of the type ‘subject–finite verb–na–object pronoun’. Out of 12 cases with a postverbal object pronoun and a postverbal secondary negator, 4 have the object in a position following negation. This is shown in (7): (7) a.

b.

ac he ne sette na hi on his setle but he not put not them on his see ‘but he did not put them on his see’ (cocathom1, +ACHom_I,_37:497.12.7330) þæt he ne geceas na him wif to meder. that he not chose not him woman to mother. ‘that he did not choose a woman as a mother for himself.’ (cocathom2, +ACHom_II,_1:4.37.28)

The flexibility in the distribution of object pronouns in Old English contrasts with the rigidity in this domain in Early Middle English and thus provides further support for the proposal that the Old English secondary negator na is not hosted by a NegP that has a rigidly fixed position in the clause structure. Haeberli & Ingham (2007) use object data to argue for a low NegP in Early Middle English. The observations made above may therefore simply mean that Old English does not have a low NegP. But given the conclusion reached in the previous subsection that there is also a high secondary negation position, there still might be the possibility that NegP is encoded in a high structural position, as suggested by van Kemenade’s tree in (2). Following the same kind of reasoning as above, we would therefore want to find distributional evidence that clearly distinguishes secondary negation from adverbs. However, the observations made by van Kemande and those made in Haeberli (2000) suggest that no such contrast can be identified. In Haeberli (2000), it is shown that adverbs occur to the left of nominal subjects in V-to-C contexts. This holds in particular also for transitive verbs with two nominal arguments as two examples in Haeberli (2000:117) involving the adverbs swa þeah (‘however, nevertheless’) and syððan (‘afterwards’) show. Furthermore, van Kemenade suggests that þa/þonne (‘then’) also behave like na in being able to occur in a position to the left of nominal subjects and that the order of the two elements is not fixed when they co-occur. It therefore seems di‰cult to identify a structural position in the higher inflectional domain that is specifically dedicated to negation. Once again, it would therefore be su‰cient to treat na as an adverb.

Looking high and low for NegP in early English

127

In conclusion, I have not been able to find empirical evidence for the occurrence of NegP in Old English based on the behaviour of the emerging secondary negator. However, I have to leave it open here whether other evidence could be used in favour of postulating a NegP in the clause structure of Old English or whether cross-linguistic and theoretical considerations force us to postulate such a projection. I will also not be able to pursue the follow-up question based on the conclusion reached here, namely how adverbs (and, hence, na if it is to be treated as an adverb) are best represented in the syntactic structure of Old English. Instead, I will turn to an additional issue that van Kemenade’s ‘‘high-low’’ analysis of Old English secondary negation raises.

4. Na: When is it high, when is it low? Having established that na can occur in a high and a low position in Old English but that these positions do not correspond to a NegP in a clear way, we may wonder now what determines the use of the low and the high negation position in Old English. Are there factors that favour the use of one as opposed to the other? Van Kemande provides a fairly precise proposal in this respect: ‘‘The first use [low na] is attested in main clauses without inversion of subject and finite verb, and is almost categorical in subclauses. The second use [high na] is restricted to root clauses with V to C movement’’ (section 1). Two issues based on this statement seem to be worth examining further: (i) Is high na available in subordinate clauses (cf. the hedge ‘‘almost categorical’’ use of low na); (ii) Is low na not available in V-to-C contexts (cf. ‘‘main clauses without inversion’’ for low na)? Let us start by considering na in subordinate clauses. I have identified 8 instances of na occurring to the left of a nominal subject in a subordinate clause in the YCOE, so these would be potential candidates for a high placement.7 However, none of them conclusively involves a high negator. Consider for example van Kemenade’s (33b), which she considers as a potential illustration of high negation in a subordinate clause:

7. Van Kemenade mentions 7 cases in her Table 4. Her selection criteria for inclusion may have led her to exclude one of the examples I have found, but it is not entirely clear to me which one could have been eliminated. I therefore consider them all here.

128

Eric Haeberli

(8) Forðæmðe na se ðorn ðære gitsunga ne wyrð because that not the thorn of greed not becomes forsearod on ðæm helme, withered in the crown ‘Because the thorn of greed does not wither in the stem’ (cocura, CP:45.341.9.2292) As (8) is a passive clause, the only argument in the clause, se ðorn ðære gitsunga, may occur in an underlying object position low in the structure. Such an analysis would be possible if we assume that (8) involves a headfinal structure with Verb Raising of the participle and extraposition of the PP. The result would be that a low na at the edge of the VP could precede the VP-internal argument.8 Similar analyses would be conceivable for five other examples: Two involve copula be and three contain verbs that can be considered as unaccusative (befeallan ‘to fall’, modigan ‘to become proud’, libban ‘to live’). This leaves us with two examples. One of them involves a one-argument verb that would generally be considered as unergative rather than as unaccusative (wepan ‘to weep’). And the second one, van Kemenade’s example (31a), involves a transitive verb:9 (9) þæt hy na sunne bescine that it not sun beshines ‘that the sun does not shine on it’ (coherbar, Lch_I_[Herb]:182.1.2618) The internal argument of the verb has moved to the left, and we therefore find a context that, as discussed in section 1.1, would allow a VPinternal subject at least from a cross-linguistic point of view (cf. Alexiadou & Anagnastopoulou 2001). If that option is available in Old English, (9) could also be analyzed in terms of low negation. By analogy, the subject of an unergative verb like wepan would also be expected to be able to stay within the VP and the same conclusion with respect to secondary 8. Within a framework that only uses head-initial projections, it would be conceivable that a low negator and the VP-internal argument undergo remnant movement to the left of the auxiliary once other material has moved out of the VP. 9. Interestingly, the preverbal negative clitic ne is absent in this example, a phenomenon that is rare in OE. It is not clear, however, whether this observation has any bearing on the issues discussed here.

Looking high and low for NegP in early English

129

negation would hold for that case. However, I will have to leave it open here whether there is independent evidence supporting the hypothesis that subjects of unergative and transitive verbs can be licensed in a VP-internal position in Old English. We are therefore left with at most two and possibly no examples for which a high secondary negator has to be postulated in Old English subordinate clauses. This will not allow us to entirely remove van Kemande’s hedge (‘‘almost categorical’’ use of low negation in subordinate clauses). But no further examples from main clauses could be added to this total, because in main clauses of the type ‘S-V. . .’, the occurrence of a secondary negator in pre-subject position could always be related to fronting to CP, a process that is clearly attested in early English (cf. e.g. Haeberli & Ingham 2007: 10). The possibility that high negation is restricted to cases of subject-verb inversion therefore seems to be worth exploring within a theoretical account of the placement of secondary negation in Old English. If we assume furthermore that subject-verb inversion with negative verbs involves V-to-C movement, we can be more specific and suggest, as van Kemenade does, that the presence of high negation is dependent on V-movement to C. Before considering the theoretical implications of this conclusion, let us turn to the second issue that van Kemenade’s proposals concerning the use of high and low negation raises. In the citation given above, the use of low negation is mentioned only in connection with main clauses without inversion and subordinate clauses. But what about clauses with inversion, i.e. V-to-C contexts? As the discussion in section 2.1 has shown, it is rather di‰cult to find conclusive evidence for a high secondary negation in V-to-C contexts in Old English. The corollary of this is that a low na could account for a large majority of examples, i.e. all clauses with na in a position following the subject (388 out of the 446 cases included in Table 3), and also all clauses involving an unaccusative verb and a pre-subject secondary negator. At first sight, it seems to be di‰cult to rule out low secondary negation in V-to-C contexts from a theoretical point of view. If we assume that Old English has systematic V-to-T movement, a derivation with V-to-C and without V-to-C movement should look identical at least up to TP. The mere occurrence of a further V-movement step beyond TP should not be able to interfere with low placement of na in any way. So the minimal assumption from a theoretical point of view would be that low na is available in V-to-C contexts but that, optionally, na can be inserted in a higher position as well.

130

Eric Haeberli

The question we may raise then is whether there are empirical reasons to assume that secondary negation should be restricted to the area above TP in V-to-C contexts. Here, van Kemenade’s information structure (IS) considerations could be relevant. If a clear IS pattern emerged that could be maintained only under the assumption that there is a high secondary negator in V-to-C contexts, that assumption would receive considerable support despite its theoretically unexpected status. However, with van Kemande’s IS account in its current form, I am not entirely convinced that (a) a su‰ciently clear-cut IS pattern emerges and (b) that what emerges could not be dealt with through a combination of high and low negation in V-to-C contexts. Given the wide range of word order patterns found in Old English, it seems very attractive to turn to IS in order to discover potential factors that may influence the choice of one word order over another one. The line of investigation that van Kemenade and others propose in this direction therefore seems promising. However, I find the specific implementation of this research programme in van Kemenade’s paper somewhat unsatisfactory. What seems to be missing in particular are very precise definitions of the IS factors that are taken into account. The general distinction that van Kemenade makes seems to be a fairly straightforward and plausible one: Elements to the left of na, þa and þonne are discoursegiven whereas those to the right are discourse-new or focused. But specific applications of these notions then look a bit problematic. In connection with example (26) where the subject precedes na and should therefore be discourse-given according to van Kemenade’s proposal, it is argued that a demonstrative can be considered as discourse-given due to its being the antecedent of a postposed relative clause. However, the referent of this complex DP does not seem to have been mentioned in the earlier discourse, so it would not typically be characterized as discourse-given. In section 3.2.2 focusing on subjects that occur after na and should therefore be discourse-new or focused, it is suggested that, even though certain subjects are discourse-given, this property becomes irrelevant for their placement once an object pronoun is present. But no explanation is given as to why this should be the case. Furthermore, for some reason, discourse givenness also seems to be irrelevant for the placement of generic subjects and subjects referring to a unique entity (God, the holy Father, the holy Ghost, the soul). Some subjects of this type follow na although they have been mentioned earlier in the discourse. Finally, two proper names and a definite DP that have previously been mentioned are also considered as suitable in the low position under the assumption that no specific reference

Looking high and low for NegP in early English

131

to the discourse antecedent is made. It remains unexplained, however, what criteria can be used to make such a judgment or why it should be relevant as the occurrence in the previous discourse should be su‰cient to qualify something as discourse-given. Given that these evaluations of what counts as discourse-given or not are not justified in any way nor independently supported by any references to the literature on IS, they have the flavour of post hoc attempts to save a generalization. The IS literature provides other distinctions that could be worth exploring in connection with the di¤erent subject positions van Kemenade identifies. In their analysis of the placement of objects in Old English, Taylor & Pintzuk (2009) consider hearer status (new, anchored, accessible, given) in addition to discourse status (mention, no mention). They conclude that, when other factors are taken into account, only hearer status has a significant e¤ect on the distribution of the object but not discourse status. However, even if other IS categories were used for the analysis of the distribution of subjects and if these categories were applied with neutral coding independent of any theoretical expectations, the most likely outcome would probably be that the pattern is not entirely sharp, contrary to what van Kemenade tries to establish. Instead, we may simply have IS tendencies that influence the choice of one word order over another, but additional factors (e.g. length, type of verb etc.) may play a role as well. Thus, the generalization to be made may rather be that discourse-given material (or something of this type) tends to occur high in the structure and, hence, to the left of na whereas discourse-new material (or something of this type) tends to be low and to its right. Let us then return to our question whether V-to-C contexts necessarily require a high secondary negator for this generalization to be accounted for. For given subjects, the answer is clearly negative. If we assume that given subjects tend to move to the highest subject position, they would end up to the left of na regardless of whether it occurs above TP or below TP. As for non-given/focused subjects, there are two scenarios to consider. If certain subjects are licensed VP-internally, it could be argued that VPinternal placement is related to discourse-newness or focus, and this position would be to the right of na even with low na. The only scenario that could potentially be problematic for an analysis in terms of the availability of low na in V-to-C contexts is the one that involves subjects that have to leave the VP (i.e. at least subjects of transitive verbs with a nominal object, but possibly subjects of transitives more generally and unergatives). Such subjects might be expected to be discourse-new and nevertheless occur to the left of na.

132

Eric Haeberli

However, it is not certain that we really face a problem here. First, if the configuration ‘SU(new/focus)-na’ does not exist at all, we might argue that IS can also influence the choice of the insertion site of na and not only the placement of the subject. In other words, if there is a general IS tendency for new/focused elements to occur towards the right and we have a choice of inserting another item either to the left or to the right of the new/focused element, then insertion to the left might be chosen so that the new/focused element can occur in the rightmost position. The order ‘SU(new/focus)-na’ would therefore be avoided. An alternative conclusion may be that such a configuration is actually not impossible. One example that could be relevant in this context is given in (10): (10) Ne forseon ða gelæredan na ða ungelæredan, ne neglect the learned not the unlearned ‘The learned to not neglect the unlearned,’ (cowulf, WHom_10a:45.805) The definite subject ða gelæredan seems to be of the generic type here rather than related to a referent already mentioned in the discourse. Such subjects are generally claimed by van Kemenade to occur to the right of na, which is not the case here. Whatever the correct conclusion may be, it is su‰cient for our purposes to observe that IS considerations do not seem to be incompatible with the hypothesis that low na is also available in V-to-C contexts. The bottom line would then be that low na is present in all syntactic contexts in Old English whereas high na is restricted to V-to-C contexts. The question that arises then is why such a restriction on high secondary negation should hold, an issue we will pursue in the next section. 5. The restriction on high secondary negation To account for the restriction of high na to V-to-C movement contexts, van Kemenade proposes that the occurrence of na in a high position is related to an unvalued Neg feature on na that is valued by a negative operator in CP. However, for a full analysis, more would have to be said. First, the occurrence of a negative operator in CP does not guarantee Vto-C movement. Within the Minimalist framework on which the concept of feature valuation is based, it would indeed be conceivable that a negative operator could occur in CP without triggering V-to-C movement. Thus, something would have to be said on the connection between the

Looking high and low for NegP in early English

133

negative operator in CP and verb movement to C. Secondly, it is not immediately clear why high na should have to be licensed by an even higher operator. In non-V-to-C contexts, na is licensed in a low position, somewhere at the periphery of the VP. The minimal assumption for low na would be that it has the same feature content as high na. In other words, low na would be expected to have an unvalued Neg feature as well. Hence, there would have to be another negative element in the clause structure that can value the Neg feature of low na. The question that arises then is why this negative element could not also license (i.e. value the unvalued Neg feature of ) high na. At first sight, it would indeed seem that whatever licenses low na should also be able to license high na and that therefore high na should be just as suitable for non-V-to-C main clauses and for subordinate clauses as it is for clauses involving V-to-C movement. Here, I will try to address these issues and make the correlation between V-to-C and high na somewhat more precise by developing some proposals made by Haegeman & Lohndal (2010) for the analysis of Negative Concord languages. Recent analyses of Negative Concord within an Agree-based Minimalist framework suggest that Negative Concord as found in early English is the result of Multiple Agree, a process whereby one interpretable negative feature values (possibly multiple) uninterpretable negative features (cf. e.g. Zeijlstra 2004). Haegeman & Lohndal (2010) show that such an approach is problematic for the analysis of Negative Concord in West Flemish and they therefore propose an alternative analysis based on binary agree. The idea here is that Agree relations are always established between two elements. In other words, if there is a sequence of features, Agree operates in pairs. Agree is defined as follows (Haegeman & Lohndal 2010: 196) (11) Agree: a Agrees with b if a c-commands b, a and b both have a feature F and there is no g with the feature F such that a c-commands g and g c-commands b. Apart from the required structural configuration in terms of c-command, (11) introduces a locality condition on Agree. A further assumption Haegeman & Lohndal make is that Agree can involve two uninterpretable features and that, in that case, it is the lower one that is deleted and the higher one that survives. If we now consider negation in Negative Concord languages, the hypothesis of a null sentential negative operator with an interpretable negative feature and negative consituents with uninterpretable negative

134

Eric Haeberli

features is adopted (cf. also e.g. Zeijlstra 2004). The process of binary Agree in Negative Concord contexts is illustrated in (12), with strikethrough indicating that only one of the [uNeg] features remains after Agree (Haegeman & Lohndal 2010: 198). (12) a. b. c. d. e. f.

[C [uNEG]] [D [uNEG]] ! Agree [C [uNEG]] [D [uNEG]] Merge [B [uNEG]] ! Agree [B [uNEG]] [C [uNEG]] [D [uNEG]] Merge [A [iNEG]] [B [uNEG]] [C [uNEG]] [D [uNEG]] ! Agree [A [iNEG]] [B [uNEG]] [C [uNEG]] [D [uNEG]] [A [iNEG]] [B [uNEG]] [C [uNEG]] [D [uNEG]]

Bearing these assumptions in mind, let us now return to the question of how the restriction on na to V-to-C contexts could be accounted for. The basic intuition underlying my analysis will be that the presence of high secondary negation might be related to whatever property in the Cdomain is responsible for negative V1 (Neg V1) in Old English. As negation seems to be a trigger of V-to-C movement in Old English, it would be plausible to say that the presence of high na, which is dependent on V-to-C, has negation in C as its ultimate source. As discussed by Ingham (2005), within a Minimalist framework, Neg V1 in early English is best captured by postulating an uninterpretable negative feature in the CP-domain. In an Agree-based system, this negative feature would have to associated with an additional EPP-type or a‰xal feature that triggers the movement of the verb. Furthermore, I propose the following additional negative elements for an Old English negative clause with a secondary negator: (a) A null operator with an interpretable negative feature (cf. above). (b) An uninterpretable negative feature related to the preverbal clitic negator ne. Given the uncertain status of NegP in the clause structure of Old English (cf. section 2), I will simply assume here that this negative feature is situated on the T head. (c) An uninterpretable negative feature on na. An issue that I have left open in the above list is the structural position of the null sentential negative operator. Haegeman & Lohndal remain relatively vague in this respect, but in one representation of a West Flemish subordinate clause (their example (38)) they situate it to the immediate right of the subject. I will follow this proposal and assume that the default

Looking high and low for NegP in early English

135

position of the null negative operator is right below the lower inflectional subject position, i.e. right below TP in Old English. As for low na, I propose that it is between the null operator and VP. Given these hypotheses, an Old English non-V-to-C negative clause with a low secondary negator has a distribution of negative features as shown in (13). (13) T [uNEG] OP[iNeg] na [uNEG] When this structure is derived, two binary Agree relations are established, first the one between OP and na and then the one between T and OP. As a result, both uninterpretable features can be deleted and we obtain a grammatical result. But what would happen if the secondary negator occupied a high position above TP? The distribution of features in such a case is illustrated in (14): (14) na [uNEG] T [uNEG] OP[iNeg] Once T is merged, an Agree relation between T and OP is established and the uninterpretable NEG feature on T is deleted. Next, na is merged and it tries to establish an Agree relation to delete its [uNEG] feature. The deleted [uNEG] feature on T cannot Agree with [uNEG] on na any more. So what about [iNEG] on OP? Adopting a proposal made by Chomsky (2001), I will assume that [uNEG] on T, although having been marked for deletion, remains visible up to the next strong phase level (i.e. CP).10 Remaining visible, [uNEG] on T acts as an intervenor for the purposes of locality as defined in (11). [uNEG] on na can therefore not establish an Agree relation with OP. As no higher [uNEG] feature is available, either, [uNEG] on na cannot be deleted and we are left with an uninterpretable NEG feature by the time the derivation reaches LF. The structure in (14) is therefore ruled out. Let us now consider V-to-C contexts. As suggested earlier, we may assume that in such structures, C also carries a [uNEG] feature. In addition, I propose that the presence of a negative feature on C also allows the null negative operator to be merged in a high position, possibly within 10. This hypothesis can be motivated by PF considerations. If the [uNEG] feature on T were eliminated entirely as soon as it has entered an Agree relation, it would not be available for Spell Out at the phase level. Its presence is therefore required if we assume that [uNEG] on T corresponds to the preverbal negative clitic ne in OE.

136

Eric Haeberli

an extended CP-domain or right below CP.11 We therefore get the following configuration with a high secondary negator:12 (15) C [uNEG] OP[iNeg] na [uNEG] T [uNEG] This structure leads to a grammatical result. First, na and T enter an Agree relation and [uNEG] on na survives whereas [uNEG] on T is marked for deletion. Then, OP and na Agree and [uNEG] on na is deleted. Finally, an Agree relation between C and OP is established and [uNEG] on C can be deleted. All uninterpretable NEG features are thus deleted. Note, finally, that the configuration in (15) would also lead to a grammatical result if the secondary negator occurred in the low position, i.e. below T. In conclusion, our discussion has shown that an extension of Haegeman & Lohndal’s (2010) framework allows us to restrict the occurrence of a high secondary negator to V-to-C contexts if we make the hypothesis that it is only in such contexts that the CP domain hosts negative features in Old English. This analysis has an important diachronic implication, however. We would expect high negation to be possible only as long as there is an uninterpretable NEG feature on C, i.e. as long as there is Neg V1. The validity of this expectation will be examined in the next section. 6. Diachronic developments In the final part of her paper, van Kemenade explores the diachronic development of secondary negation. Her main conclusion is that, although 11. In fact, if we assume, as suggested in (15) (but cf. footnote 12), that the uninterpretable NEG feature on C co-occurs with one in T, insertion of the null operator in a high position would not just be allowed, but it would be necessary to obtain a grammatical derivation in a structure with [uNEG] on C. If the null operator were below T, we would get the sequence C [uNEG] T [uNEG] OP[iNeg], meaning that two uninterpretable features occur above the interpretable one, a configuration that we ruled out in (14). 12. An alternative to assuming that both C and T carry a [uNEG] feature in V-toC contexts would be to assume that only C carries one and that it is therefore the feature on C that is spelled out as ne in such cases. This option would be attractive if the presence of Neg V1 in early English could be related to the presence of the preverbal negative clitic, i.e. if Neg V1 could only be derived as long as there was such a preverbal negative clitic. However, Ingham (2006) concludes that the two phenomena are not linked in such a way as the decline of Neg V1 seems to have preceded the decline of ne.

Looking high and low for NegP in early English

137

its phonetic form changes (OE na being replaced by not in Middle English), the syntax of the secondary and soon-to-be primary negator remains fairly stable throughout the Middle English period. As in Old English, nominal subjects can occur to the left or to the right of the secondary negator in inverted main clauses in Middle English. Van Kemenade therefore proposes that Middle English still has the high/low contrast with respect to secondary negation in that inverted (V-to-C) main clauses have not in a position above TP whereas other clauses have not in a low position. Furthermore, it is assumed that the placement of a nominal subject still depends on its IS status in that the secondary negator continues separating discourse-given material from focused material. There is one major innovation that van Kemenade observes in the syntax of not in Middle English though, which is the emergence of its use as a head in the fifteenth century. The main evidence for this comes from ‘secondary negator–subject pronoun’ orders in inverted main clauses, an order which was not possible in Old English and Early Middle English. Let us now consider the implications of these observations for the conclusion reached in the previous section, namely that high not should be lost when Neg V1 is lost. First, we have to establish when Neg V1 is lost in the history of English. On the basis of Old and Middle English religious prose texts, Ingham (2006) concludes that ‘‘Neg V1 had gone out of use by the second half of the fourteenth century’’. This conclusion is to a large extent confirmed by Ingham’s (2005) study of verse material. Ingham observes a steady decline in Neg V1 from earlier thirteenth century manuscripts to late fourteenth century manuscripts: 81.8% Neg V1 in the earlier thirteenth century,13 61.5% in the later thirteenth century, 22.0% in the earlier fourteenth century, and 19.2% in the later fourteenth century. Although the frequency of Neg V1 at the end of the fourteenth century remains non-negligible in the verse texts, the possibility of an increased use of archaic features in this type of material means that Neg V1 has probably become at best a very marginal feature by then. We can now compare this finding to van Kemenade’s findings concerning the placement of not. Van Kemenade examines texts from the PennHelsinki Parsed Corpus of Middle English (PPCME2, Kroch & Taylor 2000), which is divided into four periods: m1 (1150–1250), m2 (1250– 13. The percentages are calculated against the total possible instances of inversion. Thus, the absence of Neg V1 means that no other constituent precedes the subject so that, if inversion had take place, the verb would occur in initial position.

138

Eric Haeberli

1350), m3 (1350–1420), and m4 (1420–1500). Period m1 (Early Middle English) can be dealt with straightforwardly. The word order ‘not-subject’ can still be found in this period. As Haeberli & Ingham (2007: 20) observe, many cases of this type involve verbs that can be argued to lack an external argument and they might therefore have a subject in a low position and, hence, not in a low position as well. However, there are some examples that cannot be dealt with in this way, which suggests that not can at least occasionally occur in a high position in Early Middle English inverted main clauses. According to the proposals made in the previous section, we would therefore expect Neg V1 to be possible as well. This expectation is borne out. Ingham’s (2005, 2006) studies show that Neg V1 is still frequently attested in Early Middle English. The end of PPCME2 period m2 corresponds to the period when, according to Ingham, Neg V1 is on its decline. However, if we consider a text from the end of the PPCME2 period m2, the Earliest English Prose Psalter (circa 1350), we notice that Neg V1 still occurs relatively frequently in this text. Thus, the occurrence of ‘not-subject’ orders in period m2 is in line with the hypothesis formulated in section 4 that high negation is related to Neg V1. From PPCME2 period m3 onwards, we would not expect Neg V1 to remain productive any more according to Ingham’s observations. This is indeed to a large extent borne out by the PPCME2 data. There are generally only isolated examples of negative clauses with the finite verb/ auxiliary in initial position in the PPCME2 periods m3 and m4. Nevertheless, the order ‘not-subject’ does not disappear entirely. Some cases might again be amenable to an analysis in terms of a low subject position and low negation, but ‘not-subject’ order can sometimes still be found with transitive verbs and two nominal arguments, i.e. cases where the subject is unlikely to occur within the VP. The question that arises therefore is whether the data from periods m3 and m4 are incompatible with the proposed link between high negation and Neg V1. I would like to argue that the answer is negative. As van Kemenade observes, not develops the option of being used as a head towards the end of the Middle English period. This can be clearly seen in period m4 where not can occasionally precede a pronominal subject in inverted clauses (cf. van Kemenade’s Table 9). A head analysis would therefore be equally possible for the few ‘not-DP subject’ cases reported in van Kemenade’s Table 6 for period m4. For not as a head, two analyses would be possible. Either not is merged as the head of a NegP, or alternatively not is merged as an XP and then cliticizes on the closest higher head.

Looking high and low for NegP in early English

139

With both scenarios, not could be merged below TP and then attached to the verb as it moves up the inflectional domain. A high negation position would therefore not be necessary any more to account for ‘not-subject’ orders in this period. What remains to be examined now is PPCME2 period m3. For this period, no examples can be found in which not precedes a subject pronoun.14 Do we therefore have to conclude that not as a head is not available yet and that therefore all orders of the type ‘not-subject’ have to be analyzed as involving a high negator? There is some evidence suggesting that such a conclusion is not necessary. Another type of subject apart from pronouns that seems to occur systematically to the left of secondary negation in Old English and Early Middle English are demonstratives. This is in line with van Kemenade’s IS proposal according to which discoursegiven elements tend to precede secondary negation. What is interesting with respect to PPCME2 period m3, however, is that we can find several instances of demonstratives to the right of not. The relevant examples are given in (16). (16) a.

Is nat this a cursed vice? ‘Isn’t this a sinful vice?’

(CMCTPARS, 306.C1.734)

b. nys nat that a myry thyng and a joyful? ‘Isn’t that a merry and joyful thing?’ (CMBOETH, 429.C2.45) c.

and ellys hadde not þis kyng trowyd; ‘and otherwise this king would not have had faith’ (CMWYCSER, 306.1410)

(17) shows another clearly discourse-given element to the right of not: (17) when þe net was ful of many grete fyschys, was not þe net broken ‘When the net was full with many big fish, the net was not broken.’ (CMWYCSER, 242.347)

14. The only potential example is the following one found in Chaucer: (i)

And sire, by youre leve, that am nat I

‘And sir, if you please, that is not me.’ (CMCTMELI, 221.C1.148) However, the subject could be argued to be focused here, so the example does not seem to be relevant for our purposes.

140

Eric Haeberli

Given what seem to be unexpected word orders compared to the Old English and Early Middle English data, examples like (16) and (17) could be analyzed as first clear signs of the emergence of not as a head. Other examples with not in a position preceding the subject could then of course also be argued to involve not as a head cliticized to the verb rather than merger of not in a high position. In fact, the approach pursued here may even provide an account of why not develops its head status when it does and how this use spreads. Until around 1350, ‘not-subject’ orders in inverted main clauses can be derived through a high negator because V-to-C is triggered by negation in C (cf. example 15). In the course of the fourteenth century, Neg V1 becomes more and more marginal, and language learners start postulating a grammar without Neg V1. However, ‘not-subject’ orders may still occur in their input, and they accommodate such orders by attributing optional head status to not. Initially, not may simply be used as a head with nominal subjects to maintain the IS tendencies that applied earlier, i.e. in particular to position focused elements to the right of negation. But over time, its use is generalized to other contexts such as demonstratives in PPCME2 period m3 and finally pronouns in period m4. Given the above observations, the correlation between high negation and negative V-to-C proposed in the previous section can be maintained as periods m3 and m4 can be analyzed exclusively in terms of a low negation. One additional issue remains to be addressed, however. Van Kemenade argues that, if negation were able to occur in a low position in inverted main clauses in Middle English, it should allow object shift, i.e. the occurrence of an object pronoun to its left. According to van Kemenade, this expectation is not borne out. However, the evidence provided in support of this conclusion is problematic as it is based on the figures in her Table 7. The problem with Table 7 is that, according to my own searches in the PPCME, it seems to include clauses with finite auxiliaries although it is well known that such clauses do not allow object shift. The surface word order is therefore generally ‘Aux-subject-not-V-object pronoun’ and such examples are inconclusive with respect to high or low placement of negation at least from the point of view of the placement of the object pronouns. Inverted main clauses with a finite main verb rather than a finite auxilary are relatively rare, but in the data I considered we nevertheless get ‘object pronoun–not’ orders in all Middle English periods in the PPCME2. For period m3, there are two examples with the order ‘object pronoun-not’ (cf. (18a) for an illustration) and one with the oppo-

Looking high and low for NegP in early English

141

site order, whereas ‘object pronoun-not’ occurs once in period m4 and the opposite order twice (cf. (18b) for an example with object shift).15 (18) a.

My doghtyr, why schryues þou þe not of þat synne? my daughter, why confess you yourself not of that sin ‘My daughter why don’t you confess that sin? (CMMIRK, 95.2577)

b.

Why schewyd me not yowr lettyr be-forn? ‘Why did you not show me your letter before?’ (CMKEMPE, 137.3207)

Given the low total numbers, no quantitative conclusions can be drawn from the m3 and m4 data. But what is crucial for our purposes is that object shift is not ruled out entirely in inverted main clauses, so there is no reason to assume that negation cannot be low in such a context. The fact that object shift does not occur systematically is not unexpected, either, as variation with respect to the occurrence of object shift also seems to be found in non-inverted clauses (cf. Table 8). This suggests that object pronouns only undergo optional movement to the left. In conclusion, low negation cannot be excluded as a feature of the Middle English inverted main clause data from periods m3 and m4, and high negation is not required to account for them. The diachronic implications of the analyses presented in section 4 are therefore compatible with the data from the PPCME2.

7. Conclusions Developing some of van Kemenade’s empirical observations and theoretical proposals, I have proposed the following scenario for the development of secondary negation in early English. The Old English secondary negator na can occur in a high or a low position, with the high one being restricted to V-to-C movement contexts and the low one (contrary to what van Kemenade assumes) being available in all contexts. At this point, there is no evidence for the placement of na in a position dedicated to negation and it can therefore be analyzed syntactically as an adverb. 15. In addition, there is one conditional clause with subject-verb inversion where the object pronoun precedes not in period m4.

142

Eric Haeberli

The restriction of na to V-to-C contexts can be accounted for by relating V-to-C in negative contexts to an uninterpretable negative feature in C and by analyzing the licensing of negative elements in terms of a system of binary Agree as proposed by Haegeman & Lohndal (2010). In Early Middle English, the main properties of the secondary negator not remain the same as those of na in Old English. But the evidence discussed in Haeberli & Ingham (2007) suggests that at least low not may not have the status of an adverb any more but occupies a NegP. This situation remains stable until the fourteenth century when Neg V1 is lost. As a consequence of this development, high not cannot be licensed any longer and low not remains the only negative marker. Orders of the type ‘not-subject’ are maintained, however, through the emergence of a head use of not which is manifest first in clauses with demonstrative subjects and later in clauses with pronoun subjects.

References Alexiadou Artemis & Elena Anagnostopoulou 2001 The subject-in-situ generalization, and the role of case in driving computations. Linguistic Inquiry 32: 193–232. Allen, Cynthia L. 1995 Case marking and reanalysis. Grammatical relations from Old to Early Modern English. Oxford: Clarendon Press. Chomsky, Noam 2001 Derivation by phase. In: Michael Kenstowicz (ed.), Ken Hale: A life in language, 1–52. Cambridge: MIT Press. Einenkel, Eugen 1912 Die englische Verbalnegation. Anglia 35, 187–248. Haeberli, Eric 2000 Adjuncts and the syntax of subjects in Old and Middle English. In: Susan Pintzuk, George Tsoulas & Anthony Warner (eds.), Diachronic syntax: Models and mechanisms, 109–131. Oxford: Oxford University Press. Haeberli, Eric 2005 Clause type asymmetries in Old English and the syntax of verb movement. In: Montse Batllori, Maria-Lluı¨sa Hernanz, Carme Picallo & Francesc Roca (eds.), Grammaticalization and parametric variation, 267–283. Oxford: Oxford University Press. Haeberli, Eric & Richard Ingham 2007 The position of negation and adverbs in Early Middle English. Lingua 117: 1–25.

Looking high and low for NegP in early English

143

Haegeman, Liliane & Terje Lohndal 2010 Negative concord and (multiple) Agree: A case study of West Flemish. Linguistic Inquiry 41: 181–211. Ingham, Richard 2005 The loss of Neg V!C in Middle English. Linguistische Berichte 202: 171–206. Ingham, Richard 2006 NegV1 and secondary negation in Old and Middle English religious prose. In: Joanne Close, Alexandra Galani, Beck Sinar & Phillip Wallage (eds.), York Papers in Linguistics 2 (5): 29–49. University of York: Department of Language and Linguistic Science. Kemenade, Ans van 1997 V2 and embedded topicalization in Old and Middle English. In: Ans van Kemenade & Nigel Vincent (eds.), Parameters of morphosyntactic change, 326–352. Cambridge: Cambridge University Press. Kroch, Anthony & Ann Taylor (eds.) 2000 The Penn-Helsinki Parsed Corpus of Middle English (PPCME 2). University of Pennsylvania, Philadelphia. Pintzuk, Susan 1999 Phrase structures in competition: Variation and change in Old English word order. New York: Garland. Pintzuk, Susan & Eric Haeberli 2008 Structural variation in Old English root clauses. Language Variation and Change 20: 367–407. Taylor, Ann & Susan Pintzuk 2009 Information structure and the syntax of objects in Old English: A corpus study. Paper presented at the Philological Society workshop Corpus-based advances in historical linguistics, University of York. Taylor, Ann, Anthony Warner, Susan Pintzuk & Frank Beths (eds.) 2003 The York-Toronto-Helsinki parsed corpus of Old English. University of York, York. Zeijlstra, Hedde 2004 Sentential negation and negative concord. Ph.D. dissertation, University of Amsterdam.

Ne-drop and indefinites in Anglo-Norman and Middle English Richard Ingham 1. Introduction The history of English negation must count as one of the better-researched areas in the study of language change, yet certain aspects of it still remain unclear, even as regards what is probably the most salient change, the replacement of ne by not as the main sentential negator. This appears to have taken place some time between the middle third of the thirteenth century and the last quarter of the fourteenth, but little substantial and reliable prose data is extant for that period. The prose textual record shows that around 1220 ne was obligatory, while not served as a reinforcing negator, present around 50% of the time, but that by the late fourteenth century, the situation had flipped: now not was more or less obligatory, while ne was much rarer (Iyeiri 2001). But the nature and timing of changes during the intervening 150 years, approximately 1230–1380, is still obscure. In ordinary usage, the register in which language change is usually considered to take place, not is likely to have been in the process during that period of becoming the preponderant negative form, to judge from the near-obligatory status it attained in the written register by the late fourteenth century. But it is unclear whether not continued to rise in frequency before extensive ne-drop set in, or whether a sharp decline in the use of ne accompanied a further rise in the use of not. Frisch (1997) examined negative particles in a combined corpus of prose and verse from early to late Middle English. It showed that from 1150 to 1290 ne was supplied over 95% of the time, and 90% in 1290– 1360, but that in 1360–1430 the rate of ne use collapsed to only 20%. By this last period, sentential negation was normally with not alone, if a negative indefinite was not present. However, Frisch’s study conflates prose and verse data indiscriminately, so the finding that the Early Middle English negation pattern was maintained mostly intact until 1360 could be an artefact of using largely verse data for the period 1290–1360. Verse data is anyway problematic for studying syntax, for various reasons: the exigencies of rhyme and metre, a tendency to archaism, and the fact that the extant verse manuscripts may be partial reworkings of older material.

146

Richard Ingham

Iyeiri (2001) also studied the development of Middle English negative particles quantitatively across the time span of interest to us, though unlike Frisch she kept poetic texts separate from prose. Few of the poetic texts, unfortunately, can be securely placed within our target period 1230– 1380 in terms of composition dates and manuscript dates, so no clear account of successive temporal stages was achievable. Ingham (2005) followed Iyeiri’s initiative, analysing negation using verse data alone across the Middle English period, but went a stage further and categorised material into 50-year periods by estimated manuscript date. However, the aim of the study was not to identify characteristics of negation in a given time-period in the history of English, but to see only whether the loss of ne was correlated with the loss of Negative inversion in the same manuscript periods. There was no suggestion of taking poetic texts as witnesses of states of language at the time those manuscripts were written. In the course of this investigation, it was found that ne-drop occurred over 20% of the time in manuscripts from 1300–1350, a rather higher figure than in Frisch (1997), but the interpretive problem remains that verse manuscripts, even though more tightly linked to a particular period than in preceding research, are not really the sort of data desired in diachronic syntax. The ‘bad data or no data’ problem makes it di‰cult to choose between two theories of what happened with negation in English. One is to construe the ‘weakening’ of the short preverbal negator ne as proposed by Jespersen (1917) and others as erosion of its phonetic substance, to the point where ne was no longer audible as a surface cue to the semantic value of the clause. The other is that, though still audible, ne no longer provided a reliable semantic cue, having weakened from a morphologically negative to an expletive item. Kiparsky & Condoravdi (2006) have argued that a reinforcing negator appeared in Greek without phonological erosion being responsible for the loss of the earlier negative form. The changes that occurred in English may also have been driven by semantic/ pragmatic factors, and not, or not only, by phonology. The establishment of an originally reinforcing negator as an ordinary way of expressing negation would from this perspective have led to the displacement of ne, and the phonological weakening of the vowel played at best no more than a minor role. The extent of phonological reduction of the negator ne at this period in medieval English is hard to determine. No direct evidence from Old English/Middle English verse attests to the vowel quality of letter in unstressed syllables. Indeed, Lass (2009) has recently cast doubt on whether

Ne-drop and indefinites in Anglo-Norman and Middle English

147

Old and Early Middle English actually had a schwa vowel at all. However, in favour of an extreme reduction of unstressed final vowels written in Middle English is the fact that Old English loans into Welsh (ParryWilliams 1923) show no trace of a final syllable in words like Welsh ysber: OE/ME spere, Welsh ysteˆn: OE staena, Welsh som/siom ‘shame’: OE sceamu, suggesting that the source item was commonly pronounced even before 1150 with no determinate final unstressed vowel quality, perhaps schwa at most. The evidence that weakening of ne was phonologically driven is inconsistent. In Middle English texts of the first half of the thirteenth century there is no evidence of erosion of ne. Yet in Old English ne-drop is already seen in non-West Saxon origin texts (Ingham 2006a), and in West Saxon texts the vowel of the negator is elided to produce forms such as nis for ne is. It is not clear why thirteenth century Middle English did not show ne-drop if ne was lost because of phonological weakening, which seemingly had begun much earlier. More generally, and to borrow concepts from phonological change, the question amounts to whether the change in negation in Middle English was a case of ne being ‘pushed’ by the expansion of not, or of not being ‘pulled’ into taking the place of a weakened ne. If the use of not in Middle English rose markedly while ne remained more or less obligatory, this would favour the idea that ne did not disappear for phonological reasons, but for pragmatic ones. If, on the other hand, ne was rapidly replaced by not this would favour the idea of phonologically determined loss. However, as noted, the opportunity to answer this question is frustrated by the lack of reliable data in the ‘middle’ Middle English period. A further uncertainty concerns the relation between sentential negation and indefinites under negation. According to Jespersen (1917), European languages with a short preverbal negator (Slavic, earlier Germanic) tend to have Negative Concord, between the sentence negator and n-item indefinites, that is, indefinite expressions which stand in a negative clause, but do not themselves express clausal negation in the absence of the sentence negator. However, languages with a more substantial postverbal negator (modern Northern and West Germanic) usually do not have Negative Concord. They have indefinites that have an incorporated negative meaning, e.g. German kein, Swedish ingen. Since English went from being a language of the first type to one of the second type, it would be expected to have lost negative concord as part of that overall process. If not became the main negator in the later thirteenth century, we expect n-items such as none to become able to introduce negation by themselves.

148

Richard Ingham

Late Middle English (Ingham 2000, 2007) indeed shows plentiful examples of clauses apparently negated by single n-items, so it would seem straightforward to account for this development by saying that they incorporated negation, whereas previously, when ne was the main negator, they were NPIs. The second question, then, is therefore whether this change occurred in the feature makeup and licensing of negative clause indefinites in English in the ‘middle’ Middle English period, when, as we have seen, prose data is deficient.

2. Design of the study In this study a hitherto unexamined source of data is brought to bear on these questions. Large amounts of prose material are extant from this period, written in later Anglo-Norman, the insular variety of medieval French used in England after the Norman Conquest. They were composed at a time when English is thought to have exerted considerable substrate influence on the bilingual insular French users who produced these texts. In this study the possibility will be considered that the underlying regularities of English negation at the time in question may be discerned by analysing the ways in which Anglo-Norman negation departed from the rules of Old French negative syntax. Since no detailed study of negation in these sources has hitherto been conducted, much of this article is devoted to an empirical investigation of the relevant properties of negation in later Anglo-Norman, which will then be compared with Middle English. Undoubtedly, this is an unusual way of seeking to identify the properties of a linguistic system, and would not normally be adopted if the target period could be studied directly. However, there are a number of reasons why the approach we adopt has some prospect of success. First, the concept of the syntactic system of one language exerting contact influence over the syntactic system of another one, until recently poorly studied, has now been given a thorough framework of analysis thanks to the work of Heine & Kuteva (2005), which presents the notion of syntactic replication, and will be adopted in the present study. Secondly, the sheer amount of prose material which is closely datable both in terms of its time of composition and of the provenance of the manuscript, is favourable to periodisation in diachronic study in a way that much literary material regrettably is not, and naturally favours quantitative investigation: the number of manuscripts in Anglo-Norman in the thirteenth and four-

Ne-drop and indefinites in Anglo-Norman and Middle English

149

teenth centuries far exceeds that available in Middle English, so regularities that in syntactic research require very substantial amounts of data have a good chance of surfacing. Finally, there are very good grounds for believing that negation was a domain in which Anglo-Norman was influenced by English, in connection with the issue of phonological erosion of ne mentioned above. AngloNorman is already known to have shown ne-drop, long before this became productive in continental French (Martineau & Mougeon 2003; Martineau this volume). This might be related to another well-known feature of AngloNorman, the tendency to lose what will be called the weak ‘e’ vowel. Forms such as fra for fera, and pelrin for pelerin (Short 2007), in which the unstressed vowel is not realised at all, indicate its extreme vulnerability, and research into Anglo-Norman prosody (Du¤ell 2005) shows conclusively the metrical irregularities caused by the loss of weak ‘e’. The negative form containing the weak ‘e’ vowel was clearly more vulnerable in insular than in continental French, and a very plausible case can be made for seeing this as a reflection of English influence, since Middle English shows extensive e¤ects of vowel attrition in unstressed syllables (Lass 2006: 61). Thus Anglo-Norman ne-drop is plausibly attributable to contact with English. Although Ingham (2006b, c) found that later Anglo-Norman syntax succeeded in remaining independent of English, specifically as regards subject and object pronouns, Anglo-Norman showed substantial English phonological influence, and where this directly impinged on a grammatical element such as ne, syntax could well have been a¤ected. In the following sections we shall first consider how Old French di¤ered from Old and Early Middle English as regards the syntax of negation, then report the findings of an analysis of ne realisation with indefinites and negative particles in Anglo-Norman texts of the late thirteenth and earlier fourteenth centuries, after which the implications of these findings for the nature of n-words in Anglo-Norman and in Middle English is discussed.

3. Negation in Old French In Old French it is common to find reinforcement of ne by particles such as pas, mie etc. (Price 1997). Gregory (1997) investigated the frequency of use of reinforcing particles in Old French negation, using prose texts from

150

Richard Ingham

the twelfth century onwards.1 His overall finding is that already by the later twelfth century prose texts reinforce the sentential negator ne with pas or the like somewhat over half of the time. By the Middle French period (from c. 1300 onwards), reinforcing negative particles become the norm. Plainly, therefore, English and French were in this respect evolving along parallel lines. Yet there was an important di¤erence: Middle English was a negative concord language whereas Old French was not. That is to say that the Middle English indefinites non, neuere, and the reinforcing expression noht/nawiht, later not, were used only in negative clauses, unlike nul, jamais and pas which could be used in other clause types such as interrogatives and conditionals. The Middle English items were therefore not simply polarity items, but were ‘n-items’, i.e. expressions restricted to negative contexts. In medieval French subject-verb inversion was the norm after an initial adjunct, such as ore ‘now’ aussi ‘also’ or a prepositional phrase. However, time-adverbs under negation, especially ja(mais), onques (both meaning ‘never’), and PPs with nul ‘none’, did not trigger inversion, e.g.: (1) (Nos ne chevaucherons entre moi et toi fors de nuiz,) car por nule riens ge ne voudroie estre conneu¨z en ceste voie. (Mort le roi Artu, p. 6) This was arguably because as polarity items they needed to be licensed within IP (Ingham, to appear). This continued to be the case into Middle French,2 e.g.: (2) Onques je n’en enquis tant ‘I never asked for so much.’

(Jean D’Arras, Me´lusine, p. 190)

In early Middle English, inversion after n-items was also absent, presumably for the same reason that n-items, bearing a [uNEG feature], also needed to remain within a licensing domain for feature checking. 4. Source texts and contexts of analysis The material for this study are drawn from two non-literary sources extant in contemporary manuscripts, the Parliamentary Rolls of Medieval 1. It is somewhat di‰cult to interpret his figures, which are not tabulated, but cases of ne alone seem to account for 37.5% of contexts in Sermones in cantica, a text of the later twelfth century, and 47.1% in the somewhat earlier Psalter Commentary. 2. According to Ingham (to appear) this is because they do not topicalise and need to be licensed in the IP domain of the negative particle ne.

Ne-drop and indefinites in Anglo-Norman and Middle English

151

England (Given-Wilson 2005) and the largely Anglo-Norman Year Books for the 1290s and 1340s. The latter consist of records of successive cases, typically taking the form a transcript of the initial plea by the prosecuting lawyer, and then the ensuing debate between the justices regarding the merits of the case, the latter presenting dialogic discourse. These passages allow a comparison to be made with legal register written-mode texts, the petitions contained in the Parliament Rolls. In accordance with the issues raised in section 1, the contexts for analysis were: a) b)

negative clauses with reinforcing particles negative clauses with indefinites

In the next section we report the results of this investigation, together with observations on qualitative di¤erences that emerged between Anglo-Norman and continental French which seem relevant to the change regarding nedrop. 4.1. Erosion of ne with indefinites in Anglo-Norman In the 1290s, ne is usually maintained in both source texts, e.g.: (3) [. . .] ke yl naveyt nule disseisine fet. ‘that he had committed no disseisin.’

(YB XX Edw I, p. 5. 1292)

(4) [. . .] e pur ceo ne serra rien fet. ‘and therefore nothing will be done.’ (PROME Edw I Roll 4. c. 1290) In the 1340s, we find plenty of evidence that Anglo-Norman ne was being eroded with indefinites, e.g.: (5) Nulle ley nous mette a respondre a ceo qil ount allege. ‘No law requires us to respond to what they have alleged.’ (XV Edw III, p. 33. 1341) (6) Nous voloms rien faire. ‘We do not want to do anything.’

(Edw III, p. 313. 1340)

(7) Nous agardroms jammes execucion a luy. ‘We shall never award execution to him.’ (XX Edw III, p. 125. 1346)

152

Richard Ingham

Cases are found of ne-drop where the indefinite expression precedes, and also where it follows, the finite verb. In other words, Anglo-Norman did not adopt the pattern of modern Italian, in which the sentential negator is dropped when an n-item is preverbal, but not when postverbal. Overall frequencies of ne-drop are tabulated in Table 1 below. It can be seen that ne-drop was quite common in the 1340s, but in the late thirteenth century was still rare. Furthermore its growing use in the fourteenth century was largely confined to contexts with indefinites, especially nul. Indeed, in sentential negation contexts ne-drop never occurred in PROME in either period. The key finding in these overall results, then, is that certain Anglo-Norman indefinites, those which occurred in negative contexts, became able to introduce negation by themselves. The process was already beginning in the late thirteenth century, and became commonly followed in the 1340s. Figure 1 shows a breakdown of results by source texts. Of interest is that in the 1340s the level of ne-drop was higher in PROME than in the Year books. The latter include large amounts of transcribed dialogue, which, although not of conversational nature, was at least delivered orally. The petitions in PROME, however, are uniformly written discourse. It does not seem, then, that ne-drop as such characterised spoken language, otherwise we might expect to see more ne-drop in the Year Books, rather than the other way round. Conceivably, its occurrence in the Year Books in sentential contexts might be seen as reflecting spoken language use, but this probably depends on how convincing one finds the di¤erence between zero occurrences in PROME and a level of about 2–3% occurrences in the Year Books. In both source texts, ne-drop with indefinites in negative context reached quite substantial levels in the 1340s. This was especially true in the case of nul, where nearly half of all occurrences showed ne-drop. We have then, a clear distinction between an item that was very commonly able to express negation by itself, and other items, such as pas and mie, which generally could not. The theoretical implications of this di¤erence will be considered below. First we wish to compare these results for Anglo-Norman with those of the study by Ingham (2005) mentioned in the introduction, which analysed ne-drop in Middle English verse for the periods 1250–1299 and 1300– 1350. This investigation exclusively profiled verse data, but despite our reservations stated earlier about using such source data for diachronic syntax, it represents the only substantial source of information available about the ‘middle’ Middle English period that is specifically focused on

16

69

mie

nient

2

0

2

57

13

57

rien

jamais

onques

%

Total

93

nul

1

1

2

11

ne-drop with indefinites

%

Total

143

pas

46

5

73

149

3

62

49

0

0

4

19

0

0

0

ne

þne

þne

ne

PROME Edw I

YB 1290s

ne-drop with negative sentence particles

92.8

493

103

18

130

242

98.8

342

72

78

192

þne

7.2

38

1

1

6

30

1.2

4

2

0

2

ne

S/T1290s

9

9

47

30

2

60

289

þne

0

3

5

25

2

3

6

ne

YB 1345/6

9

2

9

87

0

8

53

þne

0

0

3

73

0

0

0

ne

PROME 1340s

Table 1. Frequency table of ne retention versus drop in AN negative clauses

64.9

202

18

11

56

117

97.4

412

2

68

342

þne

35.1

109

0

3

8

98

2.6

11

2

3

6

ne

S/T1340s

Ne-drop and indefinites in Anglo-Norman and Middle English

153

154

Richard Ingham

Figure 1. Percentage of ne-drop by context and by AN source text

the evolution of negation including negated clauses with indefinites, the latter having been excluded by Frisch (1997). It distinguished on the one hand indefinites never, no, none and on the other the sentential negator noht, so the contexts are syntactically and semantically identical in design with the contexts identified in the present study. Subject to due caution, therefore, it is worth comparing with the Anglo-Norman data. In Figure 2, the percentage values of ne-drop are presented for the two languages in the two contexts. The values obtained in the present research for Anglo-Norman are plotted against those found for the Middle English verse texts by Ingham (2005), both shown for a ‘period I’ and a ‘period II’. The time frames referred to by these labels are similar but not identical: the Middle English data were collected from verse manuscripts dated to between 1250–1299 and 1300–1350 respectively. The Anglo-Norman data, as we have seen, come from shorter time windows, though still within each of those two 50-year periods. The category label ‘Sent. Neg.’ designates contexts where a clause is negated by a sentential negator in the absence of an indefinite expression. Ne-drop with indefinites is indicated by the category label ‘indef. under neg.’ Even comparing prose with verse, as Figure 2 does, the divergence between the languages is so large that it is hardly possible to see the findings as due to the di¤erence in source genre. In Middle English no clear di¤erence is observed in the likelihood of ne-drop between sentential negative and indefinite contexts. Although the rate of ne-drop in indefinite

Ne-drop and indefinites in Anglo-Norman and Middle English

155

Figure 2. Percentage of ne-drop in later AN and ‘middle’ ME

contexts is somewhat higher,3 in both contexts it accelerates noticeably in period II. In Anglo-Norman, by contrast, a wide gap opens up between the two contexts: as we saw in Figure 1, ne-drop remained very low in sentential negative contexts, whereas in indefinite contexts it rose steeply. The disparity between the types of textual material makes it unwise to attempt a close comparison between the Anglo-Norman and the Middle English data, but there can be little doubt that in ordinary usage in the first half of the fourteenth century, Middle English noht must have been strongly gaining ground as the sentential negator, whereas there was no real sign that Anglo-Norman might show a similar evolution in favour of pas or mie. The figures from the Anglo-Norman sources analysed here leave little doubt that a change was in progress in the status of indefinites in negative clauses in the first part of the fourteenth century. In the following sections we examine two further areas where later Anglo-Norman indefinites under negation clearly diverged from continental French usage around this time. 4.2. Sentence fragments In the Year Books of the fourteenth century, the indefinites jamais and nul regularly introduced negation in a sentence fragment, in the absence of 3. In fact, it was to close again in the later fourteenth century; see Ingham (2005: 192, table 2).

156

Richard Ingham

negative clause structure. This can be seen clearly with responses to information questions, involving both NP and PP fragments: (8) Quel accion averait ele? Nule, a vostre entente. ‘What action would she have? None, according to your intent.’ (XVII Edw III, p. 57. 1342–3) (9) Qel respouns averez vous vers la demandante [. . .] autre qe vous navez ore? – Richemunde. Pur cas nule autre. ‘What answer would you have against the demandant other than you have now? – Richmond. Possibly no other.’ (XVII Edw III, p. 507. 1344) (10) Berwick: En quel parose est ceo? – Lanfar. En nule. ‘Berwick: In which parish is this? – Lanfar. In none.’ (XXX Edw I, p. 187. 1302) (11) Quei avendreit de cel issue? A nul purpos pur moy. ‘What would happen from that issue? It would be to no purpose as far as I am concerned.’ (XVII Edw III, p. 385. 1343) Similarly, jamais could be used in isolation to refute the previous proposition: (12) Par ajournement, et issi voleint faire hors de Bank le Roi ajourner icy. – Pole. Jammes. (XV Edw III, p. 335. 1340–1) (13) Votre protestacion vous sauvera de mischief. – Stouford. Jammes (XVII Edw III, p. 317. 1344) No such examples have been found by the present author in the two online prose texts in the Laboratoire de Franc¸ais Ancien, or in the printed editions of the other four prose romances used by Ingham (to appear).4 They are not found in PROME, no doubt because petitions do not contain dialogue.

4. Tobler & Lommatsch (1925–), sub nul, have only one example of nul used in a sentence fragment comparable to (10). It is from a thirteenth century poem written by the Recluse of de Moilliens, a place near the Picard town of Amiens, which is about 80 km from the then linguistic boundary with Flemish. The possiblity of contact influence on North-Eastern Old French from Flemish is one to which we return in section 6.

Ne-drop and indefinites in Anglo-Norman and Middle English

157

4.3. Subject-verb inversion after indefinites A further intriguing consequence of the changing status of indefinites under negation in Anglo-Norman is that they took on a property they did not have in continental French, as regards subject-verb inversion. In the fourteenth century Year Books, clause-initial adjunct PPs with nul trigger inversion, e.g.: (14) (Par la primer clause en le fet tout le manoir est charge de la rente,) et par nul parole apres est il descharge. ‘And by no subsequent word is it discharged.’ (XV Edw III, p. 323. 1341) (15) En nul cas avera home essone apres essone de service le Roi [. . .]. ‘In no case shall one have essoin after essoin on the King’s service.’ (XVII Edw III, p. 37. 1342–3) This appears to be some kind of focusing operation comparable to Neg Inversion in modern English (Haegeman 1995). After other focused constituents, inversion was triggered in Anglo-Norman, e.g. bone demande in: (16) Je vous ey done bon bref en les deux villes, mes bone demande vous durray jeo pas. ‘I have given you a good writ in the two vills, but I shall not give you a good demand.’ (XVII Edw III, p. 593. 1343) Fronted jamais likewise regularly triggers inversion in the later Year Book dialogues, e.g.: (17) Jammes avera il bref a supposer remainder outre. ‘He will never have a writ to suppose a remainder over.’ (XVII Edw III, p. 343. 1343) (18) Jammes ne meintendra il soun brief. ‘He will never maintain his writ.’

(XX Edw III, p. 13. 1346)

The fact that inversion has taken place with a subject pronoun guarantees that, in terms of generative syntax, we are dealing with movement of the verb to the CP domain, rather than with movement within IP (Vance 1997). Thus examples (17)–(18) align clauses with initial jamais with the Verb-Second property of Old French, even though in continental sources this syntactic behaviour is not found with jamais (Ingham to appear, a).

158

Richard Ingham

Inversion does not seem to have been correlated with ne-drop itself: numerous cases are found where jamais inverts even though ne is retained, as in (18). However, if by this time ne was treated as an expletive this fact is unsurprising. Significantly, inversion with adjuncts containing nul and jamais are absent from the Year Books of the late thirteenth century. It seems clear that inversion with polarity items is a fourteenth century phenomenon in Anglo-Norman, as with their use in sentence fragments. 5. Discussion The empirical findings from fourteenth century Anglo-Norman are that indefinites commonly introduced negation by themselves, both in clauses, and in sentence fragments, whereas reinforcing negative particles pas and mie did so only marginally, and in only one of the source texts. For indefinites, especially nul, rien, and jamais, to be able to negate a clause by themselves means that they had undergone a change of status diverging from continental French. Since nul occurs in sentence fragments, we consider it had a lexical entry in which it was not a PI, but an n-item.5 In Modern French, aucun, jamais and rien are fine in sentence fragment contexts, and they are n-items, not polarity items.6 In other words, although Anglo-Norman prefigured a change to take place in the later development of French, the reason seems to have been because indefinites were undergoing influence from English as to their syntactic and semantic make-up, gaining an entry in which they bore a negative feature. It seems to be as a result of this change of status to n-items that they became able to trigger subject-verb inversion. In continental French the only negative adverbial element that was able to do this was the inherently negative item non, e.g.: (19) Non ferait il la moie.

(Haut Livre 494)

(20) Non fit onques nul chevalier.

(Haut Livre 544)

Although pas could be fronted, it did not thereby trigger inversion. This follows from the fact that it was not a negative expression, but a polarity reinforcer. 5. PIs are not licensed in sentence fragment contexts. 6. However, inversion was lost in French before this change occurred, so inversion with n-items is not found in Modern French either.

Ne-drop and indefinites in Anglo-Norman and Middle English

159

Let us now return to the issues to do with English stated at the outset. First we asked whether ne-drop in Middle English was phonologically driven, or whether the rise of reinforcing negation was the decisive factor. The near-complete (over 97%) maintenance of Anglo-Norman ne with sentence negative particles in the 1340s indicates that the erosion of ne was not an across the board phenomenon as we would expect if it was purely driven by phonology. The petitions in PROME made no use at all of unsupported pas, mie or nient. The primary reason for the loss of ne must surely therefore be to do with the changing nature of other elements in the clause. Apparently, items such as nul, in particular, gained the status of being able to introduce negation by themselves, whereas the sentential negative particles pas and mie generally resisted this development.7 This is a significant finding, because it casts doubt on the assumption made earlier that weakening of unstressed vowels in English was the driver of the process. In Middle English, an identical process appears to have been at work in the case of all n-items. The nature of this strengthening of negation remains to be explicated, but it seems clear that a key di¤erence is that the reinforcing negator not was originally an n-item whereas pas and mie were not. Hence the short negator ne could be eroded in both types of context, without phonology playing the decisive role. The comparison between Middle English and Anglo-Norman makes it very clear that only where the reinforcing element was an n-item was ne dropped. In Table 2, continental French, Anglo-Norman, and ME are compared for the relevant traits: Table 2. Comparison of behaviour of indefinites in negative clauses Cont Fr 120-1400

Later AN

Late Middle English

(i) Indefinites appear in negative contexts accompanied by ne.

(i) Indefinites appear in negative contexts accompanied by ne.

(i) indefs appear in negative contexts accompanied by ne.

(ii) –

(ii) indefinites introduce negation

(ii) indefinites introduce negation

7. Nient as a reinforcing negative particle seems to have dropped out in the fourteenth century, at least in these data sources.

160

Richard Ingham

It is proposed that later Anglo-Norman replicated ME, but this entails that ME must already have had property (ii) for replication to have taken place. The properties of Anglo-Norman that display evidence for replication of English are already visible in the first half of the fourteenth century, so we therefore propose that ME had property (ii) at least as early as that period. Recent work by Hoeksema (2009) on the contemporary evolution of reinforcing expressions in English, and by Willis (2004, this volume) on the diachrony of Welsh provides parallels for this development. Items that were originally n-items requiring a sentence negator to licence them take on the status of being able to introduce negation by themselves. This line of reasoning can be applied to the development of none, never etc. from negative polarity items to being inherently negative expressions. Then the process of replication under language contact would have a¤ected Anglo-Norman, to produce indefinites having the equivalent property. There remains nevertheless the logical possibility that Anglo-Norman was not influenced by English but was undergoing an independent development, but this appears implausible. The same phenomenon of ne-drop with nul in late Old French and early Middle French was identified by Ingham (2011) in documents written in Saint Omer, then on the border of the French and Flemish speaking areas of modern-day Northern France. Speakers of the two languages would have been in frequent contact, and it is known that Flemish was undergoing the same evolution of the loss of the preverbal negator and the incorporation of negation in the indefinite gheen, corresponding to nul. Since the only reported instances of ne-drop in the medieval French dialect continuum occurred where there was contact with a Germanic language, the likelihood that Anglo-Norman was similarly a¤ected is very strong.

6. Conclusions In this article it has been shown that ne-drop in Anglo-Norman was sensitive to whether the clause contained a reinforcing expression or an indefinite. Furthermore, indefinites in Anglo-Norman became used in ways that were not countenanced in continental French. Thus ne-drop was conditioned lexico-grammatically, not phonologically. Assuming that it was influenced by English, therefore, Anglo-Norman negation seems to have shown the influence not so much of phonological erosion of ne, but of a shift in the properties of indefinites. When these took on incorporated

Ne-drop and indefinites in Anglo-Norman and Middle English

161

negative characteristics in the earlier fourteenth century, ne was dropped very commonly. We have argued that the shift in the properties of indefinites in Anglo-Norman was a case of replication, in terms of Heine & Kuteva (2005), of the shifting properties of indefinites in English. Anglo-Norman was already showing these properties in the first half of the fourteenth century, before any reliable prose evidence is available for English. The conclusion therefore is that the shift in the nature of negation in Middle English was not a late fourteenth century phenomenon, but took place considerably earlier than previous investigators have suggested. N-words such as none and never, would probably have been in ordinary use without ne by the early fourteenth century. The motive for this change, in which ne was lost in Middle English, was probably not, or at least not only, phonetic erosion, since phonetic erosion does not explain the pattern of retention and loss in Anglo-Norman, which is known to have been strongly influenced by English pronunciation. A tendency already observed in other languages for NPIs to become independent negative expressions su‰ces to explain the change in status of indefinites in English, and, on a contact hypothesis, in Anglo-Norman. References Sources Haut Livre 2007

Le Haut Livre du Graal [Perlesvaus]. Edited by Armand Strubel. Paris: Le Livre de Poche. Jean D’Arras, Me´lusine 2003 Jean d’Arras. Me´lusine ou La Noble Histoire de Lusignan. Edited by Jean-Jacques Vincensini. Paris: Lettres gothiques. Mort le roi Artu 1954 La mort le roi Artu, roman du XIIIe sie`cle. Edited by Jean Frappier. Geneva: Droz. PROME 2005 Parliament Rolls of Medieval England. Edited by Chris GivenWilson. Scholarly Digital Editions. YB 1866, 1873 Yearbooks of the reign of King Edward the First, Years XX and XXI. Edited by Alfred J. Horwood. London: Longman. YB 1911 Yearbooks of the reign of King Edward the Third, Years XVII– XX. Edited by Alfred J. Horwood & Luke O. Pike. London: Longman.

162

Richard Ingham

Studies Du¤ell, Martin 2005 Some phonological features of insular French: A reconstruction. In: R. Wright & P. Ricketts (eds.), Studies on Ibero-Romance linguistics dedicated to Ralph Penny, 103–125. Newark: Juan de la Cuesta. Frisch, Stefan 1997 The change in negation in Middle English: a NEGP licensing account. Lingua 101: 21–64. Gregory, Stewart 1997 Negative particles in French prose of the twelfth century. In: Stewart Gregory & David Trotter (eds.), De mot en mot, Aspects of medieval linguistics, 37–51. Cardi¤: University of Wales Press. Haegeman, Liliane 1995 The syntax of negation. Cambridge: Cambridge University Press. Heine, Bernd & Tania Kuteva 2005 Language contact and grammatical change. Cambridge: Cambridge University Press. Hoeksema, Jack 2009 Jespersen recycled. In: Elly van Gelderen (ed.), Cyclical change, 15–34. Amsterdam: Benjamins. Ingham, Richard 2005 The loss of Neg V ! C in Middle English. Linguistiche Berichte 202: 171–206. Ingham, Richard 2006a On two types of negative concord in early English. Language Variation and Change 18 (3): 241–266. Ingham, Richard 2006b The status of French in medieval England: Evidence from the use of object pronoun syntax. Vox Romanica 65: 1–22. Ingham, Richard 2006c Syntactic change in Anglo-Norman and Continental French chronicles: Was there a ‘Middle’ Anglo-Norman? Journal of French Language Studies 16 (1): 25–49. Ingham, Richard 2007 NegP and negated constituent movement in the history of English. Transactions of the Philological Society 105 (3): 1–33. Ingham, Richard to appear A derivational approach to negative polarity item licensing in Old French. In: Deborah Arteaga (ed.), Historical Romance linguistics. Amsterdam: Benjamins. Ingham, Richard 2011 L’anglo-normand et la variation syntaxique en franc¸ais me´die´val. In: Maria Iliescu, Heidi Siller-Runggaldier & Paul Danler (eds.), Proceedings of the XXVth Colloque International de Lin-

Ne-drop and indefinites in Anglo-Norman and Middle English

163

guistique et de Philologie Romane. Berlin/New York : Mouton de Gruyter. Iyeiri, Yoko 2001

Negative constructions in Middle English. Fukuoka: Kyushu University Press.

Jespersen, Otto 1917 Negation in English and other languages. Copenhagen: A.F. Høst. Kiparsky, Paul & Cleo Condoravdi 2006 Tracking Jespersen’s cycle. In: Mark Janse, Brian D. Joseph & Angela Ralli (eds.), Proceedings of the Second International Conference on Modern Greek Dialects and Linguistic Theory, 172– 197. Mytilene: Doukas. Lass, Roger 2006 Phonology and morphology. In: Richard Hogg and David Denison (eds.), A history of the English language, Volume II: 1066–1476. Cambridge: Cambridge University Press. Lass, Roger 2009 Weak segments in early Middle English. In: Donka Minkova (ed.), Phonological weakness in English: from Old to PresentDay English, 237–315. Basingstoke: Palgrave. Martineau, France & Raymond Mougeon 2003 Sociolinguistic research on the origins of ne deletion in European and Quebec French. Language 79 (1): 118–152. Parry-Williams, Thomas H. 1923 The English element in Welsh: A study of English loan-words in Welsh. London: Honourable Society of Cymmrodorion. Price, Glanville 1997 Negative particles in French. In: Stewart Gregory & David Trotter (eds.), De mot en mot, Aspects of medieval linguistics, 173–190. Cardi¤: University of Wales Press. Short, Ian 2007 Manual of Anglo-Norman. London: Anglo-Norman Text Society. Tobler, Adolf & Erhard Lommatsch 1925– Altfranzo¨sisches Wo¨rterbuch. Wiesbaden: F. Steiner.

Looking at Middle English through the mirror of Anglo-Norman Jack Hoeksema Now we see but a poor reflection as in a mirror; then we shall see face to face. Now I know in part; then I shall know fully, even as I am fully known. (Corinthians 13: 12, New International Version)

1. Introduction Many details of interest for historical linguistics may be lost forever. Even English, with its enviable abundance of texts from all periods, has significant gaps in the textual record, making it frustratingly hard to figure out precisely what went on in crucial periods of change. Perhaps not even Heaven may o¤er the retired linguist a glimpse of what went on in the year 1300 by providing a forum for conversing with his ancestors of that era, unless it turns out that linguistic change somehow stops at the Pearly Gates. Richard Ingham’s paper elsewhere in this volume on Middle English ne-drop looks at a phenomenon that must have taken place in the period 1230–1380, namely the replacement of the clitic ne by non-clitic not as the prime exponent of negation. We know it must have taken place because the period before and the period after show a major change. The main problem Ingham is wrestling with is the lack of good sources for the central period. Prose texts are the preferred source of information on syntax, for a number of reasons pointed out by Ingham. Poetic texts are often hard to date, unlike legal documents or personal letters, and may show traces of various periods and dialects, having been copied many times. Moreover, they may also di¤er from the spoken language due to the e¤ects of rhyme, meter, and stylistic preferences for archaisms (cf. for instance the discussion of SOV word order in Middle Dutch epic poetry in Weerman 1989, van den Berg 1991, held by Weerman to be an archaic pattern and by van den Berg to be due primarily to rhyming requirements).

166

Jack Hoeksema

Some of the problems raised by Ingham are more serious than others. E¤ects of rhyme can probably be ignored. Rhyme is an important factor for the study of word order variation, but in the choice between retaining or dropping clitic negation it is hard to see how rhyme would be of great importance. Meter might be a more significant factor, although its e¤ect on poetry that is not strictly metrical is likely to be fairly slight. Archaic language could be a matter of more serious concern, and dating is almost always problematic for literary texts. If there are su‰cient texts, one might try to use fixed-size parts from each, apply the best dating estimates available, and assume that for every piece of text whose estimated date is too early, another one will have an estimated date that is too late, and so on average all errors could be expected to cancel out. This is the method followed in Ingham (2005), and perhaps it could be repeated on a larger scale to gain more reliable results. Nonetheless, there is a considerable amount of inaccuracy in such e¤orts to make the best of bad data, and anyone considering the matter would have to agree with Ingham that if there were another way to study the phenomenon at hand, it would be most welcome. The alternative method of studying the sudden increase in ne-drop in the period 1230–1380 proposed by Ingham is both clever and tricky. It uses material of the right kind (prose) and precisely dated (because it involves legal texts), but in the wrong language: French, or rather, Norman-French. The operative idea is that any syntactic changes in Norman-French not found on the continent are very likely due to pressure from the dominant language on the British Isles, Middle English. And so if Norman-English shows an unexpected drop in the use of clitic negation, then, in light of the great overall similarities between the two languages (both have negative concord, both have a negative clitic ne), the most plausible hypothesis would indeed be that this reflects an earlier, or simultaneous, change in the syntax of Middle English. What I will do in this paper is to accept Ingham’s general methodology, and to comment on his findings and conclusions.1 I will do so by making comparisons with other cases of linguistic change involving negative concord. I hope that we may indeed use Anglo-French to shed light on Middle English, but what is it that we see, dimly, in this mirror? 1. I will have nothing to say about some aspects of the paper, such as the discussion of negative inversion, or the status of n-words in short answers and sentence fragments. I fully agree with Ingham on the significance of these matters, but take them to be uncontroversial (but cf. e.g. Giannakidou 2000; Penka 2007 for a di¤erent view of short answers).

Looking at Middle English through the mirror of Anglo-Norman

167

2. Indefinites versus negative particles Ingham’s findings for Norman French, presented in his Table 1, show a significant and very interesting di¤erence in the rate of ne-drop for negative sentences with a negative particle such as pas or mie, and negative sentences with negative indefinites such as rien, nul, jamais. He interprets this as evidence that the negative indefinites have been reinterpreted, by that period, as inherently negative, thus rendering ne superfluous, while the negative particles at that point were still negative polarity items, hence in need of a licensing element such as ne. The reinterpretation of rien, jamais, nul etc. as negative quantifiers could arguably be due to contact phenomena, given that the Middle English counterparts such as naught or never are firmly in the class of n-words, with inherent negative force, just like their Germanic counterparts in Middle High German or Middle Dutch. If such is indeed the case, then the onset of ne-drop in AngloNorman might reflect rather directly a similar drop in Middle English. So far, so good. A few caveats are in order, however. The proposed explanation for the di¤erence between n-words and negative particles correctly proceeds from the assumption that ne-drop requires that the counterpart of ne has inherent negative force, allowing it to become the sole exponent of negation. But this is a necessary condition for ne-drop, not a su‰cient one. Indeed, the literature abounds with cases where inherently negative elements maintain concord with a clitic negative element in spite of the redundancy inherent in such a process (cf. e.g. Labov 1972; Ladusaw 1992; Giannakidou 2000; Zeijlstra 2004; de Swart 2010). It is therefore entirely possible that a language reinterprets polarity items as negative quantifiers, and starts dropping clitic negation, while another language maintains concord between inherently negative quantifiers and clitic negation. If the former type of language is exemplified by AngloNorman, and the second by Middle English, then Ingham’s argument would fail. However, given that continental French did not show similar signs of ne-drop, and given that we should be averse to treating the Anglo-Norman change as completely coincidental, surely the most plausible scenario is the one sketched by Ingham. Another caveat concerns the few cases of ne-drop with negative particles noted by Ingham that appear in Table 1. If these negative particles are still polarity items, there should not be any ne-drop whatsoever, so these cases are problematic. One possible way out might be to assume that in fact we are dealing with a case of ‘competing grammars’ in the sense of Kroch (1994), Santorini (1992), Kroch & Taylor (1997). A small minority of speakers

168

Jack Hoeksema

might already have reinterpreted the negative particles as inherently negative elements, while the majority still treated them as polarity items which serve to reinforce negation, but cannot negate a sentence on their own.2 While this hypothesis seems to be entirely conceivable, I should note that another possibility might also be considered. Perhaps we should view ne-drop in Anglo-Norman as a syntactic variable that is not only sensitive to regional variation, or sentence type (for the latter, cf. e.g. Burridge 1993, or Hoeksema 1997 on Middle Dutch ne-drop) but also to the nature of the negative partner of ne. In particular, it might be that n-words are more prone to refrain from negative concord than negative particles, for reasons that would have to be studied further. In order to test this possibility, I decided to return to the Middle Dutch data on which Hoeksema (1997) was based, to see if anything along those lines might be detected. In that paper, I did not look at any di¤erences between negation with the negative adverb niet and negation by means of n-words, but the original database was still available (actually a slightly larger version which includes some texts I added after the publication of the 1997 paper).3 For the fourteenth century, the database yielded the following pattern of ne-drop and ne-retention: Table 1. Middle Dutch (14th century data only) þne

ne

% ne

Total

Niet (regular negation)

368

26

93,4

394

N-word

199

26

88,4

225

Total

567

52

91,6

619

2. Another option one might pursue is the possibility that n-words and negative particles are actually showing the same change, but with a time delay for negative particles. When two logistic curves are fitted for both types of sentence, rather than the straight lines of Ingham’s Figure 2, it might be that we see similar slopes, hence similar rates of change, but with di¤erent intercepts, as argued for the rise of DO-support in various contexts in Kroch (1989). Recent work by Kallel (2007) on the replacement of not þ n-word by not þ any in fifteenth and sixteenth century English however suggests that the constant rate e¤ect need not hold in all cases a single change taking place in a variety of contexts, although a weaker version might be operative. More empirical work needs to be done to establish the precise conditions under which constant rate e¤ects are to be expected. 3. The Middle Dutch material is a mix of prose texts and poetry, with a predominance of prose, both from artes sources (medical and theological texts) and legal documents.

Looking at Middle English through the mirror of Anglo-Norman

169

Comparing this with Ingham’s Anglo-Norman data, we see both differences and similarities. The main di¤erence is that the Anglo-Norman material shows far greater di¤erentiation between negation with negative sentence particles such as pas, mie and nient and negation with n-words than does the Middle Dutch data. For ease of reference, I tabulate the percentages from Ingham’s Table 1 (only the totals, not split out for subperiod or type of source) and the above Middle Dutch data: Table 2. Middle Dutch versus Anglo-Norman: retention of ne Middle Dutch

Anglo-Norman

Negative particle

93,4%

97,4%

N-word

88,4%

64,9%

Nonetheless, the di¤erence among the two types of negation points in the same direction for both languages, and this is the similarity I was alluding to: there is more n-drop with n-words than with negative particles. The di¤erence between the two categories in fourteenth century Dutch is statistically significant: p ¼ 0.0357, according to a two-tailed Fisher’s exact test. One reason why the Dutch data are interesting in the context of a comparison with Anglo-Norman is the di¤erent origin of the sentential negator niet. Whereas pas clearly has its origin in a negative polarity item meaning ‘step’, that is to say, a polarity item of the category of minimizers (cf. Bolinger 1972; Israel 2001 for discussion of this category), niet, like English not or German nicht, takes its origin a negative quantifier meaning nothing, involving an indefinite with an incorporated negation, hence something that is inherently negative from the start. For such words, the main change they need to undergo is the change from indefinite pronoun (DP) to negative adverb, not the change from polarity item to inherently negative particle. Consequently the di¤erence we found in Middle Dutch between ne-drop with niet and ne-drop with n-words cannot be due to the di¤erence between a polarity item (requiring a triggering element) and an inherently negative n-word. Presumably, the ultimate source of the di¤erence, whatever be its true nature, is not a grammatical one, since the di¤erence is not absolute, but merely a statistically significant tendency. It might be interesting to see whether the group of n-words is homogenous, or whether further variation may be found in it (cf. Ingham’s finding that in Anglo-Norman nul is particularly prone to drop its associate ne),

170

Jack Hoeksema

but for the present, my data is not su‰ciently large to investigate this matter in detail. The Middle English data presented by Ingham from his (2005) paper show a small di¤erence between the frequency of ne-drop in sentences with n-words and not that points in the same direction as the Dutch data, but at a rather higher level of ne-drop.

3. A potentially useful test There is a potentially useful test by means of which one may try to support the thesis that ne-drop in Anglo-Norman shows a di¤erentiation between n-words and negative particles. I am referring to the old almostabsolutely-test, originally due to Laurence Horn (cf. Horn 1972), adopted by many linguists (e.g. Hoeksema 1983; Zanuttini 1991) but since abandoned by Horn as an argument for distinguishing free choice and polarity sensitive any (cf. Horn 2000). The test is controversial, but in my opinion, it may still serve some purposes, and what makes it potentially interesting in the present context is the fact that it has never been used, to the best of my knowledge, in a diachronic context. In modern English, there is a clear di¤erence between the various quantifiers when it comes to modification by so-called approximative adverbs such as almost, virtually, nearly. The following paradigm will serve to make the general point: (1) a. b. c. d.

Almost everybody applied. Almost nobody applied. *Almost somebody applied. Almost anybody could apply.

The universal quantifiers every and no are fine in combination with almost, as is free choice any. The existential quantifier some on the other hand does not appear in combination with almost. A similar pattern is displayed by the modifier absolutely: (2) a. b. c. d.

Absolutely everybody applied. Absolutely nobody applied. *Absolutely somebody applied. Absolutely anybody could apply.

Looking at Middle English through the mirror of Anglo-Norman

171

Negative particles may likewise be modified: (3) He is almost not to be found. (4) He is absolutely not pleased. However, reinforcing particles typically shun away from these modifiers, perhaps, if Horn’s (2000) suggestion is correct, because almost is a positive polarity item, or, alternatively, because almost creates an intervention e¤ect of the kind first discussed in Linebarger (1980), if we follow the lead of Penka (2006) in this matter. To be sure, combinations such as the following can be found in the Internet, but they seem odd: (5) In other words none of that liquidity is getting out in the economy, almost whatsoever!!! (http://bizzxceleration.llinlithgow.com/) (6) she doesn’t look like her almost at all, something in the eyes but otherwise there’s no connection. (http://www.britneyspears.com/2008/11/thats-not-britney.php) Given these occasional hiccups, any test along these lines would need to have a statistical character. Example (6), for example, was found by googling the string ‘‘like her almost at all’’ which occurred once on the Internet. For the string ‘‘like her at all’’ over 5 million hits were found. If modification were freely allowed, we would expect a much higher number of results for ‘‘like her almost at all’’ or ‘‘like her absolutely at all’’ (which yielded zero hits). In Hoeksema (2007) I presented some corpus data from Dutch. In a collection of approximative adverbs, collected by me from various stages of modern Dutch, I found 611 combinations with universal quantifiers, 457 with negative quantifiers (niemand ‘nobody’, niets ‘nothing’ etc) and a meager 7 with a positive indefinite pronoun. Interestingly, all 7 cases involving occurrences in the scope of negation, which led me to propose that these cases involve parasitic licensing, due to a negative feature which is copied from a nearby negative element. (Indefinites denoting a precise amount or stretch of time, like een uur ‘an hour’ or een kilo ‘a kilo’ were not included in the comparison, since these freely allow modification by approximators (Horn 2000).) In view of the fact that the 2007 paper was based on data from 2004, and a substantial amount of data had been collected since then, I decided to revisit the approximators. The results are presented in Table 3. I added up the results for the approximators

172

Jack Hoeksema

bijkans, bijna, haast, nagenoeg, vrijwel, zo goed als, zowat ‘almost, nearly’ (a slightly larger, and more complete collection than the set used in Hoeksema 2007). Table 3. Approximators, Quantifiers and Negation in modern Dutch Type of Element Modified Universal

N 1275

N-word

610

Niet ‘not’

190

Indefinite

8

As we see, the stark asymmetry between positive indefinites on the one hand, and universals, n-words and negation on the other, is even more striking in the current set of data. Assuming that this type of skewed distribution is also found in French (and in Anglo-Norman), we might be able to use a rise in the frequency of approximators as signaling a shift in meaning from polarity item to negative quantifier. A collection of 6631 occurrences of polarity sensitive and free choice any (mostly collected from the Internet) likewise shows strong e¤ects when we look at the cases modified by approximators (I looked at almost, just about and nearly). Table 4 presents on overview: Table 4. FC and NPI any in various contexts Context/use

All cases

With approximator

Free choice

2015

49

Negation

2085

–

Question

780

–

Without-clause

160

1

Negative predicate

449

–

Comparative

293

6

Other negative context

849

–

6631

56

Totals

Looking at Middle English through the mirror of Anglo-Norman

173

The approximators clearly distinguish free choice from polarity sensitive any. An interesting special case is provided by comparatives. While comparatives are a classical environment for polarity items, I have argued in Hoeksema (1983) that free choice any may also show up in them. This is clearly shown by the possibility of certain collocations in comparatives which are typical of free choice any, such as any old (Carlson 1981): (7) No answer is better than any old answer. If we list the 6 cases listed under comparative as free choice cases, we have a very strong result. 55 out of 56 cases of any with an approximator are free choice, and 1 only involves negative polarity any. When I inspected the one remaining case, I concluded that even that one might be regarded as free choice (italics are mine): (8) Have you noticed that without any real intent on your part, in the middle of almost any performance, a shady little thought may creep in from the wings and attract your attention? Hence I see no reason to reject the approximator-test in a corpus setting, where we can find significant, if not always absolute, di¤erences between universal and existential quantifiers. For Middle English, I am rather pessimistic about the usefulness of such a test, given the paucity of relevant material. If the OED is anything to go by, the use of almost or absolutely with negative quantifiers does not seem to be attested before the sixteenth century.4 The examples in (7) and (8) were lifted from the OED (s.v. almost and absolutely, respectively) as the earliest attestations of the combinations in question: (9)

Bycause they were so great a company, almoost nothynge helde agaynst theym. (1523 LD. BERNERS Froissart I. ccxv. 270)

4. There might still be other ways to find negative quantifiers by looking for telltale modifiers. In Hoeksema (2009) I noted that the modifier sweet in sweet nothing transfers to sweet fuck-all, sweet diddly squat and the like, indicating that these expressions have moved from polarity item to inherently negative quantifier. (Note that sweet anything is not an idiomatic combination in the way that sweet nothing is.) The main problem with such tests is that they tend to require huge corpora, and for Middle English or Anglo-Norman the available material is not very large. In addition, the Anglo-Norman texts are o‰cial documents, with limited stylistic options. This too may stand in the way of enlisting collocations as tests.

174

Jack Hoeksema

(10) There is absolutely no pure and cleane oblation besides the sacrifice of Christes body and blood. (1565 N. SANDERS Supper of Our Lord II. iv. f. 54) For Middle Dutch, much the same is true. Perhaps our medieval ancestors did not deem it necessary to modify universal statements by hedges like almost because they had a pragmatic understanding that a few exceptions were permitted. A great deal more work would have to be done in historical pragmatics to figure out the reasons for this general lack of approximative expressions in universal statements in Middle English and Middle Dutch. 4. Conclusions The Anglo-Norman material from the Parliamentary Rolls and the Yearbooks shows an increasing amount of ne-drop, especially in sentences where the negative element is an n-word, for the period 1230–1380, a period for which good prose sources of Middle English are lacking. Ingham (this volume) interprets this as evidence that Middle English in this period is beginning to shed its system of negative concord, and that Anglo-Norman, with its structurally rather similar system of negation and negative concord, is reflecting this ongoing change. This conclusion is the more compelling since continental French is not showing comparable signs of change in the same period. Ingham takes this to be evidence that Jespersen’s (1917) theory, namely that ne-drop is mainly driven by phonological erosion, cannot be correct here, since a purely phonological account would not di¤erentiate between n-words and negative by negative particles such as pas. This point is an important one, and well taken.5 As for the suggestion, put forward by Ingham, that the di¤erence in the rate of ne-drop between n-words and negative particles shows that n-words had been reanalyzed (under the influence of Middle English) as negative quantifiers, whereas negative particles were still polarity items at the time, I noted in the above pages that this need not be the only possible 5. The reader may have noted that I have not discussed Ingham’s discussion of negative inversion. The reason is that I have nothing useful to add to his discussion of the matter. The idea of an optional NegP (cf. Ingham 2007) is interesting, but I have grave doubts about the feasibility of NegP analyses a` la Haegeman (1995) in general (cf. Hoeksema 1997). The great variety of NegP analyses is in itself a sign that something is amiss.

Looking at Middle English through the mirror of Anglo-Norman

175

explanation. Fourteenth century Middle Dutch also shows di¤erent rates of ne-drop for n-words and for negation by the negative adverb niet, yet for Middle Dutch we must assume that both n-words and niet were inherently negative. As a potential source of independent evidence for Ingham’s theory, I have proposed to use the almost/absolutely-test from Horn (1972). It remains to be seen, though, whether the Anglo-Norman material contains a su‰cient number of occurrences of these adverbs to elucidate the matter more fully.

References Bolinger, Dwight 1972 Degree words. The Hague: Mouton. Burridge, Kate 1993 Syntactic change in Germanic. Amsterdam/Philadelphia: Benjamins. Carlson, Gregory N. 1981 The distribution of free choice any. Proceedings of the Seventeenth Regional Meeting of the Chicago Linguistic Society 17: 8–23. Chicago: Chicago Linguistic Society. Giannakidou, Anastasia 2000 Negative. . .concord? Natural Language and Linguistic Theory 18: 457–523. Haegeman, Liliane 1995 The syntax of negation. Cambridge: Cambridge University Press. Hoeksema, Jack 1983 Negative polarity and the comparative. Natural Language and Linguistic Theory 1 (3): 403–434. Hoeksema, Jack 1997 Negation and negative concord in Middle Dutch. In: Danielle Forget, Paul Hirschbu¨hler, France Martineau & Maria-Luisa Rivero (eds.), Negation and polarity: Syntax and semantics, 139– 158. Amsterdam: Benjamins. Hoeksema, Jack 2007 Parasitic licensing of negative polarity items. Journal of Comparative Germanic Linguistics 10: 163–182. Hoeksema, Jack 2009 Jespersen recycled. In: Elly van Gelderen (ed.), Cyclical change, 15–34. Amsterdam: Benjamins. Horn, Laurence R. 1972 On the semantic properties of logical operators in English. Ph.D. dissertation, UCLA.

176

Jack Hoeksema

Horn, Laurence R. 2000 Pick a theory, not just any theory. In: Laurence R. Horn & Y. Kato (eds.), Negation and Polarity: Syntactic and Semantic Perspectives, 147–192. Oxford: Oxford University Press. Ingham, Richard 2005 The loss of Neg V ! C in Middle English. Linguistische Berichte 202: 171–206. Ingham, Richard 2007 NegP and negated constituent movement in the history of English. Transactions of the Philological Society 105 (3): 365– 397. Israel, Michael 2001 Minimizers, maximizers and the rhetoric of scalar reasoning. Journal of Semantics 18 (4): 297–331. Jespersen, Otto 1917 Negation in English and other languages. Copenhagen: A.F. Høst. Kallel, Amel 2007 The loss of negative concord in English: Internal factors. Language Variation and Change 19: 27–49. Kroch, Anthony S. 1989 Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Kroch, Anthony S. 1994 Morphosyntactic variation. In: Katharine Beals (ed.), Proceedings of the Thirtieth Annual Meeting of the Chicago Linguistic Society, Vol. 2: 180–201. Chicago: Chicago Linguistic Society. Kroch, Anthony S. & Ann Taylor 1997 Verb movement in Old and Middle English: Dialect variation and language contact. In: Ans van Kemenade & Nigel Vincent (eds.), Parameters of morphosyntactic change, 297–325. Cambridge: Cambridge University Press. Labov, William 1972 Negative attraction and negative concord in English grammar. Language 48, 773–818. Ladusaw, William 1992 Expressing negation. In: Chris Barker & David Dowty (eds.), Proceedings of the Second Conference on Semantics and Linguistic Theory, 237–259. Columbus: The Ohio State University. Linebarger, Marcia 1980 The grammar of negative polarity. Ph.D. dissertation, MIT. Penka, Doris 2006 Almost there: The meaning of almost. In: Christian Ebert & Cornelia Endriss (eds.), Proceedings of Sinn und Bedeutung 10, 275–286. ZAS Papers in Linguistics, Nr. 44, Berlin.

Looking at Middle English through the mirror of Anglo-Norman

177

Penka, Doris 2007 Negative indefinites. Ph.D. dissertation, University of Tu¨bingen. Santorini, Beatrice 1992 Variation and change in Yiddish subordinate clause word order. Natural Language and Linguistic Theory 10: 595–640. de Swart, Henrie¨tte 2010 Expression and interpretation of negation. Dordrecht: Springer. Van den Berg, Evert 1991 Over de syntaxis van Middelnederlandse rijmteksten. Tabu 21 (2): 67–75. Weerman, Fred 1989 The V2 conspiracy: A synchronic and a diachronic analysis. Dordrecht: Foris. Zanuttini, Ra¤aella 1991 Syntactic properties of sentential negation. A comparative study of Romance languages. Ph.D. dissertation, University of Pennsylvania. Zeijlstra, Hedde 2004 Sentential negation and negative concord. Ph.D. dissertation, University of Amsterdam.

Ne-absence in declarative and yes/no interrogative contexts: Some patterns of change France Martineau 1. Introduction The Jespersen cycle, which assumes a historical weakening and then a reinforcement process to explain the evolution of a one-term negation to a two-term negation and then to a one-term negation again (for instance, in French, from ne to ne . . . pas to pas), has been the subject of many discussions in recent years. For instance, it has been argued that the development from emphatic negation to plain negation is not related to the weakening process a¤ecting the head of the negation. This paper focuses on negation in French, examining the absence of ne in declarative and interrogative contexts, from the seventeenth century to the nineteenth century and the relation between its absence and the second term of the negation. (1) Declarative context: Mais c’est pas nan plus notre a¤aire ‘But nor is it our business’

(1764, Sarcellades)

(2) Interrogative context: Mais suis-je pas bien fat de vouloir raisonner? ‘But am I not conceited for wanting to argue? (Molie`re, 1660, Sganarelle, scene 1) In this paper, I will show the following: a) Ne-absence, as a surface phenomenon in declarative and interrogative contexts is triggered by di¤erent syntactic/phonetic factors; b) in declarative contexts, ne-deletion is phonetically triggered from the eighteenth century onwards; c) in interrogative contexts, ne is absent because pas/point/mie may behave as polarity items until the eighteenth century. They need to be licensed by an operator, especially by verb movement in C.

180

France Martineau

In the present study, I carry out the first systematic quantitative investigation of the ne-absence phenomenon, in both declarative and interrogative contexts, based on a large sociolinguistic corpus of literary and non-literary data. Ne-absence has attracted much attention, but few studies have included a broad diachronic investigation based on corpora. Moreover, apart from Martineau & Vinet (2005), no study has, to my knowledge, systematically compared ne-absence in declarative and interrogative contexts. Finally, apart from Martineau & Mougeon (2003) on ne-absence in declarative contexts, no diachronic studies have examined the influence of social stratification. As most literary texts tend to reflect normative usage, it is important for analyses of vernacular features such as the absence of ne to rely on corpora with sociolinguistic stratification: corpora containing either comedies whose authors attempted to give a representation of spoken French through characters from di¤erent social classes, or non-literary texts such as family letters written by people from di¤erent social classes, in particular by less educated people. The quantitative analyses presented in this chapter are based on two main corpora: Corpus de the´aˆtre et de textes parodiques du 17 e au 20 e sie`cles (Martineau & Mougeon 2002). Since 2007, this corpus was developed in two distinct copora: Martineau’s on French and Canadian plays and mazarinades and Mougeon’s on French plays, and Corpus de franc¸ais familier ancien (Martineau 1995; see Martineau 2009a for a description). Documents in both corpora are structured according to diatopic and diastratic parameters. The first corpus, Corpus de the´aˆtre et de dialogues parodiques du 17 e au e 20 sie`cles includes more than 300 comedies and mazarinades (pamphlets) from France, from the seventeenth to the twentieth century. All the characters are classified in terms of social class membership. The second corpus, Corpus de franc¸ais familier ancien, includes more than 15,000 private family letters from France, Canada and Louisiana from the seventeenth century to the nineteenth century, written by speakers from di¤erent social classes. Unlike literary texts, family letters are not a representation of speech but a direct reflection of the use of language, albeit through a written medium. Because they reflect written usage, they are often put aside as valuable sources for diachronic analyses. However, if the documents are selected according to sociolinguistics principles, and are representative of di¤erent social classes, they could reveal patterns otherwise concealed by more stereotyped texts, as Section 3 shows. This article is organized as follows: the first section deals with the Jespersen cycle, a hypothesis to account for the patterns of change of neabsence and the research questions it evokes in relation to the present

Ne-absence in declarative and yes/no interrogative contexts

181

study; the second and third sections discuss ne-absence in declarative and interrogative contexts, respectively. 2. The Jespersen cycle According to the Jespersen cycle, negation evolves from one-term negation (the head alone), to two-term negation (the head and an indefinite NP or an adverb), to one-term negation again (only the second part). According to Jespersen (1917: 4),1 ‘‘the original negative adverb is first weakened, then found insu‰cient and therefore strengthened, generally through some additional word, and this in turn may be left as the negative proper and may then in the course of time be subject to the same development as the original word.’’ The evolution would have followed the three following steps in French: (3) The Jespersen cycle: Step 1: a weakly stressed negative marker becomes a clitic: non ! ne; Step 2: it is then strengthened by another element: ne ! ne . . . pas/point/mie; Step 3: this process of weakening continues and ne begins to disappear. The second element then comes to be interpreted as the negative marker: ne . . . pas ! pas. The Jespersen cycle makes strong predictions with respect to the concomitance of the changes and the correlations between them. a) It presumes that the two terms (ne/pas), which are semantically and syntactically related in Modern French, share the same diachronic development. If the three steps are part of the same process, it suggests that the factors at the source of the phonetic weakening are similar throughout all three steps (steps 1–3 above); namely, that the weakening of the head (steps 1 and 3) and the reinforcement (step 2) share the same pattern of change and similar constraints. b) It also correlates the loss of nominal status of words such as pas, point or mie and the progression of their negative meaning with the weakening of ne. 1. See van der Auwera (2009) for a recent critical discussion of the Jespersen cycle; a complete bibliography is available at the Cycles of Grammaticalization website (http://www1.aston.ac.uk/lss/research/research-projects/cycles-ofgrammaticalization/).

182 c)

France Martineau

Finally, one may extrapolate from the Jespersen cycle that ne-absence is a phenomenon which is found across the board in declarative and interrogative contexts alike.

However, Kiparsky & Condoravdi (2006) for Greek and Lucas (2007) for Arabic and Berber have argued that the weakening and the reinforcement are not part of the same change. According to the former authors (p. 4–5), ‘‘reinforcement of negation by a postverbal indefinite (the ‘strengthening’) is not a response to the phonetic weakening of the head. Instead, we will follow more recent analyses of Jespersen’s cycle in seeking the driving force of the cycle in pragmatics and semantics.’’ In Martineau (2009b), I express doubts about the hypothesis that ne-loss in French is part of the same process that induces the reinforcement of negation. Indeed, I argued that the rise of pas is related to a semantic/pragmatic reanalysis, from original nominal meaning, to polarity item, to negative element. As two-term negation progressed (compared to one-term negation), grammaticalization took place; as a result, the second term lost its emphatic contrast, and a restriction on the choice of element functioning as the second term in negation emerged (in Old French, many words denoting small quantities, pas, point, mie, goutte, etc., could be used as the second term; Classical French was more or less restricted to pas and point; in the nineteenth century, pas was the preferred second term; see Price 1971, Kawaguchi 2009; Martineau 2009c). Moreover, if weakening and reinforcement were closely related, the contexts favouring the use of two-term negation would also be the ones favouring the absence of ne. Such correlated constraints have not as yet been reported. In the present article, I provide additional evidence that the weakening of ne and its reinforcement with a second term are not closely related, based on the patterns of change of absence of ne in declarative and interrogative contexts. In declarative contexts, I show that the absence of ne is unrelated to the rise of ne . . . pas, while in interrogative contexts, the absence of ne is related to word order and polarity. 3. Analysis of ne-absence in declarative contexts Concerning the step from non to ne, Kawaguchi (2009: 195) states: ‘‘il est hors de doute que la particule ne´gative ne telle qu’on l’utilise aujourd’hui existait de´ja` depuis la premie`re moitie´ du Xe sie`cle’’ [‘‘without a doubt, the negative particle ne as we use it today has existed since the first half of the tenth century’’, my translation, FM]. As regards the development from one-term negation to two-term negation, as Price (1971: 252) showed

Ne-absence in declarative and yes/no interrogative contexts

183

for Old and Middle French that ‘‘[ p]as comes from Latin passum ‘a step’ and so was perhaps first used with verbs of motion (i.e. Je ne marche pas perhaps meant ‘I don’t walk a single step’) and mie comes from mica ‘a crumb’, but by the time these words are first attested as negative particles at the beginning of the twelfth century, they had already completely lost their original meanings and are used indi¤erently with any verb: e.g. Altrement ne m’amerat il mie ‘Otherwise he will not love me’, blet n’i poet pas creistre ‘Wheat cannot grow there’ (both in the Roland).’’ The time period when absence of ne started to spread is thus important to understand the potential relation between the rise of two-term negation (reported to occur in the twelfth century) and absence of ne (step 3). Moreover, the linguistic and social conditionings favouring ne-absence may provide evidence for the phonetic weakening process. Considering that the phenomenon of ne-absence is almost categorical in vernacular Modern French (Ashby 2001 for Continental French; Sanko¤ & Vincent 1977 for Canadian French), one might expect it to have been well established in the past as well. Indeed, occurrences of ne-absence can be found, for instance, in the seventeenth century – in Heroard’s diary of the daily life of young king Louis XIII (Foisil 1989) as well as in eighteenth century letters. R R (4) Et quatre heure aprait mon ieur delongueule Etarive´, il eporte pabien ‘Four hours after Mr. de Longueil arrived, he was not feeling well’ (1755, Corpus de franc¸ais familier ancien) Based on that type of evidence, some researchers (Blanche-Benveniste & Jeanjean 1986; Dufter & Stark 2007; Haussman 1992) have gone a step further and concluded that ne-absence was a common phenomenon in Classical French, and perhaps even earlier. But do such attestations provide enough evidence to conclude that they were instances of everyday use? A historical sociolinguistic approach, based on large corpora, may help answer this question. Table 1 presents Martineau’s and Mougeon’s (2003) results for the frequency of ne-absence in dialogic texts (comedies and mazarinades), in Continental French, from the seventeenth to the nineteenth century. From Table 1, we may conclude that ne-absence in declarative contexts is a recent phenomenon. Ne-absence was exceptional in Classical French, occurring in less than 1% of cases in the seventeenth century, and in less than 2% in the eighteenth century, in all social classes. From the nineteenth century on, ne-absence increased considerably in lower-class speech, rising up to 24%.

184

France Martineau

Table 1. Frequency of ne-absence in dialogic texts from the 17th to the 19th c.2 17th c.

18th c.

19th c.

Low

0.2% (2/767)

2.8% (22/781)

38.5% (191/496)

Middle and High

0% (0/110)

0% (0/189)

0.6% (2/307)

Total

0.2% (2/877)

2.7% (22/970)

24% (193/803)

X2 for 17th–19th c.: 307, df ¼ 2, p < 0.000

Given the results in Table 1, how do we explain the high frequency of ne-absence in another type of document from the seventeenth century, i.e. Heroard’s diary of the speech of young king Louis XIII, as shown in Table 2 (from Dufter’s and Stark’s 2007 results)? Heroard was the physician in charge of the young Louis XIII; during the years he took care of the child, he wrote a well-documented diary of what his pupil said and did. Dufter’s and Stark’s results clearly show a high frequency of neabsence in pronominal subject contexts (ex.: Il ne veut pas ‘He does not want to’).3 Moreover, the nominal subject context (ex.: Le roi ne veut pas ‘The King does not want to’), which generally shows a lower rate of absence of ne than the pronominal subject context, and which is therefore linked to the last step in the evolution of the phenomenon, also shows substantial ne-absence. The discrepancy between Martineau and Mougeon’s results and Dufter’s and Stark’s (2007) is so large that it suggests some interference caused by the type of text. Both dialogic texts and Heroard’s diary are indirect sources of Classical French speech; neither type of source was intended to provide

2. Data are from Martineau & Mougeon (2003), except for seventeenth and eighteenth century middle and high social classes. For the seventeenth and eighteenth century, Martineau and Mougeon examined only lower-class characters in dialogic plays, since lower-class speakers are the locus of change for ne-absence. For the present paper, I examined additional data for middle and upper social classes for these centuries, in order to ease the comparison with interrogative structures. 3. I recalculated the percentages of Dufter & Stark (2007) to examine the rate of absence of ne in each context (not on the total number of occurrences).

Ne-absence in declarative and yes/no interrogative contexts

185

Table 2. Frequency of ne-absence in Heroard’s diary Pronominal subject

Nominal subject

Relative pronoun

PAS

Others

PAS

Others

PAS

Others

4–5 years

62.1% (277/446)

48.9% (137/280)

76% (19/25)

22.2% (2/9)

60% (9/15)

20% (1/5)

5–9 years

79.5% (202/254)

68.8% (95/138)

60% (9/15)

75% (3/4)

37.5% (3/8)

0% (0/6)

a faithful transcription of spoken French. Dialogic texts are a representation of vernacular speech by authors from the elite, who made a selection among vernacular features for stylistic purposes. Their selection depended upon the degree of saliency of features and on how a particular feature might have served to distinguish characters from one another. If the absence of ne was not socially salient enough at that time, authors may have chosen not to use it. On the other hand, not only is Louis XIII’s speech representative of seventeenth century speech, it also reflects features of the speech of a young speaker in the process of language acquisition. Studies have shown that young speakers acquiring Modern French may show a high frequency of ne-absence (see for instance Pohl 1972; see Martineau & Mougeon 2003 for discussion). As Dufter & Stark argue, a high frequency of ne-absence in Louis XIII’s speech cannot only be related to a particular feature of a certain stage of acquisition, because ne-absence increases for both age categories (4–5 years and 5–9 years), contrary to what would have been expected if the frequency of the phenomenon were due solely to acquisition problems. However, the influence of acquisition processes should not be underestimated, since linguistic features typical of young speakers are also present in the document. One way to resolve the methodological issue raised by the results from seventeenth century and eighteenth century dialogic texts and those from Heroard would be to have more direct sources for the Classical French usage. That type of comparison is now possible, using a unique corpus of Classical vernacular French, Martineau’s Corpus de franc¸ais familier ancien. I have therefore conducted an analysis of ne-absence, taking into consideration only one type of text, namely family letters, and only the social bracket and time frame where ne-absence may have started to

186

France Martineau

spread: lower-class speakers from the mid-eighteenth century to the end of the nineteenth century, since these registers may evidence earlier ne absence, which was not found in the more formal texts generally considered by language historians.4 The results are shown in Table 3. Table 3. Frequency of ne-absence in French family letters, produced by lower class speakers, from the mid-18th c. to the end of the 19th c. Mid-18th c.

Beginning of 19th c.

Mid-19th c.

End of 19th c.

PAS

20.7% (6/29)

30.5% (18/59)

40.2% (103/256)

55.4% (133/240)

POINT

25% (4/16)

27% (17/63)

41.4% (17/41)

–

Others

(1/3)

31.2% (10/32)

47.6% (40/84)

77.9% (60/77)

Total

20.8% (10/48)

29.2% (45/154)

41.9% (160/381)

60.8% (193/317)

Table 3 shows that, contrary to what Martineau & Mougeon found in dialogic texts, ne-absence was already well established in the mideighteenth century and there is a slow but steady increase from the mideighteenth century to the end of the nineteenth century; although this increase is statistically non significant between the mid-eighteenth century and the beginning of nineteenth century while it is significant for nineteenth century, the frequency in the mid eighteenth century shows that absence of ne was more than an idiosyncratic feature. Regarding the time period, the steady increase from the eighteenth century to the end of the nineteenth century points to the fact that the absence of ne started to spread in Classical French, after the emergence of the two-term negation in twelfth century; thus, it is di‰cult to argue that the weakening of ne is the cause of the reinforcement/emphasis use of the negative adverb. The results in Table 3, as well as results in Martineau & Mougeon (2003), also show that the change is initiated by lower class speakers, often the agents of phonetic reduction. Two factors may have triggered that reduction. First, schwa started to become phonetically unstable. Very

4. All the letters are by Canadian French writers. The same type of research on absence of ne could be conducted on Continental French.

Ne-absence in declarative and yes/no interrogative contexts

187

early in the history of French, schwa was elided before vowels and consonants. Enclisis with negation, leading to the deletion of schwa, also developed early (Morin 1974). Another source for the deletion of schwa is phonetic syncope, which is attested in the fifteenth century (Morin 2008), and is also mentioned by grammarians such as Oudin in the seventeenth century. This means that in all probability, the particle ne was phonetically reduced to the single consonant [n] in the seventeenth century, until its complete reduction. Moreover, in Modern French, it is argued that the subject pronoun is an a‰x to the verb (Ashby 2001, Harris 1978). If we consider that the subject pronoun was already an a‰x to the verb in the eighteenth century, we may suppose that the particle ne, being sandwiched between the subject a‰x and the verb, was in a weak position, and was deleted. Indeed, most of the occurrences of ne-absence found in the mid-eighteenth century occur in contexts where the subject is a clitic pronoun (5a); less frequently, occurrences appear after a nominal subject, as in (5b). (5) a.

quil en reviendra pas ‘that he will not come back from there’ (1756, Corpus de franc¸ais familier ancien)

b.

Nos ame´riquin on pas attire´ Boucoups de grace ‘Our Americans did not attract many thanks’ (1775, Corpus de franc¸ais familier ancien)

Turning back to the Jespersen cycle in (3), this means that the weakening process would have proceeded directly from Step 1 (non ! ne) to Step 3 (ne ! Ø). In other words, I suggest that deletion of ne in declarative structures would not have induced changes in the syntactic head-adjunct relation. Ne remains the head and in Modern French, although it is often deleted in vernacular spoken French, it is still used in more formal speech. Moreover, in written French (Coveney 2002) and in spoken French, the clitic pronoun/nominal constraints mentioned above are still working. 4. Analysis of ne-absence in interrogative contexts I will now turn to absence of ne in interrogative contexts and I will show that the pattern of change in this context is di¤erent from the one in declarative contexts, which therefore implies another process at work. In interrogative structures, I argue that polarity items trigger the absence

188

France Martineau

of ne and structural conditions should be met for polarity items to license ne-less pas/point/rien. 4.1. Declarative/Interrogative structures To investigate ne-absence in interrogative contexts, I examined occurrences of direct yes/no negative interrogative contexts in dialogic texts, from the seventeenth to the nineteenth century,5 including presuppositional rhetorical questions if they showed an element also used in a negative context ( pas, point, mie, aucun, rien, etc.). I restricted the data to direct yes/no-questions; wh-questions and indirect questions as in (6) and (7) were not examined. (6) Wh-questions Sa livre´e ! Pourquoi ne couches-tu pas avec ? ‘His livery! Why aren’t you sleeping with it?’ (Labiche, 1859, Le Baron de Fourchevif, Act 1, scene II) (7) Indirect questions Je ne sais si dans l’aˆme il ne sentiroit point une secre`te flamme. ‘I don’t know if in his soul he will not have a secret feeling’ (Molie`re, 1656, Le De´pit amoureux, Act 2, scene VI) The seventeenth century is the starting point for this analysis for two reasons: (1) before that period, yes/no questions are mostly VS and it would be di‰cult to evaluate the influence of word order (VS or SV) and (2) dialogic texts with representation of French vernacular morphosyntax are scarce before the seventeenth century. I also restricted the study to literary texts, since non-literary texts, such as letters or diaries, do not usually include occurrences of direct interrogation; when a question is asked, it usually appears in an indirect form. Ne-absence in interrogative contexts was a common phenomenon in Medieval French and in the sixteenth century. It occurs with pas, point, mie and indefinite NP such as pas or rien, as shown in (8). (8) a.

Que te semble de ma nouvelle Espousee ? Est elle pas belle et honneste sou‰sanment ? ‘What do you think of my new wife? Isn’t she nice and honest enough?’ (Griseldis, v. 2430–2; Marchello-Nizia 1997: 306)

5. 17th c. (9 plays); 18th c. (9 plays); 19th c. (21 plays)

Ne-absence in declarative and yes/no interrogative contexts

b.

189

Vous a il encores rien fait? ‘Has he still not done anything to you?’ (Cent Nouvelles Nouvelles Anonymes, p. 133; Martineau 2009d)

Table 4 shows the rate of ne-absence in declarative and interrogative structures. Results for ne-absence in declarative contexts are repeated from Table 1. First, we observe a steady increase in ne-absence in declarative structures (recall that this increase is statistically highly significant). As we have seen, ne-absence in declarative structures is associated with phonological contexts, a result I have linked to Jespersen’s weakening process hypothesis (from ne to ø). Second, if we turn to interrogative structures, results show a frequency of 8.8% for ne-absence in the seventeenth century, with a decrease in the eighteenth century followed by a significant increase of frequency in the nineteenth century. The di¤erences between the three periods are statistically significant (X 2 ¼ 34.2, df ¼ 2, p < 0.001) and point to an underlying major change. Third, the comparison between interrogative and declarative structures shows a strong statistical di¤erence between both structures in the seventeenth century (X 2 ¼ 60.6, df ¼ 1 p < 0.001), a di¤erence that disappears in the eighteenth and nineteenth centuries. Again, these di¤erences suggest that important changes took place after the seventeenth century. Fourth, in interrogative contexts, ne-absence is not restricted to lower-class speech in the seventeenth century, although there is a minor di¤erence between lower or middle and high classes. However, we see that in the eighteenth century, there is a drop in the frequency of ne-absence for all classes and by the nineteenth century, its social stratification mirrors the one in declarative contexts. Table 4. Distribution of ne-absence in declarative and interrogative contexts Declarative

Interrogative

Total

Low

Middle and High

Total

Low

Middle and High

17th c.

0.2% (2/877)

0.2% (2/767)

0% (0/110)

8.8% (16/182)

19.4% (7/36)

6.2% (9/146)

18th c.

2.7% (22/970)

2.8% (22/781)

0% (0/189)

2.7% (3/111)

3.3% (2/61)

2% (1/50)

19th c.

24% (193/803)

38.5% (191/496)

0.6% (2/307)

25.7% (129/501)

52.2% (119/228)

3.6% (10/273)

190

France Martineau

In what follows, I will provide an explanation for the relatively high frequency of ne-absence in interrogative structures in the seventeenth century (8.8%) in comparison to its (relatively low) frequency in declarative structures (0.2%) and its significant decrease in the eighteenth century. It is important to always compare the same type of documents when we use the frequencies as an indicator of language change. For instance, the high frequency of ne-absence in letters (Table 3) cannot be compared to the frequencies in comedies and mazarinades. The seventeenth century discrepancy in the frequency of ne-absence in declarative versus interrogative contexts is in line with observations made by Price (1993, 1997) to the e¤ect that pas/point/rien, etc. are not always negative in Classical French and before. Martineau & Vinet (2005) provided quantitative results to support these observations. They gave rates of ne-absence in interrogative contexts for Medieval (15% 16/108) and sixteenth century (26% 32/124) French based on a literary corpus which showed that ne-absence in interrogative contexts was not a marginal phenomenon. However, as Martineau & Vinet show, for the same periods, ne-absence in declarative structures is almost non-existent (Medieval period: 2/1691, i.e. 0.1%; sixteenth century: 4/1238, i.e. 0.3%). To explain the discrepancy between declarative and interrogative structures, these studies argue that in interrogatives, ne-absence may be interpreted as instances of polarity e¤ects, not of negation. My results in Table 4 provide additional evidence for this discrepancy between absence of ne in interrogative and declarative contexts and, crucially, show that important changes occurred in the eighteenth century. Many authors have mentioned the fact that, up to Classical French, indefinite aucun and rien could be used as polarity items ((9), (10)) (among others, Buridant 2000; De´prez & Martineau 2004; Ingham to appear; Schnedeker & Pre´vost 2002; Price 1971, 1993, 1997). Because of their polarity status, they could co-occur with negative pas and point (11). (9) tu ne serois pas bon poe¨te si tu chantois contre les reigles de la musique; ny moy bon gouverneur de ville, si je faisois aucune chose contre les loix civiles. ‘you would not be a good poet if you sang without following the musical rules; nor would I be a good governor of the city if I did anything against the civil laws’ (Deimier, 1610, L’Acade´mie de l’Art Poe´tique, p. 491) (10) je ne veux pas vous exposer a` rien de faˆcheux ‘I dont’ want anything bad to happen to you’ (Chasles, 1713, Les Illustres franc¸oises, p. 254)

Ne-absence in declarative and yes/no interrogative contexts

191

(11) Qui n’estoit point couvert d’aucunes armes ‘who did not wear any weapon’ (Audiguier V. d’, 1615, Histoire Trage-Comique p. 423) In Classical French, rien and aucun started to lose their polarity status and came to be reanalyzed as negative. One of the consequences was the decline in the co-occurrence of negative rien or aucun with pas or point (although, as shown by Hansen 2009, this possibility is not completely ruled out in Modern Continental French). From the results in Table 4 for ne-absence in interrogative contexts, we can observe the transition from polarity items in the seventeenth century and accordingly, a relatively high frequency of ne-less interrogative structures, to reanalysis of polarity items as negative elements, with a high frequency of ne-absence by lower-class speakers in the nineteenth century; the eighteenth century is thus a period of transition. The following remark from Me´nage (1675–1676) suggests that the seventeenth century grammarians were perhaps already aware of this change: ‘‘Monsieur de Vaugelas veut qu’il soit mieux de dire, Ont-ils pas fait? Sans la ne´gative, que N’ont-ils pas fait? Avecque la ne´gative. Je ne suis pas de son avis. N’ont-ils pas fait? Me semble plus e´le´gant.’’ (cited by Kawaguchi 2008 : 206) [‘‘Mr. Vaugelas considers it better to say Ont-ils pas fait?, without the negative [ne], than to say N’ont-ils pas fait?, with the negative [ne]. I disagree. N’ont-ils pas fait? seems to me more elegant.’’ My translation, FM]. Because ne is not deleted but absent in interrogative contexts, I expect to find di¤erent triggering factors for ne-absence in interrogative structures than for ne-absence in declarative structures in the seventeenth century. In Classical French, major changes took place in interrogative word order. For the purposes of this study, I have distinguished di¤erent word order contexts, as illustrated below by nineteenth century instances in (12a–f ). In (12a), the subject clitic is postverbal but adjacent to the verb. In (12b), the preverbal NP subject is coindexed with a postverbal subject clitic. In (12c), the subject is preverbal and a question marker -ti is used;6 (12d), with the marker est-ce que, appears to be similar to (12c) with respect to the preverbal position of the subject and the use of a question marker. In 6. It is di‰cult, in the construction SV-ti when the NP subject is third person singular (e.g. Pierre vous l’avait-ti dit?), to distinguish a dislocated NP from an NP subject in its canonical position. I have assumed that the spelling -ti by the authors of dialogic plays is an indication of the use of the question marker -ti.

192

France Martineau

(12e), the subject is preverbal and the phrasal intonation indicates an interrogation. Simple inversion in (12f ) was used in Medieval French and in the sixteenth century, but no occurrences were found in my corpus (see Roberts 1993 for a discussion of simple and free inversion). (12) a.

b.

Verb-Subject Pronominal Inversion Monsieur Timole´on, je ne vais jamais au bal; mais ne sommes-nous pas voisins ? ‘Mr Timoleon, I never go to the ball; but aren’t we neighbours?’ (Dupeuty & Cormon, 1842, Paris La Nuit, Act I, tableau 2, scene III) Complex inversion Mais ma me`re n’s’opposera-t-elle pas ? ‘But wouldn’t my mother object?’ (Henrion & Moreau, 1802, Les Amours de la Halle, scene III)

c. SV-ti J’vous l’avais-t’y pas dit ? ‘Didn’t I tell you?’ (Monnier, 1862, Les Bas-fonds de la socie´te´, p. 15) d.

Est-ce que SV Est-ce que j’ t’ai pas dit ousque d’meurait ma nie`ce? ‘Didn’t I tell you where my niece was staying?’ (Dupeuty & Cormon, 1842, Paris La Nuit, Act 3, tableau 4, scene VII)

e. Intonation SV Tu n’ veux pas t’en v’nir? ‘Don’t you want to come?’ (Dupeuty & Cormon, 1842, Paris La Nuit, Act 3, tableau 4, scene VII) f.

Simple inversion Et ne m’a le me´decin point ordonne´ de re´gime? ‘And didn’t the doctor put me on a diet?’ (Cent Nouvelles Nouvelles 140. 47, from Elsig 2009: 3)

Ne-absence in declarative and yes/no interrogative contexts

193

There are several studies that have examined word order in interrogative structures (for Modern French, see Coveney 2002; Hansen 2001; see Elsig 2009 for a review of the literature on diachronic and dialectal distribution of interrogation, and on the conditioning factors; Thomas 2010 for an analysis of interrogatives in Heroard’s diary) both for Continental and for Quebec French. In Modern Continental spoken French, subject inversion, which was the default structure in Medieval French, is marginal. The Intonation SV structure is the default case, followed far behind by est-ce que questions, while SV-ti is rare. Complex inversion is said to be lost, just like simple inversion. However, apart from the present study, no systematic analysis has examined the relation between the evolution of the types of interrogative structure and the presence/absence of ne. Having shown that declarative and interrogative structure behave differently according to ne-absence or presence, and having hypothesized that the change in frequency from the seventeenth to the nineteenth century may reflect a change from polarity to negative status, I now turn to the interrogatives and the licensing conditions for polarity items. I will argue that ne-less structures were strongly linked to VS word order in interrogative structures. 4.2. Ne-absence and VS word order Let us first review the general distribution of interrogative structures. Table 5 presents the percentages of types of interrogative structure for positive yes/no interrogatives (from Medieval French to the seventeenth century) from Elsig (2009: 137) and my own results for negative yes/no interrogatives (from the seventeenth to the nineteenth century). Elsig’s results for positive interrogatives and my results for negative interrogatives show a general shift from VS word order to SV word order, a general trend in the history of French (Picoche & Marchello-Nizia 1989). If we take into consideration only my results on negative interrogatives and compare VS structures (pronominal inversion, NP complex inversion and SV-ti, the two last being sometimes ambiguous) with SV structures (intonation and est-ce que), we see a gradual shift from VS to SV structures, from the seventeenth century (71.4% of VS) and the eighteenth century (51.3%) to the nineteenth century (28.6%). The eighteenth century was the critical period for the loss of the old VS structure. Elsig’s results show that VS was the most frequent structure in yes/no interrogatives, from Medieval to seventeenth century French, with a small decline in VS structure after the sixteenth century. My own results show a

194

France Martineau

Table 5. Yes/no interrogative structures From Elsig’s results (2009:135) (positive yes/no interrogatives) (headings are renamed according to Martineau’s headings) VS Pronominal inversion

DPINV

NP Complex Inversion / SV-ti

Intonation SV

Est-ce que SV

Total

Middle French

88% (168)

1% (2)

4.2% (8)

6.8% (13)

–

191

16th c.

82.9% (266)

2.5% (8)

5.6% (18)

9% (29)

–

321

17th c. Corneille Molie`re Agre´ables Confe´rences

64.7% (494)

.1% (1)

7.7% (59)

26.2% (199)

1.3% (10)

763

Martineau’s results (negative yes/no interrogatives) VS Pronominal inversion

DPINV

NP Complex Inversion

SV-ti

Intonation SV

Est-ce que SV

17th c.

62.7% (114)

–

6% (13)

1.6% (3)

26.4% (46)

3.3% (6)

182

18th c.

45.9% (51)

–

3.6% (4)

1.8% (2)

34.2% (38)

14.5% (16)

111

19th c.

20.2% (101)

–

.4% (2)

7.6% (38)

62.6% (314)

9.2% (46)

501

similar pattern for the seventeenth century: VS is the prominent structure. The decline continues throughout the eighteenth and nineteenth centuries, at which time it was no longer the dominant structure.7 In Modern Continental French, as Coveney (2002:190) remarks: ‘‘subject clitic inversion in interrogatives has been declining in French, and, in informal spoken varieties, is now found only marginally.’’ The Intonation SV pattern is said to have emerged in the thirteenth century but was rare in Medieval French. My results show that the frequency of the Intonation SV structure rises in the eighteenth century but 7. According to Elsig (personal communication), in Quebec French at the end of the nineteenth century there is a clear tendency for negative yes/no interrogatives to be used in SV structures while positive yes/no interrogatives show more variation between the types of structures.

Ne-absence in declarative and yes/no interrogative contexts

195

it is not until the nineteenth century that it becomes the dominant structure. As Coveney (2002) and Ashby (1977) show, it is still the dominant structure in Modern Continental French. The SV-ti structure, which is said to have emerged during the sixteenth century from inverted subject clitic il with -t ‘de liaison’, is a minor interrogative structure. However, its use is relatively stable from the seventeenth to the nineteenth century, if we consider together, as Elsig did, NP Complex Inversion and SV-ti, which are often di‰cult to distinguish when the subject NP is third person singular. In Modern French, the use of SV-ti is restricted to certain dialects, such as Quebec French, in which the variant SV-tu is used. In Modern French, Complex inversion, following Berhnsted (1973) cited by Elsig (2009: 18), is claimed to be extinct in the vernacular speech. The est-ce que structure may have emerged in the course of the 15th c., or perhaps earlier. Table 5 shows that in the seventeenth century, it was a minor structure, and even in the nineteenth century, it did not compete with other structures. Results from Ashby (1977: 10%) and Coveney (2002: 20%) show that this structure remains marginal in Modern Continental French. Summarizing the results, ne-absence is initially limited to interrogative structures rather than declarative structures, whereas both declarative and interrogative structures start to show comparable behaviour in the eighteenth century (Table 4). We have also seen a steady decline in VS interrogatives, both positive and negative, and that the eighteenth century is a critical period for VS word order in interrogatives (Table 5). I will now argue that the fact that there is a parallel change in both positive and negative interrogatives suggests that the absence of ne in interrogatives is related not to clause type as such but to word order. We may now consider whether there is significant correlation between ne-less interrogatives and the way the interrogatives are expressed (VS or SV). The results are shown in Table 6. In the seventeenth century, ne-absence occurred in one major type of structure, i.e. pronominal inversion (VS, NP Complex inversion).8 Moreover, it is attested in the SV-ti structure, where the number of occurrences is very small, however. What is surprising is that, although SV structures (with intonation or with est-ce que) were in use, our data contain no instances of ne-absence in these structures. 8. We will not examine NP Complex Inversion, as the number of occurrences available is too small.

196

France Martineau

Table 6. Ne-absence according to the type of interrogative structure VS pronominal inversion

NPComplex inversion

SV-ti

Intonation SV

Est-ce que SV

Total

17th c.

11.4% (13/114)

15.4% (2/13)

33.3% (1/3)

0% (0/46)

0% (0/6)

8.8% (16/182)

18th c.

1.9% (1/51)

0% (0/4)

100% (2/2)

0% (0/38)

0% (0/16)

2.7% (3/111)

19th c.

12.8% (13/101)

0% (0/2)

89.5% (34/38)

24.2% (76/314)

13% (6/46)

25.7% (129/501)

In the eighteenth century, the overall absence of ne is very rare (only 3 occurrences) and it is di‰cult to draw a conclusion. However, we see that although VS pronominal inversion structures are often used (51 occurrences), there is only one instance of ne-absence, as against the 11.4% observed in the seventeenth century. The SV -ti context is still susceptible to ne-absence, but here the total number of occurrences is, once again, very low. As in the seventeenth century, SV structures do not exhibit ne-absence. In the nineteenth century, ne-absence is favoured in SV structures, which are the most frequently used in the nineteenth century while SV-ti also continues to strongly favour ne-absence. Interestingly, ne-absence in VS pronominal inversion shows an increase in frequency, after its decline in the eighteenth century. This means that while the frequency of VS pronominal inversion shows a steady decrease from the seventeenth century to the nineteenth century (from 62.7% to 20.2%), the frequency of ne-absence in this context is more or less the same in the nineteenth century as it was in the seventeenth century. However, as we will see below, this similarity conceals an important change. In sum, ne-absence is strongly related to three structures in the seventeenth century, namely VS pronominal inversion, complex NP inversion and SV-ti, whereas in the nineteenth century, it is related to SV-ti and SV word order, as in SV intonation and Est-ce que SV. Arguably, it is not surprising that ne-absence occurs in more frequently used structures (VS structures in the seventeenth century and SV structures in the nineteenth century). However, I will argue that word order in the seventeenth century o¤ers conditions for ne-absence that are di¤erent than those for neabsence in the nineteenth century, namely that ne-absence is not the same phenomenon in the seventeenth century as it is in the nineteenth century,

Ne-absence in declarative and yes/no interrogative contexts

197

which also explains the patterns of ne-absence frequency in VS pronominal inversion interrogatives. Evidence for this di¤erence comes from the social stratification of neabsence. As Table 7 shows, all occurrences of SV -ti interrogatives come from lower-class speakers and ne-absence is almost categorical. In other words, this structure is highly stereotyped in plays. Table 7. Social stratification from the 17th to the 19th c. in SV-ti negative interrogatives Low

Middle

High

Total

17th c.

100% (1/3)

0% (0)

0% (0)

33.3% (1/3)

18th c.

100% (2/2)

0% (0)

0% (0)

100% (2/2)

19th c.

89.5% (34/38)

0% (0)

0% (0)

89.5% (34/38)

In comparison, ne-absence in VS pronominal inversion structures shows more variation, as shown in Table 8. In the seventeenth century, ne-absence is not limited to lower-class speech, and middle- and higher-class speakers also show instances of ne-absence, although less frequently. However, by the nineteenth century, the social stratification of ne-absence has changed: it is very high for lower-class speakers but rare or absent for middle- and higher-class speakers. Table 8. Social stratification from the 17th to the 19th c. in VS pronominal inversion negative interrogatives Low

Middle

High

Total

17th c.

20% (4/20)

12.5% (5/40)

7.4% (4/54)

11.4% (13/114)

18th c.

0% (0/26)

0% (0/6)

5.3% (1/19)

1.9% (1/51)

19th c.

35.7% (10/28)

5.1% (3/59)

0% (0/14)

12.8% (13/101)

198

France Martineau

Indeed, by the nineteenth century, the social stratification in VS interrogative structures reflects what is found in SV intonation interrogative structures, as shown in Table 9. In nineteenth century SV Intonation, ne-absence is very high for lower-class speakers but rare or absent for middle- and high-class speakers. Table 9. Social stratification from the 17th to the 19th c. in SV intonation negative interrogatives Low

Middle

High

Total

17th c.

0% (5)

0% (21)

0% (20)

0% (0/46)

18th c.

0% (16)

0% (4)

0% (18)

0% (0/38)

19th c.

47.8% (69/145)

5.1% (7/138)

0% (0/31)

24.2% (76/314)

Table 10 reveals that est-ce que SV shows a pattern similar to that of SV Intonation structures. That is, the frequency of ne-absence, which was nil in the seventeenth and eighteenth century, increases in the nineteenth century, but only for lower-class speakers. Table 10. Social stratification from 17th to 19th c. in Est-ce que negative interrogatives Low

Middle

High

Total

17th c.

0% (2)

0% (1)

0% (3)

0% (0/6)

18th c.

0% (15)

0% (0)

0% (1)

0% (0/16)

19th c.

35.3% (6/17)

0% (0/24)

0% (0/5)

13% (6/46)

In other words, by the nineteenth century, ne-absence shows a similar social stratification, between SV and VS interrogative structures (as shown in Tables 8 to 10), and between declarative and interrogative structures (as shown in Table 4). Moreover, in the nineteenth century, ne-absence in the

Ne-absence in declarative and yes/no interrogative contexts

199

SV interrogative context shows the same phonetic conditioning pattern as the declarative context, namely after a subject clitic, as in (13). (13) t’a donc pas ferme´ la porte de la boutique ? ‘So you haven’t closed the door to the store?’ (Dupeuty & Cormon, 1842, Paris la Nuit, Act 3, tableau 4, scene III) As in Martineau and Mougeon (2003)’s data for declarative sentences, instances of ne-absence after a nominal subject are rare in SV interrogatives. Indeed, there is no example of ne-absence after an NP subject in SV Intonation interrogative structure and only one case was found of ne-absence after a nominal subject in Est-ce que interrogatives, as in (14). (14) Est-ce que tous les rangs sont pas confondus ? ‘Aren’t all the classes mixed up?’ (Monnier, 1861, Religion des imbe´ciles, L’E´glise, p. 133) While interrogatives progressively adopted the SV word order, they also started to fall under the same constraints on absence of ne as other declarative negative sentences. Moreover, this social stratification extended to VS pronominal inversion. Thus, what remains to be explained is the fact that VS pronominal inversion, NP complex inversion and SV-ti favoured ne-absence in the seventeenth century. They all show a relation to VS word order, a word order that gradually becomes marginal in interrogatives. I therefore argue that ne-absence in seventeenth century interrogatives is related to word order, and more generally, to conditions licensing an operator for polarity items. In interrogative contexts, such as (15), the question is oriented towards a positive answer; ne is not deleted but absent, and rien is a polarity item paraphrased by anything. If we add the particle ne, the meaning changes; the question lacks a presuppositional reading and the answer could be yes or no. (15) As-tu rien vu de plus joli ? ‘Have you ever seen anything prettier?’ (Molie`re, 1665, Dom Juan, Act 2, scene II) Why are VS pronominal inversion structure and SV-ti structure the preferred structures for a presuppositional reading? After all, we may have found absence of ne in SV with intonation or est-ce que but, as we have seen, most of the cases of absence of ne are in VS and SV-ti structures in the seventeenth and eighteenth century. However, VS and SV-ti

200

France Martineau

interrogative structures are historically related, which is not the case for intonation SV and est-ce que SV structures, and it is not surprising that some common features a¤ect both of them. Let us first examine VS pronominal inversion structures. I will assume a Q feature associated with rhetorical questions to be overtly identified, by movement of the verb to the left periphery (C). In (15), rien, as a polarity item, would be licensed by this operator. The same would happen with pas or point, which may trigger positively-biased questions, as in (16). (16) Se bat-on pas toujours quand qu’on devient Cocu ? ‘Don’t we always fight when we are cheated?’ (Bre´court, 1666, Nopce de village) Interestingly, in the seventeenth and eighteenth centuries, the presence of ne in VS interrogative structures with rien and aucun renders the reading negative, while this is not automatically the case with pas or point. With pas or point, when the question is oriented towards a positive answer, the absence of ne or its presence adds nothing to the whole meaning: it could be absent as in (16), or be present as in (17), without a¤ecting the interpretation of the sentence. (17) he´ ban Janin, ne tavas-je pas pnostigue´ stavantuze, me croizas-tu un outre foua ? ‘Janin, if I hadn’t told you about that, would you have believed me next time?’ (1649, Agre´ables conferences) In (17), I suggest that ne and pas occupy their expected position in the NegP but the checking relation bleaches the negative reading of pas, which is reinterpreted as an emphatic a‰rmative marker.9 However, if ne is used alone, without pas or point, the reading of the sentence is negative, as in (18).

9. As proposed by Vinet (2000, 2001) for -tu, as in (i). (i)

Fak la`, il (*ne) part-tu pas a` crier. ‘‘Suddenly, he starts yelling’’

In her analysis, ne cannot appear with the emphatic a‰rmative operator -tu, while in our results, ne could co-occur with pas and -ti. The diachronic relation between -ti and -tu and the presence/absence of ne remains to be studied.

Ne-absence in declarative and yes/no interrogative contexts

201

(18) Vous m’avez autrefois tant aime´e: ne puis-je, en devenant votre femme, obtenir la graˆce de mon pe`re ? ‘You loved me so much before; if I marry you, could I not obtain mercy for my father?’ (Cyrano de Bergerac, 1654, Pe´dant Joue´, Act 3, scene III) When rien and aucun lost their polarity meaning, sentences such as (15) were reanalyzed with a negative reading, triggering the internal and external constraints on absence of ne. However, we still find instances of pas or point used in presuppositional structures in the nineteenth century. (19) Parbleu ! veux-tu pas t’aller coucher, feignant ? ‘My God, won’t you go to sleep, loafer?’ (Dupeuty & Cormon, 1842, Paris la nuit, Act 3, tableau 4, scene I) Sentences such as (19) may have been ambiguous: they may be interpreted either with ne-deletion and a negative reading, a feature of lowerclass speakers in the nineteenth century, or with absence of ne and presuppositional reading. Recall that in Table 8, VS pronominal inversion is used by high and middle-class speakers (72.2% P 73/101) as well as by lower-class speakers (27.7% P 28/101) in the nineteenth century. However, as shown in Table 8, in the nineteenth century, high-class speakers avoid ne-absence in the VS interrogative structure, probably because the presuppositional reading is not the main reading. In an ambiguous context, high-class speakers, and in a certain degree, also middle-class speakers, are conservative and prefer to maintain ne. Clearly, this behaviour is triggered by the rise of ne-deletion in SV declarative and interrogative contexts, while in the seventeenth century, ne-absence would not have been misinterpreted as ne-deletion. As with VS interrogative structures with pas or point, SV-ti interrogative structures allow a presuppositional reading, as in (20). (20) he´ morgue´ Georget laisse les faize, tu nan sezas ne pi ni mieux, quinze fran sonty pas bon a` gagnie´ ? ‘My God, Georget, leave them alone, you won’t be worse or better o¤, isn’t 15 francs worth having?’ (1649, Agre´ables Confe´rences) Moreover, as with VS interrogative structures, ne may be present (21) or absent (22) with pas or point in presuppositional reading in the nineteenth century.

202

France Martineau

N’avons-t’y point dit cette fois-la`, M. Faucheux, que j’e´tions le premier peuple d’la terre ? ‘Haven’t we said this time, Mr. Faucheux, that we were the first nation on earth?’ (Monnier, 1858, Bourgeois aux champs, Les Girouettes, scene I)

(21) a.

n’aviont-t’y point dine´ a la meˆme table avec le sous-pre´fet ? ‘Haven’t we dined with the sub-prefect?’ (Monnier, 1858, Bourgeois aux champs, Les Girouettes, scene III)

b.

c. C ¸ a n’vous f ’ra-t’y point d’chagrin si je l’faisons ? ‘Won’t you be sad if we do it?’ (Monnier, 1858, Bourgeois aux champs, L’esprit des campagnes, scene IV) (22) c’e´tiont-t’y point des horreurs d’payer ce que j’payons de contribution ‘Wasn’t it horrifying expensive what you pay?’ (Monnier, Bourgeois aux champs, Les Girouettes, scene I) In the seventeenth century, the SV-ti may have allowed ne-absence for the same reasons that VS structures allowed it, namely a checking relation with an operator in C: indeed, ti originates from a third person singular and it is often di‰cult to distinguish the SV-ti structure from the NP complex inversion during this period. By the nineteenth century, the two structures are distinct and it is not clear how to interpret the absence of ne in these structures used only by lower-class speakers (see Table 7). The rise of ne-deletion may have interfered with the presuppositional reading. The avoidance of ne-deletion in SV-ti interrogative structures may have been ambiguous, between a presuppositional reading and a negative reading, at a time when ne-deletion was gaining strength. 5. Conclusion Based on an historical sociolinguistic analysis, I have shown that ne-absence follows a di¤erent path for interrogatives and declaratives up to the eighteenth century. a) In declarative structures, ne-absence is ne-deletion: it is a phonetic weakening change, from lower-class speech to higher-class speech. I have shown that it began to spread in Classical French at a time when it was not salient enough to be used as a typical feature of lower-class

Ne-absence in declarative and yes/no interrogative contexts

203

speech by playwrights, hence the fact that it could only be demonstrated on the basis of large lower-class letters corpora. b) In interrogative structures, ne-absence is related not to interrogative structures as such, but to the conditions allowing polarity items to be licensed, namely VS word order (see also Ingham, this volume for an analysis of ne-absence in Anglo-Norman). c) Polarity items like rien, aucun etc. started to disappear in the eighteenth century. As a result, the frequency of negative concord (see De´prez & Martineau 2004 and De´prez, this volume) and ne-absence in interrogative structures decreased. d) There may have been some overlapping sociolinguistic constraints between the increase of ne-deletion in SV structures and the decrease in ne-absence in VS structures, particularly from the eighteenth century, when ne-deletion started to increase in SV declarative structures. Although ne-absence in interrogative and declarative structures appears superficially similar, the patterns of change are di¤erent because the phenomenon is not the same in both structures; in declarative structures, ne is deleted while in interrogative structures, either ne is absent – when polarity items are used – or it is bleached by a syntactic/semantic checking relation. However, by the nineteenth century, both interrogative and declarative structures had started to merge in SV word order, leading to the same phonetic conditioning and social stratification. The Jespersen’s cycle implies a cause-e¤ect relation between the weakening of ne and its reinforcement by negative items. In this article, I have provided evidence against a cause-e¤ect relation, leading from one-term negation to two-term negation and then to one-term negation again. Based on a sociolinguistic analysis of new vernacular data, I have shown that phonetic weakening of ne occurred after the rise of two-term negation. I have also shown that in interrogative structures, up to Classical French, VS word order, which allows the licensing of the polarity items, is the key to understanding ne-absence. Both analyses of absence of ne, in declarative and interrogative contexts, have proved the heuristic value of sociolinguistic corpora such as family letters or dialogic text corpora to follow external and internal changes. Acknowledgments This article has been funded by the Social Sciences and Humanities Research Council of Canada (MCRI project Modeling Change: The Paths of French, directed by F. Martineau). I would like to thank those attend-

204

France Martineau

ing the following conferences where I presented previous versions of this article for their suggestions: University of Florida, March 2008; Corpus and Variation in Linguistic Description and Language Education, Tokyo, May 8 and 9, 2008; Leverhulme Trust Seminars on Cycles of Grammaticalization, Birmingham, August 2008). Special thanks to the leaders of the conference, Richard Ingham and Pierre Larrive´e, to Vo¨lker Gast, and to Martin Elsig, Yves Charles Morin, Raymond Mougeon and Sandrine Tailleur for useful discussions.

References Ashby, William 1977 Interrogative forms in Parisian French. Semasia 4: 35–52. Ashby, William 2001 Un nouveau regard sur la chute du ne en franc¸ais parle´ tourangeau: s’agit-il d’un changement en cours? Journal of French Language Studies 11: 1–22. Blanche-Benveniste, Claire & Colette Jeanjean 1986 Le franc¸ais parle´: transcription et e´dition. Paris: Didier. Buridant, Claude 2000 Nouvelle grammaire de l’ancien franc¸ais. Paris: SEDES. Coveney, Aidan 2002 Variability in spoken French. A sociolinguistic study of interrogation and negation. Bristol/Portland: Elm Bank. De´prez, Viviane & France Martineau 2004 Microparametric variation and negative concord. In: Julie Auger, Clancy Clements & Barbara Vance (eds.), Contemporary approaches to Romance linguistics, 139–158. Amsterdam: Benjamins. Dufter, Andreas & Elizabeth Stark 2007 La linguistique variationnelle et les changements linguistiques ‘mal compris’: le cas du ne de ne´gation. In: Bernard Combettes & Christiane Marchello-Nizia (eds.), E´tudes sur le changement linguistique, 115–128. Nancy: Presses Universitaires de Nancy. Elsig, Martin 2009 Grammatical variation across space and time: The French interrogative system. Amsterdam: Benjamins. Foisil, Madeleine (ed.) 1989 Journal de Jean He´roard. 2 vols. Paris: Fayard. Fournier, Nathalie 1998 Grammaire du franc¸ais classique. Paris: Belin. Hansen, Maj-Britt Mosegaard 2001 Syntax in interaction. Form and function of yes/no interrogatives in spoken standard French. Studies in Language 25 (3): 463–520.

Ne-absence in declarative and yes/no interrogative contexts

205

Hansen, Maj-Britt Mosegaard 2009 N-words features in the negative cycle of French. Paper read at Cycles of Grammaticalization Seminar, Birmingham, December 6, 2008. Hand-out. Harris, Martin 1978 The evolution of French syntax: A comparative approach. London: Longman. Hausmann, Franz-Joseph 1992 L’aˆge du franc¸ais parle´ actuel: bilan d’une controverse allemande. Grammaire des fautes et franc¸ais non conventionnels. Actes du 4e colloque international organise´ par le Groupe d’e´tude en histoire de la langue franc¸aise. 355–362. Ingham, Richard to appear L’anglo-normand et la variation syntaxique en franc¸ais me´die´val. In: Maria Iliescu, Heidi Siller-Runggaldier & Paul Danler (eds.), Proceedings of the XXVth Colloque International de Linguistique et de Philologie Romane. Berlin/New York : Mouton de Gruyter. Jespersen, Otto 1917 Negation in English and other languages. Copenhagen: A.F. Høst. Kawaguchi, Yuji 2009 Particules ne´gatives du franc¸ais : ne, pas, point et mie: Un aperc¸u historique. In: Luc Baronian & France Martineau (eds.), Le franc¸ais, d’un continent a` l’autre, 193–210. Quebec City: Presses de l’Universite´ Laval. Kiparsky, Paul & Cleo Condoravdi 2006 Tracking Jespersen’s cycle. In: Mark Janse, Brian D. Joseph & Angela Ralli (eds.), Proceedings of the Second International Conference on Modern Greek Dialects and Linguistic Theory, 172– 197. Mytilene: Doukas. Lucas, Christopher 2007 Jespersen’s cycle in Arabic and Berber. Transactions of the Philologic Society 105: 398–431. Marchello-Nizia, Christiane 1997 Histoire de la langue franc¸aise aux XIVe et XVe sie`cles. Paris: Nathan. Martineau, France 1995 Corpus de franc¸ais familier ancien. Corpus and documentation, University of Ottawa. Martineau, France 2009a Vers l’Ouest : les varie´te´s laurentiennes. In: Luc Baronian & France Martineau (eds.), Le franc¸ais, d’un continent a` l’autre, 291–325. Quebec City: Presses de l’Universite´ Laval. Martineau, France 2009b Modeling change: A historical sociolinguistics perspective on French negation. In: Yuji Kawaguchi, Jacques Durand & Makoto

206

France Martineau

Minegishi (eds.), Corpus and variation in linguistic description and language education, 159–178. Amsterdam: Benjamins. Martineau, France ` distance de Paris : usages linguistiques en France et en Nouvelle2009c A France a` l’e´poque classique. In: Dorothe´e Aquino-Weber, Sara Cotelli & Andres Kristol (eds.), Sociolinguistique historique du domaine gallo-roman : enjeux et me´thodologie, 221–242. Bern: Peter Lang. Martineau, France (ed.) 2009d Corpus Mode´liser le changement: les voies du franc¸ais (MCVF). [Available at http://www.voies.uottawa.ca] Martineau, France & Marie-The´re`se Vinet 2005 Microvariation in French negation markers: An historical perspective. In: Montse Batllori, Maria-Lluı¨sa Hernanz, Carme Picallo & Francesc Roca (eds.), Grammaticalization and parametric change, 194–205. Oxford: Oxford University Press. Martineau, France & Raymond Mougeon 2002 Corpus de the´aˆtre et de textes parodiques du 17 e au 20 e sie`cles, Corpus and documentation. Martineau, France & Raymond Mougeon 2003 Sociolinguistic research on the origins of ne deletion in European and Quebec French. Language 79 (1): 118–152. Morin, Yves Charles 2008 Le Myste`re du sie`ge d’Orle´ans: le chva ornemental et l’activation des consonnes e´tymologiques dans le franc¸ais du XVe sie`cle. In: Yvan Lepage & Christian Milat (eds.), Por s’onor croistre : Me´langes de langue et de litte´rature me´die´vales o¤erts a` Pierre Kunstmann, 257–271. Orle´ans: David. Morin, Yves Charles 1974 Re`gles phonologiques a` domaine inde´termine´: chute du chva en franc¸ais. Cahier de linguistique 4: 68–88. Picoche, Jacqueline & Christiane Marchello-Nizia 1989 Histoire de la langue franc¸aise. Paris: Nathan. Pohl, Jacques 1972 Ne et les enfants. L’homme et le signifiant. Paris: Nathan. Price, Glanville 1971 The French language, present and past. Jameson. Price, Glanville 1993 Pas ( point) without ne in interrogative sentences. Journal of French Language Studies 3: 191–196. Price, Glanville 1997 Negative particles in French. In: Stewart Gregory & David Trotter (eds.), De mot en mot, Aspects of medieval linguistics, 173–190. Cardi¤: University of Wales Press.

Ne-absence in declarative and yes/no interrogative contexts Roberts, Ian 1993

207

Verbs and diachronic syntax: A comparative history of English and French. Dordrecht: Kluwer. Sanko¤, Gillian & Diane Vincent 1977 L’emploi productif du ne dans le franc¸ais parle´ a` Montre´al. Le franc¸ais moderne 45 (3): 243–256. Schnedeker, Catherine & Sophie Pre´vost 2002 Aucun(e)s/d’aucun(e)(s)/les aucun(e)s: E´volution du franc¸ais me´die´val au franc¸ais moderne. Seminar Diachronie du Franc¸ais, IUF, May 2002. Hand-out. Thomas, Dominike 2010 E´tude sociolinguistique historique du syste`me interrogatif : les interrogatives dans le Journal de Jean He´roard. M.A. Thesis, University of Ottawa. Van der Auwera, Johan 2009 The Jespersen cycle. In: Elly van Gelderen (eds.), Cyclical change, 35–71. Amsterdam: Benjamins. Vinet, Marie-The´re`se 2000 Feature representation and -tu pas in Quebec French. Studia Linguistica 54 (3): 381–411. Vinet, Marie-The´re`se 2001 D’un franc¸ais a` l’autre: la syntaxe de la microvariation. Montreal: Fides.

The early absence of the French negative marker ne Paul Rowlett 1. Introduction The diachronic development of sentential negation within Romance, and in French in particular, is a much researched topic, and one could be forgiven for thinking that there is little of interest left to investigate. Yet, as Richard Ingham and Pierre Larrive´e’s Leverhulme-sponsored series of seminars has amply demonstrated, much remains uncertain with respect to, for example, the clausal cartography of negative features and the respective contribution to the expression of negation made by di¤erent ‘negative’ markers. The present article provides a discussion of the contribution in this volume by France Martineau. The phenomenon of interest is the behaviour of ne [n‘] (< Latin non), the French pre-verbal marker of sentential negation in use since the early tenth century. This marker is well known to be on the decline. It is not occurring where it might have been expected, that is, in clauses containing one or more negative reinforcers such as plus, pas, point, rien, aucun, which even in the modern language typically co-occur with ne in the expression of sentential negation. This phenomenon characterising the transitional stage within the Jespersen Cycle might be expected to be constant across contexts. However, not all contexts are equal in this respect. In particular, variations are observed in relation to features of the grammatical context such as nominal subjects and the nature of negative reinforcers (as shown by Ingham this volume for early French). There is a further asymmetry between sentence type. Declaratives and interrogatives have been known to contrast with respect to the occurrence of ne since Price (1978, 1993). This assumption is confirmed by Martineau. She considers the patterns as to the presence of ne in negative declarative and (matrix yes–no) interrogative sentences from the seventeenth to the nineteenth century, and uses statistical data from large-scale, geographically and sociolinguistically stratified, literary and non-literary historical corpora to illustrate how the same patterns/ chronology of ne-absence are not found in the two kinds of sentence.

210

Paul Rowlett

Martineau’s data relating to the changing market shares of the five sentence types found in interrogatives and the data relating to the changing ne-absence rates found in each one show that the unexpectedly high rate of ne-absence in interrogative sentences prior to the eighteenth century (8.8% in the seventeenth century in Martineau’s data; cf. 0.3% in declaratives) is due to the fact that in interrogative sentences at this time the market share of VS word orders (with an average ne-absence rate of 11.8%) was high (69.8%), while the market share of SV word orders (with an average ne-absence rate of just 1.8%) was low (30.2%). The conclusion, therefore, is that the unexpectedly high ne-absence rate in interrogative sentences in the seventeenth century (in contrast to declarative sentences) was due, not to the pragmatics of interrogation as such (at least, not directly), but rather to the high frequency in interrogative sentences at this time of two specific structures (pronominal inversion and complex inversion), which, with their unambiguously VS word order, have a comparatively high rate of ne-absence, and are typically absent from declaratives. And the loss, in subsequent centuries, of the di¤erence in the rate of the absence of negative ne in interrogative and declarative sentences is due to the shift, in interrogative sentences, away from structures with VS word order (which are not typically found in declaratives) and toward those with SV word order (which are typically found in declaratives). In other words, the reason why interrogatives increasingly behave like declaratives with respect to their ne-absence rate is that they increasingly use the structures typically found in declaratives. This section has provided a summary of the analysis by Martineau that shows how (subtypes of ) interrogatives di¤er from assertives in the rate of non-occurrence of ne. The two next sections consider how this may be explained from the perspective of formal syntax.

2. Analysis Before embarking on a formal syntactic analysis proper of the di¤erent subtypes of interrogatives and how they impact on the rate of non-occurrence of ne as I will in section 3, this section does three things. First, I identify the relevant formal property distinguishing interrogative sentences with VS word order from those with SV word order (section 2.1.), as these are shown by Martineau to impact on the use of ne. Second, echoing Martineau, I give my take on what is meant by ‘the absence of ne’ in the relevant con-

The early absence of the French negative marker ne

211

texts (section 2.2.). Third, I show how the two might be brought together from a formal point of view (section 2.3.). 2.1. VS vs. SV interrogatives I propose a syntactic account of French interrogatives, in particular the distinction between those with VS and those with SV word orders, in line with the general framework of Rowlett (2007: 201–17). Clauses are analysed as comprising three domains, a lexical domain labelled VP*, an inflectional domain labelled IP* and a discourse domain labelled CP*. (The * notation indicates that we are dealing in each case with an intricately articulated hierarchical array of projections rather than a single projection; see Rizzi 1997 and Cinque 1999.) (1) [CP* . . . [ IP* . . . [VP* . . . ] ] ] In simple declarative sentences with SV word order in Modern French the finite verb raises (for morpho-syntactic reasons which are independent of pragmatic force) out of the lexical VP* domain to the top-most functional head within the inflectional domain of the clause, I* . From this position it can check the j features of the subject, which raises out of the lexical VP* domain to the canonical subject position, SpecIP*: (2)

Matrix yes–no interrogatives with VS word order (pronominal/complex inversion) are commonly analysed on the basis of an interrogative feature [Q] borne by the finite verb in I* . The feature [Q] endows the sentence with its interrogative force, but needs to check the relevant operator feature

212

Paul Rowlett

(here [wh]1) of an element (here a non-overt yes–no-question operator OP) in a specifier position. Since the local specifier position, SpecIP*, is occupied by the subject, the checking needs of the feature [Q] borne by the finite verb in I* cannot be met locally. Instead, the element OP[wh] must occupy a specifier position outside IP*, that is, within the discourse domain CP*. However, since I* is not in a local specifier–head configuration with a specifier position within CP*, the [Q] feature on the finite verb in I* will not be able to check OP[wh] within CP*. One way to produce the required configuration (between the [Q] feature borne by the finite verb in I* and OP[wh] within CP*) would be for the finite verb to raise, by head movement, from I* to a head position within CP*. However, the empirical data suggest that this kind of movement has not been possible in French for some considerable time: if movement of the finite verb from I* to a head within CP* were possible, we would expect to find simple inversion of the finite verb around a nominal subject, yet we do not: (3) *Est Pierre parti? Instead, in Rowlett (2007) I follow Cyrille-Thomas (2003) in proposing (a) that since the [Q] feature borne by the finite verb in I* cannot be checked locally, a representational chain is necessarily created between I* and a higher head, that is, within CP*, (b) that the existence of this chain creates a checking configuration between I* and SpecCP*, and (c) that this in turn allows OP[wh] to target SpecCP*. The required checking configuration obtains in virtue of the representational chain (linking I* and C* ) and the local specifier–head relationship between C* and SpecCP*, and so checking is possible between the finite verb in I* bearing [Q] and OP in SpecCP* bearing [wh]:

1. I have used the label wh here, although this might be thought inappropriate in the context of yes–no (as opposed to wh) questions. The precise label used is not important. The key issue is that the [Q] feature borne by the finite verb in I* needs to be checked against the relevant feature in specifier position. The issue of the nature of the distinction between wh and yes–no questions is beyond the scope of this article.

The early absence of the French negative marker ne

213

(4)

However, the fact that (the finite verb in) I* checks (OP[wh] in) SpecCP* (indirectly via a representational chain and a specifier–head relationship) means, assuming a biuniqueness condition on checking, that (the finite verb in) I* cannot now also check (the j features of the subject in) SpecIP*. To resolve this problem, the I* position (containing the finite verb) is re-endowed with the ability to check (the j features of the subject in) SpecIP* by the presence of an ‘a‰xal agreement marker’, in other words, the ‘inverted’ subject clitic proform which is characteristic of pronominal/complex inversion in (1a, b).2 Rather than being licensed by the finite verb in I* , the subject in SpecIP* is licensed by the agreement marker which is a‰xed to the finite verb in I* , as illustrated by the underlining in (5): (5)

2. For reasons to analyse the ‘inverted subject proform’ found in pronominal/ complex inversion as an a‰xal element rather than as a subject proform that the finite verb has inverted around, see Rowlett (2007: 204–6).

214

Paul Rowlett

The subtle distinction being drawn here with respect to what it is that checks (the j features of ) the subject in SpecIP* is significant and useful because the morphological richness of the agreement marker allows it to identify the subject, and thereby satisfy the conditions for pro drop. Thus, in ‘inverted’ VS contexts a nominal subject in SpecIP* (resulting in complex inversion; (1b)) alternates, not with an overt pronominal subject – note the ungrammaticality of (6) –, but with non-overt pro (resulting in pronominal inversion; (1a)). (6) *Il est-il parti? While interrogatives with VS word order (that is, pronominal/complex inversion with an a‰xal agreement marker) involve the presence of the [Q] feature on the finite verb in I* , the same is not the case of the two interrogative structures with unambiguous SV word order, namely, intonation SV, as in Pierre est parti?, and est-ce que SV, as in Est-ce que Pierre est parti? In Rowlett (2007) I adopt the common assumption that [Q] is here located on C* . The higher location of [Q] means that it can check OP[wh] within a local specifier–head configuration without the need for a representational chain and, therefore, without the need for an a‰xal agreement marker: (7)

Thus, on this analysis, the formal distinction between VS from SV interrogatives centres on the location of the [Q] feature: on I* in VS interrogatives; on C* in SV interrogatives; the other superficial di¤erences follow directly from this. 2.2. Radical absence of ne Turning now to the absence of ne in the contexts which concern us here, that is, in interrogative sentences containing a negative reinforcer, I follow

The early absence of the French negative marker ne

215

Martineau’s (this volume) approach, which goes back to Price (1978, 1993). Although he does not provide the socially-stratified quantitative data that Martineau does, Price suggests that the early absence of ne in interrogatives is not in fact related to its later emergence in declaratives in that the clauses displaying the phenomenon are not in fact negative. In other words, the early absence of ne in interrogative contexts constitutes genuine ‘absence’ (radical emptiness) rather than ‘deletion’ (underlying/ initial syntactic presence with superficial/subsequent zero phonological realisation). Price argues that the negative reinforcers found in these examples, especially point, retained a virtual/potential value and that, until the classical period, they were polarity items which could be licensed by interrogation instead of negation. In his model a question like As-tu rien vu de plus joli? presupposed a positive response. In contrast, a question like N’as-tu rien vu de plus joli? was a genuinely negative sentence and so no particular response was presupposed. Only later, as the negative value of these elements strengthened (and pas dominated) were the structures assumed to contain (deleted) ne. Thus, the key feature of the early interrogative sentences with absence ne is that they satisfied, without recourse to sentential negation, the licensing conditions for the virtual/ potential polarity items that at the time the ‘negative’ reinforcers were, following Price and Martineau. 2.3. Bringing the two together The elements needed to understand the empirical phenomenon under discussion can be set out as follows, many of which are proposed by Martineau: (a) the negative reinforcers in the early ne-less interrogative sentences are virtual/potential polarity items rather than truly negative items; (b) as such, these items have licensing requirements which can be satisfied by a wider range of items than mere negation; (c) one such licensor is the [Q] feature found in interrogative sentences; (d) in order to license a negative reinforcer qua virtual/potential polarity item, the [Q] feature must not only be present but also be ‘local’ in the sense of being instantiated within IP*; (e) this licensing condition is satisfied in interrogative sentences with VS word order, where the [Q] feature appears on the finite verb in I* , but not in interrogative sentences with SV word order, where the [Q] feature appears on C* :

216

Paul Rowlett

(8)

(9)

The configuration in (8) corresponds to the VS structures of pronominal/ complex inversion: [Q] is on I* . This is the configuration which licenses the virtual/potential polarity items point, rien, aucun, etc., in the absence of the expression of sentential negation or, therefore, the particle ne. The configuration in (9) corresponds to the SV interrogative structures: [Q] is on C* . This configuration does not license point, rien, aucun, etc. 3. Discussion The licensing of the virtual/potential polarity items point, rien, aucun, etc., and the possibility that these items might therefore appear in non-negative sentences without ne was determined above to be due to the presence of [Q] on I* . In fact, the licensing of these virtual/potential polarity items is not the only phenomenon which appears to be sensitive to the position (as opposed to the presence) of [Q]. Consider the data in (10), taken from Huot (1986) and discussed in Rowlett (2007: 150):

The early absence of the French negative marker ne

(10) a.

b.

217

Tu crois qu’il part. you believe that-he leave-pres.ind ‘You believe he’s leaving.’ Crois-tu qu’il parte? believe-you that-he leave-pres.subj ‘Do you believe that he’s leaving?’

c. Tu crois qu’il part(*e)? you believe that-he leave-pres.ind/*subj ¼ (10b) d.

Est-ce que tu crois qu’il part(*e)? is-it that you believe that-he leave-pres.ind/*subj ¼ (10b)

e. Peut-eˆtre crois-tu qu’il part(*e). maybe believe-you that he leave-pres.ind/*subj ‘Maybe you believe that he’s leaving.’ Used in a simple declarative sentence, the bridge verb croire ‘to believe’ selects a finite complement clause with indicative mood, as in (10a). Used in a matrix yes–no interrogative, it can select subjunctive mood, but only if the interrogative is expressed using an inversion structure, as in (10b); intonation SV and est-ce que SV do not license subjunctive mood, as shown in (10c, d). It is, however, not inversion per se which licenses the subjunctive mood; this is shown in (10e), where inversion is triggered in a non-interrogative context. Given the approach above to pronominal/complex inversion in interrogative sentences, the data in (10) suggest that subjunctive mood in the subordinate clause is licensed by the interrogative nature of the matrix clause containing the bridge verb (and therefore that [Q] is involved) but also, and more specifically, that for subjunctive mood to be licensed the [Q] feature must appear locally on I* rather than C* . This is precisely the situation we have in respect of the radical absence of ne. So we have two phenomena, at di¤erent stages of the development of French, where what appears to be needed is the [Q] feature located on I* rather than C* . We might wonder what further research avenues are indicated by the formal analysis above. Three suggest themselves. First, there is the issue of the underlying explanation of the contrast identified. Why is it that the phenomena considered above are subject to such a restrictive licensing condition? The explanation might lie in the grammatical function of the

218

Paul Rowlett

clausal domains. If the function of IP* is to anchor the clause with respect to such notions as tense and modality, while CP* is about discourse, then it perhaps comes as no surprise that virtual/potential polarity items, or indeed subjunctive mood, should have a licensing requirement which refers specifically to the IP* domain. Second, there is the issue of whether the diachronic penetration of the shift in the locus of the [Q] feature (in matrix yes–no interrogatives) from the IP* to the CP* domain is related to any other shift. One possible candidate parallel shift is the one towards clitic left dislocation, illustrated in (11), which sees topical subjects appearing within CP* rather than in the canonical subject position SpecIP*. (11) a.

b.

[IP* Jean part ] Jean leaves ‘Jean is leaving.’ [CP* Jean [IP* il part ]] Jean he leaves ‘Jean is leaving.’

Finally, there is the issue of the extent to which negative reinforcers retain their status as virtual/potential polarity items. If their early occurrence in interrogative sentences in the absence of ne fell as a consequence of the decline of VS word orders, as is also suggested by Martineau (this volume), then there is perhaps no need to conclude that they ever lost their virtual/ potential value. Their increasingly close association with negation may have been a consequence of the weakening of ne rather than their own negativisation. Such an approach would also sit comfortably with analyses of sentential negation in French, and in particular of phenomena like negative concord, which are based on the assumption that negative reinforcers (with the exception of pas) are (even in the modern language) not inherently negative (e.g. Rowlett 1998a).

4. Conclusion The aim of this article has been to o¤er a formal syntactic analysis of the contention demonstrated by Martineau that the early absence of ne in ‘negative’ interrogatives is a di¤erent phenomenon to its later absence in negative declaratives. The proposal is that the relevant trigger for the phenomenon was the appearance, within the narrow domain of IP*, rather than the broader CP* domain, of the interrogative feature [Q].

The early absence of the French negative marker ne

219

This opens new avenues for the consideration of how this formal change interacts with other aspects of the historical grammar of French.

Acknowledgment I would like to express my gratitude to Pierre Larrive´e and Richard Ingham for inviting me to participate in the Leverhulme Trust-funded collaborative research project Cycles of grammaticalization during 2008 and 2009. The usual disclaimers apply. References Cinque, Guglielmo 1999 Adverbs and functional heads: a cross-linguistic perspective. Oxford: Oxford University Press. Cyrille-Thomas, Odile 2003 The syntax of fronting in French and Guadeloupe Creole. Ph.D. dissertation, University of Salford. Dufter, Andreas & Elizabeth Stark 2007 La linguistique variationnelle et les changements linguistiques ‘mal compris’: le cas du ne de ne´gation. In: Bernard Combettes & Christiane Marchello-Nizia (eds.), Etudes sur le changement linguistique, 115–128. Nancy: Presses Universitaires de Nancy. Elsig, Martin 2009 Grammatical variation across space and time: the French interrogative system. Amsterdam: Benjamins. Huot, He´le`ne 1986 Le subjonctif dans les comple´tives: subjectivite´ et modalisation. In: Mitsou Ronat & Daniel Couquaux (eds.), La grammaire modulaire, 211–246. Paris: Minuit. Martineau, France 2008 Phrase structure of the negative cycle in French. Seminar, Aston University, August. Martineau, France & Raymond Mougeon 2003 Sociolinguistic research on the origins of ne deletion in European and Quebec French. Language 79 (1): 118–152. Price, Glanville 1978 L’interrogation ne´gative sans ne. Studii s¸i cerceta˘ri lingvistice 29: 599–606. Price, Glanville 1993 Pas ( point) without ne in interrogative clauses. Journal of French Language Studies 3: 191–195.

220

Paul Rowlett

Rizzi, Luigi 1997

Rowlett, Paul 1998a Rowlett, Paul 1998b Rowlett, Paul 2007 Rowlett, Paul 2010

The fine structure of the left periphery. In: Liliane Haegeman (ed.), Elements of grammar: A handbook of generative syntax, 281–337. Dordrecht: Kluwer. Sentential negation in French. Oxford: Oxford University Press. [Available at: http://usir.salford.ac.uk/2703/] A non-overt negative operator in French. Probus 10: 185–206. The syntax of French. Cambridge: Cambridge University Press. [Available at: http://usir.salford.ac.uk/1355/] Syntactic variation in French: Diglossia and language change. Invited plenary lecture, Nineteenth Greater Manchester Postgraduate Linguistics Conference, September 2010. [Handout available at: http://usir.salford.ac.uk/10075/]

Rowlett, Paul In preparation French imperatives, negative ne, and non-subject clitics.

Atoms of negation: An outside-in micro-parametric approach to negative concord Viviane De´prez 1. Contrasting theoretical outlooks: Inside-out vs. outside-in perspectives In this paper, two core perspectives on variation in negative concord are distinguished, one that focuses on the syntactic nature of negation as the central factor of variation, and one that focuses on the internal make-up of the negative dependent terms, the n-words. The first part of the paper outlines the di¤erent predictions that emerge from these di¤ering perspectives. The second part provides evidence in support of the latter perspective. Recent work in the Minimalist framework has revamped the classic idea pioneered by the nineteenth century linguist Otto Jespersen that there is ‘‘a strong correlation between the syntactic status of the sentential negation marker and the phenomenon of negative concord’’ (Zeijlstra 2008:1). For Jespersen, negative relations were subject to a cycle largely conditioned by the syntactic nature of the sentential negation marker as weak or strong. Similarly, with their core reliance on the strength or interpretability and weakness or uninterpretability of negative features, Minimalist approaches such as Zeijlstra’s (2008) (see also Zeiljstra 2004; Willis 2004; Penka 2007), which cast negative concord as a prime case of an Agree relation a` la Chomksy (2000, 2001), operate a terminological and conceptual fusion with the classic Jespersen view. Like Jespersen’s, these approaches are macro-parametric in nature, as they seek to formulate ‘‘a single theory of negative concord, that also predicts the parametric variation between negative concord languages (strict vs. non-strict NC languages) as well as the variation between NC languages and DN languages’’ (Zejlstra 2008: 3). But furthering Jespersen, these perspectives extend the cycle’s purview beyond the syntax of negation proper to include negative dependencies in a broader sense, i.e. relations between sentential negation and dependent nominal terms. In this paper, this neo-classic macro-parametric view to negative concord variation is contrasted with a distinct micro-parametric approach to NC whose core idea is to move away from the classic and neo-classic Jespersonian focus on the syntactic properties of the sentential negative

222

Viviane De´prez

marker to center instead on the syntactic properties of the n-words themselves. This approach takes the negative dependent expressions as key factors of variation in negative concord systems (Corblin, de Swart & De´prez 2004; De´prez 1997, 1999, 2000; De´prez & Martineau 2004; Tovena, De´prez & Jayez 2004; Londhal & Haegeman 2010). At the heart of this micro-parametric approach is the claim that the diverse properties of negative concord observable in a great variety of languages, synchronically or and diachronically, primarily derive from the internal micro-morphosyntax and resulting di¤erences between participating negative expressions at the syntax-semantics interface, rather than from the syntactic nature of clausal negation. On this perspective negative concord works from the outside-in, that is, from the micro-syntactic level of negative expressions to the macro-syntactic level of the clause unlike Jespersen’s and the neo-classic approaches, which work from the inside-out, that is, from the macro-level of sentential negation to the n-expressions as schematized in (1): (1)

One could perhaps counter at the outset that, given the Agree operation (Chomsky 2001) that underlies the neo-classic approaches, and the correspondence of features that checking presupposes between the internal syntactic domain of n-expressions and the larger sentential domain that contains them, the directionality argued for in our micro-parametric approach could turn out to be a mere matter of favoured perspective, so that outside-in and inside-out perspectives would in the end boil down to the same. Not so for negative concord I argue here. The paper endeavours to demonstrate that for an empirically adequate account of variation and variety in negative relations, the direction of perspective in fact matters considerably and has a number of significant theoretical and empirical consequences that are calculable and factually verifiable in non-trivial cases. Below, it is first argued that the general direction in perspective, i.e. focus on n-expressions rather than on the sentential negation marker, sets out opposite empirical expectations in regards to the internal homo-

An outside-in micro-parametric approach to negative concord

223

geneity of negative systems in particular languages. It is further argued that this also has important consequences from a diachronic point of view. In its second part, the paper then endeavours to demonstrate that, within a restricted empirical domain, the predictions of the micro-parametric approach are clearly verified. The paper is structured as follows: Section 1.2. contrasts the theoretical predictions of a Jespersonian neo-classic inside-out macro-parametric model of negative concord, like that of Zeijsltra (2008), with those of our micro-parametric outside-in approach for both synchronic comparative linguistic variation and diachronic language change. Section 3 presents comparative data from contemporary European French and French-based Creoles that underscore the importance of inner language diversity expected under the micro-parametric approach, but not under a macro-parametric model. Section 4 o¤ers a detailed study of the internal structure of the French n-word rien and personne and of their historical evolution as a further diachronic confirmation of the importance of internal structural changes for the correlated changes in the properties of negative concord. In this section, an in-depth look at diachronic changes in the modification possibility of French n-words across time is presented that allows their internal evolution to be charted. 1.2. Cross-linguistic variation and inner language homogeneity On what has been termed here a ‘neo-classic inside-out perspective’ on negative relations, variation in the negative concord properties of a language is expected to concur with variation in the syntactic nature of the sentential negation marker. Thus, if a given language presents a given type of sentential negation marker, then expectations are that this language should manifest a given type of negative relations quite generally. Such a perspective is common to a number of generative approaches, starting from Zanuttini (2001) up to the more recent works of Zeijlstra (2004) and subsequent publications. Concretely, for Zeiljstra (2008) for instance, a core generalization that his approach is designed to predict is that: (2) Every language that has a negative marker X is an NC language (provided that n-words are present). (Zeijlstra 2008: 17)1 1. In Zeijlstra’s model, a negative X0 is allowed to be phonetically null. This is unfortunate, because to a large extend, it quite significantly hollows out the empirical bite of this model, making it quasi impossible to be empirically contradicted. Indeed, any language that manifests negative concord can in principle be claimed to have a null X0 negation, thus meeting (2), so to speak, by theoretical brute force.

224

Viviane De´prez

The implication here is unidirectional, as is customary with other macroparameters proposed in the literature (Baker 2008), so that if the sentential negation marker of a language L is not an X0, no prediction is made with respect to negative concord. Note, however, and this is fundamental, that the generalization in (2) makes sense only in so far as the notion of negative concord language can be well defined independently. Assume, for the sake of the argument, the fairly uncontroversial view that a negative concord language is a language in which negative relations generally take the characteristic shape of multiple negative marking for a single negative interpretation. What is clear then, given (2), is that a language with an X0 negation should manifest negative concord throughout. That is, inner language homogeneity with respect to negative concord, with multiple negative dependent terms interpreted as a single sentential negation, is expected. Such homogeneity must, in fact, quite simply, on Zeijlstra’s perspective, be a defining criterion for what a negative concord language is, if circularity is to be avoided. What is in question, here, however, is whether such homogeneous languages are in fact ever empirically observed.2 In the realm of negative dependencies, this turns out to be, it would seem, far from obvious. If anything, after years of careful observations in the literature, what seems to be quite characteristic of languages that e¤ectively manifest negative concord is that they regularly present a rather theoretically unsavoury patchwork of surprising inner variation. That is, languages in which a single clearly uniform type of negative dependencies prevails do not seem that common; what detailed empirical linguistic observations have commonly unearthed within particular languages are rather mixed systems in which there are some negative relations that present a certain type of linguistic behaviour along with others that present distinct ones. To wit, Italian or Spanish, which are treated as negative concord languages in Zeijlstra (2004), but also have distinct NPI expressions that function quite di¤erently (Zanuttini 1991; Laka 1990). Another such language is Greek, where negative dependent terms partition into two

2. The question is put in the strongest terms here, but note that a weaker form would be su‰cient to raise doubts about the typological predictions of (2). Indeed, even if a few truly homogeneous NC languages could be found, with all the negative relations taking the same shape, this would still not show that most languages allowing NC are or this sort, as the generalization in (2) leads us to expect. The existence of only a few truly homogenous languages is compatible with a micro-parametric approach, if it can be shown that all the n-expressions share crucial internal structures.

An outside-in micro-parametric approach to negative concord

225

sets with distinct negative relations depending on their stress patterns, and which has additional NPI expressions that fall under di¤erent licensing conditions (Giannakidou 1997). Yet another such language is SerboCroatian, where two sets of negative dependent terms (ni-NPIs vs i-NPIs) with di¤erent distributions and licensing conditions are distinguished (see Progovac 1994 among others).3 Finally, even American English, which, according to Zeijlstra (2004), is a clear example of non-NC language, can manifest unquestionable examples of NC. Witness the attested uses of the expression diddly squat that clearly functions as a negative in (3a), but just as clearly manifests concord when in co-presence with negation in (3b), or with another negative quantifier as in (3c): (3) a.

b. c.

Alan Cumming To Obama: You’ve Done Diddly Squat’ For Gay Rights. (http://www.thefreelibrary.com/ Oct 25, 2010) Those shiny cars won’t mean diddly squat when you die. (http://www.urbandictionary.com/ Oct 25, 2010) Nobody knows squat about your topic. (¼ Nobody knows anything) (Internet data cited in Postal 2004)

In short, in the realm of negative relations, within-language diversity appears to be more often the rule than the exception. From a general macroparametric inside-out perspective such as the one laid out in (2), the question that arises is: what should be made of such common inner-language variation for languages in which the sentential negation form stays reasonably constant? Faced with an inner-language diverse reality, should one conclude that either we do not have a bona-fide negative concord language even if some instances of negative concord can be observed, or if the given language is classified as a negative concord one, are we then not compelled to concede that at least some negative relations fail to consistently behave as predicted by the syntactic nature of sentential negation? In sum, what 3. Although Progovac’s (1994) study is limited to characterizing the licensing conditions of negative dependencies in Serbo-Croatian, she wonders about the source of the variation she observes in the conclusion of her chapter on language variation as follows: ‘‘If negative polarity items are homogenous class, why should they select di¤erent (LF) landing sites? First of all, they might not be a homogenous class at all. [. . .]. Perhaps the clues might be sought in their morphological properties.’’ (Progovac 1994: 90). The fruitfulness of this suggested approach is precisely what we aim at defending here.

226

Viviane De´prez

is far from obvious from an inside-out macro-parametric perspective is to see how a theory of NC whose determining factor is the syntactic nature of sentential negation could account for a language in which distinct n-expressions behave di¤erently while sentential negation, in contrast, remains constant. More generally, although an inside-out perspective such as Jespersen’s or Zeijlstra’s could well be adequately equipped to deal with cross-linguistic diversity given language-internal homogeneity, innerlanguage diversity, in contrast, appears to present a serious challenge to it. That inner language diversity does in fact occur even in languages that otherwise present a fairly solid reliance on negative concord will be forcefully argued and instantiated in what follows. It is further important to note that such inner-diversity situations are precisely the ones that the micro-parametric outside-in approach to negative concord defended here is particularly well equipped to handle theoretically. Since on this micro-parametric view, it is the internal make-up of an n-expression and its interpretation at the syntax-semantic interface that centrally determines its ‘external’ behaviour at the sentence level, and thus ultimately, its concord properties, then n-expressions do not need, and are not predicted to be homogeneous within any given language. In other words, on our micro-parametric perspective, inner-language or dialect homogeneity across all n-expressions is not required. Yet, the appearance of inner-language or dialect homogeneity could of course occur, since it is always possible that a majority of n-expressions within a given language have comparable inner structure. Note, incidentally, that this demonstrates the ability of the outside-in micro-parametric perspective to handle situations of inner language homogeneity, should they ever arise. However, inner language homogeneity is not a necessity on this view. What is predicted instead on a micro-parametric-outside-in perspective is that n-expressions with a common internal make up should behave homogeneously whether within or across languages. So there is homogeneity expected here as well, but it is homogeneity of a di¤erent kind. On the micro-parametric perspective, homogeneity across all the nexpressions within a single given language is not necessary. What is expected is homogeneity across n-expressions that can be shown to have the same syntactic make-up, in whichever language, dialect, or historical stage they may happen to occur. As a result, within languages, variation is not problematic for this perspective and is in fact expected whenever it can be shown that n-expressions di¤er significantly in their internal make up. In short, inside-out perspectives, such as Jespersen’s or Zeijlstra’s, which centre on the properties of sentential negation, produce expectations

An outside-in micro-parametric approach to negative concord

227

of homogeneity within languages that have a unique sentential negation marker, thereby raising the delicate question of what a (homogeneous) negative concord language should ultimately be. In contrast, the microparametric outside-in perspective produces expectations of homogeneity across types of n-expression structures, conveniently sidestepping the very delicate question of language homogeneity, characteristically an E-language perspective problem. Within a micro-parametric approach, generalizations are defined not over negative concord E-languages, but over types of internal-structures independently of their actual proportional embodiment in particular languages, dialects, sociolects or historical stages. As a result, if the morphostructure of an n-expression were to stay constant through time, then so would its negative concord properties. In contrast, internal changes should foster changes in external behaviour and, in particular, changes in the interaction with sentential negation, which may itself well remain constant. These are the expectations under the inside-out perspective defended here, and the following discussion will detail empirical evidence gathered in support of this view. 1.3. Historical variations The direction in perspective that distinguishes an inside-in macro-parametric approach from an outside-in micro-parametric model also has important consequences from a diachronic point of view. Given the focus in the former on the properties of sentential negation as a key factor of variation, it is clearly change in the nature of sentential negation that is expected to drive changes in the properties of negative relations. This is the core idea of Jespersen’s negative cycle. For Jespersen, it is the (phonological) weakening of negation that calls for its reinforcement by added negative expressions. For Zeijlstra, given the principle in (2), a change in the ‘weakening’ syntactic status of negation, i.e. if negation becomes an X0, would also change the negative concord expectation for any given language. Unfortunately, change in the other direction, from a weakened X0 negation, to an XP one, does not lead to any clear predictions in Zeijlstra’s model. The spirit of the proposal leads to the expectation that if sentential negation changes, then negative relations will be a¤ected. But the second part of (2) only predicts the development of negative concord in case negation is or becomes an X0.2. Let us consider the case of French, which probably o¤ers one of the bestdocumented examples of diachronic evolution for negation and negative

228

Viviane De´prez

dependencies. The in-depth historical and comparative dialectal research on the evolution of the French negative operator ne and ne . . . pas conducted by Martineau & Mougeon (2003) and Martineau & De´prez (2004) shows that changes in the sentential marker do not drive changes in the nature of negative dependencies or the behaviour of n-words. By the time that ne and pas had acquired their current syntactic and semantic properties, n-words still largely behaved like dependent NPIs. They therefore acquired their current intrinsic negative meaning and their current negative concord properties well after ne had weakened and pas had been strengthened. Pas already had the properties it has nowadays in the fifteenth century, a time at which, as our work demonstrates, n-words like rien, personne and aucun were mostly not negative by themselves but instead quite clearly dependent on negation for their negative meaning. That is, they manifested NPI like properties. Moreover, they clearly also had distinct syntactic properties. In short, the acquired negative character of contemporary n-words, and hence of negative concord, cannot be directly triggered by the weakening of sentential negation (interpretable negation switching to uninterpretable negation or vice versa, however this is encoded) and so it appears equally misguided to assume that the properties of negative concord in contemporary French could causally result from a change in the nature of sentential negation. Strictly speaking, a principle like (2) simply appears to be unable to explain the historical rise of French negative concord, since it links negative concord to a weak/X0 status of negation while French negative concord and n-words did not arise until ne was strengthened by the non-X0 negation pas. Ingham’s work on the diachronic replacement of nul by aucun in medieval to Middle French and in Anglo-Norman also sharply demonstrates the occurrence of changes in negative relations and n-words that are fully independent of any changes in sentential negation. In the same vein, it is equally well acknowledged that the French negation ne already resulted from the weakening of the Latin negation non and yet, it has come to no one’s mind to claim that French negative concord derived from this first negative weakening. On the basis of these results (see also Martineau in this volume for a further study on the diachronic changes of sentential negation), it can thus be quite firmly concluded that, in French, the expected causal direction of an inside-out perspective, i.e. changes in sentential negation driving changes in the nature of n-words dependencies towards negative concord, is simply not supported by the available historical evidence.

An outside-in micro-parametric approach to negative concord

229

Interestingly, moreover, such a causal relation does not appear to be observed in the diachrony of Greek either. As explicitly noted by Kiparsky & Condoravdi (2005b: 2) there are ‘‘numerous instantiations of the (Jespersen) cycle from Medieval to Modern Greek, but the syntax of negation stays the same’’. That is, changes occurred in negation dependent elements, while negation remained constant. So here, again, diachronic evidence does not support the view that it was changes in the syntactic properties of sentential negation that drove the changes in negative relations. In contrast, if we take the outside-in perspective, defended here and in previous works, expectations are that, in negative relations, historical change should proceed from the internal morpho-structure of n-expressions to their external sentential behaviour, and not the other way around. That is, change should primarily operate from the internal micro-domain of the n-expressions to the external macro-domain of clausal negation. In other words, this view makes clear empirically verifiable predictions on the general course of historical development of negative relations positing that changes in the internal structure of n-expressions will drive changes in negative relations. Section 3.2 presents a detailed analysis of the internal changes in the structure of French n-words supporting this view. It discusses historical data on the modification of n-words. The key idea pursued here, as before (De´prez 1997, 1999, 2000, 2004) and also adopted in Roberts & Roussou (2003), is that restructuration arose in the internal structure of French n-words as a consequence of the loss of bare noun phrases, i.e. the gradual loss of null determination in Modern French, which, as argued previously in De´prez (2000, 2005), is itself a consequence of the evolution of morphological number marking in French.4 In simple terms, what is defended here, as before, is that the historical change of French negative relations is in large part a consequence of the internal syntactic restructuring of its n-expressions, which was triggered by yet a distinct change in the expression of number in the language, causing the suppression of null determiners. Although changes also occurred in 4. It is thus the weakening of number marking rather than the mere existence of bare noun phrases that De´prez (1999, 2000) proposes as a factor of change in the internal structure of French n-words. Crucially, as discussed in detail in De´prez (1999), it would be wrong to assume, as for instance Willis (2007) mistakenly does, that the claim is that the mere existence of bare nouns should coincide with negative concord. Counterexamples to such a claim abound, as noted in De´prez (1999), with English and German being quite obviously languages in which bare NPs are possible, but negative concord supposedly is not.

230

Viviane De´prez

sentential negation, the claim is that this change is not at the root of the current properties of negative concord in contemporary French. 1.4. Micro vs Macro-parametric perspectives The points made in this section relate to more general issues on language variation and the nature of parametric distinctions. In short, following in Jespersen’s footsteps, Zeijlstra’s principle (2) amounts to claiming that there is a macro-parameter that distinguishes negative concord languages from non-negative concord ones, and that the formulation of this macroparameter centers on the properties of sentential negation. In contrast, the perspective defended here is that the phenomenon of negative concord arises through the convergence of micro-parametric settings that constrain the inner structure of n-expressions with predicted consequences on their semantics and on how they interact with sentential negation. Many approaches to negative concord have, at some level, recognized the necessity of understanding the interactions between the micro-level of negative relations, i.e. the nature of the n-expressions, and the macro-level of their e¤ects at the sentential level. It seems, however, that the majority of approaches proposed in the recent generative literature have followed a Jespersonian perspective and focused on a macro inside-out perspective that concentrates on the syntactic or semantic properties of sentential negation as the locus of the crucial defining factors of the nature of negative concord relations. This was true also of approaches based on a NegP constituent and the Neg Criterion that sought to draw similarities between negative relations and wh-dependencies.5 It is both my hope and my conviction that looking at negative relations from, so to speak, the other side of the lens, i.e. from the micro to macro outside-in perspective can but bring

5. In De´prez (1999), I provide a thorough criticism of this approach, showing that the syntactic properties that characterize negative concord relations are quite distinct from those of wh-dependencies. I demonstrate in particular that negative concord is subject to locality constraints that are much closer to those of QR than to those of wh-movement. Quite strikingly moreover, question words display a language internal homogeneity in behaviour that contrasts with the diversity found with n-words, suggesting that wh-dependencies may be more amenable to a macro-parametric variations approach than negative dependencies. That is, while it is common to find negative dependent terms that di¤er in behaviour within a single language, the same is not true for wh-terms. To my knowledge, it is not common to find wh-terms in a given language that must undergo wh-movement, while others must remain in situ.

An outside-in micro-parametric approach to negative concord

231

fruitful new discoveries. Although it might seem at the outset that detailed studies of the inner structure of n-expressions could bring out awareness of greater diversity than has so far been acknowledged, with the consequence of putting some previously held generalizations on shaky grounds, I am convinced that such an approach will bring in the end the clear potential of new generalizations that will deepen our level of understanding of the atomic structure of negative relations. It is as if after spending many years successfully characterizing and refining the overall syntactic properties of negative relations, it was now time to delve into their details to try to understand how the broad characteristics of these relations, which are now rather well known, could be deduced at least in part from the microstructure and inner relations of their composite elements. 2. Testing the empirical grounds In this second part of the paper, empirical evidence of the importance of the inner structure of n-expressions is examined. We first discuss evidence of inner language diversity in a fairly strong case of a so-called ‘strict negative concord language’, namely Haitian Creole. We then turn to historical evidence of the evolution of the inner structure of French negative expressions. In both cases, what is emphasized is how much the inner nature of n-words matters for an understanding of their contribution to the fabric of negative relations. 2.1. Inner language diversity and negative concord In previous work (De´prez 1997, 1999, 2000), negative concord systems have been distinguished on the basis of 1) the internal syntax of their n-words and 2) the semantic interface that these syntactic structures imply. It was argued that n-words in languages like Haitian Creole (HC) occupy a low position in the functional structure of their nominal constituent, manifesting syntactic and semantic properties relating to those of bare nominals. Negative concord there involves an operator variable system in which negation binds a bare nominal variable. By contrast, in languages like French n-words are placed high in their DP, behaving syntactically like determiners with the semantics of quantificational expressions. Negative concord in this case is quite distinct and involves resumptive quantification between negative quantifiers that are syntactically and semantically parallel, as proposed for English in May (1989). These two systems instantiate polar extremes between which micro-variations unfolds (De´prez 1999, 2000).

232

Viviane De´prez

Motivation for this proposal was based on the observation that although n-words share the appearance of bare nominals in both languages – pesonn, anyen in HC and personne, rien in French – the general licensing conditions on bare N-arguments radically di¤er in each, bare- arguments occurring freely in HC, and not at all in contemporary French. Assuming a null determiner structure for bare-N, this di¤erence implies that null-D are fully licensed in HC, but generally banned in French, leaving for French the structure (1b), in which the bare n-words occupy the determiner position.6,7 (4) a. b.

[DP D 0 [NP pesonn ]] Haitian Creole [DP Personne [NP N 0 ]] Contemporary French

Because they contrast Haitian Creole and European French n-words, these previous works seemingly perpetuate the impression that what is compared are again homogeneous languages that have uniform characteristic properties throughout. However, this is misleading, as these previous studies clearly concentrated on a restricted set of n-words, i.e. the nominal ones and specifically focused on their syntactic properties. Recall, however, that as argued above, the micro-parametric approach defended here predicts that distinct n-words may well have distinct properties so that within language diversity is expected if there are distinct inner structures for negative dependent terms. It has often been noted in the literature that n-words in French are not all homogeneous in their ‘negative’ behaviour, and that they do not all participate in negative relations in exactly the same way. For instance, it has been commonly shown that the n-word jamais behaves somewhat differently from personne and rien, which themselves are not entirely parallel either. In particular, for many speakers including myself, jamais can still be found in formal register (mostly written) with a non-negative meaning 6. The analysis of French n-words as determiners is also independently proposed in Sleeman (1996). 7. Zanutini’s (1991) original approach di¤ered from other NegP, Neg-Criteron based approaches such as Rizzi’s and Hageman’s in attempting to correlate the properties of concord to parametric variations in the syntactic position of negation with respect to VP. Although this approach fails to identify a macroparameter between languages allowing negative concord or not, it may well play a renewed role within a micro-parametric approach as one of the factor that can influence the interpretation of negative dependencies. How this factor could be connected to the internal structure of n-words, however, remains to be explored.

An outside-in micro-parametric approach to negative concord

233

in some characteristic NPI contexts such as, for instance, the rhetorical question in (5) and the contexts in (6): (5) a. b. (6) a. b.

c.

A-t-on jamais vu pareilles choses? 8 ‘Has anyone ever seen such things?’ [. . .] mais le ferai-je jamais? ‘but will I ever do it?’ Je ne crois pas qu’il soit jamais possible [. . .]. ‘I don’t think it will ever be possible.’ [. . .] dialogue ou` Socrate se montre plus sophiste que jamais [. . .] ‘a dialogue in which Socrates reveals himself more sophist than ever’ Si jamais tu venais [. . .]. ‘If you ever came’

Such is not the case for the n-word rien for instance. 8. Note, interestingly, that such rhetorical contexts also permit remnant bare NP ( pareilles choses), a fact that we have argued goes along with an NPI interpretation of n-words under the assumption that both sport a remnant null D (De´prez & Martineau 2004). Of further interest is the form of these questions. The examples provided in the text, some of them borrowed from MajBritt Mosegaard Hansen (this volume), and thus drawn from the Frantext data-base in a period from 2000–2007, manifest characteristic stylistic features that clearly flag them as part of a formal high (written) style. These features include: inverted question structure, use of bare nouns, which are notably absent from informal oral registers. It is interesting to note in this regard, that at least in my variety of French, jamais, when used in a non-inverted yes-no question, has a negative meaning. Thus a question like (i) must be answered with si, and cannot be answered with oui, a fact that clearly flags it as negative: (i)

T’as jamais mange´ des sushis? You have never eaten sushis? Si, mais j’ai pas aime´. Yes, but I did not like it.

Questions with an inverted structure that contain an n-word cannot be so answered, since they essentially have a rhetorical meaning. (ii)

A-t-on jamais vu chose pareille? #Si. Has one ever seen such things? #Yes.

234

Viviane De´prez

(7) a. b. c.

N’a-t-il rien vu? – Si. ‘Did he see nothing? Yes.’9 Je ne crois pas qu’il ait rien fait. ‘I don’t believe he did nothing.’ Si tu ne fais rien [. . .]. ‘If you do nothing’

This illustrates the fact that distinct n-words within a given linguistic variety may have distinct properties and as a result, enter in di¤erent compositional processes in given contexts. It is entirely possible within our approach that jamais presents a more ambiguous internal structure than other contemporary French n-words. Indeed, the hypothesis made in De´prez (1999, 2000, 2004) that the change of nominal n-words in French involves the suppression of a null determiner structure does not immediately transpose to the internal structure of adverbial n-words, unless it could be argued following Larson (1985) that adverbs in general, and jamais in particular, have a hidden nominal structure. If so, the disappearance of bare nominals in French, which we argue to be at the root of the change of French n-words, may not directly concern adverbs, and it could well be that, for n-words like jamais, their historical development towards a negative quantifier was only triggered indirectly by some similarity with other n-words like rien and personne and not directly by the disappearance of bare nouns. On this view, note that the possibility that jamais still allows a structure that other n-words have eliminated is not unexpected. Of course, such conjectures, although quite plausible, remain at this point tentative, and only a careful study of the internal structure of adverbs in general and of jamais in particular, could provide support for it. Such a study, however, is quite beyond the scope of the present paper. Of direct interest to the general thesis of the paper, however, is the fact that n-words in French, whatever the French variety considered, are rarely fully uniform in their properties. To wit, despite many similarities, the very obvious fact that rien generally occurs in a non-argument position, being subject to something akin to the tous-a`-gauche construction (Kayne 1977) while personne is not, at least in the contemporary standard language.10 This could well reflect a slight di¤erence in internal structure, in 9. Yes in (5a) above imperfectly translates French si, an a‰rmative answer that can only be used in the context of a negative meaning. 10. In Geneva French, personne is apparently found in the same type of preverbal structure as rien (Ur Shlonsky, personal communication).

An outside-in micro-parametric approach to negative concord

235

relation with the fact that while personne still has a nominal counterpart in contemporary French, rien no longer does. A possible structural encoding of this di¤erence would preserve for personne the trace/copy of an internal displacement from its original noun position, while there would be none for rien as in (8): (8) a. b.

[ DP personne [. . . [ NP personne]]] [ DP rien]

With this structure, rien is closer to clitic pronouns like le that have often been analyzed internally as intransitive determiners (Postal 1966). Confronted with this well-known variability in the behaviour of French n-words, Zeijlstra (2008) suggests that this state of a¤air is symptomatic of languages in which sentential negation is currently evolving. On his view, French is exceptional because its sentential negation is currently undergoing change (cf. the loss of ne). As noted above, however, such a correlation is not supported by available historical evidence. Moreover, this view makes the prediction that only languages whose negation is currently ‘changing’ should manifest inner language variability. As demonstrated below, however, such a prediction is not confirmed either. There are clearly languages in which negation appears quite robustly constant while nonetheless manifesting intriguing inner language variation with respect to distinct n-words. Below, such a case is detailed with a careful look at the properties of the equivalent of jamais in Martinique Creole. As shown in De´prez & Martineau (2004), the behaviour of n-words like pesonn and anyen in Martinique Creole is quite uniform, supporting general description of this language as instantiating strict negative concord of the variable binding type, with n-words in a low nominal position as in (2). There are nonetheless some very interesting facts about the n-word jamn that are at odds with too broad a generalization of this picture. As some of our informants have pointed out to us, jamn does not fully behave like the other n-words in their Creole variety11. N-words such as as person, anyen, oken moun require the obligatory presence of the sentential negation pa when they occur in declarative contexts, behaving in this regard

11. The data in this section comes from original fieldwork realized in (2001) under our supervision by Timothee Montoute, a native speaker of Martinique Creole with relatives of his. This fieldwork was funded by the European Project Eurocores, granted to Peter Ford Dominey and myself.

236

Viviane De´prez

like standard NPI expressions.12 For jamn in contrast, the co-presence of pa appears to be optional: (9) a. b.

Man ( pa) janm dir u bagay kon sa. ‘I never told you things like that.’ Man janm ale Matinik. ‘I never went to Martinique.’

What is of particular interest here is that this optional co-occurrence requirement correlates for jamn with the possibility of giving rise, in certain contexts, to a double negation reading in the co-presence of pa. (10) a. b.

Man pa janm ale Matinik. ‘I didn’t never go to Martinique. (In fact, I went often.)’13 Je ne suis pas jamais alle´ a` la Martinique. (En fait, j’y suis alle´ souvent.)

In other words, in this variety of Martinique Creole jamn seems to fluctuate between an NPI like behaviour, where it combines with sentential pa to produce a single negation reading that is parallel to that of the other n-words of this language, and a negative quantifier behaviour that is closer to the one observed with French n-words and produces a double negative reading when combining with sentential negation. To our knowledge, however, there are in MC no comparable known fluctuations in the behaviour of sentential negation. Pa is a preverbal head in present day MC, and to our knowledge has always been in the available historical records 12. Haitian Creole manifest so called ‘strict negative noncord ’, which means that n-words are also found in subject position with a co-occuring negation, in contrast to so-called standard, i.e, English-like, NPIs, which cannot occur in subject position. In De´prez (2008), we proposed that n-words licensing in subject position requires the conjunction of two factors: 1) a negative operator that is a head, and 2) DP internal licensing of a null D. If one of these conditions is not met, subject/object asymmetries in NPI licensing are expected. In English, for instance, the sentential negation operator is arguably not a head. In Italian, by contrast, it is DP internal null D licensing that is not possible. Thus for both languages, NPI and n-words display subject/object asymmetries, but this same e¤ect may result from distinct constraints in the two languages. 13. Such a contrastive negative statement requires of course an appropriate context. This is the one provided by our informant: ‘You don’t even know our foods. I bet you have never ever been to Martinique!’

An outside-in micro-parametric approach to negative concord

237

(see for instance Hazael Massieux 2008). The survey we conducted of available historical texts on the internet14 has not revealed any di¤erences. What this illustrates is a characteristic inner variability within an otherwise rather homogeneous language, with a sentential negation form that has remained stable since its known formation. One could argue that this situation illustrates de-creolisation, and that jamn for our informants has been subject to influence from French. This is indeed quite likely, given the situation of diglossia that is typical of Martinique nowadays. Even so, the interesting point here is that such a decreolizing influence seems to have a¤ected only one n-word and not the whole class of these items. Nor does it concern the behaviour of negation, which in the texts surveyed by Pascal Vaillant (2009) does not occur in post-verbal position (Pascal Vaillant, personal communication). Clearly, in this variety the relation of the n-word jamn to negation is distinct from the one exhibited by the other n-words, but one would be hard pressed to argue that this can be explained by distinct properties of sentential negation. More plausible is the suggestion that it is the internal structure of the adverbial jamn that is in part di¤erent from that of nominal n-words. Incidentally, it is quite striking that this distinction should concern the equivalent of jamais, which itself, as noted above, shows a greater instability in French too. Put together, these facts clearly suggest that adverbial n-words behave somewhat di¤erently from nominal ones. What the exact structure of adverbial n-word is, and how they di¤er from nominal ones no doubt deserves further in-depth investigation. But the fact that interesting distinctions appear to group distinct n-words by category, i.e. nominal vs. adverbial ones, already provides interesting support to the view defended here that it is the internal syntax of n-words that governs the nature of the negative relation they are involved in, rather than the nature of the sentential negation maker. 2.1.1. Looking inside n-words Support for our micro-parametric approach to negative concord requires evidence of robust correlations between the inner syntax of n-words and distinctive properties of the concord relation. But for this to be shown, it is of course necessary to provide first and foremost a detailed analysis of the internal structure of n-words. The goal of this section is to provide 14. The old text consulted have been made publicly available by Hazael-Massieux at http://creoles.free.fr/Cours/

238

Viviane De´prez

such an analysis for standard contemporary European French n-words focusing on their feature composition and their modification possibilities. Once the structure is established on solid empirical grounds, the paper then o¤ers an in-depth exploration of the changes that have a¤ected the internal structure of French n-words in the course of their historical development. It attempts to chart the detailed diachronic evolution of the internal structure of n-words during their diachronic grammaticalization and discusses the correlations that these changes present with changes in the concord relation they participate in. This section is structured as follows. First, we focus on providing empirical evidence for the complex structure of French n-words as opposed to their nominal counterparts. We examine some distributional evidence, then discuss their feature composition and modification possibilities and finally we take stock of what these data mean for the internal structure of French n-words. In section 3, we turn to historical changes and provide evidence of how French n-words have evolved from a bare nominal type to their contemporary quantificational and determiner type through internal changes in their structure. We examine the time course of feature changes and changes in the modification possibilities. 2.1.2. Distributional evidence At the outset, it is important to recall that the morpheme personne still has a double life in contemporary French. On the one hand, it is a well behaved common count noun meaning ‘person’, and on the other hand, it is an n-word with an independent strong negative interpretation. A careful comparison of the two words can thus determine minimal factors that enter in their distinction. Given that the frequency of these items is roughly the same,15 how do speakers know which is which? A central distinguishing factor is that personne has a (positive) nominal interpretation whenever it co-occurs with a determiner in an argument position (11a). Conversely, its interpretation is generally that of an n-word, whenever it occurs in argument position with no determiner as in (11b).

15. Our assertion here is based on a rough estimate of occurrences of personne in the Elicop corpus. In each of the files of this corpus that we considered and tallied, the count of negative personne yielded approximately the same result as the count of the positive uses.

An outside-in micro-parametric approach to negative concord

(11) a. b.

239

J’ai rencontre´ une/cette/la personne. I-have met a/this/the personne J’ai rencontre´ personne a` cette soire´e. ‘I have met no-one at this event.’

The complementary distribution between the presence of overt determiners and the n-word interpretation of personne is quite general, apart from a few interesting cases to which we return below, indicating a competition for the same syntactic position that quite clearly supports placing the n-word personne in a determiner position. Besides accounting for (11), this also provides a straightforward explanation of how a seemingly bare nominal like personne succeeds in meeting the stringent French requirement for obligatory determination. Personne does not need a determiner, whenever it is itself in a determiner position. Worth stressing is furthermore the fact that the presence or absence of a determiner is far more reliable a cue to the interpretation of personne than the co-occurrence of a negative element like ne. As is well known, the presence or absence of ne has no impact whatsoever on the n-word interpretation of a bare personne. Similarly, adding ne to a determined personne as in (10) leaves its nominal interpretation intact: (12) ?? Je n’ ai vu une/la personne. I ne saw a person Quite clearly then, the presence of a negative operator, even the strong negator pas, does not influence the interpretation of personne. In fact, a subtle but nonetheless patent semantic distinction remains, even when a negated nomimal de personne in (13a) is compared with an n-word construction like (13b). To get a feel for it, consider the following scenario. Suppose that aliens that look like humans had invaded our world and were hunting the remaining humans they call ‘persons’ for final annihilation. Now imagine an alien hunter looking for humans in a crowded alien party. If he came up empty handed, he could truthfully utter (13a), but surely not (13b), given the crowd present at the party. (13) a. b.

Je n’ai pas vu de personne(s) a` cette soire´e. ‘I did not see people.’ ¼ (humans) Je n’ai vu personne a` cette soire´e. ‘I did not see anybody at this party.’

240

Viviane De´prez

Only in (13a) does the lexical meaning remain prevalent, as is expected if personne is indeed a nominal predicate here only. (13a) is of even further interest because of the peculiar [de NP] structure that it features. Kayne (1977) has argued that this construction sports a remnant null determiner, an analysis supported by the fact that, like other attested null D structures, such as Italian bare NPs, [de NP] constructions manifest a characteristic distributional subject/object or pre-verbal/post-verbal asymmetry, as (14) illustrates: (14) a. *Je ne crois pas que [de personnes e´trange`res] seront invite´es. I ne think not that of people foreign will be invited ‘I don’t think foreign people will be invited.’ b. Je ne crois pas qu’il sera invite´ [de personnes e´trange`res]. I don’t think that there will be invited people foreign ‘I don’t think there will be foreign people invited.’ The [de NP] structure further manifests a dependency on the presence of an overt c-commanding operator, here negation. This is expected if their remnant null D is interpreted as a variable that must be appropriately bound for interpretation. In short, distributional asymmetry and necessary operator binding are a hallmark characteristic of remnant null D. Since n-words, by contrast, manifest neither of these characteristics, this is evidence that their structure does not contain a comparable null D. This much secures the first part of an empirical demonstration that bare n-words do not have the syntax of bare nominals. To yet strengthen the argument, further16 empirical evidence is needed that n-words such as personne must be in D. Such evidence can be found, on the one hand, in the interesting exceptions to the complementary distribution mentioned above and, on the other hand, in the modification facts discussed below. Turning to the exceptions first, (15) shows a surprising instance of bare personne that can only be interpreted as positive: (15) Vous (?? ne) recevrez un livre par personne. ‘You will get one book per person.’ 16. Note that the n-word within a DP in (i) can lead to negative concord reading although the presence of ne and, hence, presumably of any negative operator (even null) is in fact banned. (i)

Le *(ne) don de rien a` personne est de nos jours presque un e´goı¨sme de rigueur. The gift of nothing to no one is nowadays quasi-standard egotism

An outside-in micro-parametric approach to negative concord

241

In (15), despite its bareness, personne cannot be coerced into being negative. Yet as (16) shows, not all PP constructions have this e¤ect on a bare personne: (16) Je (n’) y suis pour personne. ‘I am there for no-one.’ So why and how do (15) and (16) di¤er? Replacing personne by other nominals, we quickly see that, while (17a) displays the properties of a regular PP, taking all sorts of nominal as complements, this is not the case with (17b). In (17b), the range of nominal complements allowed is narrowly restricted, and in e¤ect strictly limited to nominals without determiners: (17) a.

b.

J’y suis pour les/(tous) mes/des amis/, pas pour les/(tous) mes/des ennemis. ‘I am there for the/my/some friends, not for the/my/some enemies.’ Vous recevrez un livre par *le/mon/un enfant. ‘You will receive one book per the/my/a child.’

(17b) shows that the presence of any overt determiner is simply excluded in these constructions. Yet, as (18) shows, the construction allows adjectival modification, which excludes an analysis of (15) involving N to P incorporation or direct X0 selection. (18) Vous distribuerez un cadeau par gros client fortune´. ‘You will distribute one present per big wealthy customer.’ Constructions as in (15) thus appear to strictly select bare singular NP complements17, banning any nominal phrase with a more complex determiner structure. This recalls Pereltsvaig’s (2008) observation that some Russian prepositions strictly select small nominals with no DP projection. Returning to n-words, note that if the n-word interpretation of personne were compatible with a bare NP structure, its exclusion from (15) would 17. That the complement of such prepositions must be singular is demonstrated by (i) below, where we see that a phonetically distinct form of plural is unacceptable: (i)

Vous mettrez un ballot d’avoine par cheval/*chevaux. You will put one straw pile by horse-SG/*horse-PL

242

Viviane De´prez

be mysterious. If, in contrast, it requires a complex D-structure, as defended here, the facts are entirely as predicted, since a complex D-structure fails to meet the selection restriction of this PP construction. (15) further eliminates a putative analysis of bare personne as an N-level pronoun and makes the point that not all apparent bareness can be treated equally, providing clear empirical evidence that the bare n-words of contemporary French must be D-like structures. Let us now briefly see how these observations about personne extend to other nominal French n-words and, in particular, to rien. Although rien no longer has a comparable common nominal counterpart in contemporary French, it is nevertheless not hard to find clear evidence that support extending to it the structural conclusions reached above. First, the complementary distribution with determiners is still, to some extent, observable for rien in examples like (19). (19) Il est tre`s sensible. Un rien le fait sursauter. ‘He is very sensitive. A little thing makes him jump.’ In (19), the nominal meaning of rien (thing) is still perceivable and, as expected, the presence of a determiner is required to allow it. Moreover, although (20) unlike (15) is strictly speaking not fully acceptable, it is nonetheless significantly improved by the addition of a pre-nominal modifier. (20) a.

*Vous utiliserez un sac par rien. ‘You will use one bag per little thing.’

b.

?Vous utiliserez un sac par petit rien. ‘You will use one bag per little thing.’

(20) illustrates that although the nominal equivalent of rien is no longer in use in contemporary French, it can still be coerced provided that its nominal structure is made obvious by the presence of a pre-nominal adjective, forcing rien, in this case, to be construed in a low nominal position. Note that remarkably, in this case rien loses its negative import. Hence, the reasoning applied to (15) above extends to (20), leading us to conclude that the n-word rien, just like personne, sports a complex determiner structure that makes it unsuitable in (18). The tests used above do clearly not apply to the French nominal nword aucun, which, quite plainly, must be a D. We return to this n-word below, showing that despite some di¤erences, it nonetheless shares structural similarities with the bare n-words discussed above.

An outside-in micro-parametric approach to negative concord

243

2.1.3. The features and modification of n-words This section first o¤ers a feature comparison between n-words and their nominal counterparts and then compares the modification possibilities allowed for each type of expression. Turning to features first, we consider gender and number. In (21a), the nomimal personne is feminine and can vary in number, a property here reflected by the determiner, adjectival agreement and co-referring pronouns. The n-word in (21b), in contrast, triggers no adjectival agreement and has corresponding singular and masculine bound pronouns. (21) a.

b.

Les personnes intelligentes pensent qu’elles n’ont pas toujours raison. ‘Intelligent persons think they-FEM are not always right.’ Personne d’intelligent (ne) pense qu’il a toujours raison. ‘No one intelligent thinks he is always right.’

In this respect, n-words behave like the existential quantifier quelqu’un (someone), also unmarked for gender and number. Interestingly, despite number invariability, n-words seem semantically plural, as shown by (22), where personne occurs with a verb that requires a plural subject. (22) Personne ne se rassemble plus ici. ‘No-one gathers here anymore.’ In what respect do these facts support the view that n-words occupy a high D-like position in their functional structure? Note first that the lack of gender and number features make n-words characteristically un-nominal. Moreover, syntactic invariability and/or semantic plurality appears to be characteristic of a certain type of D-elements in French such as the existential quantifier quelqu’un or the bare quantificational pronouns or question words qui and quoi. These similarities support the view that nwords share feature properties with pronouns and quantifiers, but not with nouns, hence providing evidence that they are merged in a D position.18 Further support arises from their modification properties. (23) 18. The view that n-words are pronouns has long been recognized in the classic literature (Gougenheim 1951; Grevisse 1980). The claim that n-words are in D, adds a hierarchical dimension to this claim that has consequences on adjectival modification.

244

Viviane De´prez

shows that pre-nominal adjectives are excluded with n-words, while they are possible and stackable with their nominal counterparts. (23) a. b.

*J’ai rencontre´ autre charmante petite personne. ‘I have met other charming small nobody.’ J’ai rencontre´ une autre charmante petite personne. ‘I have met an other small charming small person.’

In the spirit of Cinque’s (1994) hierarchical approach to modification, the lack of pre-nominal modification is a further indication that these n-words must occupy a position at least higher than NP in the DP structure. But it is not just pre-nominal modification that is excluded with n-words. More generally, n-words like personne and rien do not allow direct adjectival modification; they require a specific modification construction with de as in (24), which I henceforth term ‘indirect de-modification’. As (24) shows, this again is a property that n-words share with existential quantifiers like quelqu’un and bare question words like qui or quoi.19 (24) *Personne/rien/quelqu’un/qui/quoi *(d )’inte´ressant ‘No one/nothing/someone/who/what (of ) interesting’ Kayne (1994) observes that indirect de modification is more constrained than regular modification, in that it disallows stacking, as in (25): (25) *J’ai rencontre´ personne/quelqu’un/qui de charmant de petit. ‘I have met no one/someone/who/of charming of small.’ He argues that this anti-stacking constraint speaks against assigning these constructions a traditional adjunction structure and following Huot (1981), proposes a structure akin to relative clause modification with de as a complementizer-like element as in (26): (26) [D 0 [ DP/PP quelqu’un/personne [ de [ IP intelligent I 0 t ]]] (Kayne 1994) In (26), the n-word or quantifier raises from a position internal to the clause to Spec de, in a fashion paralleling that of predicate inversion in DP (den Dikken 1998). Kayne, following (Huot 1981), notes one intrigu19. As De´prez (2000) notes, indirect modification of this type is not possible with universal quantifiers, a fact that strongly suggests that n-words are like existential/indefinite quantifiers or numerals in their syntactic nature, i.e. quantificational indefinites.

An outside-in micro-parametric approach to negative concord

245

ing exception to the stacking constraint that occurs with the modifier autre: (27) Personne/rien/quelqu’un/qui/quoi d’autre d’inte´ressant ‘No one/nothing/someone/who/what of else of interesting’ D’autre modification is also possible with quelqu’un or question words, but, expectedly, not with the nominal personne, for which modification by autre must be strictly pre-nominal and direct:20 (28) Une autre personne (*d’autre) ‘An other person (of other)’ The modifier autre is peculiar in other respects that are of interest here. First, like a few other pre-nominal adjectives in French (divers, di¤e´rent), autre can sometimes play the role of a determiner and/or a pronoun. (29) Autres temps, autres moeurs, autre histoire! ‘Other times, other mores, other stories!’ Clearly as well, it must always be the highest pre-nominal modifier in any noun phrase, as it can never be preceded by any other pre-nominal adjective, like for instance petit. (30) *Une petite autre voiture/une autre petite voiture ‘A small other car/another small car’ In these respects, autre di¤ers from regular adjectives, appearing instead to share properties with numerals. In fact, like numerals, it can modify pronouns. (31) nous trois, nous autres ‘we three, we others’ Interestingly, however, autre can also modify numerals, as in (32a), but the stacking of pronouns and numerals is not allowed (cf. 32b): (32) a. b.

trois autres *nous trois autre

‘three others’ ‘*we three others’

20. Post-nominal autre is also marginally possible, but with a meaning closer to that of di¤erent: une personne autre.

246

Viviane De´prez

Furthermore, while indirect de modification is sometimes possible with regular adjectives where the noun phrase is contrastively focused as in (33), indirect modification with d’autre is always excluded. With nouns, autre must always remain pre-nominal with direct modification. (33) J’ai trois voitures de bleues/*d’autres, pas quatre. ‘I have three car of blue/of other, not four.’21 In sum, indirect modification with d’autre is only possible with n-words, existential quantifiers and question words, never with other nominal expressions. This clearly gives it a peculiar status, shared with no other adjectival modifiers. Noting this peculiarity, Huot (1981) suggested that autre involves quantity, not noun modification, and proposed the structure in (34): (34)

Huot’s suggestion in e¤ect groups modification by d’autre with other degree modifiers like de plus (more), or de moins (less) (quelqu’un/quoi de plus/de moins ‘someone/what more/less’). In a current model, a possible

21. Indirect de-modification is also possible with so-called ‘quantitative en-pronominalization’. As (ia) shows, for regular adjectives, both direct and de modification are possible. Here again, however, modification with d’autre is excluded: (i)

a.

J’en ai TROIS (de) petites, pas quatre. I of them have three of small, not four

b.

J’en ai trois (*d’ ) autres. I of them have three of other, not four

An outside-in micro-parametric approach to negative concord

247

interpretation of this insight would assume that d’autre involves NumP22 and not NP modification. To account for the above modification possibilities as well as for its similarity with numerals, we propose that autre can occupy the head of NumP, so that it can be preceded by elements that occupy either the Spec of NumP or the head of D. We have seen Kayne’s (1994) proposal for indirect de modification with regular adjectives. Adapting Huot’s suggestion, we propose to account for the stacking e¤ect by assuming that autre can be a NumP head modifier that can move along with the head out of the clausal constituent of the indirect adjectival demodification and be stranded in Spec DP, while the n-word moves even higher in the DP structure as in (35a), in a fashion akin to CP recursion. (35) a. [ FP personnei [ DP/PP de [ ti autre ti ]k [ de [IP sympatique Infl tk ]]] noone of other of friendly ‘no one else friendly’ b. [ Quii [ ti d’autre]] k as tu invite´ [de sympatique tk ] ? who of other has you invited of friendly ‘Who else friendly have you invited?’ c. [ Quii as tu invite´ [[ ti d’autre] [de sympatique ti ]] who has you invited of other of friendly ‘Who else have you invided that is friendly?’ d.

Je n’ai rieni vu [[ti d’autre] [d’inte´ressant ti]] I ne-have nothing seen of-other of-interesting ‘I have seen nothing else interesting.’

That d’autre can indeed move along with a quantity expression is confirmed by (35b), where d’autre has moved along with a wh-term to SpecCP. That movement out of the projection containing the modifier is also possible is further confirmed by (35c), where d’autre appears optionally stranded by the moved wh-term. (35d) replicates this structure with the moved n-word rien.23 (35a) accounts, on the one hand, for the exceptional stacking possibility with d’autre and on the other hand, for the 22. Throughout, we assume NumP to be the projection that hosts numerals and weak quantifiers, not number. Number in our view is a separate PlP projection as in Zamparelli’s more recent work. 23. Curiously and interestingly, stranding in this case is obligatory. I do not at this point have an explanation for this fact.

248

Viviane De´prez

restricted nature of the elements that allow it. Only elements moving out of NumP to a higher position in DP, here termed ‘FP’, manifest this DP recursion possibility. To sum up, we have shown that the modification properties of n-words are very di¤erent from those of regular nouns. The facts reviewed above provide evidence for the quantificational nature of n-words and for the proposal that they occupy a high position within DP. We have spent some time detailing the peculiar facts of d’autre modification here because, as discussed below, there are revealing changes in this respect in the evolution of the internal structure of n-words. 2.1.4. Mapping the structure The above discussion has provided some empirical evidence for the complex internal structure of n-words, but has so far remained theoretically imprecise with respect to the detailed mapping of their internal structure. This section presents an e¤ort to flesh out this internal structure in more details. For a start, we adopt Zamparelli’s (1995) proposal to subdivide the DP structure in three hierarchical quantificational zones that he termed ‘SDP’, ‘PDP’, and ‘KIP’. For Zamparelli, only weak quantifiers that are housed in PDP are compatible with ne cliticization in Italian, because only these are close enough to license the trace of ne. As Zamparelli notes, the Italian quantifier qualcuno is ambiguous in this respect. When compatible with ne cliticisation, qualcuno simply is an equivalent of some. When incompatible with it, qualcuno is restricted to human reference and is a pronominal meaning ‘someone’. To account for this ambiguity, Zamparelli proposes that qualcuno is merged as a weak quantifier in PDP/NumP and has the possibility to move or remerge higher as a strong quantifier in SDP/DP. Adapting this approach to French, we note that the ambiguity observed for qualcuno is replicated by quelqu’un vs quelques uns. Like qualcuno, which means ‘someone’, singular quelqu’un fails to allow partitive/quantitative en clitization and has the same meaning (cf. 36). Quelques uns, by contrast, has the meaning of ‘some’ and is compatible with partitive and with en cliticization and inflected for number. (36) J’en ai rencontre´ quelques uns/*quelqu’un/ *quelqu’une de mes amis. ‘I have met some-PL/someone/some of my friends.’ I thus suggest that, similarly to qualcuno þ ne, quelques uns is merged in PDP/NumP, while quelqu’un is either directly merged or obligatorily moved

An outside-in micro-parametric approach to negative concord

249

to SDP/DP. Since, as seen above, the n-words personne and rien behave in many respects like the existential quantifier quelqu’un, we now have analogical evidence to posit the same structure for these contemporary n-words. Note that, as expected on this proposal, n-words are essentially incompatible with true partitivity and with quantitative en cliticization (37): (37) a. b.

?? Je ne connais personne de tes amis. ‘I know no one of your friends.’ *Je n’en connais personne.24 ‘I know no one of them.’

Assuming that the similarities warrant a parallel treatment, I propose the following structure for the French n-words: (38)

(38) accounts for all the properties of the n-words detailed above, as it assigns them to the highest position in DP and implies that they behave essentially as strong existential quantifiers. On this view, their invariability is expected, given that they are merged above the DP layer where number agreement is syntactically negotiated, as are their modification possibilities, given that they behave essentially as quantity expressions.25 For indirect de-modification, we have adopted Kayne’s (1994) structure to which we have added a layer to allow modification by d’autre as in (33a) above. On this view, d’autre modification and the lack of partitive structure can be taken as evidence of the additional movement step towards the highest layer of SDP/DP.

24. One should not confuse true partitive constructions, which take a full DP as a complement, with pseudo-partitive ones. The latter is permitted and commonly derives apparent quantitative en contructions with rien (cf. De c¸a, il n’en reste rien/peu de chose ‘of that there is nothing/little left’ vs. ?*De tes jouets, je n’en casse rien.’ of your toys, I broke none’). 25. Recall that as posited in section 2.1, rien di¤ers from personne in being directly merged in this high position and leaving no trace of any displacement.

250

Viviane De´prez

To be complete, our study of the structure of contemporary French nominal n-expressions requires that one more expression, namely aucun N, be considered. As this expression and its historical evolution is discussed in great details in De´prez & Martineau (2004), our discussion here will remain brief. In contrast to personne and rien, aucun N clearly contains both a determiner element and an obligatory N, so that the status of aucun as a determiner is not in doubt. Unlike personne and rien, however, this nominal n-word expression allows both quantitative en cliticization, and true partitive structures. (39) Je n’ai vu aucun de tes enfants. Je n’en ai vu aucun. ‘I have not seen any of your children. I have seen none of them.’ In this respect, aucun seems to behave both like a numeral in French and like the strong quantifier chacun, which can vary in gender, but notably not in number. (40) a. b.

Aucune ide´e n’a e´te´ accepte´e. ‘No idea was accepted.’ *Aucunes ide´es n’ont e´te´ accepte´e(s). ‘No ideas were accepted.’

To account for these di¤erences as well as for the similarities between aucun and the other nominal French n-words, I propose that the determiner part aucun is merged in PDP and moves to SDP as in (39), while the nominal part continues to behave like a regular NP. (41)

Thus, like numerals, aucun participates in a strong/weak alternation, but unlike numerals, it does not allow a preceding determiner in contemporary French (les trois, vs *les aucuns). This movement to the higher layer of the DP is obligatory for this n-expression, as in the case of rien/ personne. As Martineau & De´prez (2004) have shown, this was not always true in the history of French. Expressions like *Les aucuns amis ‘the some friends’ are commonly found up to the sixteenth century. This, and the fact that aucun is invariable in number, motivates an analysis in which this n-determiner is merged in the highest D position.

An outside-in micro-parametric approach to negative concord

251

To sum up, on the basis of their distributional, feature and modification properties, I have argued in this section that all nominal French n-words occupy the highest quantificational position in the DP and, as such, behave syntactically like determiners and strong quantifiers. With respect to some of their other properties, these n-words seem to partake both in the behaviour of strong and of weak quantifiers. To account for this, I proposed that they be merged in PDP, a position equated elsewhere with NumP, and move to SDP, a position equated elsewhere with DP. Of particular interest moreover is the fact that this high position correlates with the loss of number KIP variation. All the nominal French n-words are invariable in gender and number, and consequently always trigger singular or more exactly unmarked agreement. As will be seen in section 3, the loss of feature variability historically correlates with the loss of nominal status and the development of a pronominal and quantificational status. What we observe for all French nominal n-words is the loss of variability and the rise up the DP structure to the highest position, correlating with a gain in pronominal and quantificational strength. It is these combined properties that account, in our view, for their possibility to ambiguously partake in the complex polyadic resumptive quantification that produces the negative concord readings as well as in the scopal quantification that produce double negative readings as discussed in De´prez (2000), following May (1989). 3. The internal evolution of French n-words Having discussed in detail the syntactic nature and structure of contemporary French n-words, it is now time to turn to their historical evolution, in order to chart the course of their internal change as well as to identify the micro-parametric nature of this evolution. We o¤er here a detailed examination of the evolution of rien and personne focusing on two central aspects, changes in their featural make up and changes in their modification possibilities. Our study is based on searches in the electronic databases Frantext and TFA. 3.1. Changing Features Let us first consider feature changes. Both personne and rien started out as feminine and count nouns, and both ended up on their n-word interpretation as singular and masculine nominal expressions or, more exactly, as expressions unmarked for both number and gender. Charting the course of changes of this nature in historical corpora, however, is a di‰cult under-

252

Viviane De´prez

taking, as in French, cues to determine the gender and number make up of nominal expressions are accessible only indirectly, i.e. through inspection of their dependent satellites. In Old and Middle French, rien, which derives from the Latin noun res (thing), was a feminine noun that could refer to inanimate or animate entities. (42) provides an example from Chretien de Troyes thirteenth century of an animate nominal rien preceded by a feminine definite determiner. (42) Quant la rien voit que il plus ainme [. . .] when the thing sees that he most loves ‘When he sees the person he most loves’ (Chre´tien, de Troyes, 1240, Le Chevalier de la Charrette (Lancelot)) Examples with a clearly feminine determiner as la rien can be found up to the Middle of the sixteenth century in the prose text Cent Nouvelles Nouvelles (cf. 43): (43) [. . .] qu’elle voit d’elle e´loigner la rien en ce monde dont la presence plus luy plait [. . .]. ‘that she sees leaving her the thing in this world of which she most like the presence’ (Monseigneur de La Roche, 1550, Les Cent Nouvelles Nouvelles) This appears to be the last example of this sort found in Frantext, which suggests that rien became unmarked for gender around the middle of the sixteenth century. Indeed, confirming this, it is around this time that the first examples of rien with clear instances of masculine determiners are found, as attested in (44), an example from the poet Ronsard dated from 1552: (44) [. . .] elle auroit cognoissance Qu’un rien qu’on ne voit pas, fait souvent un grand [. . .]. ‘She would have knowledge that a thing that one cannot see often makes a large [. . .]’ (Pierre de Ronsard, 1524–1585, Les Amours) That rien becomes invariable for gender does not however mean that it completely loses its nominal character. Indeed, rien could still be found with a (masculine) determiner and a positive meaning up to almost the end of the classical period, as attested by the following 1896 examples from the famous Goncourt Brothers. Interestingly, rien is now strictly inanimate in reference:

An outside-in micro-parametric approach to negative concord

253

(45) [. . .] pour moi, le rien qui m’arrive d’heureux, c’est toujours [. . .]. ‘For me, the thing that happens to me that is happy is always [. . .].’ (Edmond and Jules de Goncourt, 1822–1896, Journal ) Thus, the loss of gender and animacy reference represents the first step in the grammaticalization of this expression. Turning to number, we face the same di‰culty in determining when exactly rien stopped allowing number inflection, as it can vary even nowadays, when it is a noun clearly accompanied by a determiner (cf. des riens). To get some idea we searched our databases for examples of rien overtly marked for number without a determiner. The last example where riens occurs as a clear bare argument occurs in 1559 in Frantext. (46) Le Roy ne luy respondit riens, sinon: ‘‘m’asseurez-vous que vous l’avez [. . .]. ‘The king did not answers things, except: ‘‘confort me that you have it’ (Marguerite de Navarre, 1559, L’Heptame´ron) Based on this indirect evidence, we tentatively suggest that it is around the end of the sixteenth century that bare rien e¤ectively ceases to be variable in number. But again, rien does not immediately cease to have a nominal character, since it can be found as an invariable bare nominal expression without determiner, with a positive meaning in downward entailing contexts up to the classical period: (47) Diable m’emporte si j’entends rien en me´decine. devil takes-me if I-understand anything in medicine ‘The hell if I understand anything about medicine.’ (Molie`re, 1673, Le malade imaginaire) The continuing nominal character of rien is confirmed by evidence based on modification possibilities, to which we turn in Section 3.2. Before, however, let us first detail the time course of feature change for the n-word personne. Turning to personne, the exact time course of the loss of its features is even more di‰cult to determine from corpus searches, given the continuous current use of its nominal counterpart. We have nonetheless endeavoured to chart an approximate course by looking at examples of bare personne triggering feminine agreement with an adjective, a past participle or a co-referring pronoun. In Old French bare personne was clearly feminine, as witnessed by (48) from the Vie de St Alexis (eleventh century) with a co-referring feminine pronoun.

254

Viviane De´prez

(48) Car la sainte Ecriture si nous dit et tesmoigne, because the Saint Writing here us tells and witnesses, Qu’au monde n’a personne, pour tant qu’ that-in-the world there-isn’t (a) person for as-much-as elle se joigne. she REFL joins ‘Because the Holy Bible tells us here that in the world that there is no one that would be found.’ (Sarre Nguissaly 2003:121) The capacity to trigger feminine agreement for bare personne seems to have lasted up to the seventeenth century, as witnessed by (49) with a past participle agreement from L’Astre´e (1607–1625): (49) Je ne puis dire avec ve´rite´ que jamais personne I ne can tell with truthfulness that ever person-FEM ne fut plys ayme´e que moi. ne was more loved-FEM than me ‘I cannot tell in truth that anyone was ever loved more than me.’ (H. Durfe´, 1607–1625, Astre´e, 536, in Nguissaly 2003:126) Yet, as early as the mid fourteenth century, there are also examples without agreement: (50) Jamais nouvelle n’en seroit a personne vivant. Never news ne it-GEN be-COND-3PSG to person alive ‘Never will this be news to any person alive.’ (Monseigneur de La Roche, 1550, Cent Nouvelles Nouvelles, 387, 73) More generally, it would seem that after the Renaissance, bare personne had definitely lost its capacity to be marked for gender. It thus came to generally trigger default masculine agreement and co-reference, as is shown in (51) from Pascal 1670, where both pronominal reference and past participle agreement show masculine and singular agreement: (51) Personne n’a d’assurance – hors la foi, s’il veille ou s’il dort. ‘Not anyone has assurance – outside faith, of whether he is awake or asleep.’ (Pascal, 1670, Pense´es, 164, in Nguissaly 2003:132)

An outside-in micro-parametric approach to negative concord

255

To sum up, evidence from our corpus searches reveals that both rien and personne seem to have lost their characteristic nominal features of, first, gender by the middle of the sixteenth century and, second, number at the end of that century. 3.2. Changing Modification This section focuses on historical changes in modification possibilities. I will argue that even more than feature loss, the changes in modification provide a striking window into the evolving internal structure of n-words. Going back to its earlier occurrences, rien was modifiable by a number of pre-nominal adjectival quantifiers, like nulle and toute, all of which attest to its feminine gender. In Frantext, the last example of nulle riens ‘no things’ is found in the fifteenth century in Christine de Pisan’s work: (52) Ne deu¨st faire nulle riens, Toute fois entr’eulx une riens [. . .] ‘I had to do no thing, yet between then one thing [. . .]’ (Christine de Pisan, 1402, Le Livre du chemin de lonc estude) There are 94 further occurrences of toute rien ‘all things’ in the TFA database. Our searches in this database revealed that this feminine modification was possible up to 1381. In Frantext, the last example of this modification is found in 1550, again in the Cent Nouvelles Nouvelles. (53) [. . .] et sur toute rien luy estoit [. . .] defendu le mestier [. . .]. ‘and on everything it was forbidden to him to take up the profession’ (Monseigneur de La Roche, 1550, Les Cent Nouvelles Nouvelles) It is of course di‰cult to identify regular modifiers in a corpus, which is why we chose to focus in our searches on modification by autre, which turns out to be particularly revealing in its evolution. In some early examples, we see that autre began as a prenominal modifier of rien as in (54). This is comparable to today’s modification of regular nouns. (54) Nous ne demandons autre riens Que nous y mettre. ‘We are not asking for other things than to put ourselves to it.’ (Anonymous, 1376, Miracle de sainte Bautheuch) The last example of this kind given here in (55) occurs in 1606. Rien in (55) is clearly masculine, and also plausibly invariable in gender, but still positive and here preceded by a determiner.

256

Viviane De´prez

(55) Dieu qui de rien fit tout, et qui de tout encore peut faire un autre rien, [. . .] ‘God who out of nothing made everything and who of everything can yet make another thing.’ (J. Bertaut, 1606, Recueil Qq. Vers Amoureux) Taking the pre-nominal position of autre to be indicative of the nature of rien, examples like (54) and (55) suggest that its nominal character, as indicated by its positioning in the N projection, lasted somewhat beyond the loss of its gender and number features, up to the beginning of the seventeenth century. Yet, the fact that no example of this sort occurred after this date can be taken to mark the end of a historical phase. From 1606 on, throughout the classical period and up to the early twentieth century, our data indicate that autre has become a post-nominal modifier of rien as in (56). (56) [. . .] elle me dit adieu, et se retira sans me donner rien autre chose qu’une bague [. . .]. ‘She said goodbye and left without giving me anything other than a ring.’ (A.-R. Lesage, 1732, Histoire de Gil Blas de Sant) Despite the change of position, modification by autre remains direct at this time, i.e. with no linking de. In its shape then, this modification remains seemingly close to regular post-nominal adjectival structures. However, recall that as seen above, modification by autre of regular nouns is strictly pre-nominal up to contemporary French. Thus, in e¤ect, post-nominal direct modification by autre can be taken to indicate that rien no longer behaves like a regular noun in the language. Taking the position of autre to mark the upper limit of NP internal modification, as is suggested by its high contemporary position, examples like (56) imply that by the seventeenth century, rien had definitely left the nominal domain for the NumP domain. That is, after a period of transition where autre could occur alternatively as pre or post-nominal modifier, i.e. where on our view, movement of rien from the nominal domain to the NumP domain remained optional, rien came to be directly merged in NumP, accounting for obligatory post-nominal autre. This form of direct post-nominal modification in turn is maintained through the classical period and is found rather late, i.e. up to 1925, in the writings of Claudel. Of interest is the fact that in Claudel’s writings, as well as throughout the classical period, characteristic NPI uses of n-words in downward entailing contexts are commonly found. As shown by the following table reproduced from De´prez &

An outside-in micro-parametric approach to negative concord

257

Martineau (2004), positive uses and polarity context uses (excluding negation) of n-words amounted to about 30% during the seventeenth and 20% during the eighteenth century. Table 1. Frequency of aucun in positive, polarity and negative contexts periods

positive contexts

polarity contexts

negative contexts

17th century

3.5% (7/200)

27.5%

69%

19th century

0%

20.5%

79%

Against the background of this diachronic study, we can see that direct post-nominal modification with autre appears to be a characteristic feature of the classical period (seventeenth–eighteenth century), which coincides with the increasing use of n-words as NPIs (De´prez & Martineau). As evidenced in Martineau & De´prez (2004), Quebec French also contrasts with contemporary continental French in commonly allowing the co-occurrence of the negation pas with n-words as well as common NPI uses of these expressions. In short, there are both diachronic and comparative evidence of the correlated occurrence of direct post-nominal modification by autre and of the use of n-words as polarity items.26 The next pattern of autre modification observed in our diachronic corpus is that of indirect modification with de. Notably, this indirect modification appears to have a peculiar historical evolution. In Frantext, it is first found in the twelfth century in the works of Chretien de Troyes, but it occurs only sporadically. After this, rather curiously and quite interestingly, this pattern disappears entirely from the corpus until 1843, where we see it re-surface, after a seven century gap, in the writings of Honore´ de Balzac. (57) [. . .] la`, madame; ne vous faut-il rien d’autre? ‘There madam; do you need nothing else?’ (H. de Balzac, 1843, La Rabouilleuse) The use of this indirect pattern with autre remains fairly infrequent until about 1890, where it finally reaches about equal status with the previously discussed post-nominal direct modification rien autre and rien autre chose. Yet it is not until 1909, in the writings of Andre´ Gide, that the indirect de-modification pattern firmly dominates and definitely wins over 26. For a more detailed discussion of the NPI uses of n-words both in classical French and in Quebec French, see De´prez & Martineau (2004).

258

Viviane De´prez

the direct modification pattern. Worth stressing here is the fact that this is an interesting discovery, as our research shows, apparently for the first time, that the use of the modern indirect modification pattern with de as in rien d’autre was not firmly established until essentially the beginning of the twentieth century. Referring back to the findings of De´prez & Martineau (2004) again, this appears to coincide with the rise of the strongly negative interpretation of n-words. For further empirical evidence, let us briefly consider the modification possibilities with bare personne, for the sake of brevity concentrating on modification by autre. As in the case of rien, we observe that the first occurrences of the modifier autre with a bare personne are pre-nominal, as in (58). In Frantext, the last example of bare personne modified by prenominal autre is found in the seventeenth century. (58) [. . .] on n’entend autre personne qu’un sonneur de cloches. ‘One cannot hear anyone other than a bell ringer.’ (P. de Deimier, 1610, L’Academie de L’Art Poetique) After this date, the prenominal modifier autre is only found in the Det þ personne combination, which, as discussed above, is the nominal construction still currently in use. Thus, the modifier autre occurs strictly in postnominal position with a bare personne, as in the following characteristic example: (59) [. . .] pour le sens commun de croire qu’elle n’est apperc¸ue par personne autre. ‘[. . .] for the common sense to believe that she was not seen by anyone else’ (J.-J. Rousseau, 1776, Rousseau Juge de Jean-Jacques) With bare personne, this type of modification is found up to the early twentieth century. Again it is quite striking to discover that the current form of indirect modification with de, i.e. personne d’autre, arose only quite recently. Searching Frantext for indirect modification with d’autre, we see that it first occurred in Chretien de Troye in the twelfth century, but, as in the case of rien, the pattern completely disappears to finally resurface only very recently, in 1902 in the writings of Paul Adam. (60) [. . .] quand elle rentrait, personne d’autre ne se de´rangeant pour elle? ‘When she would come in, none else would trouble themselves for her?’ (P. Adam, 1902, L’enfant d’Austerlitz) Thus we see here again that indirect modification with d’autre is not dominant with bare personne until the early twentieth century.

An outside-in micro-parametric approach to negative concord

259

3.3. Theoretical implications and conclusions The previous section has provided detailed evidence for the evolution of the n-expressions rien and personne, attempting to establish, first, the time course of their feature loss, and second, correlated structural changes. It is now time to take stock of these observations and to consider their theoretical consequences. Summarizing and highlighting the central facts, we have observed three stages in the syntactic evolution of n-expressions. In the first stage, n-words are essentially nominal in nature, as evidenced by the use of determiners, their gender and number features as well as by the pre-nominal position of the modifier autre. By and large, we have dated the loss of features to the middle of the sixteenth century and the disappearance of the pre-nominal modification pattern by the beginning of the seventeenth century. We concluded that, structurally, during this period the n-expressions occurred within the NP layer of their nominal structure. In the second stage, the n-words lost their nominal features and became invariable with respect to both number and gender. Modification by autre is now postnominal but it remains direct, without de. We have suggested that at this stage, which essentially spans the classical period, these n-words climbed a first step within the nominal structure, reaching NumP so that they are now directly merged in this position, accounting for the strictly postnominal position of autre. The third stage corresponds in our data with the appearance of indirect de modification. As discussed in Section 2.3., this type of modification with d’autre corresponds to quantity or degree modification, like modification with de plus, or de moins, and indicates that n-words have now taken on a quantificational nature. Following Zamparelli (1995), we have assumed that this could correspond syntactically to yet another step up the structural ladder of the DP structure, reaching in Zamparelli’s terms, the level of strong quantification or SDP. Putting all these assumptions together, the following structure sums up the evolution of the n-word that our evidence has unearthed. (61)

On this view, the evolution of n-words corresponds to a gradual three-step movement up the DP-structure. Of particular interest are the parallels that

260

Viviane De´prez

can be observed between the historical course of this evolution and the findings that De´prez & Martineau (2004) and Martineau & De´prez (2004) report on the historical evolution of the meaning of n-words. Focusing on the interpretation of aucun, they have distinguished three periods that largely correspond to the ones we have distinguished here. In the first period, corresponding to Old and Middle French, the interpretation of n-words seems largely context independent and positive. Beginning with the sixteenth century and into the classical period, however, the interpretation of n-words becomes context dependent. As Martineau & De´prez (2004) argue, n-words at this period manifest characteristic features of NPI: They are licensed only in restricted contexts, i.e. the so-called ‘a¤ective’ contexts, which in addition to downward entailing contexts include interrogatives and comparatives; they can be licensed at long distance; they are not licensed in subject position and are compatible with negation, including the negation pas, which is dominant at the time at the sole exponent of sentential negative meaning. Finally, in the last period, which corresponds to the birth of Modern French at the beginning of the twentieth century, n-words come to take on a negative meaning of their own, now occurring alone, i.e. without the co-presence of the sentential negative marker pas. This evolution is well known. What had not been put together up to now is the parallelism that the meaning evolution of n-words shows with the changes that a¤ect their internal structure. This is what this section hopes to have accomplished by coupling the historical evolution of the meaning contribution of n-words at the sentential level, with a careful study of their internal change. To finalize the correlation of the two, we propose that n-words undergo the following evolution. In the first period of their evolution, their structure contains a null determiner that is essentially equivalent to the one posited for existentially or generically interpreted bare noun phrases. That is, for all intent and purposes, rien and personne function like bare indefinite noun phrases. At the beginning of this period, personne and rien are still nominal and count nouns, and their null determiner is essentially the same as that of other indefinite noun phrases in the language. Following Eckardt’s (2006) very interesting analysis of the semantic evolution of French negative elements within a focus alternative semantics model a` la Rooth (1985), we take this to be a period in which, when used in negative contexts, these expressions came to be used in emphatic denial first, simply under contingent circumstances, that is, only when their contexts license plausible alternatives to their very general predicative meaning of ‘person’ or ‘thing’. Such contexts were essentially comparable to

An outside-in micro-parametric approach to negative concord

261

what is needed in current modern French to license the denial of a regular indefinite noun phrase like une personne. That is, they are for instance comparable to the alien context we set up in (11) above to make sense of the denial of the determined expression de personne. This context is set up to contingently oppose personne to ‘aliens’, thus introducing a plausible alternative that is disjoint from the meaning of ‘person’, as required under Rooth alternative-focus semantics. Other relevant plausible contexts could set up relevant alternatives by opposing ‘persons’ to objects, plants, animals or more generally, to ‘non-person’ entities, following Eckardt (2006). Gradually, however, as the n-words lose their ability to be marked for gender and number, their nominal quality is altered. At the same time, regular indefinite null determiners start to become increasingly rare in assertive contexts, due to the general decline (phonetic erosion) of plural marking on nouns, and the consequent growing reliance on overt determiners marking for number distinctions. We would like to propose that the conjunction of these syntactic changes is what sets up the stage to alter the structure of n-words and the type of alternative they evoke for a noncontingent use of personne in emphatic negative constructions. In other words, it is at this time, following Eckardt (2006), that n-words become focused existentially quantified predicates which call for other predicates as their alternatives. Eckardt (2006: 256) formalizes this change as follows: (62) [[ personne ]] ¼ lQpx( PERSON(x) b Q(x) ) ‘some person’ Alt( personne) ¼ {lQpx(BAKER(x) b Q(x)), lQpx(PRIEST(x) b Q(x)), lQpx(BUTCHER(x) b Q(x)), . . . } Pragmatic function: Must be used in emphatic focus. In (62), the alternatives to personne are no longer non-persons, but rather subtypes of persons, such as bakers, priests, scholar, students, i.e. alternatives to predicates describing human types. A consequence of this change is that alternative propositions constituting the ‘focus semantic value’ (Rooth 1985) of these expressions are now in a logical entailment relation with the stated proposition that contains them. (Being a baker, butcher, or scholar entails being a person.) As a result, the stated proposition can meet the pragmatic requirement of being the most striking one among the space of alternatives, only under special conditions, i.e. only under negation or when occurring in downward entailing contexts (see Eckardt 2006 for precisions). During this period of gradual change, culminating in the second period of their evolution, the one starting with the end of the sixteenth century and spanning the classical period, we propose that the n-word gradually moved up their DP, to Zamparelli’s PDP part of

262

Viviane De´prez

the structure. In this step in their evolution, n-words are semantically NPIs and we have seen above that modification with autre then regularly occurs in a post-position. Now according to Eckardt (2006) again, NPIs are characterized by a lexical requirement for focus, which enforces their use in restricted environments. Eckardt, however, does not specify how this lexical requirement comes about. Given the evidence we provide of internal change at this level, we would like to suggest that the requirement for focus on rien and personne is not really lexical, as Eckardt supposes, but rather syntactic. More specifically, the proposal is that the focus requirement corresponds to a distinctive DP internal syntactic structure that encodes DP internal focus-movement. In short, we propose that the internal move of the n-word to the PDP zone both encodes and satisfies the requirement for the focus marking characteristic of NPI, and at the same time provides a way to license a null D in NumP, which now fails to be licensed under number agreement. For a brief clarification and justification of this view, let me quickly return to Kayne’s (1994) discussion of the internal structure of possessives and other [de NP] structures in French. For constructions such as (63a), Kayne proposes the complex structure in (63b), which essentially exploits and expands the strong parallelism assumed to hold in generative perspectives between the internal structure of DP and the structure of sentences. Thus DP, on Kayne’s view, contains a clause-like structure with an IP that is the complement of de, here conceived as a complementizer equivalent in the nominal domain. (63) a. Le crayon de rouge b. [D [ DP/PP [NP crayon] j de [IP [AP rouge]k [I 0 [FP [NPe] j [F 0 [e]k ]]]]]] The construction in (63a) is derived as follows. The lowest part of the structure FP in (63b) is a kind of small clause, headed by F 0, in which the adjective rouge is predicated of the NP crayon (in Spec FP at the start of the derivation). That is, FP is some kind of predicative structure akin to Le crayon est rouge ‘The pen is red’, but with a null copula. From this structure, there is first movement of the adjective rouge to Spec IP. This adjective movement is taken to be triggered by a focalization somewhat similar to what motivates predicate inversion in so-called ‘inverse copular constructions’ like ‘the cause of the riot was a picture on the wall’ (see Moro 1991). That is, the adjectival predicate rouge is ‘focus-fronted’, deriving a kind of inverse copular sentence, abstractly similar to ‘tasty, this co¤ee’, but here DP internal. Finally, the NP crayon also undergoes

An outside-in micro-parametric approach to negative concord

263

focus movement to the top of the structure, i.e, the specifier of de, the functional equivalent of Co in nominal structures.27 With this structure in hand, we can now return to our proposal concerning the internal structure of n-words and their syntactic and semantic evolution. In short, what we are proposing here is that, in the course of their evolution, the French n-words have undergone successive movements inside the structure of their nominal constituent that parallel the movements proposed by Kayne for the derivation of nominal structure like those in (63) above. Below is a sketch of the detail of this syntactic progression, as we see it: (64) [D [ DP/PP [ personne3] j de [IP [AP [ personne2] [I 0 [FP autre [NP personne1] SDP

PDP

NP

For ease of reference in the following discussion, the respective positions of personne in the structure (64) have been labelled personne1, personne2, personne3 in (64) and the correspondence with Zamparelli’s proposed quantificational structure is given directly under it. At the bottom of the structure, ‘personne1’ is just a noun. Contingently (and thus optionally) at first, personne can undergo internal focus movement to Spec IP/PDP, the position of personne2, for emphasis. We propose to view the head I 0/ PDP as some kind of null nominal functional projection, or null D, licensed under Spec-head gender/number agreement with the moved nominal personne. As a result of a gradual loss of gender and number features, agreement with this D is lost, so that movement to Spec IP/ PDP, i.e. personne2, comes to be licensed only under focus. This focus movement serves two purposes. First it structurally encodes the supposed lexical requirement for focus that Eckardt (2006) deems to be the hallmark of NPIs. Second, it licenses the null head of I0, now no longer licensed under agreement. It also has further interesting consequences. It syntactically crystallizes the change in the nature of the alternatives to which personne is now compared under emphatic negation. Following Eckardt, the semantic change is as in (62) above. Note that on this view, personne is now a predicate nominal akin to so called ‘professional bare nouns’ like priest. As a predicate nominal, it has become ‘more adjectival’ as it were, which could explain why it can undergo a focus movement that parallels that of adjectives such as rouge in Kayne’s structure (63). We can speculate

27. For a discussion of these structures, see Kayne (2004: Chapter 8).

264

Viviane De´prez

that this movement to the higher position in the structure was optional at first, becoming gradually obligatory later, to eventually lead to structural reanalysis and direct merge/insertion of personne into this PDP specifier position ( personne2). This restructuring, in e¤ect, grammaticalizes the focus that Eckardt assumed to be a lexical requirement for NPIs.28 So in the spirit of Eckardt (2006), but now as an internal structural construct, we propose that it is the reanalysis of this focus movement in tune with the pragmatic change in the focus-alternatives that eventually turns the former noun personne into a dependent NPI, i.e. an intrinsically and internally focused structure. Note of course, and this is crucial as it provides support for this view, that the post-nominal position of autre at this stage of the historical evolution now straightforwardly derives from the proposed structural change. It would not, in contrast, follow from any putative lexical requirement for focus. It is thus no accident, under our proposal, that the post-nominal position of autre coincides with an NPI interpretation of personne. In turn, we can take the post-nominal position of autre to signal the focalisation of personne that for Eckardt (2006) centrally defines what an NPI is. It is time now to turn to the consideration of the final steps of the evolution of personne. In the final steps, personne reaches the highest position in the complex structure in (64), namely the position of personne3. This again, following Kayne (1994), involves a type of focus-movement. Here as well interesting consequences follow from this structural change. First, personne is predicted to precede the C/P head de, which accounts for its obligatory indirect modification by d’autre at this stage of its evolution. Furthermore as noted by Kayne, only elements that involve quantifica28. The analysis developed here suggests an exiting syntactic generalization for NPIs: only nominal structures that encode focus movement internal to DP of the sort discussed here can be NPIs. Note that this view provides a very interesting way to explain the semantic nature of phrasal NPIs like: qui que ce soit, literally ‘who that this be’. Strictly speaking, these are grammaticalized (e.g. nominalized) instances of wh-movement to a focus position inside a structure that behaves like an NPI nominal. In contrast with characteristic lexical minimizers or generalizers that have been assumed to crosslinguistically provide good semantic source for NPI, these have none of the favoring semantics. Why they should provide such common source of NPI thus remained so far a bit of a mystery. The analysis provided here, resolves this mystery. These are such good NPI because they encode lexicalized focus movement in their internal structure.

An outside-in micro-parametric approach to negative concord

265

tion, in his terms an operator variable relation, can occupy this CP-like position. This suggests that it is only when moved to this highest position that personne acquires the full quantificational nature that is now associated with its negative meaning. Again, we can think of this movement as optional at first, explaining why indirect de modification of n-words for quite some time alternate with direct post-position modification. It then becomes – gradually – more obligatory, before eventually leading to structural reanalysis. Now, although our proposal provides quite a detailed account of how n-words have structurally and semantically acquired an NPI character in the course of their evolution, it still falls short of accounting for how they acquired their final intrinsic negative meaning. Somewhat disappointingly, Eckardt (2006), who provides such a deep and fascinating explanation of the semantic evolution of reinforcers into NPIs, does not in fact provide an interesting explanation for the acquisition of negative meaning. In her view, this comes about simply as a stylistic reanalysis. But what a stylistic reanalysis formally is, and what the exact compositional semantics of this final step entails, remains, unfortunately, far from clear. The structure we have proposed above, provides, perhaps, the beginning of an understanding of the mechanics of this final steps, but again, no precise semantic account will be o¤ered here either. Note that in (63), personne3 is still dominated by a null determiner, which on Kayne’s proposal is the Dhead that closes o¤ the entire nominal relative clause like structure. Yet as amply discussed above, n-words, in contrast to NPIs, present no distributional limitation of occurrence. They appear free of any c-commanding operator and can comfortably occur even in a subject position. As was argued above, this shows that personne must eventually come to occupy the position of this higher D, so that in e¤ect, it comes to be incorporated into it, rather than being dominated by it. This incorporation in turn suggests a plausible pathway for understanding the negative meaning of personne. Conceivably, given contextual requirements on NPIs, this highest null D in the structure came to acquire a negative meaning. Perhaps as a precursory mid-step to this final change, personne and rien first became strong NPIs, licensed only in anti-veridical, i.e. strongly negative contexts. We can speculate that it would be on this step that the null determiner acquired a negative feature, as a sort of semantic encoding of this strong dependency on negation, i.e. a feature-sharing process encoding semantic rather than syntactic agreement. If so, personne would have come to inherit and intrinsically integrate this negative feature through its

266

Viviane De´prez

final incorporation into the highest null D in the structure (64). If correct, note that this approach makes the interesting predictions that it is only when nominal n-words occupy the highest position in their DP that they could come to incorporate an essentially sentential negative feature. This is interesting because, in the current syntactic framework of Minimalism, the highest positions in constituents take on a particular status. They constitute the edge of phases, i.e. the only positions from which access to a higher domain of computation becomes possible. Within the Minimalist framework, the notion of phase conceptually recalls that of encapsulated domains. So the idea here is that to be visible from outside, the domain of a particular phase, i.e. to be visible to a higher structural domain, a feature must find itself at the edge of a phase domain. Specifically, it must not be properly contained within the boundaries of an encapsulated domain. This is in essence the core idea of Chomsky’s PIC. Elements at the edge of a phase are visible to the next domain of computation. Elements contained within a phase are not. Within this framework, it becomes quite immediately obvious that structural change within a given domain can bring about visibility or invisibility to a higher domain of computation. It is this core idea that I wish to speculate could be at the basis of the correlated structural and semantic changes we have observed in the French n-words. Applying this core idea to the case at hand, it would follow that the visibility of a negative feature, and hence its interpretability in a higher computational domain, here the sentence, should go hand in hand with its positioning at the edge of a domain. Suppose now that n-words were endowed with a negative feature from the onset of their development, perhaps for pragmatic reasons. If DP is a phase, an assumption that goes along with its similarity to CP, then as long as this negative feature occurs within the DP domain, it will essentially remain un-interpretable from outside, i.e. here specifically, at the level of the sentential domain. So a sensible interpretation of this situation is that such a buried negative feature could not have sentential scope. In the gradual change of their internal structure, n-words have, through a series of internal movements, been brought closer to the edge of DP, and in e¤ect, so has their negative feature. Once the DP edge was reached, it directly follows from the PIC that their negative feature became visible outside their nominal domain or in other words, visible to the sentential domain. In this new situation, their negative feature can now take on sentential scope, i.e. be interpretable at the sentential level. Schematically, the situation can be described as follows:

An outside-in micro-parametric approach to negative concord

(65) a. b.

267

[ TP f. . . . . [DP . . . . . [ F ] ] F invisible to higher domain ¼ uninterpretable. [TP f. . . . . [DP F [. . . . . ] ] edge F is visible to higher domain ¼ interpretable29

In (65a), F is invisible for sentence interpretation because it is inside the domain of D. From the perspective of a higher domain containing the DP, the feature F is thus uninterpretable. In (64b) in contrast, because F is at the edge of the DP, in D head or D specifier, it becomes visible to the sentential domain, because it is part of this higher computation. Edges are always computed as part of their containing domain. This leads us to the following hypothesis: (66) Feature Visibility principle: The visibility/interpretability of a feature F relates to its position inside/at the edge of a domain of Computation Or more technically: (66 0 ) Feature interpretability (uF/iF) is governed by the PIC. Within this theoretical frame, an interesting interpretation emerges for the structural changes we have observed. The hand in hand processes of structural change and semantic change we have brought evidence for here and before (De´prez & Martineau 2004) can be seen as part of a more general process of meaning change that is governed by very general principles of computation.30

4. Conclusions To conclude, Jespersonian-inspired models such as Zeijlstra (2008) propose a macro-parametric approach to negative concord that locates the central factor dividing negative concord languages from non-negative 29. The question also arise within traditional accounts of grammaticalization since it is quite unclear in such a tradition how a semantic change from positive to negative could be considered as a form of semantic bleaching that is assumed to generally accompany grammaticalization. 30. For an application of the same principle, I call Visibility at the edge, see De´prez (2007, 2008) studies of the grammaticalization of number and specificity/ definiteness features in French Based Creole noun phrases.

268

Viviane De´prez

concord languages in the syntactic properties of the sentence negation operator. The micro-parametric approach proposed in this chapter, by contrast, takes negative concord to arise from a convergence of factors that depend largely on the structural and semantic make up of the negative dependent terms, the n-words. Such a convergence of factors may distinguish only a subset of expressions in a given language, permitting as a result inner language variations and historical changes that do not target all expressions in the same way. We have provided empirical arguments in favour of this approach showing that inner language diversity is found even in a seemingly uniform language such as Haitian Creole. Moreover, we have provided detailed empirical evidence of the complex structure of contemporary French n-words and in particular of the fact that they occupy a position at the highest level of their DP structure. We have furthermore demonstrated that the historical evolution of these n-words involved a change in their internal structure that goes hand in hand with correlated changes in the types of negative relation they are involved in, and we have charted the time course of these correlated changes. We have argued that these internal changes provide evidence for a stepwise rise of n-words inside the DP domain and have suggested that a negative feature becomes interpretable at the higher sentential domain only when it has reached the edge of the nominal structure. If correct, this proposal predicts that only n-words that occupy the highest structural level in DP could behave like intrinsically negative quantifiers, i.e. like quantified expressions whose negative interpretation has scope in the sentential domain. Finally, we have sketched a model of feature interpretability that captures these correlated changes and accounts for their e¤ects through very general principles of computation. References Sources Elicop http://bach.arts.kuleuven.be/elicop/ Frantext http://atilf.atilf.fr/frantext.htm Textes de francais ancien http://www.lib.uchicago.edu/efts/ARTFL/projects/TLA/ Studies Baker, Mark 2008

The macroparameter in a microparametric world. In: Theresa Biberauer (ed.), The limits of syntactic variation, 351–374. Amsterdam: Benjamins.

An outside-in micro-parametric approach to negative concord

269

Chomsky, Noam 2000 Minimalist inquiries: the framework. In: Roger Martin, David Michaels & Juan Uriagereka (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik, 89–156. Cambridge: MIT Press. Chomsky, Noam 2001 Derivation by phase. In: Michael Kenstowicz (ed.), Ken Hale: A life in language, 1–53. Cambridge: MIT Press. Cinque, Guglielmo 1994 On the evidence for partial N-movement in the Romance DP. In: Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi and Ra¤aella Zanuttini (eds.), Paths towards Universal Grammar. Studies in honor of Richard S. Kayne, 85–110. Washington, 1994, Georgetown University Press. Clas, Andre´ 1976 Mate´riaux pour l’e´tude du francais du Canada, Ne´ologisme et canadianisme, vol 1. Montreal: De´partement de Linguistique, Universite´ de Montre´al. Corblin, Francis, Viviane De´prez, Henriette de Swart & Lucia Tovena 2004 Negative concord. In: Francis Corblin & Henriette de Swart (eds.), Handbook of French semantics, 417–452. Stanford: CSLI. Den Dikken, Marcel 1998 Predicate inversion in DP. In: Artemis Alexiadou & Chris Wilder (eds.), Possessors, Predicates and Movement in the Determiner Phrase, 177–214. Amsterdam/Philadelphia: Benjamins. De´prez, Viviane 1997 Two types of negative concord. Probus 9 (2): 103–142. De´prez, Viviane 1999 The roots of negative concord in French and French-lexicon Creoles. In: Michel DeGra¤ (ed.), Language creation and language change: Creolization, diachrony, and development, 375– 428. Cambridge: MIT Press. De´prez, Viviane 2000 Parallel (a)symmetries and the internal structure of negative expressions. Natural Language and Linguistic Theory 18: 253–342. De´prez, Viviane 2003 Concordance ne´gative, syntaxe des mots-N et variation dialectale. Cahier de linguistique franc¸aise 25: 97–118. De´prez, Viviane 2005 Morphological number, semantic number and bare nouns. Lingua 115 (6): 857–883. De´prez, Viviane 2007 Probing the structuring role of grammaticalization: Nominal constituents in French lexifier Creoles. Journal of Pidgin and Creole Languages 22 (2): 263–307.

270

Viviane De´prez

De´prez, Viviane 2008 On the structuring role of grammaticalized morpho-syntactic features. In: Emily Elfner & Martin Walkow (eds.), Proceeding of NELS 37, volume 1, 31–44. Booksurge LIc. De´prez, Viviane & France Martineau 2004 Micro-Parametric Variation and Negative Concord. In: Julie Auger, J. Clancy Clements & Barbara Vance (eds.), Contemporary approaches to Romance linguistics, 139–158. Amsterdam: Benjamins. Eckardt, Regine 2006 Meaning change in grammaticalization. An enquiry into semantic reanalysis. Oxford: Oxford University Press. Giannakidou, Anastasia 1997 The landscape of polarity items. Ph.D. dissertation, University of Groningen. Gougenheim, Georges 1951 Grammaire de la langue franc¸aise du seizie`me sie`cle. Lyon: IAC. Grevisse, Maurice 1980 Le bon usage. Paris: Duculot. Huot, He´le`ne 1981 Construction infinitives du franc¸ais. Le subordonnant de. Geneva: Droz. Haegeman, Liliane & Rafaella Zanuttini 1991 Negative heads and the Neg criterion. The Linguistic Review 8: 233–151. Hazael-Massieux, Marie-Christine 2008 Textes anciens en cre´ole franc¸ais de la Caraı¨be: Histoire et analyse. Paris: Publibook. Herburger, Elena 2001 The negative concord puzzle revisited. Natural Language Semantics 9 (3): 289–333. Kadmon, Nirit & P. Landman 1993 Any. Linguistics and Philosophy 16 (4): 353–422. Kayne, Richard 1977 Syntaxe du francais, Paris: Seuil. Kayne, Richard 1994 The antisymetry of syntax. Cambridge: MIT Press. Kiparsky, Paul & Cleo Condoravdi 2006 Tracking Jespersen’s cycle. In: Mark Janse, Brian D. Joseph & Angela Ralli (eds.), Proceedings of the Second International Conference on Modern Greek Dialects and Linguistic Theory, 172–197. Mytilene: Doukas. Larson, Richard 1985 Bare-NP adverbs. Linguistic Inquiry 14: 595–621.

An outside-in micro-parametric approach to negative concord

271

Larrive´e, Pierre 2004 L’association ne´gative: depuis la syntaxe jusqu’a` l’interpre´tation. Geneva: Droz. Martineau, France & Viviane De´prez 2004 Pas rien/Pas aucun en franc¸ais classique. Variation dialectale et historique. Langue francaise 143: 33–47. Martineau, France & Raymond Mougeon 2003 Sociolinguistic research on the origins of ne deletion in European and Quebec French. Language 79 (1): 118–152. May, Robert 1989 Interpreting Logical Form. Linguistics and Philosophy 12: 387– 435. Pereltsvaig, Asya 2006 Small nominals. Natural Language & Linguistic Theory 24: 433– 500. Penka, Doris 2006 A crosslinguistic perspective on n-words. Proceedings of BIDE05, International Journal of Basque Linguistics and Philology (Forthcoming). Postal, Paul M. 1966 On so-called pronouns in English. In: David Reibel & Sanford Schane (eds.), Modern Studies in English, 201–223. Englewood Cli¤s: Prentice Hall. Postal, Paul M. 2004 Skeptical linguistic essays. Oxford: Oxford University Press. Progovac, Ljiliana 1994 Negative and positive polarity: A binding approach. Cambridge: Cambridge University Press. Moro, Andrea 1991 The raising of predicates: Copula, expletives and existence. In: Lisa L. Cheng & Hamida Demirdache (eds.), MIT working papers in Linguistics 15. More papers on wh-movement, 183–218. Cambridge: Department of Linguistics and Philosophy, MIT. Rooth, Mats 1985 Association with Focus. Ph.D. dissertation, University of Massachusetts Amherst. Sarre, Nguissaly 2003 Diachronie des pronoms inde´finis a` base nominale du moyen franc¸ais au franc¸ais classique: Les re´manences d’un emploi nominal a` travers un emploi pronominal. Revue de linguistique romane 265–266: 117–136. Sleeman, Petra 1996 Licensing empty nouns in French. Ph.D. dissertation, University of Amsterdam.

272

Viviane De´prez

Roberts, Ian & Anna Roussou 2003 Syntactic change. A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. de Swart, Henrie¨tte & Ivan A. Sag 2002 Negation and negative concord in Romance. Linguistics and Philosophy 25 (4): 373–417. Watanabe, Akira 2004 The genesis of negative concord: Syntax and morphology of negative doubling. Linguistic Inquiry 35 (4): 559–612. Tovena, Lucia, Viviane De´prez & Jacques Jayez 2004 Polarity sensitive items. In: Francis Corblin & Henriette de Swart (eds.), Handbook of French semantics, 391–417. Stanford: CSLI. Zamparelli, Roberto 1995 Layers in the Determiner Phrase. Ph.D. dissertation, University of Rochester. Zanuttini, Rafaella 1991 Syntactic properties of sentential negation: A comparative study of Romance languages. PhD dissertation, University of Pennsylvania. Zeijlstra, Hedde 2004 Sentential negation and negative concord. Ph.D. Thesis, University of Amsterdam. Zeijlstra, Hedde 2008 Negative concord is syntactic agreement. MS, University of Amsterdam.

Viviane De´prez: ‘‘Atoms of negation. An outside-in micro-parametric approach to negative concord.’’ Discussion Maj-Britt Mosegaard Hansen The micro-parametric model of the evolution of the French negative indefinites personne, rien, and aucun that is presented in Viviane De´prez’s contribution to the present volume is an elegant, and to a large extent convincing, one. One of the particular attractions of the model (at least to my mind) lies in the fact that it explicitly allows for, indeed seems to predict, a certain amount of variability among individual n-words, as a result of di¤erent diachronic sources and trajectories, as well as di¤erences in the pace of change. Nevertheless, the proposed model also raises a number of questions that will be taken up below. As it is not my intention in this contribution (nor my brief, as I understand it) to o¤er a competing account, the questions raised should be seen merely as aspects that I believe merit further reflection. The first problem, acknowledged by De´prez herself, is that the model cannot provide a uniform explanation of the evolution of all French nwords, as the notion of movement within the DP is not immediately applicable to the adverbial n-words jamais, plus, and gue`re. The suggestion that the development of the latter may have been triggered indirectly by some similarity with the nominal n-words unfortunately does not take us very far in the absence of an account of what those similarities might be. It seems safe to say that, in terms of their potential for negative concord readings versus double negation readings, as well as the extent to which they can appear with positive meaning in non-a‰rmative polarity contexts, French n-words behave in largely similar (even if not fully identical) ways independently of their nominal vs adverbial origins (see further below). For that reason alone, one would prefer a model capable of accounting for their development by appealing to commonalities across specific items. Even within the group of nominal n-words, it must be noted that the evolution of nul is currently unaccounted for within the proposed model. Although nul may conceivably have evolved in a way that resembles the evolution of aucun, whether and to what extent that is so remains to be shown. Indeed, there does intuitively appear to be salient di¤erences

274

Maj-Britt Mosegaard Hansen

between the two items. While etymologically negative, Old French nul did not have negative meaning except in the presence of preverbal ne; unlike the other nominal n-words, it does not, however, seem to have had genuinely positive uses at any stage of French, but is found with a positive interpretation only in negative polarity contexts (Ingham 2008), suggesting that it must have been an NPI from the start, a status which – on De´prez’s analysis – aucun only acquired in the Renaissance. It is doubtful whether the NPI nul could have had the adjectival status that De´prez argues for in the case of aucun, given that a quick search of the Frantext and Textes de franc¸ais ancien data bases yielded only a single – marginally possible – example of nul preceded by a determiner (where the verb pre´sumer, which has two di¤erent valency patterns, is in fact far more plausibly interpreted as indirectly transitive, making de a preposition rather than a determiner): (1) [. . .] comme c’est asse`s pour me´rite de ne pre´sumer de nuls me´rites, [. . .]. ‘as it is su‰cient for merit to not presume any merits/to not expect too much of any merits’ (Calvin 1560) The further fact that Old French nul is, in a number of instances in the data bases, followed by the indefinite pronoun autre and/or by adjectives, suggests that it had the status of a determiner (when not used as a pronoun). If I understand De´prez’s account of aucun correctly, one would therefore expect that nul should have been capable of inherently negative uses long before we reach the stage of Modern/Contemporary French. The issue of di¤erent (types of ) n-words leads directly on to a di¤erent problem with the model, namely the assumption that negative concord in French is the result of a resumptive interpretation of a series of inherently negative n-words, such that they form a single quantifier binding several variables at the same time. This is in opposition to a sequential interpretation, whereby quantifiers have scope over one another, resulting in a double (or multiple) negation reading. In this, De´prez adopts the approach of May (1989). A resumptive interpretation of quantifiers requires that the items involved exhibit syntactic and semantic parallelisms, which is assumed to explain why combinations of n-words with sentential negation in ‘‘standard’’ French do not allow for negative-concord readings, but must be interpreted as double negation. Now, while one can readily admit that nominal n-words like rien and personne, which typically fulfil argument functions in their host clauses, are intuitively dissimilar syntactically and semantically from the sentential

Discussion of De´prez

275

negator pas, whose role is that of an operator, it is less immediately obvious that the adverbial n-words plus and gue`re, whose combination with other n-words invariably seems to give rise to concord readings, would be syntactically and semantically more similar to rien and personne than to pas: (2) Personne n’a pas ouvert la bouche. ‘No one did not open their mouth’, ‘Everyone opened their mouth’ ¼ DN (3) Personne n’a plus ouvert la bouche. ‘No one opened their mouth any more.’ ¼ NC (4) Personne n’a gue`re ouvert la bouche. ‘No one hardly opened their mouth.’ ¼ NC In other words, for the notion of (dis)similarity among items to be operational in determining the choice between resumptive and sequential quantification, we need a more precise account of what it means for negative items to be syntactically and semantically (dis)similar in the first place, and of how degrees of (dis)similarity can be measured. As noted above, the usefulness of the notions of resumptive vs sequential quantification is predicated on the assumption that French n-words are inherently negative elements, rather than (a special type of ) NPIs bound by the preverbal ne. In the case of sentential negation, few people would probably dispute that in contemporary (spoken) French, postverbal pas is not only inherently negative, but also the ‘‘real’’ negator, whereas preverbal ne, when it does occur, appears to be merely a kind of agreement marker, possibly used principally as a index of formality of register and/or conversational topic (Sanko¤ & Vincent 1977; Ashby 1981), possibly as a marker of negative emphasis (Fonseca-Greber 2007). Statistics of the occurrence of pas with or without preverbal ne in unplanned interaction seem to amply bear this out: in both real and apparent time, ne is on the decline, and in the most informal forms of conversation, it is almost completely absent, particularly among younger speakers (Sanko¤ & Vincent 1977; Ashby 1981, 2001; Coveney 1996; Hansen & Malderez 2004). Interestingly, however, the very same empirical studies referenced above quite consistently show that ne is maintained with often significantly higher frequency in clauses containing n-words rather than pas. It is di‰cult to say how much importance should be attributed to these patterns, in as

276

Maj-Britt Mosegaard Hansen

much as ne-retention in most cases remains below 50% in conversation; nevertheless, we might venture that they provide some evidence that speakers may be treating n-words as less fully negative than pas in and of themselves, hence as calling for additional negative support, in the form of preverbal ne. Another set of observations regarding the distribution of n-words in contemporary European French serves to further call into question their status as inherently negative, namely the fact that a number of the French n-words are still found quite regularly with positive interpretations in various non-negative NPI-licensing contexts. Thus, a search of the Frantext data base comprising texts published between 2000–2007 yielded examples of the following environments: Conditionals (only jamais): (5) Si jamais il disait non? ‘If ever (¼ what if ) he were to say no?’ Rhetorical exclamations introduced by comme si (‘as if ’): (6) Comme si on allait manquer de rien la`-bas ! ‘As if we were going to lack for anything there!’ Direct and indirect interrogatives. These are frequently, but – as (8) shows – not invariably, rhetorical questions ( pace De´prez): (7) Est-il rien de moins marxiste en e¤et que la de´claration des typographes de Nantes? ‘Is there, indeed, anything less Marxist than the declaration of the typesetters from Nantes?’ (8) Demande-lui si ce train se remettra jamais en marche. ‘Ask him if this train will ever start moving again.’ Following a negated matrix clause: (9) Je ne crois pas qu’elle ait trompe´ personne. ‘I don’t think she cheated anyone.’ In infinitival clauses following a lexically negative matrix verb: (10) Il a refuse´ de voir personne. ‘He refused to see anybody.’

Discussion of De´prez

277

Following the pseudo-comparative expression [trop þ AdjP þ pour] (‘too þ AdjP þ to’): (11) Il est trop content de lui pour s’apercevoir de rien. ‘He’s too self-satisfied to notice anything.’ Following the temporal conjunction avant que (‘before’): (12) Il est parti avant que personne le voie. ‘He left before anyone saw him.’ Following a comparative: (13) Il lui arrivait plus souvent qu’a` aucun autre d’eˆtre surprise par la garde. ‘She was surprised by the guard more often than anyone else.’ Following a superlative (only jamais): (14) [. . .] une galette des rois chaude qui est la meilleure galette que j’aie jamais mange´e. ‘a hot Twelfth-Night cake which is the best cake I’ve ever eaten.’ Following the preposition sans or the subordinating conjunction sans que: (15) Il est parti sans avertir personne. ‘He left without telling anybody.’ (16) [. . .] sans que celui-ci se doute de rien. ‘. . . without the latter suspecting anything.’ It is not at all clear that (5)–(16) above are examples of a particularly formal (and by implication, archaic) register. While rhetorical questions with positively interpreted n-words do, as De´prez points out, feature subjectverb inversion, the word order is not principally a stylistic marker in this instance, but is due to the fact that French interrogatives with declarative SV word order simply do not admit of a rhetorical interpretation (Hansen 2001). As De´prez’s model predicts, individual n-words vary with respect to the specific types of context in which they can occur with positive interpretations in contemporary French, as shown in Table 1 below (adapted from Muller 1991: 265). It is noteworthy, however, that in terms of their evolution towards inherent negativity, rien, personne, and aucun are, in fact, among the most conservative n-words in ‘‘standard’’ French, while nul,

278

Maj-Britt Mosegaard Hansen

which, as argued above, appears structurally identical to aucun, has gone much further. The item nulle part, which, although used in adverbial functions, takes the form of what looks like a DP, occupies an intermediate position. On the surface, at least, these facts go against De´prez’s prediction that n-words with a common internal make-up should behave homogeneously both within and across languages. These observations also call into question the idea that standard French and Que´be´cois French instantiate two di¤erent language types: although Que´be´cois n-words admit of positive interpretations in more contexts that their standard French equivalents, the di¤erence seems to be largely a matter of degree. The two dialects appear to di¤er most obviously in their acceptance of combinations of n-words and the sentential negator pas, which – although possible (as examples (17)–(18), from the ELICOP corpus of spoken French demonstrate) – are comparatively rare, and in cases like (17), stigmatized, in Europe. It is, however, of relevance in this context that the postverbal marker gue`re (‘hardly’), which in terms of its semantics functions like a downtoned version of the sentential negator pas, and which is largely confined to formal registers, was fully acceptable with personne and rien in ‘‘standard’’ French until at least the beginning of the twentieth century, and is still regularly used in combination with plus (cf. (19)–(21)): (17) [. . .] je ne loupe pas aucun fait divers ‘I don’t miss any newsworthy event.’ (18) [. . .] il disait pas bonjour ni rien du tout hein ‘He didn’t say hello or anything at all eh.’ (19) [. . .] on ne voyait gue`re personne [. . .] (1913) ‘We hardly saw anybody.’ (20) [. . .] je n’ai gue`re rien fait qui vaille. (1939) ‘I’ve hardly done anything worthwhile.’ (21) Il est vrai qu’on n’est plus gue`re supplicie´ pour cette raison. (2006) ‘It is true that one is hardly tortured for that reason anymore.’ What should be noted is that the status and interpretation of French nwords seem to have been ambiguous at least since late Middle French: thus, Martin (1966) in an in-depth study of the n-word rien, adduces a number of constructions in which Medieval and Renaissance writers regularly used this item with negative meaning in the absence of a licensing ne.

Discussion of De´prez

279

Table 1. Negative polarity uses of French N-words N-word Negative polarity context

jamais

rien/aucun/ personne

nulle part

nul

plus

Sans

Z

Z

Z

Z

Z

Plus que

Z

Z

Z

Z

Y

Trop pour

Z

Z

Z

Y

(Z)

Neg V que

Z

Z

Z

Y

Y

Neg Vinf

Z

Z

Z

Y

Y

Negative item {Vinf/que}

Z

Z

Y

Y

(Z)

Avant {que/de}

Z

Z

Y

(Z)

Y

Peu

Z

Y

Y

Y

Y

Direct (rhetorical) question

Z

(Z)

Y

Y

Y

Conditional

Z

Y

Y

Y

Y

Indirect interrogative

Z

Y

Y

Y

Y

The final issue I wish to address in this chapter is that of grammaticalization. De´prez’s paper raises two questions. Firstly, she takes her point of departure in the Minimalist conception of grammaticalization as a process whereby interpretable features become uninterpretable. This prompts her to ask the question whether the evolution of the French n-words then instantiates a di¤erent type of grammaticalization, given that their negative feature originally required the presence of a licensing element in order to be interpretable, and thus appears to have been strengthened rather than bleached. While this is a very interesting observation, I tend to think its chief importance lies in flagging up a possible inadequacy in the Minimalist definition of grammaticalization. It seems to me that the evolution from NPI to inherently negative quantifier can very plausibly be argued to be a case of bleaching: Eckardt (2003: Chapter 4), in her analysis of French negation, observes that prototypical NPIs are specialized for emphatic environments in so far as their use evokes scales of alternative entities/ states-of-a¤airs, while simultaneously highlighting the asserted entity/stateof-a¤airs as logically and/or rhetorically stronger than any of the alternatives. Thus, by choosing to deny the most striking alternative, (22) prag-

280

Maj-Britt Mosegaard Hansen

matically evokes a scale of more or less contextually likely individuals that John might have talked to (e.g. a friend, a relative, a doctor, a priest, a teacher. . .). At the same time, the logical meaning of (22) is simply that ‘John talked to nobody’: (22) John didn’t talk to a soul. In the course of time, the emphatic, scalar properties of NPIs may gradually be lost, as they clearly have been in French to a large extent, such that only the abstract feature of negativity remains. This loss of emphatic meaning may take place at di¤erent times and at di¤erent paces for di¤erent language users, resulting in variability in the range of constructions in which n-words can occur. Given that syntax is frequently observed to lag behind semantics in grammaticalization, there is, moreover, nothing to prevent the process of semantic attrition from being e¤ectively completed, even as speakers still feel compelled to combine n-words with a preverbal ne. The other question with respect to grammaticalization that is raised by the paper under discussion is the question of what triggered the diachronic changes that the French n-words have gone through (and more broadly, what triggers grammaticalization in general). According to the proposed model, it is very clearly syntactic changes that drive the changes in the semantics of n-words. De´prez rejects, however, the recent proposal by Zeijlstra (2008) that it is changes in the nature of sentential negation that lead to changes in n-words, arguing that the former significantly precede the latter in the history of French, and suggesting instead that the syntactic changes internal to the n-words themselves were driven by the loss of bare nominals elsewhere in the language. While the latter proposal is in itself an attractive idea, there are, as far as I can tell, two problems with it: One is the fact that, as already discussed, it does not generalize to n-words of adverbial origin. The other problem is that, when combined with the idea that syntactic change drives semantic change and not the other way around (as is usually assumed at least in the grammaticalization literature originating from the functional end of the linguistic spectrum), this proposal ultimately makes a puzzle of the evolution of the n-word aucun as analyzed in De´prez & Martineau (2004): if this item was already in D0 in the oldest stages of French, why would it suddenly descend to the lower adjectival position at just the time when bare nouns were being eliminated? Surely, for such a structural change to be motivated, the change in the semantics of aucun that De´prez argues for, from context-independent positive meaning to context-dependent negative-polarity meaning, must have preceded, rather than followed, the syntactic change.

Discussion of De´prez

281

Without commenting specifically on Zeijlstra’s (2008) model, I would like to query De´prez’s rejection of the (to my mind, intuitively rather plausible) idea that changes in the sentential negation may have be instrumental in triggering analogous changes in the status of n-words. While the postverbal marker pas had indubitably become fully negative by the end of the Middle Ages, that does not entail that n-words, if evolving towards inherent negativity in analogy with pas, would have to evince all the features of their new status shortly after pas did so. First of all, the fact that sentential negation is syntactically di¤erent from n-word negation may mean that it is only with some delay that speakers feel the need to bring the two forms of negation into line. Secondly, it is common in the grammaticalization literature to distinguish between reanalysis as such and its subsequent actualization (e.g. Harris & Campbell 1995: 77¤ ), i.e. the process whereby a reanalyzed item or construction comes to assume all the superficial grammatical characteristics associated with the innovative analysis. This process is frequently, perhaps even typically, a protracted one, and it may never go to full completion in any given case. Indeed, if De´prez’s analysis is correct and rien, personne and aucun have been reanalyzed as belonging in the strong-quantifier SDP-slot, the process of actualization must still be on-going in this case, despite the fact that it has been centuries since French lost the bare-noun construction as a productive strategy, in so far as the three items in question clearly have yet to fully adapt their distributional behavior to this new structural slot. Not only do they favor ne-retention and remain capable of NPI uses, as we have seen, but the results of a Google-search suggests that they do, in fact, (contra De´prez) allow quantitative en-cliticization, which, according to Zamparelli’s model of the DP, is a feature of the lower, weakquantifier, PDP-slot: thus, De´prez’s starred example (35b), repeated below as (23), is attested in a number of instances, as are corresponding constructions with rien like that in (24):1 (23) Je n’en connais personne. ‘I don’t know any of them.’ (24) Il n’en reste rien. ‘There’s nothing left of it.’

1. Indeed, the latter construction appears to be enshrined in the title of a hit song by French rock singer Eddy Mitchell, S’il n’en reste rien ‘If there’s nothing left of it’.

282

Maj-Britt Mosegaard Hansen

It is in my view important, in developing models of the evolution of negative markers, that we be careful not to idealize the database. There is ample evidence, across this and other areas of the grammar, that language users can live happily with structural and distributional variation and, indeed, ambiguities for very considerable lengths of time. In a language like French, we should probably expect such variation and ambiguities to be all the more evident due to the tension between, on the one hand, a culturally strong tradition of codification and prescriptivism with respect to the more formal, in particular written, registers, and, on the other hand, the inevitable evolution of informal, in particular spoken, registers. Actual usage suggests that French speakers are not operating with clearly separate grammars for formal versus informal registers, but are rather making contextually variable use of a (more or less extended, according to the individual) range of constructional possibilities. Our models, both of specific areas of grammar and of grammar in general, need to reflect this linguistic reality, which although it may be more complex in French than in some languages, is surely not unique to that language.

References Sources Elicop http://bach.arts.kuleuven.be/elicop/ (consulted April 2010) Frantext http://atilf.atilf.fr/frantext.htm (consulted April 2010) Textes de franc¸ais ancien http://www.lib.uchicago.edu/efts/ARTFL/projects/TLA/ (consulted April 2010) Studies Ashby, William J. 1981 The loss of the negative particle ne in French: A syntactic change in progress. Language 57 (3): 674–687. Ashby, William J. 2001 Un nouveau regard sur la chute du ne en franc¸ais parle´ tourangeau: s’agit-il d’un changement en cours? Journal of French Language Studies 11: 1–22. Coveney, Aidan 1996 Variability in spoken French. A sociolinguistic study of interrogation and negation. Exeter: Elm Bank Publications. De´prez, Viviane & France Martineau 2004 Micro-parametric variation and negative concord. In: Julie Auger, Clancy Clements & Barbara Vance (eds.), Contemporary approaches to Romance linguistics, 139–158. Amsterdam: Benjamins.

Discussion of De´prez

283

Eckardt, Regine 2003 The structure of change. Meaning change under reanalysis. Berlin: Humboldt University. (Habilitation thesis. Revised version published by Oxford University Press in 2006, as Meaning change in grammaticalization.) Fonseca-Greber, Bonnibeth Beale 2007 The emergence of emphatic ne in conversational Swiss French. Journal of French Language Studies 17: 249–275. Hansen, Anita Berit & Isabelle Malderez 2004 Le ne de ne´gation en re´gion parisienne: une e´tude en temps re´el. Langue et socie´te´ 107: 5–30. Hansen, Maj-Britt Mosegaard 2001 Syntax in interaction. Form and function of yes/no interrogatives in spoken standard French. Studies in Language 25 (3): 463–520. Harris, Alice C. & Lyle Campbell 1995 Historical Syntax in a cross-linguistic perspective. Cambridge: Cambridge University Press. Ingham, Richard 2008 The spread of indefinite aucun in later AN negative clauses. Paper presented at the 1st Cycles of Grammaticalization seminar, Aston University, June 7, 2008. Martin, Robert 1966 Le mot rien et ses concurrents en franc¸ais (du XIVe sie`cle a` l’e´poque contemporaine). Paris: Klincksieck. May, Robert 1989 Interpreting logical form. Linguistics and Philosophy 12: 387–435. Muller, Claude 1991 La ne´gation en franc¸ais. Geneva: Droz. Sanko¤, Gillian & Diane Vincent 1977 L’emploi productif de ne dans le franc¸ais parle´ a` Montre´al. Le franc¸ais moderne 45: 243–256. Zeijlstra, Hedde 2008 Negative concord in syntactic agreement. MS, University of Amsterdam.

Negative polarity and the quantifier cycle: Comparative diachronic perspectives from European languages David Willis 1. Introduction It is well known that changes in indefinite systems often lead to an item becoming restricted to more ‘negative’ contexts, whether this is through the introduction of a restriction on its distribution to weak negative polarity contexts (NPI contexts: interrogatives, conditional, comparatives, negatives etc.) or an outright restriction on an item’s use to exclusively negative environments. The development of French rien ‘nothing’ or personne ‘no one’ from earlier generic nouns (Latin rem ‘thing’ and persona ‘person’) is a well-known example of this phenomenon. In turn, new non-negative indefinites are often created (such as French quelqu’un ‘someone’), giving these patterns of development a cyclic quality. Ladusaw terms this development the ‘argument cycle’, and expounds it in terms reminiscent of the classic Jespersen cycle for sentential negation: The development of negation-expressing argument phrases from regular indefinite arguments has the following stages: first the argument is a regular indefinite argument, then it becomes a co-occurring ‘supporter’ of the clausal negation, and finally it becomes an independent expressor of negation. We could call these the ‘one thing’, ‘anything’, ‘nothing’ stages of the Jespersen argument cycle. (Ladusaw 1993: 437–438)

However, examples have also been cited of items developing in the reverse direction, developing more ‘a‰rmative’ distributions over time (Ja¨ger 2008). While on the face of it, these changes have the flavour of a ‘‘random walk’’ through the space of parameters or lexical options, they are clearly not unconstrained, and an adequate account of language change will be able to account for what we observe or do not observe. In this paper, I will examine pathways of development in indefinites in the light of work on child language acquisition, asking both how our knowledge of acquisition

286

David Willis

stages can inform research in historical linguistics and vice versa. In doing so, I aim to contribute to the task of ‘‘distinguishing possible from impossible changes, which I take to be a central task of any theory of historical change’’ (Lightfoot 1977: 192), on a principled basis grounded in language acquisition. This raises some general questions that I aim to address in this paper: (a) Is the quantifier cycle unidirectional? Do we ever find negative distributions and meanings being lost? (b) How can shift from NPI to negative quantifier (via an n-word stage) be modelled? How can simultaneous a‰rmative and negative meanings in the same contexts be dealt with? (c) What sort of theory of acquisition would account for the changes that we find?

2. Ja¨ger’s bidirectional hypothesis for the development of indefinites With the aim of modelling changes in the historical distribution of indefinites, Ja¨ger (2010) proposes a feature system to distinguish three types of indefinite using two features [AFFECTIVE] and [NEGATIVE]. A¤ective is to be understood as a cover term for those contexts (negative, interrogative, conditional, comparative etc.) that license negative polarity items, in the tradition going back to Klima (1964). She assumes lexical underspecification theory; hence, only marked features are stored in the lexicon. This means that English distinguishes the following three sets of indefinites: (1) normal and positive polarity (PPI) indefinites [ ] somebody NPI-indefinites [AFFECTIVE]

anybody

n-indefinites [AFFECTIVE, NEGATIVE]

nobody

Schematically, we can think of this feature system as providing for the following options in specifying the available distributions for indefinites:

Negative polarity and the quantifier cycle

287

(2)

(adapted from Ja¨ger 2010: 790) These features impose distributional restrictions: [AFFECTIVE] must be licensed semantically by a downward-entailing or non-veridical environment; while [NEGATIVE] is licensed by local Agree with a c-commanding [NEGATIVE] element (which may be a covert negative operator). This means that someone is free to occur anywhere, while anyone is limited to a¤ective environments and no one is limited to the scope of negation. In order to arrive at the observed distribution of these items, Ja¨ger superimposes on this feature system an Elsewhere Condition. This is itself parameterized, its strength varying from language to language, and rules out items in contexts where they are in principle licensed, because some other more specific element is also licensed in this context. This condition is more active in some languages than in others, and its strength may vary diachronically. Elsewhere Conditions are generally thought of as universal phenomena, so it is odd that an Elsewhere Condition should vary across time and space. As Biberauer & Roberts (2010) note, this complicates the account by e¤ectively introducing a form of Optimality Syntax into it, with di¤erent languages able to rank the Elsewhere Condition more or less highly. For instance, in the history of English, any-series items such as anyone spread to negative environments in Early Modern English (that is, sentences of the general type I didn’t see anyone become available in addition to I saw no one). This represents a weakening of the blocking e¤ect of the Elsewhere Condition. While this models the change successfully, it raises questions about how the strength of the blocking e¤ect is acquired, and how it can strengthen or, in this case, weaken over time.

288

David Willis

Inevitably, the shift towards Optimality Syntax introduces the need to compare derivations, as a derivation can be ruled out by reference to a more specific, more highly valued derivation. This represents a complication in the account, and, as such, needs to be fully motivated. The Elsewhere Condition is needed, among other things, for PPI-indefinites, which are featurally identical to ordinary items, having no specific features at all. Hence the ungrammaticality of (3), where, in principle, someone, having no need for licensing at all, is permitted, is due to the availability of (4), which includes a more specific licensing configuration, with an [AFFECTIVE] feature licensed in a negative environment. (3) *There wasn’t someone in the room. (4)

There wasn’t anyone in the room.

Ja¨ger applies this feature system both to account for specific developments in the history of German and to create a typology of possible historical changes. For German, she notes the following developments that amount to shifts in feature specification: (i) Middle High German dehein ‘any’ > Modern German kein ‘no’, originally an NPI-indefinite, which occurred in [AFFECTIVE] contexts, but could not express negation on its own, occurred increasingly in negative contexts and could express negation on its own, becoming an n-indefinite: [AFFECTIVE] > [AFFECTIVE, NEGATIVE]; (ii) the PPI-indefinite ete(s)waz ‘something’ > etwas ‘something, anything’ extends its distribution to weak NPI contexts, and so covers all nonnegative contexts, whether a¤ective or non-a¤ective: this is [ ] > [ ], but with weakening of the Elsewhere Condition to allow it to occur in weak NPI environments; (iii) Middle High German ioman ‘anyone’ (Modern German jemand ) extends its use in the opposite direction: it was originally only found in NPI contexts, but was extended to non-a¤ective contexts becoming a normal indefinite, replacing ete(s)wer ‘someone’, and disappearing from negative contexts: [AFFECTIVE] > [ ] with the Elsewhere Condition strengthening to prevent its use in negative environments; (iv) io mer ‘any, any more’ was an NPI-indefinite but becomes ‘normal’ immer ‘always’; ete(s)wa ‘somewhere’ also left the system, coming to mean ‘approximately’: [AFFECTIVE] > [ ]. While this successfully characterizes the changes, there are some problems inherent in the use of the Elsewhere Condition: for instance, it is

Negative polarity and the quantifier cycle

289

not a set value for the entire language, but both weakens, as in (ii), and strengthens, as in (iii), in the same language but with respect to particular items, cf. also other criticisms made by Biberauer & Roberts (2010). Thus the Elsewhere Condition must actually be a value of a feature attached to an individual lexical entry, and this value will need to be acquired by children separately for each item or series rather than for the language as a whole. It would also be desirable to link the proposed changes to an account of first language acquisition to explain how these features can be innovated and how they can be lost against a background of generally successful acquisition. In the more general domain (and evident also in the German data), Ja¨ger proposes a ‘‘random walk’’ system, with all (adjacent) possibilities instantiated: (i) (ii) (iii) (iv)

[ ] > [AFFECTIVE] [AFFECTIVE] > [AFFECTIVE, NEGATIVE] [AFFECTIVE, NEGATIVE] > [AFFECTIVE] [AFFECTIVE] > [ ]

Table 1. Possible diachronic developments in indefinite systems (Ja¨ger 2010: 816) ‘normal’ or PPI indefinite ‘positive’ ‘negative’

NPI indefinite n-indefinite

This specifically involves the possibility that the features [AFFECTIVE] and [NEGATIVE] may fail to be acquired for a given lexical item. Ja¨ger cites the following examples, which I divide into two types: loss of [AFFECTIVE] and loss of [NEGATIVE]: (5) Loss of [AFFECTIVE] NPI > normal indefinite with German ioman > jemand (see above) American English anymore American English anymore leaves the system of indefinites and acquires a meaning that a particular state of a¤airs holds at the present time (or some other reference time) but not previously, as with Anymore, we eat a lot of fish (Hindle & Sag 1975). The same shift out of the system of indefinites is observed with English anyhow and anyway, which are now primarily discourse markers.

290

David Willis

(6) Loss of [NEGATIVE] (i) Welsh neb, Old Irish nech ‘someone, anyone’ < lengthened grade of the negation particle plus an interrogative indefinite *ne-kwo-, with these items losing their negative value; (ii) Slavonic neˇ-series items (e.g. Russian nekto, Bulgarian njakoj etc. ‘someone’): the same development. In the Celtic case, Ja¨ger seems to envisage Common Celtic [AFFECTIVE, NEGATIVE] > [AFFECTIVE] > Old Irish [AFFECTIVE] > Middle Irish [ ], while Middle Welsh retains the inherited [AFFECTIVE] pattern before returning to the inherited state [AFFECTIVE, NEGATIVE]. The Irish developments look like the inverse of what we find with French personne, creating the impression that we really are dealing with random paths and movements in either direction along a cline. In the Slavonic case, the development postulated is presumably, for most languages at least, [AFFECTIVE, NEGATIVE] > [ ], giving rise to a ‘bagel’-distribution (Błaszczak 2002, Pereltsvaig 2005), that is, a distibution with an item occurring in non-negative a¤ective contexts but not actually in negative contexts themselves. Within Ja¨ger’s model, an appropriate setting for the Elsewhere Condition will prevent use of these items in negative contexts after the change. In the Slavonic case, one would presumably allow for the possibility that the development was a staged one, since it is very di‰cult to understand how children exposed to an item that occurs only in negative contexts could interpret the data as supporting the hypothesis that the item occurs everywhere except in negative contexts, the precise inverse of what they are actually hearing. As other examples of ‘backwards’ developments, one might add the following: (7) Old French negative words like nul (< Latin nullus) and nesun (< nes þ un < Latin neipsum þ unus) lost their negativity and were used in non-assertive negative polarity contexts (Catalani 2001: 113– 14). In doing so, they assimilate their distribution to such items as rien ‘anything’ and aucun ‘anyone’ (Haspelmath 1997: 232–233). For much of the rest of this paper, I will consider the historical and acquisitional evidence for these changes. We will examine the conditions under which these features can be innovated or lost. In particular I will argue that the conditions under which acquisition failure of the [NEGATIVE] feature must be quite limited.

Negative polarity and the quantifier cycle

291

3. The acquisition and emergence of negative polarity: the abduction and emergence of the [AFFECTIVE] feature 3.1. Specialization for negative polarity Today’s negative polarity items did not always have a restricted distribution. They have typically conventionalized as negative polarity items at some point in their historical development. This process is a gradual development, and individual items may show a greater or lesser degree of specialization for NPI contexts. Hoeksema (1994) shows that there are semi-NPIs, items that show a skew towards being used in (weak) NPI contexts, but are not totally restricted, at least at the current time. This is illustrated in Table 2 for four verbs of indi¤erence (note that, in Table 2, bother refers to the verb only in the sense ‘annoy’). Table 2 shows that, while mind is straightforwardly an NPI, banned from a‰rmative environments, care and matter are semi-NPIs. They show a skewed distribution, being less frequent in a‰rmative environments than an ordinary synonymous verb such as bother, but are nevertheless not completely ruled out in such environments. Crucially here, the restrictions are semantically unmotivated, in the sense that broadly synonymous verbs have radically di¤erent distributions. Table 2. Distribution of four verbs of indi¤erence in Present-day English (Hoeksema 1994) Environment

care

matter

mind

bother

n ¼ 792

n ¼ 406

n ¼ 341

n ¼ 377

negative (%)

53

57

72

35

other neg (%)

12

7

20

7

interrogative (%)

15

13

7

11

a‰rmative (%)

20

20

1

48

Some modal verbs, namely English (auxiliary) need, Dutch hoeven, German brauchen and Mandarin Chinese yo`ng have shown increasing specialization in (weak)-NPI contexts in the course of their history (Hoeksema 1994). This skewing need not be derived simply from the item’s meaning: while certain semantic fields show a (real world) predisposition to being or becoming NPIs, items with very similar meanings can show significantly di¤erent distributions. As Hoeksema notes:

292

David Willis

For largely pragmatic reasons, some verbs tend to occur in nona‰rmative contexts more than other verbs. This may set the stage for further specialization, but the main point here is, that there is no necessity for this at all. (Hoeksema 1994: 280–281)

Examples of this specialization abound. Middle English, wiht (Modern English a whit), a minimizer, is found in both fully positive and negative contexts, but in Early Modern English becomes restricted to negative environments: (8) She was falle aslepe a lite wight. ‘She had fallen asleep a little bit.’ (Reeve’s Tale 363; c. 1386. OED s.v. wight) Here wight may have development its NPI-restriction after it was reanalysed as an adverb, hence the development is noun > adverb > NPI adverb. Similarly, we find conventionalization of French du tout (Detges 2003); in Old French du tout was not specialized for negative contexts: (9) Jherusalem prendront du tout a leur commant ‘They will take Jerusalem completely under their command.’ (Chev. au cygne, cit. Godefroy, Tot, Detges 2003) Such uses die out in the classical French period, as du tout acquires an increasingly fixed relation to the negator pas, associated with the fixing of word order pas du tout. The same development seems to be observed with German u¨berhaupt ‘at all’. In Middle High German u¨ber houbet, literally ‘over head’, meant ‘taken together, in sum’, originally a term in cattle trading, where it specifically meant ‘without re-counting the heads of the individual animals’ (Duden 2001). The modern meaning was established only in the eighteenth century, and, with it, a restriction to weak NPI contexts. Some such developments have been documented in detail, although more studies are needed to allow solidly based comparative work. Hoeksema (2007) shows that Dutch enig ‘some, any’ has narrowed its distribution over the last 400 years, largely disappearing from a‰rmatives and conditionals, as shown in Table 3. He characterizes this as a disappearance from nonveridical environments and a specialization for purely downward entailing environments (enig was already restricted to nonveridical environments at the start of the development).

293

Negative polarity and the quantifier cycle

Table 3. The historical distribution of Dutch enig (%) (adapted from Hoeksema 2007) Context

before 1600 N ¼ 109

1600– 1700 N ¼ 224

1700– 1800 n ¼ 375

1800– 1900 n ¼ 656

1900– 1950 N ¼ 451

1950– 2000 N ¼ 524

2000– 2007 n ¼ 248

negation

17

29

34

33

33

38

37

comparative

12

5

6

10

16

20

19

neg. pred.

5

3

6

6

8

5

7

‘without’

16

13

14

14

19

19

22

conditional

20

16

9

6

5

3

3

a‰rmative

13

17

13

15

8

3

2

Notice that the split here would justify the splitting of the [AFFECTIVE] feature so as to make a distinction between a feature [DOWNWARDENTAILING] and a feature [NONVERIDICAL]. 3.2. Acquisition of NPI distributions Van der Wal (1996) and Koster & Van der Wal (1995) examine the acquisition of the NPI distribution of the Dutch NPIs hoeven ‘need’ (in child language also ‘want’) and meer ‘any more’.1 The patterns can be illustrated with the case of hoeven, in adult language a strong NPI limited to direct or implied negation and a few other contexts (Van der Wouden 1997: 155–156). In the earliest child language (up to around age 2 years), hoeven is found exclusively with immediately adjacent niet ‘not’, and hence seems to be treated as an unanalysable single unit: (10) ’K hoef nie(t) s(l )ape(n). I need not sleep ‘I don’t want to sleep.’

(2;0.14) (Van der Wal 1996: §2.7.1)

1. For further discussion of the child acquisition of Dutch negation, see van Kampen (2007, 2010). Van Kampen notes that Dutch children go through a phase of spontaneously introducing negative concord into their language, a fact which may go some way to accounting for the ease with which this feature seems to be innovated diachronically.

294

David Willis

After this time, creative use of hoeven emerges. Children use hoeven without niet, but give it a negative meaning ‘don’t want’: (11) Ik hoef wijkoek. I need honeycake ‘I don’t want any honeycake.’

(2;04.28) (Van der Wal 1996: §3.3)

This suggests that children posit an inherently negative meaning for hoeven. Van der Wal, for instance, notes that ‘‘if the distribution of NPIs is so narrowly defined that negation becomes inherent in their meaning, then a separate negation marker may become redundant and therefore need not be expressed.’’ (Van der Wal 1996: §3.3) Another type of error is also found in children after age 2, however. Here, children overgeneralize the use of hoeven to contexts where it is not found in adult speech. These include contexts that contain a licenser (e.g. a modal particle) that seems similar to the licensing contexts allowed in adult speech, and also emphatic a‰rmative contexts: (12) nee, dat is ’t hele harde. ik hoef van jou zachte ’n no that is the very hard I need from you soft one (2;11.20) ‘No, that’s the hard one. I need the soft one from you.’ (Van der Wal 1996: §4.4.1) (13) hoef jij ook? need you also ‘Do you need to too?’

(2;11.20) (Van der Wal 1996: §4.4.3)

Van der Wal concludes that ‘‘three year old children appear to already know that hoeven is a special verb, with a distribution restricted to certain licensing environments’’ (Van der Wal 1996: §3.14). This last pattern of generalization is interesting from a historical perspective for two reasons. First, it provides some evidence that [AFFECTIVE] is the correct characterization of one of the features involved in NPI licensing. The a¤ective feature, originally posited to account for the distribution of do-insertion in English (Klima 1964), also picks out emphatic assertion as an [AFFECTIVE] context. The fact that children overgeneralize hoeven to precisely these contexts lends credence to the postulation of the feature. Secondly, emerging markers of negation often carry special pragmatic values of precisely the types that children seem to be associating with hoeven here: either pure emphatic negation (Kiparsky & Condoravdi 2006) or emphatic cancellation of a presupposition (Schwegler 1988, Zanuttini 1997).

Negative polarity and the quantifier cycle

295

Other studies of the acquisition of negative polarity support the idea that children identify items as being restricted to downward-entailing environments and make few mistakes either in the distribution or the interpretation of these items. Crain & Pietroski (2002) note that the error-free acquisition of the basic distribution amounts to a classic example of the negative evidence problem in language acquisition, and suggest that the concept of downward entailment must be part of the innate genetic endowment of Universal Grammar. Musolino, Crain & Thornton (2000), citing earlier unpublished work by O’Leary and Crain, note that the only serious distributional error that English-speaking children make is to use the positive polarity items some(thing) within the scope of negation: (14) a. b. c. d.

He didn’t get something to eat. Well, they didn’t get some food. None people had some presents. So he didn’t get some money.

(C.E-K. 4;6) (E.E. 4;7) (E.P. 4;9) (E.G. 4;10)

This confirms earlier findings by Klima & Bellugi (2004 [1966]: 355–356). In terms of Ja¨ger’s feature system, this would mean that the presence or absence of the basic features are easily acquired, while errors are made in establishing the strength of the Elsewhere Condition. There are no corresponding errors with negative polarity items, that is, children do not overgeneralize the use of the negative polarity item any to (assertive) a‰rmative contexts. On the basis of such child language acquisition studies, we can draw the following conclusions: – from the very earliest utterances, children are aware of the special relationship between certain items (such as Dutch hoeven and English anything) and negation, although not the precise nature of this relationship: the (ultimately correct) [NEGATIVE] feature for hoeven is posited early and an [AFFECTIVE] feature for anything is posited early, securely and correctly; – children entertain positing inherently negative meanings (as with Dutch hoeven) before retreating to NPI hypotheses as the nature of the relationship with negation; – children easily posit an [AFFECTIVE] feature for negative polarity items such as hoeven, and then fine-tune the distribution by narrowing it. We should expect some link between these patterns and what we find in historical developments.

296

David Willis

4. ‘Forwards’ historical developments Forwards historical developments involve the innovation of the feature [AFFECTIVE] in an item previously unmarked, or the innovation of the feature [NEGATIVE] in an item previously marked only [AFFECTIVE]. The second of these is clearly predicted by the acquisitional studies above: children are willing to posit inherently negative meaning in negative polarity items, and are ready to restrict such items to negative contexts as a stage in acquisition. Arrested acquisition at such stages would lead to the innovation of the [NEGATIVE] feature. We have also seen that children are readily prepared to posit the a¤ective distribution. While we have no evidence of narrowing to this distribution in acquisition, it does not seem unlikely, given children’s readiness to posit this distribution in general, for an item to innovate this feature. I shall illustrate these developments from the history of Celtic, but they are readily found in other languages too. 4.1. Welsh indefinites In the course of their history, items of the main series of Welsh indefinites (neb ‘anyone, no one’, dim both ‘anything, nothing’ and ‘any, no’) show a move towards increasingly negative meaning. While these were once found in all weak negative polarity contexts, they become inherently negative and restricted to negative contexts. Taken as a whole, within the historically attested period, we see a shift all the way along the quantifier cycle from generic items with no negative features, via [AFFECTIVE] to [AFFECTIVE, NEGATIVE]. Items of this series were available in Middle Welsh (1150–1500) in all negative polarity contexts, that is, negative clauses containing a sentential negation in (15), in interrogatives in (16), in conditionals in (17), and in comparatives in (18). Hence, in Ja¨ger’s model, they can be considered to bear the feature [AFFECTIVE]. (15) ny welynt neb. NEG see.IMPF.3P anyone ‘They saw no one.’ (Pedeir Keinc y Mabinogi 52.7. Middle Welsh) (16) a weleisti neb o ’r llys yn dyuot y’m hol i? Q see.PAST.3S anyone from the court PROG come.INF after.me ‘Have you seen anyone from the court coming after me?’ (Peredur 14.6–7. Middle Welsh)

Negative polarity and the quantifier cycle

297

(17) pei kymerwn i neb y ’m kanlyn, mi a if take.COND.1P I anyone to GEN.1S follow.INF I PRT ’th gymerwn ditheu. ACC.2S take.COND.1S you ‘If I took anyone to accompany me, I would take you.’ (Ystoryaeu Seint Greal 731. Middle Welsh) (18) canys mvy y carei ef Eudaf no neb for more PRT love.IMPF.3S he Euddaf than anyone ‘for he loved Euddaf more than anyone.’ (Brut Dingestow 71.24 ¼ 99.12–13. Middle Welsh) By Present-Day Welsh, however, two changes have a¤ected the status of these items. First, they come to express negative meanings in the absence of a marker of sentential negation. Negative interpretations of neb-series items in sentence fragments (typically answers to questions) are found at least as early as the seventeenth century: (19) Scot: Pa ryw newydd, noble Crwmel? Scot what kind news noble Cromwell Crwmel: Dim ond darfod cwrs y rhyfel. Cromwell nothing but finish.INF course the war ‘Scot: What news, noble Cromwell? Cromwell: Nothing except (only) that the course of the war has ended.’ (Rhyfel cartrefol 811. After 1660) In the nineteenth century, negative interpretations appear in certain non-elliptical syntactic environments. Initially, irrealis conditional clauses, as in (20), and absolute clauses, as in (21), seem most favourable to negative interpretations. (20) tase dim arath i ’ch atal chi be.COND.3S nothing other to 2P stop.INF you ‘if there was nothing else to stop you’ (Gwilym Hiraethog, Llythyrau ’Rhen Ffarmwr 62.15–16. 1870) (21) yr oedd y pin ysgrifennu wedi mynd ar goll PRT be.IMPF.3S the pen write.INF PERF go.INF on lose a dim sgrap o bapur gwyn yn y tyˆ. and no scrap of paper white in the house ‘the writing pen had become lost and not a scrap of white paper in the house.’ (Gwilym Hiraethog, Helyntion bywyd hen deiliwr 52.1–3. 1877)

298

David Willis

Absolute clauses containing neb-series items were once interpreted nonnegatively, as witnessed by the following example from the 1588 Bible translation (retained in the 1620 Bible) where dim cıˆg noeth byw, with quantifier dim, is interpreted as ‘any live raw flesh’ rather than ‘no live raw flesh’: (22) Ac edryched yr o¤eiriad, yna, os chwˆydd gwynn [a fydd ] and look.IMPERS the priest then if swelling white PRT be.FUT.3S yn y Croen, a hwnnw wedi troi y blewyn yn wynn, a in the skin and that PERF turn.INF the hair PRED white and dim cıˆg noeth byw yn y chwˆydd; any flesh naked live in the swelling

‘And let the priest look, then, if [there will be] a white swelling in the skin, and it has turned the hair white and there is any naked live flesh in the swelling’ (Tyndale Bible: ‘and let the preast se him. Yf the rysinge apeare white in the skynne a#d haue also made the heer white, a#d there be rawe flesh in the sore also’) (Leviticus 13.10. 1588) Secondly, neb-series items lose the ability to appear in certain nonassertive environments with non-negative meanings. In interrogatives and conditionals, non-negative interpretations of them were possible in the nineteenth century, as shown with dim ‘any’ in interrogative and conditional contexts in (23) and (24) respectively. (23) A oes dim rhew ac eira yn Awstralia? Q be.PRES.3S any frost and snow in Australia ‘Is there any frost and snow in Australia?’ (Awstralia a’r cloddfeydd aur. 1852) (24) tae dim synwyr yn dy goryn di be.COND.3S any sense in 2S skull you ‘if you had any sense in your skull’ (Gwilym Hiraethog, Llythyrau ’Rhen Ffarmwr 45.11. 1851) Their function in non-negative environments has been taken over by new innovated items based either on Middle Welsh ryw ‘kind, type’ as an adnominal quantifier ‘some (kind of )’ (e.g. rhywun ‘someone’ or rhywbeth ‘something’) or based on Modern Welsh unrhyw (< un ‘one, any’ þ rhyw ‘kind’) (e.g. unrhyw un ‘anyone’ or unrhyw beth ‘anything’). The neb-series

Negative polarity and the quantifier cycle

299

has thus become marked as [AFFECTIVE, NEGATIVE]. Of the two new series, the rhyw-series bears no features, while the unrhyw-series bears the [AFFECTIVE] feature. The development of the neb-series instantiates the standard ‘forward’ type of development, found widely elsewhere, for instance, with French n-words such as personne ‘anyone, no one’ and rien ‘anything, nothing’. The reanalysis of personne, rien etc. from nouns to indefinite pronouns in French is sometimes regarded as having been triggered by the innovation of the indefinite article (De´prez 1999; Roberts & Roussou 2003: 149–156; Roberts 2007: 143–149). The innovation of dim ‘thing’ > ‘anything’ in Welsh cannot be assimilated to the same account, however, because Welsh has never had an indefinite article. Apart from this, the developments are very similar. In terms of Ja¨ger’s model, the neb-series, like personne, becomes restricted to negative environments, and therefore changes feature specification from [AFFECTIVE] to [AFFECTIVE, NEGATIVE]. However, two aspects of the development do not fit easily within this analysis: first, the emergence of independently negative interpretations, such as that in (19) from the seventeenth century, predates the disappearance of these items from non-negative environments by several centuries, witness the retention of non-negative examples from the nineteenth century in (23) and (24). This is a variant of the classic n-word problem (Herburger 2001: 289–291): these items allow both negative interpretation in the absence of another negator, and non-negative interpretations in certain environments for an extended period. It is unclear how the items are to be characterized during this period: if they are [AFFECTIVE], then we need some account of what changed to allow the independently negative interpretations to arise. One solution is to say that the Elsewhere Condition became less strong: a sequence of OpB . . . [AFFECTIVE] was previously ruled out by the Elsewhere Condition, but subsequently this condition weakened. This seems unsatisfactory, since it is unclear what the more specific option was that ruled out OpB . . . [AFFECTIVE]. An alternative would be to posit two items during the transitional period (Herburger 2001), one with the feature specification [AFFECTIVE] and one with the feature specification [AFFECTIVE, NEGATIVE]. A second issue involves the staged loss of neb-items from non-assertive environments. As in various other languages, indefinites that become negative survive in comparatives. This suggests that a more fine-grained set of features is needed, one which allows comparatives to be distinguished from other non-assertive environments. In fact, an even more fine-grained set of features is needed, since closer examination shows a hierarchy of

300

David Willis

Figure 1. Percentage of clauses where ‘anyone, no one’ is expressed using neb in Welsh texts 1760–1940

contexts for the change. Figure 1 shows the frequency with which an indefinite pronoun referring to a person appears as neb in various clauses types in Welsh texts from the eighteenth and nineteenth centuries. Use of neb declines in both interrogatives and conditionals, leading to the present-day situation where it is ruled out in both environments. During the period under consideration, it competes largely with rhywun ‘someone’ in these environments, and is gradually replaced by it. Notice, however, that replacement of neb by rhywun is more advanced and is completed earlier in conditionals than in interrogatives. Table 4. Percentage of clauses where ‘anyone, no one’ is expressed using neb in Welsh texts 1760–1940 1760–1820

1820–1880

1880–1940

conditional

63%

n ¼ 19

7%

n ¼ 14

6%

n ¼ 16

interrogative

86%

n ¼ 14

73%

n ¼ 11

9%

n ¼ 11

comparative

79%

n ¼ 14

75%

n ¼ 24

88%

n ¼ 17

negative

86%

n ¼ 136

93%

n ¼ 96

94%

n ¼ 17

Negative polarity and the quantifier cycle

301

Ideally, a feature system should therefore distinguish these four contexts, and build in predictions about their relative susceptibility to change. In this respect, Welsh shows a clearer picture than that seen for French, at least in existing research. The personne-series has disappeared from most questions and conditionals in French, but survives marginally in high-style rhetorical questions, and is used freely in comparatives, beforeclauses, indirect negation and negative contexts (Muller 1991: 265). Essentially, then, this is the same as Welsh. According to Catalani, personneseries items are still found in conditionals and interrogatives in the sixteenth century, albeit less frequently than in earlier Middle French (Catalani 2001: 141). However, the present-day situation had essentially been reached by the seventeenth century (Catalani 2001: 169). Catalani’s data provide no indication of whether these items disappeared more quickly from conditionals or from interrogatives, implying little di¤erence between the two contexts. Further research would be needed to establish whether any di¤erence could be detected upon closer examination. With respect to the development of rhywun ‘someone’, the underspecification account is actually rather attractive: in the nineteenth century, as neb was becoming more negative, its place was taken by the ordinary indefinite rhywun ‘someone’ in interrogatives and conditionals. In the twentieth century, however, rhywun has begun to lose ground, but remains grammatical, in these contexts, because a new more specific item, unrhyw un ‘anyone’ has been innovated. An account that says that rhywun always had neutral [ ] specification but expanded or contracted its distribution in response to what other, more specific items were available, seems quite attractive. The Welsh development shown in Figure 1 suggests a historical hierarchy along which narrowing proceeds, with an item disappearing more readily from the environments towards the left of the hierarchy: (25) conditional > interrogative > comparative > negation Such a hierarchy conforms to that proposed on the basis of typological (rather than diachronic) generalizations by Haspelmath (1997). Haspelmath formulates a ‘semantic map’ of indefinite pronouns which defines a set of possible synchronic systems. Diachronically, assuming something like Hawkins’s Universal Consistency in History (Hawkins 1979, 1983), Haspelmath’s semantic map predicts certain pathways of narrowing to be possible, since they give rise to intermediate systems that represent possible languages, while ruling out those whose intermediate stages are not defined as possible languages by the map.

302

David Willis

Figure 2. Hapselmath’s implicational map of possible indefinite pronoun systems (Haspelmath 1997: 64)

There in fact seems to be a partial implicational hierarchy regulating the order in which a negative polarity item will disappear from particular contexts. We observe, for instance, that, in both Welsh and French, indefinites disappear last from comparatives and ‘before’-clauses (and similar contexts). Furthermore, there are languages where items with a former wider distribution have contracted to only interrogatives, comparatives and negation, for instance, Italian nessuno ‘anyone, no one’ and Persian hicˇ- / Turkish hic¸-items (Haspelmath 1997: 262–263, 282–284, 286–287).2 4.2. Romance indefinites The data presented in the previous section demonstrate the need for a more fine-grained feature system to model the shift of indefinites towards being inherently negative. Martins (2000) addresses this need in her analysis of the development of indefinites across Romance (see also Roberts 2000). She uses a three-feature underspecification-based system, with the features [negative], [affirmative] and [modal]. The feature [modal] is used, in a somewhat nonstandard way, to refer to what Ja¨ger calls [affective]. Each feature may underlyingly have the value þ (specified), 0 (nonvariable underspecified) or a (variable underspecified). Essentially, a þ value can agree only with a polarity head of the same type; a 0 feature clashes with a þ value of the same feature on a polarity head; and an a value is compatible with both the presence or absence of a þ value elsewhere. Consequently a þ value requires an item to occur in that environment, a 2. There is some disagreement in the literature as to whether Italian nessuno is available in conditionals. Haspelmath (1997: 262–263) and Martins (2000: 208) state that it is not, while Donati (2000: 64) states that it is. Parry (forthcoming) notes that ‘‘conditional contexts [are] becoming less favourable than interrogative ones’’, a general direction of change compatible with (25).

Negative polarity and the quantifier cycle

303

0 value prevents it, and an a value permits but does not require it. This obviates the need for an Elsewhere Condition, since the feature system allows the occurrence or non-occurrence of an item in each possible environment to be stated. Martins’s own summary of the feature values required for various modern Romance languages and earlier historical states is given in Table 5. Where two specifications are given, the item is analysed as being ambiguous between two feature sets. So, for instance, Earlier Old Romance has indefinites with the feature values [0 a¤, a neg, a mod]. The feature [0 a¤ ] means that they are incompatible with (assertive) a‰rmative environments, while the features [a neg] and [a mod] permit them to occur in either negative environments or in modal (i.e. non-negative non-assertive) environments. In Modern French, on the other hand, indefinites such as personne are analysed as being ambiguous between ‘modal’ indefinites, bearing the features [0 a¤, 0 neg, þmod] and therefore occurring only in non-negative non-assertive environments, and strong NPI indefinites bearing the features [0 a¤, þneg, 0 mod] and therefore occurring only in negative environments. Table 5. Feature specifications for Romance indefinites (adapted from Martins 2000: 208) Earlier Old Romance

[0 a¤, a neg, a mod]

Romanian, Venetian

[0 a¤, a neg, 0 mod]

Later Old Romance, Catalan

Galician, Spanish, Italian, French Portuguese

weak NPIs [0 a¤, a neg, a mod] strong NPIs [0 a¤, þneg, 0 mod] modal NPIs [0 a¤, 0 neg, þmod] strong NPIs [0 a¤, þneg, 0 mod] [0 a¤, þneg, 0 mod]

Martins notes that there is a great deal of fine-grained variation in Romance in the non-negative contexts in which indefinites may be found. She deals with this crosslinguistic variation by allowing the [mod] feature, essentially the equivalent of Ja¨ger’s [affective] feature to itself have an attribute value specifying the exact list of environments in which an item appears. Hence, for instance, French personne as a modal indefinite bears the feature [þmod], which itself bears a feature specifying that a comparative context licenses this particular item. This of course requires a more articulated theory of syntactic features, but this in itself should not

304

David Willis

be an insuperable weakness in the analysis. More worrying is the degree of freedom which it allows in the space of possible grammars and, therefore, in the possible diachronic pathways. Any list of environments appears to be a legitimate set of attribute values to the [mod] feature. For instance, the following appears to be a legitimate lexical entry: (26) [0 a¤, a neg, a mod] [interrogative] This would characterize a weak negative polarity item licensed only in the scope of negation or in the scope of an interrogative operator, but disallowed, for instance, in comparatives and conditionals. It seems unlikely that such an item could exist. This is because of (a) semantic reasons: in order to occur in interrogatives an item would have to appear in the other environments with the same semantic property (e.g. downward-entailing, anti-additive etc. for semantic reasons); and (b) typological/empirical reasons: it is not attested in Haspelmath’s typological work and is ruled out by his semantic map. The same may well be true of other combinations of attribute values on the [mod] feature. Furthermore, there is no hierarchical implicational relation between the three features, hence the following is also a possible lexical entry: (27) [a a¤, a neg, 0 mod] This represents an item licensed in a‰rmative and negative environments, but not in non-assertive environments. Martins (2000: 210) herself acknowledges that this type may not exist and that ‘‘a thoroughly worked out geometry of features will be needed in order to exclude the possibility of this type of polarity item in natural language’’. The feature system itself therefore imposes few limits on possible systems, and does not lead us to expect any particular pattern of change. However, Martins does impose restrictions on change externally: she observes that, in Romance, indefinites tended to reduce their degree of underspecification, going from [a neg] [a mod] to [þneg] (i.e. narrowing their distribution from non-assertive to negative only). This general trend is accounted for via patterns of acquisition: children fail to notice the nonnegative environments in which certain indefinites are found, and hence posit [þneg] instead of [a neg] [a mod] as their specification (Martins 2000: 206–207, 215–216). However, such an approach fails to make strong enough predictions about acquisition, for instance, that if children observe a polarity-sensitive ordinary indefinite to occur in conditionals, they can

Negative polarity and the quantifier cycle

305

infer that it will also be permitted in interrogatives. A fully adequate feature system would incorporate such facts. Van der Wouden (1997: 130–45) proposes this general type of system when he suggests that there are strong NPIs (which require an antimorphic context e.g. not), medium NPIs (which require only an anti-additive context e.g. not or no one) and weak NPIs (which require only a monotone decreasing context e.g. not, no one or at most three), see Table 6. Table 6. Semantic feature specifications of various English environments. antimorphic

anti-additive

monotone decreasing

at most three

–

–

þ

no one

–

þ

þ

not

þ

þ

þ

An item compatible only with not (such as a bit happy) can then be characterized as bearing [antimorphic] [anti-additive] and [monotone decreasing] licensing features, while an item compatible with either not or no one (such as yet) has only the features [anti-additive] and [monotone decreasing] to be licensed (Van der Wouden 1997: 141–142). This allows historical change to be conceptualized in terms of lexical items gaining licensing features. In the current scenario, some elaboration of the feature system would be necessary, since contexts such as conditional and interrogative are not distinguished adequately on a semantic basis by these features. 4.3. The free relative and free choice cycles There is also a further diachronic complication, originating from the apparent existence of a separate implicational hierarchy regulating the historical development in which new indefinites arise from earlier free relatives in (i)–(iii) or free choice items in (iv)–(v): (i) French quelque ‘some’ (< quel X que [ce soit etc.] ‘whatever X [this may be etc.]’), quelqu’un ‘someone’ etc. (Foulet 1919); (ii) Breton un . . . bennak ‘some’ (< pa X bennak ‘whichever X [which X ever]’), unan bennak ‘someone’ etc. (Willis 2010, in press); (iii) Polish kto-kolwiek ‘anyone’ < Old Polish kto kolwiek (free relative) ‘whoever [may. . .]’ (Ba˛k 1966–; Błaszczak 2002, 2003; Cies´likowa 1965; Urban´czyk 1953–2002);

306

David Willis

(iv) Russian kto-libo ‘someone’ < ‘anyone you like’ < ljubo ‘dear’ (cf. ljubit’ ‘to love’) and kto-nibud’ ‘someone’ < ‘anyone it may be’ < ni ‘negative’ þ bud ’ ‘be’; (v) Welsh unrhyw un ‘anyone’, unrhyw ‘any’ < ‘any kind of ’ < un ‘one’ þ rhyw ‘kind’. The relevant pathways are: (28) free relative > conditional > interrogative free choice item > comparative > conditional > interrogative There may also be other variants and continuations: some items go on to develop into ordinary (positive) indefinite pronouns by spreading to a‰rmative contexts (French quelqu’un, Breton unan bennak, Russian ktonibud’ ), while others become full negative polarity items by spreading to negative contexts (Welsh unrhyw un). The pathway in (28) does not take these possibilities into account, and takes the items only as far the ‘bagel’ distribution (all negative polarity contexts except negation itself ) that we find for Polish kto-kolwiek and Russian kto-libo. This hierarchy allows the emergence of distributions that could not have arisen through the quantifier cycle, for instance, an item that appears in comparatives and conditionals, but not in interrogatives (perhaps Old Polish kto kolwiek, and some of the other items given above at points in their history). As a synchronic example of this distribution, consider the Finnish hyva¨nsa¨- and tahansa-series of indefinite pronouns. Items such as mika¨ hyva¨nsa¨ and mika¨ tahansa ‘anything’ have etymologies which indicate that they were originally free choice items, and to a large extent they retain this flavour in their current behaviour. Tahansa is derived from the possessive form of tahto ‘wish, intention’ (taht- þ third person possessive su‰x -nsa), suggesting a source construction of the form ‘what(ever) is his desire, what(ever) he likes’. Similarly, hyva¨nsa¨ is derived from the root hyva¨ ‘good’, again with a third person possessive su‰x, suggesting an original ‘what(ever) is his good, what(ever) is good to him, what(ever) he likes’. Su‰xation of a possessive su‰x to an adjective is widespread in Finnish, performing various functions such as expression of degree, to express a state, or to form adverbs (such as eninta¨a¨n ‘mostly’, archaic form eninta¨nsa¨). Again, then, this has a free choice source. While the free choice meaning remains, as in (29), these items can also now express indefinite pronoun meanings in equative comparison, as in (30) and in conditionals in (31) (Haspelmath 1997: 293–4).

Negative polarity and the quantifier cycle

307

(29) Kuka tahansa osaa ratkaista ta¨ma¨n ongelman. anyone can solve this problem ‘Anyone can solve this problem.’ (Haspelmath 1997: 294) (30) Ta¨a¨lla¨ on yhta¨ mukavaa kuin missa¨ tahansa muualla. here is as pleasant as anywhere else ‘Here is as pleasant as anywhere else.’ (Haspelmath 1997: 294) (31) Jos kuka tahansa saa osallistua, kuka tahansa voi if anyone can participate anyone can voittaa 100 mk. win 100 marks ‘If anyone can participate, anyone can win 100 marks.’ (Hakulinen & Karttunen 1973: 166) This leads to a synchronic distribution that could not have arisen via the quantifier cycle. An item that had proceeded along the quantifier cycle would have had to have disappeared from conditionals before it disappeared from interrogatives. In Welsh, the free choice item unrhyw ‘any’ spreads first to comparatives, where its use is already moderately advanced in the nineteenth century, and from there to conditionals, interrogatives, indirect and direct negation. Again, this is a pathway of change very di¤erent from that manifested by original NPIs. Further research is necessary to demonstrate and motivate the pathways for the development of free relative and free choice markers into ordinary indefinite pronouns. Haspelmath (1997) considers many of the relevant issues. The Finnish tahansa-items have gone a relatively short distance down the free choice cycle and further development may require semantic changes that will need to be accounted for, for instance, in some cases, the acquisition of the ability to be referential. This raises the interesting possibility that there may be no synchronic limits on the possible systems, but that synchronic distributions may be the result of the possible pathways by which these systems may arise. The di¤erence in the two hierarchies could be explained if the narrowing along the hierarchy in (25) were due to acquisition failure, and if the extension along the hierarchy in (28) were due to other factors: reanalysis of free choice meaning in negative polarity contexts as plain indefinite meaning with an NPI restriction (universal > existential quantification).

308

David Willis

In terms of feature systems, the free relative and free choice cycles pose problems for a concentric feature system in which the availability of an item at one point on the quantifier hierarchy in (25) automatically implies its availability in all contexts to the right. In fact, only Martins’s very permissive system could accommodate these developments, and even it could not o¤er much insight towards an explanation. 4.4. Conclusion In this section we have considered various instances of ‘forwards’ developments of indefinites towards becoming ‘more negative’. It appears that such developments conform to the diachronic hierarchy in (25), which mirrors in the diachronic domain the synchronic semantic map of Haspelmath (1997). Such patterns of historical development cannot be accounted for directly either by the feature systems of Ja¨ger, which is not fine-grained enough, nor by that of Martins (2000), which is too permissive. While it is tempting to adopt a feature-system of concentric licensing environments, we have also seen that the existence of patterns of development of free relatives and free choice items add complications that suggest that some of the synchronic patterns may actually be the by-products of recurrent historical pathways.

5. Backwards historical developments We can now turn to consider various proposed backwards historical developments, an important component of Ja¨ger’s bidirectional, random-walk system. I shall argue that the two most well-known ones, Celtic neb and Slavonic neˇ-series items, are not secure as evidence of the potential for loss of the [NEGATIVE] feature in language acquisition. These are a priori suspicious if we take seriously the notion that much language change is driven by acquisition failure. We saw above that children acquire knowledge of the availability of negative polarity items in negative contexts very easily indeed. This same view is implicit in Martins’s explanation for the quantifier cycle in Romance: If children rely on strong positive empirical evidence . . . in order to make decisions leading to a particular linguistic choice while building up their grammars, they may well not identify a less salient polar environment as a licensing context for a certain kind of polarity items. (Martins 2000: 206)

Negative polarity and the quantifier cycle

309

Martins identifies negatives as salient in the primary linguistic data for the acquisition of negative polarity items, while other contexts are liable to be missed because there are not salient. This reasoning would lead us to expect that the one feature of an evolving negative polarity item that would not be lost is the [NEGATIVE] feature. I will argue that this is indeed the case except in special circumstances. Other examples involve special cases: assimilation to existing series of items (hence providing children with a basis to abductively hypothesize a representation without a [NEGATIVE] feature) or else language contact (and hence, potentially, not involving child language acquisition at all). 5.1. The development of neb/nech in Celtic 5.1.1. Neb in Common Brythonic We have already seen (section 4.1) that, in the documented history of Welsh, it is clear that, in very general terms, the item neb ‘anyone, no one’ becomes more negative. In Middle Welsh, it has non-negative nonassertive uses and appears in contexts entirely unconnected with negation. However, in Present-day Welsh, it is an inherently negative item. This seems like a straightforward unidirectional development. However, problems arise when we turn to internal and comparative reconstruction. All medieval Brythonic languages share an adnominal quantifier neb, nep ‘any’. Cornish allows it freely across a‰rmative and negative contexts, while Middle Welsh and Middle Breton show more complex patterning. We have already seen that Middle Welsh neb is a weak negative polarity item, permitted in the full range of negative polarity environments, including conditionals, interrogatives and comparatives, cf. (15)–(18) above. Middle Breton nep is used only as an adnominal determiner ‘any, no’, for instance, nep den ‘any person’, but in the same range of weak negative polarity environments as Middle Welsh neb. However, in addition to this, both languages use neb as the pronominal antecedent of a free relative, even in fully a‰rmative environments. This is illustrated for Middle Welsh in (32). (32) A ’r neb a dodes hut ar y wlat, and the anyone PRT put.PAST.3S magic on the land a beris bot y gaer yma. PRT cause.PAST.3S be.INF the fortress here ‘And whoever bewitched the land caused the fortress to be here.’ (Pedeir Keinc y Mabinogi 56.4–5. Middle Welsh)

310

David Willis

Furthermore, neb is also an element within the specific unknown indefinite quantifier nebun ‘any’ (< neb þ un ‘one’), which occurs freely in both negative polarity and a‰rmative environments. Its use in an a‰rmative environment is given in (33). (33) [. . .] e wynvydedic wyry a emdangosses y nebun the blessed virgin PRT appear.PAST.3S to some yscolheic a dywedut urthav [. . .]. scholar and say.INF to.3MS ‘the Blessed Virgin appeared to some (a certain) scholar and said to him’ (Gwyrthyeu e wynvydedic Veir 14.5–6. Middle Welsh) Both these uses are archaic today. These uses of neb in Middle Welsh and Middle Breton are surprising, since these languages do not otherwise allow neb in a‰rmative contexts. When set against the fact that Cornish makes use of neb generally in a‰rmative contexts in Cornish, the Welsh and Breton use looks like the fossilized relic of an earlier more productive system. For instance, it looks as though Middle Welsh nebun was created as an item at a point when Welsh did allow a‰rmative uses of neb. This suggests that the Cornish pattern, with neb freely available in a‰rmative, non-assertive and negative contexts, is the one that should be reconstructed for the Brythonic parent language. Middle Welsh and Cornish use neb also as an animate indefinite pronoun. Middle Breton, apart from its use as an antecedent to free relatives, does not use it as a pronoun. However, such use is attested in Old Breton, and should therefore be reconstructed for the whole of the Brythonic parent language: (34) Na dimicit nep. NEG despise.IMPER.2P anyone ‘Do not despise anyone.’ (Fleuriot 1964: 262. Old Breton) We can conclude that the Brythonic ancestor of neb was both a pronoun and a quantifier, and was freely available in all environments. That is, we reconstruct no features restricting its distribution, and posit that Welsh and Breton innovated a restriction to [AFFECTIVE] contexts. This is not an unexpected development, since we know that items that do not participate at all in the negative system can enter it, developing a negative polarity distribution. That is, children sometimes posit a link with negative polarity from an apparently skewed distribution in the primary

Negative polarity and the quantifier cycle

311

linguistic data, even when no such link is present in the grammar of the adults producing the data. It is when we go back further, however, that real problems arise. Brythonic nep is cognate with the Old Irish indefinite pronoun nech (masculine and feminine), nı´ (neuter). In Old Irish, the ‘independent’ forms nech and nı´ are used as pronouns ‘anyone’ and ‘anything’ respectively. Parallel ‘dependent’ forms exist, nach (masculine and feminine) and na (neuter). These are used as adnominal quantifiers equivalent to ‘any’. These forms are found in negative polarity contexts, whether negative, as in (35), or other non-assertive environments, as in (36) (universal quantification). (35) nı´m-raib nı´ NEGþ1S-be.PRES.3S anything ‘may I not have anything’ (ZCP 7: 308 §1. Toner, Fomin, Torma & Bondarenko 2007, s.v. nı´ ) (36) cech duine shirfess ni fort every man seek.FUT.REL anything on.2S ‘every man who (whoever) shall ask anything of you’ (Leabhar Breac 462. Toner, et al. 2007, s.v. nı´ ) They also have non-negative uses rather more extensively than their Middle Welsh cognates. Nach etc. is used in a‰rmative environments to meaning ‘some’ and ‘something’: (37) ita´ na´ch cumachta fora cul na n-e´n-sa be.PRES.3S some power behind the birds-DEM ‘there is some power behind these birds’ (Serglige Conculain 7. Toner et al. 2007, s.v. nach) (38) nı´ do thabairt do neuch something.ACC to give.INF to someone.DAT ‘to give something to someone’ (glossing aliquid proferre) (Milan glosses 98.a.4. Toner et al. 2007, s.v. nı´ ) Old Irish nech may serve as the antecedent to a free relative, although in contrast to Middle Welsh usage in (32), it is not preceded by a definite article in a free relative construction:

312

David Willis

(39) comalnad neich forchanat fulfilling any.NEUT.GEN teach.PRES.3P ‘fulfilling of what they teach’ (Wu¨rzburg glosses 29a.11. Thurneysen 1946: 309) Old Irish nech and Middle Welsh neb clearly go back to a Common Celtic formation *ne-kwos. Given that we reconstruct a completely free distribution for the Common Brythonic ancestor of Middle Welsh neb, and given that we see an essentially free distribution for Old Irish nech, we can reconstruct (under identity) a completely free distribtuion for the ancestor item in Common Celtic. However, this seems to be fundamentally at odds with one interpretation of the reconstructed Common Celtic form *ne-kwos. This is clearly formed from a negative marker *ne and an interrogative *kwos ‘who’. Thurneysen interprets this as necessarily being a negative pronoun meaning ‘no one’ and, if so, it must have lost its negative force, later coming to mean ‘someone, anyone’ (Thurneysen 1946: 311). This change leads to the direct ancestor of the later insular Celtic languages (Common Brythonic and Old Irish). Thus, on Thurneysen’s view, *ne-kwos went from negative to non-negative within the development of Common Celtic, before becoming negative again in Welsh. This would involve a change in syntactic pattern of the type given in (40). (40) NEG V . . . NEG-indefinite > NEG V . . . (non-NEG) indefinite Haspelmath suggests that this is an impossible direction of change (Haspelmath 1997: 230–233). He considers the Celtic case as a possible counterexample to this generalization, and it is in fact the only possible counterexample for which he is not able to suggest an alternative account. There is, however, another account, which would not give rise to such concerns. Lewis & Pedersen (1937: 233) seem to envisage the original item to have been a free choice pronoun, suggesting that *ne-kwos was short for *kwos ne-kwos ‘someone, someone not’, which seems to imply an original meaning of ‘someone or other, anyone or other’. If so, we could reject this example as an instance of the development in (40). There are parallels for Lewis and Pedersen’s suggestion, for instance, the parallel formations of Lithuanian kas ne kas ‘something, someone’ and Hindi / Urdu koii na koii ‘somebody’ (Haspelmath 1997: 232): (41) Kas ne kas jau padaryta. (Lithuanian) what.NOM NEG what.NOM already do.PTC ‘Something has already been done.’

Negative polarity and the quantifier cycle

313

(42) Kas ne kas, o jau mano brolis (Lithuanian) who NEG who.NOM but EMPH my brother tikrai nenuvils. certainly NEG.let.down ‘If there is anyone who will not let (us) down, that is my brother.’ (‘Whoever [may let us down], my brother will certainly not let us down.’) (Keinys 1993) If Lewis & Pedersen are correct, then we have an original free choice indefinite pronoun that generalizes as an ordinary indefinite in Common Celtic, and then narrows towards negative environments in Welsh. We have already seen (section 4.3 above) that the first part of this development is attested in various languages. While it is di‰cult to demonstrate the correctness of one account over the other in cases of historical developments within fairly distant proto-languages, the Lewis & Pedersen account o¤ers a perfectly coherent alternative that is consistent with general patterns of change. 5.2. Middle Cornish ne¤ra ‘always’ Another counterexample is real, but is entirely explained by the conditions of language contact in which it arose. This concerns the Middle Cornish item ne¤ra, a loan from Old or Early Middle English næfra ‘never’. Strikingly, this item has an unexpected a‰rmative use to mean ‘always’, that is, it occurs both within the scope of negation, where it means ‘always not, never’, and outside the scope of negation, where it means ‘always’: (43) ha ne¤ra me a ’th vynyk. and ever I PRT you bless.PRES.3S ‘and I shall always bless you.’ (Bewnans Ke 791. Middle Cornish) On the face of it, then, ne¤ra has gone from negative to non-negative as it was transferred from English to Cornish. To understand why this happened, we need to look at the Middel Cornish system of indefinites more generally. Middle Cornish has a single series of indefinites with a distribution that is not sensitive to negative polarity; that is, all items are found in negative, non-assertive and a‰rmative contexts, and are thus characterized as featureless [ ]. The main forms themselves are given in (44).

314

David Willis

(44) nep-series [ ] person den / nep ‘someone, anyone’ thing nep peyth / nep tra / tra ‘something, anything’ quantity nep N ‘some, any’ time ne¤ra / bythqueth / byth ‘ever, always’ place

(in) nep pow / nep le / nep tu ‘somewhere, anywhere’

Examples in (45) and (46) show lack of sensitivity to negative polarity: nep le ‘somewhere, anywhere’ is used indiscriminately in an a‰rmative context in (45) and in a negative context in (46). (45) Hy re gafes dyhogel /dor dyseghys yn nep le. she PERF get.PAST.3S certainly earth dried.out in some place ‘She has certainly found dry land somewhere.’ (Origo mundi 1143–4. Middle Cornish) (46) rag bytqueth my ny welys /benen thy’m a for ever I NEG see.PAST.1S woman to.me REL wel plekye /wheth yn nep le better please.IMPF.3S yet in any place ‘For never have I seen a woman who pleased me more in any place.’ (Origo mundi 2107–9. Middle Cornish) To explain the distribution of ne¤ra then, we merely need to understand that the general patterning of the entire Middle Cornish system was extended to ne¤ra as soon as it entered the language. Ne¤ra is thereby assimilated by adopting the distribution of its Cornish equivalent bythqueth, and is therefore found in both negative and a‰rmative contexts. In crossing from English to Cornish then, ne¤ra has lost a [NEGATIVE] feature. However, here it is clear that the mechanism of language contact is responsible: Cornish speakers assumed that ne¤ra had the same distribution as the nearest equivalent item in their own language, namely bythqueth, which was featureless. For this reason, ne¤ra was featureless from the start in Cornish. 5.3. Scottish Gaelic ca`il, sian etc. ‘anything’ Scottish Gaelic has created a number of new indefinite pronouns from minimizers, completely bleaching the lexical content of these items in the

Negative polarity and the quantifier cycle

315

process. For instance, we have sian / sı`on ‘anything’ (Uist Gaelic) < Old Irish sı´an ‘a term applied to various kinds of continuous or prolonged sound’ including ‘howling or roaring of the wind or sea; lowing (of cattle)’, ‘humming; whistling, whirring’ etc. (cf. English a squeak); and ca`il from OIr. ca´il ‘quality, property; characteristic, trait; repute, reputation; measure, amount; way, respect’ (Toner, Fomin, Torma & Bondarenko 2007). The only plausible context for reanalysis and the relevant semantic shift in these cases is in a negative environment with these items used as minimizers: (47)

I didn’t hear [DP a squeak] > I didn’t hear [DP anything].

So, initially we have N [ ] > N [AFFECTIVE, NEGATIVE]. However, today, they are all used in a range of non-negative, non-assertive contexts: (48) Cha d’ith mi ca`il. NEG PRT eat.PAST I anything ‘I ate nothing.’

(Wentworth 2000)

(49) A bheil ca`il cea`rr? Q be.PRES anything wrong ‘Is anything the matter?’

(Wentworth 2000)

`ımpidh air daoine a (50) Bha iad a’ cur be.IMPF they PROG put.INF appeal on people REL chunnaic ca`il bruidhinn ri na polais. see.PAST anything speak.INF with the police ‘They were appealing to people who saw something/anything to speak to the police.’ Hence we have a subsequent development from N [AFFECTIVE, NEGATIVE] to D [AFFECTIVE]. The ‘backwards’ move here is from negative indefinite / strong NPI to weak NPI. This has to be explained analogically: other indefinite pronouns in Scottish Gaelic have an [AFFECTIVE] distribution, and this was extended to the new items. 5.4. The Slavonic nekto-series Finally, consider the reconstructed history of the Slavonic nekto-series of pronouns. In Russian, which is typical of the modern East and South Slavonic languages in this respect, there are two relevant series of indefinite

316

David Willis

pronouns, the nikto-series and the nekto-series.3 The nikto-series is straightforwardly negative. It is restricted to negative contexts only and, as far as we can tell, always has been so restricted: (51) Nikto ne prisˇel. (Russian) no.one NEG come.PAST.3S ‘No one came.’ (52) Esli *nikto / kto-to prixodit [. . .]. (Russian) if no.one / anyone come.PRES.3S ‘If anyone comes. . .’ Another series, the nekto-series, has the inverse distribution, being banned from negative contexts, but grammatical everywhere else: (53) *Nekto ne prisˇel. someone NEG come.PAST.3S

(Russian) (‘Someone didn’t come.’)

(54) Nekto s buketom cvetov sprosil someone with bouquet.INST flowers.GEN ask.PAST.3S nomer doma. (Russian) number.ACC house.GEN ‘Someone with a bouquet of flowers asked about the number of the house.’ (Kuznecov 1998) How did the current distribution of the nekto-series arise? Ja¨ger considers two possible reconstructions of the Slavonic nekto-series of indefinite pronouns. In the earliest attested Slavonic languages, this series had the distribution of an ordinary indefinite, occurring in a‰rmative contexts, and being unavailable in negative contexts: (55) Jako zˇe bo se neˇkto zemlju razoret’, as PRT PRT PRT someone land.ACC plough.PRES.3S drugyj zˇe naseˇet’ [. . .] other.NOM PRT sow.PRES.3S ‘Just as someone ploughs the land, another person sows it’ (Povest’ vremennyx let 67.34–5. Old East Slavonic) 3. Russian has a substantial number of other indefinite pronoun series, giving pronouns such as kto-to, kto-nibud’, kto-libo, etc., some of which were discussed above. For present discussion, however, only the nikto-series and the nekto-series are relevant.

Negative polarity and the quantifier cycle

317

(56) Isu˘ zˇe recˇe prikosno˛ se˛ mı˘neˇ neˇku˘to Jesus PRT say.PAST.3S touch.PAST.3S REFL me.DAT someone ‘Jesus said, ‘‘Someone touched me.’’ ’ (Codex Marianus, Luke 8.46. Old Church Slavonic, South Slavonic) According to one hypothesis, which Ja¨ger prefers, these derive from negative indefinites, and this would therefore be an instance of [NEGATIVE] > [ ]. According to the other hypothesis, this would be an instance of free choice pronoun > ordinary indefinite pronoun, e¤ectively [ ] > [ ]. The Common Slavonic reconstructed form of these items involves a long /e/ in the first syllable, for instance *neˇku˘to ‘someone’, clearly written in the earliest Slavonic languages, as in the Old East Slavonic and Old Church Slavonic examples above. Thus, while the second element is clearly the interrogative pronoun *ku˘to ‘who’, the nature of the first element is not clear, since it is not identical to the negative marker *ne, with a reconstructed short vowel. The first hypothesis says that the first element is the negative particle in a lengthened grade (Vasmer 1953: 209) or with ‘a¤ective’ secondary lengthening of the first vowel (Trubacˇev 1974–: xxiv.92–3). The formation therefore goes back to Indo-European. While ablaut gradations are of course a central part of Indo-European inflectional and derivational morphology, it is not clear what function an ablaut gradation of the negative particle could play, or indeed on what basis it could have arisen. Ja¨ger cites the parallel formations in Celtic and Baltic as support for this hypothesis. We have actually already seen both (Welsh neb/Irish nech and Lithuanian kas ne kas). However, all these really show is that new pronouns can be formed from a negative particle plus an interrogative. They do not demonstrate that the resulting pronoun will be restricted to negative environments, and we have seen that there is good evidence to think that the resulting pronoun might actually typically have a free choice function. The second hypothesis derives it from a Common Slavonic sequence parallel to Old Church Slavonic ne veˇ ku˘to. . . ‘I don’t know who. . .’ (Haspelmath 1991: 107; 1997: 131–2; Miklosich 1886: 214). This is a frequently attested grammaticalization path for indefinite pronouns, and Haspelmath cites parallel examples from a range of languages, including Middle High German, Old English, Old Norse and French. Since veˇ ‘I know’ actually exists in the earliest Slavonic languages, this formation would have to date to Common Slavonic itself or perhaps Common Balto-Slavonic rather than any earlier. Ja¨ger rejects this on phonological

318

David Willis

grounds, arguing that there is no evidence for the element veˇ having been present. The vowel of veˇ survives in the contraction of ne veˇ to neˇ, so in that sense there is evidence for it. However, she is right to imply that this contraction is not phonologically regular. Here it is in good company. Of all of the examples of the emergence of free choice pronouns from sequences ‘I don’t know who’ etc. cited by Haspelmath, not one is phonologically regular. This seems to be typical of cases where whole clauses undergo grammaticalization, cf. similar English cases such as whatchamacallit < what you may call it. Against this background, it would actually be surprising if this case were phonologically regular. In acquisitional terms, the first hypothesis would require acquirers of Common Slavonic to have overlooked all of the evidence for the [NEGATIVE] feature in these items, yet we have seen that this is the one aspect of the distribution of indefinites that children are very good at. On the other hand, the development of free choice pronouns into ordinary pronouns is very common crosslinguistically. We have seen other examples of it, and it does not seem to pose any acquisitional problems. 5.5. Conclusion Given that the only evidence for ’backwards’ developments involving the loss of the feature [NEGATIVE] comes from two reconstructed cases, both of which are amenable to plausible alternative analyses, it does not seem that we have good grounds for believing that changes in the development of indefinites really following a ‘‘random path’’, and that the path away from [NEGATIVE] is freely available. When it is available at all, it is under very special circumstances. Of course, the Cornish example is real, being based on attested material, but it shows us rather that such changes are possible in second language acquisition where speakers may use the grammar of their native language to form hypotheses about lexical items of another language without any regard to evidence from that other language.

6. Overall conclusions We have considered examples of the various possible directions of development in indefinite systems, both ‘forwards’ developments innovating narrower restrictions on the distribution of items, and putative ‘backwards’ developments, relaxing these restrictions. I have argued that there

Negative polarity and the quantifier cycle

319

are reasons to doubt that developments in the distribution of indefinites follow a ‘‘random walk’’ around the space of possible systems. In particular, I have doubted the possibility, during ordinary first language acquisition, of the loss of a restriction to negative environments, that is, the feature [NEGATIVE]. Not only does the evidence from first language acquisition show that this feature is easily, sometimes too easily, acquired by children, we have also seen that, while the three other developments are all amply attested, the evidence for [NEGATIVE] > [ ] is sparse. Apart from special circumstances involving individual items rather than series of pronouns, it is based almost entirely on reconstructions, never a good foundation on which to base a generalization about diachronic language change. Furthermore, these reconstructions are all open to other interpretations, based on typologically well documented developments, and so cannot be regarded as reliable evidence for this type of development.

References Ba˛k, Stanisław (ed.) 1966– Słownik polszczyzny XVI wieku. Wrocław: Zakład Narodowy im. Ossolin´skich. Biberauer, Theresa & Ian Roberts 2010 Comments on Ja¨ger ‘Anything is nothing is something: On the diachrony of polarity types of indefinites’. Natural Language and Linguistic Theory 28: 823–836. Błaszczak, Joanna 2002 What do bagels and Polish kolwiek-pronouns have in common? In: Mary Andronis, Erin Debenport, Anne Pycha & Keiko Yoshimura (eds.), Proceedings from the panels of the Thirtyeighth Meeting of the Chicago Linguistic Society: Volume 38–2, 3–18. Chicago: Chicago Linguistic Society. Błaszczak, Joanna 2003 Explaining the ‘bagel problem’: The case of Polish kolwiek-NPIs. In: Wayles Browne, Ji-Yung Kim, Barbara H. Partee & Robert A. Rothstein (eds.), Formal Approaches to Slavic Linguistics 11: The Amherst meeting 2002, 121–40. Ann Arbor: Michigan Slavic Publications. Catalani, Luigi 2001 Die Negation im Mittelfranzo¨sischen. Frankfurt am Main: Peter Lang. Cies´likowa, Aleksandra 1965 Partykuła -kolwiek w historii i dialektach je˛zyka polskiego. Prace Je˛zykoznawcze 15: 45–84.

320

David Willis

Crain, Stephen & Paul Pietroski 2002 Why language acquisition is a snap. Linguistic Review 19: 163– 183. Detges, Ulrich 2003 La grammaticalisation des constructions de ne´gation dans une perspective onomasiologique, ou: la de´construction d’une illusion d’optique. In: Peter Koch & Andreas Blank (eds.), Kognitiveromanische Onomasiologie und Semasiologie, 213–233. Tu¨bingen: Niemeyer. De´prez, Viviane 1999 The roots of negative concord in French lexifier creoles. In: Michel Degra¤ (ed.), Language creation and language change: Creole, diachrony and development, 375–427. Cambridge: MIT Press. Donati, Caterina 2000 A note on negation in comparison. Quaderni del Dipartimento di Linguistica, Universita` di Firenze 10: 55–68. Duden 2001 Herkunftswo¨rterbuch: Etymologie der deutschen Sprache. 3rd edn. Mannheim: Dudenverlag. Fleuriot, Le´on 1964 Dictionnaire des gloses en vieux breton. Paris: Klincksieck. Foulet, Lucien 1919 E´tudes de syntaxe franc¸aise: I Quelque. Romania 45: 220–249. Hakulinen, Aulil & Lauri Karttunen 1973 Missing persons: On generic sentences in Finnish. In: Claudia Corum, T. Cedric Smith-Stark & Ann Weiser (eds.), CLS9: Papers from the Ninth Regional Meeting, April 13–15, 1973, 157–71. Chicago: Chicago Linguistics Society. Haspelmath, Martin 1997 Indefinite pronouns. Oxford: Oxford University Press. Hawkins, John A. 1979 Implicational universals as predictors of word order change. Language 55: 618–648. Hawkins, John A. 1983 Word order universals. New York: Academic Press. Herburger, Elena 2001 The negative concord puzzle revisited. Natural Language Semantics 9: 289–333. Hindle, Donald & Ivan Sag 1975 Some more on anymore. In: Ralph W. Fasold & Roger W. Shuy (eds.), Analyzing variation in language: Papers from the second colloquium on New Ways of Analyzing Variation, 89– 110. Washington DC: Georgetown University Press.

Negative polarity and the quantifier cycle

321

Hoeksema, Jack 1994 On the grammaticalization of negative polarity items. In: Susanne Gahl, Andy Dolbey & Christopher Johnson (eds.), Proceedings of the Twentieth Annual Meeting of the Berkeley Linguistics Society, 273–282. Berkeley: Berkeley Linguistics Society. Hoeksema, Jack 2007 Dutch enig: From nonveridicality to downward entailment. In: Hedde H. Zeijlstra & Jan-Philipp Soehn (eds.), Proceedings of the Workshop on Negation and Polarity, 8–15. Tu¨bingen: University of Tu¨bingen Collaborative Research Center. Ja¨ger, Agnes 2008 History of German negation. Amsterdam: Benjamins. Ja¨ger, Agnes 2010 Anything is nothing is something: On the diachrony of polarity types of indefinites. Natural Language and Linguistic Theory 28: 787–822. Keinys, Stasys (ed.) 1993 Dabartine˙s lietuviu˛ kalbos zˇodynas. Vilnius: Mokslo ir enciklopediju˛ leidykla. Kiparsky, Paul & Cleo Condoravdi 2006 Tracking Jespersen’s Cycle. In: Mark Janse, Brian D. Joseph & Angela Ralli (eds.), Proceedings of the Second International Conference on Modern Greek Dialects and Linguistic Theory, 172– 197. Mytilene: Doukas. Klima, Edward S. 1964 Negation in English. In: Jerry A. Fodor & Jerrold J. Katz (eds.), The structure of language, 246–323. Englewood Cli¤s: PrenticeHall. Koster, Charlotte & Sjoukje Van der Wal 1995 Acquiring a negative polarity verb. In: Maaike Verrips & Frank Wijnen (eds.), Papers from the Dutch–German Colloquium on Language Acquisition, 109–126. Amsterdam: Institute of General Linguistics. Kuznecov, S. A. (ed.) 1998 Bol’sˇoj tolkovyj slovar’ russkogo jazyka. St Petersburg: Norint. Ladusaw, William A. 1993 Negation, indefinites, and the Jespersen cycle. In: Joshua S. Guenter, Barbara A. Kaiser & Cheryl C. Zoll (eds.), Proceedings of the Nineteenth Annual Meeting of the Berkeley Linguistics Society, 437–46. Berkeley: Berkeley Linguistics Society. Lewis, Henry & Holger Pedersen 1937 A concise comparative Celtic grammar. Go¨ttingen: Vandenhoeck & Ruprecht. Lightfoot, David 1977 Syntactic change and the autonomy thesis. Journal of Linguistics 13: 191–216.

322

David Willis

Martins, Ana Maria 2000 Polarity items in Romance: Underspecification and lexical change. In: Susan Pintzuk, George Tsoulas & Anthony Warner (eds.), Diachronic syntax: Models and mechanisms, 191–219. Oxford: Oxford University Press. Miklosich, Franz 1886 Etymologisches Wo¨rterbuch der slavischen Sprachen. Wien: Braumu¨ller. Muller, Claude 1991 La ne´gation en franc¸ais: Syntaxe, se´mantique et e´le´ments de comparaison avec les autres langues romanes. Geneva: Droz. Musolino, Julien, Stephen Crain & Rosalind Thornton 2000 Navigating negative semantic space. Linguistics 38: 1–32. Parry, Mair In press Italo-Romance. In: Anne Breitbarth, Christopher Lucas & David Willis (eds.), The history of negation in the languages of Europe and the Mediterranean. Volume 1: Case studies. Oxford: Oxford University Press. Pereltsvaig, Asya 2005 Negative polarity items in Russian and the bagel problem. In: Adam Przepio´rkowski & Sue Brown (eds.), Negation in Slavic, 153–178. Bloomington: Slavica. Roberts, Ian 2000 Some remarks on the diachrony of French negation. DELTA 16, 201–219. Roberts, Ian 2007 Diachronic syntax. Oxford: Oxford University Press. Roberts, Ian & Anna Roussou 2003 Syntactic change: A minimalist approach to grammaticalization. Cambridge: Cambridge University Press. Schwegler, Armin 1988 Word-order changes in predicate negation strategies in Romance languages. Diachronica 5: 21–58. Thurneysen, Rudolf 1946 A grammar of Old Irish. Dublin: Dublin Institute for Advanced Studies. Toner, Gregory, Maxim Fomin, Thomas Torma & Grigory Bondarenko 2007 Electronic Dictionary of the Irish Language. Royal Irish Academy & University of Ulster. http://www.dil.ie/ Trubacˇev, Oleg Nikolaevicˇ (ed.) 1974– E`timologicˇeskij slovar’ slavjanskix jazykov. Moscow: Nauka. Urban´czyk, Stanisław (ed.) 1953–2002 Słownik staropolski. Warsaw: Polska Akademia Nauk. Van der Wal, Sjoukje 1996 Negative polarity items and negation: Tandem acquisition. Ph.D. dissertation, University of Groningen.

Negative polarity and the quantifier cycle

323

Van der Wouden, Ton 1997 Negative contexts: Collocation, polarity and multiple negation. London: Routledge. Van Kampen, Jacqueline 2007 A perspective on doubling constructions in child Dutch. Nordlyd 34: 131–152. Van Kampen, Jacqueline 2010 Acquisition preferences for negative concord. In: Andrew D. M. Smith, Marieke Schouwstra, Bart De Boer & Kenny Smith (eds.), The evolution of language: Proceedings of the 8th International Conference (EVOLANG8), Utrecht, Netherlands, 14–17 April 2010, 321–328. Singapore: World Scientific. Vasmer, Max 1953 Russisches etymologisches Wo¨rterbuch. Heidelberg: Winter. Wentworth, Roy 2000 English–Gaelic dictionary. http://www.smo.uhi.ac.uk/gaidhlig/ wentworth/cead. Willis, David 2007 Syntactic lexicalization as a new type of degrammaticalization. Linguistics 45: 271–310. Willis, David 2010 Motivating the emergence of new markers of sentential negation: The case of Welsh ddim. Diachronica 27: 110–56. Willis, David In press Brythonic Celtic. In: Anne Breitbarth, Christopher Lucas & David Willis (eds.), The history of negation in the languages of Europe and the Mediterranean. Volume 1: Case studies. Oxford: Oxford University Press. Zanuttini, Ra¤aella 1997 Negation and clausal structure: A comparative study of Romance languages. New York: Oxford University Press.

Indefinite pronouns, synchrony and diachrony: Comments on Willis Johan van der Auwera & Lauren Van Alsenoy 1. Introduction This paper comments on a paper by David Willis1, which is itself, to some extent, a comment on Ja¨ger (2007a/b 2, 2008). Section 2 is a critical discussion of the typology of changes between ‘normal’ or positive indefinites, negative polarity indefinites and negative indefinites. Section 3 discusses the typology of pronouns that is assumed by the typology of the changes and we will focus on the refinements of the pronominal typology. Section 4 looks in some detail at Dutch enig, the cognate of English any, and expands both the typology of changes and that of pronouns. Section 5 is the conclusion.

2. Types of changes Ja¨ger (2007a/b) deals with polarity changes of indefinite pronouns. She accepts four types of changes. It is these four changes that are the focus of the paper by Willis. Willis essentially accepts the typology but shows that these changes are not equally natural and frequent. Table 1 represents the changes. The table has the format found in Willis (this volume), slightly di¤erent from the version in Ja¨ger (2007a: 81), and for ease of reference we add the labels in the bottom row.

1. This discussion was prepared on the pre-final version of Willis. It di¤ers from the final version, but not, we think, in ways that a¤ect this discussion. 2. It is clear that ‘our Ja¨ger (2007)’, which we call ‘Ja¨ger (2007a)’ is di¤erent from that of Willis’, which we call ‘Ja¨ger (2007b)’. His must be a longer version closer to a version that will appear in Natural Language and Linguistic Theory. If the di¤erences do not matter, we will use the label ‘Ja¨ger (2007a/b)’.

326

Johan van der Auwera & Lauren Van Alsenoy

Table 1. Possible diachronic developments in indefinite systems, 1st version ‘normal’ or positive polarity indefinite

‘positive’

f f

negative polarity indefinite ‘negative’

f f

negative indefinite Type A

Type B

Type C

Type D

Willis calls Types A and B ‘Forwards historical developments’ and Types C and D ‘Backwards historical developments’. In Table 2 we add this characterization and in accord with the horizontal metaphor of ‘forwards’ and ‘backwards’ we change the orientation of the arrows. Table 2. Possible diachronic developments in indefinite systems, 2nd version ‘positive’ ‘normal’ or positive polarity indefinite

‘negative’ negative polarity indefinite

negative indefinite

f

Type A f

Type B f

f

Type C Type D

Forwards

Backwards

Willis supplies examples of types A and B in Celtic languages and links up their existence to evidence coming from L1 acquisition. As to the examples from Celtic, they are very convincing and there is no need for comments. As to the acquisitional evidence, however, it is very indirect and much less convincing. First, the evidence does not come from the acquisition of indefinites, but rather from the acquisition of the Dutch negative polarity verb hoeven ‘need’. Second, while the evidence shows that children that start out with niet hoeven ‘not need’ can drop the negation and thus invest need with inherent negative meaning, these same children then drop the close connection with negation, but keep the negative polarity, not surprisingly thereby moving into the direction of the adult language. If the acquisitional fact that children can narrow down a negative

Indefinite pronouns, synchrony and diachrony

327

polarity meaning into a negative meaning supports a change of Type B, then the fact that the children afterwards reinstate the negative polarity meaning must be taken to in support of a change of Type C. This is a little problematic, for Willis will claim that a Type C change is dubious or exceptional. Third, if acquisitional facts about a ‘need’ verb are indeed important, it would be good to have some in support of the loss of negative polarity (Type D) too, for that is what happened in the history of the English lexical verb need. At present this verb is not polarity sensitive, di¤erent from the auxiliary need. (1) a. b. c.

He needs to go now. Does he need to go now? He doesn’t need to go now.

(2) a.

*He need go now.

b. c.

Need he go now? He needn’t go now.

However, lexical need used to be a negative polarity verb too, and it lost that very feature over a 500 year period (van der Auwera & Taeymans 2009), the length of which furthermore makes some linguists, including ourselves, wonder about the relevance of L1 acquisition facts. As to changes of type C, Willis mentions observations concerning French and Celtic. The French observation concerns Middle French nul and nesun. Willis does not discuss French, only refers to Catalani (2001: 113–114), but we take it that the French evidence is uncontroversial (see also Buridant 2000: 135–137 and Ingham, this volume). The Celtic observations are discussed in detail (this volume). One set concerns pronouns with a clear negative component and the other set derives from minimizers. As to the former, including e.g. Middle Welsh neb ‘someone’, Ja¨ger (2007b) takes these to have been negative pronouns. Willis, however, mentions that there is an alternative free choice pronoun hypothesis. On the basis of Haspelmath (1997: 230–234), who Willis also refers to, and with approval, we find the balance to be in favour of the free choice origin. As to the other type, the one deriving from a minimizer, it concerns Scottish Gaelic with e.g. sian ‘anything’, deriving from a word referring to various kinds of sounds. At present these markers occur in negative polarity contexts. Willis does not o¤er any historical data about the earlier uses, but he does claim that the minimizing uses of such words must have entered into

328

Johan van der Auwera & Lauren Van Alsenoy

the pronoun system in a negative context and then expanded to negative polarity contexts. Schematically, with English instead of Scottish Gaelic: (3) I didn’t hear a squeak. ! ‘I didn’t hear anything.’ We fail to see why this change could not have taken place in negative polarity contexts as well and at the same time. (4) Did you hear a squeak? ! ‘Did you hear anything?’ In the absence of any discussion of historical data, we claim that markers such as Scottish Gaelic sian do not constitute evidence for a Type C change. And to conclude our metadiscussion about Type C changes, Willis was sceptical about some proposals, and we are even more skeptical. Nevertheless, the French nul evidence seems clear and so Type C changes may be assumed to exist. Willis discusses changes of Type D, again from Celtic, and he accepts the German examples that Ja¨ger o¤ers, i.e., German jemand ‘someone’ and German immer ‘always’. We agree and we could add more examples: for German, the cognate to English any, viz. einig ‘some’, and for Dutch, the parallels for the three cases listed for German, and one could add ooit in the meaning ‘at some point’ (Hoeksema 1999) and also ergens ‘somewhere’ (the cognate to German irgend ). Interestingly, Willis also discusses changes that go directly from the negative to the ‘normal’ or positive pronoun. Willis does not enter this change in his Table, but it is of course easy to do. Let’s call this two-step change Type E. Table 3. Possible diachronic developments in indefinite systems, an addition ‘positive’ ‘normal’ or positive polarity indefinite

‘negative’ negative polarity indefinite

negative indefinite f

Type E

Backwards

Willis discusses Slavonic pronouns and one Celtic pronoun. The Slavonic pronouns are the nekto/nesˇto pronouns ‘someone, something’. Like the Celtic pronouns discussed for Change C, the Slavonic pronouns have an unmistakable negative component, but not a clear etymology. Once again, Ja¨ger (2007a/b) assumes that they were negative pronouns, Willis

Indefinite pronouns, synchrony and diachrony

329

approvingly discusses an alternative proposal that derives them from ‘I don’t know who’ constructions, and refers to Haspelmath (1997: 130– 132), and we find the latter more persuasive.3 The one Celtic example concerns Middle Cornish ne¤ra ‘always’. This marker is not polarity sensitive and it was borrowed from the English negative adverb never. So here we do have a direct change from negative to ‘normal or positive’. In essence, however, we agree with Willis that the development we see in Middle Cornish ne¤ra is not relevant in the discussion of what is implicitly taken as a classification of language-internal change. The Middle Cornish ne¤ra is a language contact phenomenon. We conclude that there are no good examples of the Type E change. Interestingly, there is no discussion of the mirror change, a two step change from ‘normal’ or positive to negative. We could call it a Type F change. Table 4. Possible diachronic developments in indefinite systems, a 2nd addition ‘positive’ ‘normal’ or positive polarity indefinite f

‘negative’ negative polarity indefinite

negative indefinite Type F

Forwards

Perhaps this double step is unattested and even impossible, and not implausibly, any two double step change may be impossible. Let us end this section with two more points. First, as already mentioned, Willis approvingly discusses the scenario of a free choice pronoun turning into an ordinary indefinite. Haspelmath (1997: 149–156) discussed processes of free choice uses turning into negative polarity indefinites. It is not clear whether or how these changes could be mapped in the Ja¨gerWillis tables. Second, the Ja¨ger-Willis representations allow for a scenario in which one and the same marker undergoes all of the changes, e.g. A, then B, then C and then D, or in which one and the same marker turns 3. For Willis ‘I don’t know who’ is a free choice use. This is di¤erent from Haspelmath (1997: 130–132), who takes ‘I don’t know who’ to turn directly into a ‘normal’ or positive pronoun. We see no reason to depart from Haspelmath.

330

Johan van der Auwera & Lauren Van Alsenoy

back at the first occasion, e.g. first A and then D, thus allowing a ‘cycle’ in a richer sense than the sense used by Willis.4 Table 5. Possible diachronic developments in indefinite systems, 3rd version ‘positive’ ‘normal’ or positive polarity indefinite

‘negative’ negative polarity indefinite

negative indefinite Forwards Backwards

From an acquisitional point of view, the cycle shown in Table 5 should be perfectly possible, for the simple reason that children are not historical linguists, and Willis does indeed allow for these kinds of changes as shown in his discussion of the Ja¨ger 2007b scenario in which Middle Welsh nep would have gone from negative to negative polarity and then back to negative. But this particular change is one that he has convincingly cast doubt on. So should changes that make a marker return to an earlier stage be impossible or rare,5 then this needs to be accounted for.

3. Types of pronouns The Tables discussed in the previous section give the impression that there are only 3 types of pronouns, i.e., ‘normal’ or positive polarity indefinites, negative polarity indefinites and negative indefinites. This is not quite correct, as the label ‘normal’ or positive polarity indefinites’ could already be taken to suggest. For Ja¨ger (2007a: 79–80, 2008: 151–164), this label would be appropriate both for English somebody and Swedish na˚gon ‘somebody, anybody’. The latter would be a ‘normal’ indefinite and the former a positive polarity indefinite.

4. For Willis, ‘cycle’ just refers to the changes shown in Table 1, and even if there were only forward and thus unidirectional change, he would still call the scenario a ‘cycle’. 5. Nevertheless, Middle French nul could be a case in point. It started as a negative pronoun, then widened to negative polarity, and there is some evidence that it then shrank again, at least for some speakers (Catalani 2001: 113).

Indefinite pronouns, synchrony and diachrony

331

(5) Jag har inte set na˚gon. I have not seen somebody ‘I haven’t seen anybody.’ However, they share the same slot in the typology and they are considered to be contextually unspecific; the di¤erence is taken care of by a graded application of a principle (the ‘Elsewhere condition’, Kiparsky 1973) that says a language may preclude a contextually unspecific pronoun in case it allows a more specific pronoun. English has not anybody and this is taken to block not somebody. Swedish does not have anything comparable to not anybody and so inte . . . na˚gon cannot be blocked. The ‘Elsewhere Conditionalist’ approach is problematic though, and for three reasons. First, at least for Willis, the space of operation for the pronoun types goes from ‘positive’ to ‘negative’. This is what his leftmost column says. Characterizing somebody as positive does indeed make sense, but characterizing Swedish na˚gon as positive makes much less sense, as it is equally felicitous in positive, negative polarity and negative contexts. Interestingly, Ja¨ger (2007a: 81) does not have this column, so our remark might not apply to Ja¨ger (2007a). Second, a point also made by Willis, though in di¤erent terms, both English and Swedish have an alternative, viz. the use of the inherently negative nobody or ingen. (6) Jag sa#g ingen. I saw nobody. ‘I saw nobody.’ The nobody and ingen strategies are more specific than the anybody and na˚gon strategies, yet in this case the former do not block the latter. In other words, we need yet another type of condition, this time, one that stipulates when the Elsewhere Condition applies and when it does not. Whereas a simple application of the Elsewhere Condition promises an economical and elegant account, the need for the ‘Metacondition’ makes for a more complicated account. Third, Ja¨ger and Willis agree that pronouns may act di¤erently in di¤erent negative polarity contexts (see Willis on two types of negative polarity contexts for Dutch enig or his concluding remarks of the discussion on Middle Welsh, (see also Ja¨ger 2008: 155 on the so-called ‘Neg-hierarchy’). So while a ‘conditionalist’ approach can take care of a simple distinction between ‘normal’ na˚gon and positive polarity somebody, which gives four pronoun types altogether, i.e., nobody/ ingen, anybody, somebody and na˚gon, it is an open question whether it can o¤er an account in case the full number of pronoun types distinguishable in

332

Johan van der Auwera & Lauren Van Alsenoy

terms of contexts, first and foremost the negative polarity contexts, increases significantly. This number indeed increases a lot. Haspelmath (1997: 76) gives a good idea of the magnitude of the problem. He characterizes pronoun types in terms of the nine contexts illustrated in (7) to (15). The contexts in (7) and (8) concern positive polarity, those in (9), (10), (13) and (14) are negative polarity contexts, (11) and (12) negative contexts, and (15) involves a free choice context. (7) (8) (9) (10) (11) (12) (13) (14)

Somebody called you while you were away: guess who? I heard somebody, but I couldn’t tell you who. Please ask somebody else. Has anybody told you about it? John doesn’t think that anybody will be there. I have seen nobody. If you see anybody, tell me immediately He is better than anybody else in the group.

(15) Anybody can solve this problem.

[specific known] [specific unknown] [irrealis] [question] [indirect negation] [direct negation] [condition] [comparative] [free choice]

On the basis of reasonably detailed analyses of 40 languages, about three fourths of which are European and about half Indo-European, Haspelmath (1997: 76) concludes that there are 37 pronoun types. To illustrate this point, (16) is his list of the 13 pronoun types that include the direct negation function. We also list an example for each type, each of them a translational equivalent for nobody or the anybody of not anybody; for analyses we have to refer to Haspelmath (1997). (16)

direct negation

German niemand

direct negation, indirect negation

Icelandic neinn

direct negation, indirect negation, question

Italian nessuno

direct negation, indirect negation, comparative

Maltese add

direct negation, indirect negation, question, conditional

Hebrew isˇ

direct negation, indirect negation, question, comparative

Finnish kukaan

Indefinite pronouns, synchrony and diachrony

direct negation, indirect negation, question, conditional, irrealis

333

Greek kane´nas

direct negation, indirect negation, comparative, Kannada yaaruu free choice direct negation, indirect negation, question, conditional, comparative

Catalan ningu´

direct negation, indirect negation, question, conditional, comparative, free choice

Engels anybody

direct negation, indirect negation, question, conditional, comparative, free choice, irrealis

Nanay -daa

direct negation, indirect negation, question, conditional, irrealis, specific unknown, specific known

Swedish na˚gon

direct negation, indirect negation, question, conditional, irrealis, comparative, free choice

Hindi/Urdu koi bhii

37 pronoun types may seem a lot, but in fact the number will be higher still. First, Haspelmath’s characterization of the contexts is not refined enough. Three types of refinements will be necessary. First, there are contextual distinctions widely discussed in the literature that Haspelmath chooses not to include. This is most clearly the case for the direct negation use: contrary to most other researchers, Haspelmath does not distinguish – in the typology sketched in (16) – between the inherently negative nobody, i.e., a pronoun that has no other function than that of negation, either direct or indirect, and the negatively polar anybody in the combination not anybody. The context of nobody and anybody is, of course, di¤erent: for nobody the context is positive, and for anybody it is negative. For the typology in (16) Haspelmath also does not distinguish inherently negative pronouns on the parameter of the presence of another negator (negative concord). Compare the Romanian negative pronoun nimeni ‘nobody’, which needs another negator (in a‰rmative full sentences) and the Portuguese pronoun ningue´m ‘nobody’, which needs the additional negator when the pronoun is postverbal (again in a‰rmative full sentences). (17) a.

Nimeni nu a venit. nobody not has come ‘Nobody has come.’

334

Johan van der Auwera & Lauren Van Alsenoy

b.

(18) a.

b.

Nu a venit nimeni. not has come nobody ‘Nobody has come.’ Ningue´m veio. nobody came ‘Nobody came.’ Na˜o veio ningue´m. not came nobody ‘Nobody came.’

Another context that has been central in the discussion, but that Haspelmath’s typology does not give a discriminatory role to, is that of the elliptic answer.6 It is in this context that indefinite pronouns that normally need a clausal negator to convey the negative sense, may drop the negator.7 The explanation, we propose, is Zipfo-Gricean: ellipsis means economy (Zipf ’s (1949) Least E¤ort or Grice’s (1975) 3rd Maxim of Manner ‘Be brief ’). Thus Rumanian nimeni is more economical than nimeni nu, and there is no risk of confusion. (19)

– Cine a venit? who has come ‘Who has come?’ – Nimeni. nobody ‘Nobody.’

A second reason for the need for contextual refinement and hence also for an extended typology of pronouns is the existence of contextual distinctions that have not been central in the discussion and that Haspelmath did not discuss either. The point can be illustrated by Dutch enig, the etymological counterpart to English any. As we will document in the next section, enig can corresponds to both some and any. In the any sense, it can e.g. occur in a direct negation context with niet, when enig is part of 6. This does not mean that Haspelmath (1997) does not pay attention to ellipsis; Haspelmath (1997: 194–198) is an insightful discussion. 7. Note that we do not claim that the negator has to be dropped: in Canadian French (Pierre Larrive´e, personal communication) and Brabantic Dutch (Barbiers et al. 2008: 66–68) the negator can stay.

Indefinite pronouns, synchrony and diachrony

335

a prepositional phrase, as in (20a), but not when it is a direct object, as in (20b); yet with nooit ‘never’ enig ‘any’ is always acceptable.8 (20) a.

b.

Hij is nooit/niet in staat tot enige arbeid. he is never/not in state to any work ‘He is never/not capable of any work.’ [. . .] aangezien ik zelf nooit/*niet enige ambitie heb gehad. for I self never/not any ambition have had ‘[. . .] for I myself have never/not had any ambition.’

This example shows that direct negation is too broad a notion, and that the contextual details may be quite refined. The third reason why there are more than 37 pronoun types is that pronoun types for which Haspelmath (1979) did not find any attestations or which he predicted not to exist9 are now attested. Dutch niemand ‘nobody’ illustrates the subtype of the pronoun that was predicted not to exist. Haspelmath predicted the non-existence of a pronoun that combines direct negation and question and leaves out indirect negation. Dutch niemand, however, is a counterexample. (21) illustrates the relevant uses. The indirect negation and questions do, of course, allow niemand in the negative ‘nobody’ sense, but the point is that niemand allows a negatively polar use in questions, but not in indirect negation. (21) a.

b.

Ik heb niemand gezien. I have nobody seen ‘I have seen nobody.’ *Zonder niemand gezien te hebben. . . . without anybody seen to have ‘Without having seen anyone . . .’

c. Heb je niemand gezien? have you nobody seen ‘Is it true that you saw nobody?’ or ‘Didn’t you see anybody?’ 8. The examples are taken from a corpus of contemporary spoken Dutch from both Belgium and the Netherlands, the 10 million word Corpus Gesproken Nederlands (‘CGN’, http://tst.inl.nl/cgndocs/doc_/English/start.htm). 9. The predictive power is due to the fact that Haspelmath (1997) puts all the functions on a semantic map and requires functions fulfilled by pronoun types to be adjacent. On semantic maps in general, see Haspelmath (2003) and van der Auwera (2008).

336

Johan van der Auwera & Lauren Van Alsenoy

If there are, say, 50 odd types of contexts and hence 50 odd types of pronouns, what are the consequences for an explanatory account? Can an elsewhere conditionalist approach still deal with this explosion of pronoun types? We are pessimistic, as Willis would seem to be too (see his discussion of the ‘classical n-word problem’). For a semantic map approach, however, in the style of Haspelmath (1997), though not necessarily quite the same (see van der Auwera & Van Alsenoy 2011), we are optimistic. The raison d’eˆtre for the Haspelmath approach was in fact the massive crosslinguistic polyfunctionality of indefinite pronouns. If the approach can deal with 37 types and in an elegant way, there is no principled reason why it could not be modified and extended to deal with 50. Note that we do not deny that one can generalize over pronoun types and arrive at a much smaller set of more abstract types, to wit, the three or four of the Ja¨ger-Willis account (positive, negative polarity, and negative) or, we think, a few more, for it seems appropriate to include the free choice use and its negation10 (see van der Auwera & Van Alsenoy 2011). And to the extent that one can discuss the more abstract types, one can also discuss changes from one to the other, as we have done in the wake of Ja¨ger (2007a/b) and Willis. However, to the extent the better typology of pronouns is the more complicated one, the better typology of changes will also be more complicated.11 For instance, in the abstract typology we endorsed the claim that a negative pronoun can become a negative polarity pronoun – the Type C change – and we referred to Middle French nul. In the more refined typology, we should e.g. ask whether a negative pronoun could lose its negativity in truly all contexts, so even in the elliptical answer context illustrated in (19). For nul, at least, the answer is negative, as a check-up in the Base de franc¸ais me´die´val (http://bfm.ens-lsh.fr) will show: in the elliptic answers Middle French nul remained negative (and in other contexts too, Buridant 2000: 136). 4. More types of changes and . . . more types of pronouns Willis mentions two examples of markers ‘leaving the system’, viz. the change from the German indefinite adverb ete(s)wa to an adverb meaning ‘approximately’, and the even more drastic change from anyhow and 10. The latter is sometimes called the ‘indiscriminatory’ use (after Horn 2000). I don’t go to bed with just anybody; I have to be attracted to them sexually is an example. 11. For a semantic map approach this is no problem, if, as van der Auwera (2008) claims, it is correct that the best synchronic semantic map is a diachronic one.

Indefinite pronouns, synchrony and diachrony

337

anyway to discourse markers. ‘Leaving the system’ must mean that the Tables used so far must be enriched, even when they are already refined in terms of 50 pronoun types instead of a handful. We also know that markers ‘enter the system’. French personne meant and still means ‘person’, the ancestor of Scottish Gaelic sian ‘anything’ was a sound word, and the literature contains many more examples. What these cases of markers ‘leaving’ and ‘entering’ the system show is that quantifier changes are part of larger diachronic scenarios. What we will do in this section is illustrate that quantifier changes do not only connect to other changes at the beginning and at the end of the quantifier changes, but also ‘in the middle’. We will do this with some observations and hypotheses about Dutch enig ‘some, any’ and we will enrich both the typology of changes and that of pronouns. Willis discusses enig on the basis of Hoeksema (2007). Willis’ point is that the history of enig is an illustration of the general fact that the uses of a negative polarity indefinite may change, more particularly, that over the years negatively polar enig has become restricted to fewer contexts. Hoeksema’s account is actually a little di¤erent. Willis and Hoeksema agree that enig narrowed its use from nonveridical to downward entailing contexts, but whereas Willis at least implicitly takes nonveridicality as negative polarity, Hoeksema (2007: 14) does not do that. Important for what follows is that they agree that enig disappeared from positive contexts and that it is currently a negative polarity item. Let this be our starting point. Two thirds of the occurrences reported by Hoeksema (2007: 13) and repeated by Willis are negative and one third a¤ective.12 A first interesting question is whether the ratio is constant. The answer is that it probably is not. If one studies the negatively polar enig together with another enig, which is not polarity-sensitive but which can therefore occur in negative and a¤ective contexts too (see below), one can see an evolution. In the Corpus of Spoken Dutch (‘CGN’, see note 8) there are 188 negatively polar occurrences of enig, of which 107 or 56.9% are negative and 81 or 43.1% a¤ective. In an ad hoc one million corpus seventeenth century plays culled from the Digitale Bibliotheek voor de Nederlandse Letteren (‘DBNL’, http://www.dbnl.org/), we found 177 negatively polar occurrences of enig and this time the ratio is the opposite: 107 or 60.5% are

12. With ‘a¤ective’ we refer to negatively polar uses other than negative ones.

338

Johan van der Auwera & Lauren Van Alsenoy

a¤ective and 70 or 39.5% are negative (see Table 6). If one counts occurrences of the same noun only once, the contrast is bigger still, as is shown in Table 7. Note that the new count really only a¤ects the proportions for present-day Dutch: of the 81 a¤ective occurrences, 50 are questions with the noun idee ‘idea’. Interrogative enig idee ‘any idea’ has thus become a fixed phrase, and it is the only one that the CGN and DBNL materials manifest. We can thus conclude that present-day negatively polar enig may well be on its way to become a negative indefinite. In terms of the typology of pronoun changes, we are dealing with a Type B change. Table 6. Negative vs. a¤ective contexts in the CGN and in 17th century DBN plays enig

DBNL 17th century

CGN 20th c

negative context

70 (39.5%)

107 (56.9%)

a¤ective context

107 (60.5%)

81 (43.1%)

177

188

Total

Table 7. Negative vs. a¤ective contexts in the CGN and in 17th century DBN plays enig with a specific N

DBNL 17th century

CGN 20th c

negative context

61 (40.4%)

92 (77.3%)

a¤ective context

90 (59.6%)

27 (22.7%)

151

119

Total

A second point to make is that the enig just discussed is not the only enig or, put di¤erently, if one prefers to believe that these two enig’s are the same, enig has more uses than the negatively polar uses. The negatively polar enig is taken by Hoeksema (2007), and earlier also Hoeksema & Klein (1995), to be restricted to singular count nouns, but when enig combines with mass nouns or plural count nouns, there is no polarity. Thus enig and boek ‘book’ is singular count and negatively polar, hence it cannot occur in a positive a‰rmative context. But when enig combines with tevredenheid ‘satisfaction’ or boeken ‘books’ there is no polarity sensitive, and it can occur in all contexts.

Indefinite pronouns, synchrony and diachrony

(22) a.

Niemand had enig boek meegebracht. nobody had any book brought.along ‘Nobody had brought along any book.’

b.

Had Jan enig boek meegebracht? ‘Had Jan brought along any book?’

c.

*Jan had enig boek meegebracht. *Jan had brought along any book.

(23) a.

339

Niemand merkte enige tevredenheid. nobody noticed any satisfaction ‘Nobody noticed any satisfaction.’

b.

Merkte Jan enige tevredenheid? ‘Did Jan notice any satisfaction?’

c.

Jan merkte enige tevredenheid. ‘Jan noticed some satisfaction.’

(24) a.

Niemand had enige vragen. nobody had any questions ‘Nobody had any questions.’

b.

Had Jan enige vragen? ‘Did Jan have any questions?’

c.

Jan had enige vragen. ‘Jan had some questions.’

However, the association of polarity sensitivity with the count vs. mass and singular vs. plural distinctions is not absolute. Hoeksema & Klein (1995: 167) already noted the existence of singular count positive a‰rmative occurrences of enig, they o¤ered the example in (25), and they called them ‘exceptional’. (25) We zullen op enig moment tot een beslissing moeten komen. we shall on some moment to a decision must come ‘We will have to come to a decision at some point.’ The CGN materials contain 185 positive a‰rmative occurrences of enig. 121 (65.4%) involve mass nouns, 35 (18.9%) plural count nouns, 11 (5.9%) singular count noun, and interestingly, another 18 cases (9.7%) involve nouns that can be taken as either count or mass.

340

Johan van der Auwera & Lauren Van Alsenoy

(26) a.

b.

Hij maakt kans / een kans. he makes chance / a chance ‘He has a chance.’ Hij maakt enige kans. he makes some chance ‘He has some chance.’

Given these figures and given the existence of the category that can be taken to involve either a count singular or a mass noun, calling a positive singular count any exceptional is perhaps too strong, though it is definitely infrequent. Note that plural count nouns are not really that frequent either and the typical case seems to involve mass nouns. In any case, given the absence of a clear distinction we think that it is better to posit one enig and to accept that there is strong tendency for negative polarity when enig co-occurs with a singular count noun.13 In present-day Dutch, enig is mostly used in positive contexts. Table 8 completes the CGN parts of Tables 6 and 7. Table 8. Enig in negative, a¤ective and positive contexts in the CGN CGN 20th c

enig

enig with a specific N

negative context

107 (24.2%)

92 (30.3%)

a¤ective context

81 (18.3%)

27 (8.9%)

positive context

254 (58.5%)

185 (60.8%)

442

304

Total

In the oldest stage of Dutch enig, just like its counterparts in the other old West Germanic languages (Fobbe 2004: 138–147; Hoekstra & Siebinga 13. Another reason for positing just one enig is that in negative contexts, which are hospitable to any kind of noun, the percentage of mass nouns is rather high, probably a lot higher than the average occurrence of mass nouns. The CGN contains 107 negative enig attestations, of which 35 (32.8%) involve mass nouns, 2 (1.9%) plural count nouns, 44 (41%) singular count nouns and the count or mass category has 26 cases (24.3%). The association of enig with mass nouns thus holds across all contexts, though with di¤erent strength. (There is no point in doing this count for a¤ective contexts, for here the occurrences are very infrequent.)

Indefinite pronouns, synchrony and diachrony

341

2008; Mitchell 1985: 151–152; Ja¨ger 2008: 192–198), did not occur in positive contexts at all (see the lemma in the Oudnederlands woordenboek, http://www.inl.nl/nl/woordenboeken/oudnederlands-woordenboek-onw). Comparing the oldest and the current stage of the development,14 it is clear that enig is in the midst of a type D change. Di¤erent from the Type D change that one has to posit for e.g. iemand ‘someone’ or, more relevantly even, for German einig, the change has not been completed: we have just seen that negatively polarity uses still exist, and has increased its negativity and we called it a type B change. It thus seems that a Type B and a Type D change can go hand in hand: negative polarity is being lost in favour of both positivity and negativity, though more in favour of the former than the latter. Let us call this an ‘Outwards’ change15 and if this development comes to completion, we arrive at a new type of pronoun, to be added to the list of 50 plus pronouns discussed in section 2: a pronoun that is ‘normal’ as well as negative, but not negatively polar. Table 9. Possible diachronic developments in indefinite systems, 3rd version ‘positive’ ‘normal’ or positive polarity indefinite

‘negative’ negative polarity indefinite

negative indefinite

f

Type A f

Type B f

Type C

f

Type D

f

Type B-D

Forwards

Backwards Outwards

The story of enig is not finished yet. The older Germanic languages all witnessed a confusion between the ancestor of enig and another word. In Old English, for instance, the ancestor of any is ænig and the other word 14. Place forbids a detailed diachronic analysis, but one can indeed witness the rise of the positive uses in DBNL samples. 15. If we accept an Outwards B-D change, we beg the question whether there could be an ‘Inwards’ A-C. We have no evidence, however.

342

Johan van der Auwera & Lauren Van Alsenoy

is anga, an adjective meaning ‘only, single, unique’. These words are formally similar, for both are composed of the numeral ‘one’ and a similar su‰x, and also semantically similar: as a second sense of anga, Bosworth & Toller (1898: 42) lists ‘any’ and their example is (27). (27) Secge me nu hwæðer ðu æfre gehyrdest, ðæt wisdom angum ðara eallunga þurhwunode ‘Tell me now whether you have ever heard that wisdom always remained in any?’ The form angum is unmistakably adjectival, the translation has ‘any’ and if one glosses angum with ‘even a single’, one can understand how one ends up with ‘any’. In English the ‘only, single, unique’ and ‘any’ senses managed to retain separate words: anga disappeared and was replaced by only, another formation with the numeral one. In Dutch, however – and in German, too, for that matter – the two words fell together and we end up with an ambiguity. Initially, the adjectival uses was much less frequent than the pronominal one, with e.g. 555 occurrences of enig in the seventeenth century DBNL play corpus, of which 471 (84.9%) are pronominal and 84 (15.1%) are adjectival, but nowadays, the ratio is the opposite, with 1646 enig occurrences in the CGN corpus, of which 1204 (73.2%) are adjectival and 442 (26.8%) pronominal. The adjectival sense is and was, unsurprisingly, polarity neutral. We know that pronominal enig moved from negatively polarity to neutrality, its type D change. We advance the hypothesis that the rise of adjectival enig helped push pronominal enig towards polarity neutrality.16 The point about the confusion between the adjective and the pronoun documents a category squish. And it is our hypothesis that the story of enig documents yet a second category squish. Why is it that the polarity neutral ‘some’ meaning enig uses prefer either mass or plural count nouns? We advance that this is due to a hole in the Dutch indefinite article system. As is typically the case in Standard Average European, Dutch only has a singular count indefinite article.

16. That the two senses interfered is also evidenced by the fact that a new adjectival form arose, a superlative enigst ‘only-est’, documented in the DBNL materials since the eighteenth century and currently typical for Belgian Dutch. In German there was also a new form, i.e. einzig, which started to replace einig ‘only’ in the eighteenth century (Behaghel 1923–1932: vol. 1, 421).

Indefinite pronouns, synchrony and diachrony

(28) a. b.

343

Ik merkte Ø tevredenheid I noticed satisfaction een boek a book

c. Ø boeken. books ‘I noticed satisfaction / a book / books.’ What enig does is to fill the Ø holes – see (23) and (24). enig is an indefinite article in statu nascendi. The grammaticalization is incomplete, for the use of enig is not necessary and there are alternatives, primarily Ø, but the plural also has constructions such as ettelijke, enkele and een paar ‘a couple’.17 Dutch enig thus manifests an adjective-pronoun-article squish. 5. Conclusions In this paper we first turned to the Willis-Ja¨ger discussion of possible indefinite pronoun changes. We endorsed the view that each of four changes discussed exist but, in agreement with Willis, not quite to the extent that Ja¨ger (2007a/b) proposed and, at least in the spirit of Willis, not even to the extent that Willis was willing to accept. We furthermore distinguish between one step and two step changes and advanced the claim that the latter are impossible. We then turned to the typology of pronouns underlying the hypotheses about pronoun changes and argued for a refined typology, in the spirit of Haspelmath (1997), which has room for more than 50 types of indefinite pronouns. We finally discussed the complicated history of Dutch enig and argued that both the typology of pronoun changes and that of pronoun types need to be refined even more. Acknowledgements Thanks are due to Pierre Larrive´e and Richard Ingham and the Leverhulme Trust for the organization of the Cycle Seminars, to the participants to the 17. The fact that there are more alternatives for plural count nouns than for mass nouns helps explain that present day polarity neutral enig is predominantly used for mass nouns (65.4% vs. 18.9%, see earlier).

344

Johan van der Auwera & Lauren Van Alsenoy

Cycle Seminars, and also to the Flemish and Belgian Research Funds (FWO ‘Jespersen Cycles Grant’; Interuniversity Attraction Poles ‘Grammaticalization and (Inter)subjectification Grant’ P66/4). We are also grateful to Ann Vanderheyden and Mihaila Ilioaia for their help with Old and Middle French and Romanian, respectively.

References Barbiers, Sjef, Johan van der Auwera, Hans Bennis, Eefje Boef, Gunther De Vogelaer & Margreet van der Ham 2008 Syntactic atlas of the Dutch dialects. Part II. Amsterdam: University Press. Behaghel, Otto 1923–1932 Deutsche Syntax. 4 volumes. Heidelberg: Winter. Bosworth, Joseph & Toller, T. Northcote 1898 An Anglo-Saxon Dictionary. Oxford: Clarendon Press. Buridant, Claude 2000 Nouvelle grammaire de l’ancien franc¸ais. Paris: SEDES. Catalani, Luigi 2001 Die Negation im Mittelfranzo¨sischen. Frankfurt am Main: Peter Lang. Fobbe, Eilika 2004 Die Indefinitpronomina des Deutschen. Aspekte ihrer Verwendung und ihrer historischen Entwicklung. Heidelberg: Winter. Grice, H. Paul 1995 Logic and conversation. In: Peter Cole & Jerry L. Morgan (eds.), Syntax and Semantics 3: Speech acts, 41–58. New York: Academic Press. Haspelmath, Martin 1997 Indefinite pronouns. Oxford: Oxford University Press. Haspelmath, Martin 2003 The geometry of grammatical meaning: semantic maps and crosslinguistic comparison. In: Michael Tomasello (ed.), The New Psychology of Language. Cognitive and Functional Approaches to Language Structure, 211–242. Mahwah JN: Lawrence Erlbaum. Hoeksema, Jack 1999 Aantekeningen bij OOIT, deel 2: de opkomst van niet-polair OOIT. Neder-L 20: 147–172. Hoeksema, Jack 2007 Dutch enig: From nonveridicality to downward entailment. In: Hedde Zeijstra & Jan-Philipp Soehn (eds.), 8–15.

Indefinite pronouns, synchrony and diachrony

345

Hoeksema, Jack & Henny Klein 1995 Negative predicates and their arguments. Linguistic Analysis 25: 146–180. Hoekstra, Eric & Sjoerd Siebinga 2008 Exploring the distribution of aeng, the Old Frisian counterpart of Modern English any, Modern Dutch enig. In: Rolf H. Jr. Bremmer, Stephen Laker & Oebele Vries (eds.), Advances in Old Frisian philology, 153–162. Amsterdam: Rodopi. Horn, Laurence R. 2000 Pick a Theory (Not Just Any Theory): Indiscriminatives and the Free Choice indefinite. In: Laurence R. Horn & Yusuhiko Kato (eds.), Negation and polarity, 147–192. Oxford: Oxford University Press. Ja¨ger, Agnes 2007a On the diachrony of polarity types of indefinitess. In: Hedde Zeijstra & Jan-Philipp Soehn (eds.), 78–84. Ja¨ger, Agnes 2007b On the diachrony of polarity types of indefinites. Manuscript. Ja¨ger, Agnes 2008 History of German negation. Amsterdam: Benjamins. Kiparsky, Paul 1973 Elsewhere in phonology. In: Stephen R. Anderson & Paul Kiparsky (eds.), A festschrift for Morris Halle, 93–106. New York: Holt, Rinehart & Winston. Mitchell, Bruce 1985 Old English syntax. Oxford: Clarendon Press. van der Auwera, Johan 2008 In defense of classical semantic maps. Theoretical Linguistics 34: 39–46. van der Auwera, Johan & Martine Taeymans 2009 The need modals and their polarity. In: Rhonwen Bowen, Mats Moba¨rg & Sven Ohlander (eds.), Corpora and discourse – and stu¤. Papers in honour of Karin Aijmer, 317–326. Gothenburg: University of Gothenburg. van der Auwera, Johan & Lauren Van Alsenoy 2011 Mapping indefiniteness: towards a Neo-Aristotelian approach. In: Eliza Kitis, Nikolaos Lavidas, Nina Topintzi & Tasos Tsangalidis (eds.), 19th International Symposium on Theoretical and Applied Linguistics, 1–14. Thessaloniki: Monochromia. Zeijstra, Hedde & Jan-Philipp Soehn (eds.) 2007 Proceedings of the Workshop on Negation and Polarity. http://www.sfb441.uni-tuebingen.de/negpol/negpol07.pdf Zipf, George Kingsley 1949 Human behavior and the principle of least e¤ort. Cambridge: Addison-Wesley.

Subject index actualization 281 adverb 5, 8, 11–13, 62, 81–83, 86–90, 98, 100, 116, 121–122, 125–127, 141–142, 150, 158, 168–171, 175, 181, 186, 234, 292, 306, 329, 336 Agree 63, 133, 166, 302, 328–329, 331, 337 almost test 13, 69, 170–175 anymore (positive) 289 argument/quantifier cycle 285–286, 296, 306–308 article 3, 17, 27, 89, 148, 160, 180, 182, 203–204, 209, 212, 218, 299, 311, 342, 343 binding 30–33, 37–38, 41, 48, 62, 235, 240, 274 bleaching 50, 89, 267, 279, 314 clause negation 77–79, 81–83, 100, 109, 121 conditional 7, 9–10, 31, 141, 150, 276, 279, 285–286, 292–293, 296–298, 300–302, 304–307, 309, 331–333, 336 comparative 7, 9–11, 172–173, 210, 223, 228, 257, 260, 277–278, 285– 286, 292–293, 296, 299–304, 306– 307, 309, 332–333 constituent negation 6, 12, 14–15, 17, 102, 117–118, 123–125 coordination 6, 8, 91–92, 116 count noun 238, 251, 260, 338–340, 342–343 denial 40, 45, 260–261 discourse-partitioning 87–88 double negation 15–16, 38–40, 46–47, 63, 69, 236, 273–274 elliptical answer 334, 336 emphatic negation 179, 186, 263, 275, 294

existential 4, 6, 10, 24–26, 31–34, 36– 38, 41–43, 170, 173, 243–244, 246, 249, 260–261, 307 existential closure 27, 38 focus 5, 23, 38–42, 45–46, 49–50, 64, 74, 85–87, 89–90, 93, 101, 117– 118, 130–132, 137, 139–140, 152, 157, 179, 221–222, 227, 230, 232, 238, 246, 255, 260–264, 325 fragments 7, 155–158, 166, 297 free choice 9, 34, 86, 170, 172–173, 305–308, 312–313, 317–318, 327, 329, 332–333, 336 free relatives 305–307, 308–311 goal 5, 23–24, 28–31, 33–36, 41, 43– 44, 47, 61, 65, 68, 72, 74, 237 grammaticalization 1–2, 51, 66, 78, 89, 101, 107, 109, 182, 204, 219, 238, 253, 267, 279–281, 317–318, 343–344 indefinite pronoun 85–86, 169, 171, 274, 299–302, 306–307, 310–311, 313–317, 325, 334, 336, 343 information structure 77, 86, 93, 109, 115, 130 interrogatives 7, 12, 150, 190, 193– 195, 197–199, 202, 209–211, 214– 215, 218, 260, 276–277, 285, 296, 298, 300–303, 304–307, 309 language acquisition (L1) 4, 185, 285– 286, 289, 295, 308–309, 318–319, 326–327 language contact 160, 309, 313–314, 329 mass nouns 96, 338–340, 342–343 minimizers 11, 169, 264, 314–315, 327 modal verbs 26, 68, 117, 218, 291 modal particle 87–89, 101–102, 294

348

Subject index

movement 2, 11–12, 29, 35–36, 44– 45, 48, 71, 78, 80–81, 88–90, 94, 97–101, 103–108, 115–116, 118– 119, 125, 127–129, 132–134, 141, 157, 179, 200, 212, 230, 247, 249– 250, 256, 259, 262–266, 273, 290 ne-absence 179–191, 193, 195–203, 209–210 ne-drop 3, 145–147, 149, 152–155, 158–160, 165–170, 174–175 negative concord 3, 5, 40, 43, 61–63, 65, 72–73, 133–134, 147, 150, 166, 168, 174, 203, 218, 221–232, 235, 237, 240, 251, 267–268, 273– 274, 293, 333 negative evidence 295 negation marker 77, 102, 221–224, 227, 294 low position 80, 101, 103, 104 negative operator 14–15, 35, 38, 46– 48, 61–66, 68–69, 71, 73, 89, 101, 132–135, 228, 236, 239–240, 287 negative polarity 4, 6–8, 10–13, 16, 61, 65, 160, 167, 169, 173, 225, 274, 279–280, 285–287, 290–291, 295–296, 302, 304, 306–311, 313– 314, 325–332, 336–337, 340–341 weak negative polarity context 285, 288, 292, 294, 296, 303–305, 309, 315 NegP 77–78, 80–81, 83, 102, 106–109, 115, 119–122, 126–127, 134, 138, 142, 174, 200, 230, 232 null operator 47, 134–136 number 9, 12, 28, 30, 44, 84, 86, 94, 98, 103–105, 108, 116, 118, 141, 148, 165, 171, 175, 184, 195–196, 222–223, 229, 243, 247–253, 255– 256, 259, 261–263, 267, 273–274, 276, 278, 281, 314, 316, 331–333 n-word 1, 3–15, 17, 28, 38, 43–44, 46– 50, 62–74, 149, 166–170, 172, 174–175, 221–223, 255–263, 265– 266, 268, 273–281, 286, 299, 336

object 2, 4, 28–30, 36, 44, 80–81, 83, 85–86, 91, 95, 99, 102, 104–106, 108, 117–120, 122–126, 128, 130– 131, 140–141, 149, 192, 236, 240, 261, 335 phase 16, 26–27, 29–30, 135, 256, 266, 293 positive polarity 171, 286, 295, 326, 328–332 probe 5, 23–24, 26, 28–30, 33–36, 41, 43, 45, 47, 50, 61, 63, 65, 67, 71– 72 quantifier (quantification) 12, 24, 27, 30–34, 36, 40, 43, 48, 62–64, 69, 72–73, 82, 90, 102, 116, 165, 167, 169–174, 225, 231, 234, 236, 238, 243–244, 246–251, 255, 259, 263, 265, 268, 274–275, 279, 281, 285– 286, 296, 298, 306–311, 337 weak 31–32, 34, 36, 38, 69, 73, 247–248, 250–251 reanalysis 12, 50, 78, 182, 191, 264– 265, 281, 299, 307, 315 secondary negation 77, 81, 89–90, 100, 102–104, 106, 108–109, 115– 127, 129–130, 132, 134, 136–137, 139, 141 semantic map 6, 9, 11–12, 301, 304, 308, 335–336 sentential negation see clause negation V-to-C movement 116, 118, 125, 129, 132–134, 141 V1 (Neg V1) 134, 136–138, 140, 142 verbs of indi¤erence 291 VP-internal subject 128–129, 131 word order 83, 88, 98, 100, 130–131, 138, 140, 165–166, 182, 188, 191, 193, 195–196, 199, 203, 210–211, 214–215, 218, 277, 292

Language index Afrikaans 27, 50 Arabic 182 Berber 182 Breton 305–306, 310 Old 310 Middle 309–310 Bulgarian 290 Catalan 13, 303, 333 Cornish 309–310, 313–314, 318, 329 Creole French-based 2, 4–5, 71, 267 Haitian 231–232, 267–268 Martinique 235–236 Dutch 5, 16, 27, 50, 169–172, 291– 293, 295, 325–326, 328, 331, 334– 335, 337–338, 340, 342–343 Middle 16, 165, 167–169, 174–175 English 3, 5, 17, 24–27, 30, 32, 34, 36–38, 40–46, 49–51, 64, 66–67, 77–78, 85, 102, 106, 108–109, 115, 121, 123, 129, 133–134, 136– 138, 141, 145–150, 157–161, 165– 166, 168–170, 225, 229, 231, 236, 286, 327–331, 334, 342 Old 3, 77–84, 86–90, 93, 97–102, 104, 106, 108–109, 115–123, 125– 132, 134–137, 139–142, 146–147, 149, 313, 317, 341 Middle 3, 77–81, 88–89, 101–109, 120, 122, 125–126, 137–142, 145– 150, 152, 154–155, 159, 161, 165– 167, 170, 173–174, 292, 313 Early Modern 105–107, 287, 292 Standard 5, 23, 38–44, 50, 64, Non-standard 44 Finnish 306–307, 332 Flemish 156, 160, 344 West 27, 133–134

French 1–6, 8–11, 13, 15–17, 23, 30, 42–49, 51, 61, 64–69, 71, 148– 151, 155, 157–160, 166–167, 172, 174, 179–183, 185–188, 190–195, 203, 209, 211–212, 217–219, 223, 227–229, 231–239, 242–243, 245, 248–252, 257, 260–263, 267, 273– 282, 285, 290, 292, 299, 301–303, 305–306, 317, 327–328, 334, 337 Old 11, 61, 66–69, 71–73, 148–150, 156–157, 160, 182–183, 252–253, 260, 274, 290, 292, 344 Middle 1, 150, 160, 183, 194, 228, 252, 260, 278, 301, 327, 330, 336, 344 Classical 4, 9, 12, 15, 68, 182–186, 190–191, 202–203, 257, 292 Contemporary 2, 4, 7, 67–68, 72, 223, 228, 230, 232, 234–235, 238, 242, 250–251, 256–257, 268, 274– 277 Standard 64 Quebec 4, 13, 15, 180–181, 183, 186, 193–195, 257, 334 Gaelic Scottish 314–315, 327–328, 337 Uist 315 Galician 303 German 26–27, 50, 87–89, 147, 169, 229, 288–289, 291–292, 328, 332, 336, 341–342 Middle High German 167, 288, 292, 317 Greek 5, 13–14, 26–27, 146, 182, 224, 229, 333 Hebrew 332 Hindi/Urdu 312, 333 Icelandic 332 Irish 25–26, 290, 317 Old 290, 311–312, 315

350

Language index

Italian 23, 38, 42–49, 51, 67, 152, 224, 236, 240, 248, 302–303, 332 Latin 3, 17, 183, 209, 228, 252, 285, 290 Lithuanian 312–313, 317

Slavonic 4–5, 45, 147, 290, 308, 315– 318, 328 Old Church 317 Old East 316–317 Spanish 224, 303 Swedish 147, 330–331, 333

Maltese 332

Turkish 302

Persian 302 Polish 305–306 Portuguese 303, 333

Venetian 303

Romance languages 27–28, 35, 42, 209, 302–304, 308 Romanian 45, 303, 333, 344 Russian 241, 290, 306, 315–316

Welsh 4, 8, 10–11, 25, 32, 147, 160, 290, 296–302, 306–307, 309–310, 312–313, 317 Middle 290, 296–298, 309–312, 327, 330–331