Rethinking eye contact as an autism sign

If you’ve ever been told that avoiding eye contact is a surefire sign of autism, you might want to sit down for this. A new study out of East China Normal University suggests that the truth is far more nuanced, and it might just change how we think about one of autism’s most well-known “symptoms.”

Lu Qu and Qiaoyun Liu, researchers at the university’s Shanghai Institute of Artificial Intelligence for Education, decided to take a fresh look at an old assumption: that children with autism avoid looking at people’s faces, and that this avoidance is a key marker for diagnosis. It’s an idea that dates back at least to 1943, when child psychiatrist Leo Kanner famously described autistic children as those who “never looked at any person’s face.”

That description stuck. For decades, therapists, educators, and parents have seen direct eye contact (or the lack of it) as a window into a child’s social world. In many clinics, tests like the Early Social Communication Scales (ESCS) or the Autism Diagnostic Observation Schedule (ADOS) still measure a child’s ability to engage in “joint attention,” often by counting how many times the child looks from a toy to an adult’s face and back again.

However, these tests usually happen in highly controlled lab settings that are nothing like a child’s natural play environment. The researchers wondered: what if the problem isn’t the child’s eye contact, but the way we’ve been measuring it?

To find out, they built a non-intrusive “multimodal behavior observation lab” that looked more like a playroom than a science experiment. Cameras and AI tools quietly tracked children’s gaze, gestures, and expressions while they played freely with their parents. And they didn’t just look at autistic children, they also included typically developing kids and those with developmental delays.

The results are literally eye-opening. Across the board, children (autistic or not) spent most of their time looking at toys (around 60–80% of the time) and only a sliver of their time looking at an adult’s face (as little as 6–14%). In other words, it’s perfectly normal for kids, when deeply engaged in play, to ignore your face.

A similar study published in Current Biology found almost the same thing, even using a completely different method, head-mounted eye trackers. That research showed children of all developmental profiles spent barely 1% of playtime looking at a parent’s face, but paid plenty of attention to the parent’s hands.

That detail might be the key. It turns out kids may not need to lock eyes with you to share attention. They might be reading your gestures, following your hands, and tuning into other cues entirely.

So what does this mean for autism interventions? Traditionally, many therapies have focused on helping children make more eye contact. But if joint attention can happen through other channels such as gestures, then perhaps we should be broadening our approach. Encouraging a child to watch where your hands are pointing, for example, might be more natural and just as effective for building communication skills.

Qu and Liu’s work, alongside the Current Biology findings, reminds us that understanding human behavior is particularly complex. Sometimes what we think is a “core symptom” turns out to be just one part of a much more complex picture. And in the age of artificial intelligence, we have more tools than ever to study behavior in real-world contexts rather than sterile labs.

If you want to learn more, the original article titled "Is a Child Who Doesn’t Look at People Always Autistic?—A Closer Look at Joint Attention" on ECNU Review of Education at https://doi.org/10.1177/20965311251319050