Abstract
There is convincing evidence that neurons in the medial temporal area of the macaque are involved in, even responsible for, the analysis and perception of motion. The underlying mechanisms, however, remain unclear. In psychophysical and computational studies, the reverse-phi phenomenon has been used as an argument in favour of Fourier energy models: when a rightward stepping bar reverses its contrast sign on every step, it is perceived to move to the left. This is interpreted as an argument in favour of energy models, because the contrast-reversal introduces Fourier components that move in the reverse direction. We applied this logic to individual MT cells to determine to what extent they can be considered Fourier energy detectors. We recorded from 81 MT cells in three monkeys that fixated a central dot while a phi or reverse-phi sine-wave grating was positioned in the cell's receptive field. Consistent with a Fourier energy model, the majority of cells reversed their preferred direction for the reverse-phi stimulus. We then recorded from a monkey that was rewarded for reporting the phi direction of motion for phi motion, and rewarded randomly on interleaved trials with reverse-phi motion. Both types of trials used quarter duty-cycle, square wave gratings. The monkey's behavioural responses in reverse-phi trials were predominantly in the reverse-phi direction, suggesting that his perceptual experience was similar to that of human subjects. ROC analysis of the neural responses recorded from 50 MT cells in this monkey showed that an ideal observer using these cells would report a reversed direction of motion for reverse-phi stimuli. The trial-by-trial covariation of neural responses and decisions implied that about one-third of the cells were significantly involved in the animal's directional decision process. Taken together, these findings give strong support to the claim that MT cells are well modelled as Fourier motion energy detectors.