Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Modelling the late frontal positivity: an investigation into the relationship between prediction failure and semantic learning
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.
Yiming Lu1; 1University of California, Irvine
Background: prediction is widely acknowledged as a fundamental mechanism in language comprehension. Yet, the consequences of prediction failures remain debated. Prediction failures leave electrophysiological traces in the brain. When an unexpected but plausible word disconfirms comprehenders` prediction, a post-N400 frontal positivity (short for frontal PNP) signal has been elicited by various studies (DeLong et al., 2014; Federmeier et al., 2007; Kutas, 1993; Moreno et al., 2002; Quante et al., 2018). Many proposals have been advanced to explain the underlying cognitive processes, including prediction error detection (Delong et al., 2011; Thornhill & Van Petten, 2012), inhibition of the expected words (Federmeier et al., 2007; Kutas, 1993; Ness & Meltzer-Asscher, 2018), as well as the memory enhancement of the unexpected words (Lai et al., 2021). However, the above hypotheses have been overshadowed by subsequent experiments, and the nature of frontal PNP remains elusive. In this proposed study, we hypothesize that frontal PNP reflects the updating of the probabilistic distribution of words in contexts (Rommers, 2022). Furthermore, we propose that different thematic roles modulate the updating process differentially, such that core thematic components of an event will have a larger frontal PNP reduction than non-essential ones. Methods: we use repetition priming to examine these hypotheses. 40 monolingual English speakers will be recruited. Two studies will be conducted. Contexts are biased towards an instrument in Study 1 and a manner in Study 2. All stimuli will be examined in a cloze probability test to make sure all sentence contexts are highly constraining (>70% cloze probability). Unexpected but plausible continuations are created for qualified sentences, which are subjected to plausibility judgments (5-point Likert scale, rating > 1.5). Participants read a sentence with prediction disconfirmation (e.g., the boy swept the floor with a robot, biased for instruments), which was followed by filler sentences. Then comes the critical sentence that differs from the initial sentence only in the final word. The critical sentences contain three conditions: same word/expected role (robot), different word/expected role (rag), and different word/unexpected role (smile). A linear mixed model with by-subject and by-item intercepts and slopes will be constructed for frontal PNP. Predictions: Regarding hypothesis 1, we predict that frontal PNP will be lower if the unexpected word is repeated in the second presentation (i.e., robot), compared with violations seen for the first time (i.e., rag and smile). This means that frontal PNP is sensitive to repetition priming and potentially, the perturbation of the local statistical distributions. However, if the opposite is borne out, the hypothesis is likely wrong. Regarding hypothesis 2, we predict that frontal PNP is sensitive to thematic roles. If contexts suggest a core thematic role, like instruments, the PNP amplitude for another unexpected instrument (i.e., rag) should be lower than the unexpected role (i.e., smile). This suggests that the belief updating for word association is differential depending on the kind of information encoded. However, if there is no significant difference between instrument and role, PNP is likely insensitive towards thematic roles.
Topic Areas: Reading,