Previous studies have suggested that education and race may affect performance on standardized mental status tests. In order to more clearly define these relationships, a prospective longitudinal study was devised to answer two questions: (1) whether race or level of education affects scores on the Mini-Mental State (MMS) exam in non-demented people and (2) what numerical cutpoints maximize the sensitivity and specificity of utilizing the MMS to help diagnose dementia in blacks of varying educational attainment. A total of 100 white and 258 black individuals, recruited from two city hospital primary care geriatric clinics, were evaluated and subsequently followed longitudinally over a 2 1/2 year period in order to assess accurately the presence or absence of dementia. In the non-demented, total MMS scores and performance on each item of the MMS were analyzed, revealing that people with an 8th grade or less education consistently had significantly (P less than .01) worse results than the better educated (9th grade or better) on borough, attention items, recall of table and dog, copying, sentence writing, phrase repeating, and total score. Furthermore, a total of 25% of the lower education group had an MMS score in the 18-23 range, traditionally thought to suggest dementia. There were no consistently significant differences between blacks and whites of equal education. In the better educated groups, using a score of 23 or less to define dementia maximizes the sensitivity and specificity of using the MMS in this diagnosis at 93% and 100%, respectively. In the lower education group, using 17 or less to define dementia maximizes sensitivity and specificity at 81% and 100%, respectively.(ABSTRACT TRUNCATED AT 250 WORDS)