Beyond the positive reinforcement of aggression

Abstract

Being surrounded by peers who are accepting of aggression is a significant predictor of the development and persistence of aggression in childhood and adolescence. Whereas past research has focused on social reinforcement mechanisms as the underlying processes, the present longitudinal study analysed the role of external control beliefs as an additional mediator explaining the link between peers’ acceptance of aggression and the development of aggressive behaviour. Drawing on a large community sample of N = 1,466 male and female children and adolescents from Germany aged between 10 and 18 years, results of latent structural equation modeling were consistent with the hypotheses that peer acceptance of aggression would predict external control beliefs in the social domain, which in turn, should predict aggressive behaviour over time. Additional multigroup analyses showed that the predicted pathways were consistent across gender and age groups.

Keywords

adolescence aggression aggressive peers childhood control beliefs Germany longitudinal

Affiliation with peers who show a high acceptance of aggression has been shown to be crucial for both the development and the persistence of aggressive behaviour in childhood and adolescence (Allen, Porter, McFarland, Marsh, & McElhaney, 2005; Lacourse, Nagin, Tremblay, Vitaro, & Claes, 2003; Nesdale, Durkin, Maass, Kiesner, & Griffith, 2008; Werner & Crick, 2004). Mediating mechanisms have been mostly explained from a social learning perspective, arguing that the aggressive peers’ reinforcement of aggressive and delinquent acts accounts for the high stability of aggressive behaviour throughout the lifetime (Buehler, Patterson, & Furniss, 1966; Dishion, Spracklen, Andrews, & Patterson, 1996; Mathys, Hyde, Shaw, & Born, 2013; Snyder et al., 2005). Although social reinforcement mechanisms provide a versatile approach for understanding the link between aggressive peers’ norms and the maintenance of aggressive behaviour, they imply that social interactions between peers with a high acceptance of aggression are regulated by clearly formulated and commonly shared behavioural guidelines. However, unless they take place in highly organized deviant social groups (e.g. street gangs), many dyadic exchanges between aggressive peers are unstructured, unorganized, and lacking stable stimulus–response contingencies (Dishion, Nelson, Winter, & Bullock, 2004). In the present research, we argue that this social disorganization and lack of consistency may provide a further explanatory approach for the association between a peer group environment that condones aggression and the development of aggression in childhood and adolescence. Specifically, social disorganization and lacking consistencies in social interactions between aggressive peers might promote the development of external control beliefs. External control beliefs reflect individuals’ perceptions that social outcomes are beyond their control and have been shown to be positively related to antisocial and aggressive behaviour (Anderson, 1977; Duke & Fenhagen, 1975; Han, Weisz, & Weiss, 2001; Romi & Itskowitz, 1990). In our longitudinal study, we examine the proposition that membership of a peer group in which the acceptance of aggression is high facilitates the development of external control beliefs that, in turn, contribute to the development of aggression over time.

External locus of control and aggression

Locus of control refers to an individual’s tendency to locate the sufficient causes for events within or outside the self (Connell, 1985; Rotter, 1966, 1990; Skinner, 1996; Skinner, Zimmer-Gembeck, & Connell, 1998). Accordingly, locus of control describes a set of subjective and persistent perceptions of the contingency between one’s own behaviour and the occurrence of desired and undesired outcomes. Whereas individuals with an internal locus of control tend to perceive outcomes as contingent on their behaviour, individuals with an external locus of control consider outcomes to be the result of external influences that are beyond their control. Past research has highlighted the domain specificity of locus of control and its potential to vary across different contexts. Control perceptions in the cognitive, physical, and social domain are differentiated, based on the assumption that each dimension has characteristic associative patterns with other psychological constructs as well as distinct developmental pathways (Connell, 1985; Han et al., 2001; Harter, 1982).

The perception of the self as efficient and capable of manipulating and controlling social and physical events is a significant predictor of psychological functioning and mental and physical well-being (Seligman, 1975; Skinner et al., 1998). By contrast, a lack of control is associated with both internalizing psychopathology, like depressive or anxious symptoms (McCauley, Mitchell, Burke, & Moss, 1988), and externalizing problems, such as antisocial, delinquent, or aggressive behaviour (Duke & Fenhagen, 1975; Halloran, Doumas, John, & Margolin, 1999; Han et al., 2001; Österman et al., 1999). Various variables have been found to moderate the link between control beliefs and aggression. For example, examining the domain specificity of control beliefs, Han et al. (2001) found negative associations of internal control beliefs in the physical and social, but not in the academic, domain with internalizing and externalizing psychopathology in children and adolescents aged between 7 and 17 years. Analysing the moderating role of gender, Österman et al. (1999) reported that physical, indirect, and verbal forms of aggression were significantly related to a global measure of external locus of control in male, but not in female adolescents. By contrast, a study of children aged between 8 and 11 years found significant associations between an overall score of external control beliefs and aggression for girls but not for boys (Halloran et al., 1999). For boys, the relationship even tended to be reversed, indicating that males who perceived themselves as unable to control environmental outcomes showed less aggression. Taken together, these studies not only provide evidence for significant associations between external control beliefs and externalizing psychopathology, they also suggest that this link may vary as a function of both individual attributes and domain. Accordingly, further research is needed to consider plausible moderating variables, such as gender or age, and the domain specificity of control beliefs when examining its relationship with aggressive behaviour. As aggression takes place in a social context, we argue that beliefs about the control of outcomes in the social domain are particularly relevant.

From a developmental perspective, it is argued that individuals have an innate motivation to feel competent and effective in interactions with their environment (Connell & Wellborn, 1991; Harter, 1978; Skinner, 1996; White, 1959). The subjective experience of internal, personal control, however, depends on both the individuals’ skills and capabilities as well as the structure, organization, and responsivity of their social and physical environment (Skinner et al., 1998). Research on contextual influences has shown that one significant determinant of an individual’s locus of control is the contingency of the environment (Skinner et al., 1998). Environmental contingency refers to the degree to which actions are followed consistently and discriminately by the same outcomes. Generally, the existence of contingency in the environment is a prerequisite for learning, planning, and effective problem-solving. Without a minimum of contingency, organization, and structure, individuals would not be able to anticipate future outcomes and to adequately prepare for upcoming events. Non-contingent environments are experienced as unreliable, and outcomes are likely to be perceived as arbitrary and the result of influences beyond internal control, such as fate, chance, or powerful others (Skinner et al. 1998). Accordingly, it has been shown that in warm, benevolent, and highly contingent family environments, children and adolescents tend to grow up with a greater sense of agency and internal control (Grolnick, Ryan, & Deci, 1991; Krampen, 1989; Watson, 1966; Yates, Kennelly, & Cox, 1975). In these environments, parents tend to be more sensitive to their children’s wishes and needs and to respond more consistently to their behaviours. By contrast, critical, punitive, and non-contingent family environments, in which parents are responding inconsistently and unpredictably to their children’s actions, support the development of external control beliefs as well as feelings of helplessness and ineffectiveness (Grolnick et al., 1991; Grolnick & Ryan, 1992).

Peers’ acceptance of aggression and the development of external control beliefs

The majority of studies examining contextual effects on control beliefs have focused on the family environment. However, while the socializing impact of the family is most important in early and middle childhood, its influence is gradually decreasing as the child moves into adolescence (Bierman, 2004). At the same time, the significance of social interactions with same-aged peers increases (Pardini, Loeber, & Stouthamer-Loeber, 2005). Peer interactions are especially relevant for the development of aggressive behaviour. Explanatory approaches postulate that the normative system of aggressive peer groups substantially differs from mainstream social groups in that it does not impose sanctions on aggressive behaviours. Moreover, aggressive peers even tend to reward aggressive behaviours through applause, reputation, or access to desired resources while at the same time tending to ignore, discourage, or even punish prosocial acts (Anderson, 2002; Buehler et al., 1966; Dishion et al., 1996; Mathys et al., 2013).

Despite the importance of selective reinforcement mechanisms, there is some evidence that dyadic interactions between deviant individuals are not as coordinated as commonly suggested within a contingency framework. Deviant peer interactions may be chaotic, disorganized, and lack stable contingencies between behaviours and social reactions. For example, applying a dynamic system framework, Dishion et al. (2004) observed that interactions between deviant boys and their best friends not only contained more deviant behaviours than those of well-adjusted peers, they were also less structured and organized. Dishion et al. (2004) argued that aggressive individuals’ deficits in social information processing as well as a lack of problem-solving skills may account for the observed weak structure and contingency of deviant peer interactions. Aggressive individuals often show a biased attentional focus to provoking stimuli and a tendency to misread and misattribute social cues. They tend to overreact to ambiguous social stimuli and to use inappropriate interpersonal problem-solving strategies (Crick & Dodge, 1994; Dodge, Bates, & Pettit, 1990). Accordingly, it is likely that in aggressive peer interactions, social cues are frequently misinterpreted, followed by inappropriate social reactions in the absence of clear behaviour guidelines. As a consequence, children and adolescents with an aggressive peer network may find it difficult to gauge and predict the social consequences of their own and their peers’ behaviours. Accordingly, being situated within such a disorganized peer context is proposed to constitute a risk factor for the development of external control beliefs in the social domain, thereby supporting the development of aggressive behaviour.

The current study

In the present longitudinal study, we examined external locus of control beliefs as mediators in the link between affiliation with peers who show a high acceptance of aggressive behaviour and the development of aggression. Acknowledging the domain specificity of locus of control (Han et al., 2001) and considering the interpersonal nature of aggressive behaviour, we expected an external locus of control in the social domain to be predicted by peers’ acceptance of aggression and, in turn, to predict individual aggression. Additionally, we examined whether the hypothesized pathways might be moderated by gender and age. Given the inconclusive findings on gender differences and the lack of evidence on age differences in the relationship between locus of control and aggressive behaviour, we refrained from formulating specific hypotheses.

The current study presents data on the link between peers’ acceptance of aggression, external locus of control, and aggressive behaviour in a sample of 10–18-year-old participants, who were studied at two measurement waves over a period of about 1.5 years. Although conventional tests of mediation depend on at least three data waves, Cole and Maxwell (2003) noted that two waves of data collection may be sufficient for assessing partial mediation if the causal effect between the mediator and the outcome variable may be assumed to be temporally stable (assumption of stationarity). In our case, this means that the strength of the path from external locus of control to aggression is not assumed to change over time. Although the present study does not allow us to test this assumption directly, theoretical considerations suggest that a change in causal effects is unlikely, as the socializing impact of peers is especially salient in adolescence. Accordingly, the following predictions were examined in our study:
Hypothesis 1. The higher the perceived approval of aggression in their peer network at T1, the more likely participants are to show an external locus of control in the social domain at T2.

Hypothesis 2. The more participants perceive outcomes in the social domain as non-controllable at T1, the more aggressive behaviour they show at T2.

Hypothesis 3. The path from peer approval of aggression to aggressive behaviour is mediated by an external locus of control in the social domain.

In testing Hypothesis 3, mediation is indicated by a significant product of the path coefficient from peers’ acceptance of aggression at T1 to external locus of control at T2 and the path coefficient from external locus of control at T1 to aggression at T2.

Method

Participants

The study used data from a community sample of male and female children and adolescents involved in a school-based longitudinal study conducted in different districts of the state of Brandenburg, Germany. Of the 1,466 (50% female) participants assessed at T1, 1,107 took part at T2, resulting in an attrition rate of approximately 24.5%. The mean age of the sample was 12.9 years at T1 (SD = 2.01) and 14.3 years (SD = 1.90) at T2, covering a mean time period of about 17 months between the measurement waves. At T1, participants were distributed across 181 schools, 31.8% attended primary school, 67.1% attended secondary school, and 1.1% attended other schools. In terms of parents’ educational background, 42.1% of the fathers and 42.4% of the mothers had vocational qualifications, 15.2% of fathers and 20.8% of mothers had a university entrance qualification, and 41.6% of fathers and 36.0% of mothers held a university degree. Only a small subset (1.1% fathers; 0.8% mothers,) had no or low levels of qualification.

Measures

Aggressive behaviour

Aggressive behaviour was assessed at T1 and T2 by a self-report measure developed by Krahé and Möller (2010) comprising 10 items. Participants were asked to indicate how often in the past 6 months they had shown a particular behaviour, using a five-point scale from 1 (never) to 5 (very often). Five items measured physical aggression (e.g. “I have kicked another person”), and five items measured relational aggression (e.g. “I have excluded someone from our group”). Due to the skewness of the scale distribution, we computed ordinal alpha as a measure of scale reliability (Gaderman, Guhn, & Zumbo, 2012; Zumbo, Gaderman, & Zeisser, 2007). The ordinal alphas for the total scale of aggression was high, as shown in Table 1.

Table 1.
Scale reliabilities, means, standard deviations, and gender differences

Variable N items α Range M (SD) total M (SD) males M (SD) females η²

Time 1 Peers’ acceptance of aggression 6 .88 1.00–4.00 1.65 (0.54) 1.77 (0.60)^a 1.50 (0.40)^b .06

Time 1 External locus of control 4 .67 1.00–5.00 2.13 (0.73) 2.12 (0.74) 2.14 (0.72) .00

Time 1 Aggressive behaviour 10 .87 1.00–3.20 1.39 (0.34) 1.43 (0.37)^a 1.34 (0.29)^b .02

Time 2 Peers’ acceptance of aggression 6 .86 1.00–4.00 1.63 (0.48) 1.76 (0.54)^a 1.49 (0.36)^b .08

Time 2 External locus of control 4 .67 1.00–4.50 1.98 (0.67) 1.90 (0.67)^a 2.05 (0.67)^b .01

Time 2 Aggressive behaviour 10 .86 1.00–3.10 1.34 (0.30) 1.40 (0.33)^a 1.29 (0.26)^b .03

Note. N = 1,466. ^a,b Pairwise means differ at p < . 008.

Peers’ acceptance of aggression

Peers’ acceptance of aggression was measured at T1 and T2 by participants’ appraisal of how accepted aggressive behaviour was within their peer group. Participants were asked to read a vignette describing a provocation scenario based on Möller and Krahé (2009).
Imagine one of your friends is extremely angry with one of your classmates because he/she treated your friend in a mean and unfair way in front of others in the school break. After school, your friend coincidentally meets the person again, and this time the two are alone. Immediately he/she starts quarreling with your friend again, saying nasty things.
Participants were presented with a version referring to a same-sex peer and asked to rate how acceptable most of their peers would find each of six possible reactions on a four-point scale from 1 (not at all okay) to 4 (totally okay). Three items described physical forms of aggressive behaviour (“to kick him/her,” “to punch him/her,” “to push him/her”), and three items referred to relational forms of aggression (“to spread rumours about him/her,” “to mock him/her,” “to speak badly about him/her”). Internal consistency was high at each data wave, as shown in Table 1.

Control beliefs

Participants’ locus of control was measured by the Multidimensional Measure of Children’s Perception of Control (MMPC; Connell, 1985) at T1 and T2. The MMCP is a 48-item self-report instrument, including one dimension of internal and two dimensions of external control perceptions (powerful others, unknown), each assessed in three specific behavioural domains (cognitive, social, physical) and one general domain. However, due to time restrictions, only a subset of the original items could be administered in our study, resulting in a four-item scale for external locus of control (two items from the subscale “powerful others control” in the social domain: “If I want my classmates to think I am an important person, I have to be friends with the really popular kids,” “If I want to be an important member of my class, I have to get the popular kids to like me;” two items from the subscale “unknown control” in the social domain: “When another kid doesn’t like me, I usually don’t know why,” “If somebody doesn’t like me, I usually can’t work out why”). Participants indicated the degree to which they did not know why certain outcomes occur (unknown control) and the degree to which other people brought about the respective outcomes (powerful others control), using a five-point scale from 1 (not at all true) to 5 (totally true). As measures of internal consistency are sensitive to the number of scale items (Cortina, 1993), the reliability of the control beliefs scale was acceptable, given that it only consisted of four items (see Table 1).

Procedure

All participants were tested individually by trained experimenters using paper-and-pencil questionnaires. Active consent was obtained from all students and, additionally, from parents of participants under the age of 18. Instruments and procedure were approved by the Ethics Committee of the University of Potsdam as well as the Ministry for Education in the Federal State of Brandenburg, Germany, where the study was conducted.

Plan of analysis

All hypotheses were examined by structural equation modelling using the Mplus Software, version 7.30 (Muthén & Muthén, 1998–2015). Mean scores of the two facets of physical and relational aggression and peer acceptance of aggression were used as indicators for the latent factors of the participant’s aggressive behaviour and the acceptance of aggressive behaviour in the peer group, respectively. The two facets of external locus of control were used to model the latent factor of locus of control. To account for the variance that an indicator shared with itself over time, we specified indicator-specific factors for each construct.

Missing data

Missing data analysis yielded missing data rates of 10.2% for peers’ acceptance of aggression, 4.3% for external control beliefs, and below 1% for aggression at T1. The higher proportion of missing data on the measures of peers’ acceptance of aggression and external control beliefs was due to an error in the compilation of the questionnaire which was only noticed and corrected after the data collection had started. This means that the likelihood of missing observations does not depend on any observed or unobserved values, justifying the treatment of missings as “missing completely at random” (MCAR). This assumption is further supported by the low correlations between testing date and T1 measures (all rs < .08). At T2, the rate of missing data was below 1% on all variables.

Missing data were handled by a multiple imputation approach, using the mice software package (van Buuren & Groothuis-Oudshoorn, 2011) in R. Multiple imputation is a regression-based procedure for dealing with missing data that is considered as a “state-of-the-art” missing data technique (Schafer & Graham, 2002) and has been shown to be superior to traditional techniques for dealing with missing data, such as listwise deletion, pairwise deletion, or mean substitution (Enders, 2010). In total, 25 imputed data sets were generated. Trace plot analyses showed a good convergence after 50 iterations of the imputation algorithm.¹

Clustering of observations

As participants were nested within schools, we had to deal with the non-independence of observations from participants in the same schools (Geiser, Eid, Nussbeck, Courvoisier, & Cole, 2010; Julian, 2001). Standard errors and test statistics of covariance analysis were corrected using the robust maximum likelihood estimator (MLR) together with a “type complex” modelling approach, using school membership at T1 as cluster variable.

Model fit

As the chi-value is sensitive to sample size, degrees of freedom, and violations of multivariate normality, even minor differences between the model’s implied and observed covariance matrix may lead to model rejection (Bollen, 1989; Tucker & Lewis, 1973). Accordingly, evaluation of model fit was based on the comparative fit index (CFI), the root-mean-square error of approximation (RMSEA), and the standardized root mean residual (SRMR). A good model fit is indicated by a CFI higher than .95, a RMSEA coefficient of less than .05, and a SRMR smaller than .08 (Hu & Bentler, 1998; Schermelleh-Engel, Moosbrugger, & Müller, 2003). As the same drawbacks of the absolute chi-square test apply to the chi-square difference test (Brannick, 1995), the CFI index was used to compare nested models by gender and age, for which a value smaller or equal to .01 indicates a non-significant difference (Chen, 2007; Cheung & Rensvold, 2002).

Results

Descriptive statistics and correlations

The means, standard deviations, and gender differences for all manifest constructs are presented in Table 1. To analyse gender differences, we conducted separate ANOVAs, with gender as the independent variable and each of the T1 and T2 measures as dependent variables. Inflation of alpha error caused by multiple testing was accounted for by adopting a corrected significance level of p = .05/6 = .008.

Significant effects of gender were found for peers’ acceptance of aggression at both time points, T1: F(1, 1315) = 88.25, p < .001, η² = .06; T2: F(1, 1105) = 100.20, p < .001, η² = .08. Males were more likely than females to report that their friends would accept aggression as an appropriate interpersonal behaviour in response to the provocation scenario. Additionally, males scored higher on aggression at T1 and T2, F(1, 1460) = 29.76, p < .001, η² = .02; F(1, 1104) = 32.98, p < .001, η² = .03. Only one gender difference was found on the measures of external control beliefs. Females scored higher than males on external control beliefs at T2, F(1, 1102) = 14.60, p < .001, η² = .01.

Table 2 presents the zero-order correlations among all manifest constructs as well as their links with age at T1, separately for males and females. As expected, peers’ acceptance of aggression at T1 was significantly associated with external locus of control at T2 for both males and females. Additionally, external locus of control at T1 was significantly related to aggression at T2 for males and for females. Age was significantly related to most study variables: For both males and females, age was positively correlated with peers’ acceptance of aggression, indicating that older individuals were more likely to have peers accepting of aggression. External locus of control showed consistent positive relationships with age, again, for both males and females, indicating that feelings of non-control were more pronounced in the older age cohorts. Finally, age was positively related to aggression at T1, but not at T2 for both males and females.

Table 2.
Bivariate correlations between the model variables for males (above the diagonal) and females (below the diagonal)

1. 2. 3. 4. 5. 6. 7.

1. Time 1 Peers’ acceptance of aggression – 0.15* 0.34* 0.44* 0.18* 0.27* 0.23*

2. Time 1 External locus of control 0.19* – 0.28* 0.17* 0.33* 0.28* 0.12

3. Time 1 Aggressive behaviour 0.34* 0.33* – 0.28* 0.14 0.50* 0.11

4. Time 2 Peers’ acceptance of aggression 0.37* 0.20* 0.19* – 0.21* 0.46*** 0.10*

5. Time 2 External locus of control 0.19* 0.35* 0.20* 0.18* – 0.29* 0.17*

6. Time 2 Aggressive behaviour 0.19* 0.20* 0.39* 0.40* 0.22* – 0.03

7. Age 0.13 0.19* 0.16* 0.01 0.07 −0.02 –

Note. N = 1,466. p < .05; p < .01; p < .001.

Hypotheses-testing analyses

As a first step, confirmatory factor analyses were conducted to analyse the proposed relations between the latent constructs and their manifest indicators and to test the constructs’ measurement invariance across the data waves. We specified a restricted model that constrained the factor loadings of the indicators to be equal across time and compared this model to a baseline model in which the factor loadings were allowed to differ. The restricted model showed a good fit with the data, χ²(24) = 105.84, p < .001; RMSEA = .05, CFI = .98; SRMR = .03, and did not fit significantly worse than the baseline model (ΔCFI = .004), indicating that the assumption of weak measurement invariance was tenable and the factor loadings were comparable across time. All further analyses are based on the restricted model.

To examine the mediational role of external locus of control in the link between peers’ acceptance of aggression and the development of aggression, we specified a model as displayed in Figure 1. Additionally, to control for the influence of plausible third variables on the proposed pathways, we included gender and age as covariates in the model at this stage, before following up gender and age comparisons in the multigroup analyses reported below.

Figure 1.
Standardized path coefficients for the relationships between peers’ acceptance of aggression, external locus of control, and aggression

The hypothesized model showed a good fit with the data (χ² (36) = 130.72, p < .001; RMSEA = .04; CFI = .98; SRMR = .03). In support of Hypothesis 1, peers’ acceptance of aggression at T1 significantly predicted external locus of control in the social domain at T2 (β = .15, p < .05). This path was significant even after controlling for the temporal stability of control beliefs. Additionally, external locus of control at T1 predicted peers’ acceptance of aggression at T2 (β = .14, p < .05), suggesting a reciprocal relationship between both constructs. The more participants saw events in the social domain as controlled by external forces, the more likely they were to be surrounded by peers with a high acceptance of aggression 17 months later, controlling for the stability of both control beliefs and peers’ acceptance of aggression. In line with Hypothesis 2, aggression at T2 was significantly predicted by external locus of control at T1 (β = .16, p < .05), again even after controlling for the construct’s temporal stability. Finally, and consistent with earlier research, peers’ acceptance of aggression at T1 was significantly associated with aggressive behaviour at T2 (β = .11, p < .05).

Indirect effects

To examine the indirect pathways from the T1 constructs to aggressive behaviour at T2, we employed a parametric bootstrapping approach, suggested by Hayes and Scharkow (2013). The analyses showed that only the indirect path from peers’ acceptance of aggression via locus of control in the social domain to aggressive behaviour was statistically significant, β = .02, 95% CI (.001, .058), as indicated by a confidence interval that did not include zero. This finding supported Hypothesis 3. No other indirect effects were significant.

Moderating variables

In the preceding analysis, age and gender were included as covariates in the model. To further examine gender- and age-specific variations in the proposed developmental pathways, we specified separate multigroup models with gender (dummy coded; 1 = male, 2 = female) and age (dummy coded; 1 = “late childhood,” 10–11 years; 2 = “early adolescence,” 12–14 years; 3 = “adolescence” 15–18 years at T1) as the grouping variables. For both gender and age, we first specified a model that allowed all paths to differ between groups and compared this unconstrained model with a model that restricted all paths to be equal across groups, using the ΔCFI criteria outlined by Chen (2007) and Cheung and Rensvold (2002). For both gender and age as moderating variables, the constrained models did not fit significantly worse than the unconstrained models (gender: ΔCFI = .008; age: ΔCFI = .009), indicating that all path coefficients were comparable for boys and girls and across the three age groups. In combination, these results do not provide evidence for gender- or age-specific differences in the proposed developmental pathways from peers’ acceptance of aggression via external control beliefs to aggressive behaviour.

Discussion

Most theoretical explanations of peer influences on the development of aggression in adolescence have emphasized the reinforcement of aggressive behaviour by peers who accept aggression as normative as the driving mechanism (e.g. Dishion et al., 1996; Snyder et al., 2005). The aim of the present longitudinal study was to demonstrate a further mechanism by which peers may contribute to the development and persistence of aggressive behaviour. We argued that the lack of structured interactions in aggressive peer groups may promote feelings of non-control and helplessness, which have been shown to be proximal predictors of externalizing psychopathology, such as aggressive behaviour (e.g. Halloran et al., 1999; Han et al., 2001). Studying a large community sample of male and female children and adolescents, latent structural equation modelling was used to test our hypotheses. Additional multigroup analyses allowed us to examine gender- and age-specific variations in the proposed developmental pathways.

The path from peers’ acceptance of aggression to control beliefs

In line with our hypotheses, peers’ acceptance of aggression at T1 predicted external control beliefs at T2 in the social domain, controlling for the construct’s temporal stability and the influence of plausible third variables, such as gender and age. Accordingly, affiliation with peers who endorse aggressive behaviours may give rise to generalized and persistent feelings of non-control in social interactions. Children and adolescents who are surrounded by peers with a high acceptance of aggressive behaviour may feel less able to influence the course and outcome of social interactions in accordance with their personal wishes and needs. We argued that interactions between aggressive individuals tend to be less regulated and organized, offering only weak contingencies between the behaviours shown and the subsequent social reactions. Accordingly, individuals in aggressive peer groups cannot reliably expect their actions to be followed by the same reactions across time and situations, leading them to see social outcomes as arbitrary. Additionally, being in an aggressive peer network may increase the likelihood of becoming a target of the peers’ aggressive behaviour. Peers’ aggression has been identified not only as a significant risk-factor for the development of individual aggression but also as a risk factor for victimization, suggesting that having friends with a high acceptance of aggressive behaviour may be a source of risk rather than of protection (Huizinga, Weiher, Espiritu, & Esbensen, 2003; Sampsons & Lauritsen, 1990, Schreck, Fisher, & Miller, 2004). Repeated experiences of victimization may in themselves stimulate the development of external control beliefs (Radliff, Wang, & Swearer, 2016). Interestingly, the association between peers’ acceptance of aggression and locus of control was reciprocal, indicating that individuals holding external control beliefs were attracted, over time, to peers with a higher acceptance of aggression. Thus, individuals with a strong external locus of control tend to actively select milieus that are characterized by a high acceptance of aggressive behaviour. This reciprocity demonstrates that the synergistic interplay of selective and socializing processes may gradually contribute to the persistence and amplification of external control beliefs as well as aggressive behaviour over the course of development.

The path from control beliefs to aggressive behaviour

As expected, external locus of control in the social domain was significantly related to aggression. Again, this path remained significant after controlling for age, gender, and the temporal stability of the constructs, indicating that children and adolescents who perceive themselves as less effective and competent in controlling the outcomes of their social interactions tend to show more aggressive behaviour. The link between external locus of control and aggression may be explained by negative affect that is associated with experiences of non-control (Han et al., 2001). Being repeatedly confronted with interpersonal situations that are outside one’s volitional control is frustrating and aversive, making aggressive behaviour more likely (Berkowitz, 1989; Dollard, Miller, Doob, Mowrer, & Sears, 1939). Finally, the analysis of the indirect pathways revealed that the effect of peers’ acceptance of aggression on individual aggression was partially mediated by an external locus of control in the social domain. Together, these findings suggest that being situated within a peer group in which aggression is considered as an appropriate interpersonal behaviour contributes to the development of subjective beliefs of non-control that promote future aggressive behaviour.

Gender and age differences

Although males and females differed significantly on all measures of peers’ acceptance of aggression and self-reported aggressive behaviour, the multigroup analysis yielded no evidence for gender-specific variations of the postulated developmental pathways of aggressive behaviour. Whereas some researchers (Halloran et al., 1999; Österman et al., 1999) found significant associations between external locus of control and aggression only for one gender group, the present study suggests that external control beliefs contribute to the development of aggressive behaviour similarly for males and females. From a conceptual perspective, there is no reason to assume that experiences of non-control should have different psychological outcomes for males and females. Additionally, our findings are consistent with other studies that showed that external control beliefs are related to indicators of psychosocial adjustment, such as social acceptance or measures of depression, equally for boys and girls (e.g. Connell, 1985; Weisz, Weiss, Wasserman, & Rintoul, 1987). Nevertheless, evidence on gender differences remains inconclusive, and future research is needed to clarify the moderating role of gender in the association between external control beliefs in the social domain and the development of aggressive behaviour.

Regarding age effects, the multigroup analysis comparing participants in late childhood, early adolescence, and adolescence did not reveal significant differences in the postulated pathways. Accordingly, the present study does not provide evidence that the developmental pathways from peers’ acceptance of aggression via external locus of control to aggression are moderated by age.

Strengths, limitations and perspectives for future research

The present longitudinal study aimed to analyse the mediating role of external locus of control in the link between peers’ acceptance of aggression and the development of individual aggressive behaviour. Strengths of our study are the inclusion of two data waves covering a mean interval of 17 months and the large sample size that comprised a wide age range from late childhood into adolescence.

At the same time, some limitations of our study have to be acknowledged. The first is the reliance on a half-longitudinal design with two measurement points. As noted, two waves of data collection may be sufficient to assess partial mediation if the effect between the mediator and the outcome variable is temporally stable (stationarity assumption; Cole & Maxwell, 2003). Although we argued that in the present sample a change of causal effects is unlikely because of the stable impact of peers in the covered age group, this assumption cannot be directly tested without having at least three measurement waves. A second limitation is that the measures of all constructs were based on self-reports. Although self-reports provide rich information about an individual’s beliefs and behaviour in a variety of different contexts and time periods, they are potentially subject to various sources of inaccuracy, especially when assessing behaviours that are prone to biases due to social desirability, such as aggression. Self-presentation mechanisms, such as impression management or self-deception, may have led to an underreporting of aggressive behaviour as well as acceptance of aggression in the peer group (Paulhus, 1984, 2002). Additionally, our measure of peers’ acceptance of aggression required the participants to have valid insights into the dynamics of their peer group, an assumption that might not always hold, especially for younger participants. Multi-informant measures, including self- and peer-ratings, would have provided are more stringent test of our hypotheses. Third, due to the study’s time restrictions, we could only include a limited number of items of the scale for external locus of control, which reduced the internal consistency of this measure at both time points and did not allow us to differentiate between control beliefs in relation to success and failure outcomes. Although external locus of control has generally been shown to be associated with lower psychosocial functioning and a number of disruptive behaviours, some studies suggest that an external locus of control might be of advantage in certain situations. For example, Burish et al. (1984) found that in medical settings with only little possibilities for personal control, external control beliefs were associated with less distress and negative arousal. Similarly, attributing negative outcomes of social interactions to external causes might be less frustrating and aversive, thereby decreasing the likelihood for aggressive behaviour. Future research may provide insights into such interactions between control beliefs and situational characteristics. Fourth, although we argued that the less structured organization of aggressive peers’ interactions stimulates the development of external control beliefs, we did not actually measure the degree of organization in interactions between individuals with a high acceptance of aggression.

Despite these limitations, the present findings may help to improve the understanding of the psychological processes underlying the link between peers’ acceptance of aggressive behaviour and the development of aggression in childhood and adolescence. Consistent with the view that multiple pathways may lead to the same psychopathological behaviour (Ciccchetti & Rogosch, 1996), our research indicates that external control beliefs may help to explain how peers’ acceptance of aggression contributes to the consolidation of aggression over and above the role of social reinforcement. These findings suggest that intervention programs would benefit from incorporating measures to reduce an external locus of control in the social domain. Although not explicitly focusing on aggression, several studies have shown that it is possible to experimentally modify perceptions of personal control using different procedures, such as providing contingent versus random feedback (Whitson & Galinsky, 2008), reminding participants of personal events over which they did or did not have control (Kay, Gaucher, Napier, Callan, & Laurin, 2008), or the cognitive restructuring of beliefs about controllability (Lachman, Weaver, Bandura, Elliott, & Lewkowicz, 1992). Similarly, we believe that teaching children and adolescents that they are able to influence and modify social events may enhance their sense of control and thereby reduce aggression both directly and indirectly via its link to affiliation with peers who show a high normative acceptance of aggression. The latter aim is crucial as intervention programs are less successful with individuals in peer groups that are highly accepting of aggression (Dodge, Dishion, & Lansford, 2006).

Considering the important socializing role of peer relations, more research is needed to investigate the multiple pathways through which characteristics of the peer group may contribute to the development of both functional and dysfunctional behaviour in childhood and youth. Our findings suggest a closer analysis of how external control beliefs interact with other psychological mechanisms in explaining peer influences on aggressive behaviour.

Variable	N items	α	Range	M (SD) total	M (SD) males	M (SD) females	η²
Time 1 Peers’ acceptance of aggression	6	.88	1.00–4.00	1.65 (0.54)	1.77 (0.60)^a	1.50 (0.40)^b	.06
Time 1 External locus of control	4	.67	1.00–5.00	2.13 (0.73)	2.12 (0.74)	2.14 (0.72)	.00
Time 1 Aggressive behaviour	10	.87	1.00–3.20	1.39 (0.34)	1.43 (0.37)^a	1.34 (0.29)^b	.02
Time 2 Peers’ acceptance of aggression	6	.86	1.00–4.00	1.63 (0.48)	1.76 (0.54)^a	1.49 (0.36)^b	.08
Time 2 External locus of control	4	.67	1.00–4.50	1.98 (0.67)	1.90 (0.67)^a	2.05 (0.67)^b	.01
Time 2 Aggressive behaviour	10	.86	1.00–3.10	1.34 (0.30)	1.40 (0.33)^a	1.29 (0.26)^b	.03

	1.	2.	3.	4.	5.	6.	7.
1. Time 1 Peers’ acceptance of aggression	–	0.15***	0.34***	0.44***	0.18***	0.27***	0.23***
2. Time 1 External locus of control	0.19***	–	0.28***	0.17***	0.33***	0.28***	0.12**
3. Time 1 Aggressive behaviour	0.34***	0.33***	–	0.28***	0.14**	0.50***	0.11**
4. Time 2 Peers’ acceptance of aggression	0.37***	0.20***	0.19***	–	0.21***	0.46***	0.10*
5. Time 2 External locus of control	0.19***	0.35***	0.20***	0.18***	–	0.29***	0.17***
6. Time 2 Aggressive behaviour	0.19***	0.20***	0.39***	0.40***	0.22***	–	0.03
7. Age	0.13**	0.19***	0.16***	0.01	0.07	−0.02	–

Footnotes

Funding

The authors declared receipt of the following financial support for the research, authorship,and/or publication of this article: This research was conducted as part of the Graduate College “Intrapersonal developmental risk factors in childhood and adolescence: A longitudinal perspective”, funded by the German Research Foundation (GRK 1668).

Note

References

Allen

J. P.

Porter

M. R.

McFarland

F. C.

Marsh

McElhaney

K. B.

(2005). The two faces of adolescents’ success with peers: Adolescent popularity, social adaptation, and deviant behavior. Child Development, 76, 747–760.

Anderson

C. R.

(1977). Locus of control, coping behaviors, and performance in a stress setting: A longitudinal study. Journal of Applied Psychology, 62, 446–451.

Anderson

(2002). The code of the streets. In Gabbidon

S. L.

Greene

H. T.

Young

V. D.

(Eds.), African American classics in criminology & criminal justice (pp. 293–305). Thousand Oaks, CA: SAGE.

Berkowitz

(1989). Frustration-aggression hypothesis: Examination and reformulation. Psychological Bulletin, 106, 59–73.

Bierman

K. L.

(2004). Peer rejection: Developmental processes and intervention strategies. New York: Guilford Press.

Bollen

K. A.

(1989). Structural equations with latent variables. Oxford, UK: John Wiley & Sons.

Brannick

M. T.

(1995). Critical comment on applying covariance structure modeling. Journal of Organizational Behavior, 16, 201–213.

Buehler

R. E.

Patterson

G. R.

Furniss

J. M.

(1966). The reinforcement of behavior in institutional settings. Behaviour Research and Therapy, 4, 157–167.

Burish

T. G.

Carey

M. P.

Wallston

K. A.

Stein

M. J.

Jamison

R. N.

Lyles

J. N.

(1984). Health locus of control and chronic disease: An external orientation may be advantageous. Journal of Social and Clinical Psychology, 2, 326–332.

10.

Chen

F. F.

(2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling, 14, 464–504.

11.

Cheung

G. W.

Rensvold

R. B.

(2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233–255.

12.

Cicchetti

Rogosch

F. A.

(1996). Equifinality and multifinality in developmental psychopathology. Development and Psychopathology, 8, 597–600.

13.

Cole

D. A.

Maxwell

S. E.

(2003). Testing mediational models with longitudinal data: Questions and tips in the use of structural equation modeling. Journal of Abnormal Psychology, 112, 558–577.

14.

Connell

J. P.

(1985). A new multidimensional measure of children’s perceptions of control. Child Development, 56, 1018–1041.

15.

Connell

J. P.

Wellborn

J. G.

(1991). Competence, autonomy, and relatedness: A motivational analysis of self-system processes. In Gunnar

M. R.

Sroufe

L. A.

(Eds.), Self processes and development (pp. 43–77). Hillsdale, NJ: Lawrence Erlbaum Associates.

16.

Cortina

J. M.

(1993). What is coefficient alpha? An examination of theory and applications. Journal of Applied Psychology, 78, 98–104.

17.

Crick

N. R.

Dodge

K. A.

(1994). A review and reformulation of social information-processing mechanisms in children’s social adjustment. Psychological Bulletin, 115, 74–101.

18.

Dishion

T. J.

Nelson

S. E.

Winter

C. E.

Bullock

B. M.

(2004). Adolescent friendship as a dynamic system: Entropy and deviance in the etiology and course of male antisocial behavior. Journal of Abnormal Child Psychology, 32, 651–663.

19.

Dishion

T. J.

Spracklen

K. M.

Andrews

D. W.

Patterson

G. R.

(1996). Deviancy training in male adolescent friendships. Behavior Therapy, 27, 373–390.

20.

Dodge

K. A.

Bates

J. E.

Pettit

G. S.

(1990). Mechanisms in the cycle of violence. Science, 250, 1678–1683.

21.

Dodge

K. A.

Dishion

T. J.

Lansford

J. E.

(2006). Deviant peer influences in programs for youth: Problems and solutions. New York: Guilford Press.

22.

Dollard

Miller

N. E.

Doob

L. W.

Mowrer

O. H.

Sears

R. R.

(1939). Frustration and aggression. New Haven, CT: Yale University Press.

23.

Duke

M. P.

Fenhagen

(1975). Self-parental alienation and locus of control in delinquent girls. The Journal of Genetic Psychology: Research and Theory on Human Development, 127, 103–107.

24.

Enders

C. K.

(2010). Applied missing data analysis. New York: Guilford Press.

25.

Gadermann

A. M.

Guhn

Zumbo

B. D.

(2012). Estimating ordinal reliability for Likert-type and ordinal item response data: A conceptual, empirical, and practical guide. Practical assessment, Research & Evaluation, 17, 1–13.

26.

Geiser

Eid

Nussbeck

F. W.

Courvoisier

D. S.

Cole

D. A.

(2010). Analyzing true change in longitudinal multitrait-multimethod studies: Application of a multimethod change model to depression and anxiety in children. Developmental Psychology, 46, 29–45.

27.

Grolnick

W. S.

Ryan

R. M.

(1992). Parental resources and the developing child in school. In Procidano

M. E.

Fisher

C. B.

(Eds.), Contemporary families: A handbook for school professionals (pp. 275–291). New York: Teachers College Press.

28.

Grolnick

W. S.

Ryan

R. M.

Deci

E. L.

(1991). Inner resources for school achievement: Motivational mediators of children’s perceptions of their parents. Journal of Educational Psychology, 83, 508–517.

29.

Halloran

E. C.

Doumas

D. M.

John

R. S.

Margolin

(1999). The relationship between aggression in children and locus of control beliefs. The Journal of Genetic Psychology: Research and Theory on Human Development, 160, 5–21.

30.

Han

S. S.

Weisz

J. R.

Weiss

(2001). Specificity of relations between children’s control-related beliefs and internalizing and externalizing psychopathology. Journal of Consulting and Clinical Psychology, 69, 240–251.

31.

Harter

(1978). Effectance motivation reconsidered: Toward a developmental model. Human Development, 21, 34–64.

32.

Harter

(1982). The perceived competence scale for children. Child Development, 53, 87–97.

33.

Hayes

A. F.

Scharkow

(2013). The relative trustworthiness of inferential tests of the indirect effect in statistical mediation analysis: Does method really matter? Psychological Science, 24, 1918–1927.

34.

Bentler

P. M.

(1998). Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecification. Psychological Methods, 3, 424–453.

35.

Huizinga

Weiher

Espiritu

Esbensen

(2003). Longitudinal research in the social and behavioral sciences: an interdisciplinary series. In Thornberry

T. P.

Krohn

M. D.

(Eds.), Taking stock of delinquency: An overview of findings from contemporary longitudinal studies (pp. 47–91). New York: Springer.

36.

Julian

M. W.

(2001). The consequences of ignoring multilevel data structures in nonhierarchical covariance modeling. Structural Equation Modeling, 8, 325–352.

37.

Kay

A. C.

Gaucher

Napier

J. L.

Callan

M. J.

Laurin

(2008). God and the government: Testing a compensatory control mechanism for the support of external systems. Journal of Personality and Social Psychology, 95, 18–35.

38.

Krahé

Möller

(2010). Longitudinal effects of media violence on aggression and empathy among German adolescents. Journal of Applied Developmental Psychology, 31, 401–409.

39.

Krampen

(1989). Perceived childrearing practices and the development of locus of control in early adolescence. International Journal of Behavioral Development, 12, 177–193.

40.

Lacourse

Nagin

Tremblay

R. E.

Vitaro

Claes

(2003). Developmental trajectories of boys’ delinquent group membership and facilitation of violent behaviors during adolescence. Development and Psychopathology, 15, 183–197.

41.

Lachman

M. E.

Weaver

S. L.

Bandura

Elliott

Lewkowicz

C. J.

(1992). Improving memory and control beliefs through cognitive restructuring and self-generated strategies. Journal of Gerontology, 47, 293–299.

42.

Mathys

Hyde

L. W.

Shaw

D. S.

Born

(2013). Deviancy and normative training processes in experimental groups of delinquent and nondelinquent male adolescents. Aggressive Behavior, 39, 30–44.

43.

McCauley

Mitchell

J. R.

Burke

P. M.

Moss

S. J.

(1988). Cognitive attributes of depression in children and adolescents. Journal of Consulting and Clinical Psychology, 56, 903–908.

44.

Möller

Krahé

(2009). Exposure to violent video games and aggression in German adolescents: A longitudinal analysis. Aggressive Behavior, 35, 75–89.

45.

Muthén

L. K.

Muthén

B. O.

(1998–2015). Mplus user’s guide (7th ed). Los Angeles: Muthén & Muthén.

46.

Nesdale

Durkin

Maass

Kiesner

Griffiths

J. A.

(2008). Effects of group norms on children’s intentions to bully. Social Development, 17, 889–907.

47.

Österman

Björkqvist

Lagerspetz

K. M. J.

Charpentier

Caprara

G. V.

Pastorelli

(1999). Locus of control and three types of aggression. Aggressive Behavior, 25, 61–65.

48.

Paulhus

D. L.

(1984). Two-component models of socially desirable responding. Journal of Personality and Social Psychology, 46, 598–609.

49.

Paulhus

D. L.

(2002). Socially desirable responding: The evolution of a construct. In Braun

H. I.

Jackson

D. N.

Wiley

D. E.

(Eds.), The role of constructs in psychological and educational measurement (pp. 49–69). Mahwah, NJ: Lawrence Erlbaum Associates.

50.

Pardini

D. A.

Loeber

Stouthamer-Loeber

(2005). Developmental shifts in parent and peer influences on boys’ beliefs about delinquent behavior. Journal of Research on Adolescence, 15, 299–323.

51.

Radliff

K. M.

Wang

Swearer

S. M.

(2016). Bullying and peer victimization: An examination of cognitive and psychosocial constructs. Journal of Interpersonal Violence, 31, 1983–2005.

52.

Romi

Itskowitz

(1990). The relationship between locus of control and type of aggression in middle-class and culturally deprived children. Personality and Individual Differences, 11, 327–333.

53.

Rotter

J. B.

(1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs: General and Applied, 80, 1–28.

54.

Rotter

J. B.

(1990). Internal versus external control of reinforcement: A case history of a variable. American Psychologist, 45, 489–493.

55.

Sampsons

R. J.

Lauritsen

J. L.

(1990). Deviant lifestyles, proximity to crime, and the offender-victim link in personal violence. Journal of Research in Crime and Delinquency, 27, 110–139.

56.

Schafer

J. L.

Graham

J. W.

(2002). Missing data: Our view of the state of the art. Psychological Methods, 7, 147–177.

57.

Schermelleh-Engel

Moosbrugger

Müller

(2003). Evaluating the fit of structural equation models: Tests of significance and descriptive goodness-of-fit measures. Methods of Psychological Research, 8, 23–74.

58.

Schreck

C. J.

Fisher

B. S.

Miller

J. M.

(2004). The social context of violent victimization: A study of the delinquent peer effect. Justice Quarterly, 21, 23–47.

59.

Seligman

M. E. P.

(1975). Helplessness: On depression, development, and death. New York: W H Freeman/Times Books/ Henry Holt & Co.

60.

Skinner

E. A.

(1996). A guide to constructs of control. Journal of Personality and Social Psychology, 71, 549–570.

61.

Skinner

E. A.

Zimmer-Gembeck

M. J.

Connell

J. P.

(1998). Individual differences and the development of perceived control. Monographs of the Society for Research in Child Development, 63, v–220.

62.

Snyder

Schrepferman

Oeser

Patterson

Stoolmiller

Johnson

Snyder

(2005). Deviancy training and association with deviant peers in young children: Occurrence and contribution to early-onset conduct problems. Development and Psychopathology, 17, 397–413.

63.

Tucker

L. R.

Lewis

(1973). A reliability coefficient for maximum likelihood factor analysis. Psychometrika, 38, 1–10.

64.

van Buuren

Groothuis-Oudshoorn

(2011). Mice: Multivariate imputation by chained equations in R. Journal of Statistical Software, 45, 1–67.

65.

Watson

J. S.

(1966). The development and generalization of “contingency awareness” in early infancy: Some hypotheses. Merrill-Palmer Quarterly, 12, 123–135.

66.

Weisz

J. R.

Weiss

Wasserman

A. A.

Rintoul

(1987). Control-related beliefs and depression among clinic-referred children and adolescents. Journal of Abnormal Psychology, 96, 58–63.

67.

Werner

N. E.

Crick

N. R.

(2004). Maladaptive peer relationships and the development of relational and physical aggression during middle childhood. Social Development, 13, 495–514.

68.

White

R. W.

(1959). Motivation reconsidered: The concept of competence. Psychological Review, 66, 297–333.

69.

Whitson

J. A.

Galinsky

A. D.

(2008). Lacking control increases illusory pattern perception. Science, 322, 115–117.

70.

Yates

Kennelly

Cox

S. H.

(1975). Perceived contingency of parental reinforcements, parent-child relations, and locus of control. Psychological Reports, 36, 139–146.

71.

Zumbo

B. D.

Gadermann

A. M.

Zeisser

(2007). Ordinal versions of coefficients alpha and theta for Likert rating scales. Journal of Modern Applied Statistical Methods, 6, 21–29.