A Dynamic Route Finder for the Cognitive Map

Cognitive behaviorist E. C. Tolman (1932) proposed many years ago that rats and men navigate with the aid of cognitive maps, but his theory was incomplete. Critic E. R. Guthrie (1935) pointed out that Tolman's maps lack a rule for action, a route finder. We show that a dynamic model for stimulus generalization based on an elementary diffusion process can reproduce the qualitative properties of spatial orientation in animals: area-restricted search in the open field, finding shortcuts, barrier learning (the Umweg problem), spatial "insight" in mazes, and radial maze behavior. The model provides a behavioristic reader for Tolman's cognitive map. The cognitive behaviorist Edward Tolman spent much of his career devising clever experiments to show that stimulus-response accounts of rat behavior cannot be correct. Some of his most striking demonstrations involve spatial learning. One such example is shown in Figure 1, which depicts a maze apparatus used in a famous experiment by Tolman and Honzik (1930). The maze has three paths from the Start box to the Goal box. The paths differ in length: Path 1 (heavy vertical line) shorter than Path 2 (intermediate line) shorter than Path 3 (light line). In preliminary training, the rats were allowed to become familiar with all three paths to the Goal box. They also had experience with a block at Point A, which permits access to the Goal only via Paths 2 and 3. In the test condition, the block was moved to Point B--so that only Path 3 is open. The question is, Will the rats choose Path 3 as soon as they encounter the block at B, or will they choose Path 2, which is normally preferred to Path 3--indicating that they do not know Paths 1 and 2 share a common, blocked, segment? Tolman and Honzik's rats behaved intelligently and usually went straight to Path 3 after encountering the block at B. Tolman took this as evidence that the rats knew something about the topography of the maze. They were not just operating on a fixed hierarchy of preferences ( "Path 1 better than Path 2 better than Path 3" ), nor were they responding reflexively to local cues. Tolman considered this behavior to be an example of "insight," although he did not specify exactly what that means. He did say that some kind of cognitive map