Joonsik Moon¹ (), Peter Bex¹; ¹Northeastern University

The distribution of visual cues across depths and retinal eccentricities changes with gaze direction and distance. However, how the stereoscopic system utilizes these covarying distributions to maintain a stable perception of 3D structure remains underexplored. This study uses classification image analysis with naturalistic images to examine the selectivity of stereoscopic disparity mechanisms across changes in vergence. Three observers distinguished between convex and concave targets over 1,000 to 2,000 trials. Targets were disparity-defined Gaussians (σ = 0.4°) centered ±2° left or right of fixation (compliant within 1°), 100% contrast 1/F noise, on a polarized 3D display (LG 42LM6200, Lmean=75cd/m^2), with gaze compliance enforced with a 2000Hz Eyelink. In alternating blocks of twenty 200msec trials, participants verged at near (40cm) or far (infinity) distances and reported the location of the concave or convex target (at random across subjects). In a training phase, stimulus presentation duration decreased in 5 steps from 10 s to 200 ms. Stereoscopic disparity was controlled by a 3/1 staircase, thresholds from the training phase seeded the test phase. The 2D Fourier transform amplitudes of the interocular difference images were classified based on responses. Classification images contained peaks at high spatial frequencies for vertical and oblique orientations. For near-convergence, oblique orientations were correlated with correct responses, whereas for far-divergence, vertical orientation were correlated with correct responses. This suggests that stereoscopic performance shifts from vertical to oblique content as gaze shifts from far-divergence to near-convergence. These findings provide new insights into the influence of vergence signals in orientation tuning during stereoscopic depth encoding that could be related to ocular torsion during convergence.

Acknowledgements: This research was supported by NIH grant EY029713.