Prosocial behaviors, which refer to actions that one engages in to benefit others, like helping, sharing, and comforting, is often believed to be the basis of human relationships (1). Children with autism spectrum disorder (ASD) might display fewer prosocial behaviors compared with typically developing (TD) children which are likely due to the social-communication deficits associated with an ASD diagnosis (2). For example, school-aged autistic children presented fewer prosocial behaviors in daily life than typically developing (TD) children when using parent- or teacher-reported rating scales (e.g., the Strength and Difficulties Questionnaire) (3, 4). However, in experimental settings, several studies have found that preschool autistic children do show helping behaviors (5, 6). In resource allocation tasks [e.g., Dictator Game (DG) and Ultimatum Game (UG)], autistic children showed a higher preference for equality than self-interest compared to TD children (7–10); but they tended to accept unfair offers (7, 10). Different measures might identify different situation of prosocial behaviors, resource allocation tasks such as DG is believed to be powerful to illuminate individuals' social interactions because it could examine the extent to which individuals attain their own goals while simultaneously displaying altruistic behavior toward unrelated individuals. However, directly measure of prosocial behavior in specific experimental environment may be lack of ecological validity (11). Therefore, it is particularly important to evaluate prosocial behaviors from different dimensions. In China, the study of prosocial behaviors in autistic children is limited and more studies are needed.

Empathy and Theory of Mind (ToM) are generally considered to be the major determinants of prosocial behaviors (12, 13). Empathy has been described as the ability to infer and share the emotional experiences with another (14), and consists of two components (15): the ability to understand the emotional state of others and distinguish another's feelings from one's own (i.e., cognitive empathy), and the ability to vicariously experience of the emotional experiences of others and respond emotionally in an appropriate way (i.e., affective empathy) (15). ToM, on the other hand, has been referred to as the attribution of mental states, such as emotion, desires, intentions and beliefs to others (16, 17). Given the theoretical similarities, cognitive empathy refers to a complex cognitive capacity, largely overlapping with ToM ability (12, 15). However, increasing studies have argued that ToM ability differs from empathy because ToM does not denote a sharing of another person's affective states, but rather a cognitive understanding of another person's mental states (18) (Supplementary Table S1 in Appendix A). ToM deficits have been considered as one of the major features in autistic children (19, 20). Studies have also shown impaired cognitive empathy but intact affective empathy in autistic children (21–24). Specifically, previous studies also indicated that there might be different roles of empathy or ToM ability on prosocial behaviors in children. Numerous researches have shown that prosocial behaviors, such as helping, sharing and comforting, are associated with preschool- and school-aged children's disposition to empathize with others (25–30). A meta-analysis of 6,432 children (2–12 years) revealed that children with advanced ToM abilities were more likely to show prosocial behaviors (r = 0.19), especially cooperating behaviors (r = 0.24) (13). When studying prosocial behaviors in relation to empathy and ToM simultaneously, Abrams et al. (31) found that empathy was positively associated with prosocial behaviors, but the association between ToM and prosocial behaviors was only shown in the non-competitive situation. Longobardi et al. (32) found that both empathic concern and ToM had direct positive effects on prosocial behaviors in primary school children. However, few studies focus on the associations between empathy, ToM and prosocial behaviors in autistic children with inconsistent results. For instance, a study of 20 school-aged autistic children and 20 language-age matched counterparts found that affective empathy was strongly related to peer interaction and prosocial behaviors (helping and sharing) in school-aged autistic children (33), while other studies found no associations between empathic responses and prosocial behaviors in preschool-aged and school-aged autistic children (34, 35). These earlier studies of autistic children with relatively small sample sizes only consider the role of empathy and ToM ability in prosocial behaviors separately, and to our knowledge, no studies have considered the role of empathy and ToM ability in prosocial behaviors simultaneously or defined them distinctly. In this study, we measured empathy and ToM simultaneously and evaluated prosocial behaviors via both parent-reported rating scales and experiments in autistic children and TD children in mainland China. We aimed to investigate the role of empathy or ToM ability in prosocial behaviors between these two groups.

The characteristics of the children were shown in Table 1. The mean age of the included autistic children was 7.4 ± 1.5 years and 86.5% of them were boy. Most of the autistic children were being intervened formerly and/or currently (72.9%). Compared with TD children, autistic children had lower FSIQ and scored higher on the SRS and BRIEF (t = −11.03, 18.99, and 9.50, all P < 0.01). Autistic children's mothers had low education levels with low household income compared with their counterparts (χ² = 9.71 and 38.14, all P < 0.01). There were no differences between groups in child's age (t = 1.53, P = 0.11) or maternal age t = 0.57, P = 0.56).

As shown in Figure 1 (detailed data shown in Supplementary Table S6 in Appendix A), after adjusting for potential covariates, there were no significant difference in EQ-C total score and the subscale score of affective empathy and cognitive empathy between autistic children and TD children (10.8 ± 4.5 vs. 19.4 ± 6.2, 3.4 ± 2.6 vs. 6.5 ± 3.0, 2.4 ± 1.9 vs. 5.5 ± 2.7; β = −1.96, 95%CI: −4.05~0.12, P = 0.06; β = −0.52, 95%CI: −0.27~-0.58, P = 0.40; β = −0.12, 95%CI: −1.06~0.82, P = 0.80). However, autistic children showed a lower passing rate in ToM toolkit compared to TD children (25.0% vs. 88.0%, OR = 0.10, 95%CI: 0.03–0.34, P < 0.01). There was no significant difference between groups in the score of prosocial behaviors in SDQ (4.9 ± 2.0 vs. 6.5 ± 2.1, β = 0.50, 95%CI: −0.36~1.36, P = 0.25). In DG paradigm, autistic children shared less stickers (1.7 ± 1.6 vs. 1.8 ± 1.4, β = −0.78, 95%CI: −1.42~-0.13, P = 0.02) with a lower proportion of sharing (63.5% vs. 79.6%, OR = 0.29, 95%CI: 0.09–0.88, P = 0.03) than those in TD children, and the frequencies were shown in Supplementary Table S7 in Appendix A.

Table 2 showed the associations between empathic traits, ToM ability and prosocial behaviors assessed by the SDQ subscale in autistic children. In the adjusted model, autistic children who had higher EQ-C total score and higher affective empathy score scored higher in the prosocial behavior subscale (β = 0.17, 95%CI: 0.07–0.27; β = 0.47, 95%CI: 0.33–0.60; all P < 0.001).

Table 3 showed the associations of empathic traits, ToM ability and prosocial behaviors assessed using the DG paradigm in autistic children. In the adjusted model, children who passed ToM toolkit shared more stickers compared to those who did not pass the tasks (β = 1.03, 95%CI: 0.16–1.89, P = 0.02). We did not observe significant associations of EQ-C total score and subscale score of affective or cognitive empathy with the number of shared stickers in autistic children.

The results were similar when both empathic traits and ToM ability were entered into the same adjusted model (Supplementary Table S8).

When using parent-reported measurements, we did not find differences in prosocial behaviors between autistic children and TD children; and empathic traits, especially affective empathy, was associated with prosocial behaviors in autistic children. However, we found poorer prosocial behaviors in autistic children than those in TD children when using experimental measurements, which were found to be associated with ToM ability.

Inconsistent results were previously obtained in three observational studies [one in Japan (3), two in the UK (4, 52)], indicating that parent-reported prosocial behaviors in 6–12 year-old autistic children were poorer than TD children. These earlier studies did not consider several important confounders (i.e., socio-economic and cognitive factors), while our study employed a large sample size with comprehensive information on these potential confounders. However, the inconsistency might have also resulted from cultural differences, such as culturally-specific norms and different socialization processes, thus affecting Chinese parents' evaluation of their children's prosocial behaviors (53). Unlike parent-reported measurements, we found that autistic children performed less pro-socially in experimental conditions. Similarly, several studies conducted in Western countries with small sample sizes found that autistic children can share things in resource allocation tasks (6, 10), but showed a higher preference for equality than self-interest compared to TD children (7, 54). In our study, prosocial behaviors measured by the DG paradigm is a form of strong reciprocity, which is a behavior that people may increase the fitness of unrelated individuals at a cost to themselves. This kind of prosocial behaviors is crucial for humans to establish cooperation (55, 56). In this paradigm, advanced social cognitive functions interact with others and an understanding of social norms are required when participants display prosocial behavior toward unrelated individuals (11, 55). Our results indicated that experimental measurements might be able to capture behavioral characteristics of prosocial behaviors in autistic children, and future studies should confirm our findings by using both parent-reported and experimental measurements.

When studying ToM ability and empathic traits simultaneously, we found that autistic children exhibited worse ToM ability but similar empathic traits than TD children. Consistent with previous studies, we confirmed that the ToM ability of autistic children was significantly lower than that of TD children (20, 57–59). However, autistic children had similar empathic trait compared to TD children, which was inconsistent with previous studies (43, 44, 60). When we further considered different types of empathy, we did not find differences in both affective and cognitive empathy. The results of affective empathy were consistent with most previous studies, indicating an intact affective empathy in school-aged autistic children (21, 23, 34, 61). However, parent-reported cognitive empathy in autistic children was not significantly lower than those in TD children, which was not in line with some of the previous findings (23, 61). This inconsistency might have resulted from the differences between self-reported questionnaires and behavioral assessments (21–23, 62). Moreover, most of the behavioral measurements of cognitive empathy were bundled together or conflated with ToM ability, resulting in the difficulties to distinguish cognitive empathy and ToM ability in these studies (21, 22). Fletcher-Watson and Bird (24) proposed that empathy can be broken down into four component stages: (1) noticing another's feeling; (2) correctly interpreting another's feelings; (3) feeling empathy; (4) responding in line with social norms. Although the subscale of cognitive empathy can reflect the second component of empathy, some of the items in this subscale might overlap with other components (i.e., item of 8 and 17 listed in the Supplementary Table S3 of Appendix A), limiting the power of this subscale. Therefore, more studies are needed to consider different components of empathy when studying the behavioral characteristics of autistic children.

In the DG paradigm, we found that ToM ability played an important role in the prosocial behaviors of autistic children. However, our previous study found that ToM ability did not contribute to prosocial behaviors of the DG paradigm in Chinese TD children, while other complex cognitive functions (i.e., inhibition control) may play a role in deciding resource allocation (39). In China, routinely enforced parental instruction that children should share things might eliminate the need for social insight to act as a trigger for sharing behavior for TD children in experimental conditions (13). Regrading autistic children, the development of social cognition is affected by the social motivation (63). Social Motivation Model suggested that early-onset impairments in social attention ultimately deprive the child of adequate social learning experiences, causing the imbalance in attending to social and non-social stimuli and subsequently disrupting social cognitive development (63). The resource allocation task in the context of game theory (i.e., DG paradigm) is developed based on the assumption that individuals can predict other people's actions when they understand others' motivations, preferences, and beliefs (18). Consequently, better ToM ability might contribute to prosocial behaviors of DG paradigm in autistic children. Our results might have clinical insights for behavioral interventions targeting social skills in real life since prosocial behaviors are of importance to social interaction, cooperation and adaptation (64).

In this study, we only observed a positive association between empathic traits and parent-reported prosocial behaviors in autistic children, and the associations were more pronounced when considering affective empathy. Autistic children might understand the emotions of others and respond appropriately since they have intact affective empathy (21, 34, 61). According to the Intense World Theory, autistic children might have hyper-perception and hyper-emotionality caused by hyper-functioning of local neural microcircuits (65). The vicariously emotional experience with others in autistic children might be intact or amplified, which could promote them to perform more prosocial behavior. Meanwhile, parents of ASD children tended to believe that their children behaved more pro-socially when they had higher affective empathic traits in Western countries (35). However, the results should be interpreted cautiously because empathy traits and prosocial behaviors in daily life were parent-reported, indicating the potential measurement error (34). Parents might report higher levels of empathic traits or prosocial behaviors due to exaggerated parental perceptions (66). One previous study in the Netherlands showed that the prosocial responses to peer distress were similar in autistic children and TD children in a computer task, indicating that the association between affective empathy and prosocial behaviors was less pronounced in experimental conditions (34), which was consistent with our findings. Since this is the first study to reveal the different role of empathic traits and ToM ability on prosocial behaviors in autistic children, more studies are needed to confirm our findings.

This study has several limitations. First, we used simple ToM toolkit instead of advanced ToM tasks, which might introduce a ceiling effect in TD children. However, our previous study has confirmed that most autistic children may have struggled with the advanced ToM tasks (20, 58, 59). Further research is needed to develop appropriate paradigms to compare ToM ability between autistic children and TD children. Second, we were not able to measure empathic traits and ToM ability by parent-report and experiment condition, which might introduce potential confounding of our findings. Third, we only performed one single task of the DG paradigm, and therefore it might be subject to measurement errors. Fourth, the behavioral characteristics of prosocial behaviors might be different when offering different resources (e.g., food, toys, attachment objects) (67). We only used stickers in this study as allocation resources although we have investigated the degree of sticker preference, and if the children disliked the stickers, we might not have been able to elicit prosocial behaviors. We found weak correlation between the prosocial behaviors measured by parent-report and by experiments (r = −0.10) which indicated that children might behave differently in their natural social environment than in experimental conditions. Future research is needed to study children's sharing behaviors of different resources in their natural environment. Fifth, we did not match on age and intelligence between the two groups despite of the use of statistical control in this study.

Our study found that autistic children showed less pro-sociality and ToM ability than TD children. In autistic children, empathic trait was associated with parent-reported prosocial behaviors while their ToM ability was associated with prosocial behaviors in experimental condition. Our findings indicated that better ToM ability and empathic trait might promote prosocial behaviors in autistic children.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Ethical Review Committee for Biomedical Research, Sun Yat-sen University. Written informed consent to participate in this study was provided by the participants' legal guardian/next of kin.

JJ, X-HL, and J-HG: conceptualization, supervision, project administration, and funding acquisition. XW: methodology, formal analysis, and writing-original draft preparation. BA and L-ZL: study design and manuscript revise. XW, NP, QC, J-JC, S-YL, and M-XD: data collection. L-ZL: data curation. XW, BA, NP, QC, J-JC, S-YL, M-XD, J-HG, X-HL, and JJ: writing–review and editing. All authors contributed to the article and approved the submitted version.

This work was supported by the Key-Area Research and Development Program of Guangdong Province (2019B030335001) and the National Natural Science Foundation of China (81872639, 82103794, and 82003482). BA was supported by the European Union's Horizon 2020 Research and Innovation Program under the Marie Skłodowska-Curie Grant Agreement No. 813546, the Baily Thomas Charitable Fund, and the Data Driven Innovation and the UK Economic and Social Research Council (ES/N018877/1) during the course of this work.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.