Retour vers liste

Détail de la contribution

Auteur: Viviane DÉPREZ

Co-Auteur(s): Anne Cheylus Pierre Larrivée

When and How is Concord preferred? An experimental approach

Abstract/Résumé: A longstanding debate asks whether negative polarity (1a.NPI) and negative concord (1b.NC) involve identical or distinct syntactic/semantic operations. Although French (1a&b) and cross-linguistic equivalents share the same first-order-logic interpretation (2a), disagreements remain as to how it obtains for each. Yet only (1b), ambiguously allows a double negative (DN)(2b). Taking the English paraphrase in (2b) to be likewise ambiguous, May (89) proposed that DN encodes a compositional hierarchical scope relation between its negative quantifiers (3a), while NC involves the formation of a resumptive polyadic negative quantifier (3b). Applied to French, this analysis of NC has long ranging consequences. First, NC is clearly distinguished from NPI, as n-words are cast as negative quantifiers. Second, it puts French (1b) and English (2b) under the same theoretical umbrella, questioning the validity of any NC macro-parameter. Third, how (1b) and (2b) should be distinguished arises anew, particularly if French truly favors NC, but English DN. Indeed, although the analysis elegantly allows both readings for (1b)-(2b) without lexical ambiguity, it remains surprisingly vague as to which factors favors NC over DN. Processing costs, intonation, quantifier parallelism, structural complexity, clause boundedness and discourse have all been suggested to influence DN or NC, but in effect, little is known as to how speakers resolve such ambiguities in a single language. The paper explores this question experimentally. Ambiguous sentences like (1b) were paired with two scenes, each representing one reading. Subjects were asked to read them aloud and pick the one that representing its meaning. Sound production and choice time through mouse tracking were recorded. Quantifier parallelism, structural complexity and syntactic position were manipulated to probe their effects on interpretation. The design produces experimental data on the effects of quantificational parallelism, structural complexity, syntactic position, and processing time for NC vs. DN and intonation. On a resumptive quantification analysis of NC, theoretical predictions are as follows: A) parallel simple Pro-Pro structures should produce stronger NC preference than more complex NP-NP ones; non-parallel structures should favor DN. B) DN readings should lead lengthened choice time vs NC C) NC and DN preference should manifest characteristically distinct intonation contours. The paper reports on the first results of this experiment. (1)a. Personne ne fait quoique ce soit. b. Personne ne fait rien. (2)a ¬∃x ∃y do(x,y) ‘No one does anything’. b.¬∃x,¬∃y do(x,y) ‘No one does nothing’. (3)a.(NOx (NOy do(x,y))) b.NO do(x,y)