Simple cells can show non-linear binocular combination

Current models of binocular simple cells assume that input from left and right eyes is initially combined linearly. Model cells of this type underlie the successful energy model of disparity-tuned complex cells (Ohzawa et al., 1990, Science 249:1037). An alternative hypothesis is that input from each eye is passed through a threshold non-linearity prior to binocular combination. This is physiologically entirely plausible - for instance, it would occur if some binocular simple cells receive input via monocular simple cells rather than directly from the LGN - and explains a number of observations. (i) We have shown previously (Read et al., 2000, Abstr. Soc. Neurosci. 26:1845) that using simple cells of this form in the energy model yields disparity tuning curves whose amplitude is reduced when the stimuli are anti-correlated, in agreement with experiment. The energy model predicts the same amplitude for anti-correlated as for correlated. (ii) For some cells, comparing monocular and binocular responses suggests that the input from one eye is always suppressive. This is hard to explain if binocular combination is linear. (iii) If the input from both eyes is excitatory, the pattern of the cell's response to disparate drifting gratings shows a distinctive dependence on interocular phase, changing from one burst of firing per stimulus cycle to two bursts at a particular interocular phase. Observation of such "frequency doubling" would be a strong indication that the postulated cells actually exist, but has never been reported. This may be because no one has looked for it, compounded by the difficulty of discerning such behaviour in data from awake animals. A study of our data revealed several simple cells which showed disparity-dependent frequency-doubling, supporting the view that binocular combination is non-linear in some cases.