Abstract
The mask mandate during the COVID-19 pandemic brought attention to the challenge of recognizing masked faces. Common intuition suggests that the difficulty arises from the fact that only the upper half of a masked face is visible and the shapes of chin and mouth do not contribute to recognition. This study sought to understand whether the difficulty stems from a lack of visual information or from the recruitment of new cognitive abilities rarely used in face recognition. Specifically, we investigated whether the same cognitive abilities were employed for the recognition of masked faces and non-masked faces. We selected two sets of face images from the Karolinska Directed Emotional Faces database – each comprising face images of 4 men and 4 women displaying various facial expressions (non-masked face image set). Subsequently, masked versions of these faces were generated for each set (masked face image set). Participants’ face recognition abilities were measured for each of the two non-masked image sets and two masked image sets separately. They memorized two face images portraying different identities, and then after a brief interval, selected one face image depicting an identity not present in the previous display. We found a positive correlation between recognition performance for the two masked face image sets, after controlling for the influence of non-masked face recognition. Whereas Bayesian evidence favored the absence of correlation between recognition performance for the two non-masked face image sets, after controlling for the influence of masked face recognition. These results suggest unique variability in masked-face recognition that cannot be explained solely by individual differences in face recognition and that facial features covered by a mask are not a significant source of difficulty in masked-face recognition.