Autism is a complex disorder characterized by a lack of social interaction and eye contact, and is typically diagnosed by three years of age. An earlier diagnosis — in the first few months of life — could improve outcome. With this in mind, Ami Klin, a clinical psychologist and director of the autism programme at the Yale School of Medicine, and his colleagues delved into the social development of very young children.

Young animals, from human babies to newly hatched chicks, preferentially focus their attention on the movement of living beings rather than inanimate objects — a trait that enables them to orient to a caregiver, necessary for survival. In 2000, Klin, accompanied by Yale neuroscientist and co-author Warren Jones, went to an animatronic studio in California to create animations that could be used to determine how young children respond to human movement, and how this response might differ in autistic children.

They used a technique that turns videos of human actors playing children's games, such as 'peek-a-boo', into animated dots of light able to convey human motion. Then, by tracking the eye movements of children watching the moving dots, Klin and Jones could measure the children's attention to human movement, and thus social interaction. “The eyes are the window to the soul, but also to socialization,” says Klin.

A puzzling observation focused their efforts. A 15-month-old girl whose brother had autism was shown a screen divided in two: on one side the light displays were upright; on the other they were inverted, so no longer representative of human movement, and played backwards. The girl showed no preference for upright or inverted images — with one exception. During the 'pat-a-cake' video, one of the nine animations presented to her, she focused almost entirely on the upright video. “At first, we were confused,” says Klin. However, further investigation established that this was the only animation in which the sound — in this case, clapping — was clearly synchronized to the light movement.

The duo suspected that autistic children might be more attentive to a physical stimulus (sound synchronized to motion) than to a social one (human movement). To test the idea, they showed the animations to groups of 2-year-old children with and without autism, and found that only those with autism had the same response as the 15-month-old girl (see page 257). “Then we sat back and thought we should be adventurous in order to learn the profound lesson this little girl was teaching us,” says Klin.

They and their colleagues at Yale spent the next two years coming up with a method to quantify how much audiovisual synchrony there was in different animations — and compared that measurement with children's visual behaviour. In the end they were able to predict, with 90% accuracy, the children's visual preference on the basis of even incidental bits of audiovisual synchrony present in the animations.

Klin says the little girl helped them to understand that autistic children develop in a world where preferential attention is given to physical, rather than social, attributes. On the basis of this realization, Klin and Jones are now looking at ways to pinpoint when this divergence from typical social development begins. “We want to come up with a behaviour assay that will predictably diagnose vulnerabilities for autism in the first year, if not months, of life,” Klin says.