Skip to content
PaperFren

Which part of face recognition is harder for autistic adults?

Open paper intelligence

Autistic adults were worse at perceiving faces accurately and at remembering them, but once perception was taken into account they were just as good as others at deciding whether two faces were the same person.

Source

Independent measurement of face perception, face matching, and face memory reveals impairments in face perception and memory, but not matching, in autism

Stantić M, Brown K, Ichijo E, et al. · Psychonomic bulletin & review · 2023

doi.org/10.3758/s13423-023-02304-3Read the full paper ↗7 citationscc by

Study at a glance

Design
Case-control — Study 1: correlational individual-differences study using regression of face-memory scores on separate perception and matching measures; Study 2: autistic vs matched neurotypical adults on the same tests, with group entered into the regressions
N
Study 1 analysed 1,112 neurotypical adults online; Study 2 compared 57 autistic adults with 57 neurotypical adults matched on age, gender and non-verbal IQ
Population
Adults recruited online (Study 1); clinically diagnosed autistic adults from a research volunteer database and matched Prolific participants (Study 2)
Outcome
Face perception (deviation of similarity ratings from face-recognition algorithms on the Oxford Face Matching Test), face matching accuracy (OFMT, Glasgow Face Matching Test) and face memory (Cambridge Face Memory Test)

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

What they did

In the Oxford Face Matching Test, people rate how similar two faces look and then decide whether they show the same person; comparing their similarity ratings with scores from face-recognition algorithms gives a measure of face perception, and their same/different decisions give a measure of face matching. Study 1 gave this test, another matching test and the Cambridge Face Memory Test to over a thousand adults online. Study 2 gave the same tests to 57 autistic adults and 57 neurotypical adults matched on age, gender and non-verbal IQ, and used regression to see which abilities differed after the others were held constant.

What they found

In Study 1, better face perception predicted better face matching, and perception and matching each independently predicted face memory; together they explained 27.3% of the variance in memory scores. In Study 2, the autistic group scored lower on all three tests. However, after controlling for face perception, the autistic group did not differ in face matching, while their face memory was still worse even after controlling for both perception and matching. The authors conclude autism involves impaired face perception and memory but intact face matching, unlike developmental prosopagnosia where all three are impaired.

The limits

What it doesn't show

Study 2 was small relative to Study 1, and the authors' own power analysis showed only about 50% power for modest effects, so the null result for face matching could partly reflect limited power. The approach assumes that matching-test performance is made up only of perception and matching, while other abilities such as working memory or attention probably contribute too. Autistic participants varied widely, with a sizeable minority performing better than the typical neurotypical participant, so the group averages do not describe every autistic person. Neither study was preregistered, and the design is correlational, so it identifies which sub-process differs rather than why.

Key terms

Face perception
Building an accurate visual representation of a face; here measured by how closely a person's similarity ratings of face pairs agree with algorithmic similarity scores.
Face matching
Deciding whether two face images show the same person or two different people, based on knowledge of how faces vary within and between individuals.
Face memory
Storing and retrieving faces seen before; here measured with the Cambridge Face Memory Test.
Process purity
Whether a test measures just one psychological process; most face tests do not, because poor scores can come from several different underlying problems.
Statistical control
Including a variable in a regression so the effect of another variable is estimated while holding that one constant.
Developmental prosopagnosia
A lifelong difficulty recognising faces without brain injury; in earlier work by these authors it involved impairments in all three face processes.

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 / 6

How was face perception measured in this research?

Common questions

Why can't a normal face-matching test tell perception and matching apart?

Someone might fail because they cannot see the faces accurately or because they make poor same/different decisions despite seeing them well; a single accuracy score mixes both.

Why use computer algorithms to judge face similarity instead of human raters?

Algorithms outperform humans on face recognition and avoid building in the judgements of whichever group does the rating, such as neurotypical raters when testing autistic participants.

What is the practical implication of intact face matching in autism?

The authors suggest that if face perception could be improved through intervention, face recognition should improve too, because the decision process itself appears to work normally.

More on Autism