We prospectively studied the evaluation of dementia in 107 unselected outpatients; 83 had so-called "irreversible" dementias, including 74 who had an Alzheimer-type dementia. Fifteen patients had potentially reversible dementias, of which hypothyroidism and drug toxicity were the commonest causes. Distinguishing features of reversible dementia were shorter duration, use of more prescription drugs, and less severe dementia. Almost half of the patients had other previously unrecognized treatable medical diseases. Most diagnoses were made from patient history and physical and mental status examination. Patients with reversible dementia improved but rarely reverted to normal. Objective improvement occurred in 25 patients after treating unrecognized coexistent medical and psychiatric diseases, or stopping unnecessary medication. Careful clinical observation is the most useful part of the evaluation and extensive testing may not be required for all patients. Overemphasis on distinguishing reversible from irreversible forms of dementia may detract from recognition of commoner, treatable causes of dysfunction and suffering.