Goris et al (SfN, 2012) examined the organization of spatiotemporal frequency selectivity as revealed by MT neuron responses to sinusoidal gratings whose drift direction varied with drift rate either held constant (frequency separable organization) or varying along the preferred velocity plane (velocity separable). Some MT neurons' tuning was better fit by a frequency separable model, while others were better fit by a velocity separable model. There was no clear relationship between the degree of pattern selectivity and the degree to which the velocity separable model better described the neuron's tuning.

Because MT neurons can respond differently to superpositions of gratings than expected from the responses to the component gratings measured in isolation, we extended this study to include plaids (sums of two gratings with orientation differing by 120deg). Frequency separable plaids were constructed from gratings moving at the preferred drift rate of the cell. In contrast, the drift rates of components of velocity separable plaids were chosen to lie in the preferred velocity plane, as in the single grating experiment. MT responses to velocity separable plaids were significantly stronger than those to frequency separable plaids. Moreover, pattern cells had a broader bandwidth on the velocity plane than component cells. Both of these observations are consistent with predictions of the velocity separable model. We conclude that use of stimuli organized in a velocity separable coordinate system allows for a more refined dissection of the relationship between pattern selectivity and the degree to which spatiotemporal selectivity is organized with respect to the preferred velocity plane.