Résumé
During the last two decades, neuroimaging has generated a wealth of knowledge on how number processing inserts itself into the functional neuroanatomy of the human brain. We understand quite well now what are the cortical areas involved, and the neural codes for individual quantities as perceptual entities. Still, we lack a general understanding of how quantity representations are transformed during mental computations, and how or even where results of such computations are coded in the brain. By using ultra-high-field (UHF) imaging during an approximate calculation task designed to disentangle in- and outputs of a computation from the operation, we uncovered a representation of internally generated quantities which was most prominent in higher-level regions like the angular gyrus and lateral prefrontal cortex, and the intra-parietal sulcus. Intraparietal sensory-motor integration regions were the only ones found to share the same representational space for stimulus-evoked and internally generated quantities. This suggests the transformation may occur in these regions, before result numbers are maintained for task purposes in higher-level areas in a format possibly detached from sensory-evoked inputs. Results illustrate the power of UHF imaging to finely characterize neural codes underlying human numerical abilities with non-invasive methods.