Diminished Salience of Speech Underlies Diminished Face Looking in Toddlers with ASD: An Eye-Tracking Study
Background: Under-responsivity to eye contact is considered a hallmark of autism spectrum disorders (ASD). However, real-world face-to-face social interactions often involve both eye contact and speech together. It is not clear whether children with ASD show limited salience for these features, or whether they find them aversive. Similarly, it is not clear how gaze and speech cues independently and synergistically impact looking at the face, eyes, and mouth, especially early in development, in children with ASD. Finally, it is unclear how modulation of gaze patterns in response to gaze and eye cues relate to clinical phenotype in ASD.
Objectives: To systematically examine the independent and emergent effects of direct eye contact and speech on face-looking profiles in toddlers with ASD using eye tracking.
Methods: Gaze patterns of 22-month-old toddlers with ASD (n=53) were compared to typically-developing toddlers (TD, n=47) on the Selective Social Attention 2.0 Task. This task manipulated direct gaze presence (DG+) or absence (DG-), and speech presence (SP+) or absence (SP-) in 4 randomly interleaved conditions sampled with 8 trials each. The task lasted 6 minutes total. Linear mixed effect models were used to examine group and condition effects. Eye tracking associations with clinical variables were assessed using Pearson’s correlation coefficients.
Results: Both groups looked more at faces when DG and SP cues were present, but the increase was less pronounced in ASD (Δ% in DG+SP+) – (%Face in DG-SP-), TD: d=2.32, ASD: d=.73). Toddlers with ASD differed from TD toddlers in face looking only when the actress was speaking (DG-SP+: d=1.03, p<.001; DG+SP+: d=1.65, p< .001). Proportion of time spent looking at the actress's eyes did not differ between groups in any condition, but toddlers with ASD looked less at the mouth relative to TD toddlers when the actress was speaking (DG-SP+: d=.94, p< .001; DG+SP+: d=.81, p< .001). Lower responsivity to speech and gaze cues (Δ%Face) was concurrently and prospectively (40 months) associated with greater severity of autism symptoms (r=-.62, p<.001 concurrently; r=-.49, p=<.001 prospectively) and lower verbal (r=.59, p<.001; r=.44, p<.001) and nonverbal (r=.-54, p<.001; r=.31, p=.01) developmental ability. ROC analyses using Δ%Face yielded Sensitivity=88.4%, Specificity=80.0%, PPV=80.9%, NPV=87.8%, AUC=.89.
Conclusions: No evidence for face avoidance was observed. ASD-TD face looking differences were greatest when speech cues were present, suggesting limited salience for speech. Similarly, distribution of attention between eyes and mouth was perturbed in ASD only when speech was present and was indexed by lower attention to speaker’s mouth. Responsivity to speech and gaze cues was concurrently and prospectively associated with developmental phenotype and also exhibited good group-discriminative ability, highlighting its potential as a marker for ASD. This study links poor attention to the faces of interactive partners in early stages of autism with the limited salience of speech rather than direct gaze. This work (1) suggests that early intervention in ASD may benefit from increased focus on orienting to audiovisual speech cues, and (2) may provide further motivation for research into the role of speech processing in the emergence of social disability in infancy.