Understanding Others’ Minds: Social Inference in Preschool Children with Autism Spectrum Disorder

One reviewer suggested that we include the IQ scores of the seven participants who were excluded from the actual study. We have now calculated the IQ scores of the seven participants (mean = 103.18, SD = 11.54, range 90–124), and we wish to note that the seven participants were comparable with those who participated in the study on IQ scores.

Proportion of fixations is a now standard measure in assessing both children’s and adults’ cognitive abilities like language. We wish to extend this measure to other domains of cognition like ToM. We wanted to show that proportion of fixations is also a reliable eye gaze measure for assessing children’s social cognition like ToM. This is another novelty of the present study, in addition to using a novel task.

This comparison of fixation proportions on the man versus the tree was a response to one reviewer’s concern. The reviewer asked whether the lack an effect in the ASD group as compared to the TD groups could be due to the ASD group’s reduced tendency to fixate on the man as compared to the tree. The comparison showed that no difference in fixation proportions was observed on the two agents for all the three groups (see also Fig. 6; Table 2), excluding this alternative explanation.

In an eye tracking study using the visual world paradigm, participants’ current fixation in a specific area is heavily dependent on their previous fixation, both on the temporal dimension and on the spatial dimension. In addition, proportion of fixations (based on fixation count) is categorical in nature and bounded by the values 0 and 1, which follows a multinomial distribution rather than a normal distribution. Therefore, traditional statistical methods based on the hypotheses of independent sampling and normal distribution such as t tests and ANOVAs cannot be directly applied. It has now become a standard procedure to analyse the fixation data using generalised linear mixed models (GLMMs) that do not require the assumption of normal distribution (Barr 2008; Jaeger 2008; Zhan 2018).

One reviewer suggested that we could also run t-tests on differential looking scores. We have followed the reviewer’s suggestion and analysed the differential looking scores using t tests. The t tests yielded similar statistical results and thus confirmed our observations using GLMMs. The t-tests results are provided in “Appendix A” section.

One reviewer pointed out that the recognition of this deficit in making social inferences by parents and teachers should have an effect on how they respond to an affected child with ASD in both classroom and social settings, and then suggested that feedback in a cognitive remediation paradigm might have beneficial effects on high-functioning children with ASD. The idea is to improve their abilities to pause and think if they have correctly utilized information that they know in making social inferences, and in making children aware of their deficit in the capacity to make social inferences, we might improve their abilities to socially interact and communicate. We thank the reviewer for this helpful suggestion in designing treatment plans and this is definitely a future direction that is worth exploring.