2018 Sat Poster 6449

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

When Must Children Acquire Long Distance Wh Extraction?
J. Kotfila, J. de Villiers

The literature has established that young children are capable of long distance wh movement. Yet the child often answers an argument question like “What did mom say she bought?” with what mom really bought though she said she bought something else. This failure to integrate the matrix and embedded verbs in a reality type error extends to adjunct questions containing say, tell, and think (de Villiers, Kotfila, & Klein, 2016). Children are less likely to make such an error when confronted with irrealis infinitival complements than they are with realis tensed complements (de Villiers et al., 2011). In keeping with irrealis infinitival complements, yes/no questions using think should were easier for children than think is constructions (de Villiers, 2005) The present study extends these findings to wh adjunct questions containing modal complements with an epistemic and deontic modal flavor.

Epistemic and deontic modals interact differently with tense and aspect, as deontic modals scope below tense and aspect and epistemic modals scope above (Hacquard, 2017) . Epistemic modals have realis mood and are found adjacent to TP, whereas deontic modals are adjacent to the VP and take an irrealis mood. The hypotheses were:

  1. Children will provide fewer reality type answers for modal complements than for non-modal complements, subsequently resulting in more adult-like long distance
  2. Children may provide more reality type errors for an epistemic modal (realis) than a deontic modal (irrealis).

In a between-subjects design 36 children, 12 aged 3, 12 aged 4, and 12 aged 5 with a mean age of 4;4, and 35 adult controls received either an epistemic or a deontic modal scenario, on otherwise carefully matched stories. The format of the test questions and stories mimicked the adjuncts (when, where, and why) and matrix verbs, (say, tell, and think) of de Villiers, Kotfila, and Klein (2016) to be able compare the date to that non-modal control group. All answers were contained within the stories, but in different orders to rule out a recency effect.

Using one-way ANOVAS, there was no significant effect of modal flavor on answer type for reality type answers (F(2,33)=.241, p=.7. There was a marginal age group effect as older children were less likely to produce reality type answers (F(2,27)=2.7, p=.048). As there was no effective of modal flavor on answer type, the epistemic and deontic conditions were collapsed into a single modal condition to compare to the non-modal data from de Villiers, Kotfila, and Klein (2016). Children were much less likely to give reality answers for modal than non-modal complements (F(1,80)=12.39, p=.001.)

We argue that children are treating epistemic modals and deontic modals the same, representing both as low modals in terms of their syntactic position. Because tense is irrelevant as to whether or not the modal complement has been actualized, there may be no separate Assertion and Point of View for the lower clause, making correct long distance answers easier for the child (Roeper & de Villiers, 2011).

References

de Villiers, J.G., *Kotfila, J. & *Klein, M. (2016 in press) Parsing, Pragmatics, and Representation: Children’s comprehension of two-clause questions. GALANA proceedings.

Roeper, T., & J.G. de Villiers. 2011. “The acquisition path for wh-questions”. Handbook of generative approaches to language acquisition ed by J. de Villiers & T. Roeper, 189-246. The Netherlands, Springer.

Hacquard, V. (2017). Modals: Meaning Categories?. Mood, Aspect, Modality Revisited: New Answers to Old Questions, 45