Multisensory peripersonal space develops in a maturational process that is thought to be influenced by early sensory experience. We investigated the role of vision in the effective development of audiotactile interactions in peripersonal space. Early blind (EB), late blind (LB) and sighted control (SC) participants were asked to lateralize auditory, tactile and audiotactile stimuli. The experiment was conducted with the hands uncrossed or crossed over the body midline in order to alter the relationship between personal and peripersonal spatial representations. First, we observed that the crossed posture results in a greater detrimental effect for tactile performance in sighted subjects but a greater deficit in auditory performance in early blind ones. This result is interpreted as evidence for a visually driven developmental process that automatically remaps tactile and proprioceptive spatial representation into an external framework. Second, we demonstrate that improved reaction times observed in the bimodal conditions in SC and LB exceeds that predicted by probability summation in both conditions of postures, indicating neural integration of different sensory information. In EB, nonlinear summation was obtained in the uncrossed but not in the crossed posture. We argue that the default use of an anatomically anchored reference system in EB prevents effective audiotactile interactions in the crossed posture due to the poorly aligned spatial coordinates of these two modalities in such conditions. Altogether, these results provide compelling evidence for the critical role of early vision in the development of multisensory perception and action control in peripersonal space.