Figure 5.

Experiment III: the first percept bias does not depend on occlusions cues. (a) Proportion of trials for which subjects reported the transparent percept first, for tristable (‘ambiguous’) and bistable (‘occlusion’) plaids. Thick line: average data for nine subjects; thin lines: data for each subject based on 48 trials for each condition. (b) The proportion difference was not statistically significant (sign test, d.f. = 8). Median, lower and upper quartiles are displayed. Whiskers mark the most extreme values within 1.5 times the interquartile range. Crosses denote outlier values. We computed within each trial the fraction of time the percept was reported as transparent, as a function of whether the first percept was coherent ((c) n = 596) or transparent ((d) n = 257). The luminance of the intersections had no effect (paired sign test, d.f. = 8).