Because sensory systems use different spatial coordinate frames, cross-modal sensory integration and sensory–motor coordinate transformations must occur to build integrated spatial representations. Multimodal neurons using non-retinal body-centred reference frames are found in the posterior parietal and frontal cortices of monkeys. We used functional magnetic resonance imaging to reveal regions of the human brain using body-centred coordinates to code the spatial position of both visual and somatic sensory stimuli. Participants determined whether a visible vertical bar (visual modality) or a location touched by the right index finger (somatic sensory modality) lay to the left or to the right of their body mid-sagittal plane. This task was compared to a spatial control task having the same stimuli and motor responses and comparable difficulty, but not requiring body-centred coding of stimulus position. In both sensory modalities, the body-centred coding task activated a bilateral fronto-parietal network, though more extensively in the right hemisphere, to include posterior parietal regions around the intraparietal sulcus and frontal regions around the precentral and superior frontal sulci, the inferior frontal gyrus and the superior frontal gyrus on the medial wall. The occipito-temporal junction and other extrastriate regions exhibited bilateral activation enhancement related to body-centred coding when driven by visual stimuli. We conclude that posterior parietal and frontal regions of humans, as in monkeys, appear to provide multimodal integrated spatial representations in body-centred coordinates, and these data furnish the first indication of such processing networks in the human brain.