Social synchronization during joint attention in children with autism spectrum disorder

We explored the social synchronization of gaze‐shift behaviors when responding to joint attention in children with autism spectrum disorder (ASD). Forty‐one children aged 5 to 8 with ASD and 43 typically developing (TD) children watched a video to complete the response to joint attention (RJA) tasks, during which their gaze data were collected. The synchronization of gaze‐shift behaviors between children and the female model in the video was measured with the cross‐recurrence quantification analysis (CRQA). Ultimately, we discovered that children with ASD had the ability to synchronize their gaze shifts with the female model in the video during RJA tasks. Compared to the TD children, they displayed lower levels of synchronization and longer latency in this synchronized behavior. These findings provide a new avenue to deepen our understanding of the impairments of social interaction in children with ASD. Notably, the analytic method can be further applied to explore the social synchronization of numerous other social interactive behaviors in ASD.


INTRODUCTION
Social synchronization refers to an individuals' temporal coordination during social interactions (Delahaerche et al., 2012). Particularly, it includes various social communicative processes such as joint attention, imitation, turn-taking, non-verbal social communicative exchanges, affect sharing, and engagement (Tony, 2011). Notably, social synchronization has been found to play a crucial role in social development (Fitzpatrick et al., 2017b). For instance, interactional synchronization in mother-infant interactions is imperative for the development of social relations and language learning (Condon & Sander, 1974a, 1974b. Additionally, social synchronization has been found to be associated with children's social responsiveness (Fitzpatrick et al., 2018), social rapport (Bernieri et al., 1994;Lakin & Chartrand, 2003), social perception (Miles et al., 2009), cooperation (Reddish et al., 2013;Reddish et al., 2014;Wiltermuth & Heath, 2009), positive interpersonal relationships (Feldman, 2007;Hove & Risen, 2009), as well as with a better understanding of the thoughts and actions of others (Shockley et al., 2009).
During social interaction, the social cues given by the social partner often change over time. Children need to adjust their visual attention timely to synchronize with their partners' social cues. This synchrony process is essential in the social context considering that many social cues are often ephemeral and would fleet during social interactions. The incapability of paying attention to and processing these cues timely will hinder the understanding of others' attentional focus and intention, limit learning opportunity, and ultimately lead to social impairments. This may contribute to the core symptoms of autism spectrum disorder (ASD), a neurodevelopmental disorder characterized by impairments of social interaction and communication, as well as restricted, repetitive patterns of behavior and interests (American Psychiatric Association, 2013).
However, most previous studies about social attention focused on whether children with ASD pay attention to the social stimuli and the total looking time at it, paying little attention to the dynamic changes during the social interaction. While synchrony during many social interactions is essential and the deficit of social synchronization may ultimately lead to social impairments. Thus, we propose that ASD is not only seen as a disorder of the individual visual social attention abnormality but also a "misattunement" between individuals, resulting in disturbances in reciprocal interactions. Here we aimed to explore to what extent the typically developing (TD) children, as well as children with autism spectrum disorder (ASD), can synchronize their social attention with their partners' social cues during a classical screen-based responding to joint attention (RJA) task that has been widely used and recognized (Bedford et al., 2012;Billeci et al., 2016;Vivanti et al., 2017). That is, when the partner looks at the children directly, whether the children can also look at the partner; while the partner looks at the object, whether the children can also look at the object at the same time.
The RJA task was adapted from Senju and Csibra (2008). In this task, we presented children with a video depicting a female model turning toward one of the two objects, and recorded children's eye movements by an eye-tracker when watching the video to examine their gaze-following behavior. Unlike other approaches for quantifying JA, such as coding videos of children engaged in paly designed to elicit JA (Hurwitz & Watson, 2016), the eye-tracking method is much more objective, less timeconsuming, and has the potential for less examiner error. Furthermore, eye-tracking can enhance the detection of subtle shifts in visual attention and provide more precise spatial and temporal information. Besides, eye-tracking assessments of RJA are more consistent across administrations and thus potentially useful as standardized prognostic instruments (Navab et al., 2011). Considering that social interaction is essentially a continuous process, we analyzed all the trials as a whole continuous process through the use of cross-recurrence quantification analysis (CRQA) to examine the synchronization of gaze-shift behaviors between children and the female model in the video. Notably, CRQA can quantify how similarly two observed data series unfold over time (Shockley et al., 2002;Zbilut et al., 1998). This is not to mention that it also reflects the degree of synchronization as well as the level of delay in synchronization between pairs of people (Shockley et al., 2009). The application of CRQA to social interaction such as JA can provide an avenue to study whether they can look at the appropriate social stimuli at the appropriate time. The CRQA method has been employed to measure different behavioral coordinations between individuals, such as the shared postural configurations of two people engaged in a conversation (Shockley et al., 2003), and the looking behaviors of infants and their caregivers when actively playing with toys (Yu & Smith, 2013).
In this study, we applied the CRQA to measure the synchronization of gaze-shift behaviors between children and the female model in the video during the RJA tasks. Previous research has identified that difficulty with RJA became less evident with age in children with ASD (Mundy, 2016), they even exhibited the intact ability to follow gaze (Bedford et al., 2012;Falck-Ytter et al., 2015). However, effort or efficiency of RJA may remain problematic in these children (e.g., longer latency to follow the gaze) (Falck-Ytter et al., 2012). Accordingly, we expected that children with ASD in our study would demonstrate the ability to synchronize their gaze shifts with the female model in the video during RJA tasks. Notably, however, their latency in this synchronized behavior would be longer and their level of synchronization would also be lower than TD children.

METHOD Participants
A total of 84 children aged 5 to 8 participated in the study, including 41 children with ASD (37 males, 4 females) and 43 TD children (38 males, 5 females). One child with ASD and three TD children were excluded from analysis due to poor eye movement data quality (see "Data Analysis" section for details), resulting in 40 children with ASD (36 males, 4 females) and 40 TD children (36 males, 4 females) in the final sample. They were approximately 7 years old (see Table 1). We selected children at this age since it is a potentially sensitive developmental period for gaze perception among children whose more basic visual mechanisms are presumably in place (Mihalache et al., 2020). And children with ASD at this age also showed the impairments of RJA (Thorup et al., 2017). However, the behavioral evidence of atypical RJA in children beyond preschool age is not sufficient and clear. Furthermore, many previous studies about social synchronization have found that children with ASD older than 5 had difficulty in synchronizing their movements with another person (Fitzpatrick et al., 2013(Fitzpatrick et al., , 2017a(Fitzpatrick et al., , 2017bXavier et al., 2018). So here we select children at 5 to 8 years of age to examine the social synchronization of children with ASD during joint attention. Children with ASD were recruited from Peking University Sixth hospital, and had been diagnosed by two professional child psychiatrists according to DSM-5 (American Psychiatric Association, 2013). The diagnosis was confirmed using the Chinese version of the Autism Spectrum Quotient-Children's Version (AQ-Child; Auyeung et al., 2008;Li et al., 2018). The exclusion criteria for the ASD group were as follows: (a) AQ-Child score < 73 (Li et al., 2018); (b) having severe psychiatric disorders or physical diseases such as schizophrenia and bipolar disorder, which may result in complicated clinical manifestations and is hard for us to disentangle the possible confound between ASD and other psychiatric disorders; (c) having uncorrected hearing or visual impairments; (d) unable to cooperate with the eye-tracking test. The TD sample was recruited from local communities by advertisements online as well as from local primary schools in Beijing. The two groups were matched by chronological age and non-verbal IQ score (measured by Raven's Test, see Table 1). Detailed descriptions of participant characteristics can be found in Table 1.

Ethical considerations
This study was approved by the Ethical Committee in Peking University Sixth Hospital and conducted based on the 1964 Declaration of Helsinki. All parents gave written consent and their children provided oral consent before the onset of the experiment.

Stimuli
Children were asked to watch a video (lasting for 84 s) depicting a female model sitting behind a table where two identical objects were placed ( Figure 1). The video consists of two rounds, which features different objects and each round includes four responding to joint attention (RJA) tasks. At the beginning of each round, the model's face is covered by a cartoon accompanied with a voiceover, "Look!" so as to attract the child's attention. The cartoon then disappears, revealing the model's face. The model looks straight into the camera, smiling for about 5 s. Then, the four RJA tasks/trials are presented in random order. Because the exaggerated emotional expression may have an effect on responses to joint attention in children with ASD (Franchini et al., 2017), the female model maintains a neutral facial expression during the RJA trials (Navab et al., 2011). Basing on the fact that children and their caregivers often coordinate visual attention with gaze shifts (with and without head turns) and our goal in studying children's attention to gaze cues, we chose gaze shifts with/ without head turns in the RJA tasks. Each cue appears twice in each round, once to the left and once to the right (the positions are counterbalanced across trials). In each task/trial, the model shifts gaze with/without head turn toward one of the two objects after looking straight into the camera for about 300 ms. Then, the model continues to look at the attended object for 4 s and turns back to look straight at the participant for 3 s at the end of the task where a cartoon character appears beside the attended object. Because most of the participants in our study were at school-age, this RJA task may be quite easy and boring to them. The purpose of this cartoon figure is to provide positive feedback to encourage children to better cooperate with this eye-tracking test. The whole video plays continuously, there was no intertrial time between the two rounds and trials.

Procedure
After arriving at the lab, the parents sign the consent form and complete the AQ-child questionnaire. Then, the child is taken to the eye-tracking lab and is seated in the chair at a distance of 60 cm in front of a 17-inch Note: Non-verbal IQ was measured using the Raven's standard progressive matrices (Raven et al., 1983). Abbreviations: AQ-child, the Autism Spectrum Quotient-Children's Version; ASD, autism spectrum disorder; TD, typically developing.
LCD monitor (1680 Â 1050 pixels resolution). A 5-point calibration procedure is then conducted to calibrate their eye movements. During the calibration process, the child is instructed to fixate on the five red calibration points which appear sequentially in the center and four corners of the screen. The calibration is repeated to many times necessary until both eyes achieve good mapping on all five calibration points (smaller than a 2 visual angle). After the calibration, the child is asked to watch the video and no further instruction is provided. The two rounds of the video contains a total of eight RJA tasks/ trials which are played randomly. Eye movement data were recorded throughout the entire experiment.

Measures
The Autism Spectrum Quotient-Children's Version (AQ-Child; Auyeung et al., 2008) is a parent-report questionnaire, which consists of 50 items created to measure autistic traits in children between the ages of 4 and 11. Here, we used the Chinese version of the AQ-Child (Li et al., 2018) to assess each of the five domains: (1) social skill (2) attention switching (3) attention to detail (4) communication (5) imagination. Parents rated the extent to which they agree or disagree with the statements about their child on a 4-point scale (0 represented definitely agree, 1 slightly agree, 2 slightly disagree, and 3 definitely disagree). Higher scores of AQ-Child denote higher levels of autistic traits. The AQ score has been shown to have high sensitivity (96%) and specificity (90%) at a cut-off of 73 and has good test-retest reliability (92%) (Li et al., 2018). In this sample, the mean AQ score of the ASD group was 85.8 AE 8.6.

Data preprocessing
Gaze data were recorded using a Tobii T120 eye tracker and Tobii Studio software (Tobii Technology, Stockholm, Sweden). Three rectangular regions of interest (ROIs) were defined: one covered the female model's face while the other two covered the two objects (see Figure 1). Raw gaze data (i.e., sample data) were extracted from the Tobii Studio software. Notably, trials with more than 50% invalid gaze data (defined as validity codes >1 from Tobii raw data) were excluded from the analysis. To ensure the data's quality, we excluded children with fewer than six valid trials after trial rejection from further analyses.

Cross-Recurrence quantification analysis of two gaze data streams
Cross-recurrence quantification analysis (CRQA) (Xu et al., 2020) was applied to quantify coordinated attention between children and the female model, and to measure their temporal coupling with different degrees of time lags. The precise timing information for the gaze data streams of the children and the female model can be depicted graphically in what we refer to as the "crossrecurrence plot." Here, the gaze data of one child and the female model was used during RJA tasks as a reference to explain the method. As depicted in Figure 2, the horizontal dimension represents the child's gaze stream over time while the vertical dimension represents the female model's gaze stream over time. We calculated synchrony during the RJA trials (Figure 1(c),(d)). Additionally, the central diagonal line in the cross-recurrence plot provides a measure of synchrony when simultaneously looking to the same ROI. Specifically speaking, the colored pixels on the diagonal line indicate that the eyes of the child and the female model were fixated on the same object or they were directly looking at each other (when the female model was looking directly and the child was looking at the model's face). The white pixels demonstrate that the child and the female model were not fixated on the same ROI. Meanwhile, the pixels in other diagonal lines that are parallel to the central diagonal line of the plot reflect time-shifted recurrences between the gaze of the child and the female model. These pixels depict whether the child and the female model were fixated on the same ROI or if they expressed a mutual gaze but with some delay (time unit: 1/120 s based on the sample rate of the eye tracker). To quantify the patterns of synchronization in these two data streams, we generated the cross-recurrence lag profiles through computing the percentage match (or percentage recurrence, %REC) along all the diagonal lines in the cross-recurrence plot. These profiles reflect the synchronization between two data streams at a different time lag. It should be noted that the current task was not a real social interaction task and the female model could not respond to the children's gaze, thus the percentage recurrence along the lower right diagonal lines was reported.
Three statistical analyses were applied to examine whether children with ASD exhibited atypical social synchronization patterns with the female model compared to TD children. Particularly, the synchronization patterns of the ASD and the TD groups were compared with chance separately, and then with each other. Then, we assessed whether group membership could be predicted from the percentage recurrence at different time lags through using a machine learning algorithm to validate the above inferential statistical analyses. All the detailed descriptions of the methods can be found in the Supporting Information.
F I G U R E 2 An example of cross-recurrence plot. The two gaze data streams at the left and bottom represent the ROIs that the female model and the child were looking at over time, respectively. The colors at the center region represent the fact that the child and the female model were looking at the same object (the red or green pixels) or directly looking at each other (the blue pixels) at different time lags (the numbers at the x-and y-axes represent sample points, time unit: 1/120 s)

RESULTS
Synchronization between the eye movements of the children and the female model Figure 3 depicts the average percentage recurrence in the ASD and TD group pairings for each time lag, with "0" indicating simultaneous fixations to the corresponding ROIs (including mutual gaze). The curve to the right of "0" reflects that the female model's looks to an ROI that was followed by a look to the same ROI by the child within the defined lag. The black line below depicts a lag time at which the child and the female model look at the same ROI at above-chance levels. As demonstrated in Figure 3, for the ASD group, two clusters showed abovechance levels (792-5592 ms and 6425-8000 ms time lags), Z = 18.04, p < 0.001 and Z = À3.24, p = 0.001, respectively; for the TD group, one cluster showed abovechance levels (0-4508 ms time lags, Z = 20.23, p < 0.001). Therefore, the ASD group began following the female model's look to the same ROI at abovechance levels 800 ms after the model shifted her gaze or turned her head. Meanwhile, the TD group immediately followed the female model's orientation as the model shifted her gaze or turned her head.
We then compared the synchronization patterns of children with the female model between the ASD and the TD groups. The baselines of the two groups differed ( Figure 3) due to the different gaze data loss rate (when the child looked at blank spaces or looked off the screen or blinked), making the two groups incomparable. To resolve this issue, we subtracted each child's percentage recurrence at different time lags by his/her baseline which was generated through shuffling the temporal order of each child's eye-movement sequence 1000 times and calculating the average recurrence with the female model. Figure 4 illustrates the difference of the percentage recurrence between the two groups with the adjusted baseline. The black line below indicates the lag time at which the average percentage recurrence in the ASD and TD group were significantly different. The average percentage recurrence in the ASD group was lower than that of the TD group at the lag time between 0 and 2833 ms, Z = À5.58, p < 0.001.This is not to mention that the maximum recurrence between the model and children of the two groups was different as well, with 6.57% in the ASD group and 8.59% in the TD group. The lag time at which their eye movements overlap the most, is 3292 ms in the ASD group and 2025 ms in the TD group.

The classification performance of the RJA tasks
The support vector machine (SVM) was used with linear kernel (Kafai & Eshghi, 2019), which has been widely utilized in psychiatry research as a classification algorithm to evaluate the prediction performance. Overall, classification performance resulted in 60% accuracy. The sensitivity and specificity were 63% and 58%, respectively. The area under the curve (AUC) of the receiver operating characteristic curve (ROC) was 0.63. The permutation test reflected that the prediction was higher than the random level, p = 0.042.

DISCUSSION
In this study, we explored the atypical patterns of social synchronization during response to joint attention in children with ASD through utilizing the analytical method of cross-recurrence quantification analysis (CRQA). Consistent with our hypotheses, children with ASD were identified to possess the ability to synchronize their gaze shifts with the female model in the video during the RJA tasks. Compared to the TD children, however, they displayed lower levels of synchronization and longer latency in this synchronized behavior.
First, we discovered that when compared with the baselines, both groups were looking at the same ROI with the female model at above-chance levels, indicating that they were not looking randomly during the RJA tasks. This further suggests that children with ASD can respond to social cues such as gaze shift and head turn. This finding is consistent with previous studies which demonstrate that children with ASD had the ability to F I G U R E 3 Average cross-recurrence at different time lags in the ASD and TD group pairings. The baseline of each group was generated by shuffling the temporal order of each child's eye-movement sequence 1000 times and calculating the average percentage recurrence with the female model for each child, and then averaging the percentage recurrence across the children for each group. The black line below indicated the lag time at which the child and the female model were looking at the same ROI at above-chance levels follow gaze during RJA tasks (Bedford et al., 2012;Falck-Ytter et al., 2015). Upon comparing the two groups, however, we found that the level of gaze-shift synchronization in children with ASD was lower than that in the TD children. This suggests that the ability to synchronize the gaze shifts with social partners in children with ASD, although above the chance level, is poorer than TD children. The level of synchronization here was determined by the length of time children spent on the same ROIs where the female model also looked. Here, we assume that the level of synchronization could be influenced by two factors: gaze following accuracy (the children follow the female model's gaze to the attended object), and the duration of fixations at the female model's face or the attended objects. Vivanti et al. (2017) had found that children with ASD demonstrated decrease attention to the model's face when the model looked straight during RJA tasks. Bedford et al. (2012) and Falck-Ytter et al. (2015) have also noted that the time they spent on the attended objects was shorter than that in the TD group despite the intact gaze following accuracy in the ASD group. These factors could all contribute to the low level of synchronization in the ASD group during the RJA tasks. Based on the previous study, the eye-looking time positively correlated with the subsequent attention on the object in TD children, but this is not the case in children with ASD. It reflected that these children were impaired in understanding the social meaning of gaze and could not actively utilize gaze cues to determine their attention like the TD children (Wang et al., 2020a(Wang et al., , 2020b(Wang et al., , 2020c. Ultimately, our results suggest deficient synchronization performance during RJA in children with ASD, which may result in an insufficient process of social information during joint attention. Second, a delayed social synchronization was identified during joint attention in ASD. Unlike TD children who promptly demonstrated gaze-shift synchronization in responding to the female model's gaze shift, children with ASD displayed a longer latency to respond to it. In addition, the lag time at which the ASD group had the best synchronization performance was longer than that in the TD group. Such a longer latency in the gaze-shift behavior of children with ASD was also identified in some previous studies concerning RJA (Falck-Ytter et al., 2012).These results together suggest that children with ASD had an impairment of synchronization in the dimension of time, and they were not as sensitive to social cues such as gaze shifts as compared with TD children. As responding to the social cues in a timely manner can help people effectively capture some useful but ephemeral information in social interactions, it is vital to explore the latency in social synchronization as well as the motivating reasons. Here, it was speculated that this latency could be accounted for by several factors. First, children with ASD may have difficulties in interpreting social cues (Riby et al., 2013;Vivanti et al., 2011), so they need more time to process these cues before responding to them. Second, previous studies have found that children with ASD are resistant to distraction and may need more time to disengage from the model's face or the attended object (Elison et al., 2013;Elsabbagh et al., 2013). Third, some research has suggested that children with ASD also suffer from gaze dyspraxia, suggesting that they have eye-motor difficulties in looking where another person is pointing or if they are asked to look (Gernsbacher et al., 2008). Their eye movements have been found to be inaccurate, delayed in initiation, and extremely variable in amplitude (Miller et al., 2014). As a result, it may be the incapability of gaze movement rather than the deficit in intentionality to respond to the gaze shift that underlies the delayed latency of gaze-shift synchronization. The combination of these three factors could contribute to the delay in gaze-shift synchronization of children with ASD, and more future evidence is required to support this conclusion. For example, in future studies, we should have some type of baseline measure of volitional saccades to non-social cues to examine if the ASD group has difficulty in gaze movement before inferring higher-level deficits such as problems interpreting social cues in social synchronization.
Third, we further reviewed whether group membership could be predicted from the percentage recurrence through using a machine learning method as a supplementary analysis to verify the traditional analytical methods applied in this study. Generally, it was found that the overall classification performance resulted in 60% accuracy, which differed from chance. This implies that the performance of gaze-shift synchronization could be a potential indicator to differentiate children with ASD from TD children. Meanwhile however, considering F I G U R E 4 The comparison of average cross-recurrence between the ASD and TD group pairing with the baseline adjusted. The black line below indicates the lag time at which the average percentage recurrence in the ASD and TD group were significantly different the low sensitivity and specificity of the single indicator, it is necessary to combine other characteristic indicators of ASD, such as the geometric preference (Moore et al., 2018), the abnormality in face scanning (Wang et al., 2020a(Wang et al., , 2020b(Wang et al., , 2020cYi et al., 2013), and facial affect recognition (Sasson et al., 2016) or incorporate some scales about ASD such as Autism Behavior Checklist (ABC; Marteleto & Pedromônico, 2005) to improve the diagnostic possibilities. Besides, We can further measure both RJA and IJA (initiating of joint attention) during live interactions to study the two-way synchronization of gaze shifts between children and their partners. As the deficit in IJA in children with ASD seems more severe and persisting than RJA (Billeci et al., 2016;Mundy, 2016), the alignment from both sides of the dyadic interaction might additionally contribute to better classification performance.
In our study, we explored to what extent the children with ASD could synchronize their social attention with others' social cues during joint attention. Through a classical RJA paradigm, we attained direct evidence concerning the impairment of synchronization in joint attention. Some previous studies indicated that the development of RJA was characterized by a decreased latency to respond or an increase in efficiency of RJA across age in TD children (Gredeback et al., 2010;Van Hecke et al., 2012). These results suggest that, except for the ability to follow others' gaze, the efficiency of RJA responses is also important for the TD children because social interactions can involve rapid changes and processing of information and a slower RJA responses may lead to difficulty in following the focus of the social interaction. The present study considered both latency/ efficiency and level of synchronization to reflect a complete picture of RJA and found they remained problematic in children with ASD even beyond preschool age. In previous studies, the gaze following of children with ASD was demonstrated in discrete experimental trials without considering social interaction to be a continuous adaptational and coordinated behavior. And analyses based on these discrete trials can only tell us whether children with ASD would show reduced gaze-following in comparison with TD children, but we still do not know whether they have the ability to follow other's gaze by comparing the behavior with a baseline. Our analysis revealed that children with ASD were not looking randomly during the RJA task, suggesting they have the ability to follow partner's social cues, although this ability is poorer than TD children. Besides, these studies ignored the redirection of the attention to the partner's face to form the mutual gaze after gaze following, which is suggested to be a critical component of joint attention. Furthermore, most previous studies focused on whether children with ASD pay attention to the social stimuli and the total looking time at it, paying little attention to the dynamic changes during joint attention and whether children can adjust their visual attention to synchronize with others' social cues timely. The application of CRQA to measuring social synchronization during joint attention has achieved this goal which has been ignored before. Moreover, this method can be further applied in other social interactions such as social communicative exchanges, cooperation and imitation, and will deepen our understanding of the impairments of social interaction and communication in children with ASD.
Six main considerations emerged from the current findings. First, the RJA task in our study was not a real social interaction task but rather a one-way interaction, that is, the child responded to the female model in the video. However, one-way interaction is rarely seen in a natural social interaction. Besides, as the child and the model were not interacting really, one may question whether there is discernable communicative meaning of the social cues in our paradigm. Furthermore, the gaze pattern of individuals with ASD recorded during screen-based tasks may be different from their gaze behavior during live social interactions (Grossman et al., 2019). Social interaction between two persons through a live scenario that can reflect real social situations in daily life should be further explored in the future investigations. Second, social synchronization includes various social interactive behaviors, and we solely focused on social synchronization during joint attention. Social synchronization in many other aspects of social interactions, such as face-to-face communication, imitation, turn-taking, cooperation, and joint actions, should be further examined so as to better grasp the social deficits in ASD. Third, previous studies have identified increased interpersonal neural synchronization in the frontal cortex during cooperative interactions in ASD and TD people, and weaker neural synchronization could predict poorer cooperative performance (Cui et al., 2012;Wang et al., 2020aWang et al., , 2020bWang et al., , 2020c. Here, the behavioral performance of social synchronization during joint attention was analyzed. Future studies could combine the examination of the underlying neural mechanisms of social synchronization during joint attention along with other social behaviors. Fourth, the female model in our video is a stranger to the children. Considering the relationship between the dyads may affect the social interaction (Reindl et al., 2018), it is necessary to consider different types of partners, such as parent-child, peers, and teacher-student, when examining the social interactive behaviors of children with ASD in the future investigations. Fifth, children with ASD who participated in our study had a relatively high IQ. However, it should be noted that, but in fact, there are also many children with ASD who have intellectual disabilities. So the results that were identified here cannot be applied to all the children with ASD. Notably, children with ASD who also have intellectual disabilities should be considered for future studies. Finally, while the present study only included children at 5-8 years age, an important challenge for future studies is to elucidate the emergence and development of a capacity in social synchronization during RJA in a younger sample.
We can further test whether this method could be used to screen younger children with ASD.
In conclusion, we examined to what extent the children with ASD could synchronize their social attention with others' social cues during joint attention. However, joint attention is just one of the many behaviors that involve social synchronization. Our analytic method can be applied in future studies to explore the social synchronization of many other social interactive behaviors in ASD. It can offer behavioral evidence which can be combined with the study in neural synchronization. Additionally, the impairment of social synchronization may be a new implicit indicator for evaluating ASD and can be utilized to screen children with ASD along with other indicators. Besides, such an understanding of the synchronizing impairments of different social interactions can reveal significant implications for developing new training protocols to improve interpersonal interactions and serve as an indicator for more efficient assessments in ASD interventions.