The authors summarize some repeatable findings in the personality disorder (PD) classification literature. They then point out some pitfalls in research in the area. They suggest the following changes in research strategy and procedure: (1) The emphasis should be research on basic questions (e.g., which, if any, PDs are categorical, which are regions on continua, and which represent dysfunctional interactions of normal personality traits?) more than studies on applied questions (e.g., how is the DSM concept of borderline PD best diagnosed?). (2) They urge incorporation of findings from normal personality research (e.g., the ''Big Five'' factor model) and the use of appropriate statistical methods. They recommend a corresponding loosening of ties to official PD nomenclatures, the development of instruments embodying concepts shaped through interaction of ideas with data, and quasi-parallel self- and other-report versions to allow study of agreement and discrepancies among different sources of information.