Routine determination of serum vitamin B12 levels is generally recommended as part of the screening of demented patients, based on the notion that vitamin B12 deficiency is one of the causes of reversible dementia. We studied the effects of vitamin B12 replacement therapy in a prospective longitudinal study at a memory clinic, with special emphasis on assessment of severity of dementia: not only cognitive deterioration, but also disability in the activities of daily life, behavioural problems, and the burden experienced by the caregiver were examined using instruments of proven validity. In a series of 170 consecutive patients with dementia, subnormal serum vitamin B12 levels were found in 26 cases (15%); all but one fulfilled diagnostic criteria for possible Alzheimer's disease. Cobalamin supplementation was given to all patients and the effect was evaluated after 6 months. When the size and pattern of individual change scores, and the mean change scores on all instruments were taken into account, functioning after replacement therapy was not improved. When change scores of treated patients were compared with those of patients with Alzheimer's disease (n = 69), vitamin B12 replacement did not result in slowing of the progression of dementia. Contrary to widely accepted beliefs, subnormal serum vitamin B12 levels are not a (quantitatively) important cause of reversible dementia.