Presentation

Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Poster Slams

Predictability effects on auditory word recognition with a pre-primed semantic priming task

Poster B15 in Poster Session B and Reception, Thursday, October 6, 6:30 - 8:30 pm EDT, Millennium Hall

Anne Marie Crinnion¹, James Magnuson^1,2, Emily Myers¹, Phoebe Gaston¹; ¹University of Connecticut, ²Basque Center on Cognition Brain and Language,

Understanding the mechanisms underlying context effects during language processing is key for adjudicating between neurocognitive frameworks for spoken word recognition, such as Predictive Coding (PC) and Interactive Activation (IA). Of particular interest is whether predictability effects in behavioral and neural responses arise because predicted lexical items lead to ultimately higher activation levels or because they elicit smaller changes in activation. We build on work from Blank and Davis (2016) that used visual repetition priming with noise-vocoded auditory targets (e.g., SING sing vs. XXX sing), in a design necessitating multivariate analysis of the fMRI response to auditory items in order to make differing predictions for PC and IA. Instead, in this study we measure the semantic priming that the primed vs. unprimed auditory item causes (e.g. DUCK duck GOOSE vs. XXX duck GOOSE). This paradigm allows us to analyze how predictability constrains auditory processing by revealing the impact of the visual pre-prime on not only (1) the lexical activation levels that result from hearing the auditory item but also (2) the change in lexical activation that this constitutes relative to a control. We present clear rather than vocoded speech and fully manipulate the status of the visual pre-prime as (a) matching the auditory item, (b) mismatching, or (c) neutral. 204 participants saw a pre-prime that either matched the auditory item (e.g., DUCK), did not match the auditory item (e.g., TABLE) or was neutral (XXX). Participants then heard the auditory item (e.g., duck or an unrelated control, e.g., table), and then made a lexical decision on a semantically related written target (e.g., GOOSE). We found a significant interaction between the status of the pre-prime and the status of the auditory item. In planned pairwise comparisons, the semantic priming effect caused by the auditory item (hypothesized to reflect the change in lexical activation due to hearing the auditory item) was significantly smaller with the matching pre-prime than the mismatching or neutral pre-prime. This arose because, among the control conditions, we observed significantly faster reaction times for match trials (DUCK table GOOSE) than mismatch or neutral trials (TABLE table GOOSE, XXX table GOOSE). When the auditory item and target were semantically related, reaction times were significantly faster across the board, but were not impacted by the pre-prime (i.e., no difference between DUCK duck GOOSE, TABLE duck GOOSE, and XXX duck GOOSE). This pattern raises many questions for further investigation. It is potentially consistent with an account in which typically observed facilitation effects for matching or predictable words occur because these words evoke smaller changes in activation, rather than because they ultimately reach higher lexical activation levels than mismatching or unpredictable words. The paradigm developed in this study and the findings we present here should help distinguish between predictions generated by different models of spoken word recognition and inform future neuroimaging designs, ultimately leading to a deeper mechanistic understanding of prediction and context effects in language processing.

Topic Areas: Speech Perception, Perception: Auditory