Abstract
Different combinations of depth cues are relevant for different perceptual judgments. For judgments of slant, disparity and texture gradients should be combined in a weighted sum. For judgments of texture homogeneity, the slants specified by disparity and texture should be compared; this can be accomplished by subtracting one from the other. An analogous transformation occurs in color vision where L- and M-cone signals are added in luminance channels and subtracted in color-opponent channels. We asked whether the same occurs with disparity and texture signals. Specifically, are disparity and texture actually combined in a weighted sum for slant estimation and in a subtraction for judging texture homogeneity? And is access to the disparity and texture signals themselves lost in the process? To answer these questions, we presented planes whose slants were defined by disparity and texture gradients. There were three types of trials (conducted in different sessions). 1) 3-interval “oddity”, in which three stimuli were presented, one (or two) at a base slant with no conflict between disparity and texture, and two (or one) with a conflict between disparity and texture (but the same values). Observers indicated the interval containing the “odd” stimulus. 2) 2-interval “slant”, in which two stimuli were presented, one with conflict and one without. Observers indicated the interval containing the greater slant. 3) 2-interval “homogeneity”, in which two stimuli were again presented, one with conflict and one without. Observers indicated the interval containing the texture that was more compressed on one side. The slant and homogeneity thresholds predicted the oddity thresholds. This is consistent with the hypothesis that disparity and texture cues are indeed added to estimate slant and subtracted to estimate texture homogeneity, and that access to the disparity and texture signals themselves is lost in the process
NIH Research Grant R01-EY12851, AFOSR Research Grant F49620, NIH Training Grant F32-EY07043