In a psychiatric rehabilitation study, 154 concurrent ratings were performed using the 30-item Positive and Negative Syndrome Scale (PANSS) and the 18-item Brief Psychiatric Rating Scale (BPRS). Although both instruments had excellent interrater reliability, the PANSS was consistently better: on the 18 symptom items the two instruments share, the PANSS had higher intraclassr'son 14; for the syndromes, the PANSS was higher than the BPRS on positive, negative, and total. Weighted Kappas comparing shared items revealed that most were not interchangeable, with only three coefficients in the excellent range. However, syndrome scale scores were very highly correlated and resulted in similar classification for negative schizophrenia. Ten of the 12 items of the PANSS not included in the BPRS had low zero-order correlations with BPRS items, which suggests that they measure symptoms distinct from those measured by the BPRS and should add to clinical predictive power. This proved true in our study of rehabilitation of patients with schizophrenia PANSS symptom ratings explained up to 55% of the variance on seven measures of work performance, whereas the BPRS had lower predictive power on six of the seven measures. We concluded that the PANSS may be superior to the BPRS in clinical research on schizophrenia and that most BPRS items are not interchangeable with identically named PANSS items.