Paper: Word surprisal predicts N400 amplitude during reading

ACL ID P13-2152
Title Word surprisal predicts N400 amplitude during reading
Venue Annual Meeting of the Association of Computational Linguistics
Session Short Paper
Year 2013

We investigated the effect of word sur- prisal on the EEG signal during sen- tence reading. On each word of 205 ex- perimental sentences, surprisal was esti- mated by three types of language model: Markov models, probabilistic phrase- structure grammars, and recurrent neu- ral networks. Four event-related poten- tial components were extracted from the EEG of 24 readers of the same sentences. Surprisal estimates under each model type formed a significant predictor of the am- plitude of the N400 component only, with more surprising words resulting in more negative N400s. This effect was mostly due to content words. These findings provide support for surprisal as a gener- ally applicable measure of processing dif- ficulty during language comprehension.