Sternberg's (1977) people-piece task had subjects solve picture analogies of the form A:B::C:D. The elements of the analogy were drawing of people varying four binary attributes: sex, color, height, and girth. Subjects were asked to respond whether the analogy was true or false, that is, whether the A-B (or A-C) pair differed in the same characteristics as the C-D (or B-D) pair. See Figure 10.7 in the book chapter for a sample people-piece analogy problem.
This model represents behavior in the people-piece task. For each problem, the model first decides which strategy to employ: mapping A to B, or mapping A to C. Assuming it maps A to B, the model encodes A and B, infers the mapping between them (i.e. the changed attributes), encodes C and D, applies the mapping to C to create D', compares D and D' for equality, and responds. If the model chooses to map A to C, it follows the same sequence of steps with B and C switched. This model nicely predicts the latencies reported by Sternberg for problems in which the A-B and A-C pairs differed by one or two attributes.