The ability to recognize faces varies considerably between individuals, but does performance co-vary for tests of different aspects of face processing? For 397 participants (of whom the majority were university students) we obtained scores on the Mooney Face Test, Glasgow Face Matching Test (GFMT), Cambridge Face Memory Test (CFMT) and Composite Face Test. Overall performance was significantly correlated for each pair of tests, and we suggest the term f for the factor underlying this pattern of positive correlations. However, there were large variations in the amount of variance shared by individual tests: The GFMT and CFMT are strongly related, whereas the GFMT and the Mooney test tap largely independent abilities. We do not replicate a frequently reported relationship between holistic processing (from the Composite test) and face recognition (from the CFMT)-indeed, holistic processing does not correlate with any of our tests. We report associations of performance with digit ratio and autismspectrum quotient (AQ), and from our genome-wide association study we include a list of suggestive genetic associations with performance on the four face tests, as well as with f.

The Mooney Face Test is a widely used test of face perception, but was originally designed to be administered by personal interview. We have developed a three-alternative forced-choice version for online testing. We tested 397 healthy adults between the ages of 18 and 42 (M=24 years). There was a wide range of performance (64-100% correct; M=89.6%). We observed a significant sex difference favoring males (.31 standard deviation; p =.004). In addition, independently of sex, higher 2D:4D digit ratios were significantly associated with higher scores (ρ=.14, p=.006). A genome-wide association study (GWAS) for a subset of 370 participants identified an association between Mooney performance and a polymorphism in the RAPGEF5 gene (rs1522280; p=9.68×10(-8)). This association survives a permutation test (p=.031).

A factor analysis was performed on 25 visual and auditory performance measures from 1060 participants. The results revealed evidence both for a factor relating to general perceptual performance, and for eight independent factors that relate to particular perceptual skills. In an unrotated PCA, the general factor for perceptual performance accounted for 19.9% of the total variance in the 25 performance measures. Following varimax rotation, 8 consistent factors were identified, which appear to relate to (1) sensitivity to medium and high spatial frequencies, (2) auditory perceptual ability (3) oculomotor speed, (4) oculomotor control, (5) contrast sensitivity at low spatial frequencies, (6) stereo acuity, (7) letter recognition, and (8) flicker sensitivity. The results of a hierarchical cluster analysis were consistent with our rotated factor solution. We also report correlations between the eight performance factors and other (non-performance) measures of perception, demographic and anatomical measures, and questionnaire items probing other psychological variables.

Individual differences in face recognition have been linked to a single-nucleotide polymorphism (rs237887) of OXTR, the gene that encodes the oxytocin receptor (Skuse et al., 2014). In that study, participants were assessed by Warrington's test of FaceRecognition, but performance on Warrington's test has been shown not to rely purely on face recognition processes. We administered the widely used Cambridge Face MemoryTest-a purer test of face recognition-to 370 participants. Performance was not significantly associated with rs237887, or with 16 other SNPs of OXTR that we genotyped, or with a further 75 imputed SNPs. We also administered three other tests of face processing (Mooney Face Test, Glasgow Face Matching Test, and Composite Face Test), but performance was never significantly associated with rs237887, or with any of the other genotyped or imputed SNPs, after corrections for multiple testing.Additionally, we found no associations between OXTR and autism-spectrum quotient (AQ).