Integrative and predictive processes in text reading: the N400 across a sentence boundary

During reading, words that are congruent with the prior text’s meaning can help update the reader’s mental model of the text. Although prediction is viewed as a major driver of meaning congruence, reflected by the N400, we hypothesised that at the beginnings of sentences, memory-based integration is the dominant mechanism for mental model updating. We measured integrability separate from predictability, and constructed two-sentence texts that varied them independently. Experiment 1 demonstrated that our materials could show a predictability effect in a memory probe task following a rapid serial visual presentation sentence presentation. In Experiment 2 we recorded ERPs on a critical word across a sentence boundary. Critical words that were integrable with the prior text produced reduced N400s compared to baseline. Word predictability had no additional effect on the N400. The results suggest that memory-based word-to-text integration is the default mechanism in updating a text’s meaning, with word-specific prediction playing a minor role.