Interrater reliability, retest reliability and discriminative power of the `Nurses Observation Scale for Inpatient Evaluation' (NOSIE-30) were assessed in a heterogeneous sample of psychiatric inpatients (68 <= N <= 85). Three pairs of nurses independently rated patients of their ward (25-30) twice (interval period: 7 days). Interrater reliabilities of `Social competence' (.75 <= r <= .82), `Personal neatness' (.71 <= .83), `Retardation' (.73 <= r <= .84) as well as of global NOSIE-scores (.76 <= r <= .83) were adequate. However, in one pair nurses rated patients consistently at a different level: in this sence interrater differences did occur. The `Social interest', `Irritability', `Psychosis' and `Depression'-subscales lacked agreement to a sufficient degree in at least one pair. For most subscales scores at first and second measurement correlated high, except for one rater. The ability of the NOSIE to discriminate between patients with `bad' and `better' prognoses could be determined.