Abstract
Purpose: Neurons in area V2 code for border ownership in displays of a single figure or two overlapping figures (Zhou et al., J Neurosci 20:6594, 2000). To further test the hypothesis of border ownership coding we studied displays of crossed transparent bars (or a shadow across a white bar). If transparency is properly represented, the borders of the intersection should be assigned to the intersecting bars. Because it is unlikely that border ownership assignment is reversed later on in the visual process we conjectured that V2 should resolve configurations of transparent overlay to assign border ownership correctly.
Methods: Single unit responses were recorded in alert monkeys during behaviorally induced fixation. Test stimulus T1 consisted of five squares portraying the crossing of a white bar with a dark bar of 50% transparency (see http://vlab.mb.jhu.edu/vss03fq). The gray of the central square A (the intersection) was the same as the background gray. The border between square A and one of the outer squares, B, was centered in the receptive field. Test stimulus T2 consisted of square B alone (square A was the same as in T1, but now blended with the background). Thus, stimuli T1 and T2 were identical over the region A+B in which the receptive field was centered. To measure side preference, T1 and T2 were flipped about the edge in the receptive field.
Result: Of 76 orientation selective V2 cells, 29 (38%) showed border ownership selectivity for the single square (T2), and 25 of these (86%) changed their responses significantly when T2 was replaced by T1. In 22 cells (76%) the side preference was reversed (the border was assigned to the intersecting bar), in 3 cells it was only reduced.
Conclusion: Neural signals in V2 represent border ownership for displays of simple figures as well as configurations of superimposed transparent objects, or cast shadows on objects. These results suggest that area V2 plays a fundamental role in figure-ground organization.