Reduced Engagement of Visual Attention in Children with Autism Spectrum Disorder

Abstract

A common example of social differences in autism spectrum disorder (ASD) is poor modulation of reciprocal gaze, including reduced duration of eye contact and difficulty detecting the aim of another’s gaze. It remains unclear, however, whether such differences are specific to the social domain, or are instead indicative of broader alterations in processes of visual engagement and disengagement in ASD. To assess whether children with ASD experience altered engagement of visual attention, and whether such processes are specific to social stimuli, we implemented a gap-overlap eye-tracking paradigm consisting of both social and non-social images with children with ASD (n=35) and typical development (n=32). Children with ASD demonstrated a significantly reduced overall gap effect (i.e., difference in saccade latency to peripheral stimuli between overlap and gap trials) compared to controls. This reduction spanned both social and non-social conditions. Our findings suggest that children with ASD experience alterations in general processes of engagement of visual attention, and that these alterations are not specific to the social domain, but do associate with cognitive functioning. Affected processes of visual engagement in ASD may contribute to features like poor reciprocal gaze, but social-specific symptoms of ASD likely originate from other sub-cortical processes or higher-order cognition.

Introduction

Autism spectrum disorder (ASD) is a behaviorally-defined neurodevelopmental disorder in which core symptoms fall into two categories: difficulties with social communication and interaction, and restricted, repetitive behaviors and interests (American Psychiatric Association, 2013). Individuals with ASD vary widely in the way their symptoms manifest, making the disorder extremely heterogeneous. One common difference in ASD, however, is poor modulation of reciprocal gaze (American Psychiatric Association, 2013). Individuals with ASD demonstrate this difficulty across the lifespan; reduced duration of eye contact is identifiable in children who will go on to be diagnosed with ASD as early as 2–6 months of age (Clifford, Young, & Williamson, 2007; Jones & Klin, 2013; Shic, Wang, Macari, & Chawarska, 2020) and persists throughout childhood (Sadria, Karimi, & Layton, 2019; Wang et al., 2020; Werner, Dawson, Munson, & Osterling, 2005) into adulthood (Madipakkam, Rothkirch, Dziobek, & Sterzer, 2017; Pelphrey et al., 2002). Manifestations are not limited to the ability to maintain eye contact; both children and adults with ASD experience difficulty accurately detecting the aim of another individual’s gaze (Adamson, Bakeman, Deckner, & Romski, 2009; Campbell et al., 2006; Forgeot d’Arc et al., 2017; Howard et al., 2000). These difficulties in appropriately controlling one’s gaze and recognizing that of others highlight that altered reciprocal gaze is a robust and persistent example of the social differences that characterize ASD.

A number of studies have investigated whether eye gaze differences in ASD may be related to alterations in more basic processes of visual attention. In studies implementing eye-tracking technology, children with ASD have been shown to orient more slowly to and spend less time investigating the social components of a scene (Chawarska, Macari, & Shic, 2012; Frost-Karlsson et al., 2019; Klin, Jones, Schultz, Volkmar, & Cohen, 2002), instead preferring to investigate nonsocial aspects of a visual display (Bacon et al., 2019; Klin, Jones, Schultz, Volkmar, & Cohen, 2002; Król & Król, 2020). These findings indicate that children with ASD may specifically not engage visual attention with social stimuli. However, further eye-tracking studies have shown that children with ASD experience reduced engagement and disengagement of visual attention when stimuli are entirely nonsocial (Goldberg et al., 2002; Landry & Bryson, 2004; Richard & Lajiness-O’Neill, 2015; van der Geest, Kemner, Camfferman, Verbaten, & van Engeland, 2001), and there has been debate in the literature about the extent to which differences in attention are specific to the social domain (Dawson et al., 2004; Iao & Leekam, 2014; Klin, Jones, Schultz, & Volkmar, 2003; Landry & Bryson, 2004; Lewis et al., 2014; Mottron, Dawson, Soulières, Hubert, & Burack, 2006; Mundy, 1995). These latter findings raise the question of whether children with ASD exhibit altered visual engagement and disengagement regardless of the social nature of stimuli.

A particular eye-tracking paradigm that has been crucial for clarifying processes of visual engagement and disengagement is the “gap-overlap” task (Saslow, 1967). In this paradigm, a central stimulus is followed by a peripheral stimulus on either the left or right, and the latency to saccade from the central to peripheral stimulus is measured. Stimuli are presented in two conditions: “gap” conditions present the peripheral stimulus after the offset of the central stimulus, leaving a gap in time between their presentation, whereas “overlap” conditions present the peripheral stimulus while the central stimulus remains, creating a competition for where to look. Gap conditions allow for visual attention to be disengaged from the central stimulus before the presentation of a peripheral stimulus; overlap conditions require both disengagement from the central stimulus and re-engagement to the peripheral stimulus, eliciting relatively longer saccade latencies compared to the gap condition (Fischer & Weber, 1993; Hood & Atkinson, 1993; Saslow, 1967). This paradigm is particularly relevant for investigating visual attention, as it allows for saccade latencies to be compared within each participant, thereby isolating differences in attentional engagement or disengagement from general reaction time. The gap-overlap paradigm also elicits saccades pre-consciously, allowing for investigation of innate processes of attentional engagement as opposed to conscious selection of preferable scene components (Hood & Atkinson, 1993; Matsuzawa & Shimojo, 1997; Ueda, Takahashi, & Watanabe, 2014). This paradigm does not require a threshold of cognitive ability for participation, allowing inclusion of a more broad and representative sample of individuals with ASD.

Several studies have implemented the gap-overlap paradigm to investigate visual attention in children with ASD, yielding conflicting results as to whether children with ASD experience alterations in engagement and/or disengagement of visual attention (Goldberg et al., 2002; Landry & Bryson, 2004; Sabatos-DeVito, Schipul, Bulluck, Belger, & Baranek, 2016; Schmitt, Cook, Sweeney, & Mosconi, 2014; Todd, Mills, Wilson, Plumb, & Mon-Williams, 2009; van der Geest, Kemner, Camfferman, Verbaten, & van Engeland, 2001). However, few studies have combined social and nonsocial images within the same paradigm, and those that have done so have reached further conflicting conclusions regarding visual engagement in ASD and the social nature of these processes (Chawarska, Volkmar, & Klin, 2010; Fischer, Koldewyn, Jiang, & Kanwisher, 2014; Kikuchi et al., 2011).

By utilizing a gap-overlap paradigm with both social and nonsocial images as central and peripheral stimuli, this study aimed to illuminate whether children with ASD experience innate differences in engagement and disengagement of visual attention, and if so, whether these differences are modulated by the social nature of stimuli. We expected that children with ASD would exhibit reduced visual engagement specific to the social domain, as demonstrated by increased relative saccade latency from nonsocial to social stimuli and decreased relative saccade latency from social to nonsocial stimuli. We further expected that reduced visual engagement in ASD would be associated with more severe clinical symptoms related to difficulties with social responsiveness and attention.

Materials and Methods

Participants

Eye-tracking was administered to 67 children between 6 and 14 years of age: 35 children with ASD (M:F = 27:8) and 32 typically-developing (TD) controls (M:F = 13:19). Informed, written consent was obtained from legal guardians, and assent was obtained from participants when appropriate. The Institutional Review Board at the Icahn School of Medicine at Mount Sinai approved this study (HS#: 11–00724).

Participants were enrolled following a confirmed diagnosis of ASD according to the Diagnostic and Statistical Manual of Mental Disorders – Fifth Edition (DSM-5; American Psychiatric Association, 2013); licensed clinical psychologists and psychiatrists made these diagnoses utilizing the Autism Diagnostic Observation Schedule – Second Edition (ADOS-2; Lord et al., 2012) and the Autism Diagnostic Interview-Revised (Lord, Rutter, & Le Couteur, 1994). TD children were recruited from the community and had no history of psychiatric or developmental disorder, and no siblings or close relatives with an idiopathic ASD diagnosis. Specific data on participant race, ethnicity, socioeconomic status, and educational attainment levels were not recorded.

Intellectual functioning estimates were obtained for 34 children with ASD and 26 TD children; depending on age and functioning level, participants were administered one of the following assessments: the Stanford-Binet Intelligence Scales (Roid, 2003), the Wechsler Intelligence Scale for Children – Fifth Edition (Weschsler, 2014), the Wechsler Abbreviated Scale of Intelligence – Second Edition (Weschler, 2011), the Differential Ability Scales – Second Edition (DAS-II; Elliot, 2007), or the Mullen Scales of Early Learning (Mullen, 1995). As estimated IQ score of 40 (equivalent to the minimum possible score on several of our tests) was used for children who earned below-floor scores or required out-of-level testing on the Mullen or DAS-II (n=3). Groups did not differ on age (ASD: mean=10.96 years, SD=2.72; TD: mean=10.33, SD=2.22; t(65)=1.04, p=0.30). Children with ASD had significantly lower IQ scores (mean=90.82, SD=24.09) than TD children (mean=111.96, SD=16.60; t(57)=−3.77, p<0.01). There were also significantly fewer females in the ASD group (n=8, 22.9%) compared to the TD group (n=19, 59.4%; χ²(1, n=67)= 9.27, p<0.01).

To characterize day-to-day expression of both ASD symptoms and symptoms of broader childhood psychopathology, a caregiver completed the Child Behavioral Checklist for Ages 6–18 (CBCL; Achenbach & Rescorla, 2001) for 29 ASD children and 18 TD children, as well as the Social Responsiveness Scale – Second Edition (SRS-2; Constantino & Gruber, 2012) for 30 ASD children and 23 TD children.

Apparatus

Eye-tracking data were collected using an EyeLink 1000 plus eye-tracker in head-free mode at 500 Hz. Images were displayed on a 17-inch LCD monitor (1280 × 1024 pixels, 32 bits per color) with a refresh rate of 60 Hz. Participants were seated approximately 50 cm from the monitor in either a parent’s lap, a chair, or a booster seat facing the screen.

Central and peripheral stimuli were presented as circles in grayscale measuring 208 pixels in height and width. Central stimuli were presented in the center of the screen; peripheral stimuli were presented either to the left or right along the same horizontal plane. Areas of Interest (AOI) were defined as ovals 221 pixels in height and 232 pixels in width, centered around each stimulus location. From a distance of 50 cm, the viewing angle from the center of the AOI surrounding the central stimulus to the center of the peripheral stimulus AOIs was 15°. Social stimuli were 20 faces of children, and nonsocial stimuli were 20 objects (e.g., tennis ball, the Earth, wagon wheel, etc.). Participants viewed a total of 80 trials (40 gap, 40 overlap) in a fixed, pseudo-randomized order. Gap and overlap trials were divided equally into four conditions of 10 trials each based on central and peripheral stimulus type, respectively: nonsocial-nonsocial (“N-N”), social-social (“S-S”), social-nonsocial (“S-N”), and nonsocial-social (“N-S”). The same images were used across both central and peripheral stimulus sets.

Procedure

Participants were seated in a dark room facing the presentation monitor. Prior to the start of the paradigm, participants completed a 5- or 13-point visual calibration and validation routine to confirm accurate tracking of eye movements. All study procedures were passive in nature; completion of calibration, validation, and study trials required participants only to look at stimuli on the screen.

At the start of the paradigm, participants were told only to “watch the screen and its different pictures”; instructions were intentionally limited to avoid priming participants to make particular saccades. For each trial, the central stimulus first appeared and the study administrator confirmed accurate fixation on a separate monitor. Once the participant had maintained central fixation for 1s, the study administrator triggered a peripheral stimulus, which appeared either to the left or right as per the pseudorandom task order. Peripheral stimuli appeared either with the central stimulus still on the screen (overlap trials) or 250 ms after the central stimulus disappeared (gap trials; see Figure 1). Across conditions, peripheral stimuli remained onscreen for 1.5s after onset.

Schematic of sample gap and overlap trials.

Latency to the first saccade from the central to peripheral stimulus relative to the onset of the peripheral stimulus was calculated. Saccades were defined as looking from the AOI surrounding the central stimulus to the AOI surrounding the correct peripheral stimulus. Only trials with an initial central fixation and a saccade to the correct peripheral interest area were included; trials where participants did not saccade from the central stimulus were excluded from analysis. Participants were included for analysis if greater than 50% of both gap and overlap trials were successfully completed; all 67 participants met this criterion. There was no difference in the number of included gap trials between groups (ASD: mean=32.57, SD=4.69; TD: mean=34.22, SD=4.05; t(65)=−1.53, p=0.13), but ASD children had significantly fewer overlap trials included for analysis (ASD: mean=31.11, SD=5.27; TD: mean=34.16, SD=4.63; t(65)=−2.45, p=0.02).

Gap Effect

To quantify engagement and disengagement of visual attention, “gap effect” was calculated for each participant as the mean gap saccade latency subtracted from the mean overlap saccade latency. Because gap saccades require only re-engagement of attention, and overlap saccades require both disengagement and re-engagement, the gap effect is understood to be a measure of the time required to disengage from the central stimulus. A larger gap effect signals an increased overlap latency compared to gap, indicative of reduced disengagement of visual attention from the central stimulus; a smaller gap effect signals a decreased overlap latency compared to gap, indicative of relatively faster disengagement from, and thereby reduced engagement with, the central stimulus (Fischer & Weber, 1993; Kikuchi et al., 2011; van der Geest, Kemner, Camfferman, Verbaten, & van Engeland, 2001). Gap effect was calculated for the task overall and separately for each condition.

Within conditions, individual participants were excluded if they did not make an accurate saccade on more than 50% of either gap or overlap trials. Repeated measures ANOVA testing for condition-specific differences in number of included trials between groups revealed a significant main effect of diagnosis (F(1,65)=7.10, p=0.01), but no significant main effect of condition (F(3,195)=0.55, p=0.65) or condition x diagnosis interaction (F(3,195)=0.08, p=0.97). Though more children with ASD than TD were dropped from condition-specific analyses, no more than four children with ASD (11.43%) were excluded from any condition (mean=3.0, SD=0.82).

Analysis Plan

We first tested for a significant gap effect in each group using two one-sided t-tests versus zero. We then tested for a group difference in overall gap effect using an independent-samples t-test, followed by a univariate ANOVA to confirm that differences in IQ, age, and sex did not drive group effects. If a group difference in overall gap effect was identified, a repeated measures ANOVA was planned to identify group differences in gap effect across social and nonsocial conditions. If a significant interaction between condition and diagnosis was identified here, additional independent-samples t-tests were planned to identify which conditions drove these effects. Finally, bivariate correlations within groups were used to explore relations between gap effect and clinical measures of social functioning, attention, and cognition.

Community Involvement

There is no community involvement in this study.

Results

Both ASD (mean=47.65ms, SD=53.50ms; t(34)=5.27, p<0.01) and TD (mean=108.72ms, SD=53.90ms; t(31)=11.41, p<0.01) children demonstrated a significant gap effect during the task. However, children with ASD displayed a significantly reduced overall gap effect compared to TD children (t(65)=4.65, p<0.01; Cohen’s d=1.14; see Figure 2A). This difference remained significant following univariate ANOVA with IQ, age, and sex as covariates (F(1,54)=8.45, p=0.01; partial η²=0.14). Examining gap and overlap conditions separately revealed that the relative reduction in gap effect seen in ASD was primarily driven by a significant increase in gap saccade latency (ASD: mean=262.28ms, SD=77.90ms; TD: mean=217.50ms, SD=63.69ms; t(65)=−2.56, p=0.01; Cohen’s d=0.63), whereas overlap saccade latency did not differ between groups (ASD: mean=309.93ms, SD=71.17ms; TD: mean=326.22ms, SD=77.65ms; t(65)=0.90, p=0.37; Cohen’s d=0.22; see Figure 2B).

Mean overall gap effect in the autism spectrum disorder (ASD) and typically-developing (TD) groups. Error bars represent ± one standard error. Asterisks indicate p<0.05.

Mean gap and overlap saccade latency in the ASD and TD groups. Error bars represent ± one standard error. Asterisks indicate p<0.05.

Repeated measures ANOVA was conducted to test for condition-specific differences in gap effect between groups. These analyses revealed a significant main effect of diagnosis (F(1,55)=11.60, p<0.001; partial η²=0.17), but no significant main effect of condition (F(3,165)=1.00, p=0.40; partial η²=0.018) or condition x diagnosis interaction (F(3,165)=0.95, p=0.42; partial η²=0.017), confirming that ASD children demonstrated reduced gap effect across all combinations of social and nonsocial stimuli (see Figure 3). The main effect of diagnosis on gap effect remained significant when IQ, age, and sex were included as covariates (F(1,44)=5.90, p=0.02; partial η²=0.12); likewise, the main effect of condition (F(3,132)=0.05, p=0.98; partial η²=0.001) and condition x diagnosis interaction (F(3,132)=2.32, p=0.08; partial η²=0.05) remained non-significant when including these covariates.

Mean gap effect by condition in the autism spectrum disorder (ASD) and typically-developing (TD) groups. Condition labels indicate central and peripheral stimulus type, respectively: nonsocial-nonsocial (“N-N”), social-social (“S-S”), social-nonsocial (“S-N”), and nonsocial-social (“N-S”). Repeated measures ANOVA revealed a significant main effect of diagnosis. Error bars represent ± one standard error.

Children with ASD had significantly higher CBCL Attention (mean=65.45, SD=9.79) and SRS-2 Total (mean=87.10, SD=25.66) scores than TD children (CBCL Attention: mean=51.83, SD=3.67; t(45)=−5.64, p<0.01; Cohen’s d=1.84; SRS-2 Total: mean=22.61, SD=15.10; t(51)=−10.70, p<0.01; Cohen’s d=3.06).

In children with ASD, overall gap effect did not correlate with ASD symptom severity as measured by the ADOS-2 Total score (r(35)=−0.11, p=0.52), nor with severity of social communication deficits specifically, as indexed by clinician ratings (ADOS-2 Social Communication score: r(35)=0.04, p=0.81) or parent report (SRS-2 Total score: r(30)=0.18, p=0.33). Regarding the relation between visual engagement and broader attention problems, overall gap effect in ASD also did not correlate with severity of attention problems measured by the CBCL Attention score (r(29)=0.14, p=0.46). However, overall gap effect did show a significant positive correlation with IQ (r(34)=0.48, p<0.01) within the ASD sample. There was no association between gap effect and age (r(35)=0.10, p=0.56).

In TD children, overall gap effect did not correlate with symptoms of social dysfunction (SRS-2 Total score: r(23)=0.10, p=0.66) or attentional difficulties (CBCL Attention score: r(18)=−0.16, p=0.54). Overall gap effect also did not correlate with either IQ (r(25)=−0.07, p=0.74) or age (r(32)=0.05, p=0.80).

Discussion

Our study found that children with ASD demonstrated a reduced overall gap effect compared to TD children, signifying relatively faster disengagement from, and thereby reduced engagement with, the central stimulus. The difference in gap effect was specifically driven by an elongated gap saccade latency in ASD children. This finding diverges from most previous literature in which group differences in gap effect are typically driven by either altered overlap latency in ASD with relatively spared gap latency (Kikuchi et al., 2011; Landry & Bryson, 2004) or altered latencies of both gap and overlap saccades in ASD (Goldberg et al., 2002; Todd, Mills, Wilson, Plumb, & Mon-Williams, 2009; van der Geest, Kemner, Camfferman, Verbaten, & van Engeland, 2001). One prior study also identified increased gap latency in ASD children, hypothesizing that this alteration was specific to peripheral stimuli that did not align with the child’s circumscribed interests (Mo, Liang, Bardikoff, & Sabbagh, 2019). Our findings, however, identified an increased gap latency in ASD children with interchangeable central and peripheral stimuli not chosen to match children’s particular interests. Our results clarify and extend these previous findings, suggesting that increased gap latency can occur irrespective of specific interests. Of note, previous studies vary widely in methodology; very few implement both social and nonsocial stimuli, and even fewer include a representative sample of cognitive functioning within the ASD population. By including different types of stimuli and enrolling a more representative sample, our study and its result of reduced engagement of visual attention may be more generalizable to the larger ASD population and the kinds of stimuli with which they interact.

Because gap trials allow for disengagement from the central stimulus prior to the presentation of the peripheral stimulus, they represent the latency to re-engage with the peripheral stimulus (Fischer & Weber, 1993; Hood & Atkinson, 1993; Saslow, 1967). Therefore, elongated gap latency suggests that ASD children may experience relatively slow or delayed processes of visual re-engagement. Further, because overlap saccade latency represents the sum of the latency to central stimulus disengagement and the latency to peripheral stimulus re-engagement (Fischer & Weber, 1993; Hood & Atkinson, 1993; Saslow, 1967), the absence of a group difference in overlap latency creates a significantly reduced gap effect and suggests that ASD children also experience reduced engagement with the central stimulus (see Figure 4). These findings identify alterations in processes of visual engagement in children with ASD, including both engagement with an original stimulus and re-engagement with a peripheral stimulus.

Mean overlap latency, gap latency, and gap effect in the autism spectrum disorder (ASD) and typically-developing (TD) groups. Error bars represent ± one standard error.

Gap effect differences in ASD were stable across both social and non-social central and peripheral stimuli. This finding suggests that the altered processes of visual attention in children with ASD are not specific to social stimuli, and instead are more domain-general. These findings suggest that core ASD symptoms such as difficulties with reciprocal gaze and eye contact may be driven, at least in part, by broad alterations in visual engagement processes rather than social-specific differences in visual engagement. Where visual processing of and engagement with social stimuli appear to be selectively impacted in ASD, different low-level (e.g., structural encoding of faces; Desai et al., 2019) or higher-order (e.g., facial recognition; Dawson et al., 2002) cognitive processes may instead be implicated. Our finding of domain-general alterations is consistent with several other literatures supporting broad perceptual processing differences in ASD (Iao & Leekam, 2014; Lewis et al., 2014; Mottron, Dawson, Soulières, Hubert, & Burack, 2006). Such broad processing differences may create or exacerbate differences in behavior that, in turn, result in social symptoms.

In ASD children, gap effect did not appear to be related to clinical measures of social difficulties, including the ADOS-2 Social Affect score and SRS-2 Total score. This further suggests that visual engagement alterations in ASD are not tightly linked to social domain symptoms. We also did not find that gap effect correlated with clinical measures of attention problems on the CBCL; because the CBCL is a parent-report survey that predominantly captures behaviors reflecting broad inattention in daily life, it may not directly capture symptoms specific to engagement of visual attention.

Gap effect was, however, strongly related to cognitive functioning in ASD. This finding is interesting in light of the fact that gap effect and IQ did not correlate in TD children, and group differences in gap effect remained even when differences in IQ were accounted for as a covariate. Previous studies utilizing the gap-overlap paradigm across participants with ASD, TD, and other developmental disabilities (DD) have found that general reaction time and oculomotor ability are largely unaffected by IQ for both ASD and DD; however, altered processes of attentional engagement and the relation between these processes and cognitive ability have been shown to be specific to ASD, not DD, participants (Chawarska, Volkmar, & Klin, 2010; Sabatos-DeVito, Schipul, Bulluck, Belger, & Baranek, 2016). Furthermore, in several previous studies in which ASD participants were matched for IQ with control participants, no differences in gap effect were identified (Fischer, Koldewyn, Jiang, & Kanwisher, 2014; Goldberg et al., 2002; Kikuchi et al., 2011; Schmitt, Cook, Sweeney, & Mosconi, 2014). Our findings support this previous literature by identifying an effect of IQ on attentional engagement that is specific to the context of ASD. It may therefore be the case that reduced engagement of visual attention and its underlying neural processes are more tightly linked to processes that affect cognition, rather than social processing, in ASD.

The increased prevalence of ASD among males, as well as the increased frequency of cognitive impairment in ASD (Christensen et al., 2019), limited our ability to match our ASD and TD samples on the bases of sex and IQ. However, we included these variables as covariates in our analyses and confirmed that our results were not significantly affected by group differences in these variables. Our ASD sample did have fewer accurate saccades in the overlap condition; however, there was no group difference in the number of overlap trials for which participants did not make a saccade (ASD: mean=0.69, SD=0.93; TD: mean=0.63, SD=1.07; t(65)=0.25, p=0.81), confirming that overlap saccade latency in the ASD group was not disproportionately affected by exclusion of trials in which participants did not disengage from the central stimulus. Additionally, data integrity was assessed to confirm that a majority of trials were accurate for each participant overall, as well as within each condition.

Future studies should seek to clarify the specific neural mechanisms implicated in processes of visual engagement, both to shed light on biological processes uniquely affected in ASD and potentially to identify a specific localization for targeted treatment. Previous studies have found that gap saccades are predominantly driven by subcortical pathways between the retina and superior colliculus (Chernenok, Burris, Owen, & Rivera, 2019; Csibra, Tucker, & Johnson, 1998; Farroni, Simion, Umiltà, & Barba, 1999); thus, an important next step will be to clarify whether engagement and re-engagement of visual attention utilize the same sub-cortical mechanisms, and to explore the extent to which each is affected in ASD. Alternatively, increased gap latency in ASD may be driven by reduced use of fixation stimulus offset as a cue to saccade to the impending peripheral stimulus (Wilson & Saldaña, 2019). Future studies should seek to differentiate processes of visual re-engagement from the effects of visual cueing in ASD. As altered visual engagement related specifically to cognition in our ASD sample, identifying underlying mechanisms may provide a window into ASD-specific cognitive processes.

Conclusion

Our results identify alterations in visual attention engagement in children with ASD, including processes of both engagement and re-engagement, and reveal that such alterations are not specific to the social domain but do associate with cognitive functioning. These novel findings expand upon previous eye-tracking studies in ASD and suggest domain-general differences in visual engagement and disengagement in children with ASD that likely contribute to changes in visual attention. As affected processes of visual engagement in ASD appear not to be specific to the social domain, future studies should implement research paradigms that target other subcortical and higher-order processes to identify where specifically social difficulties in ASD originate and what mechanisms may be implicated in such processes.