Defining Language Profiles in Young Autistic Children Across Verbal, Speaking, and Generative Dimensions

Language trajectories in autism are highly heterogeneous and cannot be captured by unidimensional classifications. Research and clinical practice often rely on verbal/non-verbal or speaking/non-speaking labels, but these categories, when used in isolation, obscure distinct linguistic profiles. Spoken language is also frequently used as a proxy for structural language skills, despite evidence that speech may not be generative or productive. A multidimensional perspective is therefore needed to determine whether verbal skills, spoken output, and linguistic generativity reflect dissociable aspects of language in autism, and whether variation across these dimensions can be explained by non-verbal cognition or autism characteristics. Participants were 167 autistic children aged 2-6 years from the Dutch- and French-speaking regions of Belgium. Verbal skills were assessed using standardized tests. Speaking status was quantified as the proportion of linguistic segments in spontaneous productions recorded at home. Generativity was measured using a similarity index derived from semantic embeddings. Associations with non-verbal Intelligence Quotient (IQ) and autism characteristics were examined using age-controlled linear models. Overlap between the three dimensions, Verbal, Speaking, and Generativity, was only partial. Some children with little or no speech showed relatively strong structural language abilities, whereas others produced abundant but non-generative speech and had limited structural skills. Both non-verbal IQ and autism characteristics were related to all three dimensions, with IQ explaining a larger proportion of variance. Overall, one-dimensional labels fail to capture language variability in autism; assessing Verbal, Speaking, and Generativity as complementary dimensions may better identify uneven linguistic profiles in autistic children.