Abstract
Our capacity to recognize familiar objects is mediated by both bottom up sensory information and top-down knowledge. Prior knowledge of the category of a to-be-presented stimulus decreases recognition time, and fMRI studies have shown that anticipation of a stimulus increases brain activity in corresponding category selective regions of human inferior temporal cortex. In the present study, we used magnetoencephalography (MEG) to examine the effect of top-down knowledge on the temporal dynamics of object recognition. Thirty-two participants viewed a set of objects with categorical structure (humans and animals) while their brain activity was recorded with MEG. Participants were asked to categorize the stimuli as "human" or "animal" as quickly and accurately as possible. Prior to stimuli presentation, they were given a cue that would correctly identify object category in 80% of the trials (80/20, cue valid/cue invalid). Pattern analysis was used to decode both "anticipated" stimulus category during the cue period and actual stimulus category during the stimulus period. Behaviorally, we observed higher accuracy (p<.001) and lower reaction times (p<.001) for validly cued trials. In the stimulus period, we observed higher decoding accuracy for validly cued trials (p<.001), and a negative correlation between decoding accuracy and reaction time (ρ = -.622; p<.001). During the cue period, prior to stimulus onset, we could decode the "anticipated" category, and we also observed a significant negative correlation (ρ = -.604; p<.001) between decoding performance and future reaction time performance, thus indicating that greater anticipatory neural activity produces faster reaction times. Our general findings show that category expectations activate object brain representations that in turn facilitate perception and behavior. Further analysis suggests that the brain generates an average exemplar of a category during the cue period to prime the category representation.
Meeting abstract presented at VSS 2013