The role of spatial frequency tuned channels in vergence control

Marr and Poggio’s (1979) computational theory of human stereo vision is based on the idea that disparities of different magnitudes are processed independently by differently-tuned spatial frequency channels (Felton, Richards and Smith, 1972). The theory holds that large disparities are dealt with by low spatial frequency units, with vergence movements initiated to bring into correspondence high spatial frequency channels dealing with small disparities. By limiting the size of the disparity that can be processed by any one channel to around plus/minus the width of the central region of the receptive fields of the units comprising the channel, the theory avoids the problems of having to choose between many ambiguous disparity matches. A question of some importance to the theory is how vergence movements are triggered to bring into alignment the high spatial frequency units. Marr and Poggio consider various possibilities, the most natural being to have vergence changes driven directly from the outputs of low-spatial-frequency/largedisparity units. Accordingly, we conducted an investigation of vergence movements made in response to narrow-band filtered disparate textures. Our main result is that even relatively high spatial frequencies can initiate speedy vergence shifts for disparities as large as 28’. We conclude that the triggering of vergence shifts cannot be strictly limited to low spatial frequency channels, at least for the disparity range we used.