The purpose of the present study was to investigate the relation between visual hemifields and spatial frames of reference, according to the idea that multiple representations of 3D space exist. Results from two experiments clearly show that an upper visual hemifield advantage only arises when allocentric spatial judgments are required in order to perform a location task, whereas a lower visual hemifield advantage arises when egocentric spatial judgments are required. Such a double dissociation was interpreted as due to the activity of two separate neural pathways operating specific transformations of visual input for different functional outputs: scene recognition, mostly relying on allocentric frames of reference and subserved by the ventral, occipito-temporal pathway of the visual system, and goal directed actions, mostly relying on egocentric systems and subserved by the dorsal, occipito-parietal pathway of the visual system.