Predictability shapes meaning resolution for lexicosemantic ambiguities

Spoken sentence comprehension requires listeners to integrate information across levels of representation. The predictability of linguistics features, including words and phonemes/forms, has been proposed to facilitate this process. In our MEG/EEG study (N=31), participants heard sentences like “While sailing down the river, she noticed the trees along the bank/shore/cashpoint”. On average, ambiguous words (bank) were most predictable (i.e., highest cloze probability); followed by synonyms (shore); while alternative meaning words (cashpoint) were unpredictable. We wanted to test whether predictability would alter the processing of ambiguous words in context. Research into lexicosemantic ambiguities has shown increased amplitude BOLD responses in fMRI (e.g., Rodd et al. 2005), and increased amplitude ERP/ERF (e.g., MacGregor et al. 2020) for ambiguous words over single-meaning words. Thus, neural responses may be impacted by lexicosemantic uncertainty. We hypothesized an effect of Semantic Congruency – where neural response amplitude increases for anomalous words over non-anomalous words (ambiguities+synonyms) – and Lexicosemantic (Un)Certainty – where amplitude differs between ambiguities and synonyms. Using spatiotemporal cluster-based permutation tests, we observed a significant Semantic Congruency effect: ERP/ERF amplitude was greater for anomalous words, consistent with a canonical N400. One cluster was observed in the Lexicosemantic (Un)Certainty comparison: amplitudes diverged between ambiguities and synonyms. However, this approach cannot separate ambiguity and predictability. Follow-up analyses will use mixed-effects modelling to predict ERP/ERF amplitude in cluster-defined time windows. Overall, the results demonstrate the top-down influence of context and lexical knowledge on word meaning resolution, highlighting the flexible nature of linguistic representations and adaptability of the language system.