If the model of a V1 receptive field as a Gabor filter or edge detector is reasonably accurate, sinusoidal gratings and bars are among the best possible stimuli for such a receptive field. One of the more difficult problems in vision research has been trying to build on models of V1 receptive fields. Within V1, we can attempt to model the responses of V1 neurons to more elaborate stimuli based on the Gabor filter model. In extrastriate cortex, we can examine emergent properties and try to determine how the outputs of V1 neurons are combined to produce them. In both cases, these questions are focused on determining the circuitry which underlies the response properties of visual cortical neurons.
We measured the responses of neurons in macaque V1 to dynamic, translational Glass patterns as a function of dot separation and dot-pair orientation. We computed the expected responses for a receptive field model to translational Glass patterns, and found that the complexity of our V1 tuning curves could be understood in terms of the responses of linear filters to pairs of dots. This modeling connects our understanding of V1 receptive fields as rectified, quasi-linear filters with results from psychophysical studies of Glass patterns. These results provide a basis for studying how subsequent visual areas integrate weak, local signals into global form percepts.
In macaque area V2, we attempted to replicate and expand upon our results from V1. We found that V2 neurons, like those in V1, were selective for the orientation of the dot pairs in a translational Glass pattern, and this selectivity depended on the dot separation. We then devised a stimulus that presented an optimal translational Glass pattern to the CRF, surrounded by different global forms (concentric or radial). We compared responses of neurons to this stimulus with a translational Glass pattern control. On average, cells responded slightly less to the global form stimuli, but this difference was small. There did not appear to be a significant effect of the global form under the conditions we tested. We suspect that exploration of larger receptive fields in higher cortical areas might reveal neurons that are selective for the global forms present in some concentric and radial patterns.
We also explored the responses of neurons in V1 and MT to plaid stimuli, consisting of the addition of two overlapping sinusoidal gratings. In V1, the response of a neuron to a grating at the preferred orientation can be reduced by the superposition of an orthogonal mask grating or by placing a parallel stimulus outside the CRF. These effects are called cross-orientation suppression and surround suppression, respectively. We found that the timing of cross-orientation suppression is very fast, often acting on the cell even before the excitatory response onset to a preferred stimulus. In contrast, the timing of suppressive signals from the surround tend to show some delay. This difference in timing leads us to conclude that these two suppressive signals come from separate mechanisms.
Also using plaid stimuli, we recorded from neurons in macaque area MT. We collected responses to many repeats of a dynamic stimulus in order to get a finely detailed view of the response over time. We divided cells into three categories (PDS, CDS and unclassified) and considered their responses separately. PDS and CDS cells tended to show different characteristics - PDS cells had longer latencies and their characteristic response evolved over a longer period of time. These two effects cause the population motion response of MT to evolve after motion onset. The initial portion of the response is predominantly driven by component motion, and only after some tens of milliseconds does a reliable response to pattern motion emerge. These results suggest that the neuronal circuit for a PDS neuron involves more computation than the circuit for a CDS neuron. They may also account for psychophysical observations indicating a transition from component-dominated to pattern-dominated percepts in the period immediately after motion onset.
We have used global form stimuli and pattern stimuli to test receptive field structure in V1 and V2 and the timing of cross-orientation suppression and pattern motion computation. In using Glass patterns in V1 and V2, we have made some progress toward an understanding of how visual cortex combines weak, local form signals to generate a percept of global form. Our use of dynamic stimuli to examine the timing of suppressive phenomena in V1 and pattern motion computation in MT has given us insight into the circuitry which underlies these effects. We hope that the further development of these two concepts will prove fruitful in uncovering some of the processes involved in visual perception.