Edward Tolman

In brief

Edward Chace Tolman (1886 to 1959) was an American psychologist who spent his career as a behaviorist and yet became behaviorism's most effective internal critic. Working with rats in mazes at Berkeley, he accepted the field's demand that psychology study observable behavior rather than introspected mind, but he denied its core assumption that learning is the mechanical stamping-in of stimulus-response bonds by reward. His rats, he argued, were not building chains of muscle twitches. They were building knowledge: an internal representation of the maze he called a "cognitive map." Two lines of evidence anchored the claim. In latent learning, rats allowed to wander an unrewarded maze showed, the moment food appeared, that they had learned its layout all along, which meant learning could occur without reinforcement. In spatial-orientation studies, rats trained to reach a goal by one route took a correct novel shortcut when it was offered, which meant they had learned where the goal was, not merely which turns to make. Tolman called his position "purposive behaviorism," and it kept the idea of internal, goal-directed cognition alive through the decades when strict behaviorism ruled American psychology.

The life

Tolman was born in 1886 in Newton, Massachusetts, into a comfortable, high-minded New England family. His older brother Richard Tolman became a distinguished physicist and chemist, and Edward followed the same expected path first, taking a degree in electrochemistry from the Massachusetts Institute of Technology in 1911. He disliked it. Reading William James turned him toward philosophy and psychology, and he enrolled at Harvard, earning his doctorate in psychology in 1915. The Harvard department of that era exposed him to the new behaviorism of John B. Watson, whose 1913 manifesto had insisted that psychology abandon consciousness and study behavior alone.

A formative influence came from Gestalt psychology and a period of study in Germany with Kurt Koffka. Gestalt psychology held that perception and thought are organized wholes, not sums of elementary sensations, and it left a permanent mark: Tolman would always resist the idea that behavior decomposes into isolated stimulus-response atoms. He absorbed Watson's methodological discipline without ever accepting the atomism that usually came with it.

After a brief appointment at Northwestern, Tolman joined the University of California, Berkeley, in 1918 and remained there for essentially the rest of his career, building one of the great American laboratories of animal learning. He was a pacifist, shaped by his opposition to the First World War, and a man of visible political conscience. That conscience produced the most public episode of his life. In 1949, during the anti-communist fervor of the early Cold War, the University of California required its faculty to sign a loyalty oath disavowing membership in subversive organizations. Tolman, though no communist, refused on principle, holding that the oath violated academic freedom and tenure. He became the leader of the resisting faculty and was dismissed. He carried the fight to the California courts, and in Tolman v. Underhill (1955) the state supreme court ordered the non-signers reinstated. He returned to Berkeley honored rather than broken; the campus building that housed the psychology and education faculties was later named Tolman Hall.

His science grew from the maze. Berkeley's rats gave him a controlled world in which to test what an animal actually acquires when it "learns," and across three decades he used them to mount a patient, evidence-based argument that learning is the acquisition of knowledge about the environment, not the automatic strengthening of responses. He gathered the framework in Purposive Behavior in Animals and Men (1932) and delivered its most famous statement, "Cognitive Maps in Rats and Men," as his presidential address to the American Psychological Association in 1948. He died in 1959, a decade before neuroscience began to find the cognitive map in the brain.

The full treatment

The problem it answers

By the 1920s and 1930s, American psychology was dominated by behaviorism, and the ambition of its most systematic figures, above all Clark Hull, was to explain all learning as the strengthening of stimulus-response connections through reinforcement. On this account a rat learns a maze the way a habit is drilled: at each choice point, the response that happens to lead to reward is reinforced and grows more probable, until a chain of correct turns is stamped in. The animal is essentially a switchboard; reward is the current that welds the connections. The theory was rigorous, quantitative, and, to many, complete. Tolman's problem was that it did not fit what his rats actually did.

What he claimed

Tolman argued that the rat running a maze is not assembling a chain of turns but forming a "cognitive map," a mental model of the spatial layout, of what leads to what, of where the goal lies relative to where the rat is. Behavior is guided by this internal representation and by expectancies about outcomes, and it is purposive: organized toward goals rather than triggered blindly by stimuli. He called the position "purposive behaviorism" to mark the balance. It is behaviorism, because it insists that every mental construct be tied to observable behavior and manipulable conditions. It is purposive, because it grants the animal internal states, goals, expectancies, and knowledge, that come between stimulus and response. Tolman named these "intervening variables": theoretical entities, inferred from behavior and defined by it, that a strict stimulus-response theory refused to admit.

The latent-learning demonstration

The sharpest evidence came from latent learning. In the classic study by Tolman and C. H. Honzik (1930), three groups of rats ran the same complex maze once a day. One group always found food at the end and steadily improved, cutting errors as reinforcement theory predicts. A second group never found food and drifted along making many errors. The third group found no food for the first ten days, then had food introduced on the eleventh. If learning required reinforcement, this group should have begun improving only after day eleven, slowly, from scratch. Instead, the day after food first appeared, their performance dropped almost at once to the level of the always-rewarded group, and in some measures below it. The rats had evidently been learning the maze all along, silently, during the unrewarded days. Reward did not create the learning; it merely gave them a reason to display it. Learning and performance, Tolman concluded, are different things, and learning can proceed without reinforcement. This was a direct blow to the Hullian claim that reinforcement is necessary for learning to occur.

Cognitive maps and the shortcut

The second pillar was spatial orientation. In Tolman, B. F. Ritchie, and D. Kalish (1946), rats were first trained to reach food by running along an indirect, roundabout path. Then the maze was replaced with a "sunburst" apparatus: the original path was blocked, and a fan of many new straight alleys radiated out in different directions. A stimulus-response theory predicts the rats should prefer the alley closest to the direction of their old trained response. Instead, the largest share of rats chose the alley that pointed almost directly at the true location of the food, a route they had never run. They behaved as though they knew where the goal was in space, not merely which movements had once paid off. Tolman took this as evidence for a map-like representation: the rat had encoded the goal's position, and could compute a novel path to it. He put the moral in memorable terms, warning against theories that make the animal a slave to "narrow strip maps" and arguing for "broad" cognitive maps flexible enough to support detours and shortcuts.

Related distinctions

Several of Tolman's finer distinctions carried the weight of the argument. Learning versus performance separated what an organism knows from what it does, and showed that motivation (like hunger, or the presence of reward) governs the second without being required for the first. Place learning versus response learning asked whether an animal at a maze junction learns "the goal is to the west" or "turn left here"; his group's studies favored place learning, and later work found the answer depends on conditions, but the question itself was his. And intervening variables gave the whole enterprise its method: internal states could be admitted into a science of behavior so long as each was anchored, operationally, to the observable inputs that produced it and the observable outputs it predicted.

Lineage

Tolman stood inside the behaviorist revolution begun by John B. Watson (1878 to 1958), and he never left it: he shared its insistence that psychology be built on public, measurable behavior rather than the introspection of the founding laboratories, the tradition that ran back to Wilhelm Wundt. But his refusal to reduce behavior to stimulus-response atoms came from Gestalt psychology and his study with Kurt Koffka (1886 to 1941), which taught him that organisms respond to organized wholes and relations, not isolated elements. William James's functional, purpose-centered view of mind pointed him the same way. His mature system was an attempt to fuse these currents: the rigor of behaviorism with the holism and purposiveness of the Gestalt and functionalist traditions. Downstream, his intervening variables and his image of internal representation fed directly into the cognitive revolution of the 1950s and 1960s, which reinstated internal mental processes as the proper subject of a scientific psychology, and the "cognitive map" survives as a working term in that field to this day.

The strongest case for it

The case for Tolman is that he was largely right, and demonstrably so, in an era when the fashion ran hard against him. His latent-learning and shortcut experiments are clean, replicated, and genuinely hard to accommodate on a strict reinforcement account: if reward is necessary for learning, rats should not learn a maze they are never rewarded in, and they do. If animals learn responses rather than places, they should not take a correct novel shortcut, and they do. He achieved this while honoring the methodological demand of his opponents, tying every internal construct to observable behavior, so he could not be dismissed as a mentalist smuggling the soul back in. His broader framework, expectancies, goals, internal models, intervening variables, is close to the vocabulary that cognitive psychology and cognitive neuroscience adopted once behaviorism receded. And the vindication went physical. In 1971 John O'Keefe discovered "place cells" in the rat hippocampus, neurons that fire when the animal occupies a particular location, and in The Hippocampus as a Cognitive Map (1978) O'Keefe and Lynn Nadel argued that the brain literally builds the kind of spatial representation Tolman had inferred from behavior. O'Keefe shared the 2014 Nobel Prize in Physiology or Medicine, with May-Britt and Edvard Moser, for that line of work. Few psychologists have had a metaphor confirmed by a Nobel decades after their death.

The strongest case against it

The objections were pressed hardest in Tolman's own lifetime, by rival behaviorists who thought his cognitive maps unnecessary and unscientific.

The central figure was Clark Hull (1884 to 1952), whose systematic reinforcement theory was the dominant learning psychology of the 1940s. Hull and his followers, especially Kenneth Spence (1907 to 1967), argued that the phenomena Tolman attributed to maps could be explained without them, by more careful stimulus-response modeling. Latent learning, they countered, need not show reinforcement-free learning at all: even an "unrewarded" rat gets rewards its experimenter did not intend, the reduction of exploratory drive, the relief of being removed from the maze, so reinforcement may be quietly present throughout. The neo-Hullians also produced latent-learning studies that failed to replicate or gave weaker effects, and the empirical record through the 1950s was genuinely mixed rather than a clean win for Tolman. On this view, positing an internal map was an unparsimonious leap beyond what the data forced.

A second, sharper objection targeted the explanatory adequacy of the map itself. The behaviorist Edwin Guthrie (1886 to 1959) made the famous complaint that Tolman's theory left the rat stranded: an animal equipped only with expectancies and a cognitive map is "buried in thought" at the choice point, contemplating its knowledge, with nothing in the theory to say why it ever moves. A map tells you where things are; it does not, by itself, get the legs going. Tolman's system, the charge ran, was strong on knowledge and weak on the mechanism that turns knowledge into action, precisely the gap that the response-based theories filled.

Tolman himself conceded ground. He was, by temperament, a builder of suggestive frameworks rather than a closer of every loophole, and he acknowledged that his theory was less complete and less predictive than Hull's tightly quantified system. He never produced the precise equations his rivals valued, and for a generation Hull's approach, not his, set the standard for rigor in learning theory. The verdict of history reversed that ranking, but the contemporary criticism was not a strawman: it identified real weaknesses in scope and mechanism that Tolman did not fully answer.

Where it stands now

Tolman is now read as one of the psychologists who was ahead of his time and has since been vindicated. Strict Hullian reinforcement theory faded; the cognitive revolution restored the internal representations Tolman had insisted on; and the discovery of place cells and, later, grid cells gave the cognitive map a home in the hippocampal-entorhinal system, turning a behavioral inference into a topic of mainstream neuroscience. Latent learning is a settled textbook phenomenon, and the distinction between learning and performance is basic to the field. This does not mean the old behaviorists were simply wrong about everything: reinforcement remains central to large parts of learning, and Guthrie's question about how a map issues in action is still, in effect, the problem that theories of the link between representation and behavior try to solve. What survives whole is Tolman's essential claim, that even a rat in a maze is acquiring knowledge of its world rather than a bundle of reflexes, and that a psychology which refuses to speak of internal representation will miss what learning actually is.

Test yourself

Think about how you find your way around a city you know well. When a familiar road is closed, do you retrace a memorized sequence of turns, or do you picture where your destination lies and improvise a new route toward it? If you can invent a shortcut you have never driven, you are using exactly the kind of internal representation Tolman inferred in his rats, and doing the thing a chain of stamped-in stimulus-response habits could never do.