Associations Between Verbal Reasoning, Normative Beliefs About Aggression, and Different Forms of Aggression
Abstract
The purpose of the present study was to examine the impact of sex, verbal reasoning, and normative beliefs on direct and indirect forms of aggression. Three scales from the Peer Estimated Conflict Behavior Questionnaire, Verbal Reasoning tests, and an extended version of Normative Beliefs About Aggression Scale were administered to 663 Estonian students (289 boys and 374 girls; 150 fifth, 264 seventh, and 249 ninth graders; ages 11–16). Self- and same-sex peer ratings were used to assess the frequency of aggression. Associations between study variables were examined using structural equation modeling. Results showed that boys had higher levels of physical and verbal aggression in all the grades and higher levels of indirect aggression in Grade 7. Verbal reasoning predicted negatively all the forms of aggression, except for indirect aggression in Grade 7. Normative beliefs had a positive effect on all three forms of aggression in Grade 7 and Grade 9.
Several studies have shown that aggression is widely used during middle childhood and adolescence. As a rule, boys are more aggressive than girls, at least in terms of direct forms of aggression (i.e., physical and verbal aggression). When indirect (relational) forms of aggression are taken into account, aggression of girls becomes more visible. The results, however, vary depending on who provides information (e.g., self, peers, teacher, parents), how old the participants are, and in which cultural context the study is conducted (for reviews see Archer, 2004; Archer & Coyne, 2005; Vaillancourt, 2005). Moreover, a number of studies have examined different factors that could explain individual differences in aggression. For instance, beliefs children hold about acceptability of aggression (Huesmann & Guerra, 1997) and children's verbal abilities (e.g., Giancola, 1994) have been found to be associated with aggression. However, there is a lack of studies conducted elsewhere than in the United States. In addition, there is a need for research examining the effects of verbal reasoning and normative beliefs simultaneously. Thus, the goals of the present study were to (1) investigate age and sex differences in direct and indirect aggression among adolescents in Estonia; (2) analyze the relations among different forms of aggression, verbal reasoning, and normative beliefs; and (3) determine whether these relations are moderated by grade level.
Regarding sex differences, boys use direct aggression more frequently and hold higher aggression-approval beliefs than girls (Archer, 2004; Huesmann & Guerra, 1997; Zelli, Dodge, Lochman, & Laird, 1999). In contrast, indirect aggression seems to be more common among girls; however, while some studies have shown that girls use indirect aggression more frequently than boys, other studies have either not found any sex differences or have shown that boys use both types of aggression more frequently (Archer, 2004).
As language is an important means by which one can control behavior and emotions (Cole, 2001; Vygotsky, 1934/1997), an inefficient use of language might make it more difficult to foresee the consequences of one's actions. Individuals with higher levels of abstract verbal reasoning are able to think of behavioral alternatives, consider long-term consequences of different behaviors, and, as a result, select the more appropriate way of handling the conflicts (cf. Crick & Dodge, 1994). It means that children with better verbal reasoning skills may either act nonaggressively or choose a more covert/indirect form of aggression. Positive relations between indirect aggression and verbal skills could be expected as being an effective manipulator of relationships needs cognitive skills such as planning, analyzing the consequences, and comparing the alternatives (e.g., in order to avoid being caught). As verbal reasoning and social skills develop with age, direct forms of aggression are expected to be gradually replaced by indirect forms of aggression (e.g., Björkqvist, Österman, & Kaukiainen, 1992). However, there is some evidence that indirect aggression can also be associated with impulsiveness rather than with the behavior that needs skillful planning (see also Zalecki & Hinshaw, 2004).
Several studies have demonstrated associations between direct aggression and language deficits in various age groups (Arnold, 1997; Cole, 2001; Estrem, 2005; Giancola, 1994; Seguin, Arseneault, Boulerice, Harden, & Tremblay, 2002; Stanford, Greve, Boudreaux, Mathias, & Brumbelow, 1996). Aggressive children have problems with concept formation and use, understanding of concepts, and complex language structures (e.g., Cole, 2001; Olvera, Semrud-Clikeman, Pliszka, & O'Donnell, 2005). While findings about direct aggression and verbal skills are quite concordant across studies, heterogeneous results have emerged for indirect aggression—results vary with tasks and samples. Whereas some studies have shown negative correlations between indirect aggression and language skills (Estrem, 2005), other studies have demonstrated positive relations between the two constructs (e.g., Bonica, Arnold, Fisher, Zeljo, & Yershova, 2003).
Furthermore, beliefs about the acceptability of aggressive behavior influence whether children and adolescents resort to aggressive means. The more acceptable it is, the higher the likelihood of behaving aggressively. These beliefs are cognitive representations of past experiences which influence processing new stimuli and responses. When children get older, these representations become more stable and thus more resistant to environmental changes (Guerra, Huesmann, & Spindler, 2003; Huesmann & Guerra, 1997). Previous studies have shown that normative beliefs about aggression are associated with direct forms of aggression (Huesmann & Guerra, 1997; Werner & Nixon, 2005). Moreover, beliefs about acceptability of physical aggression are related to physical aggression, and beliefs about relational/indirect aggression are associated with relational/indirect aggression (Werner & Nixon, 2005).
The current study was carried out in Estonia. Estonia was incorporated into the Soviet Union for more than 50 years and regained its independence in 1991. During the last decades—marked by great democratic changes in the society— educational reforms have been implemented in schools, with emphasis on student-centered teaching methods, flexible curricula, and greater availability of choices for students. However, although there have been advances in economics, there has been a rise in social inequality and insecurity which might also be reflected in students' attitudes and behaviors. For instance, it has been shown that Estonian adolescents are more aggressive and exhibit lower levels of social responsibility as compared with Finnish adolescents (Keltikangas-Järvinen & Terav, 1996).
The first purpose of the present study was to investigate age and sex differences in the frequency of physical, verbal, and indirect aggression in fifth, seventh, and ninth grade students. Previous studies have shown that aggression is frequently used as a common solution for handling conflicts among middle school students in Estonia (Kõiv, 2000; Peets & Kikas, 2006). According to the developmental theory of aggression (e.g., Björkqvist et al., 1992), we expected that older children would use direct forms of aggression less frequently and indirect forms more frequently. In addition, boys were anticipated to show higher levels of physical and verbal aggression (Archer, 2004). Considering the developmental trends (Björkqvist et al., 1992), we hypothesized that sex differences in indirect aggression would be dependent on the grade level, with older girls showing similar levels or even higher levels of indirect aggression than boys.
Second, we were interested in examining associations between different forms of aggression, verbal reasoning (deduction skills and an ability to form abstract relations among concepts), and normative beliefs, and in determining whether these relations would be moderated by grade. In general, attitudes about violence and punishment have been found to be more positive in the United States than in the European countries (McAlister et al., 2001). The same study showed that the attitudes of Estonian adolescents regarding using violence and physical punishment tended to be rather positive as compared with adolescents in other European countries. With regard to the current study, we expected positive links between approval of aggression and all three forms of aggression (Huesmann & Guerra, 1997; Werner & Nixon, 2005). In contrast, verbal reasoning was hypothesized to have a negative effect on direct forms of aggression (i.e., physical and verbal aggression) (Arnold, 1997; Cole, 2001). As there are mixed findings concerning the associations between verbal reasoning and indirect aggression, we expected that the link between the two constructs would be more modest. In addition to the independent effects, we examined whether verbal reasoning would moderate the link between normative beliefs and aggression. We also explored whether the associations would vary depending on the grade level.
In line with several researchers (Achenbach, McConaughy, & Howell, 1987; Peets & Kikas, 2006; Rushton, Brainerd, & Pressley, 1983) who have stressed the importance of using different informants as it enables obtaining a less biased view of the phenomenon, we used a multi-informant approach. For instance, Ladd and Kochenderfer-Ladd (2002) found that the combination of different informants for victimization (self, peers, teachers, parents) provided a better estimate of children's relational adjustment than any informant alone. In the current study, we measured aggression using self and same-sex peers as informants. Although children and adolescents often underestimate their aggression, self-reports can capture incidents which are not observable and known to other informants. In addition, the more normative the aggressive behavior is, the more likely it is that children admit resorting to aggressive means. Moreover, although peer-reported aggression is often considered a more “objective” measure, it should be kept in mind that biases can also operate at the classroom/peer group level (e.g., the child might become less aggressive over time but peers still identify him or her as aggressive due to the negative affect they feel toward that peer).
The associations were evaluated using structural equation modeling, which allowed us to take into account measurement error. The theoretical model is depicted in Figure 1. It should be pointed out that as our data were cross-sectional, the selection of variables treated as exogenous and endogenous was theoretical. Thus, the direction of arrows in the model should be interpreted with caution and tested again using a longitudinal design.
The theoretical model.
METHOD
Participants and Procedure
The sample consisted of 663 Estonian students (289 boys and 374 girls; 150 fifth, 264 seventh, and 249 ninth graders; ages 11–12, 13–14, and 15–16) from eight different municipal schools located in a middle-class neighborhood. Participating students were from 29 different classes with an average class size of 28 students (minimum 16, maximum 37, SD 6.14).
Before testing, informed consent letters were sent to parents via students. All children could participate, unless we received a clear disagreement from the parent in the form of the signature or the child himself/herself refused to take part in the study. Students filled out the questionnaires during regular school hours. At the beginning of the session, participants received a short overview of the research and questionnaires. The entire testing session lasted approximately 35 min.
Measures
Verbal reasoning. The test consisted of three parts. The first two subtests were adapted from Toomela (2003). First, students had to define six concepts. Second, students were given six pairs of words. Their task was to describe the most important similarity between the concepts. Word pairs referred to the items from the same category (dog–cat). All 12 answers were coded into lower-level concepts (coded as 0) and higher-level concepts (coded as 1). An answer was coded as a higher-level concept when (a) the relationship between words was defined hierarchically (e.g., “hammer and ax are similar because they are both tools”) or (b) the word was related to a higher-level concept in the hierarchy (e.g., “hospital is a medical institution”). The third subtest, logical reasoning, consisted of six syllogisms, using universal and particular quantifiers (see Scribner, 1997). Students had to assess whether the given conclusion was valid or not. The verbal reasoning score was calculated by summing all “higher-level” answers in the first two subtests and logically correct answers in the last subtest. The internal consistency of the Verbal Reasoning test was good (Cronbach α=.82).
Normative beliefs about aggression. Children were administered an extended version of Normative Beliefs About Aggression Scale (Huesmann & Guerra, 1997). In addition to the original version, we included items about indirect aggression. Sixteen items describe possible responses (physical, verbal, or indirect) to provocation (e.g., “Suppose a boy says something bad to another boy, Tom. Do you think it's OK for Tom to yell at him?”). These items vary on severity of provocation, severity of response, sex of provocateur, and sex of responder. Twelve questions tap the general beliefs about aggression (e.g., “It's generally OK to gossip and say bad things about others”). Respondents had to assess the degree of acceptability of each behavior on a 4-point Likert scale. Normative beliefs score was calculated as a mean across specific and general items. The internal consistency of the scale was good (α=.88).
Aggression. The participants were administered three scales from the Peer Estimated Conflict Behavior Questionnaire, which is a shortened version of the Direct and Indirect Aggression Scales Questionnaire (Björkqvist & Österman, 1998). Each form of aggression was measured with a short question describing the relevant behaviors (e.g., “Who is verbally aggressive, that is, who yells, insults, calls names, or teases others?” for verbal aggression). The respondents were asked to estimate on all three scales how frequently each classmate acted in the described way when he/she got angry with or had problems with another student in the class. Students provided their ratings on a 5-point scale (0=never, 1=seldom, 2=sometimes, 3=quite often, 4=very often) by checking the appropriate choice behind each classmate's name.
Participants were given the test sheets where all the names of students in a class were written beforehand. In order to facilitate the whole scoring procedure, the names of girls and boys were separated from each other, with the names of boys starting the list. In addition to self-ratings, we used same-sex ratings which were summed for each form of aggression (child's own rating was excluded) and divided by the number of estimators.
Data Analyses
The program Mplus 3.0 was used (Muthén & Muthén, 1998–2004). Throughout the analyses a maximum-likelihood (ML) estimator was used (Muthén & Muthén, 1998–2004). The fit of the models was evaluated using the following indices: chi-square test, comparative fit index (CFI), and root mean error of approximation (RMSEA). The cutoff criteria for the fit indices were CFI>.95 and RMSEA<.06 (Hu & Bentler, 1999). As the chi-square test is sensitive to sample size and deviations from underlying assumptions (e.g., multivariate normality), other fit indices are considered more adequate to assess the model fit (Hu & Bentler, 1998; Kline, 2005).
RESULTS AND DISCUSSION
Structural equation models were built to test the hypothesized associations among sex, verbal reasoning, normative beliefs, and physical, verbal, and indirect aggression (see Figure 1). First, a measurement model with latent constructs of physical, verbal, and indirect aggression was specified. Each factor had two manifest variables, self-estimated and peer-evaluated aggression. The first unstandardized loading was fixed to one, whereas the second loading was always freely estimated. Covariances between the latent factors were also estimated. The model did not fit the data well, χ2(6)=426.37, p<.001, CFI=.83, RMSEA=.33. Next, modification indices and values in the residual correlation matrix suggested correlating the error terms of self-estimated forms of aggression. The fit of the model improved significantly, χ2(3)=3.13, p=.37, CFI=1.00, RMSEA=.01. To test whether the unstandardized loadings were invariant across the grades, we compared the model where the second loading was unconstrained to the model where the loading was constrained to be equal across the three grades. Chi-square difference test showed loading invariance across the grades (χ2 difference=18.83 − 12.22=6.61, df=15 − 9=6, p=ns). The unstandardized loadings of self-estimated physical, verbal, and indirect aggression were .73, .64, and .47, respectively.
Next, a structural equation model was built where three latent factors were regressed on predictors, that is, on sex, normative beliefs, and verbal reasoning. First, the general model was tested. All the regression paths were statistically significant (see Table 1). The fit of the model was χ2(12)=34.97, p<.001, CFI=.99, RMSEA=.05. Then, two multigroup analyses were conducted with all the structural paths (1) constrained to be equal across grades and (2) freely estimated. The model with freely estimated paths improved the model fit significantly (χ2 difference=121.94 − 79.59=42.35, df=60 − 42=18, p<.001). The analyses relieved that some of the structural paths were statistically nonsignificant, and were thus fixed to zero for a respective grade. The final fit of the model was χ2(48)=83.76, p=.001, CFI=.99, RMSEA=.06. Parameters for the final model are also presented in Table 1.
| Sex | Verbal Ability | NormativeBeliefs | VerbalAggression | IndirectAggression | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| B | SE | β | B | SE | β | B | SE | β | B | SE | β | B | SE | β | |
| General model | |||||||||||||||
| Physical | −.84 | .05 | −1.16 | −.05 | .01 | −.07 | .24 | .07 | .34 | .27 | .02 | .37 | .19 | .01 | .54 |
| Verbal | −.53 | .05 | .76 | −.05 | .01 | −.07 | .33 | .06 | .47 | .26 | .02 | .74 | |||
| Indirect | −.10 | .04 | −.20 | −.03 | .01 | −.06 | .18 | .05 | .37 | ||||||
| Multigroup model | |||||||||||||||
| Grade 5 | |||||||||||||||
| Physical | −.81 | .07 | −.56 | −.05 | .02 | −.23 | .00 | .00 | — | .31 | .04 | .67 | .25 | .04 | .73 |
| Verbal | −.52 | .06 | −.40 | −.04 | .02 | −.19 | .00 | .00 | — | .32 | .04 | 1.04 | |||
| Indirect | .00 | .00 | — | −.03 | .01 | −.22 | .00 | .00 | — | ||||||
| Grade 7 | |||||||||||||||
| Physical | −.93 | .09 | −.59 | −.03 | .01 | −.10 | .37 | .10 | .19 | .33 | .04 | .56 | .22 | .03 | .52 |
| Verbal | −.51 | .09 | −.35 | −.03 | .01 | −.12 | .45 | .11 | .24 | .28 | .03 | .72 | |||
| Indirect | −.15 | .07 | −.14 | .00 | .00 | — | .27 | .08 | .20 | ||||||
| Grade 9 | |||||||||||||||
| Physical | −.70 | .05 | −.62 | −.03 | .01 | −.19 | .26 | .08 | .19 | .16 | .02 | .47 | .11 | .02 | .47 |
| Verbal | −.48 | .05 | −.39 | −.04 | .01 | −.23 | .40 | .09 | .26 | .17 | .02 | .68 | |||
| Indirect | .00 | .00 | — | −.03 | .01 | −.21 | .27 | .08 | .26 | ||||||
- Note. Covariances and correlations are given in italics.
We also tested whether verbal reasoning moderated the association between normative beliefs and aggression by regressing three latent aggression variables on predictors and the product of normative beliefs and verbal reasoning (normative beliefs and verbal reasoning variables were standardized before). Results showed that verbal reasoning and normative beliefs interacted on influencing indirect aggression (p<.05). Followup analyses were conducted following the procedure suggested by Aiken and West (1991). Association between normative beliefs and indirect aggression was estimated at low (−1 SD), medium (0 SD), and high (+1 SD) levels of the moderator (verbal reasoning). Sex was always controlled in these models. Findings demonstrated that when verbal reasoning skills were above the average, normative beliefs were most strongly linked to indirect aggression (B=.14, SE=.03, β=.36***). When verbal skills were at average, there was still a significant (albeit weaker) prediction (B=.09, SE=.02, β=.24***). The link between beliefs and indirect aggression became nonsignificant when children had low levels of verbal reasoning (B=.04, SE=.03, β=.11, ns). We also conducted multigroup analyses where interaction effects were tested separately for each grade. None of the effects was significant.
Physical and Verbal Aggression
Associations between our predictors and direct forms of aggression were similar across all grade levels and mostly concordant with earlier findings. Boys were physically and verbally more aggressive than girls, with the difference being larger for physical aggression (cf. Archer, 2004). In addition, children with higher verbal skills had lower levels of physical and verbal aggression (cf. Cole, 2001; Seguin et al., 2002). Normative beliefs had a positive effect on direct aggression in the whole sample, and in Grade 7 and Grade 9. Although Huesmann and Guerra (1997) found that beliefs regulate aggressive behavior already in elementary school, in our study, beliefs were not associated with aggression in Grade 5. In Estonia, the transition from elementary to middle school takes place in Grade 5. It might be that beliefs tend to change during that period. Usually, students study in one building from Grade 1 until Grade 9 or even to Grade 12. However, it is common that classrooms for younger students are located in one area at school, whereas classrooms for 5th to 9th (sometimes up to 12th) graders are in a different area. It is likely that from Grade 5 onwards when children are spending more time with older students they have also more possibilities to witness a greater amount and variety of aggressive acts than in younger grades. Guerra et al. (2003) also showed that witnessing violence influenced normative beliefs in older children. It might be possible that aggressive behavior affects the development of social cognitions in younger children. Over time, when the environment supports the cognitions children hold, social cognitions become more stable and have a greater influence on subsequent behaviors.
Indirect Aggression
In the general model, the effects and relations were similar to those found for direct aggression. However, there were differences between the grades. While boys were indirectly more aggressive in Grade 7, sex differences were not found in Grade 5 or Grade 9. In addition, the effect in Grade 7 was not especially strong (β=1.14). Thus, our findings suggest that both boys and girls display an indirect form of aggression. Moreover, it is likely that the behavioral repertoire of most of the aggressive children includes direct as well as indirect aggression. However, it might be also possible that if we had used same- and opposite-sex ratings, sex differences in indirect aggression would have been more evident.
Furthermore, while verbal reasoning had a negative effect on indirect aggression in Grade 5 and in Grade 9, a nonsignificant association was revealed in Grade 7. Thus, in contrast to the study by Kaukiainen et al. (1999), we did not find a positive linkage between verbal reasoning and indirect aggression. Still, it should be pointed out that the negative effect of verbal reasoning was the highest for physical and the lowest for indirect aggression. Whereas Kaukiainen and colleagues operationalized social intelligence as perceived performance competencies in a social context, in the present study, we assessed cognitive skills. And, although a certain level of cognitive ability is needed in order to be a skillful manipulator, interpersonal skills rather than mere cognitive abilities might be more proximal correlates of indirect aggression. Moreover, the strength of associations between normative beliefs and indirect aggression varied with the level of reasoning skills. When children were above the average regarding their verbal abilities, the link was the strongest. In contrast, normative beliefs did not predict indirect aggression for children who were below the average with regard to their verbal skills. Indirect aggression may be used to obtain certain interpersonal goals (e.g., high status, friends); however, these goals may also be reached with social and more acceptable ways of behavior. Children with higher level of reasoning skills are better capable of thinking about different ways of acting and comparing their consequences. When these children have high aggression-approval beliefs, they may choose to use indirect aggression as a hidden way to achieve their goals, while children with low aggression-approval beliefs may search for alternative ways of behaving.
Differences between grades might be related to specific peer groups. Namely, relational manipulation is the most hurtful inside the circle of close friends. Some researchers have stressed that indirect aggression is more characteristic of girls just because of the intimate nature of their friendships during adolescence (e.g., Owens, 1996). However, McNelles and Connolly (1999) showed that girls and boys used different activities to achieve intimacy. Boys achieve affective experience through shared activities, which means that exclusion from the peer group can be as hurtful for boys as it is for girls. Accordingly, it is possible that some boys may view indirect form of aggression as an effective way of hurting others.
Conclusions and Limitations
The findings of the present study show that normative beliefs and verbal reasoning have independent effects on direct and indirect aggression, and some of those effects differ by grade. These results refer to the possibility that normative beliefs are more unstable during the transition from elementary school to middle school when children face new environmental challenges (e.g., new teachers, new classrooms, and recess areas). The associations with indirect aggression were very weak (negative) or nonsignificant. Examining classroom norms might provide additional information about how individual beliefs are actualized into different forms of aggression. In conclusion, it should be stressed that as our study was cross-sectional, we could only hypothesize about the direction of the effects. Future longitudinal studies are needed to determine the developmental pathways for different forms of aggression and to examine their associations with verbal abilities and normative beliefs.
ACKNOWLEDGMENT
Research for this paper was supported by a grant from the Estonian Science Foundation (grant no. 5371).
