Search Abstracts | Symposia | Slide Sessions | Poster Sessions
The (Non-)Sentence Repetition Task: A novel assessment of expressive and receptive syntactic competence
Poster Session A - Sandbox Series, Thursday, October 24, 10:00 - 11:30 am, Great Hall 3 and 4
This poster is part of the Sandbox Series.
Jeremy Yeaton1, Shiva Upadhye1, Richard Futrell1, Gregory Hickok1; 1University of California, Irvine
Introduction. Whether and how receptive and expressive syntactic abilities are related to one another is widely debated, however, high-quality evidence bearing on this question is relatively sparse. Here, we introduce a novel task--the (Non-)Sentence Repetition Task (NSRT)--which makes apples-to-apples comparisons between expressive & receptive syntactic performance on the same items in the same individuals. We present preliminary data from healthy participants ahead of collection from individuals with post-stroke aphasia. Background. Expressive & receptive syntactic competence have typically been assessed using separate tasks with distinct materials. On the receptive side, acceptability judgment tasks are often used, which ask participants to judge whether presented sentences conform to their grammar. On the expressive side, constrained sentence repetition or elicitation tasks, or unconstrained discourse elicitation tasks can be used. Within the context of sentence repetition/elicitation, the Sentence Superiority Effect (SSE)--where a sequence of words is more accurately recalled if it is a well-formed sentence than if it is not--can provide a window into subconscious syntactic processing (Scheerer, 1981). An unanswered question is whether participants can consciously generate a grammatically correct sentence which resolves an error in a presented (ungrammatical) sentence. Methods. Trials in the NSRT occur in 3 phases: 1) judgment of a presented sentence as acceptable or unacceptable; 2) verbatim typed or oral production of the presented sentence even if it was ungrammatical; and 3) production of the corrected sentence if it was judged to be ungrammatical. The stimuli used three types of violations: determiner-noun number agreement ('two boy'), subject-verb number agreement ('two boys walks'), and tense agreement ('two boys were walk'). Each participant saw 60 trials of which 20 were grammatical. We recorded response time and accuracy for the judgments, as well as timestamped typing data for the verbatim and correction phases. We present data from 10 healthy young adult pilot participants here but data collection is ongoing in healthy young adults, older adults, and stroke survivors. Results. Overall, participants performed with both high accuracy (81.8%) and good discriminability (A'=0.90) on the acceptability judgments. Participants were significantly faster and more accurate at judging grammatical than ungrammatical sentences. They showed a significant SSE in the verbatim phase, producing more accurate repetitions in the grammatical than ungrammatical conditions. During the correction phase, participants were highly effective at correcting the presented sentence to generate a grammatical one (95%). During the verbatim phase, restarts constituted a higher proportion of corrective actions when the stimuli were ungrammatical, and fewer restarts were observed when stimuli were judged as grammatical. Additionally, more restarts were observed in the verbatim phase than in the repetition phase for ungrammatical items. Discussion. Our initial results show a significant effect of grammaticality on judgment accuracy and RT, and accuracy of verbatim repetition. They further show subtle differences between the different kinds of grammatical violations, pointing to different processing levels or strategies. Analysis of the backtracking data points to different planning and inhibition strategies in the two production phases. An outstanding question is whether these strategies would be mirrored in individuals with aphasia.
Topic Areas: Language Production, Syntax and Combinatorial Semantics