Previous work has suggested that familiarity/novelty of learned materials affects the circuitry involved in memory, primarily in the size of activations rather than the pattern of activation. Although this work has examined both recall and recognition, it has been limited to verbal material. In this study, we set out to determine if the same result applies to nonverbal memory. We used the same experimental design, but used faces as the memory task. Healthy volunteers thoroughly learned a set of 18 faces a week prior to the Positron Emission Tomography (PET) experiment (well-learned memory) and were asked to remember another set of 18 faces, to which they were exposed 1 min before the PET experiment (novel memory). During the PET session, their task was to recognize the faces learned a week before and the faces seen a minute before; the "remembered faces" were interspersed among entirely new (distractor) faces. We found that, unlike for verbal material, the retention interval and the familiarity level of the faces affected both the pattern and the size of activations. Comparing the novel and well-learned recognition tasks revealed that novel memory for faces is primarily a frontal-lobe task, while well-learned recognition memory for faces utilizes a more distributed neural circuit, including visual areas, which appear to serve as memory-storage sites.