The question of whether Mini-Mental State Examination scores should be adjusted for age and educational levels to screen for dementia in clinical populations is reexamined in the results of a recent study supporting adjustment. If the criterion is to identify the most accurate screening procedure for each sociodemographic subgroup, the evidence indicates that the unadjusted scores are preferable. Other criteria might lead to different conclusions. The validities of some of these criteria are questionable because they have the flaw that they are easily satisfied by using random decision procedures.