How do face masks affect remembering faces, and can gaze help?
Masks made unfamiliar faces harder to remember in every combination tested, most of all when a face was seen masked at one time and unmasked at the other, and people who adapted where they looked were less affected.
Source
Impact of mask use on face recognition: an eye-tracking study
Study at a glance
- Design
- Human experiment — 2 x 2 within-subject old/new face recognition task (masked vs unmasked at learning x at recognition) with eye tracking analysed by hidden Markov models, plus intelligence and executive-function tests
- N
- N=88 · 88 university students or graduates; each completed eight learning-recognition blocks
- Population
- University students and graduates aged 17-30 (51 women), viewing unfamiliar Asian faces
- Outcome
- Recognition sensitivity (d') and response time; eye movement pattern (eyes-focused vs broader) and gaze consistency (entropy)
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
Eighty-eight students memorised sets of unfamiliar faces and later judged which faces they had seen, while an eye tracker recorded their gaze. Each face appeared masked or unmasked at learning and masked or unmasked at test, giving four conditions, with mask colour and lighting changed between phases so pictures were never identical. Gaze sequences were modelled with hidden Markov models to classify each person's looking pattern and its consistency, and participants also completed tests of intelligence, working memory, planning and attention.
What they found
Compared with unmasked faces at both stages, recognition accuracy dropped substantially when masks were worn at learning only, at test only and at both stages, with a strong interaction showing the worst performance when mask status did not match between learning and test. When recognising masked faces, people looked more at the eyes and made more consistent gaze transitions, whereas gaze during learning did not change with masks. People who shifted more toward an eyes-focused pattern when recognising unmasked faces learned with masks lost less accuracy (a modest correlation), and this held after controlling for intelligence and executive abilities.
The limits
What it doesn't show
The links between gaze change and accuracy are modest correlations, so they do not show that changing gaze strategy causes better recognition; a training study would be needed to test the authors' suggestion of teaching eye strategies. Participants were young, educated adults viewing unfamiliar Asian face photos with digitally added masks, so the results may not apply to familiar faces, real-life encounters, or the children, older adults and autistic people the authors speculate are vulnerable. The eyes-focused advantage was specific to one mismatch condition, and the gaze-consistency correlation became only marginal once cognitive abilities were controlled.
Key terms
- Encoding specificity
- The principle that memory retrieval works best when the information available at test matches what was available when the memory was formed.
- d prime (d')
- A signal detection measure of how well someone distinguishes old from new items, independent of their tendency to say 'yes'.
- Hidden Markov model (HMM) of eye movements
- A statistical model that summarises a person's gaze as a set of looked-at regions and the probabilities of moving between them.
- Entropy (gaze consistency)
- A measure of how predictable someone's fixation sequence is; lower entropy means a more consistent looking routine.
- Holistic face processing
- Perceiving a face as an integrated whole rather than as separate features, which masks are known to disrupt.
Flashcards
0 of 11 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
Compared with unmasked faces at both stages, when was recognition accuracy impaired?
Common questions
Why was a mismatch between learning and test the hardest condition?
Memory is best retrieved with the same cues present at encoding. A face learned masked lacks nose and mouth information in memory, and a face tested masked hides features that were encoded, so the mismatch leaves less overlapping information.
Did people look at masked faces differently when memorising them?
No. Gaze during the learning phase did not differ between masked and unmasked faces; the changes appeared during recognition, when masked faces drew a more eyes-focused, consistent pattern.
Can we conclude that training people to look at the eyes will help?
Not from this study. It shows that people who naturally shifted toward the eyes were less impaired, but it did not test whether instructing people to do so improves recognition.
More on Face processing