Interrater reliability is considered a precondition for the validity of theoretical models and their corresponding diagnostic instruments. Studies have documented good interrater reliability for structured interviews measuring personality characteristics on a descriptive-phenomenological level but there is little research on reliability of assessment procedures on a structural level. The current study investigated the interrater reliability of the structural interview (SI) designed to assess neurotic, borderline, and psychotic personality organization according to Kernberg. Videotaped SIs of 69 psychiatric patients were randomly and independently rated by two out of three trained psychologists. Agreement between rater pairs was expressed as square weighted kappa (K-sw, 95% CI). Results indicate satisfactory interrater reliability with respect to Kernberg's tripartite classification (K-sw = 0.42, 95% Cl 0.07 to 0.77). Subdivision of the borderline category or introduction of intermediate subcategories to the tripartite system did not significantly affect reliability (K-sw = 0.55, 95% CI 0.30 to 0.80; K-sw = 0.59, 95% CI 0.34 to 0.84, respectively). The conclusion is that trained clinicians can reliably assess structural personality organization using the SI. Refining the nosological system adding subcategories did not reduce reliability.