BACKGROUND Although the assessment of cognitive functioning in the late stages of Alzheimer's Disease (AD) is important for identifying abilities that may improve communication and interactions with severely impaired patients in clinical and institutional settings and for assessing the efficacy of pharmacologic agents and behavioral interventions for the treatment of AD, few adequate instruments exist for measuring the cognitive capacities of these severely demented individuals. OBJECTIVE To evaluate the reliability and validity of the Severe Cognitive Impairment Profile (SCIP), a measure of neuropsychological functioning in severely demented patients, and compare it with other available instruments. METHODS We administered the Mattis Dementia Rating Scale (DRS), Mini-Mental State Examination (MMSE), SCIP, and Severe Impairment Battery (SIB) to 41 severely demented patients with AD participating in an AD research center. We used (1) Spearman rank correlation coefficients to assess interrater and test-retest reliability and construct validity of the SCIP; (2) one-way analysis of variance with post hoc comparisons to examine performance on the SCIP and the SIB at different levels of dementia severity; and (3) descriptive statistics to establish the sensitivity of the SCIP to cognitive functioning in a subgroup of very severely demented patients. RESULTS Interrater and test-retest reliability correlation coefficients were highly significant for total SCIP score (r=0.99 and r=0.96, respectively) as well as for all SCIP subscales. High correlations were also found between SCIP scores and two widely used tests of global cognitive functioning, the DRS (r=0.91) and the MMSE (r=0.84), suggesting good construct validity. The SCIP was able to significantly differentiate between four groups of severely impaired patients divided by level of dementia severity, while the SIB was unable to differentiate between the less severely demented groups. A subgroup of 16 very severely demented patients (DRS score, <50 points) obtained an average of 45% of total possible points on the SCIP, compared with an average of 1% and 21% of total possible points on the MMSE and DRS, respectively. After approximately 1 year of decline, 12 severely demented patients with AD were able to correctly answer an average of more than 58% of the items on the SCIP, compared with only 30% on the DRS and 20% on the MMSE. CONCLUSIONS The SCIP is a reliable, valid measure of neuropsychological functioning in severely demented patients with AD with the ability to avoid both floor and ceiling effects and to evaluate a wider range of cognitive abilities than other tests used with severely impaired individuals.