Abstract
The responses of disparity sensitive neurons can be understood as operating via two, independently adaptable, channels, which sum and take the difference between left and right eye inputs, reducing signal redundancy. We used this efficient coding approach to investigate the perception of motion-in-depth, where the ability to selectively adapt binocular encoding channels allowed for an examination of Changing Disparity (CD) and Interocular Velocity Difference (IOVD) cues. Efficient coding predicts that perceived speed-in-depth should decrease with differencing channel adaptation and increase with summation channel adaptation, with such adaptation selectively targeting the CD cue. We measured psychometric functions for speed-in-depth discrimination with binocular motion-in-depth stimuli. Discrimination performance was measured at three standard speeds (27, 53 and 107mm/s), comparing stimuli presented above and below a central fixation point in a 2AFC procedure, where participants were asked whether motion was faster in the top or bottom stimulus. To examine the role of summation and differencing channels, participants were presented with correlated and anti-correlated adaptation stimuli. Adaptors were 1/f filtered noise patterns or 1D horizontal noise. Adaptors were presented above and below fixation, with a binocularly anti-correlated adaptor in one location and a correlated adaptor in the other. Adaptor arrangements were counterbalanced across blocks. Anti-correlated adaptors targeted the differencing channel, while correlated adaptors targeted the summation channel. We measured points of subjective equality (PSEs) for cases where the standard speed-in-depth stimulus was in the same location as either the correlated or anti-correlated adaptor. In line with predictions, PSEs were consistent with a reduction in perceived speed-in-depth at the location of the anti-correlated adaptor, and an increase at the location of the correlated adaptor. Adaptation effects decreased with increasing reference speed. We consider these results in terms of the contributions of CD and IOVD cues to motion-in-depth, and the mechanisms governing their integration.