2018 Alternates 6740

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

Better to be reliable than early: Cognitive-control effects on developmental parsing
Z. Ovans, J. Novick, Y. Huang

Children parse sentences incrementally, and this leads to errors when initial misanalyses conflict with later-arriving evidence. This occurs when verbs predict misinterpretations (e.g., “Put the frog on the napkin into the box” à misinterpret PP1s as goals [1]) as well as when verbs revise misanalyses (e.g., “The dax is chased by the monkey” à misinterpret NP1s as agents [2]). These challenges are often attributed to immature cognitive-control abilities, which are associated with with poor conflict resolution in non-linguistic [3] and linguistic tasks [4].

However, cross-task correlations do not reveal how cognitive-control abilities are deployed in real time to influence syntactic parsing during development. Does cognitive-control engagement always enhance garden-path recovery [5]? Alternatively, are these effects mediated by reliance on reliable parsing cues? Having an agent-first bias offers an early cue to role assignment, but predicts incorrect roles for unaccusatives, perspective verbs, and passives. In contrast, verb- specific biases provide more reliable cues to meaning, but are often delayed in sentences like passives. Thus, if cognitive-control engagement enhances reliance on reliable parsing cues, it should increase garden-path errors when verbs are misleading prediction cues (i.e., “Put” sentences), but decrease errors they are accurate revision cues (i.e., passives).

The current study manipulates cognitive-control engagement by interleaving Stroop (Congruent vs. Incongruent) and Sentence trials (Ambiguous vs. Unambiguous, Fig.1). Following prior work [5], Incongruent Stroop trials engage cognitive-control, and this impacts subsequent sentence comprehension. Experiment 1 examined verb-initial sentences, where correct-goal looks (e.g., box) after disambiguation indicated successful revision and incorrect- goal looks (e.g., napkin) indicated misinterpretation. For Unambiguous sentences, 5-year-olds (n=64) made more correct-goal looks following Incongruent- compared to Congruent-Stroop trials. In contrast, for Ambiguous sentences, they made fewer correct-goal looks following Incongruent- vs. Congruent-Stroop trials (Fig.2). This Stroop x Sentence interaction suggests that cognitive-control engagement promotes parsing via reliable cues (p<.01): These results are consistent with engagement biasing attention to early-arriving verbs (e.g., “Put” syntactically demands a location), causing children to predict incorrect locations for Ambiguous sentences. Nevertheless, it is possible that Incongruent-Stroop trials depleted children’s limited cognitive- control, hampering parsing of subsequent sentences.

To address this possibility, Experiment 2 (n=32) examined how cognitive-control engagement affects parsing when verbs are revision cues in passives. Unlike Experiment 1, passives have late-arriving reliable verb cues, so successful interpretation will indicate reliance on the verb, even as a revision cue. After passive morphology, likely-theme looks indexed successful recovery from the agent-first bias while likely-agent looks indicated a failure to do so. Similar to Experiment 1, a Stroop x Sentence interaction emerged (p=.01). Here, children made less accurate looks for actives following Incongruent- compared to Congruent-Stroop trials. In contrast, passives yielded more accurate looks following Incongruent- compared to Congruent- Stroop trials. Together, these results suggest that cognitive-control engagement influences developmental parsing by increasing reliance on reliable cues. This improves revision when verbs are accurate but hinders revision when verbs lead children astray. Limited cognitive- control may facilitate acquisition of linguistic representations (e.g. argument structure) by pushing children to follow reliable cues that will, by definition, usually lead to correct interpretations.

References

[1] Trueswell, Sekerina, Hill, & Logrip, (1999). The kindergarten-path effect: Studying on-line sentence processing in young children. Cognition. [2] Huang, & Arnold, (2016). Word learn- ing in linguistic context: Processing and memory effects. Cognition. [3] Diamond, Kirkham, & Amso, (2002). Conditions under which young children can hold two rules in mind and inhibit a prepotent response. Dev Psychol. [4] Woodard, Pozzan, & Trueswell, (2016). Taking your own path: Individual differences in executive function and language processing skills in child learners. J Exp Child Psychol. [5] Hsu & Novick, (2016) Dynamic engagement of cognitive control modulates recovery from misinterpretation during real-time language processing. Psychol. Sci.