This paper describes tests of traditional and augmented versions of Fishbein and Ajzen's Theory of Reasoned Action (TRA). applied to condom use among high-risk heterosexually-active teens, drawn from a sexually transmitted disease clinic and a juvenile detention facility. In a longitudinal survey, predictor variables suggested by Ajzen and Fishbein (1980), plus condom use self-efficacy, attitude-norm interaction and gender, were measured at Time l. Condom use at Time 2, 3 months later, was regressed onto these variables. Support was found for the traditional TRA variables, but little support was found for the addition of self-efficacy. Implications for interventions to increase condom use among high-risk teens are discussed.