A study of adherence to treatment was conducted by 179 general practitioners in elderly outpatients with geriatric cerebral symptomatology treated with pentoxifylline. The drug was provided in 2 different randomized packages, with or without memory-aid stickers (also randomized). Compliance was assessed by pill count after 1 mo of treatment. Clinical evolution was assessed by a digit-span test, and by filling in 9 "relative" visual analogue scales of aggravation--improvement. Side effects were recorded from patient complaints. Leftover drug was brought back by 83.1% of patients, and this proportion was influenced neither by packaging type nor memory-aid stickers. Compliance was considered good (fewer than 30 tablets returned) in 62% of patients, and was not influenced by either packaging types or stickers. Peaks of pill count were evident at multiples of packaging units (10 or 40 according to type). Compliance was not related to age or sex, but was related to memory score. There was a correlation between compliance and clinical improvement, and a significant inverse correlation between the former and the frequency of side effects.