Less impairment in face imagery than face perception in early prosopagnosia

There have been a number of reports of preserved face imagery in prosopagnosia. We put this issue to experimental test by comparing the performance of MJH, a 34-year-old prosopagnosic since the age of 5, to controls on tasks where the participants had to judge faces of current celebrities, either in terms of overall similarity (Of Bette Midler, Hillary Clinton, and Diane Sawyer, whose face looks least like the other two?) or on individual features (Is Ronald Reagan's nose pointy?). For each task, a performance measure reflecting the degree of agreement of each participant with the average of the others (not including MJH) was calculated. On the imagery versions of these tasks, MJH was within the lower range of the controls for the agreement measure (though significantly below the mean of the controls). When the same tasks were performed from pictures, agreement among the controls markedly increased whereas MJH's performance was virtually unaffected, placing him well below the range of the controls. This pattern was also apparent with a test of facial features of emotion (Are the eyes wrinkled when someone is surprised?). On three non-face imagery tasks assessing color (What color is a football?), relative lengths of animal's tails (Is a bear's tail long in proportion to its body?), and mental size comparisons (What is bigger, a camel or a zebra?), MJH was within or close to the lower end of the normal range. As most of the celebrities became famous after the onset of MJH's prosopagnosia, our confirmation of the reports of less impaired face imagery in some prosopagnosics cannot be attributed to pre-lesion storage. We speculate that face recognition, in contrast to object recognition, relies more heavily on a representation that describes the initial spatial filter values so the metrics of the facial surface can be specified. If prosopagnosia is regarded as a form of simultanagnosia in which some of these filter values cannot be registered on any one encounter with a face, then multiple opportunities for repeated storage may partially compensate for the degraded representation on that single encounter. Imagery may allow access to this more complete representation.