Categorical perception effects for facial identity in robustly represented familiar and self-faces: The role of configural and featural information

Abstract

Categorical perception of robustly represented faces (self, friend) and unfamiliar faces is investigated, and the relative roles of configural and featural information are examined. Participants performed identification and discrimination tasks on morph series containing the self-face and a friend's face (self–Friend 1), two friends' faces (Friend 2–Friend 3), and two unfamiliar faces (Unfamiliar 1–Unfamiliar 2), presented in upright and inverted orientations. For upright faces, categorical perception effects were observed for both familiar morph series but not for the unfamiliar morph series, suggesting that robust representation is a requirement for categorical perception in facial identity. For inverted faces, categorical perception was observed for the self–Friend 1 morph series only. This suggests that categorical perception is tied to configural processing for familiar non-self-faces, but can be observed for self-faces during featural processing—consistent with evidence that self-face representations contain strong configural and featural components. Finally, categorical perception is not enhanced by the presence of the self-face relative to other familiar faces when upright, but shows a trend of being enhanced for self-faces when inverted, adding to the debate on the ways in which robustly represented faces can elicit categorical perception.

Keywords:

• Face processing
• Self-face
• Categorical perception
• Configural and featural processing
• Robust representation

Notes

¹It is common in the face processing literature to use the terms “configural” and “holistic” processing interchangeably. I use the term “configural” processing throughout the paper. See Rossion (2008, 2009) for a discussion of the terms “configural” and “holistic”.

²A better likeness task, as developed by Beale and Keil (1995), was chosen as a more appropriate discrimination task for facial stimuli than a classic ABX task (e.g., Etcoff & Magee, 1992), as the better likeness task encourages more holistic processing of the image. In addition, the better likeness methodology has been tested rigorously for its suitability for use in categorical facial perception tasks (see Levin & Beale, 2000).

³Following Etcoff and Magee (1992) and Calder and colleagues (1996), results were analysed according to continuum type. That is, data from each continuum type (self–Friend 1, Friend 2–Friend 3, Unfamiliar 1–Unfamiliar 2) were pooled across the individual continua. Calder and colleagues report similar categorical perception results for individual face continua to results from those pooled by continuum type.

⁴ Upright self–Friend 1 continuum, F(1, 35) = 251.75, MSE = 180.19, p < .001; upright Friend 2–Friend 3 continuum, F(1, 35) = 163.09, MSE = 257.42, p < .001; upright Unfamiliar 1–Unfamiliar 2 continuum, F(1, 35) = 152.26, MSE = 230.37, p < .001; inverted self–Friend 1 continuum, F(1, 35) = 355.13, MSE = 103.14, p < .001; inverted Friend 2–Friend 3 continuum, F(1, 35) = 44.15, MSE = 497.21, p < .001; inverted Unfamiliar 1–Unfamiliar 2 continuum, F(1, 35) = 18.94, MSE = 389.87, p < .01.