To accomplish the representation of the vast number of known faces, Valentine (1991) proposed that the brain codes faces as points in a multi-dimensional face space, where the axes correspond to facial attributes. Later researchers (O'Toole, Abdi, Defenbacher, & Valentin, 1993; Wilson & Diasconescu, 2006) have suggested that the axes are formed by extracting the principal components (PC) from a population of faces. As this theory has taken hold, various properties of face space have been examined. For instance, Wilson (2006) showed that learning could have an effect on the properties of face space. Recognition thresholds were significantly better in the regions surrounding learned faces than they were in the regions surrounding novel faces. This study demonstrates a similar effect, wherein participants were shown faces that consistently varied along a particular dimension in face space. Differences between pre- and post-learning thresholds for other faces randomly scattered about face space confirmed that it is indeed possible to transfer increased perceptual discrimination abilities not just to nearby faces, but across all of face space. Citations: O'Toole, A. J., Abdi, H., Deffenbacher, K. A., & Valentin, D. (1993). Low-dimensional representation of faces in higher dimensions of the face space. Journal of the Optical Society of America, 10, 405-411 Valentine, T. (1991). A unified account of the effects of distinctiveness, inversion, and race in face recognition. Quarterly Journal of Experimental Psychology, 43A, 161-204. Wilson, Hugh R., Diaconescu, Andreea (2006). Learning alters local face space geometry. Vision Reasearch, 46, 4143-4151.

In the world, encoding and learning of faces does not occur in isolation. We are exposed to and need to learn multiple faces, including encoding and learning in the presence of other faces. We developed the Progressive Face Learning Test to characterize individual differences in learning multiple novel faces in a short period as well as the effect of interference from other faces learned on learning rates.The test starts with a single face to be learned, which after presentation, is to be identified from a choice of faces, containing the target face and numerous foils.The number of faces to be learned is progressively increased by adding a new face after testing all the faces presented in the previous round.Faces to be learned as well as foils were chosen to be of comparable subjective distinctiveness.The test was used to characterize and compare different aspects of learning of faces like overall performance, learning rates and rate of change of learning rates for individual faces as well as with progressively increasing number of faces.Comparisons were done between people who have normal face recognition ability, above normal face recognition ability (“super-recognizers”) and below normal face recognition ability, including prosopagnosics.