Measuring children’s gaze
patterns as they watch movies of social interactions is a reliable way to
accurately identify nearly half of autism spectrum disorder (ASD) cases,
according to a new study just published in Autism
Research by BGU researchers.
“Eye tracking is likely to be one of the first technologies that will be
incorporated into clinical use for assessment of ASD symptoms, but it needs to
be optimized to identify and quantify specific ASD symptoms,” explains Prof. Ilan Dinstein, of BGU’s Departments of Psychology and Cognitive and Brain
Sciences and director of Israel’s National Autism Research Center. “This new
study takes a first important step in this direction using eye tracking
technology to compare different movies and measures within the same group of
According to the United States Center for Disease Control (CDC), one in 59
children in the U.S. have ASD. Generally, when typically developing children
watch movies of social interactions, they do so in a reliable and predictable
manner, observing faces, gestures, body movements, and objects that are
relevant to the social interaction and its narrative. In the new study, the
researchers demonstrate that children with ASD watch such movies with
significantly more variable and idiosyncratic gaze patterns.
Previous eye tracking studies have reported that children with ASD fixate less
on faces in comparison to control groups. However, children must also gaze at
actions, gestures, body movements, contextual details, and objects that are
part of the social narrative, thereby creating complex gaze patterns to
properly understand social interactions.
In the current study, the researchers presented ASD and control children with
three short movies, each shown twice. Two of the movies were animated and one
was a realistic home video; all contained social interactions between at least
two individuals. This experimental design allowed comparisons across movies,
presentations and different eye tracking measures to identify what is the best
technique for identifying ASD children based on differences in gaze behavior.
Since children watch movies in a predictable manner, the gaze pattern of
individual children is remarkably similar to the mean gaze pattern of their
group. In other words, typically developing children agree on where and when to
look at specific locations in the frame.
In contrast, children with ASD exhibited significantly more
variable/idiosyncratic gaze patterns that differed from the mean gaze pattern
of the typically developing children. Moreover, their gaze patterns were
remarkably inconsistent not only between individuals but also across movie
presentations. Hence, when children with ASD watch the same movie repeatedly,
they have more variable and inconsistent gaze patterns.
“Quantifying this gaze idiosyncrasy in individual children enabled separation
of ASD and control children with higher sensitivity and specificity than
traditional measures such as time gazing at faces,” Dinstein says. “It was also
strongly correlated with their social symptom’s severity.”
The largest differences across ASD and control groups were apparent when using
a realistic video containing a social interaction between two sisters (two and
five years-old) in a messy room with everyday objects. This suggests that
abnormal, idiosyncratic gaze patterns were most pronounced when ASD children
observed real-life unedited interactions of other children. This makes the
findings particularly relevant to real-life social situations.
“Taken together, these results demonstrate that ASD children with more severe
symptoms exhibit larger gaze idiosyncrasy,” Dinstein says. “This can aid not
only in early detection of autism but also in assessing changes in ASD
severity over time and in response to treatments. Such measures, which
objectively measure symptoms directly from the child, are critically lacking in
today’s clinical trials of autism treatments.”
Above: Example of gaze fixation points for
10 control children (blue) and 10 ASD children (green) on an individual frame
of the naturalistic movie. The mean gaze position of the control group is
marked with a cross.