A stereo advantage in generalizing over rotations in depth on a same-different successive matching task

Abstract

PURPOSE. To determine whether there is a stereo advantage in generalizing over rotations in depth on a same-different successive matching task, as a way of testing 2D template-matching models of form-matching performance. STIMULI AND RESULTS. The stimuli were (simulated) shaded, randomly shaped, closed tubes. On some trials (“Same trials, different rotation”) the second presentation consisted of the form shown first, rotated in depth around a horizontal or ‘side-to-side’ axis. In Experiment 1, performance when viewing was in (simulated) stereo was compared to performance when one eye was covered, using (total) rotations of 38 degrees (19 degrees ‘up’ and 19 degrees ‘down’). On Same trials with differing rotations there was a stereo advantage of 70.6 percent to 57.3 percent; t(17) = 5.154, p < .001. In Experiment 2, rotations of 38 degrees (total) were also used, and the nonstereo condition consisted of showing (essentially) the same image to each eye. On Same trials with differing rotations there was a stereo advantage of 64.34 percent to 56.81 percent; t(25) = 2.718, p = .006. Experiment 3 was the same as Experiment 2, except that rotations of 30 degrees (total) were used. On Same trials with differing rotations, there was a stereo advantage of 68.43 percent to 57.85 percent; t(25) = 3.991, p < .001. In each experiment, performance under nonstereo viewing (Same trials, different rotation) was significantly greater than 50 percent (p < .025, p = .025, p < .005, respectively). Signal detection analyses will also be presented. CONCLUSIONS. At least with a difficult task, and stimuli that encourage global encoding, there is a stereo advantage in generalizing over changes of rotation in depth. The results count against 2D template-matching models of subject performance.