Child Development, January/February 2008, Volume 79, Number 1, Pages 170 – 185

Affect-Congruent Social-Cognitive Evaluations and Behaviors Ka¨tlin Peets

Ernest V. E. Hodges

University of Turku

St. John’s University

Christina Salmivalli University of Turku and University of Stavanger

This study examined whether the affect children feel toward peers would influence children’s social-cognitive evaluations and behaviors. The sample consisted of 209 fifth-grade children (11- to 12-year-olds; 119 boys and 90 girls). For each child, 3 target peers (liked, disliked, and neutral) were identified via a sociometric nomination procedure. The names of the targets were then inserted into hypothetical vignettes in which the target peer’s behavior had a negative consequence for the child. After each vignette, questions about intent, outcome expectations, and self-efficacy beliefs were asked. In addition, self-reports regarding relationship-specific proactive and reactive aggression and regarding victimization were collected. The results demonstrate that children social-cognitively differentiate between the relationship types and that relationship-specific evaluations are associated with relationship-specific behaviors.

It is well documented that the way children behave in peer settings is associated with their social standing in the peer group (i.e., how much others like or dislike the child). Over time, peer reputation can become even more important than social behavior in predicting future likability (Denham & Holt, 1993). Different social-cognitive processes are proposed to be, at least partly, responsible for establishing and maintaining the link, for instance, between aggression and rejection. However, although social information processing (SIP) has been studied in children who are disliked by the peer group, there is a lack of studies examining different social-cognitive processes that could potentially contribute to the maintenance of disliking or liking someone. The purpose of this study was to examine whether children process information and behave differently depending on whether the peer is liked, disliked, or neutral. According to the SIP model (Crick & Dodge, 1994), behavioral enactment is a consequence of several mental processes, such as encoding and interpretation of cues, formation and clarification of goals, accessing possible behavioral responses from memory, and evaluating possible outcomes for the selected behavioral responses, and choosing the ‘‘best’’ one.

This work was supported by the Academy of Finland Grants 202554 and 68884 to C.S. Our special thanks go to all the children and teachers for making the present research possible. Correspondence concerning this article should be addressed to Ka¨tlin Peets, Department of Psychology, University of Turku, FI-20014 Turku, Finland. Electronic mail may be sent to [email protected].

In addition, although Crick and Dodge acknowledged the importance of social context, Lemerise and Arsenio (2000) explicitly specified that the emotional valence of the relationship between two individuals can have an effect on interpretation and other social-cognitive processes. Accordingly, provocation by a liked peer might be explained differently than the same behavior enacted by a disliked peer. Most of the studies on SIP have investigated differences in social cognitions between aggressive and nonaggressive children. For instance, it has been shown that aggressive children have a tendency to interpret intentions of others as hostile (e.g., Dodge, Murphy, & Buchsbaum, 1984; Dodge & Somberg, 1987; Waldman, 1996), are more confident in their ability to aggress (Perry, Perry, & Rasmussen, 1986), and evaluate aggression more favorably in terms of its appropriateness (Erdley & Asher, 1998) and expected consequences (Perry et al., 1986). As a result, aggressive children’s SIP style is often viewed as deficient. Moreover, Crick and Dodge (1994) proposed that ‘‘processing patterns and [behavioral] tendencies, as they are formed, come to act like personality-like characteristics that guide behavior’’ (p. 81). According to this assumption, a tendency to attribute hostility to others and aggressive behavior itself become gradually inherent characteristics of aggressive children.

# 2008, Copyright the Author(s) Journal Compilation # 2008, Society for Research in Child Development, Inc. All rights reserved. 0009-3920/2008/7901-0012

Social-Cognitive Evaluations and Behaviors

The view of trait-like processing has developed, at least partly, due to the designs used to study SIP. Namely, research on social-cognitive processes has been mainly conducted in a paradigm where children are asked to think about a hypothetical peer. However, it is likely that children’s social-cognitive evaluations of hypothetical peers are influenced by their everyday interactions with actual peers. In other words, cognitions elicited by thinking about a hypothetical peer might reflect representations of a specific peer from the child’s social environment. As we do not know whom the hypothetical peers actually represent (a generalized peer, a specific actual peer), more ecological validity could be gained by examining social-cognitive processes elicited by thinking of an actual peer. Furthermore, a few studies conducted so far suggest that the behavioral reputation of the target peer has a stronger effect on a subject’s subsequent social-cognitive processes compared to the subject’s own behavioral status. For instance, Dodge (1980) found that both aggressive and nonaggressive subjects attributed more hostile intent to aggressive targets than to nonaggressive targets, suggesting that aggressive children are not rigid in their socialcognitive evaluations. In the study by Perry et al. (1986), aggressive as well as nonaggressive children were more confident about gaining peer group approval (e.g., friends’ approval) when their aggression would be targeted toward aggressive peers than toward nonaggressive peers, and both groups of children thought that aggressive targets would display fewer signs of distress than nonaggressive targets. Moreover, emotional valence of the relationship can have a stronger effect on social-cognitive evaluations than the more objectively measured behavioral reputation of the target peers. For instance, in a study by Peets, Hodges, Kikas, and Salmivalli (2007), although children nominated more aggressive and less prosocial peers as their enemies, the relationship with the target was always the strongest predictor of hostile attributions. More specifically, children attributed the most hostility to their enemies and the least to their friends. Concordantly, a few other studies have shown that social-cognitive evaluations depend on whether the target is liked or disliked by the subject. More favorable evaluations are made about the members of the in-group (e.g., friends) than the out-group (e.g., enemies) (Guerin, 1999). For example, children tend to hold the peers they dislike more responsible for the negative behaviors than the peers they like (Hymel, 1986). A similar result was found by Guerin, where dispositional explanations were made for positive behaviors enacted by liked peers and for negative behaviors enacted by disliked peers, with

171

the opposite pattern holding for situational explanations. Similarly, aggressive children could be expected to behave the most aggressively toward the peers they dislike. It has been shown that most aggressive acts take place only in 20% of the dyads (Dodge, Price, Coie, & Christopoulos, 1990) and that aggressive nonvictims seem to be especially selective in their targets (Card, Isaacs, & Hodges, 2000). Moreover, as was stated by Cairns, Cairns, Neckerman, Gest, and Garie´py (1988), ‘‘The finding that aggressive adolescents have lower levels of general popularity and likeability . . . may serve to obscure the ‘concealed competencies’ that permit these persons to survive in particular social contexts’’ (p. 821). Therefore, many aggressive children are likely to have prosocial skills (e.g., bistrategic children; Hawley, Little, & Pasupathi, 2002), which might be actualized toward the peers aggressive children like. Thus, positive or negative affect felt toward a peer might be a crucial factor determining how the behavior of the peer is interpreted and what type of behavior is enacted as a response. Interpretation and other social-cognitive processes in turn could be regarded as tools working toward maintaining an existing representation of the peer. Once the representation is established, it might be difficult to alter it, even when the behavior of the peer changes more favorably (Denham & Holt, 1993; Hymel, 1986). Reis, Collins, and Berscheid (2000) talk about a ‘‘self-reinforcing relationship feedback loop.’’ For instance, when a child attributes hostility to another peer, even if the peer did not intend to harm, there is a high probability for the child to behave hostilely toward the peer, which in turn triggers a hostile response from the peer, thereby confirming the child’s original view of the peer. Two dimensions of hostile behavior (reactive and proactive aggression) were examined in this study. Whereas reactive aggression is defined as an angry ‘‘defensive reaction to a perceived threatening stimulus’’ (Dodge & Coie, 1987, p. 1147), proactive aggression is seen as a goal-directed harmful behavior. It has been shown that although the two forms are highly correlated, a two-factor model fits the data better than a one-factor solution (Poulin & Boivin, 2000b). Moreover, past research has supported theorized differential cognition-specific links (Crick & Dodge, 1994) with these two subtypes of aggression. For instance, whereas reactive aggression is associated with hostile attributional bias (Dodge & Coie, 1987), proactive aggression is related to positive outcome expectations (Schwartz et al., 1998). The discriminative validity of these two dimensions of aggression is further supported by findings that reactive aggression is associated with victimization (e.g., Poulin & Boivin, 2000b;

172

Peets, Hodges, and Salmivalli

Schwartz et al., 1998) but proactive aggression is positively associated with leadership (Poulin & Boivin, 2000b), and children who are proactively aggressive, but not reactively aggressive, have friends who are similar to them in their behavioral tendencies (Poulin & Boivin, 2000a). Although studies focusing on examining relationship effects with regard to social-cognitive processes in children and adolescents are still rare, during the past decade, researchers have begun to take a closer look at the unique effects of the relationship between two individuals using statistical tools such as social relations modeling (Coie et al., 1999; Hubbard, Dodge, Cillessen, Coie, & Schwartz, 2001; Simpkins & Parke, 2002). For example, Coie et al. (1999) found that the relationship between the child and the peer accounted for at least as much variance in proactive and reactive aggression as the child’s and the peer’s own aggressive dispositions. Moreover, dyadic reactive aggression was best predicted by relationshipspecific hostility rather than hostile tendencies in general (Hubbard et al., 2001). Likewise, our study was designed to examine variability in social cognitions and behaviors due to relationship types with different affective valences. More specifically, the purpose of this study was to analyze different social-cognitive evaluations (hostile attributions, outcome expectations, and self-efficacy beliefs) toward individually nominated liked, disliked, and neutral peers and to evaluate the cognition – behavior links. Hypotheses were tested using multilevel modeling (Muthe´n & Muthe´n, 1998 – 2004), which allows to partial variance in scores into two components: (a) variance due to different target types and (b) variance due to interindividual differences. More specific hypotheses were the following. On the basis of the results of the study by Peets et al. (2007), it was expected that social cognitions would show target specificity. When the disliked peer served as a target, we anticipated that children would attribute more hostility and have higher efficacy beliefs for aggression. With regard to target-specific behaviors, the present study focused on reactive and proactive aggression. It was hypothesized that higher levels of aggression would be reported toward disliked peers. In addition, we expected that children would report being victimized more frequently by their disliked peers (Card & Hodges, 2007). The second aim of this study was to test whether certain social cognitions would differentially relate to reactive and proactive aggression. In several studies, children have been divided into extreme groups on the basis of arbitrary cutoff scores on reactive and proactive aggression. This approach excludes a great

deal of original variance in the scores. In addition, in order to test whether there exist unique correlates for reactive and proactive aggression, it is necessary to partial out the common variance between the two forms of aggression (Poulin & Boivin, 2000a, 2000b). For example, when Poulin and Boivin (2000b) used this approach, they found that whereas reactive aggression was associated with a negative peer status and with more victimization by peers, proactive aggression was related to a positive status and to less victimization by peers. Therefore, also in our analyses, we controlled for reactive aggression when predicting proactive aggression and the other way around. More specifically, on the basis of Crick and Dodge’s (1994) review, as well as the results from other studies, it was hypothesized that attribution of hostility would be associated with reactive aggression, whereas proactive aggression would be linked to positive instrumental outcome expectations (Schwartz et al., 1998) and to higher self-efficacy beliefs for aggression (Yoon, Hughes, Cavell, & Thompson, 2000). Additionally, we tested whether children’s own behavioral reputation (i.e., aggression and being victimized) would explain individual differences in cognitions. Finally, most of the research on SIP has been conducted with boys (Crick & Dodge, 1994). Although studies have found that boys expect less guilt and parental disapproval for aggression (e.g., Perry, Perry, & Weiss, 1989) and are generally more aggressive (at least overtly) than girls, it is unclear whether the associations between social cognitions and adjustment differ by gender. Orobio de Castro, Veerman, Koops, Bosch, and Monshouwer (2002) conducted a meta-analysis on hostile attribution of intent and aggressive behavior. As girls were underrepresented in the studies they reviewed, they could not draw any conclusions regarding a moderation effect of gender. In addition, Huesmann and Guerra (1997) found that although boys regarded aggression as a more appropriate behavior than girls, longitudinal associations between normative beliefs for aggression and aggressive behavior did not differ for boys and girls. We expected that both boys and girls would show affect-congruent SIP. However, we did not have specific hypotheses with regard to gender moderation of the cognition – behavior links.

Method Participants The sample consisted of 209 fifth-grade students (11- to 12-year-olds; 119 boys and 90 girls) from six

Social-Cognitive Evaluations and Behaviors

public schools in the area of Turku (a town in the southwestern part of Finland, with approximately 175,000 inhabitants). The original sample included 266 students; however, 19 children (7%) did not receive parental permission and 1 child was absent from school during the period of data collection. In addition, for some of the children (n 5 24), the same peer was identified as a disliked and neutral target, and some of the subjects (n 5 13) did not nominate anyone who could fit the description of the disliked and/or neutral peer. These children were thus excluded from the subsequent analyses, resulting in a final sample of 209 students. Mean number of students per class was 19 (SD 5 6.43). Altogether, 11 classes participated in the study. As the Finnish population is socioeconomically and ethnically/racially (i.e., majority of the population is Caucasian and speaks Finnish) homogeneous, information about family socioeconomic and ethnic background was not collected. Procedure All the measures (including measures not central to the current research questions) were group administered by psychology undergraduates in April to May 2005. Children filled in the questionnaires during regular school hours. In order to assess target-specific social cognitions, target peers had to be identified beforehand. Thus, the entire testing procedure was divided into two sessions 1 – 2 weeks apart from each other. All the measures described later (except for the social-cognitive questionnaire) were administered during the first testing session. The first and second testing sessions lasted approximately 60 and 45 min, respectively. At the end of both testing sessions, children were thanked and given a small gift (i.e., candy) for their cooperation. Participants were ensured about the confidentiality of their answers and were encouraged to trust their own opinion. Measures Identification of the Target Peers Three target peers (i.e., liked, disliked, and neutral) were identified for each child by means of sociometric questions: ‘‘Who are the classmates you like the most?’’ (liked peer); ‘‘Who are the classmates you like the least?’’ (disliked peer); and ‘‘Who are the classmates you do not really like or dislike?’’ (neutral peer). All questions were read out loud by a research assistant. Girls were presented a sheet with all the names of the girls, and boys received a sheet with the names of the boys. The names were displayed across

173

the top of the sheet, whereas each item was presented in a row on the left. Participants were asked to put an X beneath the names of up to three classmates who best fit the description in an item (self-nominations were not allowed). After each of the three questions, children ranked their chosen peers, writing a number from 1 to 3 beneath the respective X-es, with 1 representing the peer who fit the best. After the data collection, for each of the three items, pairs of children with reciprocated nominations were identified. If the child had several reciprocal relationships, the peer who received a higher ranking from the child was chosen as a representative of the specific target type. For children whose nominations were not reciprocated by any of the same-sex classmates, the highest ranked, unilaterally nominated peer was chosen as a target. In our final sample, 28 (13.4%) participants did not receive a single reciprocal nomination for liking. For dislike and neutral nominations, the same frequency index was even higher, 102 (48.8%) and 121 (57.9%), respectively. Thus, participants who had only unilateral relationship with a target were also included in the analyses to preserve power. Moreover, perception of someone as liked and disliked, without necessarily being reciprocated, could be expected to trigger respective social-cognitive evaluations and behaviors (Peets et al., 2007). Thus, in our study, some of the children had a unilateral relationship, whereas others had a reciprocal relationship with the target.

Self-Reported Target-Specific Behavior A 19-item self-report questionnaire was administered to the participants. Only items assessing reactive and proactive aggression and those assessing victimization were selected to fulfill the purpose of the current study. Three reactive and three proactive items were adapted from the scales developed by Dodge and Coie (1987) (e.g., reactive aggression, ‘‘I blame . . . in fights’’; proactive aggression, ‘‘I threaten and bully. . .’’). Four victimization items were adapted from the scale by Perry, Kusel, and Perry (1988) (e.g., ‘‘I get called names by . . .’’). Girls and boys received the questionnaire forms with the names of all their same-sex classmates, and they were asked to rate how often they behaved in the described way toward each classmate or how often each of the classmates behaved in the described way toward them (i.e., victimization). For each item, a separate page with the names of the girls or boys was provided. Ratings were provided on a 5-point frequency scale (0 5 never, 1 5 almost never, 2 5 sometimes, 3 5 almost all the time,

174

Peets, Hodges, and Salmivalli

4 5 all the time) and were later averaged across the respective items. Thus, for each child, we were able to calculate self-reported aggression and victimization scores regarding each same-sex peer. However, for the purpose of this study, only scores in relation to each child’s three target peers (i.e., liked, disliked, and neutral) were used in the analyses. Peer-Nominated Behavior Although the nomination procedure consisted of 20 questions, only items measuring reactive and proactive aggression and those measuring victimization are described. The content of aggression and victimization scales was identical to the one used in assessing target-specific behaviors. Thus, three items measured reactive (e.g., ‘‘Who are the kids who blame others in fights?‘‘) and three items assessed proactive aggression (e.g., ‘‘Who are the kids who threaten and bully others?’’; Dodge & Coie, 1987). The victimization scale consisted of four items (e.g., ‘‘Who are the kids who get called names by other kids?’’; Perry et al., 1988). All items were read aloud by a research assistant. Girls had a sheet with the names of the girls, and boys received a sheet with the names of the boys. The names were displayed across the top of the sheet, and each item was presented in a row on the left. Participants were asked to put an X beneath the names of up to three classmates who best fit the description in an item. In addition, if they did not find anyone who could fit the description, they were allowed to write no one behind the item. The scales were internally consistent (Cronbach’s alpha was .90 for reactive aggression, .80 for proactive aggression, and .87 for victimization). Nominations were summed for each item and divided by the number of possible nominations. Items measuring each scale were then averaged. Social-Cognitive Questionnaire Before administering the social-cognitive questionnaire, three target peers had been identified for each child and their names hand-written into the questionnaires. Thus, each child received a questionnaire form with a unique combination of the names of their targets. The questionnaire consisted of three parts (one for each target type), with four hypothetical stories per target type. Most of the stories were adapted from the Social-Cognitive Assessment Profile used by Hughes, Meehan, and Cavell (2004). In each vignette, a target peer’s behavior had a negative consequence for the child. The intent of the target was always displayed as ambiguous. Children were asked

to pretend that the story really happened to them. Below are two examples of the stories; instead of a blank, the name of the target appeared. 1. Imagine that you are walking home after school. It is a rainy day and there are mud puddles everywhere. Suddenly . . . runs by you and hits a puddle, and mud splashes all over you. All your clothes are now dirty and wet, and you are cold. 2. Imagine that you are sitting at your desk and you see . . . writing a note and passing it around the room. When the other kids read the note, they laugh. The note does not get passed to you. After each story, children responded to questions about intent attributions, outcome expectations (relational and instrumental), and self-efficacy beliefs. The hypothetical situations were identical across the targets, but the order of the presentation of the target types varied across classrooms. In order to not confuse children, the questions were presented in the same sequence. Hostile attributions. After each story, children were asked two questions measuring intent attributions. Four alternatives were presented to the first question, asking why something happened (e.g., ‘‘Why do you think . . . splashed you with mud?’’). Two options described the target with intention to hurt the child (e.g., he/she wanted to hurt me), and two other options described a negative consequence resulting due to accidental or benign reasons (e.g., he/she was in a hurry). Children had to write an X in front of the response option they regarded as the most probable. Answers were coded as 0 (nonhostile) and 1 (hostile). The second question was always ‘‘Was . . . mean or not?’’ Answers were similarly coded as 0 (nonhostile) and 1 (hostile). Internal consistencies (Cronbach’s alphas) were .83 (liked peer), .89 (disliked peer), and .90 (neutral peer). Scores were derived by averaging the items across the four vignettes for each target type, with one being the maximum possible score. Outcome expectations. Two types of questions assessed relational (e.g., ‘‘If you pushed . . . into the puddle, would he/she like you after that?’’) and instrumental (e.g., ‘‘If you pushed . . . into the puddle would it stop him/her from splashing you in the future?’’; ‘‘If you said to . . . ‘If you don’t tell me what the note was about, I will write one about you as well,’ would you get to know what the note was about?’’) outcome expectations for using aggression. Children’s responses were coded on a scale from 1 to 4 (1 5 definitely not, 2 5 maybe not, 3 5 maybe yes, 4 5 definitely yes). Internal consistencies for relational and

Social-Cognitive Evaluations and Behaviors

instrumental outcome expectations were .80 (liked target), .79 (disliked target), and .79 (neutral target), and .67 (liked target), .64 (disliked target), and .75 (neutral target), respectively. Self-efficacy beliefs. Finally, children evaluated how easy or hard it would be for them to behave aggressively toward a target (e.g., ‘‘How easy or hard would it be for you to push . . . into the puddle?’’). Response options were 1 5 very easy, 2 5 easy, 3 5 hard, 4 5 very hard. Answers were reverse scored in order to ease the interpretation. Thus, the higher the score, the higher the self-efficacy. The three scales were internally consistent (Cronbach’s alphas were .86, .90, and .89 for liked, disliked, and neutral targets, respectively).

Results Unilateral Versus Reciprocal Targets First, we examined whether children with unilateral versus reciprocal targets differed on targetspecific cognition and behavior scores. As the majority of the participants (
87%) had a positive reciprocal relationship with the target, analyses (t tests) were only conducted regarding disliked and neutral peers. Only relational outcome expectations differed toward unilaterally and reciprocally disliked peers, t(205)5 2.03, p , .05. More specifically, children whose disliking was not reciprocated expected greater relational outcomes for aggression compared to those whose negative nomination had been reciprocated (M 5 1.80 vs. M 5 1.63). Behavioral Reputation of Targets We conducted within-subject analyses of variance to explore whether the targets differed on peer-nominated proactive and reactive aggression and on victimization scores. We found that target type had an effect on proactive aggression, F(2, 416) 5 16.03, p , .001, gp2 5 .07; reactive aggression, F(2, 416) 5 56.08, p , .001, gp2 5 .21; and victimization, F(2, 416) 5 60.58, p , .001, gp2 5 .23. Pair-wise comparisons (Bonferroni tests with p set at .05) demonstrated that disliked peers were reactively (M 5 0.31) and proactively (M 5 0.18) more aggressive compared to liked peers (Mreactive 5 0.12, Mproactive 5 0.11) and neutrals (Mreactive 5 0.14, Mproactive 5 0.09). However, there were no differences between liked peers and neutrals. With regard to victimization, disliked peers were more victimized (M 5 0.25) than liked peers (M 5 0.07) and neutrals (M 5 0.11). In addition, neutral peers had higher victimization scores compared to the peers children liked.

175

We were also interested in examining whether disliked peers who were identified as targets more than once differed from disliked peers who served as targets only once. We excluded duplicate cases and found that unique targets had lower scores on proactive aggression, t(112) 5 3.20, p , .01; reactive aggression, t(112) 5 5.31, p , .001; and victimization, t(112) 5 3.49, p , .001. Multilevel Modeling Means and standard deviations for the targetspecific variables are presented in Table 1. The data were analyzed with two-level modeling using Mplus 3.0 (Muthe´n & Muthe´n, 1998 – 2004). Multilevel models are used in the case of hierarchical data when individual observations are not independent from one another. Most common examples are when students are nested in classrooms or family members within families. Another possible application of multilevel modeling is when there are several data points for each individual (i.e., repeated measures approach). In the present study, cognitions and behaviors for three target types were nested within individuals. Thus, three targets represented the within level (target level) and individuals constituted the between level (individual level). Each individual had an identification number that was used as an indicator of the cluster. Hence, the number of clusters was equal to our sample size. Throughout the analyses, a robust estimator MLR (normality-based maximum-likelihood estimation with robust standard errors) was used to correct for the violation of normality assumption (Muthe´n & Muthe´n, 1998 – 2004). In addition, missing values on the study variables were handled using missing data method provided in Mplus 3.0 (Muthe´n & Muthe´n, 1998 – 2004). This method allows using all the available data to compute the parameter estimates. First, null models were run with all the target-specific cognition and behavior variables. These analyses yielded intraclass correlations (ICCs) and withinand between-level variance estimates. Next, two-level models were specified to examine a specific pattern of cognitions and behaviors associated with each target type. We thus focused on within level; however, intercepts were allowed to vary across individuals. Third, we examined whether cognitions would predict aggression toward a target and being victimized by a target. Associations were examined at both levels. Finally, in order to validate individual-level findings that were based on self-reported behaviors, we examined associations between peer-nominated behaviors and mean levels (i.e., intercepts) of cognitions.

176

Peets, Hodges, and Salmivalli

Table 1 Means and Standard Deviations of Target-Specific Variables Entire sample Liked Variable

M

Social Cognitions HA 0.09 ROEAG 2.49 IOEAG 2.85 SEAG 2.09 Behaviors REA 0.40 PROA 0.17 VICT 0.26

Neutral

Boys Disliked

Liked

Girls

Neutral

Disliked

Liked

Neutral

Disliked

SD

M

SD

M

SD

M

SD

M

SD

M

SD

M

SD

M

SD

M

SD

0.20 0.77 0.69 0.88

0.26 1.97 2.50 2.35

0.33 0.65 0.70 0.90

0.51 1.71 2.36 2.71

0.37 0.59 0.67 0.92

0.09 2.54 2.88 2.34

0.19 0.78 0.69 0.92

0.30 2.09 2.48 2.64

0.36 0.70 0.75 0.91

0.53 1.75 2.31 3.01

0.38 0.59 0.72 0.85

0.10 2.43 2.82 1.77

0.20 0.76 0.68 0.71

0.19 1.80 2.54 1.97

0.28 0.54 0.62 0.73

0.49 1.66 2.42 2.31

0.36 0.60 0.62 0.86

0.57 0.34 0.46

0.42 0.16 0.31

0.64 0.35 0.53

0.84 0.36 0.88

0.84 0.65 0.89

0.33 0.19 0.28

0.51 0.38 0.52

0.55 0.22 0.39

0.73 0.43 0.64

0.95 0.46 1.01

0.89 0.78 0.96

0.49 0.15 0.23

0.63 0.30 0.38

0.26 0.06 0.21

0.45 0.15 0.30

0.70 0.22 0.70

0.76 0.40 0.76

Note. HA 5 hostile attributions; IOEAG 5 instrumental outcome expectations for aggression; PROA 5 proactive aggression; REA 5 reactive aggression; ROEAG 5 relational outcome expectations for aggression; SEAG 5 self-efficacy beliefs for aggression; VICT 5 victimization.

targets. Overall, within each level, social-cognitive correlates of reactive and proactive aggression were similar in direction and magnitude. However, conclusions about differential social-cognitive correlates awaits the inclusion of appropriate controls (e.g., controlling for proactive aggression, when evaluating links to reactive aggression). In addition, as there was a significant difference in target- and individual-level variance estimates for boys and girls with regard to some of the target-specific variables (see below), we also constructed multigroup models where associations between study variables were estimated for girls and boys separately.

Bivariate correlations between study variables at the target level (see above diagonal) and individual level (see below diagonal) are displayed in Table 2. Interpretation of these correlates varies by level. For instance, regarding associations between reactive aggression and victimization, a target-level correlation (r 5 .72) indicates that the more children perceived to be victimized by a target, the more frequent reactive aggression they reported toward the same target. The individual-level correlation within the same variable pair (r 5 .58) denotes that children who perceived to be more frequently victimized across the three targets also reported more frequent reactive aggression across the Table 2 Target- and Individual-Level Bivariate Correlations Between Study Variables Variable VICT REA PROA HA ROEAG IOEAG SEAG VICTP REAP PROAP GENDER

1

2 .72***

.58* .48 — .44y .08 .01 .54** .47** .44** .37**

.67** — .24 .18 .27* .14 .37** .36** .18y

3 .51*** .66*** — .15 .13 .39** .04 .35** .45** .25**

4

5

6

7

.56*** .47*** .37***

.33*** .21*** .19** .49***

.38*** .22*** .08 .40*** .57***

.42*** .36*** .27*** .58*** .30*** .25***

— — — — — — —

.43* .64*** .26* .11 .01 .27*

.50*** .08 .10 .12 .04

.03 .11 .24** .45***

8

9

10

.34*** .11 .07

.73*** .01

.09

Note. Target- and individual-level correlations are displayed above and below the diagonal, respectively. Attribution of hostility was specified as a target-level variable. Gender and peer-nominated behaviors were specified as individual-level variables. HA 5 hostile attributions; IOEAG 5 instrumental outcome expectations for aggression; PROA 5 proactive aggression; PROAP 5 peer-nominated proactive aggression; REA 5 reactive aggression; REAP 5 peer-nominated reactive aggression; ROEAG 5 relational outcome expectations for aggression; SEAG 5 self-efficacy beliefs for aggression; VICT 5 victimization; VICTP 5 peer-nominated victimization. y p , .10. *p , .05. **p , .01. ***p , .001.

Social-Cognitive Evaluations and Behaviors

ICCs First, null models were tested to examine ICCs. In our data, an ICC equal to 1 would mean that children did not differentiate between the targets; however, there were differences between individuals in terms of their average scores across the targets; in contrast, an ICC equal to 0 would indicate that subjects discriminated between the target peers, with no significant variance, however, between subjects’ mean scores. ICCs and variance estimates are displayed in Table 3. Results showed that scores on the hostile attribution scale did not vary significantly between individuals, indicating that the whole variance in the respective scores was due to children differentiating between the three target types. In contrast, selfefficacy beliefs showed the highest ICC (among boys as well as among girls). With regard to target-specific behaviors, the majority of the variance existed between the target types. Whereas among boys, most of the variables had also significant variance at the individual level, among girls, relational outcome expectations for aggression, proactive aggression, and victimization varied only by target type. Target Specificity To test the associations at the target level, social cognition and behavior scores were regressed on dummy-coded target types (i.e., disliked peers vs. other targets, liked peers vs. other targets). For instance, when we compared disliked peers with other targets, relationship with the disliked peer

177

was coded as 1, whereas other targets served as a reference group (coded as 0). Separate models were built for comparing disliked peers with other targets and liked peers with other targets. In addition, we also compared pairs of targets (e.g., liked targets vs. neutral targets, disliked targets vs. neutral targets). For example, when we compared liked and disliked targets with neutral targets, social cognition and behavior scores were regressed on two dummycoded target types simultaneously, and thus, the relationship with the neutral target served as a reference group. Results showed that when the target was a disliked peer, children attributed more hostility to the target (B 5 .34, SE 5 0.03, b 5 .45), expected fewer positive relational (B 5 .52, SE 5 0.05, b 5 .37) and instrumental (B 5 .32, SE 5 0.05, b 5 .26) outcomes, and felt more efficacious about aggressing against the target (B 5 .48, SE 5 0.05, b 5 .38), compared to the two other target types. The pattern of results remained the same when the disliked target was compared to neutral and liked targets separately. Differences, however, were overall larger between disliked and liked peers than between disliked and neutral peers. When the liked peer served as a target, subjects inferred less hostility from the peer (B 5 .29, SE 5 0.02, b 5 .39), expected greater relational (B 5 .66, SE 5 0.05, b 5 .48) and instrumental (B 5 .42, SE 5 0.05, b 5 .35) outcomes for aggression, and felt less efficacious about aggressing against the peer (B 5 .43, SE 5 0.04, b 5 .34). Results were in the same direction when liked peers were compared to neutrals.

Table 3 ICCs and Variance Estimates Entire sample Variable

ICC

Social Cognitions HA .02 ROEAG .17 IOEAG .33 SEAG .57 Behaviors REA .22 PROA .35 VICT .12

Boys

Girls

WV

SE

BV

SE

ICC

WV

SE

BV

SE

ICC

WV

SE

BV

SE

.13*** .47*** .34*** .38***

0.01 0.04 0.03 0.04

.00 .09*** .17*** .49***

0.01 0.03 0.03 0.05

.01 .18 .28 .49

.14*** .48*** .41*** .45***

0.01 0.05 0.04 0.05

— .10** .16*** .43***

— 0.03 0.04 0.07

.03 .12 .41 .55

.11*** .46*** .25*** .29***

0.01 0.06 0.03 0.04

.00 .06 .18*** .35***

0.01 0.05 0.04 0.06

.40*** .15*** .45***

0.05 0.03 0.06

.11*** .08** .06y

0.03 0.03 0.03

.23 .36 .13

.45*** .21*** .55***

0.08 0.05 0.09

.13** .12** .08y

0.05 0.04 0.05

.18 .21 .03

.34*** .07*** .31***

0.06 0.02 0.05

.08* .02 .01

0.03 0.01 0.02

Note. In order to obtain variance estimates for boys, attribution of hostility had to be specified as a target-level variable. BV 5 between-level variance; HA 5 hostile attributions; ICC 5 intraclass correlation; IOEAG 5 instrumental outcome expectations for aggression; PROA 5 proactive aggression; REA 5 reactive aggression; ROEAG 5 relational outcome expectations for aggression; SEAG 5 self-efficacy beliefs for aggression; VICT 5 victimization by a target; WV 5 within-level variance. y p , .10. *p , .05. **p , .01. ***p , .001.

178

Peets, Hodges, and Salmivalli

Regarding target-specific reactive and proactive aggression, children reported more frequent reactive (B 5 .43, SE 5 0.06, b 5 .33), as well as proactive (B 5 .19, SE 5 0.04, b 5 .24) aggression toward the peers they disliked. In addition, they perceived to be victimized by disliked peers more often than from other peers (B 5 .59, SE 5 0.06, b 5 .44). Children made a distinction between disliked and liked peers, as well as between disliked peers and neutrals. However, no difference existed between liked peers and neutrals. Multigroup analyses yielded a very similar pattern of results among boys and girls. Analyses where we compared pairs of targets (e.g., liked peers vs. neutrals, disliked peers vs. neutrals, liked peers vs. disliked peers) revealed that for girls, there was a nonsignificant difference between liked and disliked peers with regard to proactive aggression, and between liked peers and neutrals regarding victimization. Surprisingly, girls also tended to report less frequent reactive (B 5 .23, SE 5 0.07, b 5 .19) and proactive (B 5 .08, SE 5 0.03, b 5 .14) aggression toward neutrals than toward the targets they liked. Boys reported similar levels of proactive aggression toward liked peers and neutrals. As it is possible that differences between the target types might have been due to disliked targets being generally more aggressive and victimized than other targets, we conducted additional analyses where the relationship with the target (disliked peer vs. other target peers) and target peers’ aggressiveness (reac-

tive and proactive aggression) and victimization were entered simultaneously into the equations. We found that after controlling for the target peers’ behavioral reputation, the affect felt toward the peer still remained a significant predictor (Table 4). Cognitions Predicting Behaviors Next, although many zero-order correlations between social-cognition and self-reported aggression/victimization scores were significant (Table 2), we were interested in examining whether social cognitions would explain unique variance in reactive and proactive aggression. Two separate models were constructed for self-reported reactive and proactive aggression, and parameter estimates were modeled at the target as well as at the individual level (attribution of hostility was included as a predictor only at the target level). Thus, at the target level, we examined whether cognitions in a specific relationship would be predictive of aggression in the same relationship. At the individual level, we investigated whether individuals with a certain pattern of cognitions across the three target types were also more aggressive. To control for the overlap between the two forms of aggression, reactive aggression was included as a control variable in the model when predicting proactive aggression, and the other way around. Parameter estimates are presented in Table 5. Parameter estimates are drawn from the models with all the predictors (also nonsignificant) included.

Table 4 Predicting Social Cognitions and Behaviors From the Relationship With the Target and Behavioral Reputation of the Targets Target peers’ behavioral reputation

Relationship with the target Variable

B

Social Cognitions HA .27 ROEAG .43 IOEAG .24 SEAG .39 Behaviors REA .23 PROA .08 VICT .41

SE

b

0.03 0.05 0.05 0.05

.36*** .31*** .20*** .31***

0.06 0.04 0.06

.17*** .10* .30***

Victimization B

SE

Reactive aggression b

.51

0.20

.16**

.42 .42

0.22 0.14

.13y .20**

Proactive aggression

B

SE

b

B

SE

.41 .61 .35

0.10 0.17 0.17

.24*** .19*** .13*

.53

0.21

.16*

.55

0.23

.19*

.50

0.29

.14y

1.24

0.24

.33***

b

Note. Behavioral reputation was measured by means of peer nominations. Relationship with the target was a dummy-coded variable (1 5 disliked peers, 0 5 other targets). Parameter estimates are drawn from the models with all the predictors (also nonsignificant) included but only significant associations are displayed. HA 5 hostile attributions; IOEAG 5 instrumental outcome expectations for aggression; PROA 5 proactive aggression; REA 5 reactive aggression; ROEAG 5 relational outcome expectations for aggression; SEAG 5 self-efficacy beliefs for aggression; VICT 5 victimization by a target. y p , .10. *p , .05. **p , .01. ***p , .001.

Social-Cognitive Evaluations and Behaviors

179

Table 5 Cognitions Predicting Behaviors (General Model) Target level Variable Reactive Aggression PROA HA IOEAG Proactive Aggression REA HA IOEAG SEAG Being Victimized REA PROA HA IOEAG

Individual level

B

SE

b

.96 .35 .10

0.08 0.10 0.06

.56*** .19*** – .09y

.36 .11 .07

0.05 0.05 0.04

.61*** .10* .11y

R2

B

SE

b

.80

0.17

.62***

.44

0.14

.56**

.10

0.04

.27**

.49

0.17

.53**

.49

.39

.45

.45

.60 .51 .24 .38 .20

0.08 0.13 0.09 0.07

.48*** .14y .20*** – .17**

R2

.37

Note. Parameter estimates are drawn from the models with all the predictors (also nonsignificant) included but only significant associations are displayed. HA 5 hostile attributions; IOEAG 5 instrumental outcome expectations for aggression; PROA 5 proactive aggression; REA 5 reactive aggression; ROEAG 5 relational outcome expectations for aggression; SEAG 5 self-efficacy beliefs for aggression. y p , .10. *p , .05. **p , .01. ***p , .001.

First, reactive aggression was regressed on four cognition variables at the target level and on three cognition variables at the individual level (attribution of hostility was specified as a within-level variable). At the target level, hostile attributions were associated with reactive aggression over and above the effects of proactive aggression. The more the child attributed hostility to a peer, the more frequent reactive aggression was reported toward the same peer. Moreover, the positive association between proactive and reactive aggression at the target level indicates that both types of aggression are typically delivered to the same peers. There was also a negative path from instrumental outcome expectations to reactive aggression, but this association reached only marginal significance. At the individual level, only proactive aggression was a significant predictor, suggesting that children who had higher levels of proactive aggression across the targets had also higher means of reactive aggression across the relationship contexts. A similar two-level model was run for proactive aggression. Results showed that at the target level, hostile attributions and instrumental outcome expectations explained variance in proactive aggression over and above the effects of reactive aggression. However, the path from instrumental outcome expectations was only marginally significant. At the individual level, the paths from reactive aggression and self-efficacy for aggression reached statistical significance. Accordingly, children with higher levels

of self-efficacy for aggression across the targets also reported higher frequency of proactive aggression across the targets. In addition to reactive and proactive aggression, we examined whether social-cognitive evaluations would make a unique prediction to children’s evaluation of being victimized by a target. We used the same analytical procedure as before. Victimization scores were regressed on both forms of aggression and cognitions at both levels. At the target level, reactive and proactive aggression were significant predictors of victimization; however, proactive aggression reached only marginal significance. Thus, reactive aggression toward a target was particularly linked to being victimized by the same target. Moreover, hostile attributions (positively) and instrumental outcome expectations (negatively) predicted being victimized by a target. Thus, the more children attributed hostility to a peer and the less they expected aggression directed at the peer to result in positive instrumental outcomes, the more children reported being victimized by the peer. At the individual level, reactive aggression was a significant predictor of victimization, indicating that the more children reported reactive aggression across the targets, the more they tended to report being victimized across the targets. The same paths were estimated separately for boys and girls. As the between-level variance in relational outcome expectations, proactive aggression, and

180

Peets, Hodges, and Salmivalli

aggression across the target types. In addition, proactive aggression was correlated with self-efficacy beliefs, thereby validating our finding that especially proactively aggressive children felt more efficacious about using aggressive behavior across the targets. Whereas none of the correlations turned out to be significant for girls, boys who were considered more victimized by their peers reported lower relational (r 5 .28, p , .05) and instrumental (r 5 .18, p , .10) outcome expectations. In addition, proactively aggressive boys felt more confident about aggressing at the targets (r 5 .22, p , .01).

victimization was not significant for girls, paths from/to those variables were fixed to zero. Parameter estimates are presented in Table 6. Among boys, attribution of hostility was a significant target-level predictor of all behavioral variables. In addition, boys who had lower levels of instrumental outcome expectations for aggression reported being more victimized by the target. At the individual level, boys who had higher self-efficacy beliefs across the targets reported more frequent proactive aggression across the relationship types. Among girls, attribution of hostility was a significant target-level predictor of reactive aggression and victimization. And, although there were some associations between outcome expectations and proactive/reactive aggression, those associations disappeared when we did not control for the overlap between the two forms of aggression.

Discussion The results of this study demonstrate that children’s social-cognitive evaluations and aggressive behaviors vary as a function of the affect children feel toward the peers. Thus, the current study adds to the literature on SIP and social adjustment by providing evidence that disliking and liking of someone are accompanied by affect-congruent social cognitions and behaviors. All hypotheses were tested using multilevel modeling where three target types were treated as nested within individuals. The ICCs and individual- and

Associations Between Peer-Nominated Aggression/ Victimization and Cognitions Bivariate correlations (Table 2) showed an overlap between self- and peer-evaluated behaviors. Moreover, results indicated that children who were more victimized according to peer nominations expected lower levels of positive relational outcomes for Table 6 Cognitions Predicting Behaviors (Multigroup Models) Boys

Girls

Target level Variable Reactive Aggression PROA HA ROEAG IOEAG Proactive Aggression REA HA ROEAG IOEAG SEAG Being Victimized REA HA IOEAG

B

SE

.89 .36

0.09 0.14

.58*** .20**

.41 .16

0.08 0.08

.63*** .13*

b

Individual level R2

B

SE

b

.77

0.18

.65***

.50

0.20

.59*

.54

.49*** .12* .24***

B

SE

b

1.21 .32 .15 .24

0.17 0.13 0.06 0.08

.57*** .18* .16* .20**

.31

0.06

.64***

.06 .12

0.03 0.05

.10

0.05

.23*

.61

0.17

.64***

.50 .58

0.09 0.13

Individual level R2 .47

.47

.61 0.11 0.12 0.09

R2 .42

.50

.54 .24 .29

Target level B

SE

b

R2 .17

.43

.15* .21**

.44

.63 .51*** .35***

Note. As there was a nonsignificant between-level variance in proactive aggression and victimization for girls, associations at the individual level were set to zero. Parameter estimates are drawn from the models with all the predictors (also nonsignificant) included but only significant associations are displayed. HA 5 hostile attributions; IOEAG 5 instrumental outcome expectations for aggression; PROA 5 proactive aggression; REA 5 reactive aggression; ROEAG 5 relational outcome expectations for aggression; SEAG 5 self-efficacy beliefs for aggression. *p , .05. **p , .01. ***p , .001.

Social-Cognitive Evaluations and Behaviors

target-level variance estimates revealed that most of the social cognition and behavior scores had significant variance at both levels. Yet, the majority of the variance in the scores was attributable to differences among the target types. For instance, for hostile attributions, as much as 98% of the variance in the scores was due to children differentiating between the targets, suggesting that children clearly interpret the negative behaviors enacted by liked, disliked, and neutral peers differently. This is in line with the study by Peets et al. (2007) who also found a very low ICC for the attribution of hostility (i.e., ICC 5 .13). In contrast, individual variation in the mean scores was significant and highest in the case of self-efficacy beliefs for aggression, indicating that some of the subjects had higher self-efficacy beliefs across the targets. These findings suggest that some processing mechanisms (i.e., hostile attributions) are more contextually based, whereas others (e.g., self-efficacy) are relatively more trait like and thus likely to influence behavior across interaction partners. When the target was a disliked peer, both boys and girls attributed more hostility and expected less positive relational as well as instrumental outcomes for aggression. At the same time, children held higher self-efficacy beliefs for aggression toward the peers they disliked. As expected, the greatest difference existed between liked and disliked peers. Hence, our findings show that children have different socialcognitive evaluations toward the peers they like and dislike, providing support for relationship-specific SIP. Current results are in line with other studies (Burgess, Wojslawowicz, Rubin, Rose-Krasnor, & Booth-LaForce, 2006; Hymel, 1986; Peets et al., 2007), showing that children tend to interpret the behaviors enacted by disliked and liked peers differently, giving the ‘‘benefit of the doubt’’ to friends. However, it is possible that children’s own status moderates the degree of differentiation children make between friends and other types of peers. For instance, future research could examine whether rejected children who have poor-quality friendships are less likely to give the benefit of the doubt even to their friends. In addition, children were more confident about aggression resulting in positive relational as well as instrumental outcomes in the case of liked peers. It could be argued that children believe that aggression would be less efficient when directed at a disliked peer due to the possibility of an angry retaliation on the behalf of the target. In addition, at least some children might become submissive when facing the peer they dislike. It is likely that the relationship history with a liked peer ‘‘tempers negative emotional reactions to stressful social events’’ (Burgess et al.,

181

2006, p. 380) and provides children with a confidence that problematic situations can be worked out without escalating negativity. Moreover, although children can have more positive outcome expectations for aggression targeted at a liked peer, they might be even more confident about positive consequences when more constructive strategies (i.e., assertive) were enacted. Thus, analyzing expected outcomes for aggression in reference to anticipated consequences for other possible behavioral strategies would provide additional insight into this issue. Furthermore, although children anticipated less positive instrumental and relational outcomes for aggression in the case of disliked peers, they still felt more efficacious about aggressing toward the peers they disliked. As expected, more frequent reactive and proactive aggression was targeted at the peers boys and girls disliked. In addition, in accordance with the study by Card and Hodges (2007), children perceived to be victimized more frequently by their disliked peers. Interestingly, girls tended to report slightly higher aggression toward liked peers than toward neutrals. It is possible that this result reflects aggressive girls’ friendships. Alternatively, aggression might be a more common way of handling conflicts with peers with whom children (and especially girls) have a relationship history. It might be ‘‘safer’’ to engage in aggression toward friends than toward neutrals as the probability of conflict resolution is higher among friends (see, e.g., Newcomb & Bagwell, 1995). Still, it should be pointed out that a large proportion of children did not report aggression even toward the peers they disliked. Other social behaviors, then, are likely to also be differentially distributed in the peer group based on how positive or negative children feel toward others. For instance, prosociality could be expected to be targeted more frequently at liked peers than toward disliked peers or toward neutrals. And, in interactions with disliked peers, avoidant behavior might be more frequently enacted than with friends (or neutrals) where, perhaps, direct confrontation may be used. Differentiation between liked peers and neutrals might also increase with age (Berndt & Perry, 1986) and be stronger for girls (see also Nangle, Erdley, Zeff, Stanchfield, & Gold, 2004). With regard to cognition – behavior links, several studies have analyzed only zero-order correlations between social-cognitive processes and social behaviors (e.g., Schwartz et al., 1998), which do not provide information about whether certain social-cognitive evaluations have unique associations with different behaviors. Therefore, we examined whether targetspecific social-cognitive evaluations would explain unique variance in target-specific reactive and

182

Peets, Hodges, and Salmivalli

proactive aggression. It was found that among boys, attribution of hostility was a significant predictor of reactive as well as of proactive aggression. Among girls, attribution of hostility predicted only reactive aggression. Although we did not directly study dyadic exchanges, our results give evidence for the reciprocal nature of peer interactions. For instance, the more frequently children reacted aggressively to the incidents with a target peer and the more hostility they attributed to the peer, the higher the likelihood of also being victimized by the peer. This result applied to both boys and girls. Studies on the dyadic nature of peer interactions have indeed shown that most aggressive acts take place only in a small number of dyads (Dodge et al., 1990), and dyadic hostile attributions are especially associated with dyadic reactive aggression (Hubbard et al., 2001). Moreover, among boys, the lower the instrumental outcome expectations, the more victimization was reported, providing support for the previously suggested explanation that children might know from their own experience that aggression is not effective toward the aggressive target. It should be noted that our sample sizes were smaller for multigroup analyses, resulting in less power. To overcome this limitation, larger sample pools should be employed in the future. We point out that although some unique differential associations between the social cognitions and the two forms of aggression were found, the overall pattern of cognitions specific to reactive or proactive aggression was not very strong. One possible explanation is that we did not assess social-cognitive evaluations in relation to use of each form of aggression. For example, Werner and Nixon (2005) demonstrated that normative beliefs about physical aggression were associated with physical aggression, and beliefs about acceptability of relational aggression had a unique association with relational aggression. Thus, self-efficacy, for example, might be more strongly and differentially linked to proactive aggression if children were asked about their confidence in their ability to use proactive aggression. Although target-specific aggression and victimization were based on self-reports, and the results could be affected by a shared method bias, target-specific behaviors were correlated with peer-nominated behaviors, thereby providing stronger support for the ‘‘real’’ association between the study variables. A similar method to study target-specific aggression was successfully employed by Card and Hodges (2006). They used a self-nomination inventory where each participant nominated an unlimited number of same-sex peers to whom they delivered each behavioral description (e.g., ‘‘I pick on him/her’’). Like-

wise, they demonstrated the validity of their instrument by correlating self-reported proportion of targets aggressed against with a peer-nominated measure of aggression. However, in order to reduce the mono-method bias, future studies could analyze relationship-specific behaviors using, for instance, a peer-rating technique (Coie et al., 1999) where each child rates pairs of classmates/grade-mates on the frequency of aggressive acts (i.e., how often Child A starts fights with Child B, and vice versa). We also examined whether differences among the target types would be due to disliked peers being more aggressive/victimized at the classroom level compared to other peers. As was found by Peets et al. (2007), personal affect felt toward the peer remained a significant predictor of social-cognitive processes and behaviors even when we controlled for the targets’ general behavioral reputation. Although aggression is usually associated with rejection by other peers at the classroom level, aggressive children can be liked by other aggressive peers, as children tend to like peers who are similar to them in their social status and behavioral styles (e.g., Nangle et al., 2004). However, it is possible that for children who have aggression in their behavioral repertoire, target peer’s aggressiveness, especially if directed toward the child, might be one of the potential factors that increases the likelihood of cognitions becoming actualized into aggressive behaviors toward this peer. In addition, in the present study, the peers who were identified as disliked targets for two or more children were generally more aggressive (and victimized) at the classroom level than those who were identified as disliked targets for only one child, providing support for the well-known finding that aggression is one of the main reasons during middle childhood and early adolescence why someone is disliked. Yet, there are certainly very many individual reasons for not liking or liking someone, especially when children get older (Hayes, Gershman, & Halteman, 1996). However, when only one child dislikes the peer, and when the dislike is not mutual, the affect toward the peer might be less stable compared to the situation when several children have a negative disposition toward the same peer. For instance, if an aggressive child who is disliked by several children tries to change by behaving in more accepted, perhaps prosocial, ways, affect-congruent biases held by several peers might make it extremely difficult for this child to change in social status. Thus, although it is common for children to have peers they dislike or like, and social-cognitive tendencies supporting this existing view of the peer should be regarded as a normative mechanism by means of which peer interactions

Social-Cognitive Evaluations and Behaviors

function, situations where several peers support one another’s negative representations of the particular child can be the most detrimental for the child, especially when the child does not have any supportive relationships. Thus, at the same time when practitioners teach social skills to the child and intervene in problematic behaviors, chronic representations and expectations toward this child at the dyadic or at the group level should not be left unnoticed. In addition, considering the results of the current study, researchers using peer nominations to assess, for instance, stability of aggressive behaviors should be cautious when interpreting their findings. It is possible that the results reflect more strongly the stability of the biases held by the group rather than the stability of aggressive behavior per se. As stated by Hymel (1986), ‘‘Although positive behaviors performed by disliked children will not necessarily be discounted or minimized by perceivers, such positive behaviors are likely to be attributed to less stable causes and, therefore, would not be expected to recur’’ (p. 443). Although all social-cognitive evaluations varied highly as a function of the affect children felt toward the peers, most of the variables exhibited significant variance also among individuals. Moreover, it was found that boys who had higher self-efficacy beliefs across the targets also reported more frequent proactive aggression across the relationships. The same association was found when we used same-sex peers as informants for proactive aggression. It is possible that self-efficacy functions as a trait-like characteristic, and boys with higher self-efficacy beliefs for aggression engage in more frequent proactive aggression across different interaction partners. Hence, although boys tend to be discriminative in their aggression, with higher frequency of acts targeted at the peers they dislike, there might be a subgroup of boys who aggress toward the disliked peers the most but from time to time target their aggression also toward other peers. It might be that certain socialcognitive evaluations, especially the ones that are more related to proactive aggression, are more actorand partner driven and less influenced by a specific dyadic relationship (Hubbard et al., 2001). It is also possible that girls who have higher self-efficacy beliefs for aggression use more relational/indirect forms of aggression. Additional gender differences were also found. Not surprisingly, girls had lower levels of relational outcome expectations and self-efficacy beliefs for aggression across the targets, which provides support for the finding that girls are more relationship oriented and might try to avoid conflicts in general (Burgess et al., 2006; Rose & Rudolph, 2006). With

183

regard to aggression and victimization, boys reported higher levels of aggression and victimization across the relationship types. However, there were no gender differences according to peer-nominated behaviors. It might be that as (overt) aggression is more common and accepted among boys, girls are less likely to admit behaving aggressively, especially, when the target is explicitly identified. Participants in the current study were in their early adolescence when the views of self and others start to be more differentiated (Harter, Waters, & Whitesell, 1998, p. 756). For instance, in the study by McDowell, Parke, and Spitzer (2002), kindergarten children’s goals and strategies did not vary across different social contexts (mother – child, father – child, peer – child), with the opposite being true for parental social-cognitive processing. Thus, children at an earlier developmental stage might be less sensitive to the context effects. Furthermore, in order for the relationship-specific representations to develop, children need time to interact with their peers. In the present sample, participants had been in the same classroom with most of their classmates for 5 years. It could be expected that differentiated social-cognitive evaluations are more stable when the classroom structure remains the same over time, especially if the evaluations are supported by the peer group. What happens when children move to a setting that requires formation of new relationships (e.g., transferring from one school to another)? It is well documented that past experiences within the family and among the peers affect how others are viewed (e.g., Crick & Dodge, 1994; Dodge, 1993). These general representations are likely to be activated when children meet unfamiliar peers and start to form new relationships. For instance, if a child has very many antipathetic relationships and is rejected by many peers, the child might view a new peer as hostile, which might decrease the likelihood of them becoming friends. However, it is still likely that an initial view of a peer changes with ongoing interactions. As Rabiner, Keane, and MacKinnon-Lewis (1993) stated, ‘‘It seems that although children may enter social situations predisposed to regard new peers in particular ways, these beliefs change according to the experiences they have with those children’’ (p. 241). Yet, longitudinal studies are needed to explore the factors that increase or decrease the likelihood of two unfamiliar kids becoming friends and non-friends over time. In summary, we were able to replicate and extend the results found in the study by Peets et al. (2007) using different methodology. More specifically,

184

Peets, Hodges, and Salmivalli

whereas they used individual interviews where relationship descriptions and vignettes were read to children, in the present study, we employed sociometric nominations to identify the targets and groupadministered questionnaires to assess social-cognitive evaluations. Moreover, target-specific behaviors were available in the current study. One of the shortcomings of this study was that our sample consisted of children who had either a reciprocal or unilateral relationship with the target. Although we do think that a negative versus a positive perception of the peer triggers different social-cognitive evaluations, social cognitions involved in reciprocal liking or disliking might be more intensive and acted out more frequently, especially when the child is aware of the partner’s reciprocated positive or negative affect. For instance, in the study by Card and Hodges (2007), children reported more victimization by mutual antipathies than by unilaterally disliked peers. In addition, the present findings provide evidence that the representation of a peer is activated by seeing the name of the peer in a vignette. However, in order to be able to make stronger claims about automatic activation of the target-specific representations, future studies would benefit from employing experimental designs using, for instance, different priming paradigms. Additionally, in the present study, social cognitions and behaviors were assessed only with regard to same-sex peers. If opposite-sex peer ratings/nominations were allowed, it could be expected that there would be a higher proportion of oppositesex peers identified as disliked targets and a higher proportion of same-sex peers identified as liked targets (e.g., Nangle et al., 2004). Future research could examine whether dislike felt toward same-sex peers results in different social-cognitive evaluations and behaviors compared to dislike felt toward opposite-sex peers. Finally, it would be interesting to know whether the affect felt toward the peer triggers affectcongruent behavioral responses or whether socialcognitive processes mediate this link. In conclusion, this study demonstrates that SIP (at least reflective information processing) is not a fixed phenomenon but varies with the relationship children have with different peers. Although the frequency of aggressive behaviors declines with age, social-cognitive processes supporting the affect felt toward someone are likely to be evident also in older adolescents and adults (see, e.g., Hymel, 1986). As our study was concurrent, future studies should test, for instance, whether stable dislike toward the peer is also accompanied by more stable ‘‘negative’’ social cognitions and aggressive behaviors compared to unstable dislike.

References Berndt, T. J., & Perry, T. B. (1986). Children’s perceptions of friendships as supportive relationships. Developmental Psychology, 22, 640 – 648. Burgess, K. B., Wojslawowicz, J. C., Rubin, K. H., Rose-Krasnor, L., & Booth-LaForce, C. (2006). Social information processing and coping strategies of shy/ withdrawn and aggressive children: Does friendship matter? Child Development, 77, 371 – 383. Cairns, R. B., Cairns, B. D., Neckerman, H. J., Gest, S. D., & Garie´py, J.-L. (1988). Social networks and aggressive behavior: Peer support or peer rejection? Developmental Psychology, 24, 815 – 823. Card, N. A., & Hodges, E. V. E. (2006). Shared targets for aggression by early adolescent friends. Developmental Psychology, 42, 1327 – 1338. Card, N. A., & Hodges, E. V. E. (2007). Victimization within mutually antipathetic relationships. Social Development, 16, 479 – 496. Card, N. A., Isaacs, J., & Hodges, E. V. E. (2000, March). Dynamics of interpersonal aggression in the school context: Who aggresses against whom? In A. Nishina & J. Juvonen (Chairs), Harassment across diverse contexts. Poster symposium conducted at the 8th biennial meeting of the Society for Research on Adolescence, Chicago. Coie, J. D., Antonius, H. N., Dodge, K. A., Hubbard, J. A., Schwartz, D., Lemerise, E. A., et al. (1999). It takes two to fight: A test of relational factors and a method for assessing aggressive dyads. Developmental Psychology, 35, 1179 – 1188. 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. Denham, S. A., & Holt, R. W. (1993). Preschoolers’ likeability as cause or consequence of their social behavior. Developmental Psychology, 29, 271 – 275. Dodge, K. A. (1980). Social cognition and children’s aggressive behavior. Child Development, 51, 162 – 170. Dodge, K. A. (1993). Social-cognitive mechanisms in the development of conduct disorder and depression. Annual Review of Psychology, 44, 559 – 584. Dodge, K. A., & Coie, J. D. (1987). Social information processing factors in reactive and proactive aggression in children’s peer groups. Journal of Personality and Social Psychology, 53, 1146 – 1158. Dodge, K. A., Murphy, R. R., & Buchsbaum, K. (1984). The assessment of intention-cue detection skills in children: Implications for developmental psychopathology. Child Development, 55, 163 – 173. Dodge, K. A., Price, J. M., Coie, J. D., & Christopoulos, C. (1990). On the development of aggressive dyadic relationships in boys’ peer groups. Human Development, 33, 260 – 270. Dodge, K. A., & Somberg, D. R. (1987). Hostile attributional biases among aggressive boys are exacerbated under conditions of threats to the self. Child Development, 58, 213 – 224.

Social-Cognitive Evaluations and Behaviors Erdley, C. A., & Asher, S. R. (1998). Linkages between children’s beliefs about the legitimacy of aggression and their behavior. Social Development, 7, 321 – 339. Guerin, B. (1999). Children’s intergroup attribution bias for liked and disliked peers. Journal of Social Psychology, 139, 583 – 589. Harter, S., Waters, P., & Whitesell, N. R. (1998). Relational self-worth: Differences in perceived worth as a person across interpersonal contexts among adolescents. Child Development, 69, 756 – 766. Hawley, P. H., Little, T. D., & Pasupathi, M. (2002). Winning friends and influencing peers: Strategies of peer influence in late childhood. International Journal of Behavioral Development, 26, 466 – 474. Hayes, D. S., Gershman, E. S., & Halteman, W. (1996). Enmity in males at four developmental levels: Cognitive bases for disliking peers. Journal of Genetic Psychology, 157, 153 – 160. Hubbard, J. A., Dodge, K. A., Cillessen, A. H. N., Coie, J. D., & Schwartz, D. (2001). The dyadic nature of social information processing in boys’ reactive and proactive aggression. Journal of Personality and Social Psychology, 80, 268 – 280. Huesmann, L. R., & Guerra, N. G. (1997). Children’s normative beliefs about aggression and aggressive behavior. Journal of Personality and Social Psychology, 72, 408 – 419. Hughes, J. N., Meehan, B. T., & Cavell, T. A. (2004). Development and validation of a gender-balanced measure of aggression-relevant social cognition. Journal of Clinical Child and Adolescent Psychology, 33, 292 – 302. Hymel, S. (1986). Interpretation of peer behavior: Affective bias in childhood and adolescence. Child Development, 57, 431 – 445. Lemerise, E. A., & Arsenio, W. F. (2000). An integrated model of emotional processes and cognition in social information processing. Child Development, 71, 107 – 118. McDowell, D. J., Parke, R. D., & Spitzer, S. (2002). Parent and child cognitive representations of social situations and children’s social competence. Social Development, 11, 469 – 486. Muthe´n, L. K., & Muthe´n, B. O. (1998 – 2004). Mplus user’s guide (Version 3.0) [Computer software and manual]. Los Angeles: Muthe´n & Muthe´n. Nangle, D. W., Erdley, C. A., Zeff, K. R., Stanchfield, L. L., & Gold, J. A. (2004). Opposites do not attract: Social status and behavioral-style concordances and discordances among children and the peers who like or dislike them. Journal of Abnormal Child Psychology, 32, 425 – 434. Newcomb, A. F., & Bagwell, C. L. (1995). Children’s friendship relations: A meta-analytic review. Psychological Bulletin, 117, 306 – 347. Orobio de Castro, B., Veerman, J. W., Koops, W., Bosch, J. D., & Monshouwer, H. J. (2002). Hostile attribution of

185

intent and aggressive behavior: A meta-analysis. Child Development, 73, 916 – 934. Peets, K., Hodges, E. V. E., Kikas, E., & Salmivalli, C. (2007). Hostile attributions and behavioral strategies in children: Does relationship type matter? Developmental Psychology, 43, 889 – 900. Perry, D. G., Kusel, S. J., & Perry, L. C. (1988). Victims of peer aggression. Developmental Psychology, 24, 807 – 814. Perry, D. G., Perry, L. C., & Rasmussen, P. (1986). Cognitive social learning mediators of aggression. Child Development, 57, 700 – 711. Perry, D. G., Perry, L. C., & Weiss, R. J. (1989). Sex differences in the consequences that children anticipate for aggression. Developmental Psychology, 25, 312 – 319. Poulin, F., & Boivin, M. (2000a). The formation and development of friendship in childhood: The role of proactive and reactive aggression. Developmental Psychology, 36, 233 – 240. Poulin, F., & Boivin, M. (2000b). Proactive and reactive aggression: Evidence of a two-factor model. Psychological Assessment, 12, 115 – 122. Rabiner, D. L., Keane, S. P., & MacKinnon-Lewis, C. (1993). Children’s beliefs about familiar and unfamiliar peers in relation to their sociometric status. Developmental Psychology, 29, 236 – 243. Reis, H. T., Collins, W. A., & Berscheid, E. (2000). The relationship context of human behavior and development. Psychological Bulletin, 126, 844 – 872. Rose, A. J., & Rudolph, K. D. (2006). A review of sex differences in peer relationship processes: Potential trade-offs for the emotional and behavioral development of girls and boys. Psychological Bulletin, 132, 98 – 131. Schwartz, D., Dodge, K. A., Coie, J. D., Hubbard, J. A., Cillessen, A. H. N., Lemerise, E. A., et al. (1998). Socialcognitive and behavioral correlates of aggression and victimization in boys’ play groups. Journal of Abnormal Child Psychology, 26, 431 – 440. Simpkins, S. D., & Parke, R. D. (2002). Do friends and nonfriends behave differently? A social relations analysis of children’s behavior. Merrill-Palmer Quarterly, 48, 263 – 283. Waldman, I. D. (1996). Aggressive boys’ hostile perceptual and response biases: The role of attention and impulsivity. Child Development, 67, 1015 – 1033. Werner, N. E., & Nixon, C. L. (2005). Normative beliefs and relational aggression: An investigation of the cognitive bases of adolescent aggressive behavior. Journal of Youth and Adolescence, 34, 229 – 243. Yoon, J. S., Hughes, J. N., Cavell, T. A., & Thompson, B. (2000). Social cognitive differences between aggressiverejected and aggressive-nonrejected children. Journal of School Psychology, 38, 551 – 570.