This report addresses the need for prospective studies of personality disorders, as well as some of the difficulties encountered in longitudinal studies when missing data occur due to subject attrition and variable follow-up intervals. Various statistical methods for handling repeated measurements data are reviewed. Many of these methods are quite complex and require expert statistical skills. A simpler way to handle multivariate data using single-number summary scores is proposed as an alternative which is efficient and more readily understood by professionals in many disciplines. Findings are presented from a prospective study of borderline personality disorder which utilized repeated observations over time. Individual regression models were applied to each subject's repeated measurements to obtain a summary of his or her trend on measures of mood and global functioning. The individual regressions produced separate statistics, slopes summarizing rates of change and intercepts which estimated initial levels of functioning. These summaries were then used in group analyses. Findings indicated that subjects showed mild to moderate impairment in mood and moderate impairment in overall functioning. The individual slopes indicated that little overall change was observed during the 5-year period after initial assessment. Neither presence of borderline diagnosis (definite vs. trait vs. no borderline diagnosis) nor gender predicted initial levels of functioning or rates of change. Further examination of other predictors which may influence longterm outcome, such as history of childhood trauma or presence of schizotypal personality features, is suggested. It is concluded that prospective studies are essential in establishing the validity of personality disorders and in understanding individual variation in outcomes.