It is well established that stereoacuity for conventional (1st-order) stimuli improves with increasing contrast, with slopes ranging from −0.5 to −1.0 on log - log axes (Halpern and Blake, 1988 Perception 17 483 – 495; Legge and Gu, 1989 Vision Research 29 989 – 1004). In the experiments reported here we used a variety of stimuli (Gabor patches, amplitude-modulated gratings, and 1-D noise patches) and tasks (stereoacuity and Dmax) to determine if 2nd-order stereopsis shows a similar square-root dependence, or if the contrast relationship is relatively flat as observed for monocular localisation tasks (Hess and Holliday, 1992 Vision Research 32 1085 – 1097). Where possible, stereopsis was assessed with the use of the same stimuli for both 1st-order and 2nd-order stereopsis to ensure that the resulting data would be comparable. In all cases where stereopsis was based on 2nd-order processing there was little or no influence of contrast on performance; the average slope was approximately −0.15. In comparison, more typical slopes near −0.75 were obtained under 1st-order test conditions.
The indifference of 2nd-order depth perception to changes in stimulus contrast not only supports the proposed distinctiveness of the two types of processing but places constraints on theoretical and computational models of 1st-order and 2nd-order stereopsis. Further, these data support the proposal that the same nonlinear position estimate is used for monocular localisation of nonabutting targets, and for 2nd-order stereopsis.
