Abstract
Reverse engineering involves disassembling a complex device and analyzing its components and workings in detail with the goal of understanding how the device works in its intact state. To elucidate the neural components implicated in normal face perception, we investigate the disrupted components in individuals with congenital prosopagnosia, an apparently lifelong impairment in face processing, despite normal vision and other cognitive skills. Structural and functional MRI data reveal compromised connectivity between more posterior face-selective cortical patches and more anterior regions that respond to face stimuli. Computational descriptions of the topology of this connectivity, using measures from graph theory that permit the construction of the network at the level of the whole brain, uncover atypical organization of the face network in CP. Moreover, this network disorganization is increasingly pronounced as a function of severity of the face recognition disorder. Last, we reconstruct the face images viewed by normal and prosopagnosic observers from the neural data and demonstrate the altered underlying representations in key cortical regions in the prosopagnosic individuals. This multipronged approach uncovers in fine-grained detail the alteration in information discrimination in the prosopagnosic individuals as well as the pertubations in the neural network that gives rise to normal face perception.