Abstract
Many visual attributes appear to be encoded by a norm-based code in which the stimulus is represented by how it deviates from an average or neutral value (the norm). One prediction of these codes is that neural responses should be weaker for stimuli near the norm and stronger for uncharacteristic stimuli, a pattern that has been shown for a variety of stimulus dimensions. For example, BOLD responses are weaker for average than distinctive colors, faces, or voices. We used rapid event-related fMRI to determine whether a similar pattern of BOLD responses occurs in visual cortex for image blur, which behaves functionally like a norm-based dimension representing blurred or sharpened variations relative to a norm of "in focus." We presented observers with achromatic images of natural scenes, filtered by varying the slope of the log amplitude spectra from -1 (strongly blurred) to +1 (strongly sharpened) relative to the original slope. RMS contrast was equated to the original after filtering. In V1 there was higher activation for the in focus images than for the sharpened or blurred images. Similar patterns were observed in extra-striate areas, but to a lesser degree. The higher overall activation in V1 and other visual areas for focused images is inconsistent with an explicit norm-based coding for image focus, and could imply an implicit neural signature (e.g. distribution of responses across frequency-tuned channels) or one that occurs subsequent to retinotopic visual cortex. We also performed a whole-brain GLM to search for regions showing BOLD responses consistent with an explicit norm-based model. The results of this analysis suggest that, although norm-based coding was not evident in retinotopic cortex, it may nevertheless be instantiated in higher cortical areas.
Meeting abstract presented at VSS 2014