Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Neural computations in children’s third-party interventions are modulated by their parents’ moral values


One means by which humans maintain social cooperation is through intervention in third-party transgressions, a behaviour observable from the early years of development. While it has been argued that pre-school age children’s intervention behaviour is driven by normative understandings, there is scepticism regarding this claim. There is also little consensus regarding the underlying mechanisms and motives that initially drive intervention behaviours in pre-school children. To elucidate the neural computations of moral norm violation associated with young children’s intervention into third-party transgression, forty-seven preschoolers (average age 53.92 months) participated in a study comprising of electroencephalographic (EEG) measurements, a live interaction experiment, and a parent survey about moral values. This study provides data indicating that early implicit evaluations, rather than late deliberative processes, are implicated in a child’s spontaneous intervention into third-party harm. Moreover, our findings suggest that parents’ values about justice influence their children’s early neural responses to third-party harm and their overt costly intervention behaviour.


Third-party intervention into perceived moral transgression is a key proximate mechanism enabling humans to manage conflict and maintain high levels of social cooperation1,2. Intervention (e.g. verbal reprimand, policing and protesting) is different from punishment or defensive behaviours in that it does not include retaliatory or punitive sanctions directed specifically at the wrongdoer3,4. However, intervention may often be costly, because the agent who intervenes can pay a personal cost while their actions yield no immediate personal gain1. Recent studies have shown that rudimentary abilities for costly intervention behaviours are present in infancy5 and toddlerhood6,7 well before children actually enact third-party intervention into real-life transgressions8,9,10. Although there is a growing body of literature on the development of children’s third-party intervention, little is known about the neural mechanisms underlying the early ontogeny of such behaviours. Furthermore, to our knowledge, there is no empirical work that has directly examined how child’s moral behaviours and their underlying neural processes are modulated by their parents’ values about justice. This study investigated the electrophysiological processing underlying children’s costly interventions and their time course, and how parents’ dispositions modulate the ontogeny of moral intervention behaviours.

Recent developmental work found that a preference for costly intervention behaviours emerges as early as 6 months in infancy5, and 20-month-old toddlers are willing to reward characters who intervene in third-party transgressions6,7. These findings suggest that the foundational capacity for third-party intervention, including normative understandings, emerge early in ontogeny. Beginning at around 3 years of age, young children do not just follow moral norms, but actively enforce them on others even when their own interests are not at stake11. For example, they show explicit disapproval and actively intervene against transgressions such as violations of property rights8,9,10,12,13, distribution inequity14,15 and physical harm3. Also, they often report (tattle) the perceived transgression to an authority on behalf of a third party12,16,17. Furthermore, intervention behaviours demonstrated by 3–6 years old children is nuanced, showing a higher probability of intervention into moral transgressions rather than violations of conventional norms and rules of games16. Many developmental psychologists believe that children’s protesting and reporting behaviours indicate that children have acquired a normative understanding against harm and inequity by around 3 years of age18,19.

Other scholars, however, cast doubt on whether preschoolers as young as 3 years of age can truly enact an altruistic reaction to transgression, arguing instead that what appears to be preschooler’s costly intervention behaviour is simply an indication of their impulsivity20, or their limited ability to consider the state of the transgressor’s emotions21,22, or an inability to anticipate future outcomes (e.g. retaliation, failing to complete a task) involving the self23. It has also been proposed that children’s third-party intervention behaviour is not primarily motivated by their intrinsic desire to enforce moral norms. Rather, the behaviours are more likely self-serving, for example, avoiding blame or maintaining reputation and trustworthiness within their affiliated group14,24,25.

Accumulating evidence undermines such scepticism. Children as young as 3 seem to intentionally enact third party intervention, even when they know it comes with personal cost. For example, they shut down a slide to prevent an antisocial child from playing on it, even when knowing they sacrifice their own opportunity to play on it10. Also, no significant correlation has been found between 3-year-olds’ third-party intervention and dispositional impulsivity17, suggesting that children’s intervention behaviours are not simply an impulsive ‘knee-jerk’ reaction. Moreover, many researchers point out that, when intervening in transgressions, children often use normative terms (e.g. “That’s wrong”, “You shouldn’t do that”), suggesting that young children may be “compelled to actually enforce” norms11 and promote good behaviours conducive to social cooperation3,12,15. However, whether and how children’s nascent understanding of social norms actually drives their costly intervention behaviours, cannot be understood by studying the behaviours of interest alone. Rather these need to be examined in conjunction with other independent measures26,27. Therefore, in this study, data from three independent sources were triangulated, namely: (a) electroencephalography (EEG), (b) young children’s behaviours recorded when involved in a transgression experiment and, (c) reports provided by the parents of the participating children of their child’s social and emotional dispositions.

The researchers conducting this study are aware it could be argued that when young children appear to be enforcing moral norms, they are simply imitating their parents, as children often do in other costly scenarios28. However, we question whether it is possible for children to feel compelled to imitate what they observed from parents, without ‘understanding’ what the parents were objecting to. In order to do ‘what parents did’ in a novel context, they would need to understand it was the violation of a somewhat universal norm that their parents were actually objecting to11,29,30. Previous studies have identified some strategies that parents adopt in moral induction. Parents treat moral violations (e.g. acts of harming) differently from children’s disobedience to social conventions31. For example, they often use firm-stern vocalization in response to moral transgression, and carefully explain why the act is wrong and direct the child’s attention to how specific acts might harm others’ rights or welfare32. However, it is not well understood how young children transfer moral norms learnt from parents to a novel contexts, where the parents are not present.

Meanwhile, some burgeoning work in neuroscience has begun to examine how parental dispositions are related to the neural processing of children’s moral judgement and behaviours. Some studies have shown that individuals’ disposition for justice sensitivity modulates their neural computations when they perceive interpersonal harm33,34. In the case of parents, their own moral values modulate the neural computations of their children when they perceive third-party social interactions35. In that latter study, parental justice sensitivity was predictive of infants’ and toddlers’ event-related potentials (ERP) differences in the perception of helping versus harmful scenes. ERP differences in a specific time-window (300 and 500 ms) were related to infants’ and toddlers’ reaching preference for a prosocial over an antisocial character shown after the EEG session. However, little is known about how parents’ dispositions, especially their concern for social injustice and empathy, shape children’s neural computations that subserve costly intervention into third-party moral transgressions. Therefore, this study specifically aimed to examine whether children’s implicit understanding of moral norms and values is modulated by parents’ justice sensitivity and empathy, such that it might predict their costly intervention behaviours.

Some handful of neuroscientific studies have started to build a neurobiological model of third-party intervention behaviours36. Regarding the evaluation of moral norm violation, EEG/ERPs studies show that an implicit assessment is made rapidly, within 200–250 ms after the onset of information that violates moral values and norms that one upholds37,38,39. For example, in a study with adults, a broadly distributed P2 was elicited by the onset of a target word, in a statement such as ‘euthanasia is acceptable’, when the target word ‘acceptable’ clashes with the reader’s value system37. Also, after 500–650 ms from the onset of a value-inconsistent word, another positive-going deflection (late positive potential, LPP) is observed37,39. This evidence shows that value-based judgement rapidly recruits attentional resources (as indexed by P2) and engages sustained processing within the affective system, as indexed by LPP37,39. Meanwhile, reduced P2 for the perception of harm/distress often indexes preclusion of sustained processing in prefrontal cortex, which may potentially lead to avoidant reactions40 or callous-unemotional reactions. Similar time-course of spatio-temporal responses have been detected in young children. Notably, children’s perceptual sensitivity and sustained attention to third-party moral transgressions is indexed by increased central P2 and LPP41,42,43.

There is a lack of consensus regarding how aversive reactions such as distress and anxiety, triggered by perceived norm violations, modulate subsequent prosocial actions. Some studies showed that personal distress stemming from perceived third-party harm does not motivate altruistic behaviours44,45, though, conversely, it can prohibit one’s costly interventions (i.e. self-concerned action or no action). One functional MRI study documented that personal distress, as indexed by the neural response in amygdala and periaqueductal grey does not predict costly altruism46. This network is known to trigger avoidant, withdrawal responses47. Another study found that empathic concern and costly donations were predicted by the neural response in the mesolimbic dopamine system, including the ventral striatum and ventromedial prefrontal cortex48. By contrast, other studies showed that aversive arousal guides behavioural choices in personal moral dilemmas. For instance, an EEG study with adult participants reported that individual dispositions in personal distress negatively predicted the mean percentages of utilitarian choices (e.g. killing one person to save more people) and positively correlated with the mean amplitudes of the P260, an ERP component reflecting an immediate emotional reaction during decision-making49. These results indicate that self-oriented feelings of anxiety and unease, rather than other-oriented feelings of concern, guide behavioural choices in personal moral dilemmas. Nevertheless, the jury is still out as to whether costly altruism is driven by an egoistic motivation to alleviate one’s own distress or by a genuine regard for another’s welfare48.

In ERP studies using stimuli evoking negative emotions, the frontal N2 (200–400 ms) is regarded as a marker of emotion regulation50. N2 is enhanced when participants engage with negatively valenced emotional evaluations of self and other51, and negative feedback concerning one’s performance52. The frontal N2 reflects the ‘augmentation of inhibitory controls’ when negative emotions arise50. In young children (4–6 yrs), N2 amplitudes are greater in the presence of negative emotion53,54, and more rapid when children have anxiety or fearful trait53. In a cross-sectional ERP study, Cheng and colleagues55 showed that the difference wave of N2 (third party pain vs. no pain), indexing inhibitory control over empathic arousal, has an age-related decrease between ages 3 and 9. If self-oriented feelings of anxiety and unease guide behavioural reactions45, heightened inhibitory control over personal distress, indexed by enhanced N2, may predict a child’s reaction to third-party moral transgressions.

Against the backdrop provided above, in order to elucidate how perceived norm violation and subsequent emotional arousal, elicited by observing third-party moral transgressions, lead to costly intervention in young children, we examined the relationship between children’s real-life costly behaviours and the spatio-temporal dynamics of neural processing recorded with EEG, in response to morally laden scenarios. The Chicago Moral Sensitivity Task (CMST, Fig. 1) comprises child-friendly visual scenarios depicting two characters intentionally performing either harming actions or helping actions35,41. As noted earlier, existing experimental paradigms have mostly relied on observing children’s overt behaviours or verbal responses, which are important yet limited in informing on how children’s implicit normative evaluation actually drives their costly intervention behaviours26. Meanwhile, neurodevelopmental studies, using EEG/ERP recordings, can contribute in elucidating the neural mechanisms that subserve attentional and affective processes initiated by the perception of norm violations, enabling us to understand how third-party costly intervention may be mediated by implicit moral evaluation during the early ontogeny of moral development56,57. Actually, a series of neurodevelopmental studies using the CMST35,41,42 indicated that there are distinctive contributions of both early automatic (e.g. central P2, frontal N2) and later controlled processes (e.g., central LPP) underlying children’s real-time evaluation of morally laden behaviours.

Fig. 1: Two stimulus examples from the CMST.
figure 1

Top (Third-party harm): The blue character is intentionally kicking the red character, resulting in red character being knocked down. Bottom (Third-party help): The blue character is offering help to the red character who is moving a heavy snowball, resulting in the building of a snowman. The ERPs were time locked to the onset of the second image in the sequence.

Based on the empirical work reviewed so far, we expected that children’s costly interventions in response to third-party harm would be initially driven by rapid implicit evaluation of norm violations. Specifically, we hypothesized that a difference between children who intervene in third-party harm and those who do not, might be identified by both early neural discrimination/recognition of moral norm violation (as indexed by central P2) and sustained evaluative processing (as indexed by frontal LPP). Additionally, we hypothesized that the decision to intervene in third-party harm will not be predicted by frontal N2 that reflects the inhibitory control over anxiety and distress. Testing a difference implicating the N2 response would help us elucidate whether children’s intervention behaviour is driven by impulsive reactivity20 or self-oriented feelings49. Finally, we hypothesized that parents’ dispositions in empathy and justice sensitivity would contribute to their children’s sensitivity to third-party moral transgressions, as indexed by ERPs data and their behavioural reactions in real-life settings.


Children’s reactions to a third-party moral transgression

Protest behaviour

Figure 2a shows percentages of protest responses that were observed. A total of 16 children (34.8%) explicitly protested to the adult wrong-doer while 30 children (65.3%) remained silent. Among protest behaviours, giving ‘hints of protest’ was the most prevalent form (N = 8, 50%), followed by ‘explicit normative evaluation of misdeed’ (N = 5, 31.3%) and ‘imperative protest (repeating the rule)’ (N = 3, 18.7%). Among those who did not protest, 17 children (56.7%) exhibited facial expressions of discomfort, uneasiness or anger, though these did not lead to protest behaviour. Coders could not identify any notable non-verbal signs from the remaining 13 children (43.3%), though they were all attentive to the adult transgressor’s behaviour. There was no correlation between the child’s age and the extent of protest shown (Spearman’s ρ = 0.082, p = 0.588). When children were divided into two age groups (3–4 years vs. 5–6 years, see Fig. 2b), there was still no statistical difference between two groups, χ2(4, N = 46) = 5.096, p = 0.278, though ‘explicit evaluation of misdeed’ was observable only in the older group. Child’s gender [χ2(1, N = 46) = 0.254, p = 0.61] and parent’s religious affiliation [χ2(1, N = 46) = 0.339, p = 0.560] did not associate with whether or not each child protested.

Fig. 2: Percentages of protest and behaviours that fell into each category.
figure 2

The error bars show standard error for each type of protest and report.

Reporting behaviour

Figure 2c shows percentages of report behaviours that were observed. A total of 26 children (56.6%) explicitly reported what they observed when the research assistant returned. There was a significant correlation between protest and report behaviours, χ2(8, N = 46) = 22.769, p = 0.004. Out of 16 children who protested to the transgressor, 14 children subsequently reported the misdeed. However, 2 children who had exhibited a hint of protest did not report what they had observed. Out of 30 children who did not protest, 18 children did not report what happened. However, 12 children reported what they witnessed, and 9 out of those 12 were those who exhibited non-verbal signs (e.g. discomfort, uneasiness or anger) when observing the transgressor’s misdeed. When children were divided into two age groups (3–4 years vs. 5–6 years, see Fig. 2d), there was no association between child’s age and the extent of report shown, χ2(2, N = 46) =4.301, p = 0.116. Child’s gender [χ2(1, N = 46) = 0.278, p = 0.598] and parent’s religious affiliation [χ2(1, N = 46) = 2.769, p = 0.096] did not correlate with whether or not each child reported.

Relationship between children’s ERPs and moral behaviour observed in the experiment

In order to examine whether difference between children who intervene in third-party harm and those who do not, might be identified by different ERP patterns induced by moral norm violation, we conducted repeated-measures analyses of variance (ANOVA) in the pre-selected time-windows. In each analysis, the Grouping based on children’s reaction (Protester vs. Non-protester, Reporter vs. Non-reporter) was entered as a between-subject variable. The Condition within the CMST (Help vs. Harm) was entered as a within-subject variable.

Posterior N1 component (100–175 ms)

Between Protesters and Non-Protesters, there was no Group main effect on N1 amplitudes, F(1, 40) = 0.030, p = 0.862, power = 0.06, f = 0.032. Also, no Group × Condition interaction was found, F(1, 40)=0.126, p = 0.724, power = 0.06, f = 0.055. Between Reporters and Non-reporters, there was neither Group main effect on N1 amplitudes, F(1, 40) = 0.178, p = 0.675, power = 0.07, f = 0.063, nor Group × Condition interaction, F(1, 40) = 0.214, p = 0.646, power = 0.073, f = 0.071. These results indicate that there was no difference between Groups (Protester vs. Non-Protester; Reporter vs. Non-Reporter) during the initial attention and encoding process when observing visually presented stimuli.

Central P2 component (150–350 ms)

In the entire sample, there was no Condition (Help vs. Harm) main effect, F(1, 40) = 2.291, p = 0.138, power = 0.327, f = 0.239. However, between Protesters (N = 15) and Non-Protesters (N = 27), there was significant Group × Condition interaction effect, F(1, 40) = 6.063; p = 0.018, power = 0.694, f = 0.39 (see Fig. 3). In the Non-Protest group, the mean P2 amplitude in response to viewing harming scenes (M = 2.358) was smaller than the mean amplitude of viewing helping scenes (M = 3.494), t(27) = 4.133, p < 0.001 < 0.05/2, with a medium effect size (power = 0.826, d = 0.568). In the Protest group, P2 amplitude in response to viewing harming scenes (M=3.568) was slightly bigger than helping scenes (M = 3.135), however, the difference was not statistically significant, t(14) = 0.481, p = 0.638, power = 0.08, d = 0.11. Between Reporters and Non-reporters, there was neither Group main effect on P2 amplitudes, F(1, 40) = 0.012, p = 0.915, power = 0.054, f = 0.028, nor Group × Condition interaction, F(1, 40) = 1.520, p = 0.225, power = 0.231, f = 0.193.

Fig. 3: Comparisons of waveforms, difference waveforms and range-scaled voltage spline maps of the scalp distribution.
figure 3

a Left: Grand-averaged ERPs (P2 and LPP windows highlighted) in response to perceiving helping actions (blue) and harming actions (red) at the central cluster (C3-Cz-C4) across two groups (Protester vs. Non-Protester). Difference waveforms (grey) were generated by subtracting amplitudes of helping condition (blue) from harming condition (red), with negative values plotted upwards. Right: The range-scaled spline maps of the scalp distribution of three time windows in two groups. b Group (Protester vs. Non-Protester) × Condition (Help vs. Harm) interaction effects in Central P2 and LPP. The error bars show each group’s 95% CI in two different conditions.

Frontal N2 Component (200–400 ms)

Between Protesters and Non-Protesters, there was no Group main effect on frontal N2 amplitudes, F(1, 40) = 1.545, p = 0.221, power = 0.236, f = 0.196. Also, no Group × Condition interaction was found, F(1, 40) = 0.009, p = 0.925, power = 0.05, f = 0.01. Between Reporters and Non-reporters, there was neither Group main effect on N2 amplitudes, F(1, 40) = 0.475, p = 0.495, power = 0.107, f = 0.11, nor Group × Condition interaction, F(1, 40) = 0.523, p = 0.474, power = 0.112, f = 0.115.

Central LPP Component (400–600 ms)

In the entire sample, there was no Condition (Help vs. Harm) main effect, F(1, 40) = 0.820; p = 0.371, power = 0.148, f = 0.142. However between Protesters and Non-Protesters, there was significant Group × Condition interaction effect, F(1, 40) = 4.286; p = 0.045, power = 0.545, f = 0.328 (see Fig. 3). In the Non-Protester group, the mean LPP amplitude in response to viewing harming scenes (M = 1.894) was smaller than the mean LPP amplitude of viewing helping scenes (M = 2.753), t(26) = −2.647, p = 0.014 < 0.05/2, power = 0.753, with medium effect size (Cohen’s d = 0.518). In the Protest group, the mean LPP amplitude from harming condition was slightly bigger (M = 2.530) than help condition (M = 2.163), though it was not statistically significant, t(13) = −0.670, p = 0.515, power = 0.095, d = 0.179. Between Reporters and Non-reporters, there was neither Group main effect, F(1, 40) = 0.597, p = 0.444, power = 0.122, f = 0.123, nor Group × Condition interaction, F(1, 40) = 0.600, p = 0.443, power = 0.122, f = 0.123.

Relationship between children’s ERPs, costly moral behaviour and parents’ dispositions

In order to examine the relationships between parents’ dispositions (i.e. empathy and justice sensitivity), and their child’s costly moral behaviour and her/his EEG/ERP measures, Bayesian structural equation modelling (SEM) was conducted, utilizing AMOS 2658. Figure 4 illustrates the three models that were tested in this study.

Fig. 4: Standardized estimates for the effects included in the three models.
figure 4

a Model 1: Within-child (brain–behaviour) model (ppp = 0.44, power = 0.677). b Model 2: Intergenerational (Parent–child) model 1(ppp = 0.40, power = 0.803). c Model 3: Intergenerational (Parent–child) model 2 (ppp = 0.24, power = 0.795). Bold pathway delineates significant associations among primary variables of interest. *Significant with a 90% Bayesian credible interval **Significant with a 95% Bayesian credible interval.

The Model 1 (within-child model) represents brain–behaviour relationships that are implicated in child’s protest behaviour. With regard to selecting which ERP components to include in the Model 1, the hypothesized model included only central P2diff and LPPdiff, given that they were the only components that demonstrated a significant difference between two groups (Protester vs. Non-protester). The child’s age did not show any correlation with other variables (see Table 1), hence, the age was not included in this model and onwards. In addition to the variables from the Model 1, the Model 2 (intergenerational model) includes parent dispositions, namely caregiver’s empathy and justice sensitivity. Regarding caregiver’s empathy, based on previous work41 and observed zero-order correlations (Table 1), only cognitive empathy, not affective empathy, was included in the model. Among the four distinctive justice dimensions measured from the parents, ‘Perpetrator Sensitivity’ was included in the model, with its significant correlations with difference amplitudes for both P2 (ρ = 0.378, p < 0.05) and LPP (ρ = 0.290, p < 0.075) (see Table 1). Based on the literature indicating that a child’s dispositions moderate the relations between parenting and structure/function of neural networks that underpin children’s emotion regulation59, the Model 3 (modified intergenerational model) additionally included the child’s effortful control measured with the CBQ. Descriptive statistics and bivariate correlations for the primary study variables are provided in Table 1.

Table 1 Spearman’s rho (ρ) correlations and descriptive statistics for all variables.

The posterior predictive value (ppp) for the Model 1 was 0.44, and the ppp values for Model 2 and 3 were 0.40 and 0.24, respectively, indicating the Model 1 (within-child model) has the best fit. Among the two intergenerational models, Model 2 demonstrated better fit, with smaller ppp value (ppp = 0.40, power = 0.803; see Table 2) compared to the Model 3 (ppp = 0.24, power = 0.795). Figure 4 presents all the paths included in the model test, with standardized regression weights. Significance levels are illustrated graphically by line thickness. In what follows, we discuss model pathways that directly address our primary research questions.

Table 2 Bayesian structural equation modelling results .

First, we sought to answer whether early neural discrimination of moral norm violation (as indexed by central P2) and sustained evaluative processing (as indexed by frontal LPP) predict young children’s intervention into a third-party moral transgression. After controlling for correlation between central P2diff and LPPdiff, only central P2diff was directly associated with the child’s intervention into a third-party moral transgression. An increase in central P2 amplitude differences between harm and help conditions in the CMST was associated with child’s more explicit and evaluative response to adult’s moral transgression. Previously in this paper, it was reported that protesting children and their counterparts (non-protesters) showed a significant difference in central LPP via an ANOVA test. However, the output from the SEM shows that after controlling for LPPdiff’s strong positive correlation with central P2diff, central LPPdiff (400–600 ms) does not predict the extent of a child’s behavioural response.

Second, we also examined associations between parental cognitive empathy and justice sensitivity dispositions and amplitudes difference from two ERP components. Parental justice sensitivity (perpetrator) showed a significant positive association with the child’s central P2diff. However, parental cognitive empathy showed no significant relationship with either P2diff or LPPdiff. In the Model 3, a child’s effortful control, as a stable disposition, showed a positive correlation with the parent’s cognitive empathy. However, child’s effortful control did not predict amplitude differences in either of the time windows. Moreover, as mentioned earlier, adding a pathway from a child’s effortful control to the Model 2 did not lead to better model fit.


Although existing literature documents that preschoolers intervene in third-party transgressions8,9,10,12,13, there is little consensus about what actually give rise to such actions. Does a child’s protest truly indicate her understanding of moral norms? Is the intervention behaviour driven by discomfort, or, perhaps the child simply imitates her parents’ actions? By triangulating data from three independent sources, including EEG/ERPs, children’s behaviours recorded when involved in a transgression experiment, and their parents’ observations, this study provides evidence that rapid implicit evaluations of norm violations are implicated in a young child’s intervention into third-party harm. Additionally, it indicates that parents’ values on justice sensitivity influence their pre-school aged children’s early neural responses to moral norm violations, which is then predictive of the child’s costly intervention behaviour.

Our study further provides evidence that children, as young as 3 years of age, can enact costly third-party intervention by protesting and reporting. Previous research has shown that young children from age 3 enact third-party punishment to transgressors shown in video or puppets9,10. In the present study, in the context of real-life transgression experiment, even the youngest participant (41 months old) engaged in costly intervention, by hinting disapproval to the adult transgressor (why are you doing that?) and subsequently reporting the damage when being prompted. During the experiment, confounding factors such as a sense of ‘responsibility’, were avoided by keeping the person playing the ‘research assistant’ role out of the room when the transgression occurred. Furthermore, when leaving the room, the ‘research assistant’ did not assign the children any special role to police or monitor the actions of the ‘visitor’ (who would transgress). Moreover, the transgressor was not an acquaintance of the child, and the book was said to belong to a university (not a child’s school nor researchers), hence giving little sense of in-group/out-group membership11,60. Also, the participating children would likely attribute ‘power’ and ‘authority’ to the visitor/transgressor, as an adult26. Nevertheless, in the real-life experimental context, 34.8% of children explicitly protested to the adult wrong-doer.

In line with our hypothesis, early neural discrimination of moral norm violation (as indexed by central P2) predicted children’s intervention in real-life moral transgressions. Importantly, non-protesting children exhibited significantly reduced P2 amplitudes for harming scenes. As noted earlier, in response to morally laden scenarios, centro-frontal P2 amplitude indexes automatic evaluative-categorization of information that violates moral values that one upholds37,38,39. In previous developmental studies using the CMST, P2 amplitudes were greater when perceiving harming actions than helping actions42. Furthermore, preschoolers who participated in an empathic learning intervention subsequently exhibited enhanced P2 in harming conditions43. Meanwhile, reduced P2 was observed when perceived negativity (or threat) quickly re-allocates available resources to sub-cortical regions in the preparation of ‘fight-flight responses’47,61, precluding sustained processing in prefrontal cortex40. Therefore, reduced P2 observed in non-protesting children implies that the extent of this preclusion, when viewing third-party transgression, is greater than in those who protested during the early time-course of neural processing. In protesting children, P2 amplitudes between harm and help scenes did not show a statistically significant difference. This may be due to the small sample size (N = 15). However, it might be also possible that the difference is more pronounced in older children. For example, in a previous study where P2 amplitude was larger when viewing harming actions than helping actions42, children’s average age was approximately 1 year older than the cohort of this study. Future work is warranted to elucidate the age-related development of P2 components when engaged with morally laden scenarios.

Contrary to our hypothesis that both P2 and LPP are implicated in third-party intervention behaviours, late regulatory processing, indexed by LPP, was not directly associated with a child’s costly actions. Although independent comparisons at each of the two time-window (i.e. P2 and LPP) indicate different patterns of neural computations between protesters and non-protesters, after controlling for a positive correlation between two ERP components, only P2 was directly associated with the child’s protest behaviour. This result implies that, at least in preschool age, the quality of initial implicit evaluation of moral norm violations has a strong sustained impact on later regulatory processes and, furthermore, it modulates subsequent decision-making to intervene. Additionally, the correlation across two time-windows and the identified predictive power of P2 suggests that claims for a pervasive dichotomic split between perception vs. action, and intuition vs. reasoning may misrepresent the processes that give rise to moral actions62,63. Rather, as noticed by other authors37,63, late neural dynamics are bounded by earlier temporal dynamics, meaning that salience detection and regulation processes are interwoven during the evaluation of moral transgressions and costly action.

This study adds to evidence that children’s third-party intervention is not driven by their impulsivity or an urge to release self-oriented feelings49,64. As shown above, there was no association between a third-party intervention (protest and report) and the augmentation of inhibitory control over negative emotion, indexed by N2. Also, similar to the previous findings17, children’s dispositional surgency (including impulsivity, tendency to approach and externalize) measured with the CBQ-VSF did not show any correlation with either behaviours (ρ = 0.129, p > 0.05 for protest, ρ = −0.072, p > 0.05 for report; see Table 1). Rather, the extent of children’s report behaviour showed a moderate correlation with their effortful control (ρ = 0.399, p = 0.005) and parent’s cognitive empathy (ρ = 0.338, p = 0.020). It might be possible that more sustained intervention behaviour, such as reporting, requires additional effortful control and perspective-taking, compared to the spontaneous protest to the transgressor. Therefore, the relationship between these variables needs to be further investigated and compared across different types of intervention behaviours that children engage with.

Finally, this study adds nuanced evidence on how parents shape children’s neurodevelopment that supports their moral decision-making. The study suggests that parents’ own values on justice modulate the early phase of neural computations of third-party moral scenarios in their children. However, this modulation was not observed with later controlled processes. This result adds to previous findings on parental modulation of moral development32,41 by providing new evidence that parents’ justice sensitivity is predictive of a child’s real-life moral behaviour, by shaping attentional and perceptual neural mechanisms. In other words, pre-schoolers of parents who are perceptive to harm or injustice inflicted by themselves, have not only developed a refined salience detection mechanism of third-party harm, but also are more likely to act against what they perceive to be another’s moral transgressions.

It should be emphasized that parent’s cognitive empathy was not implicated in the child’s neural computations of moral norms or their spontaneous intervention behaviour. However, parents’ cognitive empathy had a positive correlation with a child’s effortful control and their subsequent report behaviour. This distinct contribution made by two different dispositions (cognitive empathy and justice sensitivity) suggests that parenting strategies necessary to enhance a child’s moral development require both aspects: perspective-taking and understanding of moral values. Future studies that integrate functional neuroimaging methods and naturalistic observations of parent–child dyadic interactions need to be conducted in order to further elaborate the unique contributions from both parents’ empathy and justice values to children’s moral learning and development. Such studies would not only be of interest to developmental scientists, they would also have much to contribute to educational policy and practice, in providing a neurobiologically sound account of moral development63, leading to well-designed pedagogical strategies43,65.

We are aware that the relationships between spatiotemporal dynamics of neural processing and children’s costly intervention behaviours identified in this study are drawn from data at the group level. Also, due to the small sample size, analysis of ERP components, particularly the analysis of Protester group, were underpowered. Also, the participating children might have attributed ‘power’ and ‘authority’ to the adult transgressor26, preventing them from protesting and reporting even though they might have done so with a peer. Nevertheless, the findings from this data suggest potential avenues for future studies. For example, a follow-up study would be needed to examine the possible relationships between neural computations and costly behaviours using a within-individual approach to the brain–behaviour link66. Additionally, studies with different age groups and clinical populations, such as children with conduct problems or CU traits, would help elucidate whether the brain–behaviour link found in this study can be further generalized, and what factors might strengthen or weaken the pathways identified from this study.



A priori sample size was decided with a power analysis using G*Power The necessary sample size for a large effect of f = 0.64, a 0.05 probability of error, and a power of 90% for a 2 (Group: Protester vs. Non-Protester) by 2 (Condition: Help vs. Harm) repeated-measure analysis of variance (ANOVA) was 28 (14 participants per group). The effect size of f = 0.64 was estimated from the observed effect size (ηp2 = 0.29) reported in Meidenbauer et al. (2018)’s study42 on young children’s ERP responses to morally laden visual stimuli. In order to ensure a sufficient number of children with full data set (i.e. EEG, observed behaviours during the experiment, parent survey), we recruited 47 preschoolers between the ages of 41 and 72 months (M = 53.92 months, SD = 7.76; N = 16 girls) participated in the study. Participating children did not have any identified neurological (e.g. autism, ADHD) or other medical conditions. The average age of participating parents was 38.2 years old (SD=3.62; 41 mothers and 6 fathers). All responding parents were the child’s biological parent. At the time of the study, all responding parents reported they live with the child for 7 days per week. Except 2 parents, 45 parents reported they are the child’s primary caregiver. Although all participating pre-schools have multicultural intakes, all participating children had good English comprehension, with English as the compulsory language in the preschools they were attending. All children were born in Australia, and 26 children spoke English as home language (55.3%). 21 children used English as an additional language and their home language were Korean (N = 10, 21.3%), Mandarin (N = 9, 19.1%), Arabic (N = 1, 2.1%) and Thai (N = 1, 2.1%). Regarding the family’s religion, 24 parents (51.1%) responded they have no religious affiliation, followed by Christianity (N = 10, 21.3%), Catholic (N = 5, 10.6%), Islam (N = 4, 8.5%), Buddhism (N = 1. 2.1%), Coptic Orthodox (N = 1, 2.1%), Greek Orthodox (N = 1, 2.1%) and Hindi (N = 1, 2.1%).

Fliers introducing the study and inviting participation were distributed to directors of preschools, who forwarded them to prospective parents of children enroled in their preschool. Prior to giving consent, all parents were informed that at the end of the experiment their travel cost to the campus would be reimbursed and the child would be rewarded with a book ($20 equivalent) at the end of the experiment. Written informed consent was obtained from all parents and each child’s verbal assent was sought prior to commencing data collection. All data were collected in accordance with the ethical principles of the Declaration of Helsinki and the study was approved by the University of Sydney Human Research Ethics Committee (approval No. 2017/468).

Third-party moral intervention task

This study adopted a design that was previously used to investigate children’s protesting and reporting third-party harm when an unfamiliar adult confederate intentionally damaged a library book12. Parents and children were informed that researchers would collect children’s drawings as a part of the project. After receiving parental consent and the child’s own verbal assent, a female research assistant invited a child to a drawing-room, equipped with a video camera. Children were given A3-size papers and coloured pens, then were told, “The first thing we ask all our visitors to do is draw a picture of a moment when you felt very grateful to somebody.” The research assistant encouraged and motivated the child, by drawing a picture of her own. The initial phase lasted until the child had settled in, relaxed and started her/his drawing.

Then, an adult female confederate knocked the door and entered, saying, “Hello. I came to participate in the study and was told that I should wait here, as the other rooms are locked.” The research assistant provided the confederate a seat beside the child and said, “Would you also like to draw a picture while waiting?” The research assistant then told both the child and the confederate, “I am going to leave to check the other room. I will leave my things (belongings) here: my books, pen and jacket. Oh, just one thing. Please be very careful with this book (attached with a university logo and barcode) because it belongs to the library andIhave to take good care of it. I need to return it tomorrow.” The research assistant ensured this remark was heard and understood by the child, checking her/his facial expression.

After the research assistant left the room, the transgression session started. The confederate picked up the book, saying: “Oh, this looks like an interesting book. I think I will read this book instead of doing the drawing.” After skimming through a few pages, the confederate said, “Oh, I really like the picture on this page. I’m going to tear this page out of the book, the librarian won’t know if only one page is missing.” Then, she removed the page from the book in full sight of the child and put the torn page into her pocket, so that it would not be noticed by the research assistant.

During the last phase of the transgression, the research assistant returned, telling the confederate, “Hey, the other room is now open. You can go to the other room.” The confederate immediately left the room, pretending nothing had happened. During the subsequent minute, the research assistant simply resumed her drawing without saying anything. This period offered the child an opportunity to spontaneously report the transgression she/he had observed to the research assistant. If the child had not spontaneously reported the transgression, the research assistant asked the child, “Was everything OK? Please tell me what happened while I was away.”

In accordance with the approved ethics protocol, a debriefing was followed once data collection was complete. At the debriefing, the main aim of the session was clarified to both parents and children. Each child was told that the person who they met was acting out a scenario, and emphasized that books borrowed from the library should be taken care of, so taking pages out of a book is not allowed. After the debriefing, parents were given an opportunity to decide whether they would like to have the child’s data included in the study. Nobody wished to withdraw their child’s data from the study.

The video recordings were analysed with a coding scheme adapted from the previous studies9,12,26. The analysis focused on: (i) whether the child immediately protested to the adult confederate, and (ii) whether the child later reported what he/she saw to the researcher. Details of codes and rating methods are provided below alongside the data analysis procedure.

CMST and EEG recording

The CMST comprised three-scene cartoon scenarios depicting characters intentionally helping (e.g., helping another character making a snowman, helping another character after a fall) or intentionally harming another (e.g., kicking another character, destroying another character’s toy blocks) (see Fig. 1). These scenarios were validated with an independent group of American children between the ages of 3 and 8 years41. The version of the CMST used in this study included 120 trials (60 helping and 60 harming), 3 s in duration per scenario, randomly presented via E-Prime 2 (professional), with a jittered inter-trial interval of 500–1500 ms35,41. In both conditions (Helping vs. Harming), the shapes and colours of characters were counterbalanced. ERPs were time-locked to the onset of the second picture. EEG data were collected from a 32-channel active electrode system (actiCHamp+Brain Vision Recorder, Brain Products, Germany) at 2 kHz, referenced to Cz, and all electrodes impedances were kept below 30 kOhms.


Children’s social and emotional disposition were collected with the Very Short Form of the Children’s Behaviour Questionnaire (CBQ-VSF)67. The CBQ-VSF is a 36-item observer-report questionnaire, representing three broad dimensions of temperament observable in young children: (a) Surgency (12 items, α = 0.58), (b) Negative Affect (12 items, α = 0.53) and (c) Effortful Control (12 items, α = 0.62). Parents rated the degree to which each item, reflecting a type of behaviour, applied to his/her child over the past six months on a 7-point Likert scale (1 = extremely untrue to 7 = extremely true; NA = does not apply).

Parents’ dispositional empathy was assessed with the Questionnaire of Cognitive and Affective Empathy (QCAE)68. The QCAE is a 31-item self-report questionnaire, measuring (a) Cognitive Empathy (19 items, α = 0.90) and (b) Affective Empathy (12 items, α = 0.77). Items were rated by level of agreement using a 4-point Likert scale (1 = strongly disagree to 4 = strongly agree).

Parents’ values on justice were assessed with the Justice Sensitivity Short Scales (JSSS)69. The JSSS measures justice sensitivity from the Victim (2 items, α = 0.79), Observer (2 items, α = 0.70), Beneficiary (2 items, α = 0.78) and Perpetrator (2 items, α = 0.85) perspectives. Response options for the justice sensitivity scales raged from 0 (totally disagree) to 5 (totally agree).

Data analysis

Coding and analysis of children’s reaction

Children’s protest reactions were coded into five different categories and scores were assigned as illustrated in Table 3. Reliability on a random 25% of the sample was excellent (κ = 0.95 for whether or not a child protested; κ = 0.78 for the sub-category assigned for a child).

Table 3 Coding scheme for protest behaviours.

Children’s reports were coded into three categories as described in Table 4. Reliability on a random 25% of the sample was excellent (κ = 1.00).

Table 4 Coding scheme for reporting behaviours.

ERPs data analysis

EEG data pre-processing was performed with Brain Vision Analyser 2.0 software (Brain Products). Data were down-sampled offline (256 Hz) and referenced to the average of all electrodes. Following referencing, EEG data were filtered using an IIR filter of 1–30Hz. Artefact rejection was carried out using a −150 to 150 μV threshold and visual inspection. Gratton-Coles ocular artefact correction was performed on EEG data to identify and correct ocular movements, seeded off electrode Fp170. Following basic EEG pre-processing, data were segmented according to trial type (harming/helping). Epochs were created with 200 ms baselines and 1000 ms stimulus presentation following the onset of the second stimulus frame. Epochs were averaged by trial type and were baseline corrected. Corrected averages per individual were combined for a grand average per individual and ERPs were analysed across subjects by trial types and reactions shown at the experiment (protest vs. non-protest). Only children who had at least 20 artefact-free trials for each trial type (Harming and Helping) were included in the grand average and further analyses (N = 42).

In line with the hypotheses, the following latency windows were investigated: the posterior N1 (100–175 ms), the central P2 (150–350 ms), the frontal N2 (200–400 ms) and the central LPP (400–600 ms). Posterior N1, though not directly related to hypotheses, was tested in order to rule out group differences stemming from initial attention and early encoding of visual information71,72. For each region, three electrodes were included in the averaging and analyses: the frontal region included F3, Fz, F4; the central region included C3, Cz, C4; the posterior region included P3, Pz, P4 sensors. In order to avoid inflating Type, I error rate, the choice of region and electrodes was made a priori, based on existing literature41,42,43,50,71,72. Within each region, chosen three electrodes to have relatively less influence from potential artefacts (e.g. eye and facial muscle movements) and least overlap from adjacent regions. Given no statistical difference in amplitudes across three sites within each region, amplitudes were averaged for each region. With our interest in understanding how the implicit evaluation of norm violations might contribute to protesting and reporting activities, we focused our analyses on the difference of mean amplitude values in two conditions (Help vs. Harm), then compared the extent of difference between the two groups (Protester vs. Non-Protester; Reporters vs. Non-Reporters) using a repeated measure analysis of variance (ANOVA). Greenhouse–Geisser corrections were applied when assumptions of sphericity were not met. For ANOVAs, Cohen’s f was used to measure effect sizes. Post-hoc comparisons or simple effect analyses were performed with a Bonferroni’s correction and Cohen’s d was reported.

Bayesian structural equation modelling

In order to provide a comprehensive account of how parents’ dispositions are linked to individual differences in the neural computations implicated in costly moral behaviour, structural equation modelling (SEM) was conducted, utilizing AMOS 2658. In this analysis, only Protest behaviour was included as Report behaviour did not show any association with child’s ERP data. For ERP data, we focused our analyses on the difference waves generated by subtracting the helping from the harming condition (i.e. P2diff = P2Harm–P2Help). Given that the child’s protest behaviour is categorical data that does not follow the normal distribution, Bayesian SEM was undertaken in order to test three models: (a) Within-child (brain–behaviour) model that includes child’s ERP and protest behaviour only (Model 1); (b) intergenerational model that adds parent’s dispositions to the child’s brain–behaviour model (Model 2); c) refined intergenerational model that adds child’s disposition to the second model (Model 3). Model parameters were estimated using the AMOS 26 Bayesian method that uses Markov-chain Monte Carlo (MCMC) sampling to generate full posterior distributions of model parameters. We have set up 500 burn-in and tried 25,000 MCMC runs after burn-in. i.e., 25,500 total MCMC run. Regarding prior, we adapted diffuse priors for all parameters with the default prior distribution in AMOS 26, i.e., uniform from –3.4 × 10−38 to 3.4 × 1038 to each parameter58. Regarding the model fit test, posterior predictive p-values (ppp) was considered73. There is some agreement that ppp values close to 0.50 imply a good fit while values <0.05 may indicate poor fit73,74.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The datasets presented in this article are not readily available because ethics to place data in a public repository was not obtained from the children and families who participated in this study (HREC Approval No. 2017/468). The data may be made available on request to the authors after review of the request by the Human Research Ethics Committee (HREC). Requesters should submit a data analysis plan before requesting the data. Requests to access the datasets should be directed to the corresponding author.

Code availability

In this study, no custom scripts were used for the analysis. Syntax and commands used in licensed software are available at


  1. Fehr, E. & Gächter, S. Altruistic punishment in humans. Nature 415, 137–140 (2002).

    CAS  PubMed  Google Scholar 

  2. Fehr, E. & Fischbacher, U. The nature of human altruism. Nature 425, 785–791 (2004).

    Google Scholar 

  3. Kenward, B. & Östh, T. Enactment of third-party punishment by 4-year-olds. Front. Psychol. 3, 373 (2012).

    PubMed  PubMed Central  Google Scholar 

  4. Geraci, A. & Franchin, L. Is defensive behavior a subtype of prosocial behaviors? Front. Psychol. (2021).

  5. Kanakogi, Y. et al. Preverbal infants affirm third-party interventions that protect victims from aggressors. Nat. Hum. Behav. 1, 0037 (2017).

    Google Scholar 

  6. Geraci, A. How do toddlers evaluate defensive actions toward third parties? Infancy 25, 910–926 (2020).

    PubMed  Google Scholar 

  7. Geraci, A. & Surian, L. Toddlers’ expectations of third‐party punishments and rewards following an act of aggression. Aggress. Behav. 47, 521–529 (2021).

    PubMed  Google Scholar 

  8. Riedl, K., Jensen, K., Call, J. & Tomasello, M. Restorative justice in children. Curr. Biol. 25, 1731–1735 (2015).

    CAS  PubMed  Google Scholar 

  9. Vaish, A., Missana, M. & Tomasello, M. Three-year-old children intervene in third-party moral transgressions. Br. J. Dev. Psychol. 29, 124–130 (2011).

    PubMed  Google Scholar 

  10. Yudkin, D. A., Van Bavel, J. J. & Rhodes, M. Young children police group members at personal cost. J. Exp. Psychol. Gen. 149, 182–191 (2020).

    PubMed  Google Scholar 

  11. Schmidt, M. F. H. & Tomasello, M. Young children enforce social norms. Curr. Dir. Psychol. Sci. 21, 232–236 (2012).

    Google Scholar 

  12. Heyman, G. D., Chiu Loke, I. & Lee, K. Children spontaneously police adults’ transgressions. J. Exp. Child Psychol. 150, 155–164 (2016).

    PubMed  Google Scholar 

  13. Rossano, F., Rakoczy, H. & Tomasello, M. Young children’s understanding of violations of property rights. Cognition 121, 219–227 (2011).

    PubMed  Google Scholar 

  14. Jordan, J. J., McAuliffe, K. & Warneken, F. Development of in-group favoritism in children’s third-party punishment of selfishness. Proc. Natl Acad. Sci. USA 111, 12710–12715 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  15. McAuliffe, K., Jordan, J. J. & Warneken, F. Costly third-party punishment in young children. Cognition 134, 1–10 (2015).

    PubMed  Google Scholar 

  16. Hardecker, S., Schmidt, M. F. H., Roden, M. & Tomasello, M. Young children’s behavioral and emotional responses to different social norm violations. J. Exp. Child Psychol. 150, 364–379 (2016).

    PubMed  Google Scholar 

  17. Yucel, M. & Vaish, A. Young children tattle to enforce moral norms. Soc. Dev. 27, 924–236 (2018).

    Google Scholar 

  18. Dahl, A. New beginnings: an interactionist and constructivist approach to early moral development. Hum. Dev. 61, 232–247 (2018).

    PubMed  Google Scholar 

  19. Tomasello, M. How children come to understand false beliefs: a shared intentionality account. Proc. Natl Acad. Sci. USA 115, 8491–8498 (2018).

    CAS  PubMed  PubMed Central  Google Scholar 

  20. Crick, N. R. & Dodge, K. A. Social information processing mechanisms in reactive and proactive aggression. Child Dev. 67, 993–1002 (1996).

    CAS  PubMed  Google Scholar 

  21. Arsenio, W. F. & Kramer, R. Victimizers and their victims: Children’s conceptions of the mixed emotional consequences of moral transgressions. Child Dev. 63, 927 (1992).

    Google Scholar 

  22. Nunner-Winkler, G. & Sodian, B. Children’s understanding of moral emotions. Child Dev. 59, 1338 (1988).

    Google Scholar 

  23. Atance, C. & O’neill, D. The emergence of episodic future thinking in humans. Learn. Motiv. 36, 126–144 (2005).

    Google Scholar 

  24. Engelmann, J. M. & Rapp, D. J. The influence of reputational concerns on children’s prosociality. Curr. Opin. Psychol. 20, 92–95 (2018).

    PubMed  Google Scholar 

  25. Shaw, A. et al. Children develop a veil of fairness. J. Exp. Psychol. Gen. 143, 363–375 (2014).

    PubMed  Google Scholar 

  26. Brandl, J. L., Esken, F., Priewasser, B. & Rafetseder, E. Young children’s protest: what it can (not) tell us about early normative understanding. Phenomenol. Cogn. Sci. 14, 719–740 (2015).

    Google Scholar 

  27. Turiel, E. Moral development in the early years: when and how. Hum. Dev. 61, 297–308 (2018).

    Google Scholar 

  28. Lyons, D. E., Damrosch, D. H., Lin, J. K., Macris, D. M. & Keil, F. C. The scope and limits of overimitation in the transmission of artefact culture. Philos. Trans. R. Soc. Lond. B Biol. Sci. 366, 1158–1167 (2011).

    PubMed  PubMed Central  Google Scholar 

  29. Kenward, B. Over-imitating preschoolers believe unnecessary actions are normative and enforce their performance by a third party. J. Exp. Child Psychol. 112, 195–207 (2012).

    PubMed  Google Scholar 

  30. Keupp, S., Bancken, C., Schillmöller, J., Rakoczy, H. & Behne, T. Rational over-imitation: preschoolers consider material costs and copy causally irrelevant actions selectively. Cognition 147, 85–92 (2016).

    PubMed  Google Scholar 

  31. Dahl, A. & Freda, G. F. How young children come to view harming others as wrong: a developmental analysis. In Social Cognition: Development Across The Life Span (eds. Sommerville, J. A. & Decety, J.) 151–184 (Routledge, 2017).

  32. Dahl, A. & Tran, A. Q. Vocal tones influence young children’s responses to prohibitions. J. Exp. Child Psychol. 152, 71–91 (2016).

    PubMed  PubMed Central  Google Scholar 

  33. Yoder, K. & Decety, J. Spatiotemporal neural dynamics of moral judgment: a high-density ERP study. Neuropsychologia 60, 39–45 (2014).

    PubMed  Google Scholar 

  34. Yoder, K. & Decety, J. The Good, the bad, and the just: justice sensitivity predicts neural response during moral evaluation of actions performed by others. J. Neurosci. 34, 4161–4166 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  35. Cowell, J. M. & Decety, J. Precursors to morality in development as a complex interplay between neural, socioenvironmental, and behavioural facets. Proc. Natl Acad. Sci. USA 112, 12657–12662 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  36. Decety, J., Steinbeis, N. & Cowell, J. M. The neurodevelopment of social preferences in early childhood. Curr. Opin. Neurobiol. 68, 23–28 (2021).

    CAS  PubMed  Google Scholar 

  37. Van Berkum, J. J. A., Holleman, B., Nieuwland, M., Otten, M. & Murre, J. Right or Wrong: brain’s fast response to morally objectionable statements. Psychol. Sci. 20, 1092–1099 (2009).

    PubMed  Google Scholar 

  38. Sarlo, M. et al. Temporal dynamics of cognitive-emotional interplay in moral decision-making. J. Cogn. Neurosci. 24, 1018–1029 (2012).

    PubMed  Google Scholar 

  39. Leuthold, H., Kunkel, A., Mackenzie, I. G. & Filik, R. Online processing of moral transgressions: ERP evidence for spontaneous evaluation. Soc. Cogn. Affect. Neurosci. 10, 1021–1029 (2015).

    PubMed  PubMed Central  Google Scholar 

  40. Sarlo, M. & Munafò, M. When faces signal danger: event-related potentials to emotional facial expressions in animal phobics. Neuropsychobiology 62, 235–244 (2010).

    PubMed  Google Scholar 

  41. Cowell, J. & Decety, J. The neuroscience of implicit moral evaluation and its relation to generosity in early childhood. Curr. Biol. 25, 93–97 (2015).

    CAS  PubMed  Google Scholar 

  42. Meidenbauer, K. L., Cowell, J. M. & Decety, J. Children’s neural processing of moral scenarios provides insight into the formation and reduction of in-group biases. Dev. Sci. 21, e12676 (2018).

    PubMed  Google Scholar 

  43. Wu, L. & Kim, M. See, touch, and feel: Enhancing young children’s empathy learning through a tablet game. Mind Brain Educ. 13, 341–351 (2019).

    Google Scholar 

  44. Batson, C. D. The empathy-altruism hypothesis: Issues and implications. In Empathy: From Bench to Bedside (ed. Decety, J.) 41–54 (MIT Press, 2012).

  45. Williams, A., O’Driscoll, K. & Moore, C. The influence of empathic concern on prosocial behaviour in children. Front. Psychol. 5, 1–8 (2014).

    Google Scholar 

  46. FeldmanHall, O., Dalgleish, T., Evans, D. & Mobbs, D. Empathic concern drives costly altruism. NeuroImage 105, 347–356 (2015).

    PubMed  Google Scholar 

  47. Decety, J., Bartal, I., Uzefovsky, F. & Knafo-Noam, A. Empathy as a driver of prosocial behaviour: Highly conserved neurobehavioural mechanisms across species. Philos. Trans. R. Soc. Lond. B Biol. Sci. 371, 20150077 (2016).

    PubMed  PubMed Central  Google Scholar 

  48. Ashar, Y. K., Andrews-Hanna, J. R., Dimidjian, S. & Wager, T. D. Empathic care and distress: Predictive brain markers and dissociable brain systems. Neuron 94, 1263–1273.e4 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

  49. Sarlo, M., Lotto, L., Rumiati, R. & Palomba, D. If it makes you feel bad, don’t do it! Egoistic rather than altruistic empathy modulates neural and behavioral responses in moral dilemmas. Physiol. Behav. 130, 127–134 (2014).

    CAS  PubMed  Google Scholar 

  50. Lewis, M. et al. Changes in the neural bases of emotion regulation associated with clinical improvement in children with behavior problems. Dev. Psychopathol. 20, 913–939 (2008).

    PubMed  Google Scholar 

  51. Tucker, D. M. et al. Corticolimbic mechanisms in emotional decisions. Emotion 3, 127–149 (2003).

    PubMed  Google Scholar 

  52. Luu, P., Tucker, D. M., Derryberry, D., Reed, M. & Poulsen, C. Activity in human medial frontal cortex in emotional evaluation and error monitoring. Psychol. Sci. 14, 47–53 (2003).

    PubMed  Google Scholar 

  53. Lewis, M. D., Todd, R. M. & Honsberger, M. Event-related potential measures of emotion regulation in early childhood. NeuroReport 18, 61–65 (2007).

    PubMed  Google Scholar 

  54. Todd, R. M., Lewis, M. D., Meusel, L.-A. & Zelazo, P. D. The time course of social–emotional processing in early childhood: ERP responses to facial affect and personal familiarity in a go-no-go task. Neuropsychologia 46, 595–613 (2008).

    PubMed  Google Scholar 

  55. Cheng, Y., Chen, C. & Decety, J. An EEG/ERP investigation of the development of empathy in early and middle childhood. Dev. Cogn. Neurosci. 10, 160–169 (2014).

    PubMed  PubMed Central  Google Scholar 

  56. Decety, J., Michalska, K. J. & Kinzler, K. D. The contribution of emotion and cognition to moral sensitivity: a neurodevelopmental study. Cereb. Cortex 22, 209–230 (2012).

    PubMed  Google Scholar 

  57. Decety, J. & Cowell, J. M. Interpersonal harm aversion as a necessary foundation for morality: a developmental neuroscience perspective. Dev. Psychopathol. 30, 153–164 (2018).

    PubMed  Google Scholar 

  58. Arbuckle J. L. Amos (Version 26.0) [Computer Program] (IBM SPSS, 2019).

  59. Tan, P. Z., Oppenheimer, C. W., Ladouceur, C. D., Butterfield, R. D. & Silk, J. S. A review of associations between parental emotion socialization behaviors and the neural substrates of emotional reactivity and regulation in youth. Dev. Psychol. 56, 516–527 (2020).

    PubMed  PubMed Central  Google Scholar 

  60. Schmidt, M. F., Rakoczy, H. & Tomasello, M. Young children enforce social norms selectively depending on the violator’s group affiliation. Cognition 124, 325–333 (2012).

    PubMed  Google Scholar 

  61. LeDoux, J. Synaptic Self: How our Brains Become Who We Are (Viking, 2002).

  62. Kim, M., & Sankey, D. The Science of Learning and Development in Education (Cambridge University Press, 2022).

  63. Van Bavel, J. J., FeldmanHall, O. & Mende-Siedlecki, P. The neuroscience of moral cognition: from dual processes to dynamic systems. Curr. Opin. Psychol. 6, 167–172 (2015).

    Google Scholar 

  64. Gummerum, M., López‐Pérez, B., Van Dijk, E. & Van Dillen, L. When punishment is emotion‐driven: children’s, adolescents’, and adults’ costly punishment of unfair allocations. Soc. Dev. 29, 126–142 (2020).

    Google Scholar 

  65. Sankey, D. & Kim, M. Cultivating moral values in an age of neuroscience. In Neuroscience and education: A Philosophical Appraisal (ed. Joldersma, C.) 111–127 (Routledge, 2016).

  66. Bridwell, D. et al. Moving beyond ERP components: a selective review of approaches to integrate EEG and behaviour. Front. Hum. Neurosci. 12, 106–106 (2018).

    PubMed  PubMed Central  Google Scholar 

  67. Putnam, S. P. & Rothbart, M. K. Development of short and very short forms of the children’s behaviour questionnaire. J. Pers. Assess. 87, 103–113 (2002).

    Google Scholar 

  68. Reniers, R., Corcoran, R., Drake, R., Shryane, N. & Vӧllm, B. The QCAE: a questionnaire of cognitive and affective empathy. J. Pers. Assess. 93, 84–95 (2011).

    PubMed  Google Scholar 

  69. Baumert, A. et al. Measuring four perspectives of justice sensitivity with two items each. J. Pers. Assess. 96, 380–390 (2014).

    PubMed  Google Scholar 

  70. Gratton, G., Coles, M. G. H. & Donchin, E. A new method for off-line removal of ocular artefact. Clin. Neurophysiol. 55, 468–484 (1983).

    CAS  Google Scholar 

  71. Schupp, H., Junghöfer, M., Weike, A. & Hamm, A. The selective processing of briefly presented affective pictures: an ERP analysis. Psychophysiology 41, 441–449 (2004).

    PubMed  Google Scholar 

  72. Vogel, E. & Luck, S. The visual N1 component as an index of a discrimination process. Psychophysiology 37, 190–203 (2000).

    CAS  PubMed  Google Scholar 

  73. Lee, S. Structural Equation Modeling: A Bayesian Approach (John Wiley & Sons, 2007).

  74. Muthén, B. & Asparouhov, T. Bayesian structural equation modeling: a more flexible representation of substantive theory. Psychol. Methods 17, 313–335 (2012).

    PubMed  Google Scholar 

Download references


The researchers are grateful to the many families and children who participated in this project. We also thank the directors and teachers of the childcare centres, who assisted us in the recruitment of families. This research was supported by grants awarded to the first author from the Faculty of Arts and Social Sciences (FASS) at the University of Sydney.

Author information

Authors and Affiliations



M.K. and J.D. conceived and designed the study. M.K., L.W. and S.B. collected data. M.K. performed the statistical analysis and wrote the first version of the manuscript. J.D., L.W., S.B. and D.S. discussed the results and provided comments and critical editing, then approved the final version.

Corresponding author

Correspondence to Minkang Kim.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Kim, M., Decety, J., Wu, L. et al. Neural computations in children’s third-party interventions are modulated by their parents’ moral values. npj Sci. Learn. 6, 38 (2021).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI:


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing