Fig 1.
Super-recognisers’ face matching accuracy benchmarked against student controls and forensic examiners on two challenging tests of face matching.
In the facial recognition candidate list test (left), participants decided if the face pictured at the top matched any in the array (‘candidate list’) below. In the selfie-to-passport test (right), participants decided if the two images showed the same person or two different people. Super-recogniser groups were equivalent in accuracy compared to examiners on both tasks. Despite this, average performance was substantially less than 100% and no individual attained perfect accuracy.
Fig 2.
Super-recogniser performance compared to student controls and facial forensic examiners on the Expertise in Facial Image Comparison Test (EFCT; White, Phillips et al. 2015).
Participants decided whether two images, taken months apart and in different environmental conditions, showed the same person or two different people (top). Super-recognisers were more accurate than examiners when making decisions under a short time limit (2 seconds, top left panel) but not when participants were given longer to study the images (30 seconds, top right). Qualitative differences between groups were also observed in the effect of turning faces upside down (Inversion effect, bottom panels: See text for details).
Fig 3.
Super-recognisers’ superiority is largest when faces are visible.
Examples of Face Only (A) and Body Only (B) stimuli used in the Face and Body Matching Test. Super-recognisers’ superiority is largest when matching identities with the faces visible (Face Only), but they also show an advantage when matching identities using only information in the body (Body Only). Comparison of Student Controls and super-recognisers are shown using the SR-Weak criteria only, but similar patterns emerge when using the SR-Strict criteria (see S1 Appendix).
Fig 4.
Super-recognisers show superior face memory in the photoboard recognition memory test.
This test measures participants’ ability to learn faces from either one or multiple photographs. Example stimuli from the study phase (A) and test phase (B). Super-recognisers (SR-Weak) were more accurate than Student Controls when recognising faces learnt from one image or eight images but did not show any additional benefit from having access to more images.
Fig 5.
Accuracy of super-recognisers and students on the face-in place recognition memory test.
(A) Participants learned faces embedded in scenes during the study phase. (B) Participants were then shown faces embedded in either same scenes or different scenes at test. (C) Participant accuracy shown separately for two different decisions made to face-scene images at test. Super-recognisers (SR-Weak) outperformed Student Controls when recognising previously encountered faces, both when shown in same and different visual scenes (Hits) and when correctly rejecting unseen faces (correct rejections). Super-recognisers were no better at deciding whether recognised faces were shown in same or different scenes (right plot).
Fig 6.
Super-recogniser and student accuracy on the UNSW House Test.
The UNSW House Test examines whether super-recognisers’ superior face matching and memory ability generalises to non-face objects like houses. Super-recognisers (SR-Weak) were more accurate on this test than Student Controls, suggesting that some of their superiority with faces generalises to the recognition of non-face objects.
Fig 7.
Overall performance of super-recognisers across 7 follow-up face identification tests expressed as z-scores relative to control participant norms.
Positive z-scores show that super-recognisers generally outperformed standard participant groups, but this was more consistent for the groups selected using strict performance criteria (yellow markers) than for those selected using weak criteria (green markers). Dotted lines represent the weak criteria as applied to the follow-up tests (1.7 SD), to illustrate the proportion of selected SRs that continued to meet SR criteria on follow-up testing. Details of this analysis are provided in the main text.
Table 1.
Comparison of NSW Police Force staff to normative data on 3 screening tests.
See screening tests section above for sources of normative data.