Updated on
September 23, 2026
Bandura's Social Learning Theory: Evidence and Classroom Use
Bandura's social learning theory explained: the four processes, what the Bobo doll studies found, and KS2 and KS4 modelling examples with a planner.

Updated on
September 23, 2026
Bandura's social learning theory explained: the four processes, what the Bobo doll studies found, and KS2 and KS4 modelling examples with a planner.
What is Bandura's social learning theory?
Bandura's social learning theory proposes that people can acquire new patterns of behaviour by observing models. Learning depends on attention, retention, behavioural production and motivation, but observing a model does not guarantee later performance. Reinforcement can affect whether an acquired behaviour is performed, including through observed consequences, without being necessary for every act of acquisition.
Bandura's social learning theory is the idea that people learn new behaviour by watching others, a process he called observational learning. Albert Bandura set it out in his 1977 book Social Learning Theory, building on his Bobo doll studies of the early 1960s. A learner has to notice the model, remember it and be able to do it. Motivation then decides whether they use it.
That gap between learning something and showing it is the part teachers most often miss. A learner can watch you balance an equation, follow every step and still leave the page blank. Rewards may change what learners do, but they are not needed for every act of learning.
For teachers, the four processes work as a planning checklist for any demonstration: what to point at, how learners will hold on to it, when they try it and what might stop them. Among the major learning theories, Bandura's is the one that explains why a good model can matter more than a reward chart.
Social learning theory explains how people can acquire behaviour by observing other people and the consequences around them. A model may be present in person, described in words or shown through media. Learners do not reproduce everything they see.
The person and the setting shape what they notice and recall. They also shape what the person later does.
This account challenged the view that a learner must always receive a direct reward to acquire new behaviour. It did not remove consequences from learning. Instead, it split learning from performance. It also gave thought a place while still recognising the role of the setting (Bandura, 1977).
In class, a teacher can solve an equation while explaining each choice. Learners can mark the main decisions before trying a related problem. The model gives them ideas. Their notes and first attempt show what they retained and can produce.
Bandura's theory grew from early studies of imitation into a wider account of learning, thought and action. His 1977 social learning theory linked models, thought and reinforcement. By 1986, his social cognitive theory placed more weight on self-regulation, self-efficacy and human agency.
This timeline stops some common errors:
Social learning theory and social cognitive theory are related, but they are not the same. The early theory still helps explain modelling and observational learning. The later theory covers more ground.
For example, watching a classmate show a lab routine is learning from a model. The later theory covers more factors. These include the learner's view of their skill, the lab setting and their own actions.
Bandura's four component processes are attention, retention, behavioural production and motivation. They help explain whether a person learns an observed action and later performs it. People often call them “stages”, but they work together rather than forming a fixed path.
The first two processes decide whether anything is learned at all. The last two decide whether that learning shows up in what the learner does, which is why a silent learner is not always a learner who missed it.
The learner must notice the key parts of the model's action. Attention is selective. A clear model can help. So can the learner's goals, prior knowledge and setting.
Eye contact is not required for learning. A teacher can guide attention by naming the feature to watch: “Notice where I place the evidence before I explain it.” This gives learners a clear focus. It avoids assumptions about how attention must look.
The learner must hold what happened in a form they can recall. Words, images and rehearsal may help. Watching alone does not show what entered memory.
After modelling a paragraph, ask learners to rebuild its plan from memory with a partner. They can then compare their version with the original. This is a small act of retrieval practice, and it checks retention instead of assuming that quiet observation was enough.
The learner must turn the memory into action. Bandura used terms such as motor reproduction and behavioural production as his account grew. Knowing what to do does not ensure fluent performance.
Give learners a short, supported attempt straight after the model. In art, they can try one brush technique on scrap paper. The teacher can then give precise feedback. The attempt shows where learners need more help or practice.
Motivation affects whether someone performs a learned response. Expected results, observed consequences, personal standards and value may all affect action. This process means more than offering a reward.
A learner may understand a discussion routine but avoid using it when the social risk feels high. The teacher can lower that risk through pair rehearsal before whole-class talk. This changes the conditions for performance. It does not prove that motivation was the only cause.
Core idea: People can learn by observing models. Observation does not guarantee imitation.
Four component processes: Attention means noticing the relevant action. Retention means encoding and remembering it. Production means turning the representation into action, where ability and practice matter. Motivation covers the expected outcomes, goals and values that influence performance.
Attention and retention support acquisition. Production and motivation affect whether the learning is performed.
Vicarious reinforcement: Observed consequences can change the likelihood of performing an observed response. Direct reward is not required for acquisition.
Self-efficacy: This is a judgement of capability for a specific performance, not general confidence.
Triadic reciprocal causation: Personal factors, behaviour and environment influence one another.
What the 1961 Bobo study showed: Bandura, Ross and Ross studied 72 nursery-school children aged 37 to 69 months. In a laboratory task, children exposed to aggressive models produced more imitative aggressive acts. This does not establish long-term, real-world violence. The setting was artificial and the outcome was immediate behaviour.
Classroom use: Model the process and decisions aloud. Use credible, varied models. Follow observation with guided practice and feedback. Build self-efficacy chiefly through genuine mastery, then test independent transfer.
Memory line: Observe, encode, practise, choose.
Sources: Bandura, Ross and Ross (1961); Bandura (1977, 1986); Usher and Pajares (2008).
The four processes are easiest to plan with when you see them in one lesson from start to finish. Here is the same planning run through twice, once in a Year 4 maths lesson and once in a Year 11 chemistry practical.
A Year 4 class is meeting column subtraction with exchange for the first time. The teacher models 352 minus 127 under the visualiser.
Attention. Before she starts, she says: “Watch the ones column. When the top digit is smaller, I exchange one ten for ten ones.” She circles the 2 and the 7, so learners know exactly where to look.
Retention. She works the example, saying each decision aloud, and then covers it. Partners rebuild the steps on mini whiteboards: “Is the top digit smaller? Exchange a ten. Now subtract the ones.” They uncover her example and compare.
Production. Learners try 461 minus 238 on their whiteboards while her worked example stays visible. She scans the boards and gives feedback on the exchange step only, because that is the new part.
Motivation. Whiteboards let everyone show an answer at once, so nobody has to risk being wrong alone. If learners think it is too hard, she models a slip on purpose: she takes 2 from 7 in the ones column, the classic smaller-from-larger error, then stops and corrects it aloud.
Two classic studies back this kind of modelling for primary maths. Schunk (1981) found that children who struggled with division gained more accuracy when a model said the steps aloud than from explanation alone. Schunk, Hanson and Cox (1987) found that children who had struggled with fractions gained more confidence and skill from a coping model, who improved step by step, than from a single model who got everything right at once.
A Year 11 class is preparing for the titration required practical. The teacher models one run at the front, with a camera showing the burette on the screen.
Attention. He names two things to watch: reading the bottom of the meniscus at eye level, and slowing to single drops as the colour starts to linger. Everything else in the method is already on the sheet.
Retention. He stops before the second run. Learners write the steps as a numbered list without looking, then check against the method sheet with a partner. If lists leave out the rough titration, that becomes his next teaching point.
Production. Pairs carry out a rough run and then accurate runs while his setup stays at the front. He circulates and gives feedback on one thing, reading the burette.
Motivation. He shows a set of results from an anonymous past class that do not agree within 0.10 cm³ and asks what went wrong. Learners see the consequence on someone else's data, a small and safe version of the observed consequences Bandura studied.
In both lessons the teacher checks each process before moving on. The I do, we do, you do routine follows the same order, and the guide to modelling in the classroom covers the demonstration itself in more depth.
Use the planner below to build the same four moves for the next skill you will model. Type the skill, pick your key stage and print the card.
The planner gives a starting point, not a script. Watch what learners actually do in the production step and adjust the next model to match what they missed.
The Bobo doll studies found that children copied more of a model's distinctive aggressive acts after seeing the model. The phrase “Bobo doll experiment” often blends three studies into one story. The 1961, 1963 and 1965 studies used varied models, methods and consequences. Their findings must therefore remain separate.
Bandura, Dorothea Ross and Sheila Ross studied 72 children, 36 boys and 36 girls aged 37 to 69 months, from the Stanford University Nursery School. Each child saw one of three conditions. These were an aggressive adult model, a non-aggressive adult model or no model. The model was present in person, not shown on video.
In the aggressive condition, the adult spent most of a ten-minute session using distinctive physical and verbal actions towards a five-foot inflatable Bobo doll. The children later entered a room with aggressive and non-aggressive toys. Those who saw the aggressive model copied more of its distinctive aggressive acts than children in the other conditions (Bandura et al., 1961).
The controls showed virtually no imitative aggression. This does not mean they showed no aggression at all. The study measured short-term behaviour in a set lab task.
It followed an adult model and deliberate frustration. It did not establish lasting aggression in ordinary life.
The study also did not measure the four processes. Attention, memory and choice are ideas used to explain the pattern. They were not separate measures in 1961.
Teachers can use the study to discuss research design. Ask which behaviours counted as direct imitation and which counted as other aggression. Then ask what the controls allow us to infer. This is more accurate than saying the study “proved children copy adults”.
A later study of 96 children compared a live aggressive model with the same model on film and with an aggressive cartoon character. Children in all three conditions later showed more aggressive behaviour than controls. The filmed-model work belongs to 1963, not the original 1961 report (Bandura et al., 1963).
The result supports the idea that models in media can affect behaviour in the study setting. It does not show that every video teaches well. Nor does it show that media decides a child's behaviour. Content, context, prior experience and later results still matter.
In class, pause a video model at a key decision. Ask learners what the model did and why. Their answer checks attention and interpretation. The video alone is not evidence that learning occurred.
Bandura's 1965 study showed children a filmed model. The model's aggressive acts were rewarded, punished or followed by no consequence. These outcomes affected how much children copied when first tested. Children who had seen punishment performed fewer modelled acts at that point.
The researchers then offered rewards for copying the model's actions. In Bandura's own summary, the differences between the groups disappeared. This result supported a difference between learning and performance. Children could learn a response without showing it in the first test (Bandura, 1965).
A no-consequence condition is not vicarious reinforcement. Vicarious consequences are outcomes that someone sees happening to another person. Those outcomes may then affect the learner's expectations or actions.
Any classroom comparison needs care. A learner may recall a shown method but avoid using it in public. A private, low-stakes check may help separate “did not learn” from “did not perform here”. But it cannot identify the cause by itself.
Acquisition means learning an observed pattern. Performance means carrying it out. The two can happen at different times. Reinforcement and other consequences can change performance without being required for every act of learning.
This difference is one of Bandura's key contributions.
Direct reinforcement follows the learner's own action. Vicarious consequences are results that the learner sees happening to someone else. Self-produced consequences involve personal standards and self-checks. None of these terms simply means “motivation”.
This is also where Bandura differs from a simple caricature of behaviourism. Classical conditioning is not a theory of rewards and punishments. Skinner's operant account is also more complex than rewarding every correct act.
Bandura still focused on the setting and behaviour. He also added mental models, expectations and self-regulation.
Suppose learners watch a teacher balance a chemical equation. One learner can explain each step but leaves the answer blank during public questioning. This suggests acquisition without current performance. It does not reveal whether anxiety, doubt, social conditions or another factor blocked the response.
Bandura's later work went beyond modelling and reinforcement. Social cognitive theory says people help shape their own growth. They do not just receive influence from the setting. The theory studies how personal factors, behaviour and environment interact.
Bandura called this triadic reciprocal causation, and earlier reciprocal determinism (Bandura, 1978). This means that three sets of factors can affect one another. Their effects need not be equal or happen at the same time.
A learner's beliefs may affect whether they try a task. Their attempt changes the environment. Other people's responses may then affect later beliefs and actions.
Self-efficacy is a judgement about one's ability to organise and carry out action in a specific task or field. It is not general confidence, praise or a promise of success (Bandura, 1977, 1982). A learner may have high efficacy for mental arithmetic but low efficacy for explaining mathematical reasoning.
The social cognitive theory article covers reciprocal causation, self-efficacy, self-regulation and agency in depth, and the guide to self-regulated learning shows how learners start to manage their own practice. Here, these ideas show how Bandura's theory grew. They do not belong in the 1961 experiment as if that study measured them.
Bandura's theory helps teachers plan what learners need to watch, recall and practise. It does not prove that one class method will raise attainment. Each practice claim needs evidence that matches the subject, age group and use.
Show the action and explain the key choices. In history, say: “I am using this source to support the claim because the author saw the event. But I still need to test the author's motive.” Learners can then mark the claim, evidence and caution in another example.
EEF evidence asks teachers to model their thinking aloud during real curriculum tasks. This supports the class practice. Bandura's theory offers one account of why watching a model's thinking may help.
Teachers on social media describe the same moves in their own words. The table sets two routines they shared this month against the research.
Two routines teachers described on social media this month, set against the four processes. The routine and notice columns are our summary of each post; the evidence column is the research, not the post.
| Routine, as teachers describe it | What you would notice | What the evidence says |
|---|---|---|
| Live shared writing: make and fix choices aloud | ||
Watching an adult make and fix real choices in front of them tells students more about writing than any finished model on a handout. |
Learners see the decisions and the corrections, not only the finished paragraph. That is the attention and retention half of Bandura's model done in public. | Children who had struggled with fractions gained more confidence and skill from a coping model, who improved step by step, than from a model who got everything right at once (Schunk et al., 1987). |
| A mirrored screen so every learner sees the model | ||
Pupils can pinch to zoom in/out, still see live modelling via visualiser etc |
A learner who cannot see the front of the room clearly follows the live model on an iPad and zooms in on the part they need. | No study here tests a mirrored screen. In a randomised study, trainee teachers who watched a video model used retrieval practice better than those who reread written guidance (Sims et al., 2026). |
Neither routine has been tested as it stands. Each is a place to run the four checks: what learners watched for, what they kept, what they could do and what held them back.
Tell learners what to watch. Avoid giving them a long list. In physical education, the focus can be one foot's position at the point of contact.
Ask learners to name the feature before they try the action. This makes the attention check visible. It does not treat eye contact or stillness as proof.
Watching is not a substitute for doing. Give learners a short attempt while the model is still available. Then give feedback on the feature that matters. Use the scaffolding guide when support must reduce across several attempts.
For example, model how to combine sentences. Leave the two main prompts visible while learners combine one pair. Remove one prompt only after the response is secure.
More demonstration is not always better. In a trial with 73 medical undergraduates, extra coaching during hands-on practice beat extra modelling on some measures (Darici et al., 2025). A review of physical education studies found strong evidence that observation helps motor skills, when it sits alongside practice (Han et al., 2022).
How interactive a model should be is a live argument among teachers. Some want the model to stay with the teacher; others want the class writing alongside.
Four short posts from social media on the same question: should the model be the teacher's alone, or shared with the class as it happens?
I hardly ever see I do we do you do. It's almost always we do we do we do.
if you are not doing interactive shared writing then you aren't teaching writing.
"I do, we do, you do" is an annoying name but it works well for a certain profile of pupil who'd otherwise do nothing.
What are the students doing when you are modelling on the visualiser to maintain focus….
One way to read this through Bandura: keep the first model clean so attention stays on the key step, then share the next one, so learners produce it while you can still see what they kept.
A peer model is not always better than a teacher. The model must be sound, and the task must suit peer support. The learner also needs a clear focus.
EEF's peer-tutoring review reports positive average effects. But it stresses training, structured tasks and careful pairing.
Who the model is matters too. In one experiment with 157 adolescents, adult models were more effective to learn from than peer models when the content was identical (Hoogerheide et al., 2016). Learners also judge themselves against classmates, which is the focus of social comparison theory, so a peer model who seems far ahead can discourage as much as inspire.
Ask a learner to show a familiar checking routine. Do not ask them to introduce content they have not mastered. Give the class two questions: “Which check did the model use?” and “Where did it change the answer?”
Give learners more than one way to show what they learned. A private written response, paired account or hands-on attempt may reveal knowledge that whole-class questions miss. Do not assume that every non-response has the same cause.
Record the task and the response. Then change one condition. This gives teachers a better basis for decisions than labelling a learner as unmotivated.
The theory's main strength is that it explains how new behaviour can appear without much direct reinforcement for the learner. It also treats models as sources of knowledge, not just triggers. The difference between acquisition and performance remains valuable.
But the early Bobo studies were short laboratory studies. They examined a narrow set of aggressive play behaviours. They cannot establish the long-term effect of everyday models on personality, attainment or conduct. Ethical concerns and limits on generalisation must remain clear.
Critics also argue that the theory says little about biological factors and developmental readiness, and that it is hard to predict what a learner will copy when they see many models with mixed consequences. Watching a long or complex model can also overload a novice, which is why cognitive load theory favours short models broken into steps.
People sometimes apply the theory too widely. These corrections keep its claims in proportion:
In class, use the theory as a planning lens, not a diagnosis. If a learner does not copy a model, check what they noticed, remembered and could produce. Then consider the wider task and social conditions before reaching a conclusion.
Bandura's theory predicts that a well-planned model helps, and researchers have tested that directly. The evidence check below gathers controlled studies of whether watching a model helps learners pick up a new skill.
Bandura's theory is one of several you can draw on when planning; the learning theories overview sets it beside behaviourism, constructivism and cognitivism. Next lesson, pick one skill you will model, name the one thing learners must watch, and plan a quick check for each of the four processes before you start.
Attention, retention, production (often called reproduction) and motivation. Textbooks call them stages, but Bandura described processes that interact rather than steps that run in order. A learner can watch closely and remember a model, then still not perform it until their skill or reasons allow.
It did not prove a universal law. The 1961 study found that children who saw a live aggressive model copied more of its distinctive aggressive acts in a short laboratory test. Later studies examined filmed models and observed consequences.
Vicarious reinforcement occurs when an observer sees another person's behaviour followed by a favourable result. This affects the observer's expectations or performance. Seeing behaviour with no consequence is not vicarious reinforcement.
Social learning theory centres on observational learning, modelling, thought and reinforcement. Social cognitive theory is Bandura's later and broader account. It covers personal factors, behaviour and environment. It also gives more weight to self-efficacy, self-regulation and agency.
Guide attention towards a clear model. Explain the important decisions. Ask learners to rebuild what they observed, then provide guided practice.
Check learning apart from public performance. Use subject evidence when choosing the teaching method.
Albert Bandura, a psychologist at Stanford University. His imitation studies began in the early 1960s, and he wrote about learning by imitation with Richard Walters in 1963. Social Learning Theory followed in 1977, and he broadened it into social cognitive theory in 1986.
Its strengths are that it explains learning without direct reward and separates learning from performance. Its weaknesses are that the key studies were short laboratory tasks with young children, and that it says little about biology or development. It also struggles to predict which of many models a learner will copy.
No. Vygotsky's theory is a sociocultural account of how talk and guidance from someone more expert shape thinking. Bandura's theory is about learning by observing models and their consequences. Both give other people a central role, which is why they are often confused.
A teacher models a method aloud, such as column subtraction, while pointing to the one step that matters. Learners then rebuild the steps from memory and try a similar example. The teacher checks what they noticed, kept and could do before moving on.
Influence of models' reinforcement contingencies on the acquisition of imitative responses
Bandura (1965). Journal of Personality and Social Psychology, 1, 589-595.
Self-efficacy: Toward a unifying theory of behavioral change
Bandura (1977). Psychological Review, 84, 191-215.
The self system in reciprocal determinism
Bandura (1978). American Psychologist, 33, 344-358.
Self-efficacy mechanism in human agency
Bandura (1982). American Psychologist, 37, 122-147.
Social cognitive theory: An agentic perspective
Bandura (2001). Annual Review of Psychology, 52, 1-26.
Transmission of aggression through imitation of aggressive models
Bandura, Ross and Ross (1961). Journal of Abnormal and Social Psychology, 63, 575-582. Read the primary text.
Imitation of film-mediated aggressive models
Bandura, Ross and Ross (1963). Journal of Abnormal and Social Psychology, 66, 3-11.
Should medical teachers spend more time modelling or coaching students? A dual eye-tracking and randomised controlled study on peer instruction in sonography
Darici et al. (2025). Medical Education, 59, 1105-1116.
Use of observational learning to promote motor skill learning in physical education: A systematic review
Han, Syed Ali and Ji (2022). International Journal of Environmental Research and Public Health, 19, 10109.
Learning from video modeling examples: Content kept equal, adults are more effective models than peers
Hoogerheide, van Wermeskerken, Loyens and van Gog (2016). Learning and Instruction, 44, 22-30.
Modeling and attributional effects on children's achievement: A self-efficacy analysis
Schunk (1981). Journal of Educational Psychology, 73, 93-105.
Peer-model attributes and children's achievement behaviors
Schunk, Hanson and Cox (1987). Journal of Educational Psychology, 79, 54-61.
Modelling evidence-based practice in initial teacher training: effects on teachers' skills, knowledge and self-efficacy
Sims et al. (2026). Instructional Science, 54.
Sources of self-efficacy in school: Critical review of the literature and future directions
Usher and Pajares (2008). Review of Educational Research, 78, 751-796.