Perceptual separability of featural and configural information in congenital prosopagnosia

Abstract

The deficit in face recognition in individuals with prosopagnosia has often been attributed to an underlying impairment in holistic processing. Exactly what constitutes holistic processing has remained controversial, however. Here, we compare how configural information and featural information interact during face processing in a group of individuals with congenital prosopagnosia (CP) and matched controls. We adopted Amishav and Kimchi's version of Garner's speeded classification task, in which observers classify upright faces based on configural (intereyes and nose–mouth spacing) or featural (shape of eyes, nose, and mouth) information while the other dimension remains constant or varied randomly. We replicated the finding that normal observers evince symmetric Garner interference—failure to selectively attend to features without being influenced by irrelevant variation in configuration, and vice versa—indicating that featural and configural information are integral in normal face processing. In contrast, the CPs showed no Garner interference: They were able to attend to configural information without interference from irrelevant variation in featural information, and they were able to attend to featural information without interference from irrelevant variation in configural information. The absence of Garner interference in CP provides strong evidence that featural information and configural information are perceptually separable in CP's face processing. These findings indicate that CPs do not perceive faces holistically; rather, they process featural and configural information independently.

Keywords:

• Configural processing
• Congenital prosopagnosia
• Dimensional integrality
• Dimensional separability
• Faces
• Face perception
• Featural processing
• Holistic processing

Acknowledgments

This work was supported by the Max Wertheimer Minerva Center for Cognitive Processes and Human Performance, University of Haifa, to R.K., a grant from the Israel Science Foundation (ISF, 384/10) to G.A., and support from the National Science Foundation to M.B. (BCS0923763; NSF Science of Learning Center; SBE-0542013 to the Temporal Dynamics of Learning Center). We thank Michal Tanzer, Ryan Egan, Nilly Weiss, and Erez Freud for their help in data collection.

Notes

¹ It is important to note that although some researchers define configural information as local interfeature distances (e.g., Leder, Candrian, Huber, & Bruce, 2001), in our view, configuration is a consequence of the spatial relations between the facial features; it does not reside in a single interfeature distance, although a common way to manipulate configuration is to alter individual interfeature distances.

² On the surface, it appears as though the matched control group shows greater interference than the young controls tested by Amishav and Kimchi (2010). A direct comparison of these data is somewhat problematic because there are twice as many young controls as matched controls and because of the large difference in the age range. Nevertheless, a statistical comparison of the two groups reveals no significant interaction between group, task, and condition (F < 1), reflecting the presence of symmetric Garner interference in both groups