Indexing Social Focus in the Natural Conversations of Individuals with and without ASD

Background: Language is a window into people’s inner worlds. Analyzing words produced during the course of a natural conversation can shed light on a person’s thoughts, feelings, personality traits, and interests (Tausczik & Pennebaker, 2010). In this study, we tested the extent to which social focus is evident in language, by exploring personal pronoun use in individuals with social communication challenges (autism spectrum disorder; ASD) and matched individuals that are typically developing (TD). We hypothesized that individuals with less social focus would use fewer personal pronouns during natural conversation with a naïve confederate, and individuals with more social focus would use more personal pronouns. Our natural conversation task was designed to be an ecologically valid analogue for everyday experiences of meeting new people and establishing relationships; an essential challenge for individuals with ASD and a frequent target for intervention.

Objectives: Use natural language to index social focus: Compare rates of all pronouns, personal pronouns, “I”, “you”, and “we” produced by individuals with ASD and typically developing control participants. Determine whether linguistic markers of social focus are validated by expert clinical ratings of social communication impairment: Correlate rates of personal pronoun use with autism symptoms in the ASD group.

Methods: Fifty-six participants aged 5-36 with ASD (N=38, 12 female) or typical development (N=18, 12 female) engaged in a ~5 minute unstructured “get to know you” conversation with one of 13 randomly assigned undergraduate confederates (10 female). Participants in the ASD and TD groups did not differ significantly on age (mean=14y) or IQ (106, all IQs>75), but did differ on sex ratio (p=.02). This limitation (uneven sex ratios by group) will be addressed before May, 2018 through increased sample sizes. Linear mixed effects regression models included diagnosis as a fixed effect and confederate ID as a random effect to account for individual differences among confederates.

Results: The ASD and TD groups did not differ on overall pronoun use (including words like “it” and “that”), but the ASD group produced significantly lower rates of personal pronouns than the TD group (t=2.29, p=.03, Fig.1). The ASD group produced less “we” than the TD group (t=2.11, p=.04), but groups did not differ significantly on rates of “I” or “you”. In participants with ASD and ADOS evaluations (N=37), personal pronoun use correlated significantly with ADOS social affect calibrated severity scores (r=-.36, p=.03, Fig.2), but not with repetitive behaviors/restricted interests subscores (r=-.15, p=.38).

Conclusions: We found that short, naturalistic conversations with a naïve interlocutor provide an index of social focus that correlates with expert clinical ratings of social impairment. Specifically, we found that personal pronoun use is diminished in individuals with ASD relative to typical controls, and individuals with more autism symptoms produce fewer personal pronouns than individuals with fewer autism symptoms. This result holds promise as a potential metric for tracking social intervention effectiveness, via natural conversations that reveal the extent of an individual’s social focus from week-to-week or even day-to-day.