The Interactive Observation Scale for Psychiatric Inpatients (IOSPI) is characterized by the necessity of interaction between the rater and the patient during the evaluation process. The reliability, validity and factor structure of the IOSPI were evaluated by using a total of 186 sets of ratings accomplished by 2 nursing aides on 34 adult psychiatric inpatients submitted to weekly evaluations during the 3 first weeks of hospitalization. The patients were observed by the 2 nursing aides simultaneously during the morning shift (5 h). In the afternoon of the same day, they were interviewed by two psychiatrists who filled in separately the Brief Psychiatric Rating Scale (BPRS). Factor analysis of the 16 IOSPI items revealed 5 distinct clusters of items, which represented higher-order constructs of social interest, psychomotor agitation, psychoticism, neatness and irritability. It was found that the IOSPI has significant interrater reliability, estimated by an intraclass correlation coefficient of total score and factor score. The total scores of the IOSPI showed a significant correlation with the total scores of the BPRS, meeting the criteria of concurrent validity. The criteria of predictive validity were also met, since the patients who could be discharged presented IOSPI scores significantly lower than the patients who could not be discharged. In conclusion, the IOSPI met the requirements of an evaluation scale and can stimulate therapeutic attitudes of the nursing staff by its characteristics of interactive observation.