Published September 21, 2026, in eLife, this study analyzed 83 high-density EEG recordings from 44 infants aged 2.5 to 22.6 months. Twenty-five infants had a higher likelihood of autism and 19 had a lower likelihood.
A layered view of speech learning Researchers measured neural tracking of syllables and word structure, then recorded event-related responses during a word-recognition task. The higher-likelihood group showed weaker syllable tracking, which related to verbal outcomes around 20 months. Statistical learning itself appeared preserved, while late novelty-related responses were reduced.
Why longitudinal evidence helps Repeated recordings can show how processing changes across development, rather than relying on a single visit. Analyses that excluded the final visit still linked earlier EEG measures with later verbal outcomes.
The important boundary This was a small observational cohort with uneven repeated measurements, sleep-state effects and some excluded recordings. Autism likelihood is not a diagnosis, and these EEG measures are not validated screening or prognostic tests.
Source and disclosure Full text checked. Supported by the Swiss National Science Foundation; the authors declared no competing interests. Prepared with AI assistance and source-checked September 24, 2026.