The current study objective was to develop a revised version of the Measure of Criminal Social Identity (MCSI) with an increased number of indicators to more reliably capture three MCSI dimensions. Dimensionality and construct validity of the Measure of Criminal Social Identity–Revised (MCSI-R) was examined among a sample of systematically selected inmates (N = 2192). Four competing models of the MCSI-R were specified and tested using Mplus with weighted least squares with mean and variance adjustment estimation. Bifactor model with three meaningful factors (cognitive centrality, in-group affect and in-group ties) while controlling for the general factor was the best fit for the data. Good composite reliability of the three MCSI-R dimensions was established. The three subscales of the MCSI-R evidence differential predictive utility for prisonization, number of incarcerations, self-esteem and violent offending. Practical implications and directions for future research are discussed.