What the science says

Consensus

Social Cognitive Theory is best understood as Bandura's answer to a lopsided debate. Behaviourism said the environment shapes the person; other traditions emphasised inner drives. Bandura's move, set out fully in his 1986 book Social Foundations of Thought and Action, was to say all three legs, the person, their behaviour, and their environment, act on each other at once, a mutual influence he named reciprocal determinism. You are shaped by your surroundings, but your behaviour also selects and changes those surroundings, and your beliefs steer both. This is the integrative claim, and it is why the theory sits at the framework level: it tries to hold cognition, learning, and social context in one picture.

‍

Two components inside that picture are where the durable science lives. Observational learning is the finding that people acquire behaviour by watching others, without needing to be rewarded for it themselves. The famous demonstration is the Bobo doll experiment, in which children who watched an adult attack an inflatable doll went on to imitate that aggression, while children who had not seen it largely did not (Bandura, Ross and Ross, 1961). Bandura broke the process into four sub-steps that still hold up as a checklist: you have to attend to the model, retain what you saw, be able to reproduce it, and be motivated to do so.

‍

The centrepiece, though, is self-efficacy: your belief in your capability to organise and carry out the actions a specific task requires (Bandura, 1977). It is not global self-esteem and not a personality trait; it is domain-specific, so you can have high efficacy for public speaking and low efficacy for swimming. Bandura's most practically useful claim is that efficacy beliefs are built from four sources, in rough order of power: mastery experiences (actually succeeding at the thing), vicarious experience (seeing people like you succeed), verbal or social persuasion (credible encouragement), and physiological and emotional states (reading a racing heart as either readiness or dread). Between people, higher self-efficacy reliably goes with better performance; the classic meta-analysis of 114 studies found a weighted average correlation of about .38 with work performance (Stajković and Luthans, 1998), a solid effect by the standards of the field. Bandura later extended the idea to groups, arguing that a shared belief in a collective's joint capability, collective efficacy, shapes how much a team or community will attempt and endure (Bandura, 2000).

Controversies

The largest live argument is not about whether self-efficacy correlates with performance, which it clearly does, but about which way the arrow points. Almost all the reassuring evidence is between-person: people who are more confident tend to do better. But that is exactly what you would see if confidence were mostly a consequence of past success rather than a cause of future success. When researchers switched to within-person designs, tracking the same individual's confidence and results over time, the picture changed sharply. Vancouver, Thompson and Williams (2001) found that at the within-person level self-efficacy could actually relate negatively to subsequent performance, in effect that feeling on top of a task can breed a little complacency. A later meta-analysis pulled the many within-person studies together and found that once you account for a person's performance trajectory, the self-efficacy to later-performance relationship shrank to roughly zero and non-significant, while the reverse path, past performance predicting your later self-efficacy, stayed strong; the authors concluded self-efficacy is "primarily a product of past performance rather than the driving force affecting future performance" (Sitzmann and Yeo, 2013). This does not erase self-efficacy, but it seriously qualifies the folk version in which believing harder makes you do better.

‍

A second, older dispute is conceptual. Bandura's theory holds that your sense of capability (self-efficacy) is distinct from, and causally prior to, your expectations about outcomes. Williams (2010) argued there is an unresolved contradiction in how the theory handles this, and that the field ended up over-focusing on self-efficacy as the causal mover while neglecting outcome expectancies. In follow-up work, Williams and Rhodes (2014) went further, arguing that most standard self-efficacy measures are confounded: they end up capturing not pure perceived capability but a blend that includes motivation, so a lot of "self-efficacy predicts behaviour" findings may partly be motivation predicting behaviour under another name. In short, some of the construct's success may be a measurement artefact.

Limitations

The broader theory has a breadth problem. Reciprocal determinism, that person, behaviour and environment all influence each other, is almost certainly true, but it is so general that it can accommodate nearly any result after the fact, which makes it hard to test as a whole and easy to treat as a label rather than a prediction. The theory is strongest where it is most specific (self-efficacy, the four sources, observational learning) and weakest as a grand unifying claim. Self-efficacy's domain-specificity is also often ignored in practice: a generic "confidence" measure will not predict much, because efficacy only bites when it is tied to a particular task. And there is heavy conceptual overlap with neighbouring frameworks. Perceived behavioural control in the Theory of Planned Behaviour, capability in COM-B, and competence in Self-Determination Theory are all close cousins of self-efficacy, which means practitioners frequently rediscover the same variable under four names.

Open questions

The honest open question is causal: under what conditions is building self-efficacy actually a lever on behaviour, rather than a reading of behaviour that has already happened? The within-person work suggests the effect is at best modest and conditional, so the useful research is about the boundary conditions, novel tasks versus familiar ones, early learning versus mastery, where confidence helps versus where it breeds complacency. A second open question is measurement: can we assess perceived capability cleanly enough to separate it from motivation and outcome expectancy (Williams and Rhodes, 2014)? A third, more promising frontier is collective efficacy: shared beliefs about what a group can do together may carry more weight in collective and community settings than individual efficacy does, and that is comparatively under-studied.

So what

Self-efficacy is real, it is worth building, and Bandura's four sources tell you exactly how to build it, in order of power. The single most important correction to fold in is directional. Because a person's confidence is built largely by their own track record, the highest-return move is to engineer genuine early success, not to talk people into feeling capable. Mastery experiences are the strongest source; credible models who resemble the audience are second; verbal persuasion, the pep talk, is the weakest, and pumping up stated confidence without backing it with real wins does little on its own. So the applied instruction is: create small, real successes, show relatable models succeeding through specified actions, and treat encouragement as a supporting act, not the main lever.

‍

For companies, this is the design manual for any category where the customer has to acquire a skill to get value: fitness, personal finance, cooking, DIY, language learning, any health or self-improvement product. Onboarding should be built to manufacture an early, genuine win rather than to impress, because that win is what builds the efficacy that keeps people going. Models in your communications should look and sound like the target customer and should visibly succeed by doing the specific thing you want the customer to do, which is one of the most evidence-backed creative levers in this whole family. And measure efficacy for the specific behaviour, not generic confidence, if you want it to predict anything.

‍

For political parties and campaigns, the most useful piece is collective efficacy: mobilisation often turns less on individual confidence than on a shared belief that "people like us, acting together, can actually make this happen." Visible models matter here too; seeing relatable others turn out, volunteer or speak up licenses the same behaviour in the observer. The honest caveat is not to oversell it: a feeling of collective capability helps sustain effort, but it does not substitute for a real, winnable ask and the logistics to act on it.

‍

For government and public health, self-efficacy diagnostics belong in the design of any programme that asks citizens to do something difficult and repeated: manage a chronic condition, save, retrain, change a household habit. Build the intervention around mastery, small graded successes with support, rather than around exhortation. Use relatable role models from the target community. But keep the directional caveat in view when you evaluate: a programme that raises stated confidence has not necessarily changed behaviour, because confidence tracks past success as much as it drives future action, so measure the behaviour, not just the belief.

‍

There is a quieter dual-use point worth naming, and it is really about honesty and blame. The folk version of self-efficacy, "just believe in yourself", can be used to launder a structural problem into a personal failing: if outcomes are framed as purely a matter of confidence, then people who fail under real constraints, poverty, precarity, discrimination, a stacked environment, get told they simply did not believe hard enough. Bandura's own theory guards against this, because the environment is a full leg of reciprocal determinism, not a footnote. Responsible use keeps that leg in view: build efficacy where a real path to success exists, and do not sell belief as a cure for a rigged situation.

‍

The summary is that Social Cognitive Theory gave us a genuinely useful engine, self-efficacy, and a clear recipe for building it, wrapped inside a grand claim (reciprocal determinism) that is true but too broad to do much predictive work. Keep the engine and the recipe. Build capability through real mastery and relatable models, treat encouragement as support rather than substitute, remember that confidence is partly a mirror of past success and not only a driver of future success, and never let "believe in yourself" stand in for changing the conditions people are actually up against.

‍