Taking the first-person perspective (1PP) centered upon one's own body as opposed to the third-person perspective (3PP), which enables us to take the viewpoint of someone else, is constitutive for human self-consciousness. At the underlying representational or cognitive level, these operations are processed in an egocentric reference frame, where locations are represented centered around another person's (3PP) or one's own perspective (1PP). To study 3PP and 1PP, both operating in egocentric frames, a virtual scene with an avatar and red balls in a room was presented from different camera viewpoints to normal volunteers (n = 11) in a functional magnetic resonance imaging experiment. The task for the subjects was to count the objects as seen either from the avatar's perspective (3PP) or one's own perspective (1PP). The scene was presented either from a ground view (GV) or an aerial view (AV) to investigate the effect of view on perspective taking. The factors perspective (3PP vs. 1PP) and view (GV vs. AV) were arranged in a two-factorial way. Reaction times were increased and percent correctness scores were decreased in 3PP as opposed to 1PP. To detect the neural mechanisms associated with perspective taking, functional magnetic resonance imaging was employed. Data were analyzed using SPM'99 in each subject and non-parametric statistics on the group level. Activations common to 3PP and 1PP (relative to baseline) were observed in a network of occipital, parietal, and prefrontal areas. Deactivations common to 3PP and 1PP (relative to baseline) were observed predominantly in mesial (i.e., parasagittal) cortical and lateral superior temporal areas bilaterally. Differential increases of neural activity were found in mesial superior parietal and right premotor cortex during 3PP (relative to 1PP), whereas differential increases during 1PP (relative to 3PP) were found in mesial prefrontal cortex, posterior cingulate cortex, and superior temporal cortex bilaterally. The data suggest that in addition to joint neural mechanisms, for example, due to visuospatial processing and decision making, 3PP and 1PP rely on differential neural processes. Mesial cortical areas are involved in decisional processes when the spatial task is solved from one's own viewpoint, whereas egocentric operations from another person's perspective differentially draw upon cortical areas known to be involved in spatial cognition.