Norms for cognitive measures used to assess dementia are scant for minority groups, in particular for older Japanese Americans. Using the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) Neuropsychology Battery, we compared the baseline performance of demented and nondemented Japanese Americans. Participants came from two harmonized epidemiological studies of dementia which were examined separately: the Kame Project, Seattle (350 men and women; 201 nondemented), age 65 and older; Honolulu-Asia Aging Study (HAAS), Hawaii (418 men; 120 nondemented), age 71 and older. The measures examined were Verbal Fluency; abbreviated Boston Naming; constructional praxis; and Word List Learning, Recall, and Recognition. Within each study, the CERAD measures distinguished between nondemented participants and those with mild cognitive impairment. Among persons with dementia, average level of performance decreased as severity of dementia increased. Determinants of score (age, education, language of administration, stage of dementia) varied between the two studies. Among Japanese Americans, the CERAD Neuropsychology Battery distinguished nondemented persons from those with dementia, but was less consistent in distinguishing levels of severity of dementia. This battery is useful for comparative epidemiological studies of dementia in minority populations. (JINS, 2005, 11, 192–201.)