2018 Friday Poster 6663

Friday, November 2, 2018 | Poster Session I, Metcalf Small | 3pm

Lexical aspect as a crosslinguistic cue to modal meaning: evidence from Dutch
A. van Dooren, M. Tulling, A. Cournane, V. Hacquard

Functional modals like must in Dutch and English can be used to express both epistemic (likelihood) and root (obligation) necessity (Ex.1). How do children figure out that the same modal can express different flavors, especially given that their usage distribution seems heavily skewed towards root meanings^[1],[2]? Previous literature shows that  children tend to produce root meanings before epistemic^[4], echoing the skew of the input. How do children eventually pick up epistemic meanings? We argue that children exploit distributional cues stemming from aspectual properties of the modal’s complement.

While modals can express both root and epistemic flavors, the former have to express possibilities about the future^[7]: thus root meanings are incompatible with complements that force a past orientation, which arises with a perfect (John must have left) or a present orientation, which arises with the progressive (John must be leaving), or a stative (John must be happy). Hence, grammatical aspect (perfect/progressive) and lexical aspect (statives vs. eventives) could give away epistemic readings. [1] shows that while modals are heavily skewed towards root flavors in English, epistemic and root flavors differ in how they interact with both lexical and grammatical aspect.

In this corpus study, we show that the learning problem is more dire in Dutch than in English as the input has an even greater skew towards root flavors. Moreover, there are fewer grammatical aspect cues. However, lexical aspect clearly differentiates roots from epistemics.

Methods. We examined 7 child-input corpora (Groningen^[5], age range = 1;05-3;06). We extracted utterances with modals (child: 12,880/82,108 (15.7%); adult: 40,486/ 18,1003 (22.4%)). We coded modals for _SYNTACTIC__CATEGORY [lexical, functional] and _FLAVOR   [root, epistemic, future] (Ex.2). To determine _FLAVOR, we hand-examined the transcripts for 3 corpora (n=8,364). We coded all input items for _LEXICAL__ASPECT (stative, eventive) for one corpus (n=1,823), and all epistemic usages (n=102) + an equally sized random sample of the non-epistemics (root, future) for the other two corpora.

Results. Dutch children hear epistemic talk using lexical verbs (e.g. think, know) and adverbs (e.g. maybe, probably) as frequently as English children (~5% of total utterances; Ex.3). As in English^[6], children produce some lexical epistemics (Ex.7). Epistemics in Dutch are however rarely expressed by functional modals (e.g. must), which are only used epistemically 1% of the time (Ex.4, (n=8,460); c.f. 8% for English (/n=2,592)^[1]). Lexical aspect in Dutch is distributed similarly over root versus epistemic flavor as in English: 38.7% of the epistemics with a verbal complement (n=75) take an eventive complement versus 95.1% of the roots (n=924, Ex.5). A generalized linear mixed-effects model supports that lexical aspect is a significant predictor of flavor in Dutch (Ex.6).

Discussion. This study investigated how Dutch children figure out that functional modals are polysemous, given that epistemic flavors are less frequent than in English. We find additional support for the hypothesis that children may be able to exploit distributional cues in the input^[1], based on constraints on how modal flavors interact with aspect: while infrequent, epistemics tend to combine with stative complements, while root flavors mostly take eventive ones.

References

[1] van Dooren, Dieuleveut, Cournane & Hacquard (2017). Learning what must and can must and can mean. Proceedings of the Amsterdam Colloquium. [2] Mortelmans (2012). Epistemic MUST and its cognates in German and Dutch. Journal of Pragmatics 44, 2150—2164. [3] Clark (1987). The Principle of Contrast. In MacWhinney (ed.), Proceedings of the 20th Carnegie Symposium on Cognition, 1-33. Erlbaum. [4] Kuczaj & Maratsos (1975). What children can say before they will. Merrill-Palmer Quarterly of Behavior and Development 21, 89–111. [5] Wijnen & Verrips (1998). The Acquisition of Dutch Syntax, In Gillis & De Houwer (eds.), The Acquisition of Dutch. Benjamins. [6] Cournane (2015). Modal development. PhD. Thesis. UToronto. [7] Condoravdi (2002). Temporal interpretation of modals. In Beaver, Casillas Martinez, Clark & Kaufmann (eds.) The construction of meaning, 59-88. CSLI.