Neural dynamics and interactions in top-down facilitation of visual cognition

Traditional theories have described perception as being a reactive process where the brain passively receives sensory stimuli. These 'bottom-up' theories suggest that perception is accomplished by analyzing a stimulus in increasing complexity. In contrast, recent theoretical and experimental studies have demonstrated that perception is an active process in which the brain anticipates the nature of incoming information. These 'topdown' models of cognition posit that predictions about the input are formed and that these predictions are used to guide and facilitate construction of the mental representation of the stimulus.; In top-down processing, relatively abstract, 'higher level' predictions assist stimulus-driven, 'lower level' processing to analyze information more rapidly and efficiently. In object recognition top-down predictions are derived from rapidly extracted low spatial frequency (LSF) gist information that is projected to higher level regions in the orbitofrontal cortex (OFC). The predictions derived from gist information are then sent to lower level regions in the fusiform cortex where they are matched and integrated with sensory information. Neurally, the matching process is reflected in coupling between the OFC and fusiform, with greater coupling indicating greater top-down facilitation of object recognition. This process of top-down/bottom-up integration results in a final refined neural representation of the input stimulus as the output of the recognition process. Experimental results support this framework of how top-down facilitation demonstrating that recognition related MEG activity in the OFC occurs 50 ms prior to the fusiform activity. Furthermore, synchrony between the OFC and fusiform is diagnostic of recognition. Finally, both the MEG activity in the OFC and the coupling between the OFC and temporal cortex are driven by LSFs in the input image.; Interestingly, with a few physiologically plausible assumptions, similar refining mechanisms involved in object recognition also describe how the long-term learning of perceptual representations occurs. Experimental findings demonstrate that while MEG activity in both higher regions in the prefrontal cortex and lower regions in the temporal cortex decreases, coupling between these regions is enhanced. This result suggests a mechanism whereby repetition-priming leads to stronger functional connections between cortical regions, which results in both improved task performance and neural response reduction.