Unlocked edges or contradictory edges? Is there a master signal for edge locking?

In 1977, when Richard Gregory published his classic paper on equiluminance, it was conventional to model the visual system with two chromatic channels and one 'luminance' channel. The signals of the long-(L) and middle-wave (M) cones were summed in the putative luminance channel; and so this channel would be silent at an edge between surfaces of different colour but equal luminance. There was then no edge signal to corral the more diffuse signals carried by the chromatic channels. Gregory suggested that this loss of edge-locking led to the several curious features of vision at equiluminance. The passage of 30 years has revealed a much larger number of parallel channels within the visual system. At least fifteen morphologically and functionally distinct types of ganglion cell are now known (Dacey et a1 2003; Petrusca et a1 2007), each extracting different combinations of cone signals. Each type tessellates almost the whole retina and has its own specific projections within the brain. These different cells vary vastly in the extent of their dendrites and thus, presumably, in the sizes of their receptive field. On the one hand, this means that Gregory's question becomes a more general one: Is there one channel, carrying a spatially precise signal, that coordinates the signals representing other stimulus attributes, such as colour, flicker, and texture? On the other hand, it-has become unlikely that all the types of chromatically non-opponent ganglion cell will look on in silence when an equiluminant redlgreen edge is swept across the receptor array (Mollon 1980). First, the non-chromatic cells may vary in their equiluminous points. Second, their responses may not be linear: transients signalled by different classes of cone may be transmitted without complete cancellation (see eg Cavanagh 1991; Mollon 1982). Certainly, the parasol cells-traditionally taken as the substrate for the luminance channel-give a frequency-doubled response to a red-green grating traversing their receptive field, and there is no relative intensity of the component colours at which the cell is silenced (Lee et a1 1989). Only one special subset of equiluminous edges is invisible to parasol cells: these are tritan edges, those between chromaticities that give identical quantum catches in the L and M cones and differ only in the shortwave cone signal (Tansley and Boynton 1976). The midget ganglion cells might be thought to be good candidates for carrying Gregory's master signal, since they have the smallest receptive fields. In the fovea, the midget cells-certainly …