Influence of luminance gradient reversal on complex cells in feline striate cortex

The effects of reversing the polarity of luminance contrast in adjacent segments of a bar stimulus have been investigated in complex cells of area 17 in cats lightly anaesthetized with nitrous oxide/oxygen halothane mixtures. On the basis of length summation behaviour, complex cells were classified as standard (length summating) or special (optimum response to a bar much shorter than the receptive field), after Gilbert (1977), and were further subdivided into groups lacking and possessing end-stopping. For each type of complex cell, we measured the effects of adding short segments of one polarity of contrast (light or dark) to either end of a bar of fixed length and optimum orientation but of opposite contrast. In all cells the response to the central bar was depressed by short segments of reversed contrast to an extent greater than predicted from the cells' length summation characteristics. Responses were minimized or abolished at a critical segment length. Increases beyond the critical length elicited a progressive recovery in response to a plateau level. In end-stopped cells this was followed by a further decline in response up to the limits of the cells' inhibitory end-zones. Special and standard complex cells differed only in their susceptibility to reversed-contrast segments above the critical length. In standard complex cells, the recovery of response matched the cells' length summation profiles in slope and cut-off point. In special complex cells the recovery was flatter in slope and significantly more protracted than the length summation profile. Similar results were obtained for either direction of motion (orthogonal to a cell's optimum orientation) and for either polarity of contrast (dark centre, light ends or the reverse). As might be expected, all the effects were weighted in favour of the receptive field centre. Thus the upturn in response as reversed-contrast segments were progressively extended was more rapid and achieved a higher limiting level when the central bar was short. Merely interrupting the contours of a bar by a central gap, rather than a segment of reversed contrast, gave rise to no more attenuation of response than that predictable from the length summation curve. The results are compared and contrasted with our comparable data for simple cells.