Results revealed that of the three models tested, the multidimensional model was found to be the best fitting model while the unidimensional model was significantly inferior to the other two factorial models tested. Evidence suggested that the multidimensional scale of self-control could be modified to improve the goodness-of-fit. Such modification should include the elimination of variables that showed cross-loading or had low factor loadings. Tests of invariance indicated that the self-control measure was equivalent across the two disparate male groups. These findings revealed that the male prison inmate group scored significantly higher on self-centeredness and had a higher preference for tasks that required little effort, persistence, and diligence in comparison to the male student group. This study used confirmatory factor analysis to examine the theoretical construct of the latent trait, self-control. Three models were developed to examine self-control: (1) a conventional unidimensional model; (2) a second-order general factor structure; and (3) a six-factor multidimensional model. The two groups of participants were chosen based on their difference from each other in terms of expected levels of self-control. Confirmatory factor analysis was used to determine the underlying construct of self-control and to investigate whether the self-control measure was invariant across prison and nonprison groups. All univariate and bivariate statistical analyses were performed using the Statistical Package for Social Sciences (SPSS). Future research should focus on validating the self-control measure with deviant behavior using research samples that contain low self-controlled individuals. Figure, tables, references