Several studies have employed confirmatory factor analysis (CFA) to evaluate the latent structure of posttraumatic stress disorder (PTSD) assessment measures among various trauma-exposed populations. Findings have generally failed to support the current three-factor DSM-IV PTSD conceptualization, demonstrating the need to consider alternative models. The present study used CFA to evaluate seven models, including intercorrelated and hierarchical versions of two models with the most empirical support. Data were utilized from a heterogeneous trauma-exposed sample of general medical patients (n=252). Based on several indices, the three-factor DSM-IV PTSD model was shown to be inferior to alternative models. The strongest support was found for an intercorrelated four-factor model, separating avoidance and numbing symptoms into distinct factors. Validity for this model was partially supported by divergent relations between factors and external variables. Implications of the results are discussed, and a framework is proposed for resolving discrepant findings in the PTSD CFA literature.