A trainee knows the steps of a difficult conversation. Asked to handle an upset customer for the first time, she hesitates. Another trainee with the same instruction tries, adjusts and tries again. What accounts for the difference? Ability, prior experience, the situation and support all matter. Albert Bandura’s 1977 paper Self-Efficacy: Toward a Unifying Theory of Behavioral Change proposed that a person’s belief about being able to perform a specific action is one part of that explanation.
This article is a critical guide to the paper and its research legacy. It is distinct from the Library’s general self-efficacy entry, which explains the concept in current practice. Here the question is what Bandura argued in 1977, what evidence was available, how later studies tested the mechanism and where the theory is often overstated. The paper remains influential because it offers a testable link among perceived capability, effort, persistence and behavior. It does not make belief a substitute for skill or opportunity.
The problem Bandura tried to solve
Behavior-change methods can look very different. A person may practice a feared action directly, watch someone else perform it, hear a persuasive explanation or learn to interpret bodily arousal differently. Bandura asked whether these routes shared a psychological mechanism: they might change expectations about personal efficacy. If a person expects to be able to carry out an action, they may choose to attempt it, invest effort and persist through difficulty.
The distinction is task-specific. “I can speak to this customer using these steps, even if the answer is unfamiliar” is different from “I am a confident person.” It is also different from expecting a favorable outcome. A worker can believe she can submit a complaint accurately while believing the organization will ignore it. Conversely, she can value an outcome without believing she has the means to produce the action. Bandura’s theory concerns the perceived capability to execute a course of behavior under stated conditions.
The 1977 article was a theoretical synthesis drawing on experimental and clinical work, especially treatments for fear and avoidance. It did not directly test every later application in classrooms, organizations, health behavior or digital services. Its originality lay in proposing a common variable and ways to measure changes in it. A source guide should preserve that scope instead of treating every modern “confidence” intervention as a result of the original paper.
Four sources of efficacy information
Bandura proposed four principal sources of information. Performance accomplishments give a person direct evidence from doing the task. A successful attempt under meaningful conditions can make a later attempt more credible. The interpretation of success matters: an easy demonstration with all obstacles removed may tell little about a harder setting, while a manageable challenge with feedback can reveal a route to progress.
Vicarious experience comes from observing others. Seeing a relevant model handle a task can show that it is possible and reveal how. The model’s similarity and the conditions of performance matter. A polished expert demonstration may be less informative to a beginner than watching someone with similar starting skills work through difficulty. Observation also cannot provide all the practice needed to perform a demanding task.
Verbal persuasion includes encouragement and feedback about capability. It may help a person attempt a task or reinterpret a setback, but reassurance unsupported by experience is fragile. A supervisor who says “you can do anything” gives less usable information than one who identifies what the trainee has done well, what remains and what support is available. Persuasion should not conceal a missing skill or unsafe condition.
Physiological states provide cues people may interpret as signs of readiness or inability. A racing heart before a presentation can be read as evidence that the situation is impossible, or as a familiar response to a demanding but manageable task. That interpretation may shape action. It does not mean bodily signals are unreal or always safe to disregard. Pain, illness and exhaustion can signal real limits; context and, when relevant, clinical assessment matter.
These are sources of information, not four guaranteed techniques. A person can learn from practice and still be prevented from acting by lack of authority. A model can inspire while also setting unrealistic expectations. Feedback can be accurate or manipulative. The theory becomes useful when a study specifies what information changed and what behavior followed.
The original experimental setting
Bandura and colleagues’ 1977 experiment on cognitive processes mediating behavioral change tested adults with phobias. Participants received treatments emphasizing direct mastery or vicarious experience, or no treatment. The researchers assessed efficacy expectations and approach behavior. They reported stronger and more generalized efficacy expectations after mastery-based treatment than after vicarious treatment alone, and a relationship between efficacy judgments and performance across tasks of varying difficulty.
That study supports the theory in a specific setting. It does not show that every change in belief caused every change in behavior independently of the treatment. Direct practice changes more than a judgment: it can also change skill, familiarity, arousal and the meaning of the feared situation. When a treatment moves several variables together, mediation needs careful testing. The original research was important because it tried to measure expectations at a fine-grained task level, not because it eliminated all alternative explanations.
The feared tasks were also comparatively concrete. A person could be asked whether they expected to approach an object at a certain distance and then be observed doing so. Many organizational or educational tasks are less tightly specified. “Leading well” or “being creative” may contain many different actions and outcomes. A self-efficacy question becomes harder to interpret when the task boundary disappears.
Belief, behavior and the direction of cause
It is easy to find that people who perform well report higher efficacy. The difficult question is whether the belief helped produce performance, performance strengthened the belief, or both were shaped by a third factor such as preparation or support. Bandura’s model allows reciprocal relations. A successful attempt can increase efficacy; stronger efficacy can make another attempt more likely. A study needs time ordering and, when possible, interventions that separate these paths.
Sheeran and colleagues’ 2016 meta-analysis examined experimental tests in health behavior where attitudes, norms or self-efficacy were changed. It found that experimentally induced changes in those cognitions were associated with later changes in intention and behavior across the included studies. That result strengthens a causal interpretation relative to simple correlations. It is still a synthesis across varied behaviors and interventions; an average effect does not establish that a particular workshop or app will work for every person.
A 2017 smoking-cessation experiment by Shadel and colleagues illustrates a further complication. Participants were given false feedback about their chances of quitting. Those told they had higher chances were more likely to achieve an initial 24-hour quit outcome, but the authors reported that the effect was not explained by measured changes in self-efficacy. Changed outcome expectancies appeared to matter. The lesson is not that self-efficacy is irrelevant. It is that an intervention can shift several beliefs at once, and a successful behavior change does not prove the intended mediator.
For an ethical Institute application, false feedback would be unacceptable as a practical technique. The study is useful here as a test of causal explanation. A claim that “confidence caused success” needs evidence that the relevant capability belief actually changed, that the measurement matched the task and that plausible alternative pathways were addressed.
Why context cannot be omitted
A person’s sense of capability is partly a judgment about conditions. Can the task be done with the available time, tools, permission and help? A trainee may accurately believe she could manage a difficult conversation if a supervisor is available, while doubting she could do it alone at a crowded desk. Treating the latter doubt as a personal deficit would erase the structure of the task.
Outcome expectations matter too. A worker may be fully able to document a safety problem yet know that raising it may bring retaliation. A student may be able to solve a problem yet receive no accessible route to demonstrate the answer. An intervention that only urges belief can fail because the obstacles are external. Bandura’s framework distinguishes capability expectations from expected outcomes, and that distinction should be used when designing services or learning opportunities.
The difference between a realistic and an inflated judgment also matters. A learner who underestimates ability may avoid a useful opportunity. One who overestimates ability may act without preparation in a high-stakes setting. The goal is calibrated efficacy supported by evidence, practice and feedback. That means measuring actual performance alongside perceived capability rather than celebrating a survey score alone.
Zimmerman’s review of self-efficacy in learning emphasizes that efficacy is sensitive to changes in performance context and differs from broad self-concept. A student may feel capable of revising a paragraph but not of writing an entire report under time pressure. An educator can use that specificity to design practice. The review does not imply that motivation replaces instruction, resources or a fair assessment environment.
How to read an efficacy study
First inspect the wording of the efficacy measure. Does it ask whether participants can perform a defined action under specified conditions, or whether they feel generally positive? “How confident are you?” with no task description can mix ability beliefs with mood, social desirability and expected rewards. A strong measure maps onto the behavior and its obstacles.
Next inspect the timing. Was efficacy measured before the behavior, after it or both? If belief and performance are collected at one time, causal direction remains open. If the intervention changes belief before an observed behavior, the inference strengthens, but the intervention may also have supplied training or new tools. The study should report those co-occurring changes.
Then inspect the outcome. Is it an observed action, a self-report of intended action or a distant result influenced by many conditions? A person can perform an action well but fail to obtain a reward controlled by someone else. A study that treats the reward as the sole behavior outcome may underestimate capability or misattribute organizational barriers.
Finally ask who was included. Findings from one clinical fear task, one school level or one health behavior do not automatically generalize to different ages, cultures or power relations. The paper’s framework can generate hypotheses across settings, but each application needs its own design, validation and ethical safeguards.
Applications in teaching and organizations
The Institute’s synthesis is to create conditions in which efficacy can become well founded. Give a learner a task with clear success criteria, an appropriately challenging first step, a relevant model, actionable feedback and a chance to try again. Measure both the learner’s judgment and actual performance. If the learner still cannot act, examine the tools and authority available, not only the person’s belief.
In a workplace, this can mean supervised practice before independent responsibility. For a digital service, it can mean a preview, clear error recovery and a human support route. These are design hypotheses inspired by the theory. The 1977 paper did not test a modern service interface, and a usable interface should be evaluated by completion, comprehension and repair as well as by a confidence question.
The theory can also prevent misplaced blame. If people do not attempt a change, ask whether they believe they can perform the action, whether they expect the action to lead anywhere and whether the setting makes the action feasible. Those questions point to different remedies. Training may help one problem; policy or service redesign may be required for another.
What the evidence does not show
The 1977 paper does not show that belief alone changes behavior under every condition. It does not establish a stable, general “confidence trait,” nor does it justify pressuring people to persist in unsafe or impossible tasks. Its theory is about expectations for specified actions and their relationship with effort and performance. Real skill, resources and outcomes still matter.
Later experimental and meta-analytic evidence supports self-efficacy as a useful construct in some domains, while individual studies show how difficult it is to isolate the proposed mechanism. A program that increases a survey score but not performance has not demonstrated capacity development. A program that changes behavior through training, incentives or access should not attribute the entire gain to belief without testing that claim.
Nor is a lower efficacy judgment always wrong. It may be a realistic assessment of missing preparation or obstructive conditions. The ethically useful response is to understand the judgment, offer opportunities to build capability and alter barriers where possible. Accurate self-knowledge is more valuable than compulsory positivity.
Return to the trainee
The trainee practices the conversation with a model, handles a smaller real case and receives feedback on the moment she lost the thread. Her estimate of what she can do may change because she now has evidence. If the organization then removes the supervisor and doubles the workload, her estimate may change again, also for good reason.
Bandura’s paper remains valuable because it asks researchers and practitioners to name that specific judgment and test its role in action. It is most faithful to the theory to ask what someone can do, under which conditions, with which evidence and support. A belief about capability becomes useful when it stays connected to the task and to the world in which the task must be done.
Research references
- Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84(2), 191–215. View source
- Bandura, A., Adams, N. E., & Beyer, J. (1977). Cognitive processes mediating behavioral change. Journal of Personality and Social Psychology, 35(3), 125–139. View source
- Zimmerman, B. J. (2000). Self-efficacy: An essential motive to learn. Contemporary Educational Psychology, 25(1), 82–91. View source
- Sheeran, P., Maki, A., Montanaro, E., et al. (2016). The impact of changing attitudes, norms, and self-efficacy on health-related intentions and behavior: A meta-analysis. Health Psychology, 35(11), 1178–1188. View source
- Shadel, W. G., Martino, S. C., Setodji, C., Cervone, D., & Witkiewitz, K. (2017). Does self-efficacy causally influence initial smoking cessation? An experimental study. Addictive Behaviors, 73, 199–203. View source