台大心理系

How does the human brain so rapidly and reliably link complex perceptual stimuli, such as words, to the information stored in long-term memory that constitutes the "meaning" of those words? Electrophysiological data suggest that meaning is accessed through a stimulus-elicited, temporally-delimited process that binds neural activity across a distributed, multimodal brain network. In contrast to long-standing views that treat the recognition of words (and other meaningful stimuli) as occurring primarily through feed-forward processing that it relatively impervious to context, our work has shown that the language comprehension system uses context information to predict semantic and even perceptual features of likely upcoming words. The fact that some information may be available to the brain even before the word actually appears is likely an important part of what allows meaning processing to be as fast and effective as it usually is. However, because not all information can be made available in time, additional processes come online after this initial semantic binding to help deal with unexpected words that violate predictions. Moreover, in cases of ambiguity, processes are recruited to help select context appropriate meanings, and the success of these processes impacts downstream processing, as revealed both in ERPs and eyetracking measures of natural reading. Overall, our research has revealed that there is a "time for meaning".