Descriptions
Much previous research has established that listeners and readers routinely generate context-based predictions that constrain perception and interpretation of language, but the form of this top-down/bottom-up interaction is still hotly debated. One critical question is how far down the processing hierarchy predictions are propagated—e.g. if ‘I heard a dog…’ predicts the word ‘bark’, does this modulate neural responses in units that represent lower-level speech sounds and acoustic features, as well as higher-level semantic units?