Hyper-speed meaning and form predictions: An EEG-based representational similarity analysis

Abstract

Language comprehension involves predictive processing, in which comprehenders anticipate both semantic and form-related attributes of upcoming words. This predictive mechanism is crucial as it enables efficient real-time language processing, allowing listeners and readers to keep pace with rapid information streams and quickly correct potential errors. Using electroencephalography (EEG) and representational similarity analysis (RSA), we investigated whether predictions follow a hierarchical, top-down process or occur in parallel, facilitated by associative mechanisms. In this study, native Spanish-speaking undergraduate students read highly constrained sentences designed to elicit specific target words. RSA was applied to evaluate the similarity between all possible pairs of expected words, and signals were classified into semantic, form-related, or specific-word effects based on their relationship to the expected word. The results revealed a rapid transition between effects, with semantic predictions consistently preceding form-related and specific-word predictions. While the sequential order aligns with hierarchical processing, this rapid predictive transition is better understood in the context of associative mechanisms and predictive coding.