To determine interobserver reliability of the Mental Health Patient Assessment Record, two independent observers, a visiting nurse and the investigator who was a constant second observer throughout the study, made home visit observations at two specified times: the initial assessment (time I), and approximately two months later (time II). The stratified random sample consisted of 26 discharged psychiatric patients who comprised approximately 75 percent of the new admissions to the agency during a four-week period. At time I, 26 visits were completed and values were obtained, using the Pearson product moment correlation for the scores on the MHPAR: 1) overall behavior, r = .96; 2) physical behavior, r = .91; 3) psychological behavior, r = .93; 4) social behavior, r = .83; and 5) medical behavior, r = .98. The sign test applied to total and category scores revealed that only one category, social behaviors, demonstrated a significant systematic bias of the constant second observer. Of the 26 patients observed at time I, 20 were visited at time II. The following values were obtained using the Pearson product moment correlation for MHPAR scores: 1) overall behavior, r = .98; 2) physical behavior, r = .91; 3) psychological behavior, r = .97; 4) social behavior, r = .94; and 5) medical behavior, r = .99. The sign test applied to total and category scores showed no systematic bias of the constant second observer.