Abstract
We elucidate two properties of the intersection of constraints (IC) model of depth cue combination (Domini et al., 2006, Vision Research, 46, 1707–1723). First we show that, like the modified weak fusion (MWF) model (Landy et al., 1995, Vision Research, 35, 389–412), IC combines depth cues in a weighted sum that maximizes the signal-to-noise ratio of the combined cue. Thus, IC is more similar to MWF than may at first appear. Second, we show that IC measures perceived depth in terms of just-noticeable differences (JND's), and hence predicts that the perceived depth difference between two stimuli is proportional to the number of JND's separating them, regardless of what combination of disparity and motion cues are in play. We tested this prediction.
Method We created two motion-defined random dot cylinders, mC and mE, with circular and elliptical cross-sections, respectively. We also created two disparity-defined random dot cylinders, dC and dE, and we measured points of subjective equality to match the perceived depth of dC to mC, and of dE to mE. Using a 2AFC method of constants design, we then measured depth JND's for motion-defined cylinders, using mC as a baseline, and calculated the number of JND's separating mC and mE. Similarly, we measured depth JND's for disparity-defined cylinders, using dC as a baseline, and calculated the number of JND's separating dC and dE.
Results The number of JND's separating mC and mE was significantly different from the number of JND's separating dC and dE, even though the the perceived depth difference between mC and mE was the same as that between dC and dE.
Conclusions The perceived depth difference between two stimuli is not proportional to the number of JND's separating them. This finding poses no challenge to MWF, but contradicts a key prediction of the IC model.
Natural Sciences and Engineering Council of Canada (LMW)