Conférencière : Ailis Cournane, professeur à la New York University (NYU) et directrice du Child Langage Lab à NYU

Titre : Must in Toronto English: the view from acquisition and language change. 

Résumé : 

Modal verbs like must encode both root (obligations, goals) and epistemic (inferences from knowledge or evidence) meanings, and as such are variable. Variable-meaning modal verbs like must historically arise from root-only modals (e.g., wish, want, know how) by gaining additional epistemic meanings (e.g., Traugott 1989). These variable-meaning modals then very gradually lose root meanings over historical time, as the innovative epistemic use increments in the speech community (see Labov 2001; Tagliamonte & D'Arcy 2009). In Toronto English root must has largely been replaced by historically newer modal verbs like have to, though root interpretations remain present (Tagliamonte & D'Arcy 2007). I'll discuss a the role child learners play in this incrementation (increase in frequency and contexts of epistemic must) drawing on evidence from a series of experiments with Toronto preschoolers (Cournane 2014, 2015; Cournane & Pérez-Leroux, subm.).

In the focus study, a picture preference task compared interpretations for must with bare verb complements (variably deontic or epistemic with a habitual construal) versus must with grammatical aspect-marked complements (epistemic, without coercion). By 5-years-old children become adult-like for aspect-marked epistemic sentences, but for variable bare-verb sentences they significantly overgenerate epistemic interpretations relative to young adults in the same speech community. A second study confirmed that preschoolers have and maintain access to root interpretations for must in the full sample (3 to 6 years), showing they do not simply reanalyse must as categorically epistemic, but know that syntactic cues to prejacent semantics play a key role in modal verb interpretation (see Gleitman 1990; cf. Condoravdi 2003, Hacquard 2010).

I propose children overgenerate epistemic interpretations because of the mapping between syntax and semantics: must uniformly precedes grammatical aspect (must have X-ed) in the input syntax, isomorphically keying its higher epistemic interpretation at LF (above TP; Hacquard 2006). Once children reliably learn this complex construction, they overextend the epistemic interpretation to variable-meaning must sentences with a bare verb complement in an effort to regularize syntax-semantics relations in their grammar. I argue internal learning bias underlies the gradual root > epistemic incrementation over time, and further suggest child grammatical learning biases are a plausible "primary mover" for sociolinguistic incrementation phenomena (cf. vernacular reorganization, Labov 2001:415).