Self-efficacy is a person's belief in their capability to organise and execute the courses of action required to attain a designated goal. Albert Bandura introduced the construct in his 1977 paper in Psychological Review, arguing that what people can do matters less to their behaviour than what they believe they can do. A learner who judges herself able to hold a five-minute conversation in English will attempt one, persist when it falters, and recover from a stumble; a learner of equal ability who doubts herself will avoid the attempt, give up early, or read a single mistake as proof she cannot do it. The belief, not the skill, governs whether the skill ever gets used.
Bandura drew a sharp line between two judgments a person makes before acting. An efficacy expectation is the conviction that one can perform the behaviour itself ("I can give this presentation"). An outcome expectation is the estimate that the behaviour, once performed, will produce a given result ("a good presentation will raise my grade"). The two come apart. A student may be certain that fluent speaking earns high marks (a strong outcome expectation) yet doubt she can speak fluently (a weak efficacy expectation), and it is the second belief that decides whether she tries. Bandura held that efficacy beliefs do the heavier causal work, because outcomes themselves are largely read off from what one expects to be able to do.
Bandura proposed in Self-Efficacy: The Exercise of Control (1997) that efficacy beliefs are built from four sources, ranked by their power to shape them. Mastery experiences, direct success at the task, are the strongest: nothing convinces a learner she can do something like having done it. Repeated failure erodes the belief, especially before a sense of competence is established. Vicarious experience, watching a similar other succeed, raises one's own estimate ("if she can manage it, so can I"), and the resemblance between model and observer governs how much. Social or verbal persuasion, being told by a credible source that one has what it takes, lifts efficacy modestly and depends on the persuader's expertise and trustworthiness. Physiological and affective states, the racing heart and tight chest before speaking, feed back into the judgment: a learner who reads arousal as fear concludes she is not up to the task, while one who reads it as readiness does not. A teacher's review of these four sources by Ellen Usher and Frank Pajares (2008) confirmed mastery experiences as the dominant predictor across school subjects while noting that the sources operate differently across age, gender, and domain.
Self-efficacy is task- and domain-specific, not a global trait. A learner can hold high efficacy for reading academic English and low efficacy for spontaneous speech; a single "language self-efficacy" score blurs the two. This specificity separates the construct from neighbouring ideas. Self-esteem is a global evaluation of self-worth, relatively stable and tied to feelings about oneself rather than judgments of capability. Self-concept is a broader, more evaluative self-perception within a domain ("I am good at languages") that aggregates across many tasks and carries an affective, comparative charge. Self-efficacy is the narrowest and most forward-looking of the three: a prospective estimate of capability for a specific task under specific conditions. It also differs from Growth Mindset, which is a belief about whether ability is fixed or improvable; one can believe ability is malleable in general (a growth mindset) while still feeling unable to do a particular task today (low self-efficacy).
L2 self-efficacy, a learner's belief in her capacity to perform specific language tasks, predicts effort, persistence, and achievement across the four skills. It is woven into the affective machinery of the classroom: higher self-efficacy is associated with lower Language Anxiety and with greater Willingness to Communicate, the readiness to speak when given the chance. Joe Mills, Frank Pajares, and Carol Herron (2006), studying college learners of French, found self-efficacy negatively related to reading and listening anxiety and a stronger predictor of proficiency than anxiety itself. A learner who believes she can be understood is more likely to risk speaking, and each successful exchange becomes a mastery experience that feeds the belief forward, the kind of reciprocal loop that links self-efficacy to broader theories of motivation such as Self-Determination Theory.
The construct's specificity, its great strength for prediction, is also a measurement problem: efficacy items must be matched closely to the criterion task, and mismatched or vaguely worded scales inflate or deflate relationships, making cross-study comparison treacherous. The boundary with self-concept is genuinely contested. Herbert Marsh and colleagues argue that academic self-concept and self-efficacy share so much, perceived competence, domain specificity, hierarchical structure, that the supposed distinction is empirically murky, and that domain specificity of measurement, not the construct label, drives prediction. A deeper objection targets Bandura's core distinction: Charles Eastman and John Marzillier (1984) contended that efficacy and outcome expectations cannot be cleanly separated, since any judgment of what one can do is bound up with the outcomes one foresees, leaving the theory conceptually entangled. Bandura replied vigorously, but the criticism still shadows the construct. Finally, because self-efficacy and performance influence each other reciprocally, much of the evidence is correlational, and disentangling whether belief drives achievement or success simply leaves higher belief in its wake (reverse causation) remains difficult outside controlled designs.
The most efficient lever is mastery experience: structure tasks so learners succeed at a stretch they can feel, then raise the demand. Sequence speaking from short, scaffolded exchanges to freer talk so early wins accumulate before harder ones arrive. Use vicarious experience deliberately by modelling with near-peers rather than only expert models, since a visibly similar classmate persuades more than a polished teacher. Make verbal persuasion specific and credible ("your linking of ideas there was clear") instead of blanket praise, which carries little efficacy information. Teach learners to reinterpret nerves before speaking as ordinary arousal rather than evidence of inability. Because self-efficacy is task-specific, gauge and build it per skill rather than treating confidence as one global quantity.