Skip to main content

Language-Processing by Overlapping Predictions: A Predictive Coding Approach

Final Report Summary - L-POP (Language-Processing by Overlapping Predictions: A Predictive Coding Approach)

The ERC project Language Processing by Overlapping Predictions (L-POP) has used functional brain imaging methods in humans to investigate the neurophysiological mechanisms underlying predictive language processing. We have shown that already before perceiving a highly expected word, the brain represents information about its frequency of occurrence, about whether refers to an animate or an inanimate entity, and whether it represents an abstract or a concrete concept. In other words, we have provided evidence for the pre-activation of semantic information during the on-line processing of language. We have also demonstrated that such predictive processing affects the earliest levels of language perception in a top-down manner, in the sense that less bottom-up information about the perceived stimulus is encoded by the brain when a word is expected. These results provide evidence in support of the general hypothesis that our brain processes language in a pro-active way, i.e. by using general as well as contextual knowledge to predict upcoming language input at multiple interacting levels of linguistic representation. This helps to understand the efficiency of language processing.