Selectivity for orientation and direction of motion of single neurons in cat striate and extrastriate visual cortex

1. We consider the consequences of the orientation selectivity shown by most cortical neurons for the nature of the signals they can convey about the direction of stimulus movement. On theoretical grounds we distinguish component direction selectivity, in which cells are selective for the direction of movement of oriented components of a complex stimulus, from pattern direction selectivity, or selectivity for the overall direction of movement of a pattern irrespective of the directions of its components. We employed a novel test using grating and plaid targets to distinguish these forms of direction selectivity. 2. We studied the responses of 280 cells from the striate cortex and 107 cells from the lateral suprasylvian cortex (LS) to single sinusoidal grating to determine their orientation preference and directional selectivity. We tested 73 of these with sinusoidal plaids, composed of two sinusoidal gratings at different orientations, to study the organization of the directional mechanisms within the receptive field. 3. When tested with single gratings, the directional tuning of 277 oriented cells in area 17 had a mean half width of 20.6°, a mode near 13°, and a range of 3.8-58°. Simple cells were slightly more narrowly tuned than complex cells. The selectivity of LS neurons for the direction of moving gratings is not markedly different form that of neurons in area 17. The mean direction half width was 20.7°. 4. We evaluated the directional selectivity of these neurons by comparing responses to stimuli moved in the optimal direction with those elicited by a stimulus moving in the opposite direction. In area 17 about two-thirds of the neurons responded less than half as well to the non-preferred direction as the preferred direction; two-fifths of the units responded less than one-fifth as well. Complex cells showed a somewhat greater tendency to directional bias than simple cells. LS neurons tended to have stronger directional asymmetries in their response to moving gratings: 83% of LS neurons showed a significant directional asymmetry. 5. Neurons in both areas responded independently to each component of the plaid. Thus cells giving single-lobed directional-tuning curves to gratings showed bilobed plaid tuning curves, with each lobe corresponding to movement in an effective direction by one of the two component gratings within the plaid. The two best directions for the plaids were those at which one or other single grating would have produced an optimal response when presented alone. For simple cells the size of the lobes depended on the relative spatial phase of the component gratings, but this was not the case for complex cells. 6. Although the shape of tuning curves for plaids could be well predicted from responses to gratings, the plaid responses of cells in area 17 were on average one-third smaller than expected. This probably reflects inhibitory processes acting in the orientation domain. This effect was much less pronounced in LS. 7. We tested a few LS neurons for their sensitivity to the orientation of stationary gratings whose contrast was modulated in time. All showed a definite selectivity for the orientation of these targets, which closely matched the same neurons' selectivity for the orientation of moving gratings. 8. We conclude that cells in both area 17 and LS show response patterns to two-dimensional stimuli that are linked to the orientation of the stimulus' spatial components and not to the direction of motion of the whole stimulus pattern. Thus directional selectivity in these neurons appears to be secondary to orientational selectivity.