Deconstructing the Diet-Language Myth in Autism: A Divergent V-Type Model

To investigate whether eating behaviors and diet quality predict receptive and expressive language development in children with autism spectrum disorder (ASD), independent of autism severity.

Seventy-one children with ASD (mean age 55.61 ± 15.88 months) were assessed using the TEIıL (language), GARS-2-TV (autism severity), CEBQ (eating behaviors), and KIDMED (diet quality). Analyses included hierarchical regression, 5000-bootstrap mediation, moderation, and exploratory K-means cluster analysis.

Autism severity strongly predicted both receptive and expressive language (explained variance = 53%). After controlling for severity and age, eating behaviors and diet quality provided no significant incremental variance ( $ΔR² = .020 and .012) and showed no significant partial correlations with language (all p > .15). Furthermore, all 11 mediation and 6 moderation models yielded non-significant effects. Exploratory clustering revealed a paradox: children with the most severe autism and highest food selectivity exhibited the highest Mediterranean diet adherence, likely reflecting rigid parental dietary control.

Nutritional variables do not directly or indirectly predict language development in ASD when controlling for core symptom severity. We propose a divergent "V-Type" conceptual model, suggesting that autism severity independently drives both language delays and eating disruptions. Nutritional interventions and speech-language therapy should be pursued as parallel, independent targets.