SUMMARYThis study examined the measurement properties of a cognitive screening test, the Halifax Mental Status Scale (HMSS), in a population-based sample of community-living elderly subjects ( N = 361) and a clinic sample ( N = 20) of patients with Alzheimer's disease. Receiver operating characteristic analyses yielded estimates of sensitivity and specificity of approximately 90% for this test. Factor analysis confirmed a three-factor solution for the HMSS. Although both age and education were correlated with the HMSS total score, their relationship to the factor scores differed. Our study shows that comprehensive cognitive status tests do not measure a unitary construct and that adjustment of summary scores for education is unlikely to optimize sensitivity and specificity. In population-based studies, only those components of a test that reflect education should be adjusted.KEY woms-mental status; screening tests; dementia Epidemiologic studies remain an important source of information about dementia. Given our lack of understanding about causes of dementia, the need for unbiased estimates of the more recently identified types of dementia (Lund and Manchester Groups, 1994;McKeith et al., 1992) and the need for better data on previously identified types of dementia (Chui et al., 1992;Rockwood et al., 1994;Roman et al., 1993), this is likely to remain the case. Most current epidemiologic studies of dementia employ a two-stage approach to diagnosis (Rockwood and Stadnyk, 1994). In the first stage, a