Pooling signals from vertically and non-vertically orientation-tuned disparity mechanisms in human stereopsis.

To understand the role that orientation-tuned disparity-sensitive mechanisms play in the perception of stereoscopic depth, we measured stereothresholds using two sets of random-dot stimuli that produce identical stimulation of disparity mechanisms tuned to vertical orientation but dissimilar stimulation of disparity mechanisms tuned to non-vertical orientations. Either 1 or 1.5D of astigmatic blur was simulated in the random-dot images presented to both eyes, using two axis configurations. In the parallel-axis conditions, the axis of simulated astigmatic blur was same in the two eyes (0, 45 or 135 o[rientation]deg). In the orthogonal-axis conditions, the axes of astigmatic blur were orthogonal in the two eyes (LE: 180, RE: 90; LE: 90, RE: 180; LE: 45, RE: 135; and LE: 135, RE: 45). Whereas the stimulation of disparity mechanisms tuned to near-vertical orientations should be similar in the oblique parallel- and orthogonal-axis conditions, the stimulation of non-vertically tuned disparity mechanisms should be dissimilar. Measured stereothresholds were higher in the orthogonal compared to the parallel-axis condition by factors of approximately 2 and 5, for 1 and 1.5D of simulated oblique astigmatic blur, respectively. Further, for comparable magnitudes of simulated astigmatic blur, stereothresholds in the (LE: 180, RE: 90 and LE: 90, RE: 180) conditions were similar to those in the (LE: 45, RE: 135 and LE: 135, RE: 45) conditions. These results suggest that the computation of horizontal disparity includes substantial contributions from disparity mechanisms tuned to non-vertical orientations. Simulations using a modified version of a disparity-energy model [Qian, N., & Zhu, Y. (1997). Physiological computation of binocular disparity. Vision Research, 37, 1811-1827], show (1) that pooling across disparity mechanisms tuned to vertical and non-vertical orientations is required to account for our data and (2) that this pooling can provide the spatial resolution needed to encode spatially changing horizontal disparities.