The reliability of individual differences in face-selective responses in the fusiform gyrus and their relation to face recognition ability

Face recognition ability varies widely in the normal population and there is increasing interest in linking individual differences in perception to their neural correlates. Such brain-behavior correlations require that both the behavioral measures and the selective BOLD responses be reliable. The reliability of the location of the fusiform face area (FFA) has been demonstrated in several studies. Here, we address reliability of a different kind: reliability of the magnitude of responses to faces within this localized region. We calculated split-half reliability of face-selective responses within functionally defined posterior and anterior face-selective patches in the fusiform gyrus (FFA1/FFA2). We used data from two published studies that included both a functional localizer for face-selective regions and independent data suitable for quantifying face-selectivity. We found highly reliable face selectivity in both hemispheres that was highest in the centermost voxel(s) compared to larger regions of interest. Differences in face-selectivity between the two face patches within one hemisphere and across hemispheres were also reliable. Our results reveal considerable reliability of face-selective signals in and across FFA in adults. Given the good reliability of behavioral measures of face recognition, prior failures to find a relationship between the mean response to faces in FFA and behavioral face recognition in normal adult subjects are unlikely to be due to limitations of the measurements.