We apply the Health Belief Model (HBM) and the Theory of Reasoned Action (TRA) to an analysis of unsafe sexual practices (inconsistent condom use with commercial sex workers) among men living in a high HIV-prevalence area. The empirical analysis is based primarily on a survey of sexual practices that was conducted by the authors during the fall of 1991. The survey sample includes university undergraduates, soldiers, clerks, and laborers living in Chiang Mai, Thailand. There is much overlap between the two theories, and both provide useful frameworks for examining condom use in this population. However, the Bayes Information Criterion (BIC) approach provides very strong evidence that TRA is the more plausible model, given the data. The success of TRA appears to be due to its more adequate incorporation of peer group effects. Other covariates positively associated with consistent condom use in the multivariate models include knowledge about the consequences of HIV infection, knowledge about the benefits of condom use for preventing infection, and positive attitudes about condoms.