Effects of exercise on body image are not well understood. Obese women seeking weight loss were randomly assigned to groups of exercise supported by a cognitive-behavioral protocol (n=68) or typical introductory methods (n=66), and assessed on body satisfaction (Body Areas Satisfaction Scale; BASS), self-regulatory efficacy (Exercise Self-Efficacy Scale; ESE), physical self-concept (Physical Self-Concept; PSC), weight, and body composition, over 6 months. Improvements on all measures were greatest in the cognitive-behavioral support condition. Improvement in BASS scores was better predicted by changes in the two psychological measures (PSC, ESE; Radj2=.30 and .32, respectively) than changes in the two physiological measures (weight and body composition; Radj2=.07 and .00, respectively). When all four predictors were entered into a multiple regression equation, only changes in ESE and PSE demonstrated significant unique contributions to the overall explained variance in BASS change. Theory and treatment implications were discussed.