Dutch ENIG: from nonveridicality to downward entailment

Jack Hoeksema 0 0 J. Hoeksema ( ) CLCG, University of Groningen , P.O. Box 716, 9700 AS Groningen, The Netherlands In combinations with singular count nouns, the Dutch indefinite determiner enig 'some/any' shows a diachronic distributional shift from nonveridical environments in general to a strict subset thereof, namely negative, conditional and interrogative contexts. Similarities with Greek indefinites of the kanenas-series are explored, and an argument is given that at some point enig has split into two uses, one of which is currently on its way out. Nonemphatic, nonreferring enig is disappearing, whereas emphatic enig is stable within a set of environments similar to those of English any or ever. 1Both den Dikken's and Postal's papers are concerned with secondary, or parasitic licensing as well as ordinary licensing. In parasitic licensing examples, licensing of a polarity item X by a trigger Y is mediated by another polarity item Z (see also Hoeksema 2007). - Precursors of this paper were presented at the LOT Winterschool in Groningen, 2005, and at the Negation and Polarity Conference at the Eberhard Karls Universitt in Tbingen, March 2007. I would like to thank the audience at that conference, as well as two anonymous reviewers, and Doris Penka and Hedde Zeijlstra for their comments, suggestions and criticisms. The blame for all errors and omissions, however, lies entirely with me. Zwarts 1981, 1998; Kadmon and Landman 1993; Krifka 1995; Israel 1996; van der Wouden 1997; Lahiri 1998; Giannakidou 1998; von Fintel 1999; Chierchia 2006; Gajewski 2008), or a combination of both (e.g. Linebarger 1981). However, it seems evident, that these distributional patterns are determined by more than the semantic and syntactic requirements of the polarity items themselves. One additional factor which should be considered is paradigmatic in nature, namely the presence or absence of competing elements. For instance, in many Slavic languages, indefinite polarity items are found in all contexts in which polarity items may normally be found, except one: direct clause-mate negation (Progovac 1994; Pereltsvaig 2006). This may seem surprising, given that negation is presumably the prototypical environment for polarity items. Pereltsvaig calls this the bagel problem: the distribution is that of negative polarity items, but with the central core missing. The apparent source of the bagel problem is the existence, in Slavic, of negative concord, which takes precedence over alternative types of exponence involving indefinite polarity items under the scope of negation. The missing core of the bagel is the result of some sort of blocking. More generally, it has been argued for by de Swart (2010) that negative concord and the expression of negation more generally call for an optimality-theoretic treatment, which compares various alternatives and selects the optimal one according to a set of ranked constraints. Jger (2010) likewise assumes that the complementary distribution of any and some is due to competition. The item with the more narrowly circumscribed distribution, any, blocks the use of its more general counterpart some by virtue of the Elsewhere Principle (Kiparsky 1973).2 A final complicating factor which should be mentioned here is prosodic differentiation. It is clear from a number of investigations that stressed occurrences of polarity items may vary in their distribution from unstressed occurrences (Sahlin 1979; Giannakidou 1997; Hoeksema 1999). Giannakidou, for instance, shows that Greek kanenas, when stressed, appears in a small subset of the environments where its unstressed variant can be found, namely negation, clauses introduced by xoris without or by prin before. Sahlin, in an early study using a prosodically marked-up corpus of spoken English, notes substantial differences between stressed and unstressed occurrences of any. Hoeksema, finally, reports on several prosodic differences between 2While this type of explanation is certainly appealing, there are also some recalcitrant problems still waiting to be resolved. For instance, while the distribution of some and any is largely complementary, there are some areas of overlap as well. In conditionals, questions and other not strictly negative environments, either determiner is acceptable: (i) Do you want some coffee? (ii) Do you want any coffee? The two questions are sometimes claimed to be pragmatically distinct (Borkin 1971; Ladusaw 1979), in that the latter, but not the former, question assumes a negative answer. However, the matter seems to be more subtle than that, and questions such as (ii) are certainly different from rhetorical questions. Likewise, environments of double negation are acceptable for both items as well (cf. Baker 1970, and for more contexts of double negation Szabolcsi 2004): (iii) I cant believe that you dont want some coffee. (iv) I cant believe that you dont want any coffee. polarity-sensitive and nonsensitive ooit ever in Dutch (possibility of comma intonation (in the case of nonsensitive ooit) and possibility of emphatic lenghtening of the vowel (cf. for discussion of this phenomenon van Ommen et al. 2007) in the case of polarity-sensitive ooit), compare the following sentences, where the spelling o-o-i-t indicates emphatic lengthening: a. Ik geloof niet dat ik o-o-i-t zo gelachen heb. I believe not that I ever so laughed have I dont believe I have ever laughed so much. b. Ik geloof dat ik ooit zo gelachen heb. I believe that I once so laughed have I believe I have once laughed so much. c. *Ik geloof dat ik o-o-i-t zo gelachen heb. I believe that I once so laughed have I believe I once laughed so much. In this paper I will discuss an almost completed change in the distribution of a Dutch polarity item, enig. I will argue that the narrowing of the distribution of this item is compatible with the hypothesis that the item has been reanalyzed from a nonreferential indefinite to a more emphatic type of indefinite with the property of pragmatic strengthening in the sense of Kadmon and Landman (1993). This reanalysis is accompanied by a change in distribution from nonveridical environments toward downward-entailing environments. The evidence on which the diachronic claims are based, comes from a collection of more than 100,000 occurrences of Dutch negative polarity items, collected by the author from various sources, including electronic texts such as the ones in the Digital Library of Dutch Literature and Language (see www.dbnl.org) and other Internet sources, the digital newspapers at the website of the Dutch Royal Library (www.kb.nl), but also from books and magazins read by the author over a long period. This collection is currently large enough to study diachronic changes in the more frequent polarity items, such as Dutch enig.3 2 Dutch enig The West Germanic languages all have determiners derived from the numeral one by means of the affix -ig (which developed into -y in English), comp (...truncated)