This study presents data on the interrater reliability of a Dutch version of the Structured Clinical Interview for DSM-III-R Personality Disorders (SCID-II). Seventy outpatients were interviewed before the start of their treatment by one rater, while a second rater observed. Both raters were instructed to make independent ratings and the second rater was not allowed to participate in the discussion. On criterion level, interrater reliabilities appear to be very good, with a few exceptions (most reliabilities are higher than 0.75). However, all 5 observation criteria had poor interrater reliabilities. Agreement on personality disorder, on the whole, was excellent (overall kappa = 0.80). The possible reasons why relatively lower reliabilities are found with some criteria are discussed. Finally, problems encountered during the interviews are addressed and possible adjustments of the SCID-II are suggested.