Cortical Coding of Extended Coloured Figures

Abstract

We have studied the neural coding of extended coloured objects. Single cells were recorded from parafoveal regions of V1 and V2 in alert, fixating monkeys. In V1 most of our recordings were from the upper layers. A square figure (usually 4 deg × 4 deg, that is, much larger than the response field) of uniform colour on a gray background was placed at various positions relative to the receptive field of the cell. Orientation and colour were optimised for each cell.

The vast majority of cells (75/89 in V1, 43/47 in V2) responded only when the receptive field was positioned on or near an edge, but not inside the figure (‘edge-only response’). Of the few cells that did respond with the receptive field inside the figure (‘surface response’), only some were colour selective (V1: 6/14, V2: 2/4); that is, the cell response clearly differentiated colours of similar luminance. This lack of surface responses is surprising, since most LGN cells, namely type I and type II cells respond to large uniform stimuli of some colour. On the other hand, many of the ‘edge-only’ cells were colour selective by the same criterion, some responding exclusively within a small volume of colour space. Most ‘edge-only’ cells were selective for orientation, and also many for edge polarity. V1 and V2 may encode colour of extended figures by edge signals rather than an isomorphic representation. In this case, colour filling-in would result from decay of the colour-specific edge signals rather than spreading of the colour-specific surround activity into the figure.