In order to evaluate the prevalence of non-suppression after dexamethasone in psychogeriatrics and to further verify the reliability of the neurobiological information obtained, dexamethasone suppression test was carried out in a geriatric population composed of patients with dementia of Alzheimer type, vascular dementia, stroke and age and sex-matched controls. Basal cortisol levels did not differ among groups and was positively correlated to age. Prevalence of non-suppression, defined according to Carroll's criterion was high in the pathological groups studied, and relatively high in controls, showing no diagnostic value. Unlikely suppressors, the time course of mean plasma corticol levels of non-suppressors was highly heterogeneous in each group considered, especially in the pathological groups. More restrictive criteria for the definition of non-suppressors are proposed, in order to increase the specificity of the test when applied to psychogeriatrics.