Abstract
In congenital prosopagnosia (CP), the anatomy and activation patterns in ventral occipital temporal cortex (VOTC) have been shown to be grossly normal (Hasson et al., 2003), though finer-grained analyses have demonstrated subtle functional and structural abnormalities in VOTC (Avidan et al., 2005; Behrmann et al., 2007). CP's mostly normal VOTC suggests that they possess an intact neural infrastructure for face processing that may, with the right type and amount of training, be able to support improvements in face identification. We designed a training task that targeted CPs deficits at spatially integrating multiple features. In particular, CPs were trained to integrate spatial information from two areas of the face, the mouth region and the eye region, in order to make a perceptual judgment. After 10 days of training 750 trials each day, several prosopagnosics significantly improved on the standardized Cambridge face memory and face perception tests as well as on memory and matching tasks using novel faces. These individuals also reported face recognition improvements in everyday life. Without any additional training, improvements on the face assessments lasted for a period of months before declining. In one successful case, we compared event-related potentials (ERP) and functional MRI (fMRI) before and after training. ERP results revealed that although the N170 component was not selective for faces before training, its selectivity after training was normal. fMRI demonstrated that training increased functional connectivity between VOTC face-selective regions (right occipital face area and right fusiform face area) and between face-selective regions and an extended network of regions. Together, these results demonstrate that intensive training can improve face identification abilities in CPs and this appears to occur through strengthening of normal neural mechanisms.