Increased observation of predictable visual stimuli in children with potential autism spectrum disorder

Abstract

Children with autism spectrum disorder (ASD) often exhibit social communication impairments and restricted, repetitive behaviors (RRB). Previous studies have shown that children with ASD prefer observing repetitive movements over random movements, reflecting RRB symptoms, but the developmental timeline of this preference remains unclear. New evidence suggests that children with ASD may develop predictive processing abilities for repeated behaviors, providing insight into how they recognize and respond to predictable patterns. This study employed a preferential-looking paradigm to examine whether children with potential ASD demonstrated longer observation durations for predictable movements compared to typically developing (TD) children. Participants were presented with pairs of stimuli featuring predictable and unpredictable movements, which they freely observed side-by-side. Results showed that children with potential ASD spent significantly more time observing predictable movements, particularly during the latter part of the stimulus presentation. These findings suggest that a gradual increase in attention to predictable movements may reflect difficulties in learning cause-and-effect relationships between movement trajectories and the anticipation of complete shapes. This study highlights the potential utility of predictable movement stimuli as a behavioral marker for early ASD screening. It underscores the essential need for further research into predictive processing in children with ASD.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-89171-1.

Introduction

Children with autism spectrum disorder (ASD) often exhibit varying degrees of restricted and repetitive behaviors (RRB) and social communication impairments. To identify these ASD symptoms in early childhood, researchers commonly use eye-tracking techniques, which are regarded as potential biomarkers¹. For example, Pierce et al.¹ reported that adults with ASD prefer geometric stimuli over social stimuli compared to typically developing (TD) adults and adults with other developmental disabilities. Wen et al.² extended these findings, showing that by around 2 years of age, children with ASD prefer observing non-social stimuli over social stimuli, which may contribute to delays in language development. Similarly, other studies^3,4 have shown that children with ASD prefer observing non-social stimuli—such as geometric shapes, toys, and color patterns—over social stimuli, such as human interactions. This preference aligns with the DSM-5-TR diagnostic criteria for ASD, which describe persistent deficits in social communication and interaction across multiple contexts⁵.

However, focusing solely on preferences for social versus non-social stimuli, may overlook other critical ASD symptoms, mainly restricted, repetitive, and/or sensory behaviors. RRBs involve stereotyped and repetitive patterns, such as a preference for sameness, inflexible adherence to routines, or ritualized behavior patterns⁵. Compared to social communication deficits, RRB symptoms are less frequently measured using eye-tracking methods. For example, Robain et al.⁶ found no significant differences in social orientation ratios between children with ASD and TD children when viewing non-social stimuli. Since the DSM-5-TR⁵ requires the presence of both social communication impairments and RRB symptoms for an ASD diagnosis, it is essential to explore how non-social stimuli influence the expression of RRB symptoms more closely.

Previous studies have attempted to address this gap. Gong et al.⁷ examined RRB symptoms by presenting children with ASD (approximately 6 years old) and age-matched TD children with social and non-social stimuli that had identical speeds and repetitive movement patterns. Their findings indicated that children with ASD spent more time observing non-social repetitive movements than TD children. Furthermore, younger children with ASD were more likely to focus on non-social repetitive movement stimuli compared to older children with ASD. Similarly, Wang et al.⁸ investigated preferences for repetitive movements in 3-year-old children with ASD and age-matched TD children. In their study, two identical non-social cartoon stimuli moved in repetitive or random patterns with the same speed and spatial parameters. Using a forced-choice, preferential-looking paradigm, Wang et al.⁸ found that children with ASD demonstrated a clear preference for repetitive movements, as evidenced by longer-looking times. In contrast, TD children exhibited no such preference. Additionally, the preference for repetitive movements in children with ASD increased during the latter part of the stimulus presentation. Participants with severely restricted interests showed even higher observation ratios for repetitive movements. These findings^7,8 suggest that children with ASD spend more time observing non-social repetitive movements compared to other types of moving stimuli, which is consistent with the RRB characteristic of ASD. However, it remains unclear how children with ASD differentiate between repetitive and random movements. For example, the stimuli in the study by Wang et al.’s⁸ included black dotted lines indicating the movement paths, which may have helped participants detect repetitive movement more easily. To address this limitation, the current study utilized geometric shape stimuli that traced their surroundings either with a single-stroke sketch or a multi-stroke sketch, making it impossible to identify repetitive movements until the motion was complete. Furthermore, it remains essential to investigate whether children with ASD can predict repetitive movement patterns.

As reviewed by Cannon et al.⁹, the ability of children with ASD to predict stimuli or responses varies depending on the context. Specifically, children with ASD find learning predictable cause-and-effect pairings more challenging than engaging in predictive processing of repetitive behaviors⁹. Goris et al.¹⁰ reported that TD children with high autistic traits preferred predictable tone stimuli (e.g., la-la-si-si-si-si-si) over unpredictable stimuli (e.g., la-fa-do-sol-re-si-mi), showing a positive correlation between the selection and aesthetic grading of predictable stimuli and the participants’ autistic traits. In an eye-tracking study focusing on preferences for predictable stimuli, Vivanti et al.¹¹ compared the number of fixations on repeated and novel visual stimuli presented side by side. They found that children with ASD decreased fixations on novel visual stimuli with each additional presentation, whereas the comparison groups exhibited an increase in fixations. Additionally, children with ASD required more presentations of stimulus pairs to habituate to the repeated stimuli than other children. These findings suggest that children with ASD do not exhibit an attentional preference for novel, unpredictable stimuli when presented alongside repeated, predictable stimuli. Tan et al.¹² also reported that children with ASD developed anticipatory responses to goal-directed actions more slowly than TD children. These results indicate that children with ASD may prefer predictable stimuli because they are less sensitive to unpredictable stimuli^11,12.

Children with ASD likely prefer observing predictable stimuli over unpredictable ones^10–12, similar to their preference for repetitive stimuli^7,8. The RRB symptoms observed in ASD may stem from the development of predictive processing in response to repetitive behaviors⁹. Moreover, because children with ASD are less sensitive to unpredictable stimuli^11,12, these findings provide insights into why children with ASD may initially observe repetitive and random stimuli equally but shift their focus to repetitive stimuli in the latter part of simultaneous presentations⁸. If a gradually increasing preference for predictive stimuli can be observed in children with ASD, the results would provide potential biomarkers for detecting autistic traits based on RRB symptoms whereas previous studies^1–4 have mainly focused on the social communication deficits. By comparing preferences for predictable versus unpredictable movements of non-social stimuli, without presenting social consequences or verbal interactions, the results could be applicable to children with ASD worldwide, including those with limited verbal abilities.

This study examined whether children with potential ASD would exhibit longer observation durations for predictable movement stimuli than TD children, particularly during the latter part of the stimulus presentation. Participants were instructed to freely observe pairs of predictable and unpredictable movement stimuli presented side by side using a preferential-looking paradigm. I hypothesized that children with potential ASD will observe both types of stimuli equally in the first half of the presentation but will gradually shift their eye movements toward predictable stimuli in the second half^8,11,12. In contrast, TD children were not expected to change their observation patterns across time or stimuli. Additionally, I anticipate that participants will not exhibit significant group differences or interactions in their observation of non-social stimuli⁶, showing similar observation patterns for each stimulus during the presentation. Thus, the preference for predictable stimuli is expected to manifest not as a difference between groups but as a change in how each group observes the stimuli over time.

Methods

Participants

I conducted a power analysis based on the effect size reported in previous study’s time course analyses⁸ to determine the required sample size. The analysis indicated that a minimum of 14 participants would be required to detect a partial η² of 0.24 or higher with two-way mixed factorial analyses of variance (ANOVAs) at p = 0.05 and a power of 95%. To meet this requirement, 31 children (26 boys and 5 girls; mean age = 3.32 years; age range = 1.50–5.83 years) participated in this study. All children and their parents were Japanese citizens, and none had experienced an imminent preterm delivery. None of the participants had been diagnosed with any neurodevelopmental disabilities. Moreover, there was no familial history of ASD. Most parents did not report concerns about their children’s language or communication abilities. However, 10 children exhibited language delays and problematic behaviors during routine medical checkups at 18 or 36 months. Additionally, nursery teachers expressed concerns regarding some children’s communication difficulties with peers. Teachers also assessed visual abilities, reporting that no children experienced trouble focusing on objects during school activities.

Children were assessed with three questionnaires: the Modified Checklist for Autism in Toddlers, Japanese Version (M-CHAT)¹³, the Social Responsiveness Scale, Second Edition, Japanese Version (SRS-2)¹⁴, and the Japanese MacArthur Communicative Development Inventory (J-CDI)¹⁵. The M-CHAT is a parent-report checklist comprising 23 yes/no questions designed to detect ASD in children aged 18–24 months. A score of three or more failures on the checklist is the cutoff for ASD risk. For children aged 24 months or older, parents were asked to recall and evaluate their children’s behavior at the 18-month checkup using the checklist.

The SRS-2 is a 65-item parent-report rating scale used to assess autistic traits in children aged 2.5–4.5 years and toddlers aged 4–18 years. Each item is rated on a 4-point Likert scale, ranging from “not true” to “almost always true.” The scale is divided into five subscales: social awareness, social cognition, social communication, social motivation, restricted interests, and repetitive behavior (RRB). A T-score of 60 or higher on each subscale is the cutoff for significant traits. Raw scores were converted into T-scores using child-specific norms for children aged 30 months or younger.

The J-CDI is a 448-item parent-report questionnaire assessing language development in children aged 6–18 months. It evaluates listening comprehension, expressive vocabulary, comprehensive vocabulary, and gestures. For children aged 18 months or older, parents were asked to recall and score their children’s abilities at the 18-month checkup.

Based on their SRS-2 T-scores, participants were divided into two groups. Those with T-scores higher than 60 on more than one subscale were assigned to the potential ASD group (n = 19), while the remainder were classified as typically developing children (TDC, n = 12) group. All 10 children with concerns at their medical checkups were included in the potential ASD group. Three participants were excluded from the statistical analyses: two from the children with potential ASD and one from the TDC group. These participants were excluded because their total observation time for each stimulus pair was less than 2 s during the first and second halves of the presentation combined. One included participant observed the stimuli for less than 1 s in the 1st half (0.84 s) but observed for more than 1 s in the 2nd half (1.30 s), resulting in a total observation duration of 2.14 s, meeting the inclusion criteria.

Table 1 presents the results of the parent-report questionnaires for all participants. Independent t-tests indicated no significant differences between the groups in chronological age, M-CHAT scores, or J-CDI scores, except for comprehensive vocabulary. Participants in the potential ASD group had significantly higher scores on all the SRS-2 subscales than the TDC group. They showed lower comprehensive vocabulary scores on the J-CDI. The SRS-2 T-scores for the potential ASD group revealed mild autistic traits in the full-scale and social communication and interaction subscale scores. Additionally, restricted interests and repetitive behavior subscale scores indicated moderate autistic traits.

Table 1.

Parent-reported Questionnaire results for all participants.

Group	TDC		Potential ASD		p-value
Sex	m = 9, f = 2		m = 13, f = 4		–
Age	3.34	(1.01–5.11)	3.47	(1.06–5.10)	0.42
M-CHAT	1.18	(0–3)	1.29	(0–4)	0.41
SRS2: awareness	49.73	(37–58)	56.82	(40–76)	0.003 ***
SRS2: cognition	47.45	(33–57)	56.88	(42–71)	0.00 ***
SRS2: communication	48.73	(41–59)	63.71	(46–90)	0.00 ***
SRS2: motivation	49.36	(35–59)	56.47	(40–79)	0.04 *
SRS2: social communication and interaction	45.55	(38–56)	61.29	(46–85)	0.00 ***
SRS2: restricted interests and repetitive behavior	48.09	(37–57)	67.12	(37–90)	0.00 ***
SRS2: full scale	47.91	(37–57)	63.53	(49–90)	0.00 ***
J-CDI: listening comprehension	26.45	(18–28)	23.59	(0–28)	0.14
J-CDI: expressive vocabulary	343.18	(75–448)	292.59	(0-448)	0.20
J-CDI: comprehensive vocabulary	412.00	(88–448)	322.29	(2-448)	0.06
J-CDI: gesture	55.27	(30–64)	48.18	(10–64)	0.10

M-CHAT¹³ Japanese version of the Modified Checklist for Autism in Toddlers. SRS2¹⁴ Social Responsiveness Scale Second Edition. J-CDI¹⁵ Japanese MacArthur Communicative Development Inventory. Scores in italics indicate values exceeding the cutoff scores. Scores in parentheses represent the range of these values. P-values reflect the results of independent t-test comparing the TDC and potential ASD groups for each measurement.

The protocol for this study was approved by the Waseda University Institutional Review Board (IRB; approval number: 2022-513). After receiving IRB approval, participants were recruited in Tokyo. Before the study, written and oral informed consent was obtained from all participants and their parents, who signed the appropriate consent forms.

Apparatus

An eye tracker (Tobii Pro Fusion 120 Hz; Tobii Technology) was used to record participants’ eye movements. The stimuli were presented on a display (Dell P2419H, 1920 × 1080 resolution). Participants were seated approximately 70 cm from the eye tracker during the recording. A 5-point calibration procedure was performed before the stimulus presentation to ensure accurate data collection. The presentation of stimuli was controlled using a laptop computer (Dell Precision 5540, Windows 11). Eye movement coordinates were calculated using Tobii Pro Lab software (Tobii Technology). Statistical analyses, including calculating significant probabilities, effect sizes, and analysis power were conducted using IBM SPSS Statistics (version 29.0.2.0).

Video stimuli

Two types of video stimuli were developed for each of six geometric shapes (circle, triangle, square, cross, star, and octagon), categorized based on movement regularity as predictable (P-video) and unpredictable (U-video), resulting in a total of 12 stimuli. Each video lasted 10 s and was counterbalanced, with P- and U-videos of the same shape presented on opposite sides of the display. As shown in Fig. 1, the P-videos depicted a geometric shape being traced in a single-stroke, predictable motion. Simultaneously, the U-videos displayed the same geometric shape traced in a random, unpredictable sequence. Each video stimulus was presented for 10 s before transitioning to the next.

Fig. 1.

Procedure for predictable and unpredictable video stimuli. The right side shows the predictable (P) video stimuli, depicting a one-stroke sketch of a geometric shape with its line being traced. The left side shows the unpredictable (U) video stimuli, where geometric shape’s lines are traced randomly. Each video stimulus lasted 10 s with the first 5 s constituting the first half of the trial and the last 5 s constituting the second half. Areas of interest (AOIs) were defined over the white squares containing the geometric shapes, with both AOI size standardized to 430 × 240 pixels.

Procedure

Participants’ parents were first asked to complete the three questionnaires described earlier. After completing these questionnaires, the children were seated approximately 70 cm from the display. The two videos (P- and U-videos) were then shown side-by-side in a preferential-looking paradigm to compare their observations. No chinrest was used during the experiment. Calibration was conducted by instructing participants to observe nine red dots on the display for calibration. Following calibration, participants were presented with either the P- or U-video for 10 s. After each video presentation, a blank black slide was displayed for 1 s before the next stimulus appeared. Participants viewed six pairs of P- and U-videos twice, with the presentation order randomized across participants. No specific feedback was given except instructions to observe their preferred stimulus freely.

Dependent variables and data analysis

The software defined the areas of interest (AOIs) to assess observational behaviors. Data were collected using Tobii Pro Lab software with the Tobii Fixation filter. The AOIs were confined to areas containing geometric shapes, with both AOIs size were equalized in 430 × 240 pixels. During eye-tracking measurements, all participants recorded at a data quality rate of 65.00% (range = 42–93%). The data quality was deemed sufficiently high; thus, all collected data were included in the analysis. The dependent variables included: (1) mean number of fixations, representing the frequency with which participants observed the stimuli within the AOIs; (2) mean fixation duration, indicating the average time spent observing per fixation; and (3) total fixation duration, representing the total time spent observing the stimuli in a single trial. From the total fixation duration data, the ratio of screen-looking time was calculated by summing the total fixation durations for the P- and U-videos and dividing by the total stimulus presentation duration. The preference for predictable stimuli was determined by calculating the predictable movement observing duration ratio, defined as, the total fixation duration for the P-video divided by the sum of total fixation durations for the P- and U-videos. A ratio below 50% indicated no preference for predictable movements, while a ratio above 50% suggested a preference for predictable movements⁸.

A temporal course analysis⁸ was conducted to examine when preferences for visual stimuli emerged. Wang and colleagues⁸ divided their stimuli into three phases (early, middle, and late), each lasting approximately 31 s. However, since the P- and U-videos in this study were presented for only 10 s per trial, the stimuli were divided into two phases: the first and second halves of the presentation. Four dependent variables were calculated from these phases, focusing on the first and second half scores. Independent samples t-tests were performed between the TDC and potential ASD groups for each questionnaire measure. Additionally, mixed factorial analyses of variance (ANOVAs) were conducted on the number of fixations, mean fixation duration, and total fixation duration for the remaining 28 participants. The ANOVA design was a 2 (Group: TDC vs. potential ASD; between-subjects) × 2 (Time Course: first half vs. second half; within-subjects) × 2 (Video Condition: P vs. U; within-subjects) factorial design comparing, which compared eye movement patterns across the two groups. Another 2 (Group: TDC vs. potential ASD; between-subjects) × 2 (Time Course: first half vs. second half; within-subjects) ANOVA was performed for the ratios of screen-looking time and observing duration for predictable movements. Post hoc analyses were conducted as needed using Bonferroni’s multiple comparison tests. Finally, Pearson’s correlation analyses were performed to investigate the relationship between the observing duration for predictable movements and parent-reported questionnaire data, exploring how observing predictable movements impacted the participants’ profiles.

Results

Parents report

Table 1 presents the results of the parent-report questionnaires for all participants. As previously described, analyses were conducted by dividing participants into two groups based on their SRS-2 scores. Independent t-tests revealed no significant differences between the groups in chronological age, M-CHAT scores, or J-CDI scores, except in comprehensive vocabulary. Participants in the potential ASD group had significantly higher scores on all SRS-2 subscales than the TDC group and lower comprehensive vocabulary scores on the J-CDI. The SRS-2 T-scores for the potential ASD group indicated mild autistic traits in the full-scale, and social communication and interaction subscale. In addition, scores on the restricted interests and repetitive behavior subscale suggested moderate autistic traits.

Eye-movement patterns

Figure 2 illustrates the screen-looking time ratio for the two groups. The TDC group observed the screen for 51.46% (range 21.27–79.37%) in the first half and 52.92% (range 26.09–90.93%) in the second half of the stimulus presentation. Children with potential ASD observed the screen for 51.60% (range 16.71–80.01%) in the first half and 53.68% (range 10.52–97.14%) in the second half. Two-way ANOVAs revealed no significant main effects or interactions between groups and time courses, F (1, 26) = 0.10, p  = 0 0.750, η² = 0.004, power  = 0 0.061, indicating that both groups observed the screen for an equal duration during the stimulus presentation.

Fig. 2.

Screen-looking time ratios in the two groups.

Figure 3 displays the eye-movement patterns of the 28 participants in the two groups observing P- and U-videos. Figure 3a shows the mean number of fixations for both groups. The TDC group fixated on the P-video an average of 3.75 times (range 1.33–5.83) in the first half and 3.67 times (range 1.75–7.58) in the second half. Moreover, they fixated on the U-video 4.15 times (range 1.42–7.58) in the first half and 4.03 times (range 0.42–6.83) in the second half. Children with potential ASD fixated 3.48 times (range 0.92–6.75) in the first half and 4.39 times (range 0.25–7.67) in the second half on the P-video, while they fixated 3.88 times on the U-video (range 0.67–6.92) in the first half and 2.91 times (range 0.33–6.17) in the second half. A significant three-way interaction was found among the group, time course, and video condition, F (1, 26) = 5.96, p = 0.013, η² = 0.214, power = 0.726. Post hoc analyses revealed significant simple-simple main effects of predictable movements within the potential ASD group between the first and second halves of the presentation, F (1, 26) = 13.60, p = 0.001, η² = 0.343, power = 0.944, and for unpredictable movements between time courses, F (1, 26) = 6.90, p = 0.014, η² = 0.210, power = 0.715. Additionally, a significant simple-simple main effect of the video condition was observed for the second half of the stimulus presentation within the potential ASD group, F (1, 26) = 11.16, p = 0.003, η² = 0.300, power = 0.895. However, there were no significant group interactions. These findings suggested that participants in the potential ASD group fixated more on predictable movements during the second half of the presentation than the first half while they fixated more on unpredictable movements during the first half than the second half. Furthermore, within the second half, participants with potential ASD significantly fixated more on predictable movements than on unpredictable movements. In contrast, eye movement patterns in the TDC group did not differ across time courses or video conditions.

Fig. 3.

Eye-movement patterns when observing predictable and unpredictable movements. (a) Mean number of fixations for stimuli with predictable (P) and unpredictable (U) movements. (b) Mean fixation duration for P- and U-stimuli. (c) Total fixation duration per trial for P- and U-stimuli. Significance levels: p < 0.05, **p < 0.005, ***p < 0.001.

Figure 3b presents the mean fixation duration for all participants. In the TDC group, the average fixation duration on the P-video was 198.98 ms (range 88.22–353.10) in the first half and 189.89 ms (range 87.83–357.81) in the second half. For the U-video, the average fixation duration was 197.99 ms (range 97.82–378.03) in the first half and 186.91 ms (range 54.53–373.95) in the second half. In contrast, children with potential ASD had an average fixation duration of 167.66 ms (range 74.29–282.98) in the first half and 198.49 ms (range 91.09–442.90) in the second half for the P-video and 168.53 ms (range 63.59–310.32) in the first half and 151.79 ms (range 27.02–359.36) in the second half for the U-video. A significant three-way interaction was found for mean fixation duration among group, time course, and video condition factors, F (1, 26) = 4.42, p = 0.045, η² = 0.145, power = 0.526. Post hoc analyses revealed significant simple-simple main effects within the potential ASD group for predictable movements between the first and second halves of the presentation, F (1, 26) = 5.68, p = 0.025, η² = 0.179, power = 0.631. A significant simple-simple main effect for video condition within potential ASD group was found in the second half of the stimulus presentation, F (1, 26) = 5.00, p = 0.034, η² = 0.161, power = 0.576. However, no significant group interactions were observed. These findings suggest that participants in the potential ASD group had longer mean fixation durations on predictable movements in the second half of the presentation compared to the first half. They also exhibited longer mean fixation durations on predictable movements than unpredictable ones in the second half.

Figure 3c depicts the total fixation duration per trial for all participants. In the TDC group, the total fixation duration for the P-video was 0.91 s (range 0.21–1.55) in the first half and 0.90 s (range 0.22–2.06) in the second half. For the U-video, the total fixation duration was 0.95 s (range 0.23–1.93) in the first half and 1.03 s (range 0.06–2.26) in the second half. In children with potential ASD group, the total fixation duration for the P-video was 0.74 s (range 0.17–1.50) in the first half and 0.99 s (range 0.08–2.11) in the second half. For the U-video, the total fixation duration were 0.84 s (range 0.10–1.79) in the first half and 0.61 s (range 0.03–1.43) in the second half. A significant three-way interaction was found among the group, time course, and video condition factors, F (1, 26) = 9.75, p = 0.004, η² = 0.273, power = 0.852. Post hoc analyses revealed significant simple-simple main effects within the potential ASD group for predictable movements between the first and second halves of the presentation, F (1, 26) = 13.98, p = 0.001, η² = 0.350, power = 0.949, and for unpredictable movements between time courses, F (1, 26) = 4.59, p = 0.042, η² = 0.150, power = 0.541. A significant simple-simple main effect for video condition was observed in the second half of the stimulus presentation within the potential ASD group, F (1, 26) = 7.25, p = 0.011, η² = 0.224, power = 0.752. However, there were no significant group interactions. These results indicate that participants in the potential ASD group fixated longer on predictable movements during the second half of the presentation than in the first half. They also fixated less on unpredictable movements in the latter than in the first half. Moreover, during the second half of the presentation, participants with potential ASD spent more time observing predictable movements than unpredictable ones.

Figure 4 shows the observation duration ratio for predictable movements in the two groups. The TDC group observed the P- and U-videos for nearly equal durations in the first half (50.11%, range 36.34–66.46%) and the second half (51.48%, range 24.97–77.65%). Participants with potential ASD observed predictable stimuli for 48.11% (range 34.68–66.04%) in the first half and 63.69% (range 42.06–89.61%) in the second half. Two-way ANOVAs revealed a significant interaction between group and time course, F (1, 26) = 7.59, p = 0.011, η² = 0.226, power = 0.756. Post hoc analyses showed a significant simple main effect of the group during the second half of the presentation, F (1, 26) = 4.81, p = 0.037, η² = 0.156, power = 0.561, and a significant simple main effect of time course within the potential ASD group, F (1, 26) = 22.95, p = 0.000, η² = 0.469, power = 0.996. These results indicate that participants with potential ASD group fixated more on predictable stimuli during the latter part of the P- and U-video presentations compared to the first half.

Fig. 4.

Ratio of observing predictable movements across the two groups. the gray dotted line represents equal observation of both predictable and unpredictable stimuli.

Correlation analysis

Table 2 presents the correlation analyses between the observing duration ratio for predictable movements and parent-report questionnaire scores. The observing ratio for the first half showed a negative correlation with the listening comprehension (r = − 0.41, p = 0.02) and gestures (r = − 0.35, p = 0.03). In contrast, the observing ration for the second half was positively correlated with SRS-2 scores, including awareness (r = 0.42, p = 0.01), communication (r = 0.33, p = 0.05), social communication and interaction (r = 0.38, p = 0.02), restricted interests and repetitive behavior (r = 0.32, p = 0.05), and full scale scores (r = 0.37, p = 0.03). Additionally, the second-half observing ratio showed negative correlations with J-CDI scores for listening comprehension (r = − 0.45, p = 0.01), expressive vocabulary (r = − 0.49, p = 0.004), comprehensive vocabulary (r = − 0.49, p = 0.004), and gestures (r = − 0.54, p = 0.001). These results suggest that a higher observation ratio for predictable movements in the second half of the stimulus presentation is associated with higher autistic traits and lower language comprehension and gesture scores.

Table 2.

Pearson’s correlation analysis between ratios of observing predictive movements and parent-reported measurements.

Ratios of observing predictable movements	1st half		2nd half
Parents’ reports	r	p	r	p
Chronological age	0.12	0.27	− 0.20	0.16
M-CHAT	0.27	0.08	0.22	0.13
SRS2: awareness	0.10	0.31	0.42	0.01*
SRS2: cognition	− 0.08	0.34	0.28	0.08
SRS2: communication	− 0.08	0.34	0.33	0.05*
SRS2: motivation	0.09	0.32	0.21	0.14
SRS2: social communication and interaction	− 0.03	0.44	0.38	0.02*
SRS2: restricted interests and repetitive behavior	0.04	0.43	0.32	0.05*
SRS2: full scale	0.01	0.48	0.37	0.03*
J-CDI: listening comprehension	− 0.41	0.02*	− 0.45	0.01**
J-CDI: expressive vocabulary	− 0.23	0.12	− 0.49	0.004***
J-CDI: comprehensive vocabulary	− 0.29	0.06	− 0.49	0.004***
J-CDI: gesture	− 0.35	0.03*	− 0.54	0.001****

Discussion

This study examined whether children with potential ASD would exhibit longer observation durations for predictable movement stimuli than TD children, particularly during the latter part of the stimulus presentation. The initial hypothesis predicted that participants would display similar observation patterns for both stimuli during the presentation without significant group differences or interactions in their P- and U-video stimuli observations. I anticipated preferences for predictable stimuli to manifest as changes in how each group observed the stimuli over time rather than as group differences. The findings demonstrated that children with potential ASD show longer observation durations for predictable stimuli than TDC group, particularly in the latter half of the stimulus presentation. Participants were instructed to freely observe paired stimuli—a one-stroke sketch movement and a random stroke sketch movement—presented side by side using a preferential-looking paradigm. Figures 3 and 4 illustrate that TDC did not significantly alter their eye movement patterns over time. They maintained similar observation patterns across the first and last five seconds of the stimulus presentation for predictable and unpredictable movie stimuli. In contrast, children with potential ASD gradually increased fixation duration, fixation count, and the observing ratio for predictable stimuli compared to unpredictable stimuli, especially in the latter half of the presentation. These results align with the findings of Wang et al.⁸, who reported that children with ASD display a preference for predictable movements over random movements, particularly during the second half of the stimulus presentation. This study extends those findings by demonstrating that children with potential ASD who have not been clinically diagnosed also show a preference for repetitive and predictable movements over random movements. Such responsiveness to predictable stimuli could serve as an early indicator of ASD symptoms in 3-year-old children.

In Wang et al.’s⁸ study, participants with ASD showed a screen-looking time ratio of 53.29%, whereas TD children exhibited a higher ratio of 64.30%. In this study, the screen-looking time ratios for TD children and children with potential ASD were approximately 52.19% and 52.64%, respectively. These ratios indicate that both groups maintained visual attention during the presentations, with no significant differences between groups. Notably, the looking time ratio for TD children in this study was lower than those reported in previous studies^7,8. However, Robain et al.⁶ found no significant differences in looking time ratios between children with ASD and TD children when viewing non-social stimuli. Whereas previous studies^7,8 often reported longer looking time ratios for TD children than children with ASD, the current findings suggested both groups observed the stimulus pairs for comparable durations. Therefore, the observed preference among children with potential ASD did not emerge from the total time spent looking at the stimuli but rather from how they allocated their attention to predictable versus unpredictable movements.

In analyzing eye movement patterns, previous studies^6–8 have emphasized the significance for total stimuli presentation duration rather than trial-by-trail differences. However, since this study presented stimulus pairs on a trial-by-trail basis, the focus was on how eye movement patterns varied within individual trials. Based on Fig. 3, children with potential ASD fixated for approximately 0.37 s longer in total, 46.70 ms longer on average, and 0.97 times more frequently on P-video than U-video in the second half of the presentation. These differences may appear minor. However, they could represent meaningful opportunities for children to observe predictable stimuli within a 5-s window highlighting their preference for predictability. Oka and Omori (2023)¹⁶ used an eye tracker to investigate the progression of children’s communication from triadic to polynomial social relationships. Their task required participants to focus on face-looking stimuli, regardless of the accompanying auditory stimuli. Despite each trial lasting only 2 s, their trial-by-trial analyses revealed subtle differences in fixation patterns, demonstrating, that observing visual stimuli without auditory distractions can facilitate language development and enhance understanding of social communication patterns. Similarly, while the differences in fixation durations in the current study may be statistically significant, future research is still necessary to explore how such differences reflect real-life behaviors and interactions in children with or without ASD.

The current study also found group differences in the temporal changes of eye movement patterns for predictable stimuli, suggesting decreased sensitivity to unpredictable stimuli^11,12. Vivanti and colleagues¹¹ reported that children with ASD gradually reduced their fixation counts on novel, unpredictable visual stimuli and required more trials to habituate predictable stimuli. Similarly, Tan and colleagues¹² found that children with ASD need more trials than TD children to learn anticipatory responses. These findings suggest that children with ASD develop a preference for predictable stimuli over time, but this preference emerges over and diminishes more slowly than in TD children. As shown in Fig. 4, children with potential ASD increased their observation ratio for predictable movements in the second half of a trial. In contrast, TDC maintained a stable observation ratio across time. The current study extends these findings by demonstrating that children with potential ASD prefer predictable movements within a single trial. However, like previous studies^11,12, this research did compare observation ratios across trials, leaving room for further investigation.

One explanation for why children with potential ASD were more sensitive to predictable stimuli in the latter half of the presentation could be difficulties in understanding the cause-and-effect relationship between movement trajectories and their resulting shapes⁹. Cannon and colleagues reported that children with ASD experience challenges in acquiring and anticipating predictable pairings due to an inflexible use of anticipatory strategies, particularly in uncertain contexts¹². Similarly, Tan and colleagues¹² reported that children with ASD exhibited fewer fixations and shorter fixation durations on anticipatory stimuli than TDC in uncertain situations. Vivanti and colleagues¹¹ found that children with ASD did not decrease their fixations on repeated stimuli across trials when repeatedly presented with the same predictable stimulus. According to Cannon and colleagues⁹, children with ASD can learn predictive processing through repetition, which may help them respond in predictable situations. In the present study, participants were presented with pairs of predictable and unpredictable stimuli involving different geometric shapes across trials. Participants detected predictable movements by observing these stimuli in a task with an intermediate level of certainty compared to those used in the abovementioned studies. Understanding how children observe shapes with predictable and unpredictable movements may contribute to the early identification of children at risk for developing ASD, although further research is needed.

Compared with previous studies^{1,2,7,8,11,12}, this study found that a preference for predictably moving stimuli reflects not only RRB symptoms but also social communication and interaction deficits, as well as a comprehensive vocabulary deficit (see Table 1). Prior research^1,2 has shown that children with ASD prefer observing non-social stimuli, consistent with the DSM-5-TR criteria of persistent deficits in social communication and interactions across multiple contexts⁵. Wen et al.² reported that children with ASD who spent more time observing non-social stimuli exhibited higher autism severity, lower verbal abilities, and reduced adaptive skills than children who spent less time observing such stimuli. In the current study, children with ASD gradually increased their preference for predictable stimuli over time. These results could serve as a potential biomarker reflecting the DSM-5-TR⁵ criteria for RRB symptoms in ASD. Other studies^7,8,11 have demonstrated that a preference for repetitive or predictable stimuli in children with ASD is associated with RRB symptoms, adaptive behavior, or verbal abilities. For example, Wang et al.⁸, conducted correlation analyses showing that children with ASD gradually increased their observation of repetitive movements over time. In contrast, TDC did not exhibit such changes, highlighting RRB characteristics. This study’s correlation analyses revealed significant relationships between observing ratios for predictable stimuli and parent-report measures. As shown in Table 2, the observation ratios for predictable stimuli in the second half of the presentation positively correlated with several T-scores on the SRS-2 subscales, including awareness, communication, social communication and interaction, RRB, and the full scale. Conversely, the observation ratio in the first half negatively correlated with vocabulary scores. Additionally, the second-half ratio was sensitive to developing expressive and comprehensive vocabulary. These findings align with the report by Wen et al.² that children with higher autism severity and lower verbal abilities spend more time observing non-social stimuli. This study also found that children with higher autistic traits and lower expressive and comprehensive vocabulary were more likely to prefer predictable stimuli in the latter part of the presentation. Children with potential ASD can detect predictable movements in non-social stimuli, suggesting the method’s applicability to children with limited verbal skills. This approach does not require social consequences or verbal interactions. Therefore, it could be particularly effective for assessing children with ASD who exhibit reduced verbal communication. Although further research is needed, the observed gradual preference for predictably moving stimuli may represent an ASD trait encompassing both social communication deficits and RRB symptoms. Moreover, it may provide valuable insights into developing expressive and comprehensive vocabulary skills, further supporting the utility of predictable stimuli as a potential diagnostic and developmental tool for early ASD screening.

This study has several limitations. Firstly, the participants had not been clinically diagnosed with ASD. It is essential to investigate whether children with clinically confirmed ASD also demonstrate a gradual preference for predictable stimuli. While the participants in this study were assessed using the M-CHAT and SRS-2, autism severity was not evaluated using the ADI-R or ADOS-2, as the necessary licenses were unavailable. Furthermore, most participants were older than two years, making the M-CHAT less suitable for their age group. Future research should incorporate assessments of autism severity and traits using the ADI-R and ADOS-2. This would allow for comparisons between children with clinically diagnosed ASD and those with broad autism phenotypes who do not meet the diagnostic criteria for autism. Secondly, children with potential ASD showed moderate social communication and interaction subscale scores and mild RRB. These findings may be related to parents experiencing greater difficulty communicating with their children. Table 1 shows that children with potential ASD had a smaller comprehensive vocabulary than TD children, whereas both groups displayed similar levels of listening comprehension, expressive vocabulary, and gestures. During the 36-month development checkup in Japan, children with potential ASD often exhibit vocabulary deficits. Parents may also be more sensitive to language development delays than behavioral symptoms. Future research should control SCI and RRB scores to determine whether the preference for predictable movements is related to specific symptoms or their severity. Thirdly, there is a potential for recall bias in parent responses to the M-CHAT and J-CDI questionnaires. While the SRS-2 effectively distinguished autistic traits between children with potential ASD and TD groups, the M-CHAT did not show significant group differences. Most participants were older than 18 months, requiring parents to retrospectively assess their child’s behavior at an earlier age. This recall requirement might have contributed to the lack of group differences in M-CHAT scores. Additionally, J-CDI scores exhibited a ceiling effect, potentially reflecting biases related to the children’s current developmental status rather than their behavior at 18 months. Unlike the SRS-2, which is age-appropriate for the participants, no suitable questionnaire or behavioral task was used to evaluate verbal abilities specifically. Future studies should include participants within the appropriate age range for these assessments and ensure the application of developmentally appropriate tools. Additionally, demographic data were not collected in this study. Differences in demographic characteristics between the groups could have influenced the findings. Future research should collect demographic data to enhance the generalizability of the results.

Finally, future studies would benefit from recruiting children with potential ASD younger than 18 months and focusing on a narrower age range. The small sample size and wide age range of participants in this study limit the generalizability of the findings. The effects of these results may vary significantly across different developmental stages. Kurasawa et al.¹⁷. reported that the mean age of ASD diagnosis in Japan is 7.2 years (SD = 4.2) with a median of 6.0 years (range 4.0–10.0). The findings of this study may be particularly relevant for children at risk for ASD who remain underdiagnosed until age seven, offering a method to facilitate earlier ASD detection in Japan. Similar to prior studies, future research should aim to recruit approximately 20 children within a specific age range^6,8. A focused age range would help reduce recall bias, enabling parents to provide more accurate responses to questionnaires without relying on retrospective assessments of their child’s behavior. Although the ideal age for diagnosis is around three years, earlier identification and support are crucial to optimizing developmental outcomes. This study’s procedure, which involved observing pairs of stimuli without requiring verbal responses, could be adapted for younger populations. Notably, even children under 18 months old in this study completed the experiment successfully without dropping out, indicating that the method is feasible for younger participants. In Japan, most children undergo standardized developmental checkups at 18 and 36 months, with over 96% participating in the 18-month checkup¹⁸. Introducing a brief, two-minute video observation task during these checkups could help identify children at risk for ASD and support their parents. This approach could enable earlier screening and detection of ASD traits. Future research should explore developmental differences in the gradual preference for predictable stimuli by comparing children at risk for ASD at 18 and 36 months. Early diagnosis is critical for implementing effective interventions, as earlier detection increases the likelihood of leveraging brain plasticity to support development. This study’s results highlight the potential of predictable stimuli as an early screening tool, offering promise for improving early diagnosis and intervention for children at risk for ASD.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

The author was solely responsible for the conception and design of the study, the acquisition, analysis, and interpretation of data, and the drafting and revision of the manuscript.

Funding

This research was supported in part by Grants-in-Aid for Scientific Research (KAKENHI) from the Japan Society for the Promotion of Science (JSPS), Grant number 23K17293.

Data availability

Data is provided within the supplementary information files.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval

All procedures involving human participants were conducted in accordance with the ethical standards of the 1964 Helsinki Declaration and its subsequent amendments or comparable ethical standards. The study was approved by the Institutional Review Board of Waseda University (Approval No. 2022 − 513).

Consent to participate

Written and oral informed consent was obtained from all participants and their parents.