Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Verbs as privileged cues for sentence retrieval
Poster Session E, Sunday, September 14, 11:00 am - 12:30 pm, Field House
This poster is part of the Sandbox Series.
Sebastián Mancha1, Ellen Lau1; 1University of Maryland, College Park
Intro: In a sentence recall task, how might cues of different roles affect the retrieval of a sentence bound in episodic memory? Following work on structural priming and theories of lexicalized syntactic knowledge (Joshi & Schabes 1992, Segaert et al. 2013), we propose a privileged connection between the verb and the clause it heads, and aim to test this theory by examining sentence retrieval signals in MEG. Predictions: Verbs will be more informative cues to structure and meaning, controlling the combinatorics of the clause, e.g. differences in transitivity, verbs like 'notice' canonically require a direct object, whereas verbs like 'sit' prohibit one. Noun cues will only serve to assist retrieval of their argument position, which do not limit combinatorial possibilities as much. The relative structural informativeness of verbs then should lead to more information recollected post-cue comprehension compared to noun cues, which can be measured by activity in left Angular Gyrus (Rugg & King 2018). Because retrieved verbs and their associated structure encompass more of the ‘pre-compiled’ sentence, we hypothesize they will facilitate subsequent production, possibly suppressing activity in left posterior inferior frontal gyrus and left posterior middle temporal gyrus, relieved from some of their typical production computations (Matchin & Hickok 2019, Giglio et al. 2022). Materials: 150 transitive sentences and 150 intransitive sentences will be created, intransitives will include prepositional phrases for length and balancing the number of nouns present across sentences. [Transitive - The strict teacher noticed the noisy student // Intransitive - The lonely professor sat near the old dean] There will be three cue types: null (XXXX), subject, verb. Sentence-Cue pairs will be controlled for Mutual Information derived from a dependency-parsed corpus, such that any difference between conditions cannot be attributed to the statistical informativeness of the cues. Methods: Data will be collected from 25 participants using MEG. In critical trials participants will hear a sentence, complete a spoken arithmetic task to suppress phonological loop effects, then be visually cued to recall the sentence. Spoken responses will be recorded to facilitate speech-onset analyses and recall accuracy scoring. Evoked fields will be analyzed between cue display and speech onset. Previous work on memory retrieval electrophysiology points to a time window around 500-800ms post-cue presentation (Vilberg et al. 2006, Evans & Wilding 2012) in left AG that displays a parametric response to informativeness of a retrieved memory, which we expect to correlate with cue type. We then predict any suppressed activity in left MTG and IFG should follow the onset of this time window, ~600-800ms. Interpretations, Future work: Given a reliable effect of verb-cueing we would take this as evidence for a privileged connection between a verb and its clause, though the results would equivocate between a conceptual/event representational origin or a grammatical origin. Should this methodology prove successful in eliciting different neural markers, it can be extended to other constructions that may help to distinguish between these, such as passives, transitive sentences with cognate objects, etc.
Topic Areas: Language Production, Syntax and Combinatorial Semantics