Hyper-Focus, Sticky Attention, and Springy Attention in Young Autistic Children: Associations with Sensory Behaviours and Cognitive Ability

Abstract

The autistic-developed monotropism account suggests that atypical, domain-general attentional hyper-focus on interests is a central aspect of autism, but domain-general attention differences in autism can manifest differently. Prior research suggests autistic children are often slow to disengage attention from stimuli – a pattern often called “sticky attention” – and that they can show reduced novelty preference. These attentional patterns could influence sensory experiences and learning. We used eye-tracking to investigate novelty preference and sticky attention” in young autistic children; we also examined whether attentional patterns were related to cognitive abilities and caregiver-reported sensory responsiveness. 46 autistic and 28 non-autistic participants, aged between 2–4 years, provided usable data. We found no evidence that autistic children exhibited greater “sticky attention” than non-autistics, but “sticky attention” in autism was associated with more caregiver-reported sensory hyper-responsiveness, seeking/interests, and enhanced perception. Autistic children also non-significantly trended towards exhibiting reduced novelty preference. Unexpectedly, the time-course of this trending novelty preference difference implied it was not driven by reduced orienting to novelty, but increased returning to already-familiarized stimuli: what we call “springy attention.” Exploratory analyses of data from the attentional disengagement task suggest autistic participants may have exhibited greater “springy attention,” though further research with paradigms optimized for measuring this construct should confirm this. Importantly, “springy attention” was robustly related to reduced cognitive abilities and greater caregiver-reported hypo-responsiveness. Thus, this study illuminates two distinct domain-general attentional patterns, each with distinct correlates in young autistic children, which could have important implications for understanding autistic children’s learning, development, and experiences.

Lay Summary

We tracked young autistic children’s eye gaze in order to gain insights about how they attend to and experience things around them. Autistic children who were slower to “unstick” their attention from images showed, according to their caregivers, increased sensory responses, enhanced sensory perceptions, and high rates of sensory interests or sensory seeking behaviors. In other words, autistic children’s tendency to get “stuck” on things could lead to an intensified sensory or perceptual experience, which might be pleasant or unpleasant. We also found that autistic children who looked at newly-presented things for less time before “springing” back to what they looked at before, had lower cognitive abilities, which might be due to children missing opportunities to learn from exploring their environments.

Introduction

Autism Spectrum Development (ASD)¹ is a common neurodevelopmental disability characterized by difficulties with social communication and interaction (at least in some contexts; Crompton, Ropar, et al., 2020; Crompton, Sharp, et al., 2020), as well as a wide variety of non-social characteristics. These non-social characteristics often include repetitive motor movements, difficulty coping with uncertainty and lack of control, intense or focused interests, and differences in sensory reactivity (Grove et al., 2021).

A promising approach towards the study of ASD emphasizes how early differences in attention can have cascading effects on learning and development (Bradshaw et al., 2022; Murray et al., 2005; Vivanti et al., 2022). Indeed, as attention allocation shapes one’s experience of the world, understanding atypical attention appears highly consequential for understanding autistic development and autistic people’s day-to-day experiences. Although some accounts of autism emphasize social attention, atypicalities of autistic attention appear fairly domain-general (R. Landry & Bryson, 2004; Venker et al., 2021). For example, the autistic-developed monotropism account suggests that autism is, at its core, characterized by a tendency to narrowly hyper-focus attention, in a sort of “tunnel,” towards topics or stimuli of interest to the individual (Murray et al., 2005).

However, many questions remain regarding hyper-focus and monotropism in autism and other neurodevelopmental differences, such as ADHD. For example, although hyper-focus has been regarded as synonymous with flow (Ashinoff & Abu-Akel, 2021) – or states of intense positive engagement in tasks that are challenging but within a person’s ability (Beard & Csikszentmihalyi, 2015) – measures of hyper-focus and flow are inversely related in ADHD adults (Grotewiel et al., 2022). This could suggest that “hyper-focus” is not unitary but reflects multiple processes, though these processes have yet to be clearly distinguished and described. Moreover, increased hyper-focus appears to be associated with increased distractibility (Dwyer et al., 2024), suggesting that attention capture and hyper-focus might share some common roots. To clarify this conceptual confusion regarding domain-general attention in neurodivergent people, further mapping of distinct manifestations of atypical attention is necessary.

Sensory Processing in Autism

But attention is not the only factor influencing early development in autism. Like atypical attention, atypical sensory processing appears to have cascading, domain-general effects on autistic children’s development (Cascio et al., 2016). Atypical sensory processing is apparent in infancy (Baranek, 1999; Kolesnik et al., 2019), and longitudinal studies of infants at heightened autism likelihood suggest sensory differences can predict later social/communication differences (Baranek et al., 2018; Damiano-Goodwin et al., 2018; Feldman et al., 2022).

Many different frameworks exist for conceptualizing and describing autistic sensory processing (e.g., Belek, 2018; DeBoth & Reynolds, 2017; Ward, 2018; Williams et al., 2021). Reviewing these is beyond the scope of this study, but one caregiver-reported measurement model includes hyporesponsiveness (HYPO); hyperresponsiveness (HYPER); sensory interests, repetitions, and seeking (SIRS); and enhanced perception (EP) (Ausderau et al., 2014).

Some accounts suggest atypical attention leads to some autistic sensory behaviours/experiences (Dwyer, Takarae, et al., 2022; Dwyer et al., 2023b; Murray et al., 2005; Thielen & Gillebert, 2019). Theoretically, experiences of sensory distress and hyper-reactivity could reflect hyper-vigilance towards and attention capture by aversive stimuli, sensory interests could reflect intense focus on engaging or pleasurable stimuli, and hypo-reactivity might sometimes reflect stimuli falling outside attentional awareness. But regardless of its mechanism, sensory processing in autism can be very impactful in daily life. For example, studies show that it is related to autistic people’s activity participation (Ismael et al., 2018; Little et al., 2015), sleep (Dwyer, Ferrer, et al., 2022; Tzischinsky et al., 2018), quality of life (Lin & Huang, 2019), and mental health (Neil et al., 2016).

Novelty Preference in Autism

Prior research suggests one form of atypical attention in ASD could be diminished novelty preference (Arora et al., 2022; Vivanti et al., 2018). When playing, autistic children handle fewer objects than typically-developing children (Fanning et al., 2021). In many theories of human development, the attentional exploration of novel stimuli from the environments of infants and young children is crucial to learning and cognitive ability (Bornstein & Sigman, 1986; Piaget, 1954; Sternberg, 1981); indeed, in the general population, infant novelty preference has been used to predict adult cognitive ability and academic achievement (Fagan et al., 2007). Naturalistic-developmental behaviour interventions actively endeavour to recruit young autistic children’s attention to pleasurably engage them in novel experiences and learning opportunities (Frost et al., 2021; Schreibman et al., 2015; Vivanti et al., 2022). Yet surprisingly, although novelty preference appears to be associated with social and non-social autistic behaviours (Arora et al., 2022; Vivanti et al., 2018), prior autism studies have not found links between novelty preference and cognitive ability (Arora et al., 2022; Fanning et al., 2021; Vivanti et al., 2018). Given the theoretical importance of novelty to learning and cognitive ability, it is possible that other tasks, such as those that have predicted cognitive ability in non-autistic populations (Fagan et al., 2007), might yield different results. Currently, it is unclear whether novelty preference in autism is related to sensory experiences or reactivity.

“Sticky Attention” in Autism

Prior research suggests that another manifestation of domain-general differences in autistic attention may be slow attention disengagement, or “sticky attention” (Keehn et al., 2019; R. Landry & Bryson, 2004; Sacrey et al., 2014). Slowness to disengage attention appears consistent with the monotropism account, that is, the theory that autism is fundamentally characterized by the narrow focusing of attention a small number of targets, resulting in other stimuli being relatively less attended to (Murray et al., 2005; Murray, 1996). Notably, the theory’s developers explicitly identify slow set-shifting as a possible manifestation of monotropism (Murray et al., 2005). Furthermore, “sticky attention” emerges early; prospective infant studies suggest that those later diagnosed as autistic exhibit atypical average attentional disengagement at least by 12 months, and possibly earlier (Sacrey et al., 2014). “Sticky attention” is commonly quantified using “gap-overlap” tasks, in which participants are first presented with a central stimulus, then a peripheral target. In “gap” trials, the first stimulus disappears before the target appears, allowing participants to orient to the peripheral target with relatively little delay. In “overlap” trials, the peripheral target appears while the first stimulus remains onscreen, requiring participants to disengage attention from the first stimulus before focusing on the new target, possibly delaying fixation on that target.

Furthermore, some prior research suggests attention disengagement and autistic sensory behaviours are linked. Although few studies have searched for such relationships, one study of autistic children and children with intellectual/developmental disabilities found that slower gap-overlap disengagement (“stickier attention”) was concurrently associated with hypo-responsiveness and sensory seeking, whereas faster disengagement was associated with hyper-responsiveness (Sabatos-Devito et al., 2016). Another study, using aggregated caregiver-report and observational measures of disengagement from infancy, found that “sticky attention” predicted later sensory seeking (Baranek et al., 2018).

“Springy Attention” in Autism

One outstanding question is what might happen after study participants look away from the previous focus of their attention in a gap-overlap task or a novelty preference task and begin to focus on a newly-presented stimulus. Do participants often return to their previous focus of attention? Even in enjoyable leisure activities, such as playing video games, humans can exhibit momentary mind-wandering (Varao-Sousa, 2019), suggesting that momentary inattention need not prevent people from quickly returning to a focused state. Thus, we explore whether autistic and non-autistic children orient to a newly-presented stimulus then disengage from it to return to what they were looking at before, a pattern that we call “springy attention.”

Present Study

In this study, part of the larger Brain Research in Autism Investigating Neurophenotypes (BRAIN) project at the UC Davis MIND Institute, we collected passive eye-tracking data from young children between 2–4 years of age. We used a novelty preference task and a gap-overlap task to measure “sticky attention” (attention disengagement). We expected that:

1. In the gap-overlap task, autistic participants would fixate on the peripheral target more slowly than non-autistic participants in the overlap condition, reflecting a pattern of slow disengagement, or so-called “sticky attention”;

2. In the novelty preference task, autistic participants would show less novelty preference than non-autistic participants;

3. In autistic participants, reduced novelty preference and slower attention disengagement would be associated with lower cognitive abilities, increased sensory hyper-responsiveness, and more sensory seeking.

After viewing preliminary results, we conducted additional analyses to explore “springy attention,” or the propensity to return to a previous object of attention after briefly orienting to a newly-presented target.

Methods

Participants

Participants were recruited from the community via social media posts, fliers placed in local public libraries, pediatrician offices, and community health fairs. Participants were also recruited from the MIND Institute Participant Research Registry. 95 young children (65 autistic, 30 non-autistic) between 2–4 years old visited the laboratory for eye-tracking data collection between 2018 and 2023. However, gap-overlap and novelty preference data were sometimes not collected due to technical issues (n=2, both autistic) or participant movements/unhappiness/inability to track eyes (n=13, 11 autistic). In four more cases (all autistic), only novelty preference data were attempted due to movements/unhappiness, and in two cases (both autistic), only gap-overlap collection was attempted for the same reason. After data processing, a total of 46 autistic and 28 non-autistic participants provided usable data on at least one task (Table 1): 42 autistic and 26 non-autistic in the gap-overlap task, and 33 autistic and 22 non-autistic in the novelty preference task (Supplementary Table 1).

Table 1.

Characteristics of autistic and non-autistic participants who had usable data on at least one of the two eye-tracking tasks. For categorical variables (sex), p-values are based on Fisher’s exact tests, with Cramér’s V as an effect size; an accompanying Bayes Factor was calculated using the contingency table function of the R package ‘BayesFactor’ (Morey et al., 2022) under Poisson sampling using default values of the prior concentration parameter. For continuous variables (age, MSEL, Vineland, SEQ, SCQ, SRS, and ADOS), p-values are based on Wilcoxon-Mann-Whitney tests with Cliff’s δ as an effect size; ranked Bayesian t-tests (Morey et al., 2022) are provided to further contextualize results using default priors (noninformative Jeffreys prior for variance, standard Cauchy prior for effect size). Bayes Factors > 3.00 are generally considered to provide evidence of effects, and Bayes Factors < 0.33 are generally considered to provide evidence against effects.

	Autistic		Non-Autistic		p BF 10	Effect size [95% CI] (Cramér’s V, Cliff’s δ)
	Mean (SD)	Range	Mean (SD)	Range
Sex	40 male 6 female		20 male 8 female		.13 1.59	.19 [.01, .43]
Age (months)	36.20 (5.15)	25.30 – 45.24	37.70 (6.21)	26.22 – 51.22	.25 0.44	−.16 [−.41, .11]
MSEL DQ (43 ASD, 26 non-ASD)	65.37 (20.00)	29.27 – 118.73	107.36 (11.19)	88.27 – 123.64	<.0001 >999,999	−.92 [−.97, −.76]
Vineland-2 Adaptive Behaviour Composite (39 ASD, 21 non-ASD)	77.08 (12.68)	48.00 – 105.00	107.29 (12.07)	81.00 – 132.00	<.0001 >999,999	−.91 [−.97, −.77]
SEQ-3.0 HYPER (31 ASD, 17 non-ASD)	2.27 (0.51)	1.52 – 3.90	1.67 (0.40)	1.19 – 2.48	.0002 432.83	.67 [.36, .84]
SEQ-3.0 HYPO (31 ASD, 17 non-ASD)	2.24 (0.55)	1.42 – 3.95	1.37 (0.31)	1.00 – 2.18	<.0001 652,624	.86 [.64, .95]
SEQ-3.0 SIRS (31 ASD, 17 non-ASD)	2.92 (0.56)	1.69 – 4.16	1.90 (0.54)	1.16 – 3.38	<.0001 67,034	.81 [.50, .94]
SEQ-3.0 EP (31 ASD, 17 non-ASD)	1.96 (0.49)	1.17 – 3.00	1.97 (0.48)	1.33 – 3.00	.93 0.30	.02 [−.32, .35]
SCQ Total (39 ASD, 20 non-ASD)	18.74 (6.25)	4.00 – 33.00	5.65 (3.23)	0.00 – 16.00	<.0001 >999,999	.93 [.76, .98]
ADOS Calibrated Scores (33 ASD)	7.27 (1.86)	4.00 – 10.00
ADI-R Reciprocal Social Interaction (43 ASD)	17.47 (4.64)	9.00 – 25.00
ADI-R Communication (43 ASD)	11.77 (3.48)	5.00 – 20.00
ADI-R Repetitive Behaviour (43 ASD)	6.33 (2.34)	2.00 – 12.00

Abbreviations: MSEL DQ – Mullen Scales of Early Learning Developmental Quotient; SEQ-3.0 – Sensory Experiences Questionnaire, version 3.0; HYPER – hyperresponsiveness; HYPO – hyporesponsiveness; SIRS – sensory interests, repetitions, and seeking; EP – enhanced perception; SCQ – Social Communication Questionnaire; ADOS – Autism Diagnostic Observation Schedule; ADI-R – Autism Diagnostic Interview-Revised.

Note that ADI-R scores reflect the comprehensive diagnostic (not current behaviour) algorithm (Lord et al., 1994).

Autistic participants had previous community autism diagnoses, which were confirmed by clinician judgement, per DSM-5 (American Psychiatric Association, 2013), based on assessments including administration of an observational measure to all participants, which could be either the Autism Diagnostic Observation Schedule-2 (ADOS-2; Lord et al., 2012) or, as a COVID-19 adaptation, the Brief Observation of Symptoms of Autism (BOSA; Dow et al., 2021) or an ADOS in which the examiner and participant wore face masks (preventing scoring). Furthermore, the Autism Diagnostic Interview-Revised (Lord et al., 1994) was also collected from 44 of the 46 autistic participants (including two incomplete interviews).

The Social Communication Questionnaire (SCQ, Berument et al., 1999) was used to screen general population comparison participants for autism; in the present study’s age range, it appears to show good sensitivity and specificity in distinguishing autistic children from members of the general population (Marvin et al., 2017). Non-autistic participants with SCQ scores ≥ 11 were only included if they were assessed by clinicians as non-autistic in an evaluation including ADOS or BOSA administration. When the SCQ was unavailable, clinician judgement (based on visits involving administration of the Mullen Scales, ADOS, and/or other standardized measures) ruled out autism. One non-autistic participant had neither SCQ scores nor clinical encounters.

Measures

Sensory Experiences Questionnaire

The Sensory Experiences Questionnaire, Version 3.0 (SEQ-3.0; Ausderau et al., 2014) is a 105-item, parent-report sensory behaviour questionnaire. The first 97 items measure behaviour frequency on a 5-point Likert scale, loading into several modality factors (auditory, visual, tactile, gustatory/olfactory, and vestibular/proprioceptive), two context factors (social and nonsocial), and four sensory response pattern factors: hyporesponsiveness (HYPO); hyperresponsiveness (HYPER); sensory interests, repetitions, and seeking (SIRS), and enhanced perception (EP). Hyperresponsiveness refers to increased reactivity to or avoidance of stimuli, consistent with experiences of sensory distress; hyporesponsiveness refers to reduced or delayed responses to stimuli; and SIRS refers to intense fascinations towards or craving of stimuli (Ausderau et al., 2014). Enhanced perception is conceptualized as “superior acuity in the awareness of specific sensory stimuli” (Ausderau et al., 2014), but prior research suggests that it is not related to sensory acuity thresholds measured in laboratory settings (Dwyer, 2023; Dwyer, Takarae, et al., 2022), implying it may have more to do with noticing or responding to subtle stimuli in the real world.

The SEQ-3.0 was added to the study after data collection began, so it was only available from caregivers of 31 autistic and 17 non-autistic participants. Missing items (0.37%) were imputed using datawig’s SimpleImputer.complete function, imputing missing values in each column from all other columns; only the numeric Likert scale items were used in the imputation process (Bießmann et al., 2019).

Mullen Scales of Early Learning (MSEL)

The Mullen Scales are a standardized, norm-referenced measure of cognitive ability for children aged 0–68 months (Mullen, 1995). In this study, subscales indexing Receptive Language, Expressive Language, Visual Reception, and Fine Motor were administered and used to calculate a Developmental Quotient (DQ) comparable to IQ, reflecting the average of the age equivalent scores from the different subscales, divided by chronological age, multiplied by 100, as in prior research (Nordahl et al., 2011). This procedure prevents floor effects caused by the limited range of MSEL standard scores (see Akshoomoff, 2006; Ostrolenk & Courchesne, 2023; Solomon et al., 2018).

Vineland Adaptive Behaviour Scales, 2^nd edition (Vineland 2)

In this study, the Vineland-2 (Sparrow et al., 2005) was administered as a caregiver-report questionnaire. It is a standardized, norm-referenced measure of adaptive behaviour and everyday functioning. We included the communication, daily living skills, socialization, and motor skills domains.

Eye-tracking Apparatus

Binocular infrared eye-tracking data were recorded at 60 Hz using Tobii X-60 eyetrackers, or for a subset of participants, loaner Tobii X2–60 devices. Eye-trackers were kept in dimly lit, quiet rooms at the UC Davis MIND Institute and the UC Davis Center for Mind and Brain. Tobii Pro Lab software presented stimuli and recorded data; recordings were calibrated using a five-point calibration procedure offered by the software. Stimuli were presented on 58.4 cm, 1920×1080-pixel Dell monitors; a 1440×1080 area was used to present stimuli against a white background, leaving black bars at the corner of the screen. Participants sat in a caregiver’s lap approximately 60–70 cm in front of the eye-tracker and monitor.

Procedures

The gap-overlap and novelty preference passive eye-tracking tasks described in this manuscript were collected alongside 2–3 other eye-tracking tasks, with data collection lasting ~20–30 minutes in total. One measured receptive vocabulary (described in another sample by Yoo et al., 2017), one was a dot probe task (described in another sample by Burris et al., 2017), and the last (collected from a subset of participants) was an abbreviated version of the intersensory processing task developed and described by Bahrick et al. (2018).

Gap-Overlap Task

The gap-overlap task (Figure 1A) used parameters previously reported by Chernenok et al.(2019). It included 36 trials (18 gap, 18 overlap), each lasting 3500 ms. In half of the trials, the peripheral target appeared on the left, and in the other half, on the right. In both conditions, trials began with presentation of a central fixation stimulus (one of 6 black-and-white, high-contrast shapes) for 1000 ms. In the gap condition, the central fixation object then disappeared, and a blank screen was presented for 500 ms. A peripheral target (one of 18 colourful toys of stuffed animals) then appeared on either the left or the right side of the screen for 2000 ms. In contrast, in the overlap condition, the central fixation stimulus never disappeared; instead, the peripheral target appeared alongside it for 2500 ms.

Figure 1.

Example trials illustrating the design of the gap-overlap (Panel A, left) and novelty preference (Panel B, right) tasks.

Novelty Preference Task

The novelty preference task (Figure 1B) was adapted from the Fagan Test of Infant Intelligence (Fagan et al., 2007; Fagan & Detterman, 1992). 4 trials were collected, each lasting 20 s. The first 10 s was a familiarization phase, in which two black-and-white, high-contrast figures were presented, one on the left of the screen and one on the right. Stimuli were positioned in two areas subtending approximately 14° in width and height, occupying the right and left portions of the utilized area of the display monitor as depicted in Figure 1B. After 10 s, one of the images disappeared and was replaced by a novel, unfamiliar black-and-white, high-contrast figure. Both the familiarized and novel stimuli then remained visible for another 10 s.

Data Processing and Analysis

Eye-tracking fixations were defined using the default parameters of the velocity-based Tobii I-VT fixation filter (Olsen, 2012). Briefly, when gaze velocity was less than 30°/s, they were considered part of a fixation. To reduce noise in gaze data, moving medians were calculated based on windows of 3 data points. Short fixations (<60 ms) were discarded, but adjacent fixations (<0.5°, 75 ms) were merged. Raw fixation coordinates at consecutive gaze samples were exported from Tobii Pro Lab and analyzed using eyetrackingR (Dink & Ferguson, 2015) and custom R code. Rectangular areas of interest (AOIs) were defined around the familiarized and novel stimuli from the novelty preference task as well as around the central stimulus and the peripheral target from the gap-overlap task. Only data samples evaluated by Tobii Pro Lab as being valid for at least one eye were included in the analysis.

Gap-Overlap Task

“Sticky Attention.”

In the gap-overlap task, in both the gap and overlap conditions, we excluded trials in which valid gaze data, within the AOIs, was not obtained in at least 40% of samples collected in the 2000 ms period following onset of the peripheral target, as well as trials in which participants did not fixate to the peripheral target during this 2000 ms period. Participants providing fewer than 4 usable trials in either the gap or overlap condition were then excluded, yielding a revised sample of 42 autistic and 26 non-autistic participants (autistic participants with unusable data had statistically elevated ADI-R communication and repetitive behaviour scores, but there was statistical evidence against differences in ADI-R social interaction scores and in overall ADOS calibrated scores – see Supplementary Table 2). Latency to first fixation on the peripheral target was calculated for the remaining trials, providing a metric of “sticky attention.” Trial counts did not statistically differ across groups (Supplementary Table 3).

“Springy Attention.”

Valid trials in which participants fixated to the peripheral target within 1000 ms of its onset were then used to calculate a metric of “springy attention.” Only participants who still provided at least 4 usable trials in each condition were retained, leaving 40 autistic and 26 non-autistic participants. Visual preference to the central stimulus 500–1000 ms following first fixation to the peripheral target was extracted (i.e., the proportion of looking time directed to the central AOI) separately in the gap and overlap conditions. Trial counts did not statistically differ across groups (Supplementary Table 3).

Data Analyses.

Two-way ordinal regression, a cumulative link mixture model and analysis of deviance (Christensen, 2022), was used to examine effects of diagnosis, condition, and their interaction on participants’ mean latencies to first fixation on the peripheral target (for the “sticky attention” analysis) and on visual preference to the central stimulus in the 500–1000 ms following fixation to the peripheral target (for the “springy attention” analysis). Cliff’s δ (Torchiano, 2022) was calculated as an effect size for the comparison across diagnostic groups, while the matched-pairs rank biserial correlation (Mangiafico, 2016) was used as an effect size for the comparison of the gap and overlap conditions. As an additional check, Bayesian ranked linear models, with default priors, and including random/varying effects of participant, were compared against one another in order to calculate Bayes Factors (Morey et al., 2022), further contextualizing the cumulative link mixture model effects.

In autistic participants, Spearman’s ordinal correlation coefficient indexed associations between eye-tracking metrics (the difference between overlap and gap latencies to first fixation, a metric of “sticky attention”; and visual preference to the central stimulus in the overlap condition 500–1000 ms after first fixation on the peripheral stimulus, a “springy attention” metric) and other variables, namely sensory behaviours (the four SEQ sensory pattern scores) and cognitive ability (MSEL DQ). A false discovery correction for five multiple comparisons was applied separately for each eye-tracking metric (Benjamini & Hochberg, 1995). Bayesian analyses are also provided for comparison; these were calculated using the “correlation” R package’s default “medium” priors (Makowski et al., 2020). Correlations in non-autistic participants are provided in Supplementary Table 4.

Novelty Preference Task

Novelty Preference.

In the novelty preference task, we excluded trials in which valid gaze data, within the AOIs, was not obtained in at least 40% of samples collected in the 2000 ms following onset of the novel target. Participants providing fewer than 2 usable trials were then excluded, yielding a revised sample of 33 autistic and 22 non-autistic participants (characteristics of autistic participants with/without usable data on this task did not statistically differ – see Supplementary Table 2). Visual preference to the novel target over the 2000 ms following its onset was extracted as a metric of novelty preference. Trial counts were statistically equivalent (Supplementary Table 3).

“Springy Attention.”

Furthermore, out of these trials, trials in which participants fixated to the novel target within 1000 ms of its onset and in which data loss was no more than 30% for the following 2000 ms, were used to calculate a metric of “springy attention.” Again, participants were excluded if fewer than 2 trials were usable under these criteria, yielding a sample of 30 autistic and 22 non-autistic participants. Visual preference to the familiarized nontarget stimulus in the period 500–2000 ms following first fixation on the novel target was calculated. Trial counts did not statistically differ across groups (Supplementary Table 3).

Data Analyses.

For comparisons of visual preferences across groups, Wilcoxon Mann-Whitney tests were used with Cliff’s δ as an effect size metric. Ranked Bayesian t-tests with default priors were added as an additional check on results validity (Morey et al., 2022).

Associations between visual preferences and sensory behaviours (SEQ sensory pattern scores) and cognitive ability (MSEL DQ) were examined using Spearman’s ordinal correlation coefficient, with a family-wise error correction for five comparisons (separately for novelty and “springy attention” visual preferences). Bayesian correlations were calculated with default priors (Makowski et al., 2020). Correlations in non-autistic participants are provided in Supplementary Table 4.

Results

“Sticky Attention”

An effect of task condition on latencies to first fixation on the peripheral target, Χ²=7.38, p=.007, r_rb=−.39, 95% CI: [−.64, −.10], reflected slower latencies in the overlap condition than the gap condition: delays due to disengagement of attention from the central stimulus in the overlap condition, i.e., “sticky attention” (Figure 2). Surprisingly, there was neither an effect of diagnostic group, Χ²=1.21, p=.27, δ=.14, 95% CI: [−.06, .33], nor an interaction between diagnostic group and task condition, Χ²=0.24, p=.62. Ranked Bayesian linear models provided evidence of an effect of condition, BF₁₀=3.79, and against a group*condition interaction, BF₁₀=0.30; there was no decisive evidence against a main effect of diagnostic group, BF₁₀=0.48.

Figure 2.

Latencies to first fixation on the peripheral target in the gap-overlap task, as a function of diagnostic group and task condition. Slower latencies in the overlap condition reflect delays due to the need to disengage attention from the central stimulus: in other words, “sticky attention.” However, there was no group by condition interaction, suggesting groups did not appear to differ in the extent of this “sticky attention.”

As reported in Table 2, some variables were significantly associated with differences between autistic participants’ gap and overlap latencies, an index of “sticky attention.” In autistic participants, slower attention disengagement was associated with greater sensory hyper-responsiveness, Spearman’s ρ=.43, corrected p=.03 (Figure 3A); with more caregiver-reported enhanced perception, ρ=.44, corrected p=.03 (Figure 3C); and with more caregiver-reported sensory interests and seeking, ρ=.51, corrected p=.02 (Figure 3D). However, there were no significant associations between speed of attention disengagement and either sensory hypo-responsiveness or cognitive ability, corrected p≥.19 (Table 2).

Table 2.

Among autistic participants only, ordinal correlations (Spearman’s rho, ρ) between eye-tracking task indices and other variables, namely caregiver-reported sensory behaviours and MSEL-assessed overall cognitive abilities. Correlation coefficient point estimates and 95% confidence intervals are derived from frequentist analyses. Frequentist p-values are reported both before uncorrected and after, separately for each eye-tracking index, a false discovery rate correction for five multiple comparisons (five correlations) was applied. Furthermore, Bayes Factors (BF₁₀) are provided; values less than 0.33 would generally be considered evidence in favour of no effect, while values greater than 3.00 would be considered evidence supporting the existence of a statistical association between variables. Finally, the Bayesian probability of direction (PD) is reported, which is a probability that a parameter demonstrates a statistical effect (in a positive or negative direction, hence probability of direction); PD values >97.5% are consistent with frequentist p<.05 (Makowski et al., 2019).

	Gap-Overlap Task								Novelty Preference Task
	“Sticky Attention” (Overlap – Gap Latency)				“Springy Attention” (Overlap VP to Centre)				Novelty Preference (Initial VP to Novel AOI)				“Springy Attention” (Later VP to Familiar AOI)
	ρ [95% CI]	p raw p corrected	BF10 PD	n	ρ [95% CI]	p raw p corrected	BF10 PD	n	ρ [95% CI]	p raw p corrected	BF10 PD	n	ρ [95% CI]	p raw p corrected	BF10 PD	n
SEQ HYPER	.43 [.07, .69]	.02* .03*	2.39 97.55%*	29	.09 [−.30, .44]	.66 .83	0.45 66.22%	28	−.37 [−.69, .06]	.09 .38	1.47 93.92%	22	.54 [.13, .79]	.01* .07	4.99* 98.62%*	20
SEQ HYPO	.27 [−.10, .58]	.15 .19	1.45 94.90%	29	.50 [.16, .74]	.006** .02*	9.36* 99.40%*	28	−.32 [−.65, .12]	.15 .38	1.04 90.42%	22	.32 [−.15, .66]	.18 .44	0.98 89.70%	20
SEQ SIRS	.51 [.18, .74]	.005** .02*	13.37** 99.78%**	29	.17 [−.21, .51]	.38 .64	0.57 78.17%	28	.06 [−.37, .47]	.79 .79	0.47 59.13%	22	−.23 [−.61, .23]	.32 .54	0.70 82.12%	20
SEQ EP	.44 [.09, .70]	.02* .03*	5.10* 99.02%**	29	−.02 [−.39, .35]	.91 .91	0.41 54.27%	28	−.12 [−.51, .32]	.60 .79	0.51 68.83%	22	−.17 [−.57, .29]	.46 .58	0.59 74.22%	20
MSEL DQ	−.05 [−.36, .27]	.76 .76	0.39 66.72%	39	−.68 [−.82, −.46]	<.001*** <.001***	4430*** >99.99% ***	37	.08 [−.29, .42]	.68 .79	0.42 65.35%	31	.08 [−.29, .44]	.66 .66	0.44 66.42%	29

Figure 3.

Associations between sensory and cognitive variables and gap-overlap attention disengagement difference scores (positive difference scores reflect slower fixations to the peripheral target in the overlap condition than the gap condition, i.e., slower attention disengagement or “sticky attention”). Note that these plots depict linear slopes, whereas the correlations reported in the main text use ordinal slopes that are not unduly influenced by outliers.

Panel A. Autistic participants slow to disengage their attention from the central stimulus had more sensory hyper-responsiveness.

Panel B. The association between attention disengagement and sensory hypo-responsiveness did not attain statistical significance among autistic participants.

Panel C. Slow autistic attention-disengagers showed more enhanced perception.

Panel D. Slow autistic attention-disengagers exhibited more sensory interests, repetitions, and seeking behaviours.

Panel E. There did not appear to be any association between the speed of attention disengagement and cognitive ability.

Novelty Preference

Novelty preference trended towards differing across diagnostic groups, Wilcoxon p=.07, δ=−.29, 95% CI: [−.55, .01], with autistic participants exhibiting seemingly – but not significantly – lower visual preference to the novel stimulus in the 2000 ms following its onset (Figure 4); the Bayes Factor of 1.17, obtained from a ranked Bayesian t-test, suggested no substantial evidence for or against statistical effects.

Figure 4.

Visual preference to the novel target stimulus in the 2000 ms before and following its onset (onset marked by the red line at 10000 ms). A moving average window of 100 ms was used to generate the visual preference time sequence data. Error bars represent the standard error of the mean. The time window used for statistical analyses of visual preference is highlighted in grey.

No associations between novelty preferences and either cognitive ability or caregiver-reported sensory responsiveness attained significance (Table 2, Supplementary Table 4).

Springy Attention

Novelty Preference Task

Inspection of the time course of visual preferences in the novelty preference task (as shown in Figure 4) suggested to us that autistic children might not be orienting less to the novel stimulus immediately after its onset, but that they might be switching their attention back to the previously-familiarized stimulus more quickly: i.e., displaying “springy attention.” To explore this possibility, as described in the “Methods,” we quantified “springy attention” as visual preference to the familiarized stimulus 500–2000 ms after first fixation on the novel stimulus (Figure 5). Diagnostic group differences approached, but did not reach, significance, Wilcoxon p=.10, δ=−.27, 95% CI: [−.05, .54]; however, a ranked Bayesian test yielded a Bayes Factor of 0.86, indicating no evidence of effects.

Figure 5.

Visual preference to the novelty preference task’s familiarized stimulus in the period following fixation to the novel target (“springy attention”). The visual preference analysis window (500–2000 ms after fixation to target) is highlighted in grey. A moving average window of 100 ms was used to generate the visual preference time sequence data; error bars depict standard errors.

After correction for multiple comparisons, there were no statistically significant associations between “springy attention” in the novelty preference task and sensory responsiveness or cognitive ability (Table 2, Supplementary Table 4).

Gap-Overlap Task

However, we then conducted a similar “springy attention” analysis on gap-overlap data, examining visual preference to the central stimulus 500–1000 ms after first fixation on the peripheral target. A robust effect of task condition, Χ²=95.29, p<.0001, r_rb=−.98, 95% CI: [–1.00, −.95], with a corresponding ranked Bayesian BF₁₀=9.14, was driven by greater visual preference to the central AOI in the overlap condition than the gap condition, reflecting the continued presence of the central stimulus in the overlap condition, versus its absence in the gap condition (Figure 6). There was also an effect of diagnostic group, with autistic participants being more likely to return to the central AOI, Χ²=5.75, p=.02, δ=.13, 95% CI: [−.05, .31], though there was no evidence of this effect in comparisons of Bayesian ranked linear models, BF₁₀=0.99. Frequentist ordinal analyses found no interaction between diagnostic group and task condition, Χ²=0.17, p=.68, and Bayesian ranked analyses found no evidence of such an interaction either, BF₁₀=0.51; nevertheless, follow-up analyses (corrected for two comparisons) suggested that diagnostic groups differed only in the overlap condition, Wilcoxon corrected p=.047, δ=.33, 95% CI: [.05, .56], and not in the gap condition, p=.47, δ=.06, 95% CI: [−.09, .20]. Ranked Bayesian t-tests did not find evidence of group differences in either the overlap, BF₁₀=1.03, or gap, BF₁₀=0.40, conditions.

Figure 6.

Visual preference to the gap-overlap task’s central stimulus in the 1000 ms following first fixation to the peripheral target. The overlap condition is represented by red dashed lines and the gap condition by black dashed lines; diagnostic groups (autistic in gold, non-autistic in blue) are indicated by the colour of the standard error bars. The analysis window (500–1000 ms after fixation to target) is highlighted in grey.

A moving average window of 100 ms was used to generate the visual preference time sequence data. As such, there was no time window after 1000 ms for the data to connect to, making the sequence appear to end early.

Exploratory associations suggested “sticky attention” in the gap-overlap task was not statistically related to metrics examined from novelty preference task (Supplementary Tables 5–6).

Interestingly, among autistic participants, correlations between sensory/cognitive variables and “springy attention,” defined as visual preference to the central target in the overlap condition 500–1000 ms after first fixation on the peripheral target (Figure 7), showed nearly the opposite pattern of statistical significance as correlations with “sticky attention” (Figure 3). Autistic participants displaying more “springy attention” reportedly exhibited more sensory hypo-responsiveness, Spearman’s ρ=.50, corrected p=.02 (Figure 7B), and were assessed as displaying lower cognitive abilities, Spearman’s ρ=−.68, corrected p<.0001 (Figure 7E). There were no significant associations with hyper-responsiveness, sensory interests/seeking, or enhanced perception, all p≥.64 (Table 2).

Figure 7.

Associations between sensory/cognitive variables and visual preference, in the overlap condition of the gap-overlap task, towards the central stimulus in the 500–1000 ms following first fixation to the peripheral target. Note that these plots depict linear slopes, whereas the correlations reported in the main text use ordinal slopes that are not unduly influenced by outliers.

Panel A. There was no apparent association between sensory hyper-responsiveness and tendency to return to exploring the central target (i.e., ‘springy attention’).

Panel B. Autistic participants who showed more tendency to return to exploring the central target exhibited more sensory hypo-responsiveness.

Panel C. There was no apparent association between enhanced perception and tendency to return to exploring the central target.

Panel D. There was no apparent association between sensory interests/seeking and tendency to return to exploring the central target.

Panel E. Autistic participants who showed more tendency to return to exploring the central target were assessed as having lower cognitive abilities.

Discussion

The present study describes two forms of atypical attention in young autistic children: “sticky attention,” the well-studied tendency to disengage attention slowly from stimuli, and what we describe as “springy attention,” the tendency to return to a previous focus of attention after briefly orienting to a newly-presented stimulus. These attentional patterns could be regarded as distinct forms of hyper-focus in autism, which might help begin to clarify some of the conceptual confusion surrounding hyper-focus, monotropism, and how they specifically manifest and relate to one another.

In contrast to some prior studies (Sacrey et al., 2014), we did not observe heightened “sticky attention” in the autistic group. However, among autistic participants, we observed robust associations between “sticky attention” and heightened caregiver-reported sensory hyper-responsiveness, sensory interests/seeking, and enhanced perception. Meanwhile, we observed a nonsignificant trend towards diminished novelty preference in autism, and though the Bayesian analysis did not suggest it provided evidence to suggest this effect was robust, this prompted us to examine whether it might reflect a “springy attention” pattern rather than failure to orient to novelty. When we then returned to the gap-overlap task, exploring whether similar “springy attention” was evident there, group differences in “springy attention” achieved statistical significance in frequentist analyses, as did associations between autistic participants’ “springy attention” and lower cognitive abilities as well as greater sensory hypo-responsiveness.

“Sticky Attention”

Our failure to find group differences in the speed of attention disengagement was contrary to our hypotheses, and inconsistent with some prior studies finding that autistic participants disengage attention more slowly than typically-developing controls (reviewed by Sacrey et al., 2014). However, null findings in this literature appear to be more common when intervals between stimulus onsets are long, perhaps ≥800 ms (O. Landry & Parker, 2013; Sacrey et al., 2014). In this study, the interval between the onset of the central and peripheral stimuli was long, lasted either 1000 ms (overlap condition) or 1500 ms (gap condition), such that autistic participants may have been more prepared to saccade to the peripheral stimuli, eliminating any group difference.

Nevertheless, in this study, slower attention disengagement in autistic participants was associated with caregiver reports of greater sensory hyper-responsiveness, enhanced perception of background stimuli, and sensory interests/seeking, which was broadly consistent with our predictions. It was also consistent with reports of slower attention disengagement in individuals exhibiting more sensory seeking (Baranek et al., 2018; Sabatos-Devito et al., 2016), but inconsistent with a report that faster disengagement is associated with hyper-responsiveness (Sabatos-Devito et al., 2016). The report linking faster disengagement to hyper-responsiveness used highly salient central stimuli and very non-salient peripheral stimuli (Sabatos-Devito et al., 2016); so, fast orienting to the peripheral stimulus in that study might have required heightened responsiveness to subtle changes, whereas the present study’s peripheral stimuli were much more salient.

In the context of this study, slower disengagement from the central stimulus in the overlap condition may have been one manifestation of monotropic/hyper-focused attention, especially if one recalls that hyper-focus generally appears to coexist with heightened vulnerability to distraction and attention capture (Dwyer et al., 2024). Although this study’s gap-overlap task stimuli are intended to be non-aversive, people who generally struggle to disengage the focus of their attention from stimuli capturing their attention, might well have more overwhelming and negative experiences of any stimulus that is aversive, leading to sensory hyper-responsiveness. Other stimuli might be experienced as more pleasant, in which case reduced attention disengagement might manifest as persistent and monotropic interests in and seeking of sensory stimuli, that is, sensory interests, repetitions, and seeking. Thus, the present results appear consistent with the idea that a monotropic focus style, reflected by a reduction in attention disengagement, is related to some of the patterns of sensory differences experienced by many autistic people.

Finally, this study’s finding that enhanced perception is related to speed of attention disengagement might suggest that people who are slow to disengage from background stimuli may dwell upon, and behaviourally respond to, stimuli that most would quickly pass over and ignore. This suggestion that attention, as opposed to more basic differences in sensory acuity, contributes to enhanced perception appears consistent with the prior finding that enhanced perception is not related to sensory detection thresholds (Dwyer, Takarae, et al., 2022). It is also consistent with prior research suggesting the construct of enhanced perception, as operationalized in the SEQ, is closely related to sensory repetitions and seeking (Dwyer, 2023).

“Springy Attention”

In this study, we expected to find reduced preference for novel stimuli in autistic participants, relative to non-autistics. We did observe a strongly-trending, but nonsignificant, effect in this direction. However, interestingly, inspecting the time-course of participants’ looking towards the stimuli suggested these trends were not caused by reduced orienting towards the initial appearance of the novel stimulus, but rather a reduced tendency to maintain attention towards it. In a follow-up analysis to confirm that this tendency was driven by autistic participants returning to the familiarized stimulus after initially fixating the novel stimulus, we again found a trending but nonsignificant group difference. This tendency to return to one’s previous focus of attention could be called “springy attention.”

Admittedly, subsequent analyses suggest, in a Bayesian framework, that the results did not provide substantial evidence for or against the existence of effects. This may reflect low numbers of participants and power, perhaps due to the high-contrast nature of the black-and-white stimuli from the novelty preference task; these might have sometimes been experienced as aversive, decreasing looking and reducing the number of participants with valid data.

Regardless, to further explore this attention pattern, we analyzed whether autistic participants in the overlap condition of the gap-overlap task (in which the old central stimulus remained onscreen alongside the new peripheral target), were more likely to return to the central stimulus. We found that autistic participants were, indeed, significantly more likely to display this more “springy” pattern, at least per our original frequentist analyses. This “springy attention” metric was not statistically related to “springy attention” in the novelty preference task, suggesting that whether and to what extent stimuli attract re-engagement of attention is likely to vary depending on factors such as stimulus type and duration of prior exposure.

However, within the gap-overlap task in the present study, one can question what causes this pattern of “springy attention.” The peripheral and central stimuli are not identical; the central stimuli were abstract, high-contrast, black-and-white shapes, while the peripheral stimuli were colourful pictures of stuffed animals, which sometimes had anthropomorphized characteristics (e.g., smiling facial expressions) that might arouse social interest. Therefore, the autistic participants’ “sticky attention” could be interpreted as simply a response to these stimulus properties. While some studies suggest that young autistic children are more attracted than their non-autistic peers to stimulus salience (Amso et al., 2014; Venker et al., 2021), many autistic individuals complain of overly-bright and aversive visual stimuli (Jones et al., 2003; MacLennan et al., 2022; Smith & Sharp, 2013). Perhaps even more tellingly, autistic people often display reduced attention towards social stimuli from young ages (Falck-Ytter et al., 2022; Hedger et al., 2020), so one might argue that the group differences in “springy attention” we observed in the gap-overlap task can be most parsimoniously explained as a consequence of on-average reduced attention towards stimuli that may have social relevance.

However, while social attention may well have contributed some variance towards our observations of “springy attention” in the gap-overlap task, this explanation does not apply to the trends towards group differences in “springy attention” from the novelty preference task, as all its stimuli were similar, black-and-white, high-contrast figures. Thus, considering both of this study’s tasks together, it seems possible that many autistic children may simply be more likely to return to their previous focuses of attention after having briefly disengaged from them.

One important question about “springy attention” is whether it has analogues in prior autism and attention research. Interestingly, humans often display inhibition of return – a tendency to be slower in shifting attention and responding to stimuli when they are presented in a location one recently attended to (Klein, 2000; Posner et al., 1985) – which appears superficially similar to “springy attention,” insofar as both are concerned with the speed of returning to a location after previously attending to it. However, inhibition of return is generally studied in paradigms wherein a cue attracts attention before a different target onsets in that location (Klein, 2000). Indeed, temporal overlap between cues and targets can result in facilitation, instead of inhibition, of reaction times (Maruff et al., 1999; McConnell, 2003), but these facilitation effects were obtained when participants were presented with a novel target in the same location as the cue. In the present study’s passive paradigms, “springy attention” is the tendency to, when a new stimulus appears, glance away from but return to a stimulus that was already being continuously displayed in a location. This process appears quite distinct, and may reflect sustained attention, briefly interrupted, to the continuously-displayed stimulus (Rose et al., 2017). Some prior research in tasks with explicit instructions and/or multi-minute dynamic stimuli suggests sustained attention in autism is diminished relative to typical development (Chien et al., 2015; Vivanti et al., 2017; cf. Johnson et al., 2007), but this is quite distinct from returning attention to an old stimulus after a novel stimulus appears. Moreover, autistic children’s time looking at stimuli may depend on interest (Sasson et al., 2011), consistent with the monotropism account. Notably, autistic children often explored stimuli longer in a task wherein trials ended, and new trials commenced, when participants looked away for ≥1 s (Sacrey et al., 2023). This appears consistent with the idea that autistic people can be content to continue exploring old stimuli for longer durations than is typical.

Moreover, a result that is potentially more important, and statistically more robust, than any group differences in “springy attention” came from our exploration of correlates of “springy attention.” Autistic participants who displayed a “springier” attention phenotype in the gap-overlap task – more often returning to the central stimulus in overlap trials – were more likely to have lower cognitive abilities per the Mullen Scales. Though additional research using tasks designed to measure “springy attention” is needed to validate this interpretation of our findings, this association suggests that “springy attention” style could play an important role in development. One heuristic approach towards understanding autistic development suggests that young autistic children often display relatively less “accommodation,” or modification of cognitive schemas based on new information, and relatively more continued “assimilation” of information and explorations that can be fitted into existing schemas (Vivanti et al., 2022). It is possible that heightened “springy attention” in autism – persistent returning to previous focuses of attention – could be a manifestation of this assimilation-heavy style, potentially leading many autistic children to miss key opportunities for learning from novelty, especially in environments optimized for neurotypical children’s learning styles. This might account for the strong association between autistic participants’ cognitive abilities and gap-overlap “springy attention”: participants who do not readily engage with and learn from novelty might already have missed many opportunities for such learning by ages 2–4, this study’s age range, reducing their performance on cognitive assessments.

These results should not be taken as “proof” for a monotropism account of autism, and this paradigm notably did not directly manipulate interest in stimuli. However, some individuals’ particularly strong tendency to focus attention narrowly and persistently, instead of exploring novelty, appears to be a form of hyper-focus, consistent with a monotropism account. While this study links this form of hyper-focus to reduced performance on cognitive assessments, this does not detract from other research suggesting hyper-focus can be a strength in some contexts (Rapaport et al., 2023; Russell et al., 2019). Moreover, it is unclear whether this association, even if replicated by prior research, is obligatory, or whether adaptations to support autistic development and learning styles could result in different outcomes (Vivanti et al., 2022).

Intriguingly, in this study, we also found that more “springy attention” in the gap-overlap task was associated with greater sensory hypo-responsiveness in autistic participants. This might suggest that autistic children who persistently return to their previous focus of attention are less likely to engage with and respond to stimuli which happen to fall outside the direction of their current attentional focus. The stimuli falling outside this focus would be ignored, and consequently, the individual might appear hypo-responsive to them.

Limitations

While this study features a well-characterized sample, using multiple eye-tracking tasks and novel analyses to probe distinct forms of attention in autism, it also has some limitations, including disruptions to collection of certain measures due to changing study plans and the COVID-19 pandemic. In particular, due to the lack of novelty preference and SEQ data, our achieved power to detect correlations of ρ=.40 between novelty preference and sensory reactivity in autistic participants was only 43–47%, creating a high risk of false negative results in that analysis.

Another major limitation of this study is its cross-sectional nature, which makes the causal direction of relationships unclear. For example, although the discussion emphasizes how differences in attention could shape sensory experiences, one could alternatively emphasize how atypical sensory experiences might lead young people to change how they allocate their attention, in order to avoid aversive sensory experiences and seek pleasurable ones. Although this study’s data came from a longitudinal project, the COVID-19 pandemic unfortunately disrupted data collection from subsequent visits even more heavily than at the time-point described here. Thus, exploring the longitudinal impacts of these attentional phenotypes, and particularly of “springy attention,” remains an important direction for future research.

A second limitation is the design of the gap-overlap task, which was intended for examining “sticky attention” rather than “springy attention.” As a result, it is unclear how much of the “springy attention” observed specifically in the gap-overlap task (rather than in the novelty preference task) reflects the influence of stimulus characteristics such as salience and social relevance, which differed between central and peripheral stimuli, and how much of it reflects an inherent tendency to return to previous focuses of attention. Future studies examining “springy attention” should match stimulus characteristics to prevent such ambiguity.

The present study sample is predominantly male, to a degree greater than expected based on the estimated male:female ratio in autism (Loomes et al., 2017). Unfortunately, this prevents examination of any possible sex differences in these early attentional phenotypes.

As is common in studies with young children, there was a high rate of unusable data, especially in the autistic group. Notably, ADI-R communication and repetitive behaviour scores were higher among autistic participants who did not provide usable gap-overlap task data than those who did provide usable data. However, groups did not statistically differ in sensory or cognitive variables that were the focus of this study’s correlation analyses.

Summary

This study describes distinct forms of atypical attention, and presumably hyper-focus, in young autistic children: “sticky attention,” which has been widely studied, and “springy attention,” a tendency to return to previous focuses of attention after briefly looking away. We did not observe group differences in “sticky attention,” which might reflect the slow timing of stimuli in this study, but we found heightened “springy attention” in autistic people in a gap-overlap task, and more “springy attention” at a trend level in a novelty preference task. “Sticky attention” in autistic participants was robustly associated with caregiver reports of sensory hyper-responsiveness, enhanced perception, and sensory interests/seeking, suggesting slow disengagement of attention from stimuli might result in more intense perceptual experiences, with either negative or positive valences depending on stimulus aversiveness. Meanwhile, “springy attention” in young autistic children appeared to be associated with sensory hypo-responsiveness, suggesting that a persistent attentional focus on certain stimuli might lead to other sensory stimuli being ignored. Crucially, we observed a robust association between “springy attention” and lower cognitive abilities in autistic children, suggesting that this persistent focus might prevent opportunities for learning and have cascading effects on development. Future research with longitudinal data would be needed to confirm the directionality of this effect and of other associations observed in the present study. Nevertheless, this study findings indicate that “sticky attention” and “springy attention” are each related to important factors in the experiences of young autistic children, emphasizing the importance of attention, hyper-focus, and monotropism in shaping autistic experiences.

Data Availability

Data and stimuli will be made available on reasonable request to the corresponding and senior authors.