Two hundred and four patients, 70-85 years old, were included in a double-blind (Duxil/placebo), controlled, multicentric study. The inclusion criteria were a subjective complaint of a cognitive deficiency and a cognitive deficit objectively determined using the Folstein mini-mental state test and the Sandoz geriatric clinical evaluation score. The patients, treated for 6 months, were examined at the onset of the study (T0), then 3 (T3) and 6 months (T6) later. The assessment criteria included: a visual self-evaluation test measuring cognitive function and the following psychometric tests: trail making A (TMA), memorization of a shopping list, verbal fluidity, letter identification, repetition of a story, immediate recall of numbers and immediate visual memory. Anxiety and depression evaluations were also used to assess the effects of Duxil on the affective state. Statistical analysis of the observations made on the entire population did not reveal a significant difference between the treated group and the control placebo group, in terms of assessment criteria, between T0 and T6. However, this lack of a difference could be explained, in part, by the very wide variation in the initial psychometric performance scores of the subjects. In an attempt to control this possible bias, another statistical analysis was made for each psychometric test, after the patients had been divided into 3 classes based on their initial performance scores. The results of this second analysis showed that Duxil was able to improve memory performances in TMA and number retention better than the placebo. However, this effect was limited to the group of patients whose initial scores were in the intermediate class for TMA. These findings suggest that Duxil improves the concentrating ability of patients with light to moderate deficits in this function.