Onset, persistence and reset of border ownership signals

One of the many remarkable features of our visual system is constancy, or the stability of our perception of the visual world. We have studied the neural representation of figure-ground structure. It is known that signals in visual cortex swiftly decay after the stimulus is turned off. Is figure-ground organization retained after the figure-ground cues are removed? Cells in macaque area V2 respond differentially to an edge depending on which side of the edge is the figure and which side is the ground (border ownership coding). We recorded responses of V2 neurons in situations where a figure-ground display changed to an ambiguous display. (1) A contrast defined square with the RF centered on one edge (giving a particular border ownership  assignment) was presented for 0.5s and followed by 0.5s presentation of an ambiguous edge - a contrast border dividing a circular field in half. We compared this with the presentation of the same initial display (a contrast defined square) which was then switched to a display with the figure on the opposite side. In both cases the local contrast in and around the receptive field remained constant. In this way we could measure the duration of the border ownership  signal (the difference in firing rates between conditions with the two initial sides of figure) when the figure-ground cues changed in different ways. We found that during the ambiguous edge presentation, the border ownership  signal decayed slowly, with a time constant of about 400 ms. In contrast, when the figure flipped to the other side, the signal decayed quickly, with a time constant as small as 20 ms. (2) To control for a possible effect of an afterimage, we also tested the ambiguous edge after presenting a flickering display in which the colors of figure and ground were alternated rapidly, thus preventing or reducing the formation of an afterimage. The border ownership  signal again showed a slow decay (time constant 590 ms), indicating that the persistence of the border ownership  signal was not due to an afterimage. (3) When the display was switched to 'figure' again after the ambiguous edge period, the border ownership  signal assumed the corresponding value with the same short time constant as observed for figures flipping sides. In summary, in the absence of a cue, the border ownership  signal persists with slow decay, but it is reset quickly when new figure-ground information is presented. These features are interesting as they indicate a mechanism of short term storage. We speculate that this mechanism contributes to the stability of figure-ground organization despite the rapid fluctuations that characterize the retinal image in normal vision. PDF