Thesis defense Jakub Szewczyk (Donder series 230)

28 June 2016

Promotor: prof. dr. H. Schriefers

The mechanisms of prediction in language comprehension

This thesis focuses on two topics related to language perception: the role of animacy in language processing, and the mechanisms of language prediction.

Animacy is one of the most basic distinctions in our conceptualization of the world, because it provides rich information about the object perceived. Animacy has also profound consequences for the grammatical and thematic processing of the language. As found in the first study, the language processing system is sensitive to the animate/inanimate distinction, even when animacy does not bear any consequences for the syntactic or thematic interpretation of the sentence. Thus, out of the many possible semantic distinctions, animacy has a special status in language processing.

Two subsequent studies focused on predictions in language processing. Predictions enable comprehenders to preactivate possible upcoming words. Until now, researchers have been able to demonstrate that comprehenders use predictions only in rare situations when only one word can reasonably continue the sentence. In our study, we extended these findings and showed that comprehenders predict whole semantic categories of words, even when no specific word is particularly likely. This discovery considerably broadens the range of situations in which predictions can support comprehension. In the following study, we demonstrated when and how predictions are implemented. By artificially inducing predictions, we showed that predictions can operate only when the syntactic structure of the sentence provides sufficiently strong constraints. This indicates that predictions usually concern the next word or the next phrase. We also showed that predictions are implemented via lexical access, i.e. through the same process that takes places when a word is actually perceived.

The findings of the three studies allowed me to propose a model of how predictions operate in language comprehension. The model distinguishes between active and passive predictions, each having a specific role in language comprehension.