Abstract

Previous work¹ suggests that perceived depth from multiple cues is a weighted average of the depth estimated by using each cue. A cue's weight is decreased if the cue is less reliable or is strongly discrepant from other cues. These ideas were confirmed empirically by using perturbation analysis: perceived depth was measured in a multi-cue stimulus in which depth portrayed by one of the cues was varied. We suggest that this entire framework (the statistical model of cue combination measured by using perturbation analysis) may be applied to judgments of spatial location involving multiple location cues. Rivest and Cavanagh (ARVO, 1991) showed that additional edge cues can increase localization precision. Our statistical model predicts this. It also predicts perceived location when cues are inconsistent. In our experiments, stimuli consisted of multiple line segments. An edge was indicated by a gradient in line length, orientation, contrast, blur, and/or density. The edge location indicated by each cue was separately manipulated, as was the reliability of each cue (by noise manipulations). Preliminary results indicate that perceived spatial location is a weighted average of the locations indicated by individual cues and that the most reliable cues are given more weight. When one cue is strongly discrepant, it is no longer averaged with the other cues and two edges are perceived. Dependence between cues is also seen (e.g., lowering contrast or shortening length decreases orientation cue reliability).

© 1992 Optical Society of America