The Selective Social Attention Task in Children with ASD: Results from the Autism Biomarkers Consortium for Clinical Trials (ABC-CT) Feasibility Study
(Full-text capture 2026-09-21; web artifacts lightly stripped; truncated.)
Frederick Shic
1.Center for Child Health, Behavior and Development, Seattle Children’s Research Institute, Seattle, Washington, USA
2.Department of General Pediatrics, University of Washington School of Medicine, Seattle, Washington, USA
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
1,2,3, Erin C Barney
Erin C Barney
1.Center for Child Health, Behavior and Development, Seattle Children’s Research Institute, Seattle, Washington, USA
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
1,3, Adam J Naples
Adam J Naples
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
Kelsey J Dommer
1.Center for Child Health, Behavior and Development, Seattle Children’s Research Institute, Seattle, Washington, USA
Shou An Chang
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
3, Beibin Li
Beibin Li
1.Center for Child Health, Behavior and Development, Seattle Children’s Research Institute, Seattle, Washington, USA
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
4.Paul G. Allen School of Computer Science & Engineering, University of Washington, Seattle, Washington, USA
1,3,4, Takumi McAllister
Takumi McAllister
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
3, Adham Atyabi
Adham Atyabi
2.Department of General Pediatrics, University of Washington School of Medicine, Seattle, Washington, USA
5.Department of Computer Science, University of Colorado - Colorado Springs, Colorado Springs, Colorado, USA
2,5, Quan Wang
Quan Wang
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
6.Xi’an Institute of Optics and Precision Mechanics, Chinese Academy of Sciences, Xi’an, China
3,6, Raphael Bernier
Raphael Bernier
7.Department of Psychiatry & Behavioral Science, University of Washington School of Medicine, Seattle, Washington, USA
Geraldine Dawson
8.Duke Center for Autism and Brain Development, Duke University, Durham, North Carolina, USA
8, James Dziura
James Dziura
9.Emergency Medicine, Yale University, New Haven, Connecticut, USA
9, Susan Faja
Susan Faja
10.Department of Pediatrics, Harvard Medical School, Boston, Massachusetts, USA
11.Department of Pediatrics, Boston Children’s Hospital, Boston, Massachusetts, USA
10,11, Shafali Spurling Jeste
Shafali Spurling Jeste
12.Department of Psychiatry & Biobehavioral Sciences, University of California Los Angeles, Los Angeles, California, USA
13.Department of Neurology, Keck School of Medicine of USC, Los Angeles, California, USA
14.Division of Neurology, Children’s Hospital Los Angeles, Los Angeles, California, USA
12,13,14, Michael Murias
Michael Murias
15.Department of Medical Social Sciences, Northwestern University, Evanston, Illinois, USA
15, Scott P Johnson
Scott P Johnson
16.Department of Psychology, University of California Los Angeles, Los Angeles, California, USA
Maura Sabatos-DeVito
8.Duke Center for Autism and Brain Development, Duke University, Durham, North Carolina, USA
Gerhard Helleman
12.Department of Psychiatry & Biobehavioral Sciences, University of California Los Angeles, Los Angeles, California, USA
17.Department of Public Health, University of Alabama at Birmingham, Birmingham, Alabama, USA
12,17, Damla Senturk
Damla Senturk
18.Department of Biostatistics, University of California Los Angeles, Los Angeles, California, USA
Catherine A Sugar
18.Department of Biostatistics, University of California Los Angeles, Los Angeles, California, USA
18, Sara Jane Webb
Sara Jane Webb
2.Department of General Pediatrics, University of Washington School of Medicine, Seattle, Washington, USA
7.Department of Psychiatry & Behavioral Science, University of Washington School of Medicine, Seattle, Washington, USA
2,7, James C McPartland
James C McPartland
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
Katarzyna Chawarska
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
3; The Autism Biomarkers Consortium for Clinical Trials
- Author information
- Article notes
- Copyright and License information
1.Center for Child Health, Behavior and Development, Seattle Children’s Research Institute, Seattle, Washington, USA
2.Department of General Pediatrics, University of Washington School of Medicine, Seattle, Washington, USA
3.Yale Child Study Center, Yale University School of Medicine, New Haven, Connecticut, USA
4.Paul G. Allen School of Computer Science & Engineering, University of Washington, Seattle, Washington, USA
5.Department of Computer Science, University of Colorado - Colorado Springs, Colorado Springs, Colorado, USA
6.Xi’an Institute of Optics and Precision Mechanics, Chinese Academy of Sciences, Xi’an, China
7.Department of Psychiatry & Behavioral Science, University of Washington School of Medicine, Seattle, Washington, USA
8.Duke Center for Autism and Brain Development, Duke University, Durham, North Carolina, USA
9.Emergency Medicine, Yale University, New Haven, Connecticut, USA
10.Department of Pediatrics, Harvard Medical School, Boston, Massachusetts, USA
11.Department of Pediatrics, Boston Children’s Hospital, Boston, Massachusetts, USA
12.Department of Psychiatry & Biobehavioral Sciences, University of California Los Angeles, Los Angeles, California, USA
13.Department of Neurology, Keck School of Medicine of USC, Los Angeles, California, USA
14.Division of Neurology, Children’s Hospital Los Angeles, Los Angeles, California, USA
15.Department of Medical Social Sciences, Northwestern University, Evanston, Illinois, USA
16.Department of Psychology, University of California Los Angeles, Los Angeles, California, USA
17.Department of Public Health, University of Alabama at Birmingham, Birmingham, Alabama, USA
18.Department of Biostatistics, University of California Los Angeles, Los Angeles, California, USA
✉
Correspondence concerning this article should be addressed to: Katarzyna Chawarska, Yale School of Medicine, Child Study Center, 300 George St., New Haven, CT 06510, katarzyna.chawarska@yale.edu
Issue date 2023 Nov.
PMCID: PMC11003770 NIHMSID: NIHMS1979568 PMID: 37749934
The publisher’s version of this article is available at Autism Res
Abstract
The Selective Social Attention (SSA) task is a brief eye-tracking task involving experimental conditions varying along socio-communicative axes. Traditionally the SSA has been used to probe socially-specific attentional patterns in infants and toddlers who develop autism spectrum disorder (ASD). This current work extends these findings to preschool and school-age children.
Children four-to-twelve-years-old with ASD (N=23) and a typically-developing comparison group (TD; N=25) completed the SSA task as well as standardized clinical assessments. Linear mixed models examined group and condition effects on two outcome variables: percent of time spent looking at the scene relative to scene presentation time (%Valid), and percent of time looking at the face relative to time spent looking at the scene (%Face). Age and IQ were included as covariates. Outcome variables’ relationships to clinical data were assessed via correlation analysis.
The ASD group, compared to the TD group, looked less at the scene and focused less on the actress’ face during the most socially-engaging experimental conditions. Additionally, within the ASD group, %Face negatively correlated with SRS Total T-scores with a particularly strong negative correlation with the Autistic Mannerism subscale T-score.
These results highlight the extensibility of the SSA to older children with ASD, including replication of between-group differences previously seen in infants and toddlers, as well as its ability to capture meaningful clinical variation within the autism spectrum across a wide developmental span inclusive of preschool and school-aged children. The properties suggest that the SSA may have broad potential as a biomarker for ASD.
Keywords: Autism Spectrum Disorder, Child, Eye-Tracking Technology, Social Attention, Biomarkers
Lay Summary
Previous work found that an infant’s or toddler’s performance on a simple eye-tracking task was different depending on if they had a diagnosis of ASD or not. This paper shows that the same differences exist in 4-12-year-olds and shows that performance on this task is different for those with higher ratings of autistic traits. This is an important step in being able to use a quick, easy technology like eye tracking to help clinicians identify risk or track possible changes in behavior.
Introduction
One of the core features of ASD in infancy and early childhood is limited attention to faces and gaze of social partners ( American Psychiatric Association, 2013; Luyster et al., 2009). Children with ASD show reduced looking to faces during the prodromal ( Chawarska et al., 2013; Shic et al., 2014) and early syndromal stages of the condition ( Chawarska et al., 2012, 2015; Nakano et al., 2010; Shic et al., 2011; Vivanti et al., 2017) in comparison to typically developing (TD) and developmentally delayed (DD) children. This decreased attention is context-dependent and most pronounced when children observe a person trying to engage their attention using eye contact and speech, but not when a person engages in a solitary activity where neither eye contact nor speech are present ( Chawarska et al., 2012; Shic et al., 2020), nor when perceptually salient dynamic distractors are included ( Chawarska et al., 2012).
The Selective Social Attention (SSA) version 1.0 task ( Chawarska et al., 2012) – a brief eye-tracking assay that includes experimental probes that use child-directed speech and eye contact as a bid for infant attention in the face of non-social distractors – has been found to discriminate infants, toddlers, and preschool children with ASD from TD and DD controls ( Chawarska et al., 2012, 2013; Wall et al., 2022). Further, the SSA 1.0 task has been found to be associated with a range of clinical features associated with ASD ( Murias et al., 2018) and to subtype toddlers with ASD into clinically-meaningful subgroups that differ both concurrently and prospectively in clinical presentation ( Campbell et al., 2014).
It is not clear, however, whether SSA 1.0 results evident from infancy through early preschool would also be observed at school-ages, given its focus on early developmental social-attentional constructs of dyadic engagement and spontaneous social attention ( Chawarska et al., 2015; Dawson et al., 2004). Many children with ASD show significant developmental gains in adaptive, behavioral, cognitive, social skills between early and middle childhood ( Anderson et al., 2009; Clark et al., 2017; Flanagan et al., 2015; Waizbard-Bartov et al., 2022), and thus the persistence of early attentional signatures of ASD in unaltered form through school age cannot necessarily be presumed ( Lord & Jones, 2012; Mundy & Bullen, 2022). This may be particularly true for children who receive a diagnosis of ASD after the age of five years, many of whom may not have received a diagnosis of ASD at the preschool age ( Ozonoff et al., 2018). As noted by Mundy & Bullen (2022), social attention comprises at least two distinct phenomena: (1) social orienting, the allocation of attention to faces, face features, and other social sources of information; and (2) joint attention, the response to and direction of others’ gaze and attentional cues in a coordinated interplay of dyadic, triadic, and referential information processing. While the SSA 1.0 includes embedded experimental conditions of both emulated dyadic engagement as well as joint attention, its primary dependent measure is the proportion of time spent looking at the face, a target most closely aligned with dyadic joint attention – one of the simpler foundational elements of joint attention as a whole, and the only component of joint attention also considered an expression of social orienting. While prior studies have shown diminished looking at faces in children with ASD at school age or early adolescence, even with eye tracking ( Chevallier et al., 2015; Riby & Hancock, 2009; Rice, 2007), these stimuli are typically depicted as passive viewing tasks, as opposed to as emulated infant-directed dyadic interactions, and often include more dynamic activities, backgrounds, and theatrical components. As stimulus complexity has long known to impact eye tracking group differences between children with ASD and control groups, with simpler imagery such as presentation of single faces yielding muted differences ( Speer et al., 2007), express validation of the SSA 1.0 for this older population is still needed.
Further, because the SSA 1.0 was developed with a focus on an earlier age range, with an actress who spoke directly to the camera using highly prosodic, infant-directed speech, it was unclear whether older children with ASD would tolerate the assay for several reasons. First, older children with ASD might find the stimuli exceptionally irrelevant and uninteresting, exacerbating decreased motivation for social information or social reward evident in some individuals with ASD ( Chevallier et al., 2012; Chita-Tegmark, 2016; Clements et al., 2018). This could lead to decreased viewing of this social scene in excess of that observed in much younger toddlers with ASD ( Chawarska et al., 2012). Second, the stimulus could be perceived by some individuals as infantilizing and offensive and, given high prevalence rates of co-occurring psychiatric conditions (including disruptive behaviors) in 6-to-12-year-old children with ASD ( de Bruin et al., 2007; Gadow et al., 2005; Kaat & Lecavalier, 2013), could lead to non-compliance or off-task behaviors impacting data acquisition. Third, and relatedly, some children with ASD exhibit social-evaluative fears, especially when they possess higher IQ or intact Theory of Mind skills ( Hunsche et al., 2022; Pugliese et al., 2013). Concerns regarding viewing of video content perceived as developmentally juvenile relative their current age, especially in the presence of strangers, could lead to challenging or avoidant behaviors leading to lost data.
This study specifically examines whether the SSA 1.0 can be successfully deployed in school-age children with and without ASD, with a goal of obtaining preliminary evidence regarding the potential of the SSA 1.0 to serve as an eye-tracking biomarker of ASD. A biomarker by definition is “A defined characteristic that is measured as an indicator of normal biological processes, pathogenic processes, or biological responses to an exposure or intervention…” ( FDA-NIH Biomarker Working Group, 2021). Eye tracking enables objective, quantifiable, dense, and repeatable measurement of oculomotor indices as the overt expression of neurobiological substates driving attentional processes ( Posner & Petersen, 1990). There is a growing body of work examining the potential of eye tracking as a biomarker in conditions ranging from traumatic brain injury ( Hunfalvay et al., 2021; Samadani et al., 2016, 2017; Bin Zahid et al., 2020; for review see McDonald et al., 2022), to Parkinson’s disease ( Ba et al., 2022; Blekher et al., 2009; Brien et al., 2023; Li et al., 2023; Tsitsi et al., 2021), to autism ( Bradshaw et al., 2019; Del Valle Rubido et al., 2020; Frazier et al., 2021; Loth et al., 2017; Mason et al., 2021; Murias et al., 2018; Pierce et al., 2015; Shic et al., 2022; Wen et al., 2022), though discussions regarding nomenclature are ongoing, as they are for other similar classes of biomarkers ( Ba et al., 2022; Hidalgo-Mazzei et al., 2018; Li et al., 2023; Tsitsi et al., 2021).
Extending our understanding of the SSA 1.0 to older children would thus allow us to (1) better characterize the stability of its indexed mechanistic constructs (in support or refutation of the developmental persistence of diminished selective social attention in ASD); (2) evaluate its feasibility for use with older children, potentially establishing it as a practical social-attentional biomarker with broad acceptability across early childhood – an important step in its consideration for use in longitudinal studies of early development; and (3) characterize its relationships with clinical and behavioral phenotypes, providing evidence regarding its potential for revealing clinically-relevant insights at the level of an individual.
The results presented here are based on the feasibility phase of the Autism Biomarkers Consortium for Clinical Trials ( McPartland et al., 2020), which included the SSA 1.0 task among a battery of candidate biomarkers administered to pre-school and school-age children diagnosed with ASD and in TD controls. This paper examines: (1) feasibility of SSA 1.0 task in a well-characterized sample of school-age children with ASD and TD controls; (2) group and condition effects on attention to faces; and (3) associations between performance on the SSA 1.0 task and clinical phenotypes in school-age children with ASD and typically developing (TD) controls1.
Methods
Participants:
Participants included 25 (18 male) children with ASD and 26 (16 male) TD controls. The participants were recruited at five sites of the ABC-CT study, and enrolled after consideration of exclusions including the presence of serious medical or neurological conditions that would interfere with eye-tracking data interpretation. Informed consent/assent was obtained from all families after procedures were fully explained and any questions answered. The protocol was overseen by Yale University IRB.
ASD/TD group membership was based on a standardized process harmonized across sites (for details see McPartland et al., 2020; Shic et al., 2022; Webb et al., 2020, 2022). Briefly, group membership was confirmed by trained and licensed clinicians based on DSM-5 criteria ( American Psychiatric Association, 2013) as guided by gold-standard methods including the clinician-administered Autism Diagnostic Interview – Revised (ADI-R; Rutter et al., 2003) and the Autism Diagnostic Observation Schedule, 2nd Edition (ADOS-2; Lord et al., 2012) (generating total autism symptom calibrated severity score as well as social affect (SA) and restricted interest/repetitive behaviors (RRB) comparison scores). Additional clinical information was obtained using the clinician-administered Differential Abilities Scale-II (DAS-II; Elliott, 2007a) (generating Full Scale, Verbal, and Nonverbal IQ scores) and the parent-report Social Responsiveness Scale, 2nd edition (SRS-2; Constantino & Gruber, 2012) (providing standardized T scores reflecting severity of autism symptoms across multiple dimensions).
There were no differences between groups in sex ratio, though TD controls were recruited into the study at younger ages. The groups differed in a predictable manner with regard to IQ, ADOS-2, and SRS-2 scores (see Table 1 for sample characteristics).
Table 1.
Participant Characterization
| ASD | TD | p | |||
|---|---|---|---|---|---|
| N Participants with Analyzable Data | 25 | 26 | |||
| N Participants with Valid Data | 23 | 25 | .972 | ||
| N Male Participants with Valid Data | 18 | 16 | .442 | ||
| M (SD) | [min,max] | M (SD) | [min,max] | ||
| * * * | |||||
| Age at Enrollment (years) | 8.1 (2.2) | [5,11] | 6.6 (2.0) | [4,11] | .019 |
| * * * | |||||
| DAS-II Full Scale IQ | 94.0 (17.3) | [54,123] | 114.0 (9.5) | [96,133] | <.001 |
| DAS-II Verbal IQ | 92.6 (19.2) | [51,137] | 115.4 (14.1) | [88,148] | <.001 |
| DAS-II Nonverbal IQ | 95.6 (17.1) | [47,124] | 110.8 (8.1) | [98,124] | .001 |
| * * * | |||||
| ADOS-2 Calibrated Severity Score | 7.7 (1.6) | [4,10] | 1.2 (0.4) | [1,2] | <.001 |
| ADOS-2 SA Comparison Score | 7.7 (2.1) | [2,10] | 1.4 (0.7) | [1,3] | <.001 |
| ADOS-2 RRB Comparison Score | 7.2 (2.5) | [1,10] | 1.7 (1.7) | [1,7] | <.001 |
| * * * | |||||
| SRS Total T Score | 74.6 (10.8) | [55,90] | 42.4 (4.5) | [37,51] | <.001 |
| SRS-2 Social Awareness T Score | 73.6 (12.2) | [48,93] | 42.3 (7.4) | [32,57] | <.001 |
| SRS-2 Social Cognition T Score | 73.1 (10.4) | [55,90] | 43.0 (4.3) | [39,52] | <.001 |
| SRS-2 Communication T Score | 74.7 (11.8) | [51,90] | 43.3 (4.7) | [37,54] | <.001 |
| SRS-2 Social Motivation T Score | 64.5 (12.5) | [44,90] | 43.6 (5.5) | [38,58] | <.001 |
| SRS-2 Autistic Mannerisms T Score | 73.2 (12.5) | [46,90] | 44.0 (3.7) | [41,53] | <.001 |
ADOS: Autism Diagnostic Observation Schedule; DAS: Differential Abilities Scale; RRB: Restricted Interests and Repetitive Behaviors; SA: Social Affect; SRS: Social Responsiveness Scale
Experimental Procedures:
Gaze patterns were recorded using a 500 Hz SR Eyelink 1000 Plus eye tracker. Eye-tracking stimuli were presented with Neurobehavioral Systems Presentation 18.1. The SSA 1.0, a free-viewing eye-tracking task, consists of a three-minute video including four conditions: the actress engages viewer’s attention using direct eye contact and speech (Dyadic Bid), initiates acts of joint attention using gaze and speech cues (JA), performs an activity (Sandwich), or looks at moving toys (Animal). The Dyadic Bid and JA conditions were designed to be the most socially-engaging conditions of the SSA task (for detailed description, see Chawarska et al. (2012, 2015)). Participants were seated in front of a 24” 1920 x 1200 pixel LED monitor for stimulus presentation in a quiet, dimly lit room at a distance of 65 cm. During set-up, child-friendly animated movies were shown onscreen to attract the child’s attention.
Data Reduction:
Data reduction and analysis was conducted by the Data Acquisition and Analysis Core of the ABC-CT. Data were processed using custom software written in MATLAB™ which provided standard processing of eye-tracking data including blink detection, outlier detection, eye-tracking calibration and recalibration, measurements of experimental error, and region-of-interest analysis ( Shic, 2008; Shic et al., 2022). The gaze data was down-sampled from 500 Hz to 60 Hz for speed of data processing. Data from participants were included if they met quality-control requirements for eye-tracking data sessions including (1) greater than 50% valid eye-tracking data collected over the course of the stimulus presentation (i.e. the eye tracking system was able to successfully obtain data from one or both eyes of the participant for at least half of the time the stimulus was presented); and (2) the overall calibration uncertainty was less than 2.5 visual degrees (i.e. there was less than 2.5 visual degrees of uncertainty in the evaluated location of the gaze, on average, during the time the participant was recorded by the eye tracker as looking at the stimulus). For additional details regarding eye-tracking data processing, see Shic et al. (2022).
Statistical Approach:
Dependent measures were overall looking time at the scene divided by scene presentation time (%Valid) and proportion of time looking at the face standardized by the overall looking time at the scene (%Face). Data for both %Valid and %Face were analyzed using linear mixed models (LMM) ( Bates et al., 2014) with diagnosis (group, i.e. ASD or TD), condition (Dyadic Bid, Joint Attention (JA), Sandwich, or Animal), and their interaction as factors; full-scale IQ (FSIQ) and age as covariates; and a random intercept per participant. Planned comparisons focused on replication of %Valid and %Face between-group condition effects as highlighted in Chawarska et al. (2012). Effect sizes were calculated as estimated marginal mean differences divided by the square root of the variance components sum ( Cohen, 1988; Westfall et al., 2014). Associations between %Face and ADOS-2 severity of autism symptoms, IQ, and Social Responsiveness T-Scores (SRS-2) in the ASD group were examined in a top-down fashion (assessing primary scale domains first, followed by exploration of sub-scale contributions if the primary scale was significant) using Spearman’s rank correlation coefficients with age as a covariate. As in Chawarska et al. (2012), examination of phenotypic associations was limited to the most social conditions (Dyadic Bid and Joint Attention combined).
Results
Feasibility of the SSA 1.0 Task in Preschool and School-age Children:
Valid eye tracking data were collected from 23/25 (92%) children with ASD and 25/26 (96.3%) TD controls ( Χ2(1,N=51)=.001, p = .972).
Group Differences: ( Table 2).
Table 2.
Estimated marginal means, confidence intervals, and between-group differences between ASD and TD groups by condition for eye-tracking dependent measures of general scene looking (%Valid) and looking at the face (%Face) after controlling for IQ and age.
| Condition | ASD M [95% CI] | TD M [95%CI] | Test | p | d | |
|---|---|---|---|---|---|---|
| %Valid | Dyadic Bid | 81.9% [76.3%, 87.6%] | 95.7% [90.7%, 100.0%] | t(68.0)=−3.28 | .002 | −1.239 |
| Joint Attention | 86.0% [80.4%, 91.7%] | 93.7% [88.8%, 98.7%] | t(68.0)=−1.85 | .068 |
(Truncated: full text at source URL.)