Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Dissociating the effects of semantic predictability and plausibility in dynamic environments: A computational trial-by-trial EEG analysis
Poster D67 in Poster Session D, Saturday, October 26, 10:30 am - 12:00 pm, Great Hall 4
Meihua Zhao1,2, Wenjia Zhang3, Qi Chen4, Suiping Wang1,2; 1South China Normal University, Guangzhou, China., 2Philosophy and Social Science Laboratory of Reading and Development in Children and Adolescents (South China Normal University), Ministry of Education, Guangzhou, China., 3Xi’an International Studies University, Xi’an, China., 4Shenzhen University, Shenzhen, China.
Environmental statistical structures play a significant role in language prediction processing. Previous sentence comprehension studies have found the larger N400 effect in higher global predictive validity blocks, suggesting the top-down prediction strategy is reinforced. However, the N400 is not a direct index of a lexical prediction, also reflect prediction violation and bottom-up prediction error correction. It remains an ongoing debate how Environments influence semantic prediction. Furthermore, no study has contrasted how the statistical structures of environments influence two distinct semantic predictability and plausibility processing. Here, we addressed these issues by conducting a comprehensive trial-by-trial computational modeling analysis of EEG data. We first employed a normative Hierarchical Gaussian Filter (HGF) model to simulate individual learning trajectories in an uncertain and dynamic environment. Then, we performed generalized linear regression (GLM) analyses to clarify the association between single-trial EEG signals (ERPs and time frequencies) and model parameters. We observed that the N400 component was associated with 2nd-level belief during sentence predictability processing. In contrast, the P600 component was linked to 2nd-level belief in plausibility processing. Additionally, the gamma-band (30-40Hz) was influenced by 2nd-level belief in both semantic processing. Specifically, in plausibility processing gamma-band was modulated by 2nd-level precision weighted prediction error (pwPE). Together, these findings indicated dissociation neural effects underlying the processing of predictability and plausibility in dynamic environments, with prediction-driven strategy in predictability processing, but prediction error-driven mechanisms in plausibility processing. Our computational modeling study goes beyond classical ERP analyses, highlighting the hierarchy and flexibility of prediction mechanisms in sentence comprehension in dynamic environments.
Topic Areas: Computational Approaches, Reading