Space is an important supramodal feature for linking inputs from different sensory modalities. This chapter reviews recent findings demonstrating that the visual modality is essential for the setting-up of spatial representations used as a default for multisensory interactions in later life. Thus, people blind from birth show no or a markedly reduced multisensory binding based on spatial features. By contrast, late blind people's default spatial coordinate systems are indistinguishable from those of sighted individuals. Possible consequences for crossmodal plasticity are discussed. Finally, people born totally blind whose sight has subsequently been restored, show a marked impairment in audio-visual integration. Results using the visual deprivation approach as a retrospective developmental approach to study the development of multisensory functions thus suggests that there are sensitive or even critical periods during which normal visual input must be available in order to allow for the development of full multisensory capabilities.