EEG Study of 400 Children Shows Autistic Face Responses Lack Age-Related Specialization

McPartland’s team re-analyzed ABC-CT data using all 128 electrodes rather than the usual handful of occipito-temporal sites. Machine-learning classifiers trained on the broadband signals could decode whether a child was viewing a face or an object far more accurately in neurotypical than autistic participants. The age-by-diagnosis interaction was decisive: neurotypical accuracy rose steadily across childhood while autistic accuracy remained flat, indicating the typical trajectory of face specialization is absent rather than merely delayed.

Earlier Yale work had reported reduced fusiform activation; the new distributed analysis reveals the difference is not simply weaker activation but a less categorical spatial-temporal pattern. This reframes the phenotype from “hypoactivation” to “reduced representational distinctiveness,” consistent with predictive-coding accounts that autistic cortex maintains broader priors for social stimuli.

The finding aligns with two independent lines of evidence: longitudinal eye-tracking showing persistently reduced face looking and structural MRI meta-analyses documenting atypical development of the fusiform gyrus. Together they suggest the divergence begins before the age range captured here and is not explained by attention or compliance alone.

Next steps include testing whether the same distributed EEG signature predicts response to social-skills interventions in an ongoing ABC-CT extension trial; if confirmed, the metric could serve as a stratification biomarker rather than a diagnostic one.