2018 Friday Session C 1430

Friday, November 2, 2018 | Session C, Terrace Lounge | 2:30pm

Doing what you must: child actuality inferences in modal comprehension
D. Veselinovic, A. Cournane

We report experimental work testing whether children use morphosyntactic cues to interpret the Bosnian/Croatian/Serbian(BCS) modal morati ‘must’ as deontic (e.g., obligation) or epistemic (e.g., inference). Morati poses the same learning challenge as must: one form maps to more than one abstract meaning[1,2,3]. English studies with must show children from age 3 may use grammatical aspectual cues to distinguish modal meanings[4,5,6], but prefer epistemic interpretations of must regardless of presence or absence of these cues by age 5 (Fig.1)[5,6]. Our results show that despite cross-linguistic differences in the most frequent input meaning[7] and structural cues[8], children behave similarly with morati as must, showing a non-adult epistemic bias by the age of 5. We argue over-adherence to normative actuality inferences (X must V ® X is V-ing)[9] explains child patterns.

Morati and must offer a natural comparison. Must is primarily epistemic in English input[3], while morati is overwhelmingly root in BCS input[7]. English deontic-biased constructions (e.g., Scott must wear his rainboots) allow habitual epistemic readings[5,6], while in BCS, syntax categorically disambiguates epistemic structures (BICLAUSAL) from deontic (MONOCLAUSAL) (Confirmed via grammaticality study with 25 adults; Fig.2). Thus, BCS morati allows us to test three hypotheses for child modal interpretations:

  1. Syntax-driven: adultlike structure-to-meaning mapping;
  2. Input-driven: overgeneration of deontic interpretations;
  • Pragmatics-driven: overgeneration of epistemic

Methods. The study uses the picture-choice task from[3,8]. Four training items preceded ten (5 BICLAUSAL (EX.1), 5 MONOCLAUSAL (EX.2)) test items and five fillers in random order. The screen side of pictures was randomized. On each trial, participants heard one of the sentence types, and chose between two pictures (Fig.2), EPISTEMIC vs. DEONTIC. Participants were preschool-aged children: three-year-olds (n=14, M=42mos, SD=3.7mos), four-year-olds (n=15, M=54mos, SD=3.4mos), five-year-olds (n=15, M=66mos, SD=3.9mos), and six-year-olds (n=12, M=77.5mos, SD=3.3mos); and adults (n=9, M=34.6ys, SD=10.46) from Sarajevo.

Results. Adults behaved as expected with BICLAUSAL, choosing epistemic pictures on average M=4.78(/5) times, but not with MONOCLAUSAL, choosing the unexpected (ungrammatical[8]) epistemic pictures M=1.67(/5) times. With BICLAUSAL, children get more adult-like (x-axis, Fig.3). For MONOCLAUSAL, 3yos were the most adult-like (epistemic pictures M=2.21(/5)), then children become progressively less adult-like (moving up the y-axis, Fig.3), with 6yos choosing non-target epistemic pictures M=3.5(/5) times.

Discussion. Our results support a pragmatic account which, unlike input-frequency or cue-based structural accounts, predicts the late-preschool epistemic bias of our BCS results and the extant English results. The findings for English speakers were ascribed to sentences like ‘X must V’ having habitual epistemic interpretations available[5,6]. We find the same pattern in BCS, where (Ex.2) is ungrammatical in epistemic contexts. We suggest BCS monoclausal structures with only root interpretations can be felicitously associated with EPISTEMIC pictures if participants draw an inference of the type X must V ® X is V-ing[9]. We argue world knowledge underlies this behavior – obligations normatively imply realization of the obligatory action, allowing participants to match root-biased X must V to epistemic pictures implying X’s normative behavior. Older preschool-aged children adhere more strongly to this inference than adults[cf 10]: if one must, one does.