Abstract
The neural representation of an object is constructed hierarchically. Elementary features are detected in early visual cortex and are progressively integrated throughout higher levels of processing. This hierarchical computation should be reflected in interactions between lower and higher visual areas. Furthermore, disrupting feature integration should reduce these neural interactions. We used letter crowding to disrupt feature integration, testing whether crowding affects the interactions between visual cortical areas. We used fMRI to measure neural activity in multiple visual areas while observers viewed closely-spaced letters (8o eccentricity, presented at 1 Hz for blocks of 15–21 s, separated by 15–21 s blocks of no stimulation). Adjacent letters were displayed in alternation in the uncrowded condition and simultaneously in the crowded condition. We performed a control psychophysics experiment to confirm that letter identification was impaired by crowding under these stimulus conditions. However, during the fMRI experiment, observers performed a demanding contrast discrimination task at fixation to ensure that attention was diverted from the letter stimuli. In each observer, we defined sub-regions in retinotopic visual areas that corresponded to the letter locations, and the visual word-form area (VWFA) in inferotemporal cortex. For each area, we removed the mean response using orthogonal projection, and then computed pairwise correlations between the residuals for both conditions. We found that crowding reduced correlations between several pairs of visual areas. This effect was particularly strong between retinotopic visual areas (V1, V2, V4) and VWFA. This effect was eliminated in a control experiment using Gabor patches, which are elementary features that do require feature integration. Crowding reduces correlations between early visual areas and higher visual areas, presumably by disrupting feature integration. We conclude that interactions between early feature-selective areas and higher object-selective areas reflect the feature integration process.