Abstract
While much is known about the processing of features in visual cortex, less is known about how separate features are bound together. Many theories of feature binding have proposed a role for feedback processing in visual cortex. One function of reentrant processing may be to confirm the correct binding of features. The initial feedforward representation may be sufficient for feature identification, but correct binding may require feedback. To test this theory, we used a form of masking, Object Substitution Masking, thought to selectively disrupt feedback processing in visual cortex. Subjects (n=10) viewed an array of six elements for 75 milliseconds. Each element was composed of a pair of crossed vertical and horizontal bars. One of the bars was white and the other was red, green, or blue. The orientations of the white and colored bars were chosen randomly for each item. The target item was cued by four small dots surrounding the crossed bars. In the unmasked condition the dots disappeared with the target, but in the masked condition the dots persisted for 300 milliseconds after the target disappeared. Subjects' task was to identify the color and orientation of the colored bar in the target item. In this task, identifying the target bar's orientation requires correct feature binding, but identifying its color does not. The mask had a small but non-significant effect on the color judgment; subjects were 99% correct in the unmasked condition and 94% correct in the masked condition. On the other hand, the mask had a large effect on the orientation judgment; subjects were 89% correct in the unmasked condition and 64% correct in the masked condition (p[[lt]]0.001). These results suggest that feature identification may be possible without reentry, but the binding of visual features relies on feedback processing.