Prediction efficiency and incremental processing strategy during spoken language comprehension in autistic children: an eye-tracking study

Language difficulties are common in autism, with several theoretical perspectives proposing that difficulties in forming and updating predictions may underlie the cognitive profile of autism. However, research examining prediction in the language domain among autistic children remains limited, with inconsistent findings regarding prediction efficiency and insufficient investigation of how autistic children incrementally integrate multiple semantic elements during language processing. This study addresses these gaps by investigating both prediction efficiency and incremental processing strategy during spoken language comprehension in autistic children compared to neurotypical peers. Using the visual world paradigm, we compared 45 autistic children (3-8 years) with 52 age-, gender-, and verbal IQ-matched neurotypical children. Participants viewed arrays containing a target object and three semantically controlled distractors (agent-related, action-related, and unrelated) while listening to subject-verb-object structured sentences. Eye movements were recorded to analyze fixation proportions. We employed cluster-based permutation analysis to identify periods of sustained biased looking, growth curve analysis to compare fixation trajectories, and divergence point analysis to determine the onset timing of predictive looking. Both groups demonstrated predictions during spoken language comprehension and employed similar incremental processing strategies, showing increased fixations to both target objects and action-related distractors after verb onset despite the latter's incompatibility with the agent. However, autistic children exhibited reduced prediction efficiency compared to neurotypical peers, evidenced by significantly lower proportions of and slower growth rates in fixations to target objects relative to unrelated distractors, and delayed onset of predictive looking. Reduced prediction efficiency was associated with higher levels of autism symptom severity in the autistic group and increased autistic traits across both groups, with autism-related communication difficulties showing the most robust associations. Our sample included only autistic children without language impairments, limiting generalizability to the broader autism spectrum. The task employed only simple sentence structures in controlled experimental settings, which may not fully capture language processing patterns in naturalistic communication contexts. While autistic children employ similar incremental processing strategies to neurotypical peers during language comprehension, they demonstrate reduced prediction efficiency. Autism symptom severity and autistic traits varied systematically with prediction efficiency, with autism-related communication difficulties showing the strongest associations. These findings enhance our understanding of language processing mechanisms in autism and suggest that interventions targeting language development might benefit from addressing prediction efficiency, such as providing additional processing time and gradually increasing the complexity of semantic integration tasks.