How Children with Autism Reason about Other’s Intentions: False-Belief and Counterfactual Inferences

This test was administered by the child’s psychologist and the scores made available to us for screening before the study was carried out.

This test is part of the verbal comprehension subscale and measures children’s knowledge of words and formation of concepts (Wechsler 2002). There are 36 items: 4 picture items and 32 verbal items.

We included a fourth question about the conjectural future, e.g., ‘What if next time Anne’s mother does not tell her to tidy her room, what will be the reason that Anne is picking up toys?’. However, such conjectural inferences about future intentional relations are often not deterministic, e.g., the reason Anne picks up toys today may be to find her ball, but tomorrow the reason may be to find her doll or to look at a teddy, or any number of different reasons and so we do not include responses to these questions in the analyses.

The analysis showed a main effect of group, F(1, 68) = 11.54, p < .001, np² = 0.15, age, F(2, 68) = 45.51, p < .001, np² = 0.58, and reasoning task, F (1, 68) = 33.59, p < .001, np² = 0.33. Group did not interact with age, F (2, 68) = 1.92, p < .155, or reasoning task, F (1,68) = 1.69, p < .198; however age and reasoning task interacted, F (1, 68) = 12.74, p < .001, np² = 0.27 and the interaction of the three variables was not significant, F (2, 68) = 2.69, p < .075, np² = 0.07. The decomposition of the non-significant three-way interaction, with a Bonferroni corrected alpha of p = .004, showed that at 6 years of age, children with autism made as few correct inferences as typically developing children for false belief inferences, t < 1, and counterfactual inferences t (68) = 1.25, p = .217; however at 8 years of age children with autism made fewer correct inferences than typically developing children for false belief inferences, t (68) = 4.54, p < .001, d = 1.31, and counterfactual inferences, although the latter was not significant on the corrected alpha, t (68) = 2.08, p < .04, d = 0.60; by 10 years of age, children with autism made as many correct inferences as typically developing children for false belief inferences, t (68) = 1.44, p = .15 and counterfactual inferences, t < 1. The comparisons also showed that typically developing children made more correct counterfactual inferences than false belief ones at 6 years of age, t (68) = 5.57, p < .001, d = 1.46, but this gap was closed by 8 years and 10 years, t < 1 in both cases. In contrast, children with autism made more correct counterfactual inferences than false belief ones at 6 years of age, t (68) = 4.50, p < .001, d = 1.29, this difference persisted at 8 years, t (68) = 3.61, p < .001, d = 1.04, and the gap was closed at 10 years t < 1.