Boutonnet, B. 1, 2 , Meelen, M. 1 , Cheng, L. 1 & Schiller, N. 1, 2
1 Leiden University Centre for Linguistics
2 Leiden Institute for Brain and Cognition
Brains are nowadays commonly seen as prediction machines rather than passive information processors. Prediction in language comprehension has been demonstrated in a number of studies, which show that when processing highly predictive sentences, word anticipation can be measured several constituents in advance. One crucial unanswered question, albeit relevant in today?s multilingual contexts, is whether the brain predicts as strongly when the signal is potentially less reliable (i.e. coming from a non-native speaker) since committing to strong expectations may lead to increased processing demands when those are not met. A recent study shows that integrating unexpected words is less costly in foreign-accented speech, as reflected by weaker modulations of the N400.
In the present study, we presented Dutch-speakers both native- and foreign-accented sentences containing a highly expected noun. While we replicated classic findings of word anticipation for native sentences, we observed that the same speakers had lower expectations when processing foreign-accented speech (as indicated by modulations of the determiner time-locked Phonological Mismatch Negativity) suggesting that the previously measured word integration flexibility may be subserved by weaker predictions. Our findings have implications for higher-education contexts where it is growingly common for complex information to be delivered and integrated by non-native language users.