Our aim was to assess the reliability and validity of nursing staff members' assessments of pain in cognitively impaired nursing home residents. Participants were 57 cognitively impaired nursing home residents and 52 nursing staff members. Twenty-eight residents had mild/ moderate cognitive impairment and 29 were severely impaired. Nursing staff members were asked to evaluate their patients' pain using 2 scales: The Nurse Global Scale of pain (NGS) and the Present Pain Inventory (PPI). Independent measures of pain were obtained from patient interviews, physician assessments, and from the Minimum Data Set (MDS) items; these were used to measure validity of the nursing staff's perception of pain. Test-retest and inter-rater reliabilities were also computed. Significant correlations were obtained between most of the physician-derived measures and the PPI for the moderately impaired group, but not for the severely impaired group. Both measures of nursing assistants' pain assessments correlated significantly with MDS-based measures; these correlations were higher for the less cognitively impaired group. Finally, the PPI correlated significantly and positively with resident-derived measures for those with moderate rather than severe cognitive impairment. Nursing staff members' ratings of pain were correlated with residents' level of cognitive functioning and with intake of pain medication. The results underscore the difficulty in assessing pain in those with severe cognitive impairment.