Study: Masks Impair the Wearers' Ability to Recognize Faces

Since the pandemic began, there have been dozens of published research articles detailing the difficulty of recognizing masked faces. I reviewed some of these findings in two prior posts, here and here. Generally, it has been found that masked faces are more difficult to recognize across several tasks, including identity recognition and emotional expression recognition.

However, researchers Erez Freud, Daniela Di Giammarino, and Carmel Camilleri from York University set out to ask a different question: Does the observer's mask-wearing status also influence face recognition? They hypothesized that when an observer is wearing a face mask, they may experience more difficulty perceiving and recognizing faces (whether or not those faces are masked themselves).

To test this, they conducted four online experiments, which were published in this month's issue of Cognitive Research: Principles and Implications. In each experiment, participants were asked to wear a face mask for half of the experiment and take it off for the other half. (The order of mask-wearing was counterbalanced across participants, such that half completed the first part with their mask on, and half completed the first part with their mask off.)

In the first experiment, participants completed the Glasgow Face Matching Test (GFMT; Burton et al., 2010) of face recognition, in which they had to identify faces across changes in view. None of the face images had masks; only the observers did (in half of the trials). Their findings revealed a robust impairment of face recognition performance during mask-wearing. When observers were wearing a face mask, their performance (computed as d-prime, which is a function of hits and false alarms) was approximately 2.54; when they completed the task without a mask, performance was approximately 2.82, a significant improvement.

The researchers considered the possibility of a congruence effect. Since the face stimuli were all unmasked, it could be that the congruence between observers' unmasked faces and the unmasked face stimuli led to better performance. To test this, the second experiment presented all face stimuli with masks. The findings again showed that performance was significantly better when participants were not wearing a mask (d-prime = 2.48) compared to when they were wearing a mask (2.16). This suggests that the findings from the first experiment were not driven by a congruence effect.

Next, Freud and colleagues considered whether mask-wearing might lead to general impairments in perception over and beyond face recognition. One reason for this is that mask-wearing is known to affect breathing patterns, which could in turn lead to changes in cognitive and perceptual performance (see Kluger et al., 2021). It could also be that face masks interfere directly with visual perception, or that participants with glasses may have had masks fog up their view of the stimuli. To test whether mask-wearing led to general deficits in visual perception, or specific deficits in face recognition, the researchers tested performance on a novel object recognition task. The task was otherwise identical to the face recognition task, except that the stimuli were novel three-dimensional objects presented at slightly different views. Interestingly, the findings showed no difference in performance based on observers' mask-wearing status.

Finally, the researchers tested whether simply having a mask touching their face was the key factor. They repeated the first experiment, but instructed participants to wear their masks over their foreheads in the mask-wearing condition (rather than over their mouth and nose). This experiment found that wearing a mask over one's forehead made no difference in performance in the face recognition task.

Overall, the researchers concluded that wearing a face mask (specifically over one's mouth and nose) specifically impairs face recognition. Why might this be the case?

Previous studies have shown that face perception depends on more than just visual input and can be affected by bodily factors. For example, lying sideways on a bed affects how we perceive faces at different orientations compared to sitting upright (see Davidenko and Flusberg, 2012). This is related to the concept of embodied cognition, wherein the state of one's own body and physical affordances influences perceptual and cognitive processes. In the case of wearing a face mask, it could be that one's awareness of one's own covered face during a face recognition task impairs perceptual processes related to face recognition. Thus, the findings of Freud and colleagues add to a growing body of evidence that relates observers' perception and cognition to their bodily states.