Recognising the Same Face in Different Contexts: Testing within-Person Face Recognition in Autism

Background:  

Difficulties recognising faces are well established in autism. Yet research into face recognition has focused exclusively on the ability to tell people apart, rather than on recognising the same face in different contexts. One recent study highlighted the scale of ‘within-person variability’ by demonstrating that typical adults frequently perceive differing images of the same person as different people. Effective within-person face recognition relies on using within-person variability to build stable representations of a person’s appearance. According to current theoretical accounts of autistic perception, people with autism might be less likely to make use of this variability in the generation of robust internal working models of the world. We therefore expected that children with autism would have problems extracting the variability among different images of the same face, making identity judgments about the same person across contexts difficult.

Objectives:  

To extend research on within-person variability in face recognition to autism, by comparing autistic children and typically developing children on their ability to recognise the same identity across different images.

Methods:  

Twenty-one children with autism (M age=10.62; SD=2.80) and 21 typical children (M age=10.51; SD=2.43), of similar age and intellectual ability, participated. Following a previous study, children were given 40 grayscale photographs of two distinct male identities (20 of each face; taken at different ages, from different angles and in different lighting conditions) and asked to sort them by identity within a ten-minute time limit. Importantly, children were never told how many identities to expect; they were simply to inspect the images and group them into as many identities as they perceived. We recorded the number of identities children produced (out of 40) and the number of misidentification errors per perceived identity. We predicted that children with autism would find it difficult to map diverse photos onto the same face, resulting in solutions that contained more identities than were actually presented and greater numbers of misidentification errors.

Results:  

Just like typical adults, children in both groups mistook images of the same person as images of different people, subdividing each individual into many perceived identities. While children with autism reported a similar number of identities (median=17, range=2 – 39) to typical children (median=16; range=3 – 33), they nevertheless made significantly more misidentification errors than typical children, incorrectly placing images of different identities in a single pile. Further analysis of these errors showed that, out of 400 possible mismatches between 20 images of 2 different identities, children with autism made a total of 363 mismatches compared to only 88 for typical children.

Conclusions:  

Within-person variability poses a considerable challenge to face recognition, yet has so far been neglected by autism researchers. Results suggest that children with autism may have greater difficulties than typically developing children in recognising the same facial identity shown across contexts – just as they do in telling different faces apart. We suggest that these difficulties arise as a result of problems extracting the perceptual commonalities from different images of the same person and building stable representations of facial identity.