Task Structure Shapes How Autism Language Differences Show Up in Girls and Boys

When clinicians assess the spoken language of autistic children, the structure of the task they use may determine whether subtle social differences appear at all. A new study of 111 verbally fluent autistic and non-autistic youth, published in the Journal of Autism and Developmental Disorders, found that sex and diagnostic differences in social word use emerged clearly during a loosely structured picture description activity but vanished entirely during a more tightly structured storytelling task. The findings suggest that the choice of elicitation task is not a neutral methodological detail but an active factor shaping what researchers and clinicians can detect in autistic children’s pragmatic language.

The research team, led by Meredith Cola of the Children’s Hospital of Philadelphia Center for Autism Research, focused on social words: words that refer to other people, such as “mom,” “boyfriend,” or “them.” Social word production has previously been used as a computational index of social focus, and prior work has shown that such measures can differentiate autistic from non-autistic speakers in unstructured conversations and diagnostic interviews. What remained untested was how these markers behave across tasks with different levels of structure, and whether girls and boys respond differently to those structural demands.

To answer this, the researchers analyzed language produced during two tasks from the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2) Module 3, a widely used diagnostic assessment. In the Picture Description task, children were shown a cartoon map of the United States and asked to talk about it, with the clinician encouraged to build conversation from the child’s responses. This task offers relatively little verbal and visual structure. In the Telling a Story From a Book task, children narrated the wordless picture book “Tuesday” by David Wiesner, with the clinician modeling the story for the first two pages. The clear narrative arc of the book provides substantially more support.

The sample comprised verbally fluent youth aged 6 to 15, all with full-scale and verbal IQ estimates above 70. Autistic and non-autistic groups were matched on mean age and IQ, and autistic girls and boys were matched at the group level on age, IQ, and autistic traits measured by the ADOS-2 and the Social Communication Questionnaire. Autism diagnoses were confirmed by expert clinicians using a clinical best estimate approach informed by DSM-5 criteria, rather than relying solely on ADOS-2 or SCQ cutoff scores. Participant sex was characterized by parent-reported sex assigned at birth.

Recordings of each task were orthographically transcribed by annotators blind to diagnostic status and study hypotheses, then processed with LIWC-22, a computerized language analysis program that counts words in predefined categories. Because the non-autistic group produced significantly more total words than the autistic group in both tasks, the researchers standardized social word counts per 1,000 words to avoid confounding social focus with overall talkativeness. Generalized linear regression models then predicted social word production from age, full-scale IQ, sex, and diagnostic group, with sex-by-diagnosis interactions tested and retained only when statistically significant.

The results revealed a striking task-dependent pattern. During the Picture Description task, there was a significant interactive effect of sex and diagnosis, driven by non-autistic girls, who produced more social words than any other subgroup. Non-autistic girls outproduced both autistic girls and autistic boys, and non-autistic boys also produced more social words than both autistic subgroups. Notably, autistic girls and autistic boys did not differ significantly from each other, and neither did non-autistic girls and boys. In the Telling a Story From a Book task, by contrast, there were no significant effects of sex or diagnosis on social word production at all.

The study also examined whether social word use related to clinician-rated social difficulties within the autistic group. Social word production during the Picture Description task significantly predicted ADOS-2 Social Affect scores: children who used fewer social words during this less structured task were rated by expert clinicians as having greater social difficulties. When the analysis was run separately by sex, this association held for autistic girls but not for autistic boys, a result the authors flag as exploratory and in need of replication. Social word production during the storytelling task did not significantly predict Social Affect scores for the group overall or for either sex.

The authors interpret these findings as evidence that narrative structure supports autistic children’s social language. When a wordless storybook provides a clear storyline and explicit directions, both autistic and non-autistic youth attend to and discuss the socially salient elements of the narrative, narrowing group differences. When the visual stimulus lacks a narrative arc, as with the map picture, children must decide for themselves what is worth talking about, and differences in social focus become detectable. The researchers suggest this responsiveness to structure may relate to a preference for predictability, and they note that the map illustration contains more images of people than the storybook, a stimulus difference that may also have contributed.

The practical implications extend beyond the clinic. The authors suggest that autistic youth may find greater social success in structured, activity-based settings where conversation topics are predictable, such as a stargazing group or birding club, compared with open-ended social situations. Clinically, the results indicate that clinicians administering the ADOS-2 should pay particular attention to social language during less structured activities like Picture Description when forming gestalt ratings, and should consider evaluating a child’s social language relative to same-sex peers, since baseline sex differences exist in non-autistic youth as well. Knowing which tasks do not elicit group differences, such as the storytelling task, may help clinicians focus on other behavioral observations during those activities.

The findings come with important caveats. The sample was exclusively verbally fluent youth aged 6 to 15, predominantly White and non-Hispanic, with most mothers educated beyond high school, limiting generalizability across age, race, ethnicity, and socioeconomic status. The frequency-based method captures how often social words appear but not their context or appropriateness, and all language was produced during a diagnostic assessment with an adult clinician rather than with peers. Task order was not randomized, sex was measured as sex assigned at birth rather than gender, and the per-1,000-words normalization may affect participants with very high or very low total word counts. Even so, the study makes a clear methodological point: the structure of an elicitation task can enhance or suppress the detection of diagnostic and sex-related differences in social language, and researchers and clinicians who ignore that context risk drawing conclusions that reflect the task as much as the child.

Subject of Research: Sex and diagnostic differences in social word use among verbally fluent autistic children across ADOS-2 tasks with differing levels of structure

Article Title: Context Matters: Assessing Sex and Diagnostic Differences in Autistic Children’s Language Across Tasks With Differing Levels of Structure

Article References: Cola, M., Lyons, M., Schillinger, S., Lee, A., Faulk, C., Fulop, L., Knox, A., Howard, N., Franke, H., Mulqueen, A., Cho, S., Liberman, M. Y., Flanagan, A. Y., Pandey, J., Schultz, R. T., Armstrong, S. L., & Parish-Morris, J. (2026). Context Matters: Assessing Sex and Diagnostic Differences in Autistic Children’s Language Across Tasks With Differing Levels of Structure. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-026-07562-3

Image Credits: AI Generated

DOI: 10.1007/s10803-026-07562-3

Keywords: autism, pragmatic language, social words, ADOS-2, task structure, autistic girls, language assessment, LIWC-22, children, social communication, narrative tasks, sex differences