This study investigated the automatic and effortful memory encoding model of Hasher and Zacks (1979) and the potential it may hold for aiding in the differentiation between aging-related memory decline and dementia. College students, normal elders, and dementia patients were compared on a 96-item picturebook task utilizing measures of free recall, recognition accuracy, memory for location, and memory for frequency. There were no differences between students and elders on any of the dependent measures. However, differences were found between elders and patients on each measure, and a discriminant function correctly classified the two groups with 93.3% accuracy. Subsequent discriminant analysis found patients could be correctly classified into diagnostic subgroups, i.e., dementia of the Alzheimer's type (DAT), multi-infarct dementia (MID), and Korsakoff's disease (KD) with 80.8% accuracy. The model holds promise as a guide for clinicians who are asked to make differential diagnoses of memory-impaired individuals.