Capturing specific abilities as a window into human individuality: The example of face recognition

Abstract

Proper characterization of each individual's unique pattern of strengths and weaknesses requires good measures of diverse abilities. Here, we advocate combining our growing understanding of neural and cognitive mechanisms with modern psychometric methods in a renewed effort to capture human individuality through a consideration of specific abilities. We articulate five criteria for the isolation and measurement of specific abilities, then apply these criteria to face recognition. We cleanly dissociate face recognition from more general visual and verbal recognition. This dissociation stretches across ability as well as disability, suggesting that specific developmental face recognition deficits are a special case of a broader specificity that spans the entire spectrum of human face recognition performance. Item-by-item results from 1,471 web-tested participants, included as supplementary information, fuel item analyses, validation, norming, and item response theory (IRT) analyses of our three tests: (a) the widely used Cambridge Face Memory Test (CFMT); (b) an Abstract Art Memory Test (AAMT), and (c) a Verbal Paired-Associates Memory Test (VPMT). The availability of this data set provides a solid foundation for interpreting future scores on these tests. We argue that the allied fields of experimental psychology, cognitive neuroscience, and vision science could fuel the discovery of additional specific abilities to add to face recognition, thereby providing new perspectives on human individuality.

Do People Have Insight into their Face Recognition Abilities?

“…However, this conclusion is only valid if the behavioural tests used truly reflect daily-life face recognition abilities. Regarding this point, many of the tests used have been shown to measure both individual differences in ability (e.g., the CFMT, Wilmer et al, 2012) and are used to diagnose CP (e.g., the CFMT, Duchaine & Nakayama, 2006). Importantly, these tests also appear to identify cases of acquired prosopagnosia (e.g., CFMT, Liu-Shuang et al, in press; 4AFC tests used in Study 4, Busigny et al, , 2014Rossion et al, 2003).…”

“…The CFMT has been validated on the basis of its ability to diagnose people with acquired prosopagnosia (Liu-Shuang, Torfs & Rossion, in press;Susilo, Yovel, Barton, & Duchaine, 2013) and the much lower scores observed for inverted as compared to upright faces (Duchaine & Nakayama, 2006). Also, performance on the CFMT correlates with performance on a face matching test without any learning component (i.e., CFPT) (Bowles et al, 2009;Duchaine et al, 2007) but the test displays only modest correlations with measures of nonface visual memory (Dennett et al, 2012;Wilmer et al, 2010) and even weaker correlations with measures of verbal memory (Wilmer et al, 2012). The CFMT is also reliable, as measured with Cronbach's alpha, a measure of internal consistency (α = .89, Bowles et al, 2009;α = .83, Herzmann, et al, 2008;α = .90, Wilmer et al, 2010).…”

“…A novel feature of this study is that we administered five tests of face recognition ability. Two of the tests are well-established: the CFMT (Duchaine & Nakayama, 2006) is a valid and reliable test of face learning and memory that is sensitive to variation in typical adults (e.g., Bowles et al, 2009;Wilmer et al, 2012) and the BFRT (Benton, Sivan, Hamsher, Varney, & Spreen, 1983) involves simultaneous face matching . Three other tests were newly developed: the CFMT-Films, which is based on the CFMT format but differs in that the study stage involves watching short film clips of people interacting (see Supplementary Materials) and two sequential face-matching tests, one in which front view faces are matched and another in which the faces vary in viewpoint from target to test.…”

Item response theory analyses of the Cambridge Face Memory Test (CFMT).

“…Here we focused on gender and age because these factors have been found to affect performance on face recognition tasks, including the CFMT, and the correct interpretation of such effects depend on the validity of comparing these individuals on the same scale. Gender differences have sometimes been reported to be particularly large for female faces (Lovén et al, 2011; Lewin & Herlitz, 2002), but a small 0.15 SD difference favoring females was also found for the CFMT, which was developed using only male faces to minimize this gender difference (Wilmer et al, 2012). So far, gender effects have been interpreted as reflecting quantitative differences on the same unidimensional construct, but the sort of analyses required to test this assumption have not been conducted.…”

Normative data for accuracy and response times at the computerized Benton Facial Recognition Test (BFRT-c)

“…Second, there is a relatively recent and growing interest in interindividual variability of face processing abilities, which is generally claimed to be much wider than initially thought (e.g., Bowles, McKone, Dawel, Duchaine, Palermo, Schmalzl, Rivolta, Wilson, & Yovel, 2009;Herzmann et al, 2008;Megreya & Burton, 2006;Palermo, Rossion, Rhodes et al, 2017;Wilmer, Germine, Chabris, Chatterjee, Williams, Loken, Nakayama, & Duchaine, 2010;Wilmer, Germine, Chabris, Chatterjee, Gerbasi, & Nakayama, 2012).…”