This study sought to develop a self-efficacy scale to predict smoking behavior and diagnose the most vulnerable factors associated with its onset. A sample of 800 seventh and eighth grade students received the survey on three separate occasions over an 8-month period. The instrument was comprised of 50 self-efficacy items, including social and emotional variables, and a self-report of smoking behavior. The data revealed an association between the total self-efficacy score and reported smoking behavior, implying the predictive capacity of the instrument. Factor analysis indicated several recurring item groups: social opportunities to smoke, emotional stress, and peer influences. Discriminant analysis of the subscales provided significant prediction of smoking behavior. Onset smoking behavior was also predictable over 3- and 5-month intervals. Peer influence persisted as the primary determinant of smoking activity. To confirm these results and ascertain the generalizability of this instrument, it is recommended that it be administered in other communities, and in conjunction with biochemical analysis to validate the self-reported smoking behavior. The instrument has potential application as a predictor of smoking behavior, to permit the design of more appropriate prevention programs, and as an effective evaluation measure for such interventions.