The set test was validated in 138 outpatients (mean age 74 years, 65% women) against a global clinical assessment using the DSM-III-R criteria for dementia. The test proved to have low sensitivity even at cutpoints close to a full score, and 12 of 72 patients diagnosed with dementia had a full score. We were unable to confirm the hypothesis that the set test can distinguish between vascular and Alzheimer's type of dementia. Used in its original form, the set test does not seem to be an appropriate diagnostic tool in the setting of a memory clinic.