Abstract
So-called super-recognizers (SRs) possess exceptional face processing abilities. Yet, the factors underlying their heightened abilities remain unknown, owing to overreliance on underspecified diagnostic tests. In general, these employ simple accuracy measures, yielding at best piecemeal correlations between various sub-processes. Given the inter- and intra-personal heterogeneity across sub-processes, we argue here—across four experiments in two independent studies—that consistency of face processing distinguishes superior face cognition abilities. In Study 1, we analysed sensitivity to spatial frequency content of face stimuli across viewing distances, bandwidths and orientations in two one-to-ten face matching experiments for 11 SRs and 31 normal observers. The first simulated variations in viewing distance systematically; the second instead varied the bandwidths of horizontal and vertical spatial frequency content separately (c.f., Nador et al., 2021). Across these experiments, we find that controls and SRs are sensitive to the same range of identity-diagnostic information (i.e., with equal absolute thresholds), which are, however, exploited more consistently by SRs (i.e., with different psychometric slopes). In Study 2, we analysed sensitivity to face memorability in two old/new recognition experiments among 10 SRs and 22 normal observers, by adapting a split-half “with-or-without-you” resampling technique for pattern analysis (Nador et al., 2020). In Experiment 1, recognition performance was solicited surreptitiously: after completing an initial gender categorization task, observers’ recall of those same images (among novel probes) was tested. In Experiment 2, observers explicitly memorized a set of faces, and were subsequently tested on different views of the same facial identities (again, among novel probes). Here, we find that SRs’ recall performance is more consistent between image- and identity-based memorability. Overall, our results suggest that intra-observer consistency is a critical behavioral marker for identification of individuals with superior face processing skill.