Mild to moderate hearing loss has been hypothesized to decrease performance on verbally-administered cognitive tests as an artifact of testing. To evaluate this hypothesis, we conducted a randomized trial of a written version of the Mini-Mental State Examination (MMSE), a cognitive screening instrument which, in its standard form, is primarily verbally administered. After baseline standard MMSE testing, 71 outpatients with Alzheimer's type dementia, 39 of whom (55%) had mild to moderate hearing deficits, and 32 of whom (45%) had normal hearing, were randomly assigned to receive either a written or standard MMSE. Hearing-impaired patients exhibited lower standard MMSE scores than hearing-unimpaired patients at baseline (P = .005). Contrary to expectation, however, on experimental administration, written MMSE scores were somewhat lower than standard MMSE scores in hearing-impaired patients (P not significant). Furthermore, written MMSE scores were slightly higher than standard MMSE scores in hearing-unimpaired patients (P not significant). These results suggest that the diminished cognitive performance associated with mild to moderate hearing loss is not necessarily an artifact of cognitive testing. In addition, these results provide preliminary evidence that a written MMSE is comparable to the standard MMSE and, thus, deserves further consideration for cognitive screening of profoundly hearing-impaired individuals.