Autistic Children's Face-Processing Brain Signals Diverge From Neurotypical Peers Earlier Than Thought

Each child in the study sat while pictures of faces and everyday objects appeared in front of them. Around their scalps, 128 electrodes recorded the brain's electrical activity. The question was whether a computer, looking only at those recordings, could tell what the child was seeing.

For neurotypical children, it could do so more clearly than for autistic children. That is the central finding of a study led by researchers at Yale's Child Study Center, drawing on roughly 400 autistic children. The electrical signals tied to viewing faces were simply less distinguishable in the autistic group. First author Jason Griffin, now an assistant professor of psychology at the University of Houston and formerly a postdoctoral researcher in the lab of James McPartland, put it this way: the neural activity measured at the scalp is just not as distinct in autistic children.

The study also tracked age. Among neurotypical children, the ability to identify what a child was viewing from brain activity improved as they got older, which fits the idea that the brain's response to faces becomes more refined and specialized with time. That improvement was not seen in autistic children. In Griffin's words, their face-specific processing is not following the same trajectory.

The work builds on decades of research. Some autistic people look at faces less often, make less eye contact, and may find it harder to remember faces or read expressions. In the early 2000s, Yale researchers reported that brain regions responsible for face processing were less active in autistic individuals than in others. McPartland, who is the Harris Professor in the Child Study Center and directs the Center for Brain and Mind Health, studies this with electroencephalography, or EEG. Earlier EEG studies generally looked at limited regions of the scalp. McPartland compares that to trying to understand a city's traffic by watching a single downtown street. Using all 128 electrodes is closer to viewing the whole city by satellite. He says this is the first time the full volume of information in an EEG recording has been used to study face perception in autism.

The data came from the Autism Biomarkers Consortium for Clinical Trials, which includes autistic and neurotypical children across different ages. The team applied machine learning to the recordings to see whether patterns could reveal what each child was looking at.

The timing of the differences may be the most surprising result. Much earlier research has focused on the N170, a change in electrical activity that typically occurs about 170 milliseconds after someone sees a face, and which tends to arrive later in autistic individuals. By examining a wider span of time, the researchers found evidence that differences appear before the N170. That suggests autism-related differences arise at earlier stages of visual processing than previously established. McPartland says that looking at the full temporal range shows differences occurring even earlier.

The researchers also argue the pattern points to developmental characteristics of autism, not just the accumulated effects of living with it, a distinction that has long been hard to draw. Because the differences show up in development, support to help autistic children interpret facial emotions might be especially useful if introduced earlier in childhood. The source notes that some clinicians once tried to address face-perception difficulty by requiring eye contact, and research has since shown that doesn't work. The researchers also stress that many neurodivergent people experience little or no impairment and may neither need nor want support. For others, McPartland says, trouble perceiving information in a face can complicate everything from a playground to a job interview.

There is a diagnostic angle too. Autism is currently diagnosed mainly by evaluating behavior, with no reliable biological test for routine use. Scientists are investigating several possible biomarkers, and these face-related brain patterns could contribute, perhaps by helping identify who might benefit from particular supports such as social skills therapy. More research is needed before any clinical use. McPartland calls the work a first step in introducing a new biomarker.