Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Semantic and Social Prediction and Its Effects on Memory
Poster Session C, Thursday, October 1, 10:45 am - 12:45 pm, Wangari Maathai
Yi Hao1, Emmanuel Biau1, Ana Bautista2, Clara Martin2, Julian Pine1, Francesca Branzi1; 1University of Liverpool, UK, 2Basque Center on Cognition, Brain & Language (BCBL), Spain
Background: Semantic and Theory of Mind (ToM) processing are rarely studied together, yet emerging evidence suggests they may share underlying cognitive and neural mechanisms. In this EEG study, we used time–frequency analyses (TFA) to examine whether predictive coding and prediction error rely on similar neural processes across semantic and ToM domains, and whether prediction error modulates subsequent memory formation. Hypotheses: We hypothesised that semantic and ToM predictive coding would rely on shared neurocomputational mechanisms, with alpha–beta oscillations reflecting semantic activation during predictive coding, and theta oscillations reflecting memory updating during prediction error. We further hypothesised that prediction error would enhance episodic memory for word stimuli. Method: Thirty native English speakers took part in an EEG sentence-reading task in which prediction domain (Semantic vs ToM), sentence contextual constraint (high vs. low) and target word congruency (congruent vs. incongruent) were manipulated. Each EEG session was followed 24 hours later by a word recognition task. Results: Across both domains, prediction error elicited alpha-beta desynchronisation and theta synchronisation, which potentially reflect the activation of task-relevant information and representation updating, respectively. The effects of prediction, however, differed across domains. Specifically, alpha–beta effects showed similar spatial distributions but differed in timing and direction, characterised by power decreases in the semantic domain and power increases in the ToM domain. Memory performance showed better recognition for congruent than incongruent words in the semantic domain, but the opposite pattern was observed in the ToM domain. The effect observed in the semantic domain is consistent with the ‘schema-based account’, which proposes that information coherent with prior context or existing knowledge is encoded more efficiently. Importantly, prediction-error-related theta power increases predicted better memory performance for incongruent words in both semantic and ToM domains, suggesting that prediction error processing also facilitates memory encoding.
Topic Areas: Meaning: Discourse and Pragmatics, Meaning: Lexical Semantics