Behavioral compliance or adherence is a fundamental problem in health care, and diagnosis of dropout proneness represents a first step in adherence facilitation. A paper-and-pencil scale was developed to assess self-motivation, conceptualized as a behavioral tendency to persevere independent of situational reinforcements. Scale refinement (N = 401) yielded a logically valid and reliable measure with a high internal consistency (alpha = 0.91). Repeated measurements (1 to 5 months) indicated a high degree of scale stability (r's ranged from 0.86 to 0.92), reinforcing the conception of self-motivation as relatively enduring and trait-like. Convergent evidence for construct validity was provided by the Self-Motivation Inventory's (SMI) relationship with the Thomas--Zander Ego-Strength Scale (r = 0.63) and other more behaviorally specific attitudes (r's ranged from 0.47 to 0.58). Discriminant validity was evidenced by minimal overlap with motivational measures of social desirability, health locus of control, and achievement tendency; in each case, less than 10% of the variance in self-motivation was explained. Predictive validity was demonstrated in a variety of naturalistic settings in which perseverant behavior was easily quantified. These included habitual exercise programs for exercise training (N = 64), preventive medicine (N = 66), and acute exercise (N = 48). The diversity of these settings supported the situational invariance of self-motivation. Collectively, these findings provide considerable support for the valid and reliable assessment of self-motivation and for its distinctive utility in enhancing the prediction of perseverant behavior, specifically, adherence to therapeutic exercise.