Neurons in visual area MT are selective for spatial and temporal frequencies of motion stimuli and therefore their properties are characterized in a 3D frequency domain spanned by two spatial (x, y) and temporal (t) frequencies. A full description in the 3D domain, i.e., a 3D spectral receptive field is important, because a subset of MT neurons is tuned for velocity of motion stimuli. A tilted response plane passing through the origin of the domain indicates the velocity tuning, and such a 'velocity plane' cannot be defined in any 2D sections of the 3D domain.

Here, we developed a new subspace mapping method to see full 3D spectral receptive field of MT neurons with a reasonable recording time. Our stimuli consist of 6 superimposed drifting gratings and each has different combinations of spatial and temporal frequencies. Each grating has a short lifetime and is replaced by another one randomly chosen from a pool that covers the 3D frequency domain. We recorded activity of single MT neurons from anesthetized monkeys and applied standard reverse correlation to the spike trains in response to the stimuli.

For some MT neurons, we observed the 'velocity plane' and surrounding suppressive planes that improve the velocity tuning. Other MT neurons showed blob-like response profiles separable for spatial and temporal frequencies. These results are consistent with previous studies; MT neurons are not homogeneous with respect to the velocity tuning, some are tuned for velocity and others are not. Our new method successfully captures 3D spectral receptive field of MT neurons irrespective of being tuned for velocity.