Purchase this article with an account.
Luca Vizioli, Fraser Smith, Junpeng Lao, Lars Muckli, Roberto Caldara; Neural face coding is shaped by race. Journal of Vision 2011;11(11):623. doi: https://doi.org/10.1167/11.11.623.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Human populations can be categorized by using salient phenotypic traits, a visual process defining the social concept of race. Race markedly impairs one of the most critical and specialized skill humans possess: the recognition of conspecifics. Humans are significantly better at recognizing Same-Race (SR) than Other-Race (OR) faces, feeding the popular belief that OR faces all look alike. Theoretical (Valentine, 1991 – figure 1a) and computational (e.g., Caldara & Abdi, 2006 – figure 1b) Norm-Based Multidimensional Face-Space Models (NBMDFSM) have provided a consistent account for this universal Other-Race Effect (ORE). In NBMDFSM, efficient SR face identification is achieved by sparser coding across diagnostic dimensions for SR compared to OR face-exemplars, a by-product of visual experience. Neural evidence for NBMDFSM coding has been found with fMRI in humans and single-cell recordings in monkeys. Surprisingly, whether and where such neural face coding subserves the laws predicted by NBMDFSM of the ORE is unknown. We measured Western Caucasian (WC) and East Asian (EA) observers' BOLD signals in functionally defined face-selective Region of Interest (ROI – Fusiform Face Areas (FFA) and Occipital Face Areas) while they perceived normalized WC and EA faces (10 identities per race). We then computed Representational Dissimilarity Matrix (RDM – Kriegeskorte et al., 2008) in each ROI independently, by correlating the BOLD response elicited by each identity within a race with the remaining 9 across the multidimensional voxel population (figure 1c). We found significant higher RDM r-values for OR compared to SR faces in the FFA, relating to prototypical (less distinctive) activation patterns for OR faces in both groups of observers. We used RDM to link neural face representations with psychological and computational NBMDFSM of the ORE. This multidimensional voxel mapping quantified coding efficiency in the FFA for individual SR faces. This finding has profound implications for the understanding of the ORE and face perception.
This PDF is available to Subscribers Only