Skip to content
PaperFren

Can you recognise someone you met in a mask?

Open paper intelligence

Masks hurt face memory, but the damage depended on how the study and test situations were set up, and seeing the same masked face twice did not reliably help.

Source

That person is now with or without a mask: how encoding context modulates identity recognition

Garcia-Marques T, Oliveira M, Nunes L · Cognitive research: principles and implications · 2022

doi.org/10.1186/s41235-022-00379-5Read the full paper ↗9 citationscc by

Study at a glance

Design
Human experiment — Three experiments crossing mask vs no mask at study and at test: between-subjects pure lists (Exp 1), within-subjects mixed lists (Exp 2), mixed lists plus an instruction manipulation (Exp 3)
N
Separate samples: 169 students in Experiment 1, 34 students in Experiment 2, and 70 online participants in Experiment 3.
Population
Portuguese university students (Exps 1-2) and young adults recruited online via Prolific (Exp 3)
Outcome
Recognition hit rates, sensitivity (d') and response criterion (c) for studied vs new faces

Structured fields used in claim comparison tables when every cited study has a complete layer.

What they did

Participants studied 32 photographed faces, each shown for 10 seconds, with or without a surgical-style mask, then after a filler task judged whether faces were people they had seen, which could appear masked or unmasked at test. Experiment 1 gave each person only masked or only unmasked faces; Experiments 2 and 3 mixed masked and unmasked faces within one list; Experiment 3 also varied whether instructions warned that masks might be added or removed. Recognition was analysed with signal detection theory.

What they found

In Experiment 1, masks at study reduced sensitivity, and faces studied unmasked were recognised much better unmasked than masked (d = 0.77), but a clear matching benefit for masked-masked faces did not appear. With mixed lists, whether a face was masked at study no longer mattered, while masks at test consistently made people more lenient, saying 'I recognise' more readily. Warning instructions had no reliable effect, and where study and test interacted the pattern was that masks at both phases were worst, the opposite of the predicted matching effect.

The limits

What it doesn't show

It does not show how well people recognise real acquaintances in masks: the same photograph was used at study and test, which confounds recognising a person with recognising a picture, as the authors acknowledge. The within-subject signal detection analysis required an unplanned random split of new faces to create separate false-alarm rates. Experiment 2 had a small sample, several key effects were marginal, and results differed between experiments, so the pattern is not fully consistent.

Key terms

Encoding specificity
The principle that memory is best when the cues available at retrieval match those present when the information was learned.
Holistic face processing
Perceiving a face as an integrated whole rather than as separate features; masks disrupt it by hiding the lower face.
Sensitivity (d')
A signal detection index of how well someone discriminates studied items from new ones, independent of their willingness to say 'yes'.
Response criterion (c)
A signal detection index of response bias; a lower c means a more lenient tendency to say an item is old.
Mixed vs pure lists
Whether different item types are intermixed in one study list or presented in separate lists; mixing can change encoding strategies.

Flashcards

1 / 10

0 of 10 answers reviewed

Research intelligence for this paper

See its role on concept claims, tensions it is part of, placement history, and related discoveries.

Open paper intelligence

Quiz yourself

1 / 5

What did masks at test consistently do to responding?

Common questions

Did seeing someone masked at both study and test help recognition?

Not reliably. The expected matching advantage never clearly appeared, and in some conditions masked-masked was actually the worst combination.

Why use d' and c instead of just counting correct answers?

Hits alone mix memory with bias; d' separates true discrimination, and c showed people were more willing to say 'yes' to masked faces at test.

Why does it matter whether lists are mixed?

When masked and unmasked faces were mixed, masks at study stopped harming later recognition, suggesting people adapt how they encode faces to a changing environment.

More on Face processing