When combined with cognitive investigations, functional neuroimaging methods such as positron emission tomography allow to depict the neural substrates that underlie the neuropsychological alterations in Alzheimer's disease. Capitalising on the variance in both cognitive performances and resting cerebral metabolic rate of glucose (CMRGlc) in Alzheimer's disease, it is possible to correlate these two quantitative variables on a pixel-by-pixel basis and to generate maps showing the significant correlations in stereotaxic space. Some examples using this approach in the domain of memory disorders are presented in this brief review. We notably show that the localisation of the significant correlations differs from one memory system to another, as evaluated by clinical memory tasks. This approach also unravels the compensatory mechanisms that take place with evolution of the disease. Over and above its interest in clinical neuropsychology, this method constitutes a new source of inferences complementary to the classic activation paradigm in normal subjects, as the latter identifies the cerebral structures that are involved with, but not necessarily indispensable for, the normal execution of the task. This approach highlights the interest of combining functional neuroimaging and neuropsychology to better understand the neural substrates of cognitive deficits in both patients with memory disorders and elderly normal subjects.