Latent Learning¶

Acquisition of information during exposure without direct reinforcement for that information, revealed by later performance when a goal or test makes it relevant.

Version: v2 · 2026-10-03 · History
Domain-specific #: 13374
Domain group: Social Sciences
Origin domain: Psychology & Behavioral Sciences
Subdomain: Learning → Psychology & Behavioral Sciences

Core Idea¶

Latent learning is acquisition of information that is not apparent in the learner's initial performance but becomes evident when a later goal or test makes it relevant. The target information was encountered without direct reinforcement for acquiring or using that information. Tolman's delayed-food rat-maze account and Stevenson's child incidental-object experiment show different versions of this acquisition–expression separation; neither directly observes a unique stored representation.[^{ref-22f7a65648c2][}ref-8654f5a4f076]

Scope of Application¶

In Tolman's rat maze, a delayed-reward group first ran without food, then found food from day 11 and showed a sharp error drop against earlier performance and comparison groups. In Stevenson's child task, searching for a key was rewarded, but knowledge of non-key objects was incidental and later tested. Thus an entire episode need not be unrewarded; the crucial absence is direct reward for the tested content during exposure.[^{ref-22f7a65648c2][}ref-8654f5a4f076]

Observational Learning and Reinforcement are neighbors, not substitutes for this identity. An everyday route anecdote is illustrative, not an additional evidenced experimental setting.

Clarity¶

Acquisition and expression need not occur together. A later jump in performance is evidence of earlier learning only to the extent that timing and controls limit the explanation that learning occurred solely after reward or during the test. Tolman's cognitive-map account is an interpretation of maze behavior, not a direct reading of the stored format.[^ref-22f7a65648c2]

Manages Complexity¶

Separate the learner, target information, prior exposure, early measure, later eliciting context and comparative evidence. This prevents early errors from automatically meaning no knowledge, while preventing later success from automatically proving when or how it was acquired. Reward schedules and object-test arrangements are evidence tools, not constitutive of every latent-learning case.[^{ref-22f7a65648c2][}ref-8654f5a4f076]

Abstract Reasoning¶

Specify what content was later used. Ask whether the learner encountered it before direct reward for that content, what could be observed then, and what later incentive or probe changed. Compare the later response with plausible new-learning explanations. In the rat case the target is maze structure; in the child case it is non-key object information acquired during a rewarded different task.[^{ref-22f7a65648c2][}ref-8654f5a4f076]

Knowledge Transfer¶

The same structural question crosses rat navigation and children's incidental object knowledge: did earlier experience supply information whose use was not initially visible? Food at a maze goal and a later object probe make different capabilities visible. Their control designs and specific representations do not transfer automatically; neither case requires an observed demonstrator or proves that all unrewarded exposure yields a cognitive map.[^{ref-22f7a65648c2][}ref-8654f5a4f076]

[^ref-22f7a65648c2]: Edward C. Tolman, “Cognitive Maps in Rats and Men,” Psychological Review 55, no. 4 (1948), 189–208, section “(1) Latent Learning Experiments,” printed pp. 194–196. Original text reproduced by York University: https://psychclassics.yorku.ca/Tolman/Maps/maps [^ref-8654f5a4f076]: Harold W. Stevenson, “Latent learning in children,” Journal of Experimental Psychology 47, no. 1 (1954), 17–21, DOI 10.1037/h0060086. Bibliographic record: https://pubmed.ncbi.nlm.nih.gov/13130805/ ; original abstract reproduced at https://www.researchgate.net/publication/10410881_Latent_learning_in_children . Full article not independently inspected.

Relationships to Other Abstractions¶

Current abstraction Latent Learning Domain-specific

Parents (1) — more general patterns this builds on

Latent Learning is a kind of Learning Prime

Earlier exposure yields retained information expressed in later behavior, a specific form of learning.

Hierarchy paths (2) — routes to 2 parentless roots

Latent Learning → Learning → Adaptation

Show alternative path (1)

Neighborhood in Abstraction Space¶

Latent Learning sits in a moderately populated region (57^th percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Language, Mind & Meaning-Making (57 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08