Abstract
Pattern-selective (PDS) neurons in macaque area MT signal the true direction of motion of plaids made by summing two component gratings drifting in different directions; component-selective (CDS) neurons respond independently to the two components. To learn more about the mechanisms responsible for this pattern motion computation, we studied the spatial and temporal limits over which signals from the two component gratings can be combined.
We recorded single units from area MT in opiate-anesthetized, paralyzed macaques. We estimated the temporal precision of the pattern computation by presenting the two component gratings in alternating time segments. We presented stimuli on a high-resolution monitor at a frame rate of 120 Hz, and varied alternation frequency from 7.5 Hz to 60 Hz. As expected, PDS neurons lost their pattern selectivity as we lowered the alternation frequency. The frequency at which PDS neurons lost pattern direction selectivity on average was between 20 and 30 Hz, corresponding to presentation durations of 16.7 – 25 ms. We made analogous measurements in the space domain by presenting the two drifting gratings in alternating spatial partitions. Consistent with previous results, when each partition covered half the receptive field, PDS neurons lost their pattern selectivity. As we increased the number of partitions to 4 or more within the receptive field, pattern selectivity returned. The scale of the partitions at the transition was roughly 1.5 times the neuron's optimal spatial period.
We conclude that the pattern motion computation in MT neurons occurs at a relatively fast time scale and small spatial scale; in other words, for pattern motion to be signaled, the component patterns must be closer to one another in space and time than the spatial and temporal scale of the receptive field itself.
This work was supported by NEI Grant EY07158-06.