Retour vers liste

Détail de la contribution

Auteur: Lyn TIEU

Negative polarity and domain widening: A learnability perspective

Abstract/Résumé: NPI 'any' widens the domain of quantification in downward-entailing (DE) environments (Kadmon & Landman, 1993; Chierchia, 2006), unlike 'a', which generally involves a contextually restricted domain. This paper addresses the following learnability challenge: if negatively quantified any-statements appear interchangeable with equivalent a-statements, how does the learner pick up on 'any'’s domain widening (DW) properties? Their relative distribution represents a subset-superset schema: 'any' is licensed in a subset of environments compatible with 'a'. Based on an analysis of the Warren CHILDES corpus (5704 parent utterances), I propose that the child might pay attention to the proper superset of environments compatible only with 'a', as well as the proper subset. (i)If DW and DEness are indeed connected, an input distribution in which an item appears only in DE environments could trigger knowledge that it is a widener. Indeed we find many NPI-incompatible uses of 'a', e.g. positive declarative sentences. (ii)Within DE environments, 63 occurrences of 'a' involve environments that are in principle NPI-licensing environments, but that nevertheless don’t tolerate 'any'. 13/63(=21%) involve questions containing generic subjects (What does a duck say?); free choice 'any' might appear in such positions (Any owl hunts), but the corpora reveal no such occurrences. The remaining 50/63(=79%) involve predicative/property meanings (That’s not a duck), which 'any' resists. (iii)In environments compatible with both, the learner might also identify differences in their contexts of use; what is at-issue may differ for any- vs. a-statements. 30 instances of 'any' and 66 instances of 'a' appear in environments where either is grammatical. Many instances of these negative a-statements involve a previously mentioned object(1); such previous mention is less common with 'any'. The contextual conditions that lead up to the negative assertions can thus differ. Another potential cue lies in number: in many cases, 'a' is used in contexts where only one entity is relevant; 'any' always quantifies over a plural set. Finally, we find potentially unambiguous evidence of widening; when one child says a rabbit makes the rabbit noise “oo”, the mother replies with (2). In summary, we can identify a number of distributional as well as contextual cues that likely all contribute to triggering knowledge of DW. (1)CHI: you have a wabbit! FAT: do I have a rabbit? I don’t have a rabbit. (2)no, I don’t think the rabbit makes any noise. Selected References Chierchia, Gennaro. 2006. Broaden your views. Implications of domain widening and the “logicality” of language. Linguistic Inquiry 37(4):535-590. Kadmon, Nirit & Fred Landman. 1993. Any. Linguistics and Philosophy 16:353-442.