The Neo-Behaviorism of Edward Tolman

Tinklepaugh (1928), one of Tolman's students, using a delayed reaction task, demonstrated that monkeys reacted negatively when they were shown a banana being hidden under a cup, but discovering that lettuce had been substituted there in their absence. As a control, Tinklepaugh had also left the lettuce hidden and untouched in other trials. Here is his description of the monkey’s reactions (p. 229) during the control (a) and experimental conditions (b):

"(a) Subject rushes to proper cup and picks it up. Seizes lettuce. Rushes away with lettuce in mouth, paying no attention to other cup or to setting. Time, 3–4 seconds.
(b) Subject rushes to proper cup and picks it up. Extends hand toward lettuce. Stops. Looks around on floor. Looks in, under, around cup. Glances at other cup. Looks back at screen. Looks under and around self. Looks and shrieks at any observer present. Walks away, leaving lettuce untouched on floor. Time, 10–33 seconds."

He and Tolman interpreted these results as evidence that the monkey expected lettuce in the control condition and received it, but expected a banana in the experimental condition and received lettuce instead. They interpreted the monkey’s different reaction in each case as behavioral evidence of different cognitive states. Tolman believed he had demonstrated expectancy in a non-human animal.

One of Tolman’s most famous contributions was the cognitive map, a concept he derived from Gestalt psychology. In a series of experiments, Tolman and his coworkers demonstrated that rats learned the spatial relationships between themselves and a food item. After learning an initial forced roundabout relationship in a maze between themselves and food located in a specific place, rats would successfully choose a new path in a different maze but one in which the food was in the same place it had been in the previous maze. That new path, usually, was a direct line to the food item. To Tolman, that indicated that the rats knew where the food was located in a cognitive map that they carried in their heads. He wrote describing one of his studies (Tolman, 1948, p. 204):

"the rats in this experiment had learned not only to run rapidly down the original roundabout route but also, when this was blocked and radiating paths presented, to select one pointing rather directly towards the point where the food had been."

Many students possess detailed and accurate cognitive maps of their preferred routes to class or nearby stores. But, their cognitive maps for places they seldom visit are more fuzzy and inaccurate. Tolman would have argued that people’s cognitive maps developed with experience. The more experience they had with a route or location, the better their map, he thought. An old New Yorker cartoon in which there are two very different cognitive maps of the same area. The policeman’s map is very clear while the tourist’s is not. Carroll (2017) argued that Tolman’s research on cognitive maps, although using only observable behaviors, has led to the now burgeoning field of cognitive neuroscience. He wrote (p. 216):


"Tolman demonstrated how the methods of the behaviorist could be used to study more complex and interesting psychological phenomena. Psychologists’ recognition of the potential of his approach to pursue long-neglected topics—coupled with events in artificial intelligence, linguistics, and neuroscience—played a central role in the resurgence of interest in cognitive processes in the latter half of the twentieth century."

Another famous concept of Tolman’s was latent learning. That line of research led to his distinction between learning and performance. Following earlier work in his laboratory, Tolman and Honzik (1930) decided to investigate whether learning required a reinforcer (food, in this case). They trained three groups of rats in the same complex maze using the average number of errors per trial as the criterion for learning. Those errors should drop over time, exhibiting a classic learning curve. One group, the control (HNR), was never fed. Another group was always fed (HR), and the final group (HNR-R) was only fed after the 11th day of training. If learning required a food reinforcer, then that last group should begin learning the maze only after commencing to eat on the 11th day.

That is not what happened. Instead, once the last group began to be fed, their errors dropped precipitously. Tolman and Honzik interpreted these data by saying that those rats had already learned the maze, but that it took the presence of food in the goal box for them to perform or demonstrate their learning. They called the phenomenon latent learning because the learning, they argued, had already taken place. They argued that the new presence of the food reinforcer now changed the situation, causing their errors to go down accordingly. In their theorizing, reinforcement was not necessary for learning.


Tolman created a viable neobehaviorist alternative to Watson’s earlier scheme. To do so, Tolman endowed his rats (and, by extension, people) with intervening variables or variables that lay between a physical stimulus and an observable behavior. Those intervening variables were the actual cause of the behavior, but were unobservable. Yet, they were still amenable to experimental analysis via the doctrine of operationism that had come to psychology via physics. One of the best examples of a psychological intervening variable is hunger.

To an old-time introspective psychologist, hunger would be an aspect of consciousness, a kind of feeling. However, no one could estimate or compare degrees of hunger made by different hungry people using introspective methods. Tolman (and nearly all psychologists since) operationalized hunger by providing descriptions of how to obtain hunger, namely by withholding food. Thus, a rat that had not eaten for 12 hours was hungrier than one that had not eaten for six hours. Operational definitions allowed neobehaviorists to describe internal states without using the methods of traditional introspection.

Tolman created a liberal compromise between the introspective methods of the past and Watson’s extreme behaviorist position. In place of S-R theory, he introduced S-S (stimulus-stimulus) theory. Learning, for Tolman, was all about the relationships between stimuli, not the response to a given stimulus. His position, however, was too close to the older kind of psychology for many psychologists. Clark Hull, in particular, became Tolman’s main theoretical rival. Hull, too, sought to remedy the problems of Watson’s Behaviorism. His solution was to keep Watson’s central idea intact; behavior could be controlled and predicted without using any reference to cognitive concepts such as expectancies or cognitive maps. He tried to explain learning via a complex and ambitious overarching theory full of mechanistic variables.


Back to Main Page