Human spatial representations have been postulated to use various spatial reference frames. These fall broadly into two categories: egocentric (body-centered) reference frames, where locations of objects in space are represented relative to one's body; and allocentric (object-centered) reference frames, where an object's location is represented relative to another object. “Egocentric” can further be divided into various subcategories, e.g. eye-centered, head-centered, or hand-centered. Recordings in macaques have revealed object-centered representations in area SEF (Olson, 2003).

To extend Olson's findings to humans, we used fMRI to investigate the neural correlates of egocentric and allocentric spatial representations in the context of eye movements. During the experiment, subjects fixated on a central fixation cross while a target appeared on a peripheral horizontal bar displayed at random locations in the visual field. The target and bar then disappeared and the bar without a target reappeared at a different retinal location. In the allocentric condition, subjects saccaded to the location on the bar where the target had appeared, thus using an object-centered spatial representation. In the egocentric condition, subjects saccaded to the former location of the target relative to their fixation point, thus using eye-centered coordinates. During baseline, subjects continuously maintained fixation.

Both egocentric and allocentric eye movements activated the intraparietal sulcus, FEF, SEF, MT, precuneus, V1, and ventral premotor areas. Although the two conditions yielded very similar activation patterns, some differences were observed in SEF, posterior parietal cortex and the anterior cingulate. Differences and similarities between egocentric and allocentric activations and postulated representations will be discussed.