Kayla Kusel¹ (), Hanxinyi He^1,2, Leah Kirsch¹, Jorge Morales³, Alison Campbell^1,2,4, Jeremy Wilmer⁵, Laura Germine^6,, Sarah Bate⁷, Joseph DeGutis^1,4,6; ¹Boston Attention Learning Lab, VA Boston Healthcare System, ²Department of Psychological and Brain Sciences, Boston University, Boston MA, ³Department of Psychology, Northeastern University, Boston MA, ⁴Department of Psychiatry, Boston University Chobanian and Avedisian School of Medicine, Boston MA, ⁵Department of Psychology, Wellesley College, Wellesley MA, ⁶Department of Psychiatry, Harvard Medical School, Boston MA, ⁷Institute for Technology and Psychiatry, McLean Hospital, Belmont MA, ⁸Department of Psychology, Bournemouth University, Poole UK

One important aspect of metacognition is awareness of one’s own cognitive abilities or performance. Lower-performers often overestimate their ability while higher-performers underestimate their ability, but it is unclear whether this ‘Dunning-Kruger effect’ is domain-specific or universal. To date, no studies have thoroughly examined whether face recognition ability, a specific ability where individuals have moderate insight, shows a Dunning-Kruger effect. Using a web-based sample (N=3,590), we first examined global metacognition, awareness of overall ability, by computing metacognitive sensitivity (correlation) and bias (difference score) using self-reported (Cambridge Face Memory Questionnaire-CFMQ) and objective (Cambridge Face Memory Test-CFMT) face recognition assessments. We found no evidence for differences in global metacognitive sensitivity (r’s from lowest-to-highest ability=.28/.32/.32/.33/.30/.25) or bias across abilities (p’s>.54). We next examined whether local, trial-to-trial, metacognition (sensitivity: m-ratio, bias: type-2 criterion) differed across face recognition ability by comparing 94 developmental prosopagnosics, 138 controls, and 18 super-recognizers performing an Old/New face recognition task with confidence ratings. For overall performance, super-recognizers outperformed controls, who outperformed prosopagnosics. Compared to controls, prosopagnosics were also more biased to say “new,” while super-recognizers were more biased to say “old.” Notably, when examining m-ratio we found no significant sensitivity differences between prosopagnosics (M=.98, SD=.72), controls (M=.91, SD=.74), or super-recognizers (M=.93, SD=.48). When examining response-specific type-2 criteria, no group differences were observed when subjects classified a face as “new.” However, significant differences emerged in confidence-rating patterns when subjects responded “old” (DPs: M=-1.34, SD=.41, controls: M=-1.0, SD=.44, super-recognizers: M=-.63, SD=.58, p’s<.002). In other words, prosopagnosics were more conservative when making high-confidence judgments whereas super-recognizers were more liberal. Together, these findings are inconsistent with a Dunning-Kruger effect, but rather reveal that face recognition ability affects confidence biases when encountering previously studied faces.

Acknowledgements: This work was supported by a grant to JD from the National Eye Institute (R01 EY032510-02).