Figure 4. Top panel: Bivariate normal distributions represent the likelihoods that various rectangles were presented as exemplars of fuzzy categories A or B. Each point in the space over which the distributions are defined corresponds to a possible rectangle displayed to a pigeon. Any rectangle could be presented as an exemplar of either category, but most rectangles were more likely to belong to one category than to the other.
Bottom panel: Contours of equal likelihood and the corresponding linear optimal decision bound, according to which a rectangle should be categorized as an A or a B depending on whether the rectangle is taller than wide or wider than tall, respectively. See Figure 1 and related text in Herbranson, Fremouw, and Shimp (1999) for additional details.