Social Learning

Abstract

Social Learning is a type of learning: acquiring behavior, knowledge, or skill by observing or interacting with others rather than through solitary trial and error. Miller and Dollard first cast imitation as reinforced behavior, and Bandura’s Bobo-doll studies then showed that mere observation can install new responses even when the observer is never rewarded. Comparative and developmental work distinguishes the mechanisms that make copying possible from the strategies that decide when and whom to copy. Formal tournaments show that selective, high-fidelity copying outperforms both asocial learning and indiscriminate imitation, and at scale such faithful transmission supports cumulative culture — the ratcheting accumulation of improvements across generations that is uniquely elaborated in humans. Three interactive demonstrations model vicarious reinforcement, a social-learning-strategies tournament, and the cultural ratchet.

Keywords: social learning, observational learning, imitation, social learning strategies, cumulative culture

What Social Learning Is

Social learning is the acquisition of behavior, knowledge, or skill through observation of, or interaction with, other individuals or their products. It stands in contrast to asocial (or individual) learning, in which an organism discovers a contingency for itself through direct experience — trial and error, or the pairing of its own actions with outcomes. The defining feature is the source of the information: a social learner exploits what others have already learned, rather than paying the full cost of discovery alone.

That shortcut is the reason social learning matters. A naive forager who watches a knowledgeable one avoids the poisonous plant without ever eating it; a child who hears a word once can use it without deriving its meaning from scratch. Because the information travels between individuals, a behavior can spread through a population and persist across generations far faster than genetic evolution could install it — the basis of what biologists call a second, cultural inheritance system that runs alongside the genetic one.

Cognitive psychology treats social learning as a family of processes rather than a single faculty, and two questions organize the field. The first is mechanistic: what psychological process actually transfers the behavior — is the observer copying the exact motor actions, or merely being drawn to the same objects and re-discovering the solution privately? The second is strategic: given that copying is available, when is it worth doing, and whose behavior is worth copying? The sections that follow take these in turn, after locating social learning within the wider taxonomy of learning.

Types of Social Learning

The Medical Subject Headings (MeSH) vocabulary files Social Learning (D000067570) as a narrower kind of learning, and it carries one narrower descriptor of its own. MeSH is an indexing classification built for retrieving literature, not a theory of the mind, so its single child captures only one institutional locus of social learning — the others discussed on this page (imitation, emulation, observational learning) are treated by MeSH as coordinate descriptors elsewhere in the tree rather than as subtypes here. The categories below are therefore orthogonal to the mechanism-based taxonomy of the next section: a community of practice is a setting in which social learning happens, not a distinct copying process.

Community of Practice has no dedicated article on this site yet, so it is named here without a link. The upward direction of the same hierarchy is more informative: Social Learning’s parent is learning itself, which places social learning as one route by which the general capacity to profit from experience is realized — the route that exploits other organisms as the source of that experience.

Mechanisms of Social Learning

The single word imitation hides a set of psychologically distinct processes, and telling them apart is the central problem of the comparative literature. Cecilia Heyes (1994) set out the categories that still organize the field, distinguishing what an observer learns and how. In local enhancement, the model merely draws the observer’s attention to a place; in stimulus enhancement, to a kind of object. The observer then learns the task for itself, on its own, but is steered to the right part of the environment — so no behavior is really copied, only where to look is transmitted.

Higher up the ladder, the observer learns something about the task itself. In emulation, the observer grasps the goal or the affordances of the objects — that the box can be opened, that food is inside — and then achieves that outcome using whatever actions it likes, which may differ entirely from the model’s. In true imitation, by contrast, the observer copies the specific actions the model used, reproducing the form of the behavior and not merely its result. The distinction matters because only action-copying can transmit an arbitrary or opaque technique whose logic the observer does not understand — the kind of high-fidelity copying that cumulative culture later turns out to require.

This mechanistic ladder cuts across the folk notion of “monkey see, monkey do.” Many apparent cases of animal imitation dissolve, on analysis, into enhancement or emulation: the animals converge on the same solution not because they copied an action but because attention was directed to the same objects and each re-solved the problem privately. Heyes later argued that the mechanisms of social learning are themselves partly learned — cognitive gadgets built during development by general associative processes and cultural experience, rather than dedicated instincts — so that even the capacity to imitate is, in part, culturally inherited.

From Imitation to Social Cognitive Theory

The first systematic psychology of social learning was behaviorist. Neal Miller and John Dollard (1941), working within Hullian learning theory, treated imitation as ordinary instrumental behavior: an observer copies a model, the copying is reinforced, and the tendency to match the model’s behavior is strengthened like any other response. On this account imitation is matched-dependent behavior, learned because it pays — there is nothing special about the social nature of the cue.

Albert Bandura broke decisively with that view. In the Bobo-doll studies, children who watched an adult model attack an inflatable doll later reproduced the aggression themselves — including novel acts they could not have been reinforced for, since they had never performed them before (Bandura, Ross, & Ross, 1961). Observation alone had installed the behavior, with no reinforcement of the observer at all. This is the crux of Bandura’s challenge to behaviorism: learning and performance are separable, and a great deal is learned by watching that is not immediately, or ever, acted out.

A follow-up made the separation explicit. When children saw the model rewarded, punished, or left with no consequence, those who saw punishment performed fewer of the aggressive acts spontaneously — but when all children were then offered an incentive to reproduce what they had seen, the differences vanished: every group had learned the same behavior, and only their willingness to perform it had differed (Bandura, 1965). The consequence of the model’s behavior to the observer is vicarious reinforcement: we adjust our own behavior according to outcomes we merely witness. Bandura synthesized these findings into social learning theory — later broadened to social cognitive theory — which places observation, symbolic representation, and self-regulation at the center of how humans acquire behavior (Bandura, 1977).

Social Learning Strategies

Granting that animals can copy, a further question follows: should they, and whom should they copy? Indiscriminate imitation is a bad idea — copying an out-of-date, maladaptive, or simply unlucky model spreads error as efficiently as it spreads skill. Kevin Laland (2004) reframed the field around this insight, proposing that natural selection favors social learning strategies: evolved or learned rules specifying the conditions under which an individual copies, and which models it copies. He grouped them into when strategies (e.g. copy when asocial learning is costly, copy when uncertain, copy when dissatisfied) and whom strategies (copy the majority, copy successful individuals, copy kin or the old).

The decisive evidence came from a tournament. Rendell and colleagues (2010) invited entrants to submit strategies for a simulated agent that could, on each move, innovate asocially, observe another agent, or exploit what it already knew — then ran the strategies against one another in an evolutionary competition. The result overturned the intuition that more copying is better. Winning strategies copied heavily but selectively: they relied on social learning almost to the exclusion of costly individual innovation, but they timed their observations to acquire high-payoff behaviors and discarded information as it went stale. Blind, constant imitation did poorly; so did stubborn asocial learning.

Later work turned from whether to copy to the finer question of whom, and integrated it with metacognition. Rachel Kendal, Laland, and colleagues (2018) catalogued the transmission biases that steer real copying — payoff-based, frequency-based (conformist), and model-based (prestige, age, similarity) biases — and argued they bridge animal behavior, developmental psychology, and cultural evolution. Heyes (2016) added that many human social learning strategies are metacognitive: rather than hard-wired rules, they are flexible policies that weigh the reliability of a source, deciding whom to trust much as one decides how much to trust one’s own memory.

Cumulative Culture and the Ratchet

Social learning becomes historically consequential when its products accumulate. Cumulative culture is the process by which behaviors are transmitted, modified, improved, and re-transmitted, so that later generations inherit refinements no single individual could have invented — the ratchet effect, in Michael Tomasello’s term, because faithful transmission acts like a ratchet that holds each gain in place while the next is added. A modern smartphone, a legal system, or a mathematical proof is the endpoint of thousands of such small, retained improvements.

The ratchet has two requirements, and both are demanding. Transmission must be high-fidelity — faithful enough that improvements are preserved rather than degraded on each copy, which is why true action-copying imitation, not mere emulation, is thought to be the critical ingredient. Direct evidence for that human bias comes from over-imitation: shown a demonstration that includes causally irrelevant steps, children faithfully reproduce even the pointless actions, whereas chimpanzees prune them and emulate only the goal-relevant outcome (Horner & Whiten, 2005) — the young of our species err toward copying actions, the very bias cumulative culture exploits. And there must be innovation to ratchet: variation on which the retention can act. Tomasello argued that what distinguishes human culture is not raw intelligence but shared intentionality — the capacity to attend to a problem together, which supports teaching, and high-fidelity imitation. Other species have traditions, but their cultures rarely ratchet: chimpanzee foraging techniques persist without steadily accumulating complexity.

That other species have traditions at all was itself a hard-won finding. Andrew Whiten and colleagues (1999) documented dozens of behavior patterns — tool use, grooming, courtship — that vary from one chimpanzee community to another in ways that ecology and genetics cannot explain, and are best understood as socially transmitted chimpanzee cultures. Two decades of subsequent work extended socially learned traditions across mammals, birds, fish, and insects, so that culture is now recognized as a widespread feature of animal life (Whiten, 2021) — even as the cumulative, ratcheting form remains conspicuously elaborated in our own species.

Worked Example

The ratchet’s power is easiest to see as a simple recurrence. Suppose each generation of learners inherits the best solution the previous generation achieved, and improves on it by a fixed fraction — but only if transmission is faithful enough to preserve what was inherited. Let V_n be the value (say, the efficiency of a tool) at generation n, let g = 0.15 be the proportional improvement an innovating generation adds, and let f be the fidelity: the probability that the inherited value survives to be built upon.

With perfect fidelity (f = 1), each generation retains all of the last and adds 15%, so V_n = V₀ · (1.15)ⁿ — exponential accumulation:

- Generation 0: V = 1.00 - Generation 1: 1.00 × 1.15 = 1.15 - Generation 2: 1.15 × 1.15 = 1.32 - Generation 5: 1.00 × 1.15⁵ = 2.01 (the tool has doubled) - Generation 10: 1.00 × 1.15¹⁰ = 4.05

Now impose imperfect fidelity. Model each generation as retaining a fraction f of the inherited value before adding its 15% innovation: V_n = f · V_n−1 + g · f · V_n−1, i.e. V_n = V_n−1 · f · (1 + g). At f = 0.90 the per-generation multiplier is 0.90 × 1.15 = 1.035, so value still grows, but reaches only 1.41 by generation 10 — a third of the faithful case. At f = 0.85 the multiplier is 0.85 × 1.15 = 0.9775, below one: the culture now loses ground each generation and the ratchet runs backward, decaying to 0.80 by generation 10 despite every generation innovating exactly as much as before.

The lesson is that innovation is not the bottleneck — fidelity is. A modest change in copying accuracy flips accumulation into decay, which is why the mechanistic question of Section 3 (emulation versus true imitation) turns out to decide the macro-scale question of whether a species has cumulative culture at all. The interactive ratchet demonstration below varies g and f and shows the accumulation curve crossing from growth into collapse.

Discussion

Social learning occupies an unusual position in cognitive psychology: it is simultaneously a topic in the psychology of the individual and the engine of a supra-individual process, culture. The two levels are tightly coupled. Whether a population accumulates culture depends on a fact about individual minds — how faithfully they copy — and, conversely, the strategies an individual deploys are shaped by the cultural environment those minds have built. Heyes’s cognitive-gadgets proposal pushes the coupling to its limit, arguing that even the mechanisms of social learning are cultural products, learned anew by each generation rather than inherited as instincts.

The field’s progress has largely come from dividing what once looked unitary. “Imitation” split into a ladder of mechanisms; “copying” split into a space of strategies indexed by when and whom; “culture” split into the widespread capacity for tradition and the far rarer capacity for cumulative improvement. Each division replaced a yes/no question (can animals imitate? do animals have culture?) with a graded, mechanistic one, and in each case the graded question proved more tractable and more revealing.

Social learning also connects outward to neighboring constructs. It presupposes attention to the right model, reinforcement in its vicarious form, and, for the highest-fidelity human variety, a theory of mind that lets a learner represent what a model intends. Bandura’s social cognitive theory folded these together into an account of human agency in which people are neither driven by inner drives nor shaped passively by the environment, but learn, symbolically and vicariously, from a social world they in turn help make.

Current Directions

The most active current thread treats human social learning as metacognitive and selective rather than automatic. Building on the strategies framework, recent work asks how learners decide whom to trust — weighting a model’s past accuracy, confidence, and similarity to themselves — and models this as a rational, evidence-weighing computation rather than a fixed bias. Kendal and colleagues’ (2018) synthesis of transmission biases is representative: it reframes “who copies whom” as a set of empirically separable, individually measurable policies that link developmental psychology, animal behavior, and cultural evolution under one account, and it has driven a wave of experiments isolating each bias.

A second front is the reach of animal culture. Whiten’s (2021) survey documents socially transmitted traditions in an ever-widening range of taxa, sharpening — rather than dissolving — the puzzle of why cumulative culture is so unevenly distributed. If tradition is common but the ratchet is rare, the explanatory weight falls on whatever supports high-fidelity transmission and cumulative innovation: teaching, language, shared intentionality, or the culturally built imitation mechanisms Heyes describes. Disentangling which of these is necessary, and which merely helpful, is now the field’s central open question.

Common Misconceptions

“Social learning is just imitation.”

Imitation — copying a model’s specific actions — is only the highest rung of a ladder that also includes local and stimulus enhancement and emulation, in which no action is copied at all. Much apparent imitation in animals is really enhancement or emulation (Heyes, 1994).

“Observing a behavior means it will be performed.”

Learning and performance are separable. Children who watched a punished model had learned the behavior just as well as others, but performed it only when later given an incentive to do so (Bandura, 1965).

“More copying is always better.”

Indiscriminate copying spreads out-of-date and maladaptive behavior. A strategy tournament found that winners copied selectively — when asocial learning was costly and from high-payoff models — not constantly (Rendell et al., 2010).

“Animals have culture, so human culture is nothing special.”

Many species have socially transmitted traditions, but these rarely accumulate complexity across generations. The cumulative, ratcheting culture that builds smartphones and legal systems remains conspicuously elaborated in humans (Whiten et al., 1999).

Glossary

Asocial learning.

Learning a contingency through one’s own direct experience — trial and error — rather than from others; the contrast class to social learning.

Cognitive gadgets.

Heyes’s proposal that the mechanisms of social learning, including imitation and mindreading, are built during development by general learning and cultural experience rather than being genetically dedicated instincts.

Cumulative culture.

The accumulation of behavioral improvements across generations through faithful transmission and repeated innovation, yielding artifacts no individual could invent alone.

Emulation.

Social learning in which the observer copies the goal or result of a behavior, achieving it by any means, without copying the model’s specific actions.

Fidelity.

The accuracy with which a behavior is reproduced on transmission; high fidelity is the precondition for cumulative culture, since low fidelity degrades inherited improvements faster than innovation adds them.

Imitation.

Social learning in which the observer copies the specific actions a model used, reproducing the form of the behavior and not merely its outcome; also called true imitation.

Local enhancement.

The lowest-level social influence, in which a model’s presence draws an observer’s attention to a particular location, where it then learns for itself.

Matched-dependent behavior.

Miller and Dollard’s term for imitation treated as ordinary reinforced behavior, in which matching a model is learned because it is rewarded.

Over-imitation.

The tendency, marked in young children, to copy a model’s actions faithfully even when some are causally irrelevant to the outcome; chimpanzees instead drop the superfluous steps and emulate only the goal.

Ratchet effect.

Tomasello’s metaphor for how faithful transmission holds each cultural improvement in place while further improvements are added, preventing backward slippage.

Shared intentionality.

The capacity to attend to a goal jointly with others and pursue it collaboratively; Tomasello’s candidate for what makes human culture cumulative.

Social learning strategies.

Rules specifying when an individual should copy (e.g. when asocial learning is costly) and whom it should copy (e.g. the majority, or successful individuals).

Stimulus enhancement.

Social influence in which a model draws the observer’s attention to a kind of object or stimulus, rather than a place, after which the observer learns the task itself.

Transmission bias.

A systematic tendency in whom or what is copied — payoff-based, frequency-based (conformist), or model-based (prestige, age, similarity) — that shapes which variants spread through a population.

Vicarious reinforcement.

A change in an observer’s behavior produced by witnessing the consequences — reward or punishment — of a model’s behavior, rather than by direct reinforcement of the observer.

Key Researchers

Albert Bandura (1925–2021). Originated social learning theory, later social cognitive theory; the Bobo-doll studies are its founding demonstration that observation alone can install new behavior. Wikipedia - Wikidata - Stanford Faculty - Google Scholar

Cecilia Heyes (b. 1960). Distinguishes the mechanisms of social learning from its inputs; her cognitive-gadgets account treats imitation and mindreading as culturally inherited rather than instinctive. ORCID - Wikipedia - Wikidata - Oxford Faculty

Kevin N. Laland (b. 1962). Formalized social learning strategies — the when- and whom-to-copy rules that make copying adaptive; now publishes as Kevin N. Lala. ORCID - Wikipedia - Wikidata - St Andrews Faculty

Neal E. Miller (1909–2002). With Dollard, gave the first systematic learning-theory treatment of imitation, casting it as instrumentally reinforced matched-dependent behavior. Wikipedia - Wikidata

Michael Tomasello (b. 1950). Cultural-learning account of human cognition: imitation plus shared intentionality supports the cumulative “ratchet” that other apes lack. ORCID - Wikipedia - Wikidata - Duke Faculty

Andrew Whiten (b. 1948). Provided field and experimental evidence for socially transmitted traditions in chimpanzees and, more broadly, the widening reach of animal culture. ORCID - Wikipedia - Wikidata - St Andrews Faculty

Frequently Asked Questions

What is social learning?
Social learning is acquiring behavior, knowledge, or skill by observing or interacting with others, rather than by discovering it through one’s own trial and error. It lets a behavior spread through a population without every learner paying the cost of inventing it.

How is social learning different from imitation?
Imitation — copying a model’s specific actions — is just one, high-level form of social learning. Others, such as local and stimulus enhancement and emulation, transmit only where to look or what goal to pursue, without copying any action (Heyes, 1994).

What were the Bobo-doll studies?
In Bandura’s experiments, children who watched an adult model behave aggressively toward an inflatable doll later reproduced that aggression, including novel acts, showing that observation alone can install a new behavior without reinforcing the observer (Bandura, Ross, & Ross, 1961).

What is vicarious reinforcement?
It is a change in an observer’s behavior caused by witnessing the consequences of a model’s behavior. Children who saw a model punished performed fewer of its actions spontaneously, but had still learned them (Bandura, 1965).

What are social learning strategies?
They are rules governing when to copy and whom to copy — for example, copy when individual learning is costly, or copy the majority or the most successful. A tournament found that copying selectively, not constantly, wins (Rendell et al., 2010).

Do animals have culture?
Many species show socially transmitted traditions that vary between groups for reasons ecology and genetics cannot explain, such as the community-specific behaviors of chimpanzees (Whiten et al., 1999). What is rare outside humans is cumulative culture that accumulates complexity over generations.

What is cumulative culture?
It is the accumulation of improvements across generations, made possible by transmission faithful enough to preserve gains while new ones are added — the ratchet effect. It is thought to require high-fidelity imitation and shared intentionality.

Why does copying fidelity matter so much?
Because cumulative culture only grows if inherited improvements survive to be built upon. A small drop in fidelity can flip accumulation into decay, so the accuracy of copying, not the rate of innovation, is usually the limiting factor.

References