Motivation to engage in health-promoting exercise has been ascribed to global personality traits such as self-esteem or athletic self-confidence. Self-Efficacy Theory challenges this view by proposing that highly specific estimates of personal capabilities mediate adoption of new or difficult behavior patterns. We tested this assumption by measuring self-efficacy perceptions in 40 men with coronary artery disease who participated in an experiment evaluating effects of circuit weight training (CWT). Specific self-efficacy estimates were assessed during baseline strength/endurance testing and after 10 wk participation in CWT or volleyball. Correlational analyses of self-efficacy in relation to performance on strength/endurance tests strongly supported the contention that adoption of novel activities is governed by highly specific self-perceptions. Participation in CWT produced greater strength and endurance gains than did volleyball, and these changes were accompanied by increased self-efficacy in CWT subjects for activities resembling the training tasks. The assertion that self-efficacy perceptions directly mediate involvement in challenging physical activities was supported by multiple regression analyses. These revealed that pre-training self-efficacy judgments predicted post-test strength gains even after controlling for baseline strength, type of training and frequency of participation in exercise sessions.