J Abnorm Child Psychol (2008) 36:539–552 DOI 10.1007/s10802-007-9198-0
Distinguishing Proactive and Reactive Aggression in Chinese Children Yiyuan Xu & Zengxiu Zhang
Published online: 13 December 2007 # Springer Science + Business Media, LLC 2007
Abstract This study examined proactive and reactive aggression and their relation to psychosocial adjustment in three samples (N=767, 368 girls, M age=10.03) of Chinese school age children. Confirmatory factor analyses revealed that a two-factor model which distinguished both proactive and reactive aggression fit the data reasonably well, and also fit the data better than a single-factor model in all three samples. The distinction between proactive and reactive aggression was found for both boys and girls. Reactive aggression was more strongly related to reciprocated friendship (negatively), peer victimization, emotion dysregulation, hostile attributions of others’ behavior in ambiguous social situations, and self-reported loneliness and social anxiety (positively) than was proactive aggression. Proactive aggression was related to positive outcome expectancies and efficacy beliefs of aggression for boys but not for girls, but the significant gender difference was only found for positive outcome expectancies. The findings suggest that proactive and reactive aggression represent two distinct forms of aggression which are associated with specific adjustment outcomes in Chinese children. Keywords Proactive aggression . Reactive aggression . Adjustment . Chinese
Y. Xu (*) Department of Psychology, University of Hawaii at Manoa, Honolulu, HI 96822, USA e-mail:
[email protected] Z. Zhang East China Normal University, Shanghai, China
Introduction Abundant research has shown that aggressive behavior predicts a variety of adjustment problems, such as peer rejection, school dropout, delinquency, and substance use (Rubin et al. 1998). In addition, aggressive and antisocial behaviors make up the largest category of childhood disorders that are referred to mental health services (Dumas 1989). However, childhood aggression is not a unitary phenomenon, and its emergence and maintenance involve distinct developmental processes that are related to varying psychosocial adjustment outcomes (Vitaro and Brendgen 2005). For instance, recent studies on the nature of children’s aggressive behavior suggest that there are two subtypes: proactive and reactive aggression (Dodge and Coie 1987). Proactive aggression, which is generally unprovoked, deliberate, goal-oriented, and motivated by the anticipation of reward (Dodge 1991), is influenced by social learning theory (Bandura 1973) and embodies the instrumental function of aggression. On the other hand, reactive aggression which refers to defensive and retaliatory behavior in response to threat or provocation (Dodge 1991), is consistent with the frustration–anger theory of aggression (Berkowitz 1962). The two subtypes were associated with different antecedents, concurrent adjustment, and social information-processing patterns (Schwarzt et al. 1998; Vitaro et al. 2002), and the distinction has shed new light on the attempts at sub-classification of childhood conduct disorder which might in turn have intervention implications (Dodge et al. 1997; Kempes et al. 2005). To date, most research on proactive and reactive aggression has been conducted in Western settings. Given that proactive and reactive aggression appear to have distinct antecedents and consequences (Dodge 1991; Vitaro
540
et al. 2002), it is important to determine whether they reflect a universal pattern of behavior characteristic of children across diverse cultural contexts or they are specific to Western populations. The purpose of the current study was to examine proactive and reactive aggression in Chinese school age children and to explore the relation of these two subtypes to children’s psychosocial adjustment. Aggressive behavior is incompatible with the traditional Chinese value system because it disrupts the social order and interferes with group functioning (Bond and Wang 1982). Therefore, it is highly prohibited in school settings. Chinese children who are aggressive generally experience serious school maladjustment and difficulties in peer relations (Schwartz et al. 2001). Moreover, aggressive children are generally viewed as “problem kids” and are closely monitored and severely punished by teachers and officials (Chen 2000). Unlike their North American counterparts who often develop biased self-perceptions (Asher et al. 1990; Boivin et al. 1995), aggressive Chinese children typically perceive themselves negatively, and report high levels of depression (Chen et al. 1995). It has been speculated that while the negative outcomes associated with aggression are similar for Western and Chinese children, the consequences of aggressive behavior are worse for Chinese children (Chen 2000; Schwartz et al. 2001). These strong cultural sanctions raise the question as to whether there is a discernible distinction between proactive and reactive aggression in a society where childhood aggression is prohibited and possibly viewed negatively regardless of subtype. Despite considerable overlap (Vitaro and Brendgen 2005; Vitaro et al. 2006b), the distinction between proactive and reactive aggression has received support in studies of Western children. For instance, Dodge and Coie (1987) assessed these two subtypes of aggression using teachers’ ratings and behavioral observations of boys’ social interaction in small playgroups. Psychometric analyses revealed a high internal consistency of teacher-rated proactive and reactive aggression as measured by the Proactive and Reactive Aggression Scale (Dodge and Coie 1987), and significant inter-rater agreement using observations. In addition, teachers’ ratings of proactive or reactive aggression were positively related to observations of each subtype respectively. Exploratory and confirmatory factor analyses also yielded two factors that were consistent with the two subtypes, particularly among boys (Day et al. 1992; Little et al. 2003; Price and Dodge 1989; Poulin and Boivin 2000a; Salmivalli and Nieminen 2002; Vitaro et al. 2006a). For example, Poulin and Boivin (2000a) examined aggression in school age boys using the Proactive and Reactive Aggression Scale (Dodge and Coie 1987) as rated by teachers and peers. Both subtypes were similarly correlated with peers’ nominations of physical aggression which
J Abnorm Child Psychol (2008) 36:539–552
supported the convergent validity of proactive and reactive aggression. Confirmatory factor analyses revealed that a two-factor model, despite the substantial correlation between the latent factors, fit the data better than a singlefactor model, supporting the distinction between proactive and reactive aggression among boys. Although the factorial distinction between proactive and reactive aggression has been replicated in boys, little is known about whether this two-factor aggression model can be applied to girls. Using a measure adapted from Dodge and Coie’s (1987) scale, Vitaro et al. (2006a) found that the two-factor aggression fit the data reasonably well for a large sample of kindergarten children which included both boys and girls. They also identified relatively comparable patterns of relations across gender between the two subtypes and children’s negative emotionality, harsh parenting, and family revenue.1 These findings suggest that the distinction between proactive and reactive aggression may be similar for both boys and girls. However, Vitaro et al. (2006a) did not directly test whether the two-factor aggression model applied similarly to both gender, or whether measurement invariance or equivalence (Little 1997) can be established for subtypes of aggression across gender. Using Dodge and Coie’s (1987) Proactive and Reactive Aggression Scale, the invariance of the two-factor aggression model can be tested by examining whether two distinct factors (i.e., proactive and reactive aggression) are found for both genders and whether the item loadings for each factor are comparable for boys and girls. Therefore, the first purpose of the current study was to test the twofactor aggression model using the Proactive and Reactive Aggression Scale, a well-validated and the most commonly used measure, with Chinese boys and girls. Building on previous findings (e.g., Vitaro et al. 2006a), we expected that both subtypes of aggression would be found and distinguished in Chinese boys and girls. However, because Dodge and Coie’s (1987) measure mainly taps physical aggression, boys were expected to have higher mean levels of proactive and reactive aggression than were girls. Consistent with the conceptual differences between the two subtypes, studies have found a distinct and theoretically consistent pattern of relations among Western children’s proactive and reactive aggression and various psychosocial correlates. Reactive aggression is generally
1 Vitaro et al. (2006b) also identified some gender differences concerning the correlations between reactive and proactive aggression (0.51 for boys, 0.46 for girls), and between family revenue and reactive aggression (−0.14 for boys, −0.19 for girls). However, these differences did not compromise the distinction between proactive and reactive aggression for both boys and girls.
J Abnorm Child Psychol (2008) 36:539–552
accompanied by anger and emotional outbursts, and has been associated with impulsivity and attention deficits (Crick and Dodge 1996; Dodge and Coie 1987; Dodge et al. 1997; Price and Dodge 1989; Schwartz et al. 1998; Shields and Cicchetti 1998; Vitaro et al. 2002). Reactively aggressive children appear to have poor social skills and display less cooperative and prosocial behavior (Price and Dodge 1989; Poulin and Boivin 2000a). Because these negative attributes are influential in peer group functioning, children who engage in reactive aggression are often disliked, have very few friends (Poulin and Boivin 2000a, b), and are sometimes victimized by others (Pellegrini et al. 1999; Schwartz et al. 1998). Over time, negative peer group experiences may develop into negative perceptions of social situations and contribute to the development of internalizing problems, such as loneliness and social anxiety (Boivin et al. 1995; Raine et al. 2006). Proactive aggression has been associated with disruptive classroom behavior (Dodge and Coie 1987), initiation of fights, and juvenile delinquency (Connor et al. 2003; Pulkkinen 1996; Raine et al. 2006; Vitaro et al. 1998, 2002). However, under some circumstances, proactive aggression may be effective in controlling resources due to its instrumental nature, and may be considered by some peers as being “cool.” Studies have shown that proactive aggression was associated with leadership and having a sense of humor, and was not associated with peer rejection or victimization (Dodge and Coie 1987; Poulin and Boivin 2000a). Moreover, because proactive aggression serves social goals, it has also been associated with the formation and the development of friendships (Poulin and Boivin 2000b). Dodge and Coie (1987) proposed that distinct social cognitive processes also characterize proactive and reactive aggression. Due to its angry and retaliatory nature, reactive aggression has been associated with a tendency to attribute hostile intent to peers’ behavior in ambiguous situations (Crick and Dodge 1996; Dodge and Coie 1987; Orobio de Castro et al. 2002, 2005; Schippell et al. 2003). Hostile attribution bias may lead to anger, frustration, and emotion dysregulation, which in turn may elicit strong negative reactions from peers, such as rejection or victimization. In contrast, studies have shown that positive evaluations of the likely consequences of aggressive behavior were associated with proactive aggression, which is largely based on external contingencies, particularly in conflict situations (Crick and Dodge 1996; Hubbard et al. 2001; Schwartz et al. 1998; Smithmyer et al. 2000). Positive outcome expectancies and efficacy beliefs about aggression may lead to inflated self-perceptions and partly explain why proactively aggressive children, unlike their reactive peers, rarely exhibit internalizing problems. The second purpose of this study was to investigate Chinese children’s proactive and reactive aggression in relation to peer relationships with regard to social preference,
541
reciprocated friendship, and peer victimization, aspects of socio-emotional functioning, such as prosocial behavior, emotion dysregulation, and self-reported loneliness and social anxiety, and socio-information processing including global evaluation, outcome expectancies, efficacy beliefs, and hostile attribution of aggression. Because aggressive behavior is highly prohibited in the Chinese context (Chen 2000), we expected that both subtypes would be negatively associated with social preference and prosocial behavior. In addition, based on the conceptual distinction between reactive (frustrated and retaliatory behavior) and proactive aggression (instrumental and deliberated actions), we expected to find differences between the two subtypes in relation to the psychosocial correlates. Specifically, we predicted that reactive aggression would be more strongly associated with peer victimization, emotion dysregulation, self-reported loneliness and social anxiety, and hostile attribution, whereas proactive aggression would be more strongly associated with reciprocated friendship, and global evaluation, outcome expectancies, and efficacy beliefs about aggression. Three independent samples were used for two reasons. First, due to a lack of data on proactive and reactive aggression in non-Western contexts, examining the twofactor aggression model across multiple samples would ensure the robustness of our findings. Second, we were interested in a large number of psychosocial correlates of proactive and reactive aggression that are theoretically relevant for the distinction between the two subtypes and it would be difficult to assess all of them in one sample. Therefore, we tested the two-factor model of proactive and reactive aggression in all three samples, and examined subtypes of aggression in relation to social preference in Sample I, social preference, reciprocated friendship, peer victimization, emotion dysregulation, and prosocial behavior in Sample II, and social information-processing and self-reported internalizing problems (loneliness and social anxiety) in Sample III.
Method Participants Each of the three samples consisted of fourth and fifth grade children recruited from an elementary school in Shanghai, PRC. This age range represents a developmental period where the nature and function of aggressive behavior are relatively organized and stable (Dodge and Coie 1987; Poulin and Boivin 2000a). All the three schools were from the same school district in Shanghai. Two schools (Sample I and Sample III) had three 4th and three 5th grade classes, with approximately 30–40 students per class. Another
542
J Abnorm Child Psychol (2008) 36:539–552
school (Sample II) had four 4th and four 5th grade classes, with approximately 40 students per class. Parents were contacted by teachers before the data collection, and were given information regarding the study goals and procedures. The consent rates were 95.9, 95.7, and 95.7% for the three samples. Seventeen children were absent during questionnaire administration. Therefore, data were available for 767 children (M age=10.03 years): 226 children (6 classes), for Sample I, 327 children (8 classes) for Sample II, and 214 children (6 classes) for Sample III. Table 1 summarizes the descriptive information about the three samples. Because of the “one-child” policy in Mainland China, most participants were the only children in their family. In addition, more than 90% of the children were from two-parent nuclear families. The three samples were compared on children’s age, parents’ education levels, percentage of the only children, percentage of the nuclear families, and teachers’ rated proactive and reactive aggression. The samples did not differ in any of these variables. Procedure In each school, teachers rated children’s proactive and reactive aggression. In addition, teachers’ ratings, peers’ nominations, and children’s self-reports were used to assess aspects of children’s peer relationships, socio-emotional functioning, and socio-information processing (see details below). All of the questionnaires had been translated and back translated by paid language consultants (e.g., Schwartz et al. 2001) or by graduate students (e.g., Xu et al. 2007) who were bilingual native Chinese speakers, and had been used with Chinese children in previous studies (e.g., Schwartz et al. 2001; Xu et al. 2007). All the questionnaires were administered in Mandarin, the official language in Mainland China. Measures for Samples I, II and III Teachers’ Ratings of Children’s Proactive and Reactive Aggression Teachers rated children’s proactive aggression
(e.g., this child threatens or bullies other children in order to get his/her own way; α=0.85, 0.84, 0.84 for the three samples, respectively), and reactive aggression (e.g., when this child has been teased or threatened, s/he gets angry easily and strikes back; α=0.93, 0.91, 0.91) on a 5-point scale (1=never, 5=almost always), using the measure adopted from Dodge and Coie (1987). The ratings within each scale were averaged to form teachers’ ratings of proactive and reactive aggression. Residualized Proactive and Reactive Aggression Scores Consistent with previous studies that identified moderate to high correlations between proactive and reactive aggression (e.g., Day et al. 1992; Dodge and Coie 1987; Poulin and Boivin 2000a; Raine et al. 2006), analyses in the current study revealed moderate correlations of 0.46, 0.49, 0.40 between the two subtypes in three samples. Following Raine et al. (2006), residualized proactive and reactive aggression scores were created to form relatively “pure” proactive and reactive aggression scores. Specifically, each aggression variable was regressed onto the other and the residuals were saved. “Pure” or residualized reactive aggression represented the part of the reactive aggression scores that could not be predicted from proactive aggression; “pure” or residualized proactive aggression represented that part of proactive aggression scores that could not be predicted from reactive aggression. Both raw and “pure” (residualized) proactive and reactive aggression scores were examined in relation to the psychosocial correlates (see details below).
Measures for Sample I Peers’ Nominations of Children’s Aggression A peer nomination questionnaire for physical aggression that did not specifically differentiate proactive and reactive aggression, was group administered to children to examine convergent validity of two subtypes of aggression. Each child was asked to nominate all of the peers in his/her class
Table 1 Descriptions of the three samples
Number of boys (girls) Children’s age in years M (SD)
Boys Girls
Mothers’ educationa M (SD) Fathers’ educationa M (SD) Percentage of the two-parent nuclear families (%) Percentage of the only-children (%) a
Sample I (N=226)
Sample II (N=327)
Sample III (N=214)
119 (107) 10.14 (0.62) 10.09 (0.66) 2.15 (0.84) 2.39 (0.78) 92.0 95.1
169 (158) 9.91 (0.48) 9.91 (0.43) 2.04 (0.84) 2.26 (0.79) 92.4 93.4
111 (103) 10.16 (0.60) 10.09 (0.67) 2.07 (0.82) 2.21 (0.77) 95.8 94.2
Education was coded as: 1=less than or equal to 9 years (equivalent to junior high school); 2=between 10 and 12 years (equivalent to senior high school); 3=between 13 and 16 years (equivalent to some college education or with a Bachelors degree); and 4=more than 16 years (equivalent to some graduate education or with a Masters or Ph.D. degree
J Abnorm Child Psychol (2008) 36:539–552
543
who best fit behavioral descriptors (e.g., kids who hit or push others; α=0.91) adapted from Schwartz et al. (2001). The total nominations each child received for these items were standardized within class to adjust for the differences in the number of nominators, and were averaged to form peers’ nominations of aggression.
kids who share with peers; α=0.83) using the Prosocial Behavior Subscale adapted from Schwartz et al. (2001). The total number of nominations was standardized within each class and treated as peers’ nomination of prosocial behavior.
Social Preference A peer nomination questionnaire for peer acceptance/rejection was group administered to children (Schwartz et al. 2001). Children nominated all of the peers in their classes whom they liked most, and whom they liked least. The total nominations each child received for these two items was calculated and standardized within class. A social preference score was generated from the standardized difference between the liked most and liked least scores.
Measures for Sample III
Measures for Sample II Social Preference Children’s social preference was assessed and calculated using the same procedure in Study 1. Reciprocated Friendship Following Schwartz et al. (2001), children were asked to nominate all of their friends in their class with no limit. The number of reciprocally nominated friendships was standardized within each class and treated as a rating of reciprocated friendship. Peer Victimization Children were asked to nominate all of the peers in their class who best fit behavioral descriptors (e.g., kids who get pushed or hit by others; α=0.85) using the Victimization Subscale adopted from Schwartz et al. (2001). The total number of nominations was standardized within each class and treated as peers’ nominations of victimization. Emotion Dysregulation Teachers completed the Emotion Regulation Checklist (ERC; Shields and Cicchetti 1998), a 24-item rating scale that assesses aspects of children’s capacities for emotional self-regulation, such as affective lability, intensity, valence, flexibility, and situational appropriateness (e.g., “Can recover quickly from episodes of upset or distress”; α=0.94). Items were rated on a 5-point rating scale (1=never true, α=almost always true). Following Schwartz and Proctor (2000), we calculated an emotion dysregulation summary variable based on the mean of the 24 items, with positively valanced items reverse-coded. The ERC has been shown to have high internal consistency and strong construct validity in previous studies with Chinese children (Chang et al. 2003). Prosocial Behavior Children were asked to nominate all of their classmates who best fit behavioral descriptors (e.g.,
Social Information-Processing Interview The Chinese version of the revised Social Information Processing Interview (SIPI; Quiggle et al. 1992; Schwartz and Proctor 2000) was group-administered to the children. The Chinese SIPI contained three scenarios describing ambiguous peer provocations (e.g., getting bumped while waiting in line). Children were read each scenario aloud and asked to imagine that the situation was actually occurring to them. They were first asked to rate their attribution of the peer’s behavior (e.g., “How much do you think that classmate who bumped into you was trying to be mean?”) on a 5-point scale (1=not at all, 5=very much). Higher scores indicated endorsement of more hostile attribution. Children were then offered an aggressive solution to the situation (e.g., “One thing you could do is hit or push the kid who took your place in line…”) and asked to respond to a series of items assessing their global evaluation of the aggressive response (e.g., “How good an idea do you think this would be?”), positive outcome expectancy for the behavior (e.g. “Would you get your place back in line if you hit the other kid?”), and efficacy beliefs regarding aggression (e.g., “How easy or hard would it be for you to do this?”). Children also rated these items on a 5-point scale. Higher scores indicated more aggressive social–cognitive styles. Children’s mean scores across the relevant items for the three scenarios constituted their hostile attributions (α=0.76), global evaluation of aggressive strategies (α=0.70), positive outcome expectancies (α=0.85), and efficacy beliefs scores (α=0.87). The SIPI has shown high internal consistency and construct validity in previous Western and Chinese studies (e.g., Schwartz and Proctor 2000; Toblin et al. 2005; Xu 2007). Self-reported Loneliness and Social Anxiety Children’s loneliness at school was assessed using Asher and Wheeler (1985) Loneliness Scale. Children rated 16 statements (e.g., I have nobody to talk to) on a 5-point scale (1=not at all true; 5=always true about me). Children’s responses were summed to form self-reported loneliness (α=0.82). This measure has showed high internal consistency and construct validity in studies of Western (Asher and Wheeler 1985) and Chinese children (Xu et al. 2007). Parkhurst and Asher (1992) Interpersonal Concern Scale (ICS) was used to assess children’s social anxiety related to their interpersonal concern. Children rated seven items
544
J Abnorm Child Psychol (2008) 36:539–552
Table 2 Means and standard deviations of all the variables by gender in samples I, II, and III Boys
Sample I (N=226) Proactive aggression (T) Reactive aggression (T) Undifferentiated aggression (P) Social preference (P) Sample II (N=327) Proactive aggression (T) Reactive aggression (T) Social preference (P) Reciprocated friendship (P) Peer victimization (P) Emotion dysregulation (T) Prosocial behavior (P) Sample III (N=214) Proactive aggression (T) Reactive aggression (T) Global evaluation of aggression (S) Outcome expectancies of aggression (S) Efficacy beliefs of aggression (S) Hostile attributions (S) Loneliness (S) Social anxiety (S)
Girls
F
M
SD
M
SD
2.28 2.24 0.28 −0.52
0.83 1.21 1.11 1.57
1.56 1.57 −0.31 0.58
0.51 0.80 0.47 1.44
58.43** 23.72** 25.48** 29.92**
2.22 2.37 −0.32 −0.25 0.16 2.72 −0.15
0.83 1.28 1.63 0.84 0.97 0.91 0.73
1.66 1.60 0.34 0.27 −0.17 2.36 0.17
0.62 0.89 1.51 1.07 0.58 0.85 0.95
48.57** 38.61** 14.18** 23.99** 14.01** 13.74** 11.80**
2.24 2.27 1.63 1.87 1.90 2.07 2.50 1.82
0.88 1.26 0.59 0.88 0.67 0.98 0.72 0.70
1.50 1.79 1.43 1.62 1.55 2.05 2.51 1.88
0.50 1.06 0.71 0.76 0.59 0.97 0.66 0.72
56.74** 8.97** 5.24* 4.85* 16.85** 0.03 0.00 0.31
T Teachers’ ratings, P Peer nominations, S Self reports *p