Abstract
This review estimated the strength of the relationship between self-reported and objectively measured face identity recognition ability and examined potential moderators of this relationship. CINAHL, Medline, PsycInfo, Scopus, and ProQuest Dissertation and Theses databases were searched on April 16, 2025, for two concepts: insight and face recognition. Studies were included if they were written in English, comprised adult samples, included an objective face recognition task and a measure of self-reported face recognition ability, and reported the relationship between the two. Forty-four studies ( N = 11,074) comprising 141 effect sizes were included in a random effects meta-analysis with robust variance estimation, which revealed a significant, positive relationship between self-reported and objective face recognition ability of medium strength ( r = .26, 95% CI [.22, .30]). Moderator analyses revealed a stronger effect size for samples comprising a combination of neurotypicals and developmental prosopagnosics ( r = .64, 95% CI [.47, .80]) compared to studies comprising only neurotypicals ( r = .27, 95% CI [.22, .32]), developmental prosopagnosics ( r = .15, CI [−.04, .35]), or super-recognisers ( r = .16, 95% CI [.05, .28]). A stronger effect size was observed for face memory tasks ( r = .30, 95% CI [.26, .34]) compared to face perception tasks ( r = . 21, 95% CI [.15, .28]). Considerable heterogeneity was evident between studies comprising neurotypical samples, which may be attributable to methodological differences in exclusion criteria and participant characteristics related to face recognition ability, which were generally poorly reported across studies. Collectively, findings indicate people have modest insight into their face recognition ability, and self-report measures offer limited utility for identifying face recognition difficulties.