The recognition of cognitive impairment by day and night nursing staff was studied in an acute geriatric unit. Seventy-six patients were randomly selected from a prospective sample of admissions. DSM-III-R diagnoses were established on all patients. Day and night staff were interviewed about each patient's clinical condition and asked to state whether or not they thought they were cognitively impaired or confused. Day staff were reasonably good at differentiating cognitively unimpaired from those with dementia and or delirium [kappa = 0.62, 95% confidence interval (CI) = 0.46-0.78]. All patients thought by day staff to be cognitively impaired were found to be so, although day staff did fail to identify some patients with cognitive impairment. Night staff performed less well (kappa = 0.37, 95% CI 0.18-0.57) and identified cognitively normal patients as being cognitively impaired, as well as failing to identify patients who were cognitively impaired. Night nursing interviews were not thought to have contributed to the management of any patient. The usefulness of night-time nursing interviews for research and general inpatient management purposes is questioned and the importance of daytime nursing interviews emphasized.