Representation of visual feature conjunctions in the superior parietal lobule

Abstract

Color and form are initially processed in parallel in separate units of the visual system. Feature binding describes the process that integrates the distributed information of various feature dimensions which occur at one location. While patient studies suggested that posterior parietal cortex is necessary for binding of features from different visual dimensions, imaging evidence has been inconclusive in the past. One major problem was that the comparison of visual search for feature conjunctions is more difficult than single feature search. To circumvent this confound, we compared the processing of features and feature conjunctions in an identical search task. Functional data of the the occipital and parietal cortex were acquired in 1.4 mm isotropic voxels using a 7T MR-scanner. As visual search displays, configurations of five Gabor patches were briefly presented to the subjects. These configurations differed in the combination of color and spatial frequency on fixed radially arranged positions. Two values per dimension - red and green, high and low spatial frequency - formed four specific Gabor patch types. The twelve subjects performed a singleton-conjunction search task during brain imaging. This task demands the integration of both feature dimensions at a specific location. In a multivariate pattern analysis a support vector machine was applied on the BOLD signal using a searchlight algorithm. Individual accuracy maps for classification of displays were obtained which just differ in their colors, spatial frequencies or the conjunction of both feature dimensions. In the group analysis we found superior parietal and parieto-occipital regions which encode single feature dimensions. One area of the left superior parietal lobule (MNI-coordinates × −26, y −54, z 68) shows superadditive classification accuracy for feature conjunctions compared to the individual feature classification accuracies. We conclude that this area is involved in the representation of 'true' conjunctions of visual feature dimensions.