Childhood Theory of Mind and Cognitive Flexibility and Early Developmental Change in Planning Skills Predict Later Behavioural Outcomes in Autistic Adolescents: A 12-Year Prospective Study.

Background: Longitudinal studies of autistic people show that the behavioural features of autism generally endure into adulthood. It remains unclear, however, whether individual differences in early cognitive skills are longitudinally related to individual differences in specific behavioural outcomes.

Objectives: Here, we test the predictive utility of measures of childhood theory of mind (ToM) and executive function (EF) as well as the role of early developmental change in these cognitive constructs on emerging autistic symptomatology and adaptive behaviour in a cohort of autistic youth over a 12-year period.

Methods: Twenty-six cognitively able young autistic people (2 female) took part in a prospective longitudinal study. Participants were assessed on two key components of EF (planning and set shifting) and on a battery of ToM tasks (1st- and 2nd-order false belief) at Time 1 (M=5 years; 7 months, SD=11 months) and again at Time 2, 3 years later (M= 8 years; 4 months, SD= 12 months). We also measured participants’ core autistic features (as indexed by the ADOS-2) and adaptive behaviour (as indexed by the Vineland-II) at Time 3, 9 years later (M=17 years; 10 months, SD= 14 months). Regression analyses were conducted to test whether early levels of cognitive skills (Time 1 ToM and EF) and the rates of developmental change in ToM and EF (between Time 1 and Time 2) could explain unique variance in (1) autism symptomatology and (2) adaptive behaviour in late adolescence.

Results: With regard to predicting autistic symptomatology, individual differences in Time 1 age, verbal and non-verbal ability failed to significantly predict variance in adolescents’ ADOS-2 scores, F(3,18)=.02, p=.99, R²=.003. The addition of Time 1 set-shifting ability and ToM significantly improved the fit of the model, F(5,16)=5.26, p=.004, R²=.50. Finally, developmental change in planning ability between Time 1 and 2 explained a further 18% of unique variance in core autistic features, F(6,15)=5.26, p=.004, R²=.68. With regard to adaptive behaviour, Time 1 age, verbal and non-verbal ability failed to significantly predict variance in adolescents’ Vineland-II scores, F(3,18)=1.96, p=.16, R²=.25. When we entered the early cognitive variables into the model, only Time 1 set-shifting ability significantly improved model fit, explaining an additional 20% of the variance in adolescents’ adaptive behaviour, F(4,17)=3.43, p=.03, R²=.45.

Conclusions: We show for the first time that both ToM and EF measured in childhood – and critically, early developmental changes in EF – predict specific aspects of autistic adolescents’ behavioural outcomes 12 years later. These findings suggest that early-emerging cognitive atypicalities could cause behavioural disruptions that persist into early adulthood, possibly even persisting beyond the cognitive atypicalities themselves. The predictive power of these cognitive variables both adds weight to the veracity of these models as explanatory tools in autism research and underscores the importance of childhood cognition as a candidate target for early intervention.