Evidence for separate processing of facial identity and expression information in an acquired prosopagnosic

In the two most well-known models of face processing (Bruce & Young, 1986; Haxby et al., 2000), facial identity and facial expression are processed separately, and the Haxby model further proposes that identity computations are performed in a ventral pathway that includes FFA while expression is represented in a dorsal pathway that includes pSTS-FA. However, in an influential review, Calder and Young (2005) proposed identity and expression processing within the visual system depend on the same processes, and no brain-damaged patients have shown unequivocal visual dissociations between identity and expression recognition. Here, we present findings from a 53-year-old right-handed woman we’ll refer to as Alma-Jean who, following encephalitis that lesioned her right temporal lobe, experiences severe difficulties with facial identity recognition but no loss of expression recognition. A dynamic localizer revealed intact face-selective areas in Alma-Jean’s left hemisphere. In the right hemisphere, OFA and pSTS-FA were spared but FFA and aSTS-FA were absent and within the lesioned area. Across seven tests of facial identity, Alma-Jean showed clear impairments. In contrast, she scored normally (z scores ranging from −1.0 to 0.5) on five tests of expression recognition. Alma-Jean’s normal expression scores appear to depend on typical processes, because like controls, she showed much better performance with upright faces than inverted faces. To confirm that her identity recognition deficits are due to visual problems rather than higher-level deficits, we showed that she performs normally on voice identity tests. In summary, Alma-Jean’s results provide the first evidence from a thoroughly tested brain-damaged case for a visual dissociation between facial identity and facial expression, and Alma-Jean’s results are consistent with the Haxby model’s ventral/dorsal distinction for identity and expression processing.