The primary purpose of this report was to investigate whether characteristics of subjects with borderline personality disorder observed at baseline can predict variations in outcome at the 2-year follow-up. Hypothesized predictor variables were selected from prior studies. The patients (N=160) were recruited from the four clinical sites of the Collaborative Longitudinal Personality Disorders Study. Patients were assessed at baseline and at 6, 12, and 24 months with the Structured Clinical Interview for DSM-IV Axis I Disorders; the Diagnostic Interview for DSM-IV Personality Disorders, a modified version of that instrument; the Longitudinal Interval Follow-Up Evaluation; and the Childhood Experiences Questionnaire-Revised. Univariate Pearson's correlation coefficients were calculated on the primary predictor variables, and with two forward stepwise regression models, outcome was assessed with global functioning and number of borderline personality disorder criteria. The authors' most significant results confirm prior findings that more severe baseline psychopathology (i.e., higher levels of borderline personality disorder criteria and functional disability) and a history of childhood trauma predict a poor outcome. A new finding suggests that the quality of current relationships of patients with borderline personality disorder have prognostic significance. Clinicians can estimate 2-year prognosis for patients with borderline personality disorder by evaluating level of severity of psychopathology, childhood trauma, and current relationships.