Skip to content
Psych Dictionary

Latent learning

  • Concept · Learning theory
  • Mixed or contested
  • Published October 19, 2026
  • 5 sources
  • 5 min read

In short

Latent learning is learning that occurs without any evident reward and does not appear in behavior until there is a reason to use it. Tolman’s maze experiments made it the center of a 30-year dispute over whether reinforcement is necessary for learning, a dispute that was never settled.

Key facts

Field
Learning theory
Originators
H. C. Blodgett, who named it in 1929; Edward C. Tolman and Honzik, who repeated and extended the experiments
First described
1929, University of California, Berkeley

State of the evidence

Mixed or contested

The basic finding was replicated in the Berkeley laboratory and elsewhere: rats that explored a maze without food made far fewer errors as soon as food was introduced[1]. What it shows was disputed for about 30 years and never resolved, because reinforcement theorists produced their own accounts of the same results[2]. The cognitive map that Tolman proposed to explain it is now a working idea in neuroscience, though most of the neural evidence comes from spatial navigation[3].

Evidence labels summarise the research literature; they are not recommendations. How labels are assigned

Latent learning is learning that takes place without any evident reward and does not show in behavior until there is a reason to use it. The term was introduced by H. C. Blodgett, whose maze experiments at Berkeley, published in 1929, found that rats that had wandered a maze without food performed far better as soon as food was placed at its end[1]. Edward C. Tolman made the finding a cornerstone of his view that learning consists not in strengthening connections between stimuli and responses but in building something that functions like a map of the environment[1].

The concept rests on a distinction between learning and performance: what an animal has learned may differ from what it does at a given moment. The same distinction appears in later research on observational learning, in which children who had seen a model punished imitated less but, once offered incentives, showed that they had learned as much as other children[4]. That study is described in the entry on the Bobo doll experiment.

Blodgett’s and Tolman’s maze experiments

In the typical maze experiment of the period, a hungry rat ran the maze once a day, and learning was measured as a decline in errors, meaning entries into blind alleys, and in the time taken to reach the food[1]. Blodgett ran three groups of rats through a six-unit alley maze. The control group found food in the goal box on every trial. A second group found no food in the maze for the first six days, being fed in their home cages about two hours later, and found food at the end of the maze from the seventh day. A third group was treated the same way but found food from the third day[1].

While they were not finding food, the experimental groups showed little sign of learning; their error curves did not drop. On the days after they first found food, their errors fell sharply. Blodgett called the learning that had taken place during the unrewarded trials, without showing in their performance, latent learning[1].

Tolman and Honzik repeated the experiments in 1930 with a 14-unit T-maze and larger groups[1]. One group of hungry rats found food at the end of the maze from the first day, a second never found food, and a third found food only from the eleventh day[2]. The third group’s errors dropped and its speed rose almost overnight once food appeared[2]. In a further experiment at the University of Iowa, Spence and Lippitt trained rats that were neither hungry nor thirsty in a Y-maze with food at the end of one arm and water at the end of the other. When the rats were later made hungry or thirsty, the hungry rats went more often toward the food and the thirsty rats toward the water[1].

Error curves of Tolman and Honzik’s three groups (schematic)A schematic chart of errors in the maze, from few at the bottom to many at the top, across daily runs, with no values plotted. The group that found food from the first day makes fewer errors from day to day, its curve falling steadily as the maze is learned. The group that never found food, drawn dashed, keeps making many errors, showing little sign of learning. The group that found food from day 11, drawn as the heaviest line, follows the never-fed group until a dashed marker at day 11, labeled food first appears, and then its errors drop sharply, almost overnight, to a low level. The chart is labeled schematic, not data.SCHEMATIC, NOT DATAERRORSManyFewDAILY RUNS THROUGH THE MAZEDay 11: food first appearsFood from thefirst dayNever foundfoodFood fromday 11

Figure 1Schematic error curves for Tolman and Honzik’s (1930) three groups of rats in a 14-unit maze: food from the first day, never any food, and food only from the eleventh day, after which that group’s errors dropped almost overnight. The shapes illustrate the reported pattern and are not data. After Tolman (1948) and Jensen (2006).

Text description

A schematic chart of errors in the maze, from few at the bottom to many at the top, across daily runs, with no values plotted. The group that found food from the first day makes fewer errors from day to day, its curve falling steadily as the maze is learned. The group that never found food, drawn dashed, keeps making many errors, showing little sign of learning. The group that found food from day 11, drawn as the heaviest line, follows the never-fed group until a dashed marker at day 11, labeled food first appears, and then its errors drop sharply, almost overnight, to a low level. The chart is labeled schematic, not data.

Cognitive maps

Tolman used these results against what he called the telephone switchboard school, which likened the nervous system to a switchboard connecting incoming stimuli to outgoing responses. For his own group, which he called field theorists, the central office was more like a map control room: incoming information is worked over into a tentative, cognitive-like map of the environment, indicating routes, paths and relationships, and this map determines what the animal finally does[1]. He distinguished narrow strip maps, which link the animal’s position to the goal by a single path, from comprehensive maps, which represent a wider area and allow a new route to be chosen when the old one is blocked[1].

Tolman closed his 1948 paper by applying the idea to people. He argued that excessive motivation or frustration narrows the cognitive map, and he treated regression, fixation and the displacement of aggression onto outgroups as examples of narrowing[1]. This was a speculative extension of the rat findings, not something the experiments tested.

The dispute with reinforcement theorists

The latent learning experiments became the main battleground in a dispute over whether reinforcement is necessary for learning. According to Jensen’s history, Clark Hull, Edward Thorndike and others held that learning requires a motivating state and some form of reinforcement, while Tolman, Guthrie, Spence and others denied that reinforcement is needed to produce the changes underlying learning[2]. Reinforcement theorists did not accept that the experiments refuted them. Hull showed that his own system, using what he called the fractional antedating goal reaction together with the change in incentive when food was introduced, predicted Tolman and Honzik’s results, and Guthrie explained maze learning by the contiguity of stimulus and response[2].

The debate lasted about 30 years, and by the mid-1960s many psychologists regarded the question as dead. Jensen argues that it ended not because one side won but because the issues remained unsettled and no empirical resolution seemed possible, given the hypothetical nature of the variables involved[2].

Skinner and the textbook version

Introductory textbooks have often told a simpler story. In a review of 48 textbooks published between 1948 and 2004, Jensen found repeated claims that the experiments show unequivocally that learning needs no reinforcement, that behavioral theories could not explain them, and that the cognitive map is the only explanation[2]. He also found B. F. Skinner described as a stimulus-response theorist who held reinforcement necessary for learning, a description Jensen argues is wrong on both counts[2].

Skinner stood apart from both sides. In 1950 he argued that maze measures, such as the ratio of right to wrong turns, cannot give a continuously changing measure of learning in a single organism, and he proposed the rate of a freely repeatable response, such as pressing a lever or pecking a key, as the better datum[5]. Jensen notes that behavior analysts have published no research on latent learning, partly because they did not regard the maze as useful for studying basic learning processes[2]. The textbook account in which the experiments defeated operant conditioning therefore runs together different behaviorisms, from the stimulus–response theories that grew out of classical conditioning to Skinner’s analysis of operant behavior.

Where the concept stands now

The phenomenon itself is not in doubt. Tolman wrote that other latent learning experiments in Berkeley and elsewhere had mostly confirmed the findings[1]. What remains open is the interpretation, and in particular whether exploration without food involved no reinforcement at all. Jensen points out that the unrewarded rats still ran the maze and still made fewer errors over time, and he proposes behavior-analytic variables, including exploratory behavior and features of the maze itself, that could account for the results[2].

Tolman’s second idea has had the longer life. Cognitive maps are now studied in neuroscience as representations of the relationships between things in the world that support flexible behavior. A 2018 review noted that most of the neural evidence for such maps comes from studies of spatial navigation, and that recent work has found parallels between spatial and non-spatial behavior, suggesting general mechanisms for organizing knowledge of many kinds[3]. That research uses Tolman’s term for a broader program and does not settle the original dispute about reinforcement.

Sources5

Numbered in order of first citation.

  1. 1.Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55(4), 189–208. https://doi.org/10.1037/h0061626
  2. 2.Jensen, R. (2006). Behaviorism, latent learning, and cognitive maps: Needed revisions in introductory psychology textbooks. The Behavior Analyst, 29(2), 187–209. https://doi.org/10.1007/BF03392130
  3. 3.Behrens, T. E. J., Muller, T. H., Whittington, J. C. R., Mark, S., Baram, A. B., Stachenfeld, K. L., & Kurth-Nelson, Z. (2018). What is a cognitive map? Organizing knowledge for flexible behavior. Neuron, 100(2), 490–509. https://doi.org/10.1016/j.neuron.2018.10.002
  4. 4.Bandura, A. (1965). Influence of models' reinforcement contingencies on the acquisition of imitative responses. Journal of Personality and Social Psychology, 1(6), 589–595. https://doi.org/10.1037/h0022070
  5. 5.Skinner, B. F. (1950). Are theories of learning necessary?. Psychological Review, 57(4), 193–216. https://doi.org/10.1037/h0054367

Further reading

  • Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55(4), 189–208. https://doi.org/10.1037/h0061626
  • Jensen, R. (2006). Behaviorism, latent learning, and cognitive maps: Needed revisions in introductory psychology textbooks. The Behavior Analyst, 29(2), 187–209. https://doi.org/10.1007/BF03392130
  • Behrens, T. E. J., Muller, T. H., Whittington, J. C. R., Mark, S., Baram, A. B., Stachenfeld, K. L., & Kurth-Nelson, Z. (2018). What is a cognitive map? Organizing knowledge for flexible behavior. Neuron, 100(2), 490–509. https://doi.org/10.1016/j.neuron.2018.10.002

Cite this entry

  • APA 7

    Psych Dictionary. (2026, October 19). Latent learning. In Psych Dictionary. https://www.psychdictionary.com/concepts/latent-learning

  • MLA 9

    “Latent Learning.” Psych Dictionary, 19 Oct. 2026, www.psychdictionary.com/concepts/latent-learning.

Written by Editorial team. First published October 19, 2026. Editorial standards

Educational reference material. Nothing on this site is a diagnosis, a treatment recommendation or a substitute for care from a qualified clinician.