Family Matters: Children and Adolescents with ASD Talk More about Family Than Friends
Objectives: Measure how often children with ASD, typically developing controls (TDC) and interviewers use “friend” and “family” words during the conversation and reporting section of the Autism Diagnostic Observation Schedule (ADOS), and correlate rates of “friend” and “family” word usage with social functioning (ADOS calibrated severity scores and scores on the Social Responsiveness Scale).
Methods: Research reliable PhD-level clinicians administered the ADOS to 65 participants (ASD: 48, TDC: 17). Participants were matched on full-scale IQ and sex ratio, with a slightly younger ASD group (Table 1). Trained annotators transcribed the words produced by clinicians and participants during the conversation and reporting section of the ADOS (~20 minutes). We processed each transcript using the Linguistic Inquiry and Word Count software (Tausczik & Pennebaker, 2010), then extracted the ratio of words categorized as “friend” words and “family” words for each speaker, relative to the total number of words produced. There were no significant correlations between age and word use between- or within-groups, so age was not included as a covariate in our primary analyses. Due to significant correlations between interviewer and participant word use (.3-.4), we controlled for interlocutor language in our omnibus tests.
Results: We conducted an ANCOVA for each speaker type, with word category (friend, family) as a repeated measure and diagnosis as a between-group factor. There was a trending interaction between participant word category use and diagnosis, F(1,61)=2.77, p=.10, and no significant effects of interviewer language. Paired t-tests showed that participants in the ASD group produced a significantly higher percentage of family words than friend words, p=.01, while TDCs used family and friend words at equal rates (Figure 1). Interviewers in both groups used significantly more friend than family words, ps<.01. After partialing out interviewer language, we found a significant positive correlation between the rate of “family” words used by all participants and their SRS t-scores, and a significant negative correlation between ADOS calibrated severity scores and “friend” words.
Conclusions: Natural language processing holds great promise as research tool. It has the potential to provide highly granular, quantitative data that varies dimensionally in the majority of individuals with ASD. In this study, we explored one metric of social functioning that emerged in the language produced by in children with ASD and non-ASD peers. The difference between “friend” and “family” words held even after controlling for interlocutor language, suggesting that participants with ASD may engage in less linguistic accommodation than peers. Future research with a larger cohort will explore age effects in older adolescents.