The present study focused on the consequences of cut-off scores in personality disorder diagnoses for their association with criminal behavior. Using ICD-10 personality disorder criteria eliminating offence-related symptoms, we studied the distributions of categorically diagnosed personality disorders and of dimensional personality disorder scores in a group of offenders and a noncriminal control group. Whereas the dimensional scores of the offender group differed significantly from those of the control group for all personality disorders under study, the frequency of categorical diagnoses differed significantly for two personality for two personality disorders only. Moreover, prediction of group membership (offenders vs. nonoffenders) from personality disorder scores was substantially more precise than prediction from categorical diagnoses. It is concluded that a dimensional approach to personality disorder diagnosis is not only superior theoretically but also yields more precise information about the specific associations with criminal behavior.