BACKGROUND: Acute and chronic pain is common in hospitalized demented elderly people, yet there are limited data about the performance of pain assessment tools in this population. The aim of this study was to evaluate the feasibility and reliability of four pain self-assessment scales in this population and compare their performance to an observational pain rating scale. METHODS: Our prospective clinical study was conducted in an acute-care and intermediate-care geriatric hospital on 160 consecutive inpatient referrals to the dementia consultation who met Diagnostic and Statistical Manual of Mental Disorders-IV criteria for dementia. Exclusion criteria were delirium, terminal care, and severe sensory impairment. Four unidimensional self-assessment tools--the verbal, horizontal visual, vertical visual, and faces pain scales--were administered in randomized order to mild, moderate, and severely demented patients. An observational pain rating scale was independently completed by the nursing team. RESULTS: Only 12% of the 160 patients (mean age 85 years, 71% women) understood no scale. Respectively, 97%, 90%, and 40% of patients with mild, moderate, and severe dementia understood at least one scale (p <.05). There was a nonsignificant trend toward poorer comprehension of the faces scale. Test-retest reliability was high for all four self-assessment scales, and the correlation between these scales was very strong (Spearman's r(s) = 0.81-0.95; p <.001). Observational rating correlated moderately with self-assessment and tended to underestimate pain intensity (r(s) = 0.31-0.40; p <.05). CONCLUSIONS: Self-assessment pain scales can be used reliably in the vast majority of older hospitalized patients with mild to moderate dementia and in nearly half of those with severe dementia. Observational pain rating scales correlate only moderately with self-assessment and should be reserved for those few patients who have demonstrated that they cannot complete a self-assessment.