The Dutch Behaviour Observation Scale-Geriatrics (GOS-G) is designed to give a detailed description of the functioning of geriatric patients. This article reports the psychometric properties of the GOS-G for use in homes for the elderly. Analysed were data on 71 residents of 11 homes (they participated in a study on withdrawal of benzodiazepines). They were free of serious physical and psychological complaints. For 55 persons from 10 different homes the observation scale was filled out by two different care takers at three different occasions: twice at the start of the study with a time interval of one to three weeks, and once more after eight months. Sixteen residents from another rest home were evaluated with the GOS-G by two different care takers at one occasion. The findings in this selective sample suggest the total scale to be a reliable and valid measure of the daily functioning in homes for the elderly. Moreover, the total scale appears sensitive to change. The scores on the subscales 'psychotic behaviour' and 'disoriented behaviour' were unreliable. Observation of a more heterogeneous and larger sample is needed to confirm the usefulness of the GOS-G in homes for the elderly.