The study evaluates the efficacy of a procedural memory stimulation programme in mild and mild-moderate Alzheimer's disease (AD). Twenty basic and instrumental activities of daily living have been selected, and divided into two groups, comparable for difficulty. Ten normal elderly subjects (age 68.0 +/- 4.8 years; MMSE score: 28.7 +/- 0.9; education: 7.6 +/- 3.5 years) were asked to perform the two groups of daily activities and the time required to perform the tasks of each group was recorded and used as a reference. Ten mild and mild-moderate AD patients (age 77.2 +/- 5.3 years; MMSE score: 19.8 +/- 3.5; education: 7.3 +/- 4.7 years) without major behavioural disturbances constituted the experimental group. Patients were evaluated in all 20 daily activities and the time employed was recorded at baseline and after a 3-week training (1 h/d, 5 d/week) period. Five patients were trained during the 3 weeks on half of the 20 daily activities and the other five patients were trained on the remainder. This procedure was adopted in order to detect separately the improvement in "trained" and "not trained" activities, allowing to control better the effects of the intervention. The assessment of the functional impact of the training was directly measured, through the variation of time employed to perform tasks before and after training. After 3 weeks of training a significant improvement was observed for the trained activities, from 3.6 to 1.9 standard deviations below the performance of the normal elderly controls (P < 0.05). AD patients improved also in not-trained activities from 3.5 to 2.7 standard deviations below the controls' performance (P < 0.05). The rehabilitation of activities of daily living through developing procedural memory strategies may be effective in mild and mild-moderate AD patients.