In 2006, Bruce and I had suggested that the vertical disparity could exert its effects on perception via reducing the effective interocular correlation, rather than being detected and encoded directly. As explained above, we subsequently tested and disproved this idea. However, I now realised that a more sophisticated version of the same basic idea was still viable. In this paper, I simulated the activity of neurons tuned to purely horizontal disparity, and showed that they actually also encode both the magnitude and the sign of vertical disparity. This model is consistent with all psychophysics that I know of. In fact because it works so well, as far as I can see it doesn't predict any characteristic errors of perception and so can't be falsified by psychophysics. I'm currently waiting for the physiologists to tell us whether we do have neurons tuned to a range of vertical disparities at each point in the retina.