Edward Tolman

Introduction

Edward Chace Tolman (14 April 1886 – 19 November 1959) was an American psychologist who spent his career demonstrating a proposition that most of his colleagues considered unscientific: that rats in mazes learn where things are, not what to do.

This sounds obvious now, and that is largely because of him. In the 1930s and 1940s the dominant theories of learning held that animals acquire stimulus–response connections, strengthened by reinforcement. Tolman's experiments showed that animals can acquire knowledge of a layout with no reinforcement at all, and then deploy it the moment they have a reason to. Learning, he argued, is the acquisition of knowledge; performance is something else.

He was a behaviourist by method and not by conviction — he insisted on objective, experiment-based evidence and refused to speculate about consciousness, but he used that method to argue for the existence of internal representations. The position earned him a famous sarcasm from Edwin Guthrie, who said Tolman left the rat "buried in thought" at the choice point. It also made him, in retrospect, the bridge between behaviourism and the cognitive psychology that replaced it.

He is also the only major figure in mid-century learning theory whose ideas are currently gaining ground rather than receding, largely because the neuroscience of spatial navigation caught up with him.

Early Life and Education

Tolman was born in West Newton, Massachusetts, into a prosperous and intellectually formidable family. His father was among the first graduating class of the Massachusetts Institute of Technology and became a successful business executive; his mother was a Quaker, and the moral seriousness she instilled — plain living, principled conduct, and an obligation to examine one's own life — stayed with him and later surfaced in unexpected places.

His older brother Richard, five years his senior, became a distinguished theoretical chemist and physical chemist who worked on the Manhattan Project. The family expectation was that Edward would follow his father to MIT, and he did, enrolling in 1906.

He was not interested in engineering. In his senior year he read William James's Principles of Psychology and switched course, taking his degree in electrochemistry in 1911 and then entering Harvard's graduate programme in philosophy and psychology that autumn. The following summer he went to Giessen in Germany to study with the Gestalt psychologist Kurt Koffka — an encounter that shaped everything he later did, because Gestalt psychology's central claim, that experience is organised into structured wholes rather than assembled from elements, is what Tolman spent his career importing into the study of animal learning.

He completed his doctorate at Harvard in 1915 under Hugo Münsterberg and Herbert Langfeld, with a dissertation on retroactive inhibition in memory.

Academic and Professional Career

Tolman taught at Northwestern University from 1915. He was dismissed in 1918 — university records cited lack of teaching achievement, but the real reason was his publication of an article in a pacifist student magazine, at a moment when the United States was at war. His Quaker upbringing and his opposition to conscription made him conspicuous. He was, however, a volunteer for military service in the First World War and was offered an army commission.

In 1918 he joined the University of California, Berkeley, and remained there for the rest of his career, building an animal laboratory devoted to the study of learning. His major theoretical statement, Purposive Behavior in Animals and Men, appeared in 1932.

He was elected president of the American Psychological Association in 1937 and to the National Academy of Sciences the same year. During the Second World War he served for two years in the Office of Strategic Services, the forerunner of the CIA — a striking posting for a pacifist, though consistent with a pattern in which he opposed war in principle and worked against fascism in practice.

In 1949 the Regents of the University of California demanded that every faculty member sign a loyalty oath or face dismissal. Tolman, a longtime member of the ACLU's national board, led the faculty opposition — not out of any sympathy for communism but because, as he put it, the oath infringed academic freedom and compelled political affirmation as a condition of employment. He was dismissed in 1950 and spent the intervening years teaching at the University of Chicago and Harvard.

He sued. In Tolman v. Underhill (1955) the California Supreme Court invalidated the oath and ordered the reinstatement of everyone who had refused to sign. Tolman returned to Berkeley, received the APA's Distinguished Scientific Contribution Award in 1957, and retired in 1959 — the year he died, on 19 November, in Berkeley. Shortly before his death the university awarded him an honorary doctorate, which is either an apology or an irony, and possibly both.

In 1963 the university named its new Education and Psychology building Tolman Hall.

Major Contributions to Psychology

Latent learning

The experiment that made his reputation was run with C.H. Honzik in 1930. Three groups of rats ran the same maze for a series of trials. One group found food at the end every time; one never did; a third found nothing for the first ten trials and food from the eleventh onward.

The first group's error rate fell steadily. The second group's did not. The third group's behaved like the second group's for ten trials — and then, from the eleventh, dropped abruptly to the level of the always-rewarded group.

The conclusion was that the third group had been learning the maze all along. What they lacked was not knowledge but a reason to use it. Reinforcement, Tolman argued, governs performance, not learning — a distinction that had not existed before and that reorganised the field's dependent variables.

Cognitive maps and place learning

Tolman's account of what the rats had learned was that they had acquired a representation of the maze's layout: a cognitive map. In a companion line of work he showed that rats trained to run a maze to reach food, and then required to swim a flooded version of it, performed well above chance — they had learned where the food was, not a sequence of turns.

The decisive test came in 1946, in work with Ritchie and Kalish using a maze shaped like a sunburst, with a dozen arms radiating from a common platform. Rats were trained to run from one particular arm to the goal. At test, the apparatus was rotated and the animals were released from a different arm. A response learner — one who had acquired "turn right at the choice point" — should keep turning right. A place learner should head for where the food is. The overwhelming majority headed for the place. The same design has been run in countless variants since, and the result is consistent: under most conditions, in a layout the animal has had time to survey, place wins.

The idea was formalised in his 1948 paper "Cognitive Maps in Rats and Men," which argued that the same representational capacity underlies human spatial knowledge and, by extension, more abstract forms of organised knowledge — the passage in which he wrote that in the course of learning something like a field map of the environment gets established in the rat's brain.

Intervening variables

Tolman's methodological innovation was the intervening variable: a construct posited between the independent variable (environmental conditions) and the dependent variable (observed behaviour), defined operationally rather than speculated about. Hunger, for instance, could be defined as hours of food deprivation and manipulated accordingly.

This was the mechanism by which a behaviourist could talk about expectations, purposes, and representations without abandoning objective method. It is now such standard practice that its radicalism is invisible; at the time it was the central point of contention between Tolman and the Hullian programme, which sought to derive behaviour from drive and habit strength without any representational middle term.

Purposive behaviourism

His systematic position, set out in 1932, held that behaviour is molar rather than molecular: the unit of analysis is a goal-directed act, not a muscle twitch. Behaviour is purposive, selective, and follows a least-effort principle, and it is governed by sign-gestalt expectations — learned expectations that a given sign leads to a given significate within a behaviour space.

This was, in effect, Gestalt psychology rendered in the vocabulary of animal learning, and it put him permanently at odds with Clark Hull's drive-reduction theory, which dominated the period.

Behaviour genetics

A less-remembered line of work: from the 1920s Tolman selectively bred rats for maze-learning ability, producing "maze-bright" and "maze-dull" lines. This was among the first attempts to demonstrate a heritable component in a specific learning ability, and it anticipates the later field of behaviour genetics. It also complicated his own position — he was arguing that learning builds internal representations, while simultaneously showing that the capacity to do so is partly inherited.

Major Publications

Purposive Behavior in Animals and Men (1932) — the systematic statement, dedicated to the white rat
"The Determiners of Behavior at a Choice Point" (1938, Psychological Review) — the independent/intervening/dependent variable framework
Drives Toward War (1942) — a psychoanalytically inflected account of the motives behind warfare
"Cognitive Maps in Rats and Men" (1948, Psychological Review) — where the term is formalised
Collected Papers in Psychology (1951)
"Principles of Performance" (1955) — his late answer to the charge that his rats never actually moved

Impact and Legacy

Tolman lost the argument in his own lifetime and won it afterwards. By the late 1950s his influence had collapsed. Skinner's 1950 paper "Are Theories of Learning Necessary?" persuaded a generation that it was more productive to study behaviour directly than to infer mental states from it, and the S–S versus S–R debate had become, in the standard description, convoluted and sterile.

What reversed the verdict was evidence he could not have anticipated. The discovery of place cells in the hippocampus, and then of grid cells, gave the cognitive map a neural substrate; O'Keefe and Nadel titled their 1978 synthesis The Hippocampus as a Cognitive Map, explicitly taking up Tolman's term, and the 2014 Nobel Prize in Physiology or Medicine went to the work that established it. Spatial and abstract knowledge are now known to be represented in the same hippocampal system, which is a stronger version of what Tolman claimed in 1948. Of the three dominant figures of mid-century animal learning — Tolman, Hull, and B.F. Skinner — it is Tolman's research programme that is currently the most active.

His concepts are also used far more widely than they are attributed. "Cognitive map" is now standard vocabulary in geography, urban planning, organisational theory, and artificial intelligence, and most of its users have no idea it was coined to explain rats in a maze.

Criticism. The objection that dogged him throughout was that his theory predicted too little. Guthrie's gibe — that Tolman's rat was left "buried in thought" at the choice point, unable to respond — was unfair but not empty: Tolman's account of what was learned was more developed than his account of when and whether it would be expressed. His 1938 and 1955 papers were written specifically to answer this charge.

Hull's followers, notably Kenneth Spence, were able to produce alternative explanations of the latent learning results in terms of competing habit strengths and incentive motivation, and for a time the argument looked winnable on their terms. It was precisely because both sides could accommodate the data that the debate became unprofitable — which is what Skinner's intervention exploited.

There is also a substantive point about the experimental design. Latent learning is inferred from a performance change that occurs only after reinforcement is introduced, which means the construct is defined by the very manipulation said not to cause it. The inference is sound, but it is an inference, and it depends on the assumption that no other process accounts for the abrupt improvement.

The intervening variable — the device that was supposed to keep Tolman's mentalism honest — attracted the most searching criticism, and from sympathetic quarters. MacCorquodale and Meehl argued in 1948 that a distinction had to be drawn between an intervening variable, which is no more than a shorthand summary of the observed relations, and a hypothetical construct, which carries surplus meaning about an unobserved entity. On that analysis Tolman's position was unstable: either his intervening variables were mere abbreviations, in which case they explained nothing and his theory was not really a theory, or they were surplus-meaning constructs, in which case he was doing exactly the mentalism his method was designed to avoid. Tolman's defenders reply that the ambiguity was the point — that the intervening variable was a way of building a bridge from behaviour to representation one empirical anchor at a time, and that demanding the bridge be either fully present or fully absent would have stopped the project at the start.

Finally, his behaviour genetics work has aged awkwardly. Selective breeding for maze ability demonstrates heritability of a trait within a specific environment, but "maze-bright" and "maze-dull" lines turned out to differ on many things besides maze learning — emotionality, for one — and the field moved away from the interpretation he put on it. Drives Toward War, written in the same period, is more awkward still: a psychoanalytic account of the motives behind warfare from a man whose central methodological commitment was to operational definition, and it reads as the work of someone applying a discipline he had spent his career arguing against.

Learn More

Clark Hull — his great theoretical rival, on the S–R side of the debate
B.F. Skinner — whose 1950 critique redirected the field away from theory
Behaviourism — the tradition Tolman worked inside and against
History of Psychology — the behaviourist era and the cognitive revolution that followed

References

Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55(4), 189-208.