Sixty-four nondepressed, carefully selected senile demented inpatients underwent two 1-mg overnight dexamethasone suppression tests (DSTs) separated by 7 days. These patients were in a clinically stable and drug-free state for at least 6 months before testing. The Modified Ischemic Score and ICD-9 criteria for Alzheimer's disease were used to dichotomize subjects into Alzheimer's and multi-infarct types of dementia. Patients with vascular dementias were significantly more likely to evidence DST nonsuppression. Furthermore, DSTs in this group were less reproducible from week to week than DSTs in Alzheimer's patients.