Abstract
We perceive binocular surfaces in depth when viewing random dot stereograms (RDS) only when the horizontal disparity is limited. Several researchers have reported the upper limit of fusion using RDS to be much larger than Panum's fusional area (Julesz and Fender, 1967; Erkelens, 1988; Piantanida, 1986), while others supported the classical limit fusional (Duwaer, 1983). To explore these conflicting conclusions, we developed a RDS requiring dichoptic color fusion. Dichoptic color mixing occurs when the dots are fused to form a surface in depth, and the dots return to their intrinsic colors when they are not fused. The appearance of depth is undisturbed by dichoptic color mixing. We find that the upper limit of fusion for the RDS is comparable to the classical Panum's fusional area, while the limit of depth perception is much greater.
We applied dichoptic color fusion to Panum's limiting case in lines and in RDS. In Panum's limiting case, a single vertical line in one eye corresponds to a pair of vertical lines in the other, and gives rise to the perception of two lines at different depths. Different explanations have been offered for this perception including double-duty matching, a violation of the uniqueness constraint posited in several stereo algorithms. In our study, the lines in the two eyes had different colors and a distinctly different fused color. We found color fusion for only one of the two lines, the second retaining its monocular color. Different observers had either a consistent nasal or temporal fusional bias. However, this was not the case when we used RDS with extra monocular dots of a different color placed beside binocularly matched dots. In this case observers perceive surfaces at two depths, one of which corresponds to the matched pairs and the other of which has the fused color due to fusing the differently colored dots, in apparent violation of the uniqueness constraint. This may be due to double-duty image matching.