Individual Differences in Prototypical Moral and Conventional Judgmentsand Children’s Proactive and Reactive Aggression

Marc Jambon and Judith G. SmetanaUniversity of Rochester

This article examined links between 4- and 6-year-olds’ (n = 101; Mage = 5.12 years, SD = 0.67; 53% male) abil-ity to distinguish moral and conventional transgressions along different criteria and teacher ratings of proac-tive and reactive aggression. Latent difference score modeling revealed that moral transgressions were judgedmore unacceptable and wrong independent of rules and authority than conventional violations, but significantvariability in moral–conventional distinctions was also observed. Proactive aggression was associated withless—and reactive aggression was associated with greater—differentiation in moral and conventional concepts.Proactive aggression was not associated with deficits in moral knowledge when other common assessments ofearly moral understanding were employed, highlighting the importance of using theoretically informed mea-sures of moral judgments and aggression.

Young children consider moral transgressionsinvolving harm and unfairness to be wrong basedon the negative consequences for others (Smetana,Jambon, & Ball, 2014). Nevertheless, many childrenengage in behaviors that harm others (Eisner &Malti, 2015). Despite clear conceptual overlap, rela-tively little contemporary research has examinedlinks between children’s moral judgments andaggression (Arsenio & Lemerise, 2004). Further-more, existing studies have typically failed to findmeaningful associations between aggression andchildren’s basic understanding of moral norms,leading researchers to downplay the importance ofprototypical moral judgments for children’s aggres-sive behavior. However, past research has oftenrelied on conceptual definitions and methodsincommensurate with contemporary approaches tomoral development and aggression. Integratingconcepts and methods from social domain theory(Smetana et al., 2014; Turiel, 1983) with functionalapproaches to aggression (Crick & Dodge, 1996; Lit-tle, Henrich, Jones, & Hawley, 2003), the presentstudy examined whether individual differences in

4- to 6-year-olds’ ability to distinguish moralityfrom convention along theoretically derived criteriawere differentially associated with proactive andreactive aggression.

Moral Judgments and Aggression

A central tenet of social domain theory (Smetanaet al., 2014; Turiel, 1983) is that, through socialinteractions, children develop an understanding ofmorality (i.e., norms pertaining to issues of harm,justice, and rights) as distinct from social conven-tions (i.e., agreed-upon expectations and rules foracceptable behavior within a given social context).To test this assertion, domain researchers havedrawn on philosophical and psychological defini-tions of morality and convention to examinewhether children use different criteria to evaluateevents in the two domains (referred to as criterionjudgments). This research (reviewed in Killen &Smetana, 2015; Smetana et al., 2014) has shown thatby 3–4 years of age, children judge straightforward,prototypical moral transgressions (e.g., hitting aninnocent victim) to be generalizably wrong acrossdifferent social contexts, even in the absence ofrules (rule independence) and adult prohibitions (au-thority independence). In contrast, conventional viola-tions (e.g., breaches of etiquette) are considered

Marc Jambon is now at the Department of Psychology, Univer-sity of Calgary.

We would like to thank the day-care directors and school prin-cipals for their enthusiasm and cooperation with this project, aswell as the teachers and children who participated. This researchis based in part on the Marc Jambon’s dissertation, submitted inpartial fulfillment of the requirements for the PhD at the Univer-sity of Rochester. This research was partially funded through aUniversity of Rochester Departmental Dissertation Grant.

Correspondence concerning this article should be addressed toMarc Jambon, Department of Psychology, University of Calgary,Calgary, AB, Canada, 1235D, T2N 1N4. Electronic mail may besent to marc.jambon@ucalgary.edu.

© 2017 The AuthorsChild Development © 2017 Society for Research in Child Development, Inc.All rights reserved. 0009-3920/2017/xxxx-xxxxDOI: 10.1111/cdev.12757

Child Development, xxxx 2017, Volume 00, Number 0, Pages 1–17

acceptable if the context is changed or if the rulesor prohibitions against them are removed.

Despite evidence demonstrating the earlyemergence of moral knowledge, many childrennevertheless harm others. Individual differences inaggression are evident during the preschool years,become increasingly stable with age, and are associ-ated with a wide range of psychosocial deficits(e.g., poor peer relationships, low academic achieve-ment, maladjustment; Eisner & Malti, 2015). Socialinformation processing (SIP; Dodge, Coie, &Lynam, 2006) research has also shown that biasesin children’s and adolescents’ understanding andprocessing of social information contribute to thedevelopment and maintenance of aggression. Forinstance, aggressive youth are more sensitive toexternal threat cues than their nonaggressive peers,often believing that others possess hostile intentionsin ambiguous conflict situations (i.e., hostile attribu-tion biases).

Relatively less research has examined whetheraggression is linked to variations in moral cogni-tions. Studies have commonly assessed moral eval-uations in response to hypothetical stories depictingharmful behaviors in different situations. A consis-tent finding is that moral judgments of straightfor-ward, unprovoked harm are not associated withindividual differences in aggression (Astor, 1994;Gasser & Keller, 2009; Gasser, Malti, & Gutzwiller-Helfenfinger, 2012; Malti, Gasser, & Buchmann,2009); both aggressive and nonaggressive childrenconsider such acts to be unacceptable. When pre-sented with more contextualized scenarios, how-ever, aggressive youth are more likely to consideracts of harm in response to perceived threat orprovocation to be acceptable and justified (Astor,1994; Gasser et al., 2012; Huesmann & Guerra,1997). Similar results have been observed in chil-dren and adolescents exposed to extreme politicalviolence (Posada & Wainryb, 2008).

Researchers from diverse perspectives havedrawn on these findings to conclude that, by thepreschool years, nearly all children share the coremoral belief that intentionally harming innocentothers is categorically wrong (Arsenio & Lemerise,2004; Gasser et al., 2012; Nunner-Winkler, 2013).Some have gone further to claim that judgments ofprototypical moral transgressions reflect a purelycognitive or informational knowledge of rules thathas no relevance for children’s social behavior(Hawley & Geldhof, 2012; Nunner-Winkler, 2013).Consequently, the focus has shifted away from pro-totypical moral judgments to understand how otherfactors, such as social-cognitive biases, deficits in

moral motivation and emotions, and atypical pat-terns of reasoning about complex moral events maylead children to engage in harmful acts despite“knowing” it is wrong to do so (Arsenio & Lemer-ise, 2004; Eisner & Malti, 2015).

The conclusion that all children (including thosewho harm others) possess the same basic moralknowledge may be premature, however, given thetypes of measures used in previous research toassess moral judgments. Specifically, studies exam-ining links between prototypical moral judgmentsand aggression have typically asked children aboutthe general acceptability of unprovoked moral vio-lations (e.g., “is it ok to hit someone?” Astor, 1994;Huesmann & Guerra, 1997; Malti et al., 2009) orhave assessed moral criterion judgments indepen-dent of conventional evaluations (e.g., “is it ok tohit if there were no rules/if a teacher allowed it?”;Gasser et al., 2012; Malti, Gasser, & Gutzwiller-Hel-fenfinger, 2010).

From a domain perspective, however, theseassessments do not adequately capture children’sunderstanding of morality as a distinct form ofsocial knowledge. Because prototypical moral trans-gressions entail violations of accepted norms andpractices, simply asking whether it is generallywrong to harm others does not allow us to deter-mine whether children’s judgments are based onmoral or nonmoral concerns (e.g., fear of punish-ment; Killen & Smetana, 2015; Turiel, 1983). Obtain-ing justifications for judgments provides importantinsights when studying older youth, but preschooland early school-age children often have difficultyproviding elaborated and codable responses (Sme-tana et al., 1999). Criterion judgments therefore pro-vide an important window into young children’sconceptual understanding of morality.

Yet focusing solely on criterion judgments ofmoral events—rather than the ability to differentiatemorality and convention—may also limit our capac-ity to determine whether children understand moralobligations to be distinct from other norms. Eventoddlers clearly recognize that moral violations areprohibited and considered unacceptable by peersand adults (Dunn, 2014). Given that young childrenalready have difficulty with counterfactual reason-ing and attributing unusual beliefs to others(Conry-Murray, 2013), a child who equates moralitywith external rules and authority commands mayerr on the side of caution and “disapprove” ofharm even if told that the normal restrictions oradult prohibitions have been removed. Indeed,moral development researchers have long expressedsimilar concerns with socially desirable responding

2 Jambon and Smetana

when studying children’s sociomoral judgments(Dunn, 2014; Piaget, 1932/1965; Rutland, Cameron,Milne, & McGeorge, 2005). However, a child whoconflates moral and conventional rules is likely toevaluate both types of acts in a similar manner(Blair, 1995). Thus, focusing on the ability to distin-guish moral and conventional events along differentcriteria may provide researchers with an importanttool for addressing this challenge.

The hypothesis that aggression may be linkedto deficits in children’s moral understanding, oper-ationalized as the capacity to differentiate moralityand convention, implies that children differ fromone another in this ability. To date, however, indi-vidual differences and their implications for socialfunctioning have received little systematic atten-tion. Deficits in prototypical domain-related judg-ments have been associated with disruptive andantisocial behaviors at later ages (Blair, 1995; Blair,Monson, & Frederickson, 2001; Nucci & Herman,1982). Adolescents and adults who are high inpsychopathic traits—characterized by empathicdeficits and antisocial behavior—evaluate prototyp-ical moral and conventional transgressions to bewrong and deserving of punishment, yet theyhave considerable difficulty distinguishing theseevents in their criterion judgments (Blair, 1995;Blair et al., 2001; Dolan & Fullam, 2010). Impor-tantly, these deficits are not because psychopathsconsistently treat acts of harm and injustice as con-ventional issues; several studies have found thatpsychopaths judge both moral and conventionalviolations to be categorically wrong even in theabsence of rules and prohibitions (i.e., as moralissues; Blair, 1995; Dolan & Fullam, 2010). As Blair(1995) has noted, individuals who fail to distin-guish moral and nonmoral rules may simplyrespond in a manner they deem socially desirableor expected.

These findings, although limited to atypical sam-ples of adolescents and adults, lend support to theassertion that focusing on the ability to conceptu-ally distinguish between morality and conventionmay provide a more appropriate assessment of chil-dren’s basic moral knowledge than initial moralevaluations or moral criterion judgments alone.This ability is considered a milestone in children’ssocial-cognitive development that is believed toprovide an early foundation for later moral devel-opment (Killen & Smetana, 2015; Nucci, 2009; Sme-tana et al., 2014; Turiel, 1983). Research examiningassociations between domain distinction ability andaggressive behavior in younger, less extreme popu-lations is greatly needed.

Functional Approaches to Aggression

In addition to using appropriate assessments ofmoral understanding, it is also critical to considerthe underlying purpose, or function, that aggressionmay serve (Crick & Dodge, 1996; Little et al., 2003).Although past studies have often examined moraljudgments in relation to children’s general tenden-cies toward aggressive or disruptive behavior, dif-ferentiating between reactive and proactivefunctions of aggression may be crucial for under-standing links with moral functioning (Arsenio &Lemerise, 2004). Reactive aggression reflects a “hotblooded” response to perceived threat or frustra-tion, driven by intense affect and emotional dysreg-ulation. In contrast, proactive aggression entails the“cold blooded,” goal-directed use of aggressionaimed at achieving rewards or dominating others.

Although the two types of aggression often co-occur (Cui, Colasante, Malti, Ribeaud, & Eisner,2016; Poulin & Bouivin, 2000), they are neverthelessuniquely correlated with different social-cognitivemechanisms, affective processes, and sociobehav-ioral outcomes (Card & Little, 2006; Hubbard,McAuliffe, Morrow, & Romano, 2010; Vitaro,Brendgen, & Barker, 2006). For instance, reactiveaggression is associated with physiological over-arousal and dysregulated emotions (particularlyanger), biased SIP (e.g., hostile attribution biases),and poor social relationships (e.g., peer rejection).These findings align with the traditional “social-skills deficit” conceptualization of antisocial behav-ior (Eisner & Malti, 2015; Sutton, Smith, & Swetten-ham, 1999). In contrast, when studies control forlevels of reactive aggression, proactive aggression isoften uncorrelated with indicators of maladjustmentor is positively associated with social competence,such as intelligence, popularity, and social skills(see Card & Little, 2006). A number of scholarshave drawn on evolutionary and ethological frame-works to suggest that the effective use of aggressionto successfully obtain resources, rewards, andpower in social groups can be considered adaptiveand socially competent (Hawley, 2014; Stoltz, Cil-lessen, van den Berg, & Gommans, 2016; Suttonet al., 1999).

Moral Judgments and Proactive and ReactiveAggression

Although reactive aggression may be more clo-sely tied to maladaptive social functioning thanproactive aggression, the findings differ with regardto moral functioning. Children possessing social-

Domain Distinctions and Aggression 3

cognitive biases and dysregulated anger and frus-tration may respond to provocations and (mis)per-ceived threats with force, yet there is little evidencethat reactively aggressive children lack empathy ora basic appreciation for the wrongness of unpro-voked harm (Arsenio & Lemerise, 2004; Arsenio,Adams, & Gold, 2009; Orobio de Castro, Verhulp,& Runions, 2012). Indeed, when others’ intentionsand motivations are clear, typically developing chil-dren also consider aggression performed in self-defense or to protect others as more acceptable andjustified than selfishly motivated harm (Jambon &Smetana, 2014). In contrast, studies have found thatchildren and adolescents who engage in proactiveaggression are more likely than other youth to pri-oritize their own selfish desires at the expense ofothers, have greater confidence in their ability touse aggression to obtain rewards, and are less likelyto feel guilt or remorse after wrongdoing (Arsenioet al., 2009; Crick & Dodge, 1996; Hubbard et al.,2010). Longitudinal research has also found proac-tive aggression to be uniquely associated with aheightened risk of developing psychopathic traitsand violent tendencies later in life (reviewed inHubbard et al., 2010). Given that the deliberate useof coercion and harm for self-gain represents a pro-totypical violation of moral norms, children exhibit-ing proactively aggressive tendencies may also beless advanced in their understanding of morality.However, the few studies to have empirically testedthis hypothesis have resulted in mixed findings.

Drawing on an evolutionary framework, Hawleyand Geldhof (2012) examined links betweenpreschoolers’ use of coercive strategies to obtainand control resources (i.e., proactive aggression)and their moral functioning. They found that proac-tively aggressive youth displayed low levels of guiltand were adept at manipulating others (e.g., tat-tling; being nice to teachers after caught transgress-ing), yet they were also more advanced thannonaggressive children in their judgments ofstraightforward moral transgressions. The research-ers concluded that proactive aggressors’ moreadvanced moral knowledge reflects an “enlightenedself-interest” that allows them to successfully navi-gate the social hierarchy. The findings regardingguilt were interpreted as evidence that emotions aremore important than cognitions for guiding chil-dren’s social behavior (also see Hawley, 2003).

Although this interpretation is intriguing,methodological limitations call it into question.Specifically, Hawley and Geldhof (2012) ascribedadvanced moral reasoning to children who judgedprototypical moral transgressions such as stealing

to be unacceptable based on any explanation,including nonmoral ones, that could be classified asrule or emotion based (“it is against the rules”; “itwas not nice”). In contrast, children were scored asless advanced if they judged transgressions to bepermissible or could not provide an explanation.Given that preschoolers rarely judge prototypicalmoral violations to be generally acceptable andoften have difficulty providing verbal justifications,these findings must be viewed with caution.

Although judgments were not assessed, Arsenioet al. (2009) examined whether proactive and reac-tive aggression were differentially associated withSIP abilities, moral emotion attributions, and rea-soning in a sample of low socioeconomic status(SES) adolescents. Proactive aggression was associ-ated with higher verbal skills, greater expectationsfor rewards and positive emotions followingaggression, and fewer references to others’ welfarewhen explaining their emotions. In contrast, reac-tive aggression was linked with greater hostile attri-bution biases and delays in attention and verbalability, but it was not associated with deficits inmoral emotions and reasoning.

Adopting a person-centered approach, a recentlongitudinal investigation (Cui et al., 2016) exam-ined whether moral reasoning was associated withdistinct developmental trajectories of proactive andreactive aggression from middle childhood to earlyadolescence. The researchers found that low levelsof moral reasoning at 8 years of age predictedmembership in a class of youth exhibiting high,stable levels of reactive and proactive aggressionover time. In contrast, participants who showedelevated rates of reactive (but not proactive)aggression did not differ from nonaggressiveyouth in their reasoning about prototypical moralviolations.

These studies provide some support for thenotion that proactive (but not reactive) aggressionmay be associated with deficits in moral knowl-edge, but this hypothesis remains to be tested withmore precise and theoretically grounded measuresof moral understanding. That proactive aggressorsare often verbally and socially skilled (Arsenioet al., 2009; Hawley, 2014) may help to account forthe more advanced moral judgments observed inpreschoolers (Hawley, 2003; Hawley & Geldhof,2012). Children who lack a basic appreciation forissues of harm, justice, or rights, yet who areattuned to the wants and needs of others when itsuits their desires, may be particularly adept at “ap-pearing moral” when assessments that fail to cap-ture their conceptual understanding are used.

4 Jambon and Smetana

Studies employing appropriate methods areneeded to better understand the role of moral judg-ments in the development of proactive and reactiveaggression in early childhood. This is criticallyimportant, given that individual differences inmoral cognitions and their implications for chil-dren’s broader social functioning are poorly under-stood. Although antisocial behavior has beenextensively studied in disadvantaged and high-riskpopulations, researchers have increasingly empha-sized the importance of considering the develop-mental processes associated with aggressive andcoercive behaviors in more normative, typicallydeveloping populations (Arsenio & Lemerise, 2004;Hawley, 2014). Gaining greater insight into theunique correlates of proactive and reactive aggres-sion in early childhood may also help inform inter-vention efforts aimed at targeting the specificdeficits underlying the aggression subtypes (Eisner& Malti, 2015; Vitaro et al., 2006).

The Current Study

The primary goal of the present study was toexamine whether young children’s ability to distin-guish prototypical moral and conventional eventsalong theoretical criteria was differentially associ-ated with teacher ratings of proactive and reactiveaggression. We focused on 4- to 6-year-olds becausethis constitutes an important period for the devel-opment of moral concepts (Smetana et al., 2014)and individual differences in proactive and reactiveaggression (Murray-Close & Ostrov, 2009). Consis-tent with past social domain research, we expectedthat children would distinguish moral and conven-tional norms in their criterion judgments. Impor-tantly, however, we also expected to findsignificant variability in this ability.

Past research has shown teacher reports of proac-tive and reactive aggression to be highly correlated(Card & Little, 2006; Poulin & Bouivin, 2000).Although the two subtypes often co-occur, much ofthis overlap stems from the fact that questionnairesassessing reactive and proactive functions use simi-lar wording and include items reflecting the sameunderlying behavioral form (e.g., hitting; Littleet al., 2003). Despite this methodological concern,the two functions consistently load as distinct con-structs and, when controlling for their shared vari-ance, demonstrate unique patterns of associationwith various behavioral and social-cognitive factors(Arsenio et al., 2009; Card & Little, 2006; Hubbardet al., 2010; Poulin & Bouivin, 2000). We thereforeexpected that, after controlling for reactive

aggression, children rated higher in proactiveaggression would make less of a distinctionbetween morality and convention. However, wedid not expect domain distinction ability to be asso-ciated with reactive aggression after controlling forproactive aggression.

A second important goal of the study was toexamine whether domain distinction ability pro-vided a more appropriate assessment of prototypi-cal moral understanding compared to the two mostcommon measures used in past research: (a) initialmoral judgments and (b) moral criterion judgmentsexcluding evaluations of conventions. Consistentwith prior studies utilizing these methods (e.g.,Astor, 1994; Gasser et al., 2012; Huesmann &Guerra, 1997; Malti et al., 2009), we did not expecteither assessment to be significantly associated withproactive aggression.

Finally, prior research has found little evidencefor gender differences in prototypical moral andconventional judgments (Smetana et al., 2014).Similarly, although boys typically exhibit highermean levels of overt aggression than girls, linksbetween aggression subtypes and other measuresof children’s social and emotional developmentalso do not systematically differ by gender (Card& Little, 2006). We therefore tested effects for ageand gender but did not expect to find differencesin boys and girls or for younger and older chil-dren.

Method

Participants/Sample

The sample consisted of 101 four- to six-year-olds (53% male; Mage = 5.12 years, SD = 0.67) andtheir teachers (n = 18; 94% female; Mage =35 years, SD = 15). Participants were recruitedfrom nine schools and day-care centers servinglower to middle SES families in a moderatelysized city in the northeastern United States. Datawere collected between May 2014 and October2015. The racial and ethnic composition of thesample reflected the demographics of the popula-tion from which they were drawn. Children were65% White, 23% Black, 9% biracial, 1% Asian,and 2% unknown; 14% were of Hispanic/Latino(a) origin. Two additional child participantsfailed to pass the interview scale training (seebelow) on multiple occasions and were subse-quently dropped from the study. Teachers were75% White, 19% Black, and 6% Pacific Islander;6% were Hispanic/Latino(a).

Domain Distinctions and Aggression 5

Social Rules Interview

Drawing on valid and reliable methods used inpast social domain research (see Smetana et al.,2014), structured interviews were conducted toassess children’s moral and conventional judg-ments. Interview scripts and all visual stimuli areincluded in Appendix S1.

Story Stimuli

Children were presented with four hypotheticalvignettes depicting everyday moral and conven-tional transgressions (two each). Hypothetical trans-gressors (matched to the participant’s gender) weredepicted as having clearly defined intentions, andall events were described as occurring in a class-room setting.

Each vignette presentation consisted of four storypictures: rule introduction, actor goal, transgression,and outcome. The rule introduction picture showedthe specific transgression and included a red “X”over the image. This was to ensure that all childrenunderstood that the behaviors (particularly the con-ventional events) were considered rule violations.Children were told that the act was prohibited (e.g.,“In this class, kids are not supposed to hit.”). Theactor goal picture described character(s) and theirmotives (e.g., wanting a toy that another child isplaying with). The transgression picture showed thecharacter committing the transgression to achievethe goal (e.g., hitting the child to get the toy). Theoutcome picture showed the goal achieved and theresulting consequences (e.g., the victim wasdepicted as hurt, and the actor was shown holdingthe toy).

Moral stories included physical harm (hit: a childhitting another child to obtain a desired toy) andproperty theft (steal: a child taking food fromanother child after finishing their own snack). Theconventional transgressions included violations ofclassroom organization (backpack: a child putting hisor her backpack on the ground instead of placing itin a cubby to play with a desired toy) and schoolcustom (hand: a child getting up to get anothersnack from the teacher’s desk without raising theirhand and asking permission first).

Initial Judgments

After each transgression, children’s initial judg-ments of the act’s acceptability were assesseddichotomously: “Is it ok or not ok for [transgressor]to ___?” This question was accompanied by two

images, one of a green “thumbs up” reflecting ajudgment that the act was acceptable (ok, scored as0) and a red “thumbs down” reflecting a judgmentthat the act was not acceptable (not ok, scored as 1).

Initial severity ratings were then assessed by ask-ing, “Do you think it was not bad for [transgressor]to ___, a little bad to do, or very bad to do?” Theanswer choices were represented by a “badnessmeter,” depicted using three rectangles. Not badwas represented by an empty rectangle, little badwas represented by a rectangle half filled withorange, and very bad was depicted by a rectanglecompletely colored in. Responses were scored on a3-point scale ranging from 0 (not bad) to 2 (verybad).

Criterion Judgments

Following initial judgments, additional pictureslides depicting authority or rule contingency werepresented. The pictures were designed to aid chil-dren’s understanding of hypothetical, counterfac-tual scenarios where normal rules andexpectations do not apply. Presenting childrenwith detailed scenarios depicting rule and author-ity independence, rather than simply asking if itwould be wrong if there were no rules/prohibi-tions, is consistent with the original proceduresused by domain researchers to assess criterionjudgments (Weston & Turiel, 1980). Children werepresented with only one criterion per storybecause pilot testing indicated that assessingacceptability and severity judgments of bothauthority and rule contingency for each story, inaddition to initial judgments, made the interviewexcessively long and tiresome. The criterion chil-dren received for a particular story was counter-balanced within domain. Thus, children evaluatedone story within each domain involving authoritycontingency and one depicting rule contingency,resulting in a total of four different story–criterioncombinations.

Authority contingency was illustrated by pre-senting an image of a female teacher shrugging hershoulders, with her palms turned up to indicatenonchalance. Above the teacher’s head was athought bubble depicting the target transgression,indicating that she was explicitly thinking about theact. This was accompanied by the verbal explana-tion:

What if it was a different class with a differentteacher. This teacher (point to teacher) does notcare if kids ___ in her class. See, she doesn’t

6 Jambon and Smetana

mind at all. So these questions are about what[transgressor] did if (s)he had this teacher whodidn’t care/mind and let kids ___.

For rule contingency, the rule introduction pic-ture containing the red “X” from the first storyslide was reintroduced at the top of the screen.Directly below this was the same picture, nowwithout the red “X,” positioned next to an image ofa school. Children were told, “Some schools haverules against ___, and say that kids aren’t supposedto ___” while the experimenter pointed to the toppicture. Pointing to the bottom picture, the experi-menter explained, “but in this school, there is norule against ___. Kids are allowed to ___ if theywant to. So these questions are about what [trans-gressor] did if (s)he went to this school where chil-dren are allowed to ___.”

Acceptability and severity judgments (asdescribed above) were then reassessed. Each judg-ment question was preceded by the statement, “Ifthe teacher doesn’t care/mind if kids ___ and letsthem do it . . .” (for authority contingency), or “Ifthere was no rule against ___and they let them doit . . .” (for rule contingency). For all judgments,higher scores indicated more “moral” evaluations(i.e., more unacceptable/wrong).

Aggression Measures

Reactive Aggression

Teachers reported on children’s reactive aggres-sion using six items from the physical and relationalsubscales of the Preschool Proactive and ReactiveAggression–Teacher Report (PPRA–T; Ostrov &Crick, 2007). Items were scored on a 5-point scaleranging from (1) never or almost never true to (5) al-ways or almost always true of the child. The subscalesdemonstrated good internal consistency (reactive-physical a = .97, reactive-relational a = .84) and havebeen shown to reliably predict observational assess-ments of reactive aggression (Murray-Close &Ostrov, 2009; Ostrov & Crick, 2007).

Proactive Aggression

Teachers also completed proactive subscalesfrom the PPRA–T. However, consistent with arecent study using the same measure (Poland,Monks, & Tsermentseli, 2016), preliminary analysesrevealed that the proactive and reactive scales werehighly correlated (r = .93) and did not load as sepa-rate factors. This likely stems from the high degree

of similarity in item wording. For example, itemsreflecting reactive aggression include “if other chil-dren anger this child, s/he will often hit, kick, orpunch them” (reactive-physical) and “when thischild is upset with others s/he will often ignore orstop talking to them” (reactive-relational), whereasproactive items include “this child often hit, kicks,or punches to get what s/he wants” (proactive-phy-sical) and “to get what this child wants, s/he oftenwill ignore or stop talking to others” (proactive-relational). Given that the PPRA did not differenti-ate reactive and proactive functions, the proactiveitems were dropped from the study.

Proactive aggression was therefore assessedusing teacher ratings of six items reflecting coerciveresource control (Hawley, 2003; Hawley & Geldhof,2012). Despite differences in terminology, coercivecontrol and proactive aggression reflect the sameunderlying construct (i.e., instrumental aggressionaimed at obtaining rewards). Furthermore, smalldifferences in content and wording in the coercivecontrol scale allowed us to model the proactive andreactive functions as separate factors. Items werescored on a 7-point scale ranging from (1) hardlytrue to (7) mostly true of the child, with higherscores reflecting a greater use of aggressive, coer-cive means of controlling others to accomplish goals(e.g., this child “. . .gets what s/he wants by bully-ing others”; “. . .tricks others to get what s/hewants”). The scale demonstrated good internal con-sistency (a = .93) and was highly correlated withproactive aggression ratings from the PPRA–T(r = .82), but was less highly correlated with reac-tive aggression (r = .78) compared to the PPRA–Tproactive scale.

Procedures

Children between 4 and 6 years of age attendinglocal day cares and schools and their teachers wereinvited to participate in the study. Letters describ-ing the study, along with parent permission forms,were given to all eligible families through their cen-ters/schools. Participating children were inter-viewed individually in a separate area of theirschool or day care. Most teachers (14 of 18) com-pleted online surveys regarding children’s behavior;the rest completed paper surveys.

Child interviews were conducted on an iPadusing the online survey software SurveyGizmo. Thisallowed for an interactive experience where chil-dren were able to point and choose differentanswer responses (described above). To train partic-ipants to use the device, children were first shown

Domain Distinctions and Aggression 7

each scale (e.g., “This is our ‘badness’ meter. Youcan use this to tell me how bad you think some-thing is”), and the experimenter demonstrated itsuse. Children were then shown the scales again andasked (in a random order) to select pictures corre-sponding to different judgments. The interviewscommenced only after children clearly understoodhow to use the scales.

All responses were automatically recordedonline. Judgment questions were presented in thefixed order described above, whereas the softwarepresented the stories and judgment responseoptions in a randomized order. The specific story–criterion combination was counterbalanced withingender. Each interview took approximately 20–25 min, and participants were offered breaks asneeded. Children were given a sticker after eachstory and received a colorful pencil at the end ofthe interview. Teachers received a small paymentranging from $10 to $30 depending on the numberof surveys completed.

Analysis Plan

Unless otherwise noted, structural equation mod-eling with latent variables in Mplus 7.3 (Muth�en &Muth�en, 1998-2012) was utilized to test ourhypotheses. Adapting standard procedures for ana-lyzing change scores, domain distinction abilitywas operationalized as a latent difference score(LDS; McArdle, 2009). This latent distinction con-struct represented the difference between criterionjudgments of events in the two domains, withhigher scores reflecting evaluations that moraltransgressions were less acceptable and morewrong independent of rules and authority than con-ventional violations (see Appendix S2 for a com-plete description of the procedures and exampleMplus syntax).

LDS modeling provided a number of benefitsover approaches commonly used in moral develop-ment research (e.g., analysis of variance, regression).We were able to explicitly test the hypothesis thatvariations in the ability to distinguish morality fromsocial convention reflect meaningful individual dif-ferences rather than resulting from measurementerror. It also allowed us to directly examine linksbetween domain distinctions and aggression ratherthan requiring complex tests of interaction effects.Consistent with past research (Arsenio et al., 2009;Poulin & Bouivin, 2000), separate models weretested to compare the unique associations betweenjudgments and proactive and reactive aggressionwhile controlling for the nonfocal function.

Results

Preliminary Analyses of Moral and ConventionalJudgments

Preliminary inspection of the data revealedthat nearly all children initially evaluated bothmoral and conventional transgressions as unac-ceptable and bad. Scores for individual storieswithin domains were similar for both authorityand rule contingency evaluations. Criterion judg-ments were therefore aggregated across storieswithin domain (see Table 1; complete descriptivestatistics by story and criteria are presented inAppendix S3).

One-way analyses of variance examiningwhether criterion judgments differed as a functionof the story–criterion combination revealed no sig-nificant effects (ps = .25–.97), suggesting that thespecific combination of story and criteria (e.g.,authority contingency of hitting vs. rule contin-gency of hitting) did not affect judgments. Addi-tionally, authority and rule contingency judgmentswere significantly correlated for acceptability(moral r = .54; convention r = .50, ps < .001) andseverity ratings (moral r = .44; conventionalr = .36, ps < .001) and were therefore combinedwithin domain. For ease of interpretability, crite-rion acceptability judgments were summed,whereas criterion severity judgments were aver-aged. Judgment ratings therefore ranged from 0 to2, with higher scores reflecting acts were viewedas less acceptable/more wrong. These aggregatedcriterion judgments of acceptability and severitywere used as manifest indicators in all subsequentanalyses. Descriptive statistics and manifest correla-tions for all study variables can be found inTable 2.

Table 1Descriptive Statistics for Moral and Conventional Criterion Judgments(Collapsed Across Story)

Judgment

Moral Conventional

M SD M SD

Authority contingencyAcceptabilitya 0.55 0.50 0.41 0.50Severityb 1.09 0.95 0.73 0.86

Rule contingencyAcceptabilitya 0.56 0.50 0.48 0.50Severityb 1.14 0.85 0.79 0.86

Note. aScores range from 0 (ok) to 1 (not ok). bScores range from 0(not bad) to 2 (very bad).

8 Jambon and Smetana

Missing Data

There was relatively little missing data (range =0%–7.8%), either in child interviews or teacherreports. Six children left their day cares before inter-views could be conducted, and five additional chil-dren did not have teacher reports (three due toteachers not completing surveys, and two becauseteachers left the center before data could be collectedand did not provide contact information). Little’smissing completely at random (MCAR) test wasnonsignificant, v2(59, N = 101) = 48.71, p = .828.Missing data were therefore accounted for based onthe MCAR assumption using the full informationmaximum likelihood (FIML) estimator with robuststandard errors.

Links Between Domain Distinctions and Aggression

Domain Differentiation as a Latent Construct

A confirmatory factor analysis with two latentfactors was conducted on moral and conventionalcriterion judgments. Each factor consisted of twoindicators (criterion acceptability and severity). Thisinitial model provided a relatively poor fit to thedata, v2(3) = 19.26, p < .001, root mean square errorof approximation (RMSEA) = 0.239, comparative fitindex (CFI) = 0.862, standardized root mean squareresidual (SRMR) = 0.053. Modification indices sug-gested that moral and conventional acceptabilityjudgments shared considerable error variance.Because these evaluations were measured using dif-ferent stimuli as the severity ratings, the error termswere allowed to covary. This modified two-factormodel fit the data well, v2(2) = 1.25, p = .58,RMSEA = 0.00, CFI = 1.00, SRMR = 0.031, and pro-vided a significantly better fit to the data than aone-factor model, Dv2(1) = 10.76, p = .001. An LDSmodel was then estimated to capture children’sability to differentiate morality and convention (seeAppendix S2).

Parameter estimates for the LDS model areshown in Figure 1a. Consistent with previousdomain research, the latent mean of the distinctionfactor was significantly different from zero (Cohen’sd = .94, p < .001), confirming that overall, childrensignificantly distinguished between morality andconvention in their criterion judgments. However,the variance of the distinction factor was also signif-icantly different from zero (p = .006), providingsupport for the hypothesis that children would dif-fer from one another in this ability. Conventionaljudgments were not significantly correlated withT

able

2Descriptiv

eStatisticsandBivariate

Correlatio

nsforAllStud

yVariables

MSD

Ran

ge1

23

45

67

89

1011

1213

1.Age

5.11

0.68

4–6.80

1�0

.07

0.24

*0.12

0.07

0.11

0.08

�0.10

�0.01

0.02

�0.01

�0.02

0.03

2.Gen

der(m

ale)

0.53

0.50

0–1

1�0

.01

0.07

�0.17†

�0.13

0.03

�0.13

0.17

†0.00

0.10

0.04

�0.11

3.Race(W

hite)

0.68

0.42

0–1

10.13

�0.06

0.02

0.27

*�0

.04

�0.05

0.07

0.13

0.02

0.05

4.Initial

moral

accept

1.93

0.32

0–2

10.43

**0.00

0.01

0.18

†0.13

�0.01

�0.16

�0.13

0.02

5.Initial

conv

accept

1.92

0.35

0–2

10.16

0.04

�0.04

0.12

�0.14

�0.13

�0.13

0.02

6.Initial

moral

sev

1.83

0.31

1–2

10.13

�0.14

�0.11

�0.07

�0.06

�0.12

0.09

7.Initial

conv

sev

1.52

0.55

0–2

10.04

0.05

0.25*

0.29**

�0.10

�0.13

8.Criterion

moral

accept

1.10

0.88

0–2

10.66**

0.68**

0.50**

�0.09

�0.14

9.Criterion

conv

accept

0.91

0.86

0–2

10.52**

0.56**

�0.13

�0.06

10.C

riterion

moral

sev

1.12

0.76

0–2

10.65**

�0.07

�0.11

11.C

riterion

conv

sev

0.76

0.71

0–2

1�0

.08

�0.14

12.R

eactiveag

gression

2.25

1.00

1–5

10.78**

13.P

roactiv

eag

gression

2.08

1.42

1–7

1

Note.

All

estim

ates

based

onob

served

man

ifest

variab

les.

Accep

t=acceptab

ility

judgm

ents;sev=seve

rity

judgm

ents;conv

=conv

entio

nal.Gen

der:0=female,

1=male.

Race:

0=no

n-White,1

=W

hite.*

p<.05.

**p<.01.

†p<.10.

Domain Distinctions and Aggression 9

domain distinction scores (p = .48). The inclusion ofdemographic covariates revealed that conventionaljudgments and domain distinction ability did notsignificantly differ according to gender (conven-tional: r = .02, p = .55; distinction: r = .17, p = .26),race (conventional: r = .10, p = .37; distinction:r = �.01, p = .94), or age (conventional: r = .00,p = .98; distinction: r = �.03, p = .79).

Aggression Measurement Model

Given the modest sample size and the largenumber of items, parceling was used to createaggregated manifest indicators for proactive andreactive aggression (Little, Rhemtulla, Gibson, &Shoemann, 2013). Because the focus of the presentstudy was on the different functions, items withsimilar wording from the physical and relationalsubscales of reactive aggression were combined tocreate three manifest parcels. Similarly, theory andmodification indices were used to create three

manifest parcels reflecting proactive aggression (seeAppendix S4 for a description of items and parcel-ing procedures).

As shown in Figure 1b, the resulting two-factormodel evidenced adequate fit: v2(8) = 12.68, p = .12,RMSEA = 0.078, CFI = 0.98, SRMR = 0.03, and pro-vided a significantly better fit than a one-factormodel, Dv2(1) = 8.45, p = .004. The inclusion ofdemographic covariates revealed that neither func-tion differed by gender (proactive: r = �.07, p = .43;reactive: r = �.03, p = .70), race (proactive:r = �.09, p = .88; reactive: r = .04, p =. 80), or age(proactive: r = �.07, p = .54; reactive: r = �.03,p = .78). These variables were subsequentlydropped from the analyses.

Domain Distinctions and Aggression (Models 1 and 2)

Parameter estimates for all structural models canbe found in Table 3. To test the hypothesis thatdomain distinctions would be uniquely associated

Figure 1. Measurement models for (a) latent moral/conventional difference scores and (b) reactive and proactive aggression. Factorloadings and covariance estimates are standardized. Residual variance estimates are unstandardized. **p < .01.

10 Jambon and Smetana

with proactive aggression, this factor was regressedonto domain distinctions, conventional judgments,and reactive aggression (see Figure 2a). Covarianceswere estimated between all predictors. Conventionaljudgments were not associated with proactive aggres-sion (b = �.01) and the path was constrained to zero.

As expected, after controlling for reactive aggres-sion, domain distinctions were negatively associatedwith proactive aggression (p = .014); children ratedas more proactively aggressive were less able to dif-ferentiate moral and conventional transgressions intheir criterion judgments. We then tested a secondmodel predicting reactive aggression scores fromproactive aggression and domain distinctions (seeFigure 2b). Unexpectedly, after controlling for proac-tive aggression, domain distinction ability was posi-tively associated with reactive aggression (p = .046);children rated as more reactively aggressive werebetter able to differentiate moral and conventionaltransgressions in their criterion judgments.

Follow-up analyses were conducted to examinewhether associations between domain distinctionsand aggression differed by age or gender. Thelatent variable interaction for age was tested usingthe XWITH command in Mplus, whereas multi-group analyses tested for gender differences. Effectsdid not differ by age or gender (ps > .30).

Comparing Domain Distinction Ability to OtherCommon Judgment Assessments

To test the second aim of the study—whetherdomain differentiation provided a more appropriateassessment of children’s prototypical moral under-standing than assessments commonly employed in

past research—two additional models examininglinks between proactive aggression and initialmoral judgments and moral criterion judgmentswere estimated (see Table 3).

Initial Moral Judgments (Model 3)

Only four children initially judged either of themoral transgressions to be acceptable. Given thislack of variability, initial moral severity ratings (av-eraged across the hitting and stealing stories) wereused. Proactive aggression was regressed on reac-tive aggression and observed initial moral severityscores, v2(12) = 17.03, p = .15, RMSEA = 0.06,CFI = 0.99, SRMR = 0.03. Interestingly, initial moralseverity ratings were positively associated withproactive aggression (p = .001); although most chil-dren judged moral transgressions to be wrong,proactively aggressive children rated them to bemore wrong than less aggressive youth.

Moral Criterion Judgments (Model 4)

Proactive aggression was then regressed on reac-tive aggression and a latent variable representingmoral criterion judgments, v2(18) = 18.18, p = .15,RMSEA = 0.01, CFI = 1.00, SRMR = 0.03. Moral cri-terion judgments alone were not significantly asso-ciated with proactive aggression (p = .19).

Discussion

Drawing from a community sample of diverse,lower to upper middle SES 4- to 6-year-olds, we

Table 3Parameter Estimates for Structural Models

Unstandardized Standardized

Coefficient(b)

Standarderror

LowerCI.95

UpperCI.95

Coefficient(b)

LowerCI.95

UpperCI.95

Model 1Domain dist ? Proactive �0.916 0.401 �1.580 �0.260 �.220 �.368 �.073Reactive ? Proactive 1.196 0.139 0.967 1.425 .835 .738 .933

Model 2Domain dist ? Reactive 0.532 0.276 0.077 0.987 .184 .032 .337Proactive ? Reactive 0.613 0.052 0.527 0.699 .879 .790 .984

Model 3Initial moral sev ? Proactive 0.930 0.256 0.509 1.35 .186 .089 .283

Model 4Moral crit ? Proactive �0.221 0.090 �0.239 0.058 �.096 �.215 .024

Note. Dist = distinction; proactive = proactive aggression; reactive = reactive aggression; sev = severity; crit = criterion judgments.

Domain Distinctions and Aggression 11

employed theoretically derived and validated mea-sures to test the hypothesis that variations in youngchildren’s ability to distinguish prototypical moraland conventional events along theoretical criteriawould be differentially associated with teacher rat-ings of proactive and reactive aggression. Proactiveaggression was uniquely associated with deficits inchildren’s ability to differentiate moral and conven-tional events, whereas, unexpectedly, reactiveaggression was linked with more advanced moraljudgments. However, proactive aggression was notassociated with deficits in moral cognitions whenother commonly used measures of moral knowl-edge were employed.

Individual Differences in Moral and ConventionalConcepts

The understanding that morality represents adistinct domain of social knowledge is considered anormative, universal feature of young children’s

social-cognitive development (Killen & Smetana,2015). Due to the limited cognitive and linguisticabilities of preschool and early school-aged chil-dren, criterion judgments represent the best avail-able method of assessing this conceptual distinctionin early childhood. Consistent with a large body ofpast social domain research, we found that whencomparing mean levels, children distinguishedmoral and conventional transgressions in theirjudgments. However, we also found significantinterindividual variation in this ability. Althoughlittle attention has focused on individual differencesin early moral judgments, this finding is neverthe-less fully consistent with a constructivist view ofdevelopment.

Children are predisposed toward care and con-cern for others from the earliest of ages (Eisenberg,Spinrad, & Morris, 2014), and observationalresearch has shown that naturally occurring moraland conventional events elicit different types ofreactions from peers and adults (Smetana et al.,

Figure 2. Structural models predicting (a) proactive aggression and (b) reactive aggression from domain distinctions. All parameter esti-mates are standardized. The covariance paths between the conventional factor and (a) reactive aggression (r = �.12, p = .13) and (b)proactive aggression (r = �.09, p = .13) were included in the models but are not depicted in the figure. *p < .05. **p < .01.

12 Jambon and Smetana

2014). Resulting from these experiences, childrendevelop different ways of thinking about moralityas compared to other social norms (Turiel, 1983).However, children also differ markedly from oneanother in their capacity for empathy (Eisenberget al., 2014) and the ability to read and pick up onsocial cues (Dodge et al., 2006), which may haveimportant implications for how they interpret andprocess morally relevant information. The use ofLDS modeling to capture variations in children’sdomain distinction ability could provide futureresearchers with a powerful and flexible tool toexamine the developmental processes associatedwith individual differences in early moral under-standing.

Moral Judgments and Proactive and ReactiveAggression

From a moral developmental perspective, inten-tionally harming others for self-gain constitutes anunderdeveloped moral orientation. Consistent withthis assertion, we found that proactive aggressionwas associated with a reduced capacity to differen-tiate moral and conventional norms, an ability thattypically emerges during the preschool years andprovides an early foundation for later moral devel-opment (Killen & Smetana, 2015; Nucci, 2009; Sme-tana et al., 2014; Turiel, 1983). Nevertheless, paststudies have found that proactively aggressive chil-dren are often powerful and popular within theirpeer groups (Hawley, 2014; Stoltz et al., 2016), andmay not exhibit the same signs of maladjustmentobserved in other aggressive youth (Poulin & Boui-vin, 2000; Sutton et al., 1999). These findings haveled theorists to construe some types of aggression,when flexibly employed to accomplish goals andobtain resources, as reflecting social competenceand adaptation (Hawley, 2014). Although this per-spective has garnered strong empirical support andprovides an important alternative to the traditionaldeficit-based view of aggression (Sutton et al.,1999), it fails to adequately account for the negativeimpact such behaviors have on others. Given thatproactive aggression may be uniquely associatedwith serious violence, delinquency, and psycho-pathic traits later in life (Hubbard et al., 2010), ourfindings underscore the need to consider moralfunctioning as a critical yet distinct aspect of earlysocial competence (Arsenio & Lemerise, 2004; Gas-ser & Keller, 2009).

These results have important theoretical implica-tions for how researchers conceptualize linksbetween morality and aggression. Moral

development researchers have shifted away fromstudying prototypical moral judgments to under-stand why some children deliberately harm othersdespite “knowing” it is wrong to do. Althoughother factors (e.g., empathic deficits, atypical pat-terns of reasoning regarding complex moral dilem-mas) are certainly important to consider (Arsenioet al., 2009; Eisner & Malti, 2015), our findings callinto question the assumption (Arsenio & Lemerise,2004; Nunner-Winkler, 2013) that proactive aggres-sors possess a similar understanding of morality asless aggressive children. This also provides supportfor the assertion that prototypical moral judgmentsdo reflect sincere beliefs regarding obligations torefrain from harm (Smetana et al., 2014) rather thana purely cognitive knowledge of moral rules (Haw-ley & Geldhof, 2012; Nunner-Winkler, 2013).

Because early detection and intervention is keyto preventing problematic behavioral trajectoriesfrom emerging (Vitaro et al., 2006), these resultsmay also help to inform efforts to prevent orreduce proactive aggression in young children. Todate, researchers have stressed the need for inter-vention strategies that specifically target empathicdeficits unique to proactive aggression (Arsenio &Lemerise, 2004; Hubbard et al., 2010). Our findingssuggest that focusing on children’s thinking andreasoning about issues of harm, justice, and rightsmay also be important, particularly during the pre-school and early school-age years when moral con-cepts and individual differences in aggression firstemerge. For instance, parents and educators couldbe trained to identify and respond to differenttypes of misbehaviors with domain-appropriatemessages that explicitly highlight the moral or con-ventional nature of the transgressions (Nucci, 2009).Additionally, controlled interventions utilizing con-cepts and methods from contemporary moraldevelopment research could provide invaluableinformation regarding the extent to which moraljudgments and reasoning play a causal role in chil-dren’s social behavior (Eisner & Malti, 2015; Vitaroet al., 2006).

In contrast to the findings for proactive aggres-sion, we found that children rated higher in reac-tive aggression were better able to distinguishmoral and conventional events (after controlling forproactive aggression). Although this finding wasunexpected, it may stem from the types of socialinteractions and social-cognitive biases that areindicative of reactively aggressive youth. Resultingfrom past experiences (e.g., victimization, harshfamily environments), reactive aggressors are par-ticularly likely to believe that others possess hostile

Domain Distinctions and Aggression 13

intentions and retaliate with force (Dodge et al.,2006). Given that moral experiences (e.g., withaggression, unfairness) are believed to facilitate theemergence of moral concepts (Smetana et al., 2014;Turiel, 1983), reactive aggressors may be uniquelyattuned to the implications and wrongness ofunprovoked harm. Indeed, reactively aggressiveyouth often consider aggression in response toprovocation and threat as a necessary moral imper-ative aimed at what they perceive as an injustice(Orobio de Castro et al., 2012). Given the relativelysmall effect size and unexpected nature of the find-ing, however, future replication is needed beforedrawing any firm conclusions.

Nevertheless, greater research is needed tounderstand how everyday experiences of conflict,victimization, and aggression color children’s con-ceptions of right and wrong. To date, moral devel-opment and aggression researchers have primarilyrelied on hypothetical stories to document chil-dren’s social and moral reasoning (Dodge et al.,2006; Killen & Smetana, 2015). It is less clear howchildren and adolescents understand their ownharmful behaviors after they have occurred orwhether adult and peer responses influence thisinterpretive process. Because engaging in actionsthat negatively affect others is an inevitable part ofsocial life, the extent to which individuals self-reflect on their wrongdoing is a critical yet under-studied aspect of moral development (see Wainryb& Recchia, 2014).

Assessing Prototypical Moral Judgments

Our findings underscore the importance of con-sidering moral evaluations in a theoretically con-sistent manner. In line with with past research,few children stated that moral transgressions weregenerally acceptable, and moral criterion judg-ments were not significantly associated withproactive aggression. Consistent with Hawley’sfindings (Hawley, 2003; Hawley & Geldhof, 2012),however, we found that proactively aggressivechildren initially rated moral transgressions asmore wrong than less aggressive youth. This mayreflect the nature of proactive aggression and thesocial desirability concerns inherent in asking chil-dren about socially unacceptable behaviors andbeliefs (Dunn, 2014; Piaget, 1932/1965; Rutlandet al., 2005).

Proactively aggressive children are often intelli-gent and socially skilled. Thus, they may knowhow to appear “moral” and maintain that acts ofharm would be wrong even if external sanctions or

rules were removed. Yet they appear to overgener-alize this strategy when evaluating conventionalevents as well. This interpretation is consistent withresearch showing that adolescents and adults highin psychopathy may at times treat all rule viola-tions as morally wrong (Blair, 1995; Dolan & Ful-lam, 2010). Operationalizing young children’srudimentary moral understanding as the ability todifferentiate moral and conventional norms maytherefore provide a useful method for counteringsocial desirability concerns.

Limitations and Future Directions

Although the present study addresses importantgaps in our understanding of moral judgments andaggression, several limitations should be men-tioned. First, aggression was assessed using teacherratings. Although teacher reports have been shownto reliably predict observational measures ofaggression subtypes (Murray-Close & Ostrov, 2009;Ostrov & Crick, 2007), future research shouldinclude other informants. Additionally, coercivecontrol was used to operationalize proactive aggres-sion because the PPRA did not adequately distin-guish between the two functions. Given thatcoercive control and proactive aggression reflect thesame construct (aggression aimed at obtainingrewards), are measured using similar items (harm-ing others “to get what they want”), and coercivecontrol—but not the PPRA proactive items—loadedseparately from reactive aggression, we believe thisdecision was justified. Nevertheless, research thatemploys both experimental and observationalassessments is warranted.

Given the ubiquity of questionnaire assessmentsof children’s social behavior in developmentalresearch, however, more valid and reliable ques-tionnaires aimed at differentiating between aggres-sive subtypes are needed (Hubbard et al., 2010).Although procedures have been developed to limitthe confounding of aggressive forms and functionsin behavioral ratings (Little et al., 2003), the calcu-lating, callous, and selfish nature of proactiveaggression is largely absent in existing measuresasking whether children behave aggressively to“get what they want.” A child who angrily lashesout against a threatening peer has certainly accom-plished what they set out to do, yet this does notconstitute an act of proactive or instrumentalaggression as it is typically conceptualized (Crick &Dodge, 1996; Hubbard et al., 2010; Little et al.,2003; Murray-Close & Ostrov, 2009). To betterunderstand links to moral functioning, future

14 Jambon and Smetana

studies could therefore benefit from shifting thefocus to the characteristics unique to proactiveaggression. For instance, proactive aggression hasbeen associated with the presence of callous–emo-tional (CU) traits entailing low levels of empathyand guilt and a tendency to use others for personalgain (Frick & White, 2008), and children and ado-lescents exhibiting CU traits and proactive aggres-sion show similar social-cognitive, affective, andbehavioral profiles (see Frick & White, 2008). Futureresearch could therefore examine whether deficitsin moral judgments and reasoning are morestrongly associated with aggression in childrenexhibiting high levels of CU traits (i.e., proactiveaggressors) compared to children low on thisdimension. This approach could avoid some of thelimitations of existing questionnaires while shed-ding light on how different social-cognitive andaffective mechanisms may contribute to the sameobservable behaviors.

Consistent with previous social domain research,we focused on hypothetical moral vignettes depict-ing straightforward instances of harm that closelyresemble proactive aggression. As such, reactiveaggression may have been linked with lessadvanced judgments if situations entailing reactiveaggression (e.g., provocation or retaliation) hadbeen included. Although these types of complex sit-uations have been the subject of much previousresearch (Arsenio et al., 2009; Astor, 1994; Gasseret al., 2012), they do not constitute prototypicalmoral transgressions and were therefore not thefocus of the present study. However, futureresearch should examine whether proactive andreactive aggression are differentially associated withevaluations of hypothetical situations that arematched in kind to ensure that these findings arerobust.

We also relied on a relatively small subset ofmoral and conventional vignettes and focused on alimited age range of participants. Although the sto-ries and age range were appropriate for the currentinvestigation, links between judgments and aggres-sion would be expected to differ depending on thetypes of stories and developmental period exam-ined. For example, judgments of more covert moralevents (e.g., social exclusion) may be more stronglyassociated with proactive aggression that is rela-tional or social in nature. Taking into account thedifferent forms of aggression may be particularlyimportant at later ages. Physical aggression, com-mon in young children, normatively declines withage and is associated with a host of problematicoutcomes later in life (Eisner & Malti, 2015). As

such, proactive aggression entailing serious physicalviolence in older youth is likely indicative ofbroader psychosocial deficits rather than a lack ofmoral values. Understanding developmental shiftsin the relation between moral judgments and differ-ent forms and functions of aggression is an impor-tant avenue for future study.

Finally, greater research is needed to under-stand the developmental antecedents and conse-quences of variations in children’s understandingof right and wrong. To date, research on individ-ual differences in moral development has primarilyfocused on affective (guilt, empathy) and behav-ioral constructs (prosociality, compliance; see Kil-len & Smetana, 2015). Yet intrapersonal (e.g.,temperament), interpersonal (e.g., parenting prac-tices), and contextual factors (e.g., neighborhooddisadvantage) all influence how children makesense of the world. Furthermore, although deficitsin prototypical moral understanding have beendocumented in adolescents and adults exhibitingextreme antisocial behavior (Blair, 1995; Blair et al.,2001; Nucci & Herman, 1982), we do not knowwhether impairments in early social and moraljudgments have implications for later moral func-tioning in typically developing populations. Exam-ining the developmental processes associated withindividual differences in moral cognitions wouldprovide a more integrated and nuanced under-standing of children’s moral and social develop-ment.

References

Arsenio, W., Adams, E., & Gold, J. (2009). Social informa-tion processing, moral reasoning, and emotion attribu-tions: Relations with adolescents’ reactive and proactiveaggression. Child Development, 80, 1739–1755. doi:10.1111/j.1467-8624.2009.01365.x

Arsenio, W., & Lemerise, E. (2004). Aggression and moraldevelopment: Integrating social information processingand moral domain models. Child Development, 75, 987–1002. doi:10.1111/j.1467-8624.2004.00720

Astor, R. (1994). Children’s moral reasoning about familyand peer violence: The role of provocation and retribu-tion. Child Development, 65, 1054–1067. doi:10.2307/1131304

Blair, R. (1995). A cognitive developmental approach tomorality: Investigating the psychopath. Cognition, 57,1–29. doi:10.1016/0010-0277(95)00676-p

Blair, R., Monson, J., & Frederickson, N. (2001). Moralreasoning and conduct problems in children with emo-tional and behavioral difficulties. Personality and Individ-ual Differences, 31, 799–811. doi:10.1016/s0191-8869(00)00181-1

Domain Distinctions and Aggression 15

Card, N., & Little, T. (2006). Proactive and reactiveaggression in childhood and adolescence: A meta-ana-lysis of differential relations with psychosocial adjust-ment. International Journal of Behavioral Development, 30,466–480. doi:10.1177/0165025406071904

Conry-Murray, C. (2013). Young children’s understandingof beliefs about moral and conventional rule violations.Merrill-Palmer Quarterly, 59, 489–510. doi:10.13110/merrpalmquar.59.4.0489

Crick, N., & Dodge, K. (1996). Social-information process-ing mechanisms in reactive and proactive aggression.Child Development, 67, 993–1002. doi:10.1037/0033-2909.115.1.74

Cui, L., Colasante, T., Malti, T., Ribeaud, D., & Eisner, M.P. (2016). Dual trajectories of reactive and proactiveaggression from mid-childhood to early adolescence:Relations to sensation seeking, risk taking, and moralreasoning. Journal of Abnormal Child Psychology, 44, 663–675. doi:10.1007/s10802-015-0079-7

Dodge, K., Coie, J., & Lynam, D. (2006). Aggression andantisocial behavior in youth. In W. Damon (Series Ed.)& N. Eisenberg (Vol. Ed.), Handbook of child psychology:Social, emotional, and personality development (Vol. 3, 6thed., pp. 719–788). New York, NY: Wiley.

Dolan, M., & Fullam, R. (2010). Moral/conventional trans-gression distinction and psychopathy in conduct disor-dered adolescent offenders. Personality and IndividualDifferences, 49, 995–1000. doi:10.1016/j.paid.2010.08.011

Dunn, J. (2014). Moral development in early childhoodand social interaction in the family. In M. Killen, & J.Smetana (Eds.), Handbook of moral development, 2nd ed.(pp. 135–159). New York, NY: Psychology Press.

Eisenberg, N., Spinrad, T., & Morris, A. (2014). Empathy-related responding in children. In M. Killen, & J. Sme-tana (Eds.), Handbook of moral development, 2nd ed. (pp.184–207). New York, NY: Psychology Press.

Eisner, M., & Malti, T. (2015). Aggressive and violentbehavior. In R. Lerner (Editor-in-Chief) & M. Lamb(Vol. Ed.), Handbook of child psychology and developmentalscience: Socioemotional processes (Vol. 3, 7th ed., pp. 794–841). New York, NY: Wiley-Blackwell.

Frick, P., & White, S. (2008). Research review: The impor-tance of callous-unemotional traits for developmentalmodels of aggressive and antisocial behavior. Journal ofChild Psychology and Psychiatry, 49, 359–375. doi:10.1111/j.1469-7610.2007.01862.x

Gasser, L., & Keller, M. (2009). Are the competentmorally good? Perspective taking and moral motivationof children involved in bullying. Social Development, 18,798–816. doi:10.1111/j.1467-9507.2008.00516.x

Gasser, L., Malti, T., & Gutzwiller-Helfenfinger, E. (2012).Aggressive and non-aggressive children’s moral judg-ments and moral emotion attributions in situationsinvolving retaliation and unprovoked aggression. Jour-nal of Genetic Psychology, 173, 417–439. doi:10.1080/00221325.2011.614650

Hawley, P. (2003). Strategies of control, aggression, andmorality in preschoolers: An evolutionary perspective.

Journal of Experimental Child Psychology, 85, 213–235.doi:10.1016/s0022-0965(03)00073-0

Hawley, P. (2014). The duality of human nature: Coercionand prosociality in youths’ hierarchy ascension andsocial success. Current Directions in Psychological Science,23, 433–438. doi:10.1177/0963721414548417

Hawley, P., & Geldhof, J. (2012). Preschoolers’ socialdominance, moral cognition, and moral behavior: Anevolutionary perspective. Journal of Experimental ChildPsychology, 112, 18–35. doi:10.1016/j.jecp.2011.10.004

Hubbard, J., McAuliffe, M., Morrow, M., & Romano, L.(2010). Reactive and proactive aggression in childhoodand adolescence: Precursors, outcomes, processes, expe-riences, and measurement. Journal of Personality, 78, 95–118. doi:10.1111/j.1467-6494.2009.00610.x

Huesmann, L., & Guerra, N. (1997). Children’s normativebeliefs about aggression and aggressive behavior. Jour-nal of Personality and Social Psychology, 72, 408–419.doi:10.1037//0022-3514.72.2.408

Jambon, M., & Smetana, J. (2014). Moral complexity inmiddle childhood: Children’s evaluations of necessaryharm. Developmental Psychology, 50, 22–33. doi:10.1037/a0032992

Killen, M., & Smetana, J. (2015). Origins and developmentof morality. In R. Lerner (Editor-in-Chief) & M. Lamb(Vol. Ed.), Handbook of child psychology and developmentalscience: Socioemotional processes (Vol. 3, 7th ed., pp. 701–749). New York, NY: Wiley-Blackwell.

Little, T., Henrich, C., Jones, S., & Hawley, P. (2003).Disentangling the “whys” from the “whats” of aggres-sive behaviour. International Journal of Behavioral Devel-opment, 27, 122–133. doi:10.1080/01650250244000128

Little, T., Rhemtulla, M., Gibson, K., & Shoemann, A.(2013). Why the items versus parcels controversyneedn’t be one. Psychological Methods, 18, 285–300.doi:10.1037/a0033266

Malti, T., Gasser, L., & Buchmann, M. (2009). Aggressive andprosocial children’s emotion attributions and moral reason-ing. Aggressive Behavior, 35, 90–102. doi:10.1002/ab.20289

Malti, T., Gasser, L., & Gutzwiller-Helfenfinger, E. (2010).Children’s interpretive understanding, moral judg-ments, and emotion attributions: Relations to socialbehavior. The British Journal of Developmental Psychology,28, 275–292. doi:10.1348/026151009x403838

McArdle, J. (2009). Latent variable modeling of differ-ences and changes with longitudinal data. AnnualReview of Psychology, 60, 577–605. doi:10.1146/annurev.psych.60.110707.163612

Murray-Close, D., & Ostrov, J. (2009). A longitudinalstudy of forms and functions of aggressive behavior inearly childhood. Child Development, 80, 828–842. doi:10.1111/j.1467-8624.2009.01300.x

Muth�en, L., & Muth�en, B. (1998-2012). Mplus user’s guide(version 7). Los Angeles, CA: Author.

Nucci, L. (2009). Nice is not enough: Facilitating moral devel-opment. Upper Sadle River, NJ: Pearson.

Nucci, L., & Herman, S. (1982). Behavioral disorderedchildren’s conceptions of moral, conventional, and

16 Jambon and Smetana

personal issues. Journal of Abnormal Child Psychology, 10,411–426. doi:10.1007/bf00912330

Nunner-Winkler, G. (2013). Moral motivation and thehappy victimizer phenomenon. In K. Heinrichs, F.Oser, & T. Lovat (Eds.), Handbook of moral motivation:Theories, models, applications (pp. 267–288). Rotterdam,the Netherlands: Sense.

Orobio de Castro, B., Verhulp, E., & Runions, K. (2012).Rage and revenge: Highly aggressive boys’ explana-tions for their responses to ambiguous provocation.European Journal of Developmental Psychology, 9, 331–350.doi:10.1080/17405629.2012.680304

Ostrov, J., & Crick, N. (2007). Forms and functions ofaggression during early childhood: A short-term longitu-dinal study. School Psychology Review, 36, 22–43.Retrieved from http://http://naspjournals.org/loi/spsr

Piaget, J. (1965). The moral judgment of the child. NewYork, NY: Free Press. (Original work published 1932)

Poland, S., Monks, C., & Tsermentseli, S. (2016). Cool andhot executive function as predictors of aggression inearly childhood: Differentiating between the functionand form of aggression. The British Journal of Develop-mental Psychology, 34, 181–197. doi:10.1111/bjdp.12122

Posada, R., & Wainryb, C. (2008). Moral development ina violent society: Colombian children’s judgments inthe context of survival and revenge. Child Development,79, 882–898. doi:10.1111/j.1467-8624.2008.01165.x

Poulin, F., & Bouivin, M. (2000). Reactive and instrumen-tal aggression: Evidence of a two-factor model. Psycho-logical Assessment, 12, 115–122. doi:10.1037/1040-3590.12.2.115

Rutland, A., Cameron, L., Milne, A., & McGeorge, P.(2005). Social norms and self-presentation: Children’simplicit and explicit intergroup attitudes. Child Develop-ment, 76, 451–466. doi:10.111/j.1467-8624.2005.00856.x

Smetana, J., Jambon, M., & Ball, C. (2014). The socialdomain approach to children’s moral and social judg-ments. In M. Killen & J. Smetana (Eds.), Handbook ofmoral development (2nd ed., pp. 23–45). New York, NY:Psychology Press.

Smetana, J., Toth, S., Cicchetti, D., Bruce, J., Kane, P., &Daddis, C. (1999). Maltreated and nonmaltreated

preschoolers’ conceptions of hypothetical and actualmoral transgressions. Developmental Psychology, 35, 269–281. doi:10.1037/0012-1649.35.1.269

Stoltz, S., Cillessen, A., van den Berg, Y., & Gommans, R.(2016). Popularity differentially predicts reactive andproactive aggression in early adolescence. AggressiveBehavior, 42, 29–40. doi:10.1002/ab.21603

Sutton, J., Smith, P., & Swettenham, J. (1999). Bullyingand “theory of mind”: A critique of the “social skillsdeficit” view of antisocial behavior. Social Development,8, 117–127. doi:10.1111/1467-9507.00083

Turiel, E. (1983). The development of social knowledge: Moral-ity and convention. Cambridge, UK: Cambridge Univer-sity Press.

Vitaro, F., Brendgen, M., & Barker, E. (2006). Subtypes ofaggressive behaviors: A developmental perspective.International Journal of Behavioral Development, 30, 12–19.doi:10.1177/0165025406059968

Wainryb, C., & Recchia, H. (Eds.) (2014). Talking aboutright and wrong: Parent–child conversations as contexts formoral development. New York, NY: Cambridge Univer-sity Press.

Weston, D., & Turiel, E. (1980). Act-rule relations: Chil-dren’s concepts of social rules. Developmental Psychol-ogy, 16, 417–424. doi:10.1037/0012-1649.16.5.417

Supporting Information

Additional supporting information may be found inthe online version of this article at the publisher’swebsite:

Appendix S1. Social Rules Interview Stimuli andScript

Appendix S2. Estimating a Latent Domain Dis-tinction Factor

Appendix S3. Full Descriptive Statistics forMoral and Conventional Judgments, by Story andCriteria

Appendix S4. Parceling of Proactive and Reac-tive Aggression Items

Domain Distinctions and Aggression 17