This study was aimed at comparing the memory performance in a group of normal elderly controls and a group of patients in the first stages of probable Alzheimer's disease, with a paradigm including controlled encoding and enhanced cued recall (adapted from Grober and Buschke, 1987). The use of such experimental conditions allows a perfect discrimination between normal controls and AD patients: patients' performances show a deficit of encoding, and an impairment of free recall (poor recall and lack of recall consistency between trials); moreover, AD patients, unlike normal aged subjects, show little if any sensitivity to categorical cues, and produce a great number of extra-list intrusions; discriminability at recognition is impaired. At the same time, the data reveal a relatively large degree of heterogeneity in AD patients.