Signal-to-noise ratio and not simulated depth predicts perceived 3-D structure from stereo and motion stimuli

To recover absolute depth in terms of a Euclidean inverse-projection model of 3-D shape perception, scaling parameters like absolute distance and object rotation are required to scale the information provided by binocular disparities and retinal velocities. In the psychophysical literature, however, large and systematic departures from veridical perception have been reported for each cue shown in isolation, suggesting that these scaling parameters are poorly estimated by human observers. Here we contrast two theories of 3-D shape perception. The inverse-geometry theory predicts that two stimuli are perceptually matched when they simulate the same depth. The intrinsic constraint theory (Domini, Caudek, & Tassinari, 2006) predicts that the signal-to-noise Ratio (SNR) of velocity and disparity signals is the best predictor of human performance. In this study, we found support for the second theory since observers reported that (i) perceived depth was equal for stimuli having the same SNR but different simulated depths; (ii) perceived depth was different for stimuli with different SNR magnitudes but same simulated depth.