Attenuated Attention to People but Not Humanoid Puppet Figures in Children with Autism Spectrum Disorder

Background: Previous eye-tracking studies have found attenuated attention to faces in young children with ASD, relative to typically-developing (TD) and developmentally delayed (DD) groups (Klin et al., 2009; Chawarska et al., 2012; Pierce et al., 2018). While social attention varies by activity type (Chawarska et al. 2012) and social load (1 vs. multiple humans) (Moore et al., 2018; Speer et al., 2007), less is known about how simpler, humanoid figures (e.g., puppets) may influence viewing patterns in ASD. Research using artificial intelligence in interventions has found that simplified, predictable humanoid figures may facilitate learning in ASD (Scassellati et al., 2018). A direct examination of perception of humanoid vs veridical human social figures in ASD can critically inform such intervention practices.

Objectives: We investigated attention to the faces of a puppet and person engaged in a conversation and its association with clinical features. If puppet and person are perceived as equivalent social agents, then children with ASD may exhibit attenuated looking to both figures compared to TD and DD groups. If the puppet is less socially ostensive than the person, then between-group differences may be present for the person but not puppet.

Methods: 81 children with ASD (N = 29, mean age=46.92 months), TD (N = 27, mean age=39.83 months) and DD (N = 25, mean age= 51.29 months) participated in a free-viewing eye-tracking paradigm consisting of a video depicting a puppet and an actress engaged in conversation (Figure 1). Percent of time spent viewing the social scene (%valid), percent of valid viewing time looking at the puppet or person (%Face), and a puppet-person ratio (PP_ratio) (e.g., relative time spent viewing puppet vs person) were calculated. Participants were administered the ADOS-2 and a cognitive test (e.g., Mullen or DAS).

Results: Children in the ASD group had lower %Valid than TD or DD groups (F2,77=8.26,p=.001, pη2=.17; ASDvsTD: p<.001; ASDvsDD: p=.004). %Face-Puppet did not vary by diagnosis (F2,77=0.44, p=.65, pη2=.01), but %Face-Person was attenuated in ASD (F2,77=3.84, p=.026, pη2=.09; ASDvsTD:p=.01; ASDvsDD: p=.05; Figure 2). Using the puppet-person ratio, greater relative attention to puppet vs person in the ASD group was associated with greater social impairment (r=.422, p=.032). Across all participants, greater attention to the person was associated with lower ASD symptom severity (r=-.374,p=.001), and better verbal (r=.409,p<.001) and nonverbal cognitive skills (r=.376,p<.001). Attention to the puppet was not significantly associated with any clinical features.

Conclusions: Consistent with prior work, children with ASD exhibited attenuated attention to human-speaker face, but their attention to the puppet-speaker face was comparable to DD and TD groups suggesting that simpler, perceptually salient humanoid form may serve as an enhancer of social attention in children with ASD. This was particularly true of children with greater severity of autism symptoms who spent proportionally more time monitoring the puppet speaker’s face than the human speaker’s face. This suggests that puppet-like agents may provide a particularly effective way of engaging attention of children with more severe symptoms, though whether this enhancement facilitates social learning remains to be determined.