Torture is thought to confer worse mental health than other war-related traumatic events. However, reliability of torture assessment and validity of torture constructs as indicators of poor mental health have not been systematically evaluated. Study aims were to assess the psychometric properties of 2 common torture constructs. Refugees were assessed for having experienced torture by 1 subjective and 1 objective criterion. A subset was interviewed about definitions and effects of torture. Reliability and validity of torture constructs were assessed with general linear models. Interview data were analyzed for consistency of themes. Reliability of torture constructs was moderate, which paralleled inconsistencies in interview themes. Both torture criteria had similar dose-dependent relationships to mental health. Multivariate analyses showed that torture was not an independent predictor of poor mental health when controlling for the number of war-related events. Further work is needed to define torture from distinct medical and legal perspectives to improve reliability and validity.