2018 Sat Poster 6732

Saturday, November 3, 2018 | Poster Session II, Metcalf Small | 3:15pm

Learning the force of modals: Sig you guess what sig means?
A. Dieuleveut, A. van Dooren, A. Cournane, V. Hacquard

This study investigates how children learn the force (‘possibility’ vs. ‘necessity’) of modals like can or must. Results from prior comprehension studies suggest that preschool children have not mastered the force of modals: they tend to both accept possibility modals when necessity modals are more appropriate[1] – a behavior attributed to their difficulties with Scalar Implicatures -, and to accept necessity modals in possibility situations[2], a more surprising result. Learning force from the input presents at least two challenges: first, from a logical point of view, necessity entails possibility; second, speakers often use possibility modals to soften commands. How do children figure out which modals express possibility and necessity, more specifically, that necessity modals are stronger than possibility modals? We hypothesize negation may play a key role.

We conducted an English corpus study, to assess how force is naturalistically distributed in the input, and a Human Simulation Paradigm (HSP)[3] study, using contexts from the corpus, to assess first the informativity of context regarding force. Corpus results show that necessity modals are much less frequent than possibility modals in the input, and negation is more frequent with possibility than necessity modals, especially in children’s production. HSP results show that adults can better identify necessity than possibility, which suggests that contexts in which necessity modals are used may be more ‘informative’.

Corpus study. We examined modal production of 12 child-mother pairs (1;09-3;00) from the Manchester corpus[4] (564,625 total utterances), on CHILDES[5]. 26,326 (adult: 20449; child: 5877) modal utterances were coded for force (possibility vs. necessity) and negation. Results. We find a usage force asymmetry for both children and adults: possibility modals are much more frequent than necessity modals (Table.1). Likewise, negation is more frequent with possibility than necessity. This difference is significantly higher for children (Table.2). Children’s first occurrences of negated necessity appear later (2;5) than negated possibility (2;0), which can appear before plain possibility[6] (Table.3).

HSP. To assess informativity of the context for force, we ran an HSP[4] using IBEX. 248 participants recruited on MTurk guessed redacted modals. 400 contexts were randomly extracted from our corpus (Table.4) for 4 different modals (can, could; must, should) (between subjects); all contexts consisted of a target sentence with a blank, and 7 preceding utterances. Each participant saw 4 dialogues per modal. Results. Overall, participants were better at guessing necessity than possibility modals (Figure.2). Most errors with possibility occurred with in questions subjects guessed do instead of can. Performance improved over the course of the four dialogues.

Discussion. While contexts for necessity modals seem more informative, necessity modals are not frequent in the input. Child production may reflect the priority of possibility modals, and the potential helpfulness of negation: they use possibility modals both for possibility contexts (can) and necessity contexts (can’t). Necessity modals, especially negated ones, appear delayed. This may explain earlier behavioral results: if children do not fully grasp necessity modals, they will both accept possibility modals in situations where necessity modals are more appropriate, and accept necessity modals in possibility contexts.

References

[1] Noveck, I. 2001. [2] Ozturk, O. & Papafragou, A. 2015. [3] White, A. 2016. [4] Manchester Corpus.

Theakston, A et al. 2001. [5] MacWhinney, B. 2000. [6] Papafragou, A. 1998.