This study presents information regarding the diagnostic efficiency and hierarchical functioning of the newly revised DSM-IV borderline personality disorder (BPD) criteria. Past research using the DSM-III-R version of BPD showed that the individual BPD criteria differ greatly in their diagnostic utility and that in some clinical situations, fewer than five of eight criteria could efficiently make the diagnosis. Two groups of inpatients, a BPD group (N= 33) and a non-BPD group (N= 43), were rated on the DSM-IV BPD criteria. Acceptable interrater reliability (Kappa estimates) was obtained for the presence or absence of the nine individual BPD criteria. Diagnostic efficiency statistics, sensitivity, specificity, false positive and false negative rates, positive predictive power, negative predictive power, overall classification rate and Kappa with the clinical diagnosis were obtained for all nine BPD criteria. The diagnostic efficiency data indicated that BPD criteria 1 (abandonment) and 2 (unstable relationships) functioned best in our inpatient sample, whereas criterion 9 (stress-related paranoia) performed the poorest. A stepwise logistic regression showed that the combination of BPD criteria 2, 1, 6 (unstable affect), and 3 (identity) (presented in their order of entry into the regression equation) provided the best prediction of group membership (BPD or non-BPD). The results suggest that a further refinement of both the BPD criteria set and the diagnostic decision rules may be needed.