Face masks versus sunglasses: limited effects of time and individual differences in the ability to judge facial identity and social traits

Abstract

Some research indicates that face masks impair identification and other judgements such as trustworthiness. However, it is unclear whether those effects have abated over time as individuals adjust to widespread use of masks, or whether performance is related to individual differences in face recognition ability. This study examined the effect of masks and sunglasses on face matching and social judgements (trustworthiness, competence, attractiveness). In Experiment 1, 135 participants across three different time points (June 2020-July 2021) viewed unedited faces and faces with masks, sunglasses, or both. Both masks and sunglasses similarly decreased matching performance. The effect of masks on social judgements varied depending on the judgement and whether the face was depicted with sunglasses. There was no effect of timepoint on any measure, suggesting that the effects of masks have not diminished. In Experiment 2, 12 individuals with developmental prosopagnosia (DP) and 10 super-recognisers (SRs) completed the same tasks. The effect of masks on identity matching was reduced in SRs, whereas the effects of masks and sunglasses for the DP group did not differ from controls. These findings indicate that face masks significantly affect face perception, depending on the availability of other facial information, and are not modified by exposure.

Keywords: Face coverings; Face recognition; Prosopagnosia; Super recognisers; Trait judgements.

Fig. 1

Examples of stimuli used in the matching and ratings tasks. A An unedited comparison image; B An unedited target image; C An unedited distractor image; D Sunglasses only image; E Mask only image; F Sunglasses and mask image

Fig. 2

Signal detection theory measures from the matching task. A, B show mean discrimination (A); C, D show mean bias (b) for each image condition. A, C show data separated across timepoints; B, D show data averaged across timepoint. Error bars represent 95% within subjects confidence intervals, calculated according to the Cosineau-Morey method (as reported in Baguley, 2012). In B, D, lines with *** indicate that Bonferroni-corrected pairwise comparisons between conditions were significant, p < .05

Fig. 3

A Mean accuracy and B mean response time (correct trials only) in matched identity (“Same”) and mismatched identity (“Different”) trials. Error bars represent 95% within subjects confidence intervals, calculated according to the Cosineau-Morey method (as reported in Baguley, 2012)

Fig. 4

Mean A trustworthiness; B competence; and C attractiveness ratings for each image condition, averaged across timepoint. Error bars represent 95% within subjects confidence intervals, calculated according to the Cosineau-Morey method (as reported in Baguley, 2012). Lines with ** indicate that Bonferroni-corrected pairwise comparisons between conditions were significant, p < .05

Fig. 5

Raincloud plots of A discrimination (A) and B) bias (b) in the matching task for individuals with DP and the matched control group in each image condition. Coloured points show performance of each individual in each condition; box plots display median and upper/lower quartiles for each group in each condition, and histograms represent the distribution of data in each condition

Fig. 6

Raincloud plots of proportion correct responses in A matched identity and B mismatched identity trials in the matching task for individuals with DP and the matched control group in each image condition. Coloured points show performance of each individual in each condition; box plots display median and upper/lower quartiles for each group in each condition, and histograms represent the distribution of data in each condition

Fig. 7

Raincloud plots of A discrimination (A) and B bias (b) in the matching task for SRs and the matched control group in each image condition. Coloured points show performance of each individual in each condition; box plots display median and upper/lower quartiles for each group in each condition, and histograms represent the distribution of data in each condition

Fig. 8

Raincloud plots of proportion correct responses in A matched identity and B mismatched identity trials in the matching task for SRs and the matched control group in each image condition. Coloured points show performance of each individual in each condition; box plots display median and upper/lower quartiles for each group in each condition, and histograms represent the distribution of data in each condition