This article reviews current evidence on the hemispheric specialization hypothesis for two types of spatial relations representations; categorical versus coordinate [Psychol. Rev. 94 (1987) 148; J. Exp. Psychol.: Percept. Perform. 15 (1989) 723]. Categorical representations capture general properties of the spatial structure of a visual stimulus, without defining the exact metric properties. Coordinate representations specify precise spatial locations of objects or parts in terms of metric units. It is claimed that a hemispheric difference in contribution to the computation of both types of spatial relations representations exists, in which the left hemisphere is specialized for the computation of categorical spatial representations while the right hemisphere is specialized for the computation of coordinate ones. Several forms of research (experimental, computer simulations, patient studies and neuroimaging studies) are reviewed. In general, there is convergent evidence for a conceptual separation of coordinate and categorical processing, with strongest indications for a relative right hemisphere advantage in encoding coordinate spatial relations, and weaker support for left hemispheric categorical specialization. The pattern appears to be critically linked to receptive field properties of the two hemispheres and as such is modulated by certain elementary visual characteristics of the displayed stimuli.