Possible neural substrates for binocular rivalry

Binocular rivalry is the perceptual consequence of dichoptic input which is not congruent between both visual inputs. There is some evidence, both theoretical and empirical, that the perception of binocular rivalry is mediated by interactions between binocular neurones, rather than by interactions between monocular neurones. This evidence suggests also a model of perception which predicts binocular rivalry as a consequence of normal interactions between binocular neurones in a retinotopic array. This model accounts for rivalry without postulating any additional interconnections beyond those already thought to exist between binocular neurones simply assumes an orderly mapping of tuning characteristics across groups of cells, as is typically observed in visual cortex. On the basis of this model, and findings already reported, it was hypothesised that binocular rivalry reflects extrastriate rather than area V1 processing (no process so far attributed to area V1 has yet been reported to be affected by binocular rivalry). It was hypothesised also that area V2 was the most likely area in which such processing first arises. Area V2 has been associated with the perception of 'purely subjective contours'. It has been shown that some cells in area V2 are tuned for such contours, which are characterised by the absence of Fourier components at the orientation of the perceived contour, while no cells in area V1 have been found to be similarly sensitive (von der Heydt and Peterhans 1989). This characteristic of area V2 neurones enables purely subjective contours to be used to test the two hypotheses described above. Real contour tilt aftereffects, which are thought to arise in area V1, are not affected by rivalry during their induction. If purely subjective contour tilt aftereffects (Paradiso, Shimojo and Nakayama 1989) are subject to the same types of processing as their real contour counterparts, as suggested by the rationale and model of von der Heydt and Peterhans (1989), interactions between subjective contour tilt aftereffects and binocular rivalry should indicate the role, if any, of area V2 in rivalry. It was found that purely subjective contour tilt aftereffects (Experiment One) and tilt illusions (Experiment Four) exhibit angular functions like those observed for real contour tilt aftereffects and illusions. Just as for real contour effects, these functions can be described in terms of direct effects (Experiment Two) and indirect effects (Experiment Three), suggesting purely subjective contours are processed as if they were real contours. Unlike real contour direct effects, purely subjective contour direct and indirect effects are reduced in magnitude by periods of rivalry during their induction (Experiment Five). In keeping with their suggested extrastriate locus (eg. Wenderoth, van der Zwan and Williams 1993), the magnitude of a real contour indirect effect is also reduced by periods of rivalry occurring during its induction (Experiment Six). These results suggest that rivalry does arise first in area V2. If this is true then complete interocular transfer of the purely subjective aftereffect, induced with or without rivalry, should occur because area V2 is almost exclusively binocular. This proved not to be the case, however, suggesting the ocular dominance observed in most binocular cells has to be taken into account in any explanation of rivalry (Experiment Seven). This was tested using real contours and found to be the case. These last results suggested also that rivalrous interactions occur between groups of binocular neurones only in extrastriate cortex (Experiment Eight). This hypothesis was tested by examining the effect of binocular rivalry on the duration of the plaid motion aftereffect, which is thought to arise no earlier than area MT, a visual cortical area which is also thought to be almost exclusively binocular. It was found that rivalry did reduce the duration of plaid motion aftereffects but not linear motion aftereffects, and that the impact of rivalry might be linked to plaid sensitive cells in area MT, although this last conclusion is tenuous (Experiments Nine and Ten). Finally, it was shown also that the magnitude of the reduction in duration of the aftereffect was proportional to the predominance of the plaid stimulus during rivalry, a finding which supports the mechanism of rivalry suggested by the binocular model. The results together suggest that binocular rivalry does arise through binocular interactions, but that such interactions cannot be attributed to a single cortical area. All groups of binocular neurones may be subject to the processes that ultimately give rise to the perception of rivalry, a conclusion which does not invalidate the binocular model of rivalry. This has some consequences for binocular vision, particularly stereopsis, which might occur qualitatively during binocular rivalry