Competition and Prediction in Word Processing: MEG Studies of Visual and Auditory Word Recognition

Abstract: Recent experimental evidence supports the view that entropy (uncertainty) over the representations consistent with the linguistic input and surprisal of processed input relative to this entropy drives brain responses in language processing – not the competition between representations consistent with the input. Thus, for example, high cohort entropy (and thus high competition among members of a cohort, i.e., the words consistent with auditory input up to the current point in time) correlates with less neural activity, rather than more neural activity. Morphological structure interacts with morpheme-cohort effects in ways consistent with a full decomposition view of complex word recognition. The emerging picture allows us to use neurolinguistic evidence to test certain representational claims in the linguistic literature, for example the claim that all nouns and verbs are morphologically complex. Thus theoretical claims associated, e.g., with Distributed Morphology, can guide neurolinguistic experimentation, in line with a more integrative approach to Linguistics.