The inter-situational, inter-rater and temporal reliability of a schedule for rating personality disorders is described. In an initial study with a simplified form of the schedule in patients from different wards of a psychiatric hospital inter-situational reliability between raters was higher for patients with personality disorders than with no personality disorder. Using the full schedule, inter-rater reliability, using audiotaped and separate interviews, and temporal reliability at interviews conducted a mean of 12.5 months apart all reached a satisfactory level, suggesting that the schedule may be a useful instrument for measuring deviant personality traits. The interview may be used with a subject or an informant but agreement between ratings made with informants and psychiatric patients during illness was low, and the schedule is not recommended for use with patients alone during acute episodes of illness.