Criteria for diagnosis of personality disorders have been established in the two international classificatory systems, ICD-10 (World Health Organization, 1992) and DSM-IV (American Psychiatric Association, 1994). Both systems are atheoretical, i.e. based not on any causative explanatory paradigm but on expert consensus. Their approach to diagnostic classification has problems that are so serious that, in our experience, many practitioners question the value of making a diagnosis of personality disorder at all.

Given that both ICD-10 and DSM-IV are in the process of revision, we begin with deficiencies of the current systems that have been identified as being especially important. First, the current systems are neither theoretically sound nor empirically validated (Livesley, 2007; Tyrer et al, 2007). Second, they pose problems not only of overlap (an individual might satisfy several personality disorder diagnoses) but also of inadequate capture of important clinical aspects of personality pathology (e.g. passive–aggressive and sadistic traits) (Westen & Arkowitz-Westen, 1998). Furthermore, they are not sufficiently discriminating, so a substantial number of individuals are classified as having a ‘personality disorder not otherwise specified’ (Verheul & Widiger, 2004). Third, clinical assessments of personality disorder have been shown to be very unreliable and self-report inventories have been shown to generate too much psychopathology (Zimmerman, 1994). Although semi-structured instruments show an acceptable level of reliability, their administration is cumbersome and often requires considerable training.