Up to one third of dementia sufferers eat an increased quantity of food, compared with their premorbid intake, at some stage during the dementia. A proportion of these eat extraordinarily large quantities if food intake is not restricted. In order to investigate this phenomenon in detail, a reliable and standardized method of quantifying the degree of hyperphagia is required. We report the development of such a method. Twenty-six people with dementia, who were reported by their carers to be hyperphagic, were compared with 14 matched non-hyperphagic controls with dementia and 14 matched normal elderly. Subjects were offered two standardized meals, under specified conditions, ad libitum. One meal consisted of a single food, the other of a mixture of foods. The total energy intake provided a reliable measure of the degree of hyperphagia. The single food meal was more reliable but the mixed meal was a more sensitive measure of the hyperphagia.