Background. Current systems of describing personality pathology have significant shortcomings. A polydiagnostic approach is used to study the relationship between psychological, psychoanalytical and psychopathological models of personality.

Methods. The subjects were 256 patients enrolled in treatment studies of major depression and bulimia nervosa. Subjects were assessed using the Temperament and Character Inventory (TCI), the Defense Style Questionnaire (DSQ) and the Structured Clinical Interview for DMS-III-R personality disorders (SCID-II).

Results. Subjects had high rates of DSM-III-R personality disorders with 52% having at least one personality disorder. Cluster A personality disorders were correlated with low reward dependence, high harm avoidance and low self-directedness and cooperativeness. Cluster B personality disorders were related to high novelty seeking and low self-directedness and cooperativeness. Cluster C personality disorders were correlated with high harm avoidance and low novelty seeking and low self-directedness. Immature defences were related to DSM-III-R personality symptoms, but individual defences were not related to personality clusters in a predictable way. Immature defences were strongly related to low self-directedness and cooperativeness. Both TCI self-directedness scores and immature defence scores were moderately predictive of the presence and number of personality disorders.

Conclusion. A widely accepted clinical nosology (DSM-III-R personality disorders) rated using a clinical interview correlates reasonably predictably with two theoretical models derived from different paradigms and rated using self-reports. This might be seen as providing concurrent validity for all three models. However, serious methodological shortcomings confront studies of this type, including sample selection and measurement of personality dysfunction. One way to begin to resolve these problems is to study which personality measures are best related to treatment response and prognosis.