attributions for social failure and adolescent aggression

14
AGGRESSIVE BEHAVIOR Volume 19, pages 421-434 (1993) Attributions for Social Failure and Adolescent Aggression Nancy G. Guerra, L. Rowell Huesmann, and Arnaldo Zelli Department of Psychology, University of Illinois at Chicago, Chicago, Illinois ........................................................ ........................................................ In the present study, 119 high school boys and 79 institutionalized delinquent boys of the same age range were assessed on their own aggressive behavior and on their tenden- cies to attribute social failure to controllable, external, stable causes, anticipate a hos- tile affective response, and endorse aggressivebehavioral responses to hypothetical social situations. While the two populations of boys did not differ detectably in their attribu- tional tendencies, the relations between an individual’s aggressiveness and an individu- al’s attributions differed considerably across the two populations. In particular, among delinquent but not among nondelinquent boys, the tendency to attribute one’s social failures to stable and controllable causes predicted stronger hostile emotional responses to failure and a tendency to endorse physically aggressive responses following such fail- ure. These hostile emotional responses to failure and this preference for a physically aggressive response, in turn, predicted greater actual aggression within the population of delinquent boys. Neither of these links could be demonstrated for nondelinquent boys. However, in the nondelinquent sample, attributing social failure to external and con- trollable causes predicted endorsement of aggressive responses only indirectly through increased hostile affect. It was concluded that the specific relations between cognitive and affective responses to social failure may be a contributing factor to the serious physical aggression displayed by some delinquents and to the less serious aggression of nondelinquents. 0 1993 Wiley-Liss, Inc. ........................................................ ........................................................ Key words: social failure, aggression, adolescent, delinquents, hostile emotional response INTRODUCTION There is a growing body of evidence which has demonstrated that aggression is a relatively stable behavior among children and youth (Huesmann et al., 1984; Olweus, 1984). To the extent that this stability is due to early learning experiences, it has been hypothesized that, with development, patterns of aggression are increasingly governed by specific cognitive processes [Eron, 1987; Guerra and Slaby, 1990; Huesmann, 1988; Received for publication October 2, 1992; accepted June 1, 1993 Address reprint requests to Nancy G. Guerra, Department of Psychology, Box 4348, MIC 285, University of Illinois at Chicago, Chicago, IL 60680. 0 1993 Wiley-Liss, Inc.

Upload: nancy-g-guerra

Post on 06-Jun-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

AGGRESSIVE BEHAVIOR Volume 19, pages 421-434 (1993)

Attributions for Social Failure and Adolescent Aggression Nancy G. Guerra, L. Rowell Huesmann, and Arnaldo Zelli

Department of Psychology, University of Illinois at Chicago, Chicago, Illinois

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . In the present study, 119 high school boys and 79 institutionalized delinquent boys of the same age range were assessed on their own aggressive behavior and on their tenden- cies to attribute social failure to controllable, external, stable causes, anticipate a hos- tile affective response, and endorse aggressive behavioral responses to hypothetical social situations. While the two populations of boys did not differ detectably in their attribu- tional tendencies, the relations between an individual’s aggressiveness and an individu- al’s attributions differed considerably across the two populations. In particular, among delinquent but not among nondelinquent boys, the tendency to attribute one’s social failures to stable and controllable causes predicted stronger hostile emotional responses to failure and a tendency to endorse physically aggressive responses following such fail- ure. These hostile emotional responses to failure and this preference for a physically aggressive response, in turn, predicted greater actual aggression within the population of delinquent boys. Neither of these links could be demonstrated for nondelinquent boys. However, in the nondelinquent sample, attributing social failure to external and con- trollable causes predicted endorsement of aggressive responses only indirectly through increased hostile affect. It was concluded that the specific relations between cognitive and affective responses to social failure may be a contributing factor to the serious physical aggression displayed by some delinquents and to the less serious aggression of nondelinquents. 0 1993 Wiley-Liss, Inc.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Key words: social failure, aggression, adolescent, delinquents, hostile emotional response

INTRODUCTION

There is a growing body of evidence which has demonstrated that aggression is a relatively stable behavior among children and youth (Huesmann et al., 1984; Olweus, 1984). To the extent that this stability is due to early learning experiences, it has been hypothesized that, with development, patterns of aggression are increasingly governed by specific cognitive processes [Eron, 1987; Guerra and Slaby, 1990; Huesmann, 1988;

Received for publication October 2, 1992; accepted June 1 , 1993

Address reprint requests to Nancy G. Guerra, Department of Psychology, Box 4348, MIC 285, University of Illinois at Chicago, Chicago, IL 60680.

0 1993 Wiley-Liss, Inc.

422 Guerra et al.

Huesmann and Eron, 1984; Slaby and Guerra, 19881. That is, what children are learn- ing is a characteristic way of thinking which governs social behavior and becomes re- markably resistant to change [Huesmann, 19881.

Many studies exploring the link between cognition and aggression have demonstrated a relation between aggressive behavior and deficits in children’s quantitative social problem-solving skills such as generation of alternative solutions [Richard and Dodge, 1982; Shure and Spivack, 1976; Slaby and Guerra, 19881, or between aggressive be- havior and cognitions such as aggressive daydreams and beliefs in a violent world [Huesmann et al., 1984a; Huesmann et al., 1984bl. Recent elaborations within this social problem-solving framework have sought to extend the range of cognitive vari- ables studied. It has been recognized that a child’s ability to generate and carry out socially competent responses depends upon a range of cognitive factors affecting the way the child understands and interprets social events. Attributional processes have been introduced into this framework by Dodge and his colleagues, who have identified a hostile attribution bias which characterizes the way that aggressive children interpret ambiguous intent cues [Dodge, 1980, 1986; Dodge and Frame, 19821. However, while the relation between a child’s intent attributions about someone else’s behavior under ambiguous conditions and the child’s aggression has received much attention, rela- tively little attention has been paid to the child’s causal attributions about the negative outcomes the child experiences in interactions with others. Yet on other dimensions of psychopathology, e .g . , depression, such causal attributions have been invoked as a pri- mary cause of the psychopathology [Abramson et al., 1978; Seligman, 19751.

Although little research exists connecting dimensions of causal attributions for nega- tive social outcomes to aggressive behavior, the evidence has mounted that aggressive children experience a large number of negative social outcomes and poor peer relations [Dodge et al., 1990; Huesmann et al., 1984b; Kupersmidt and Coie, 19901. While it is probable that aggressive children experience social failure to some extent simply be- cause they are aggressive, it is also possible that their attributions regarding the causes of their failures when interacting with others contribute to their aggressive behavior, just as causal attributions influence behavior in other social domains, including help- ing behavior [Meyer and Mulherin, 19801 and reaction to social stigmas [Weiner et al., 19881.

Although earlier studies of the attribution/behavior relation typically linked charac- teristic attributional styles with specific behaviors, more recent studies have focused on the role of affect in this relation. Specifically, attribution theorists have proposed that perceived causality of an outcome gives rise to a variety of affective reactions, and influences the perceived expectancy of goal attainment, which then influences subse- quent behavior [Weiner, 19861. In fact, several recent studies have reported that while attributions have both a direct and indirect (through affect) relation to behavior, the indirect effects are often considerably larger [Meyer and Mulherin, 1980; Weiner, 19801. Thus, attributions of causality are seen as guiding behavior primarily (but not exclu- sively) by guiding affect which, in turn, directs behavior.

How might attributions about the causes of social failure influence affective reac- tions and increase aggressive responding? Clearly, attributions which stimulate anger and hostility should lead to more aggressive behavior. Weiner [1982, pp. 190-1911 proposed that the attribution of “a negative, self-related outcome or event to factors controllable by others” should stimulate anger, particularly when the cause is also per-

Attributions and Aggression 423

ceived as stable. However, very little research has been undertaken to test this assertion regarding negative self-related outcomes. Rather, most studies have focused on how people’s attributions about another person’s social outcomes influence their own affec- tive reactions and responses to the person.

In this regard, variations in attributions of controllability have been found to relate to variations in empathic or hostile affect when judging the actions of others. For exam- ple, Meyer and Mulherin [1980] found that individuals reported more empathy and less anger when a hypothetical other’s social outcome was due to uncontrollable rather than controllable causes. Similarly, Weiner [ 19801 found that ascribing causality to factors controllable by another person maximized subjects’ reported negative affective reactions (e.g., anger and disgust) and minimized their subsequent willingness to help. More recent studies have also shown that social stigmas which are seen as controllable at onset by the stigmatized person and more likely to elicit anger than stigmas which are seen as uncontrollable [Weiner et al., 19881.

These effects should be further exacerbated when the other’s behavior also results in a negative outcome for the self, as occurs in social failure situations. For instance, seeing a person who is needy because of controllable factors such as drunkenness may produce anger and an unwillingness to help. However, one’s anger should be greater if the drunk person is also instrumental in causing a negative outcome for the self, such as being asked to leave an important social event because one’s partner is drunk. In short, aggression may represent a hostile response towards someone who is seen as being responsible for the occurrence of a negative self-related outcome. In addition, aggression may also signal an unwillingness to give up one’s efforts to achieve control over self-relevant outcomes [Wortman and Brehm, 19751.

Furthermore, if the other’s actions that cause the negative outcome are also seen as stable over time, there would be little point in responding prosocially to convince the other person to change his or her behavior. In this manner, a child who perceives a negative social outcome as due to the other person’s willful and unchangeable behavior may feel anger along with a low expectancy for changing the other’s behavior, and, in turn, respond with aggression. Thus, attributing one’s own social failure to external, stable, and controllable causes should lead to a hostile affective reaction and stimulate aggressive behavior.

In a preliminary investigation of these hypotheses [Guerra et al., 19901 we had as- sessed the relation between aggression and attributions about social failure in a sample of institutionalized delinquent boys. We found that delinquents who reported engaging in more physically aggressive behavior were more likely to attribute social failure to controllable causes, although no relation was found between physical aggression and the causal dimensions of externality and stability. We also found some evidence of a relation between attributions of controllability for the cause of a social failure and an- ticipation of greater hostile affect, and attributions of stability for the cause of a social failure and the endorsement of an aggressive response to the failure. However, the in- terpretation of those results was made somewhat problematic because of the lack of a nondelinquent comparison group of boys, and because we had only evaluated the de- linquents’ attributions independently on the three dimensions of controllability, exter- nality, and stability.

In the current study, we compare a sample of nondelinquent high school boys of similar age with our sample of delinquents. We examine the extent to which attribu-

424 Guerra et al.

tions for social failure differ between the two populations, and we examine the good- ness of fit in each population of a model that links attributions about social failure to hostility and aggression. Our hypothesis is that delinquents will show a greater ten- dency to attribute social failure to controllable, external, stable causes, and that the tendency to make such attributions will be related to hostility and aggression in both populations.

METHOD Subjects

The initial sample of delinquent subjects were 92 males incarcerated in a state cor- rectional facility who ranged in age from 15 to 19 years (Mean age = 16.8). They represented all the population of this facility who were able to participate during the scheduled sessions for whom permission could be obtained (93% of the available pool), The sample of 79 who completed all of the measures are the subjects described in the current study. About 67% of the subjects were of African-American or Hispanic ori- gin. Approximately 88% of the boys incarcerated in this facility were from lower or lower-middle class backgrounds. Because this was a state correctional facility, youth who are incarcerated typically have engaged in both repetitive and serious delinquent behaviors ranging from serious property crimes to murder. Furthermore, this specific facility housed the most chronic and serious offenders in the state. Although we did not differentiate the most violent delinquents from the other boys, the delinquent subjects represented the most serious offenders in the state, and were likely to be products of a long history of delinquent behavior, including at least some acts of aggression. The nondelinquent subjects were 119 males, between ages 15 and 19 (Mean age = 17.0), from a large urban high school in an ethnically diverse lower-middle class neighbor- hood as reflected by school data on median family income. Approximately 57% of the nondelinquent subjects were of African-American or Hispanic origin. They were re- cruited from several physical education classes in the high school, and were all sub- jects in these classes for whom permission could be obtained (96% of available subject pool). While some boys in this “nondelinquent” sample may, in fact, have engaged in delinquent behaviors, we defined the sample as nondelinquent by virtue of the fact that the boys were attending classes in a regular high school and were not institutionalized in a facility for delinquent youth. In contrast, we operationally defined our delinquent subjects as those boys who were incarcerated in a facility for seriously delinquent youth. This method of distinguishing between nondelinquent and delinquent groups based on school attendance or incarceration is consistent with procedures used in similar stud- ies [e.g.,Oyserman and Markus, 19901. While we were not able to obtain IQ informa- tion, we would expect the IQ’s of the delinquent sample to be lower than those of the nondelinquent sample [Hirschi and Hindelang, 19771. However, this study was not in- tended to address the extent to which low SES or low IQ play contributing roles in the psychological processes underlying aggression and delinquency, but to investigate the role of attributions in delinquent’s and nondelinquent’s aggressive behavior regardless of the possible origin.

Measures Level of aggression. To assess level of aggression, subjects were administered the

Physical Aggression Scale [Huesmann et al., 1984al. The Physical Aggression Scale is

Attributions and Aggression 425

a 13-item self-report measure which includes a 3-item self-report of general aggressive behavior (e.g., “I get angry and smash things”) with subject’s responses ranging from “never true” (scored as 0) to “almost always true” (scored as 4); a 5-item scale meas- uring the number of times during the past year where the subjects engaged in serious physical aggression against males (e.g., “How many times in the past year have you hit or punched another male?,” or “. . . knifed or threatened with a knife?”); and the same 5-item scale measuring physical aggression against females. Responses on these two scales range from “never” (scored as 0) to “four or more times” (scored as 4). Final scores are obtained by summing scores from each scale and dividing by three, so that the self-report of specific aggressive behavior scales contribute more to the over- all score than the general aggression scale, with a maximum possible score of 17 points. Both the delinquent and nondelinquent subjects varied widely in their level of self- reported aggression, with scores in both groups ranging from 0- 17.

This measure has been used in numerous previous studies with adolescents and young adults and has been found to be highly reliable and to correlate very significantly with non-self report measures, e.g. , peer nominations of aggression, measures of aggres- sive personality, and criminal justice data [Huesmann et al., 1984a; Lefkowitz et al., 19771. For example, coefficient alpha for the Physical Aggression Scale has been re- ported between .73 and .77. In a sample of 21 1 19-year-old high school males in New York (Huesmann et al., 1984a), scores on the scale correlated .42 ( P < .0001) with peer nominations of aggression and .28 ( P < .001) with MMPI scales F + 4 + 9, which have been shown to distinguish delinquents from nondelinquents (Huesmann et al., 1978). In a study of 198 30-year-old males [Huesmann et a]., 1984a1, scores of Physi- cal Aggression Scale correlated .46 (P<.OOOl) with the subject’s spouse’s reports of being physically abused by the subject. Even more impressive for the validity of the Physical Aggression Scale was the fact that it correlated .51 (P<.OOOI) with the num- ber of times the subject had been arrested according to records in the New York State criminal justice archives. Many of the questions on this scale have also been used in other validated scales used to measure aggression [e.g., Strauss et al., 19791.

Assessment for social failure (ASF). In order to assess all three dimensions of causal attributions (locus of causality, stability, and controllability) as well as subse- quent endorsement of affective and behavioral responses following social failure within a single measure, we developed the Assessment for Social Failure (ASF) questionnaire. This measure assessed these dimensions of causal attributions using the questions from the Causal Dimension Scales [Russell, 19821. However, the stimulus stories were con- structed to portray social failure instead of academic failure. The subject was asked to answer several questions about each of four stimulus stories involving social situations in which another person frustrates the subject. In two of the situations, subjects were exposed to failure involving peer-group entry with either a same gender or opposite gender peer. The other two situations involved blocked attainment of an instrumental goal by either a same gender or opposite gender peer. In previous research, these two types of situations have been identified as being social problems frequently encoun- tered by adolescents [Guerra and Slaby, 1990; Larsen, 19841. The phrasing of all re- sponses was also altered slightly from story to story in order to help prevent response set.

In administering the ASF Questionnaire, four different random orders of the four stories were used with approximately equal numbers of subjects to control for order

426 Guerra et al .

effects. Similarly, to minimize order effects between attributions and hostility, two of the four stories were followed by the three assessment scales in the order attribution, hostility, response endorsement and two were followed by the scales in the order hostil- ity, attribution, response endorsement.

Attributions for social failure. Following presentation of each story, subjects were asked “Why would this happen to you? (State the one reason which is most impor- tant).” They were then asked to rate the reason they gave on 9-point Likert scales de- veloped by Russell and his colleagues [Russell, 1982; Russell and McAuley, 19861 to measure whether the cause is viewed as controllable (three questions, e.g., “Is the cause- Controllable by you or others . . . Uncontrollable by you or others”), external (three questions), and stable (three questions). A score for each of the three subscales (con- trollable, external, stable) was calculated by summing the ratings (with proper rever- sals) on the three questions in the subscale ( 1 -9 possible points), and taking the average scores across the four stories, with a possible range of scores from 3-27. A major ad- vantage of this procedure is that subjects provide their own ratings of causal dimen- sions rather than using outside raters to code subject’s responses on these dimensions.

To control for response bias, the order of responses was reversed on one-third of the items. Since the stimulus stories for this measure were new, we first checked their in- ternal consistency by administering the measure to 61 adolescents not included in the present study. Coefficient alphas for each scale were: locus of causality (.61); stability (.71); and controllability (.70). In the present study, we rechecked the coefficient al- phas in both the delinquent and the high school sample and they were considered adequate.

As our specific hypotheses concerned the effect of attributing social failure to causes that were controllable and external and stable, we also investigated the interactive ef- fect of the three dimensions on the other variables.

Hostile reactions to failure. To assess endorsement of a hostile affective response, subjects were asked to rate on a 10-point Likert scale how they would feel if the out- comes described in each of the four ASF scenarios happened to them. They were asked to rate their feelings on each of four dimensions: guilt, shame, anger, and rage. For example, anger was assessed by asking subjects to rate “I would feel-Not at all angry . . . Very angry.” For each type of affective response, scores were averaged across the four stories, with a final possible score ranging from 0-9. Because the hypotheses of the present study concerned the relation between attributions for social failure and af- fective reactions indicating hostility, we created a hostile affective response scale by taking the mean ratings for anger and rage.

Endorsement of responses to failure. The subjects were asked to indicate the be- havioral responses they would likely make if the outcomes described in the ASF stories happened to them. They were presented with four possible actions they could take, and were asked to rate on a 5-point Likert scale the likelihood that they would chose each response. One of these actions would always be a physically aggressive action, e.g., “push or shove the girl,” one would always be a verbally aggressive action, e.g., “make a nasty comment to the girl,” one would be an assertive prosocial action, e.g., “ask again, explaining why. . . ,” and one would be a passive withdrawal action, e.g., “just forget it and wait. . . .” We derived a behavioral endorsement score for each of these behaviors by using the mean rating for each type of behavior across the four stories,

Attributions and Aggression 42 7

with possible scores ranging from 0-4. However, the passive scale was not included in subsequent analysis because of its low reliability (Coefficient alpha < .50).

Procedure Testing of all subjects was conducted in small groups by two female graduate stu-

dents during the regular school day. Two testing sessions were held separated by 2 weeks. To minimize the tendency of subjects to give socially desirable responses, all subject materials were identified by ID numbers only (institutional ID for the delinquent sub- jects and teacher-assigned IDS for the high school subjects), and subjects were informed that the experimenters conducting the testing sessions did not have access to subject names. The self-report measure of physical aggression was administered during the first session. The Attributions for Social Failure Questionnaire (ASF) was adminis- tered in the second session. For all measures, the experimenter read all items aloud in a standardized manner. After testing, the experimenter placed all assessment materials in a manilla packet and sealed the packet. All subjects were debriefed. No concerns about the gender or race of the experimenters or the confidentiality of subject’s re- sponses were expressed, and no difficulties in understanding the items were mentioned. Subjects were also asked to refrain from discussing their participation in the study with their peers. The same procedures were followed for both the delinquent and the high school samples.

RESULTS In Table 1 the means for the delinquent and nondelinquent samples are compared on

the variables of interest. A multivariate analysis of variance enabled us to reject the null hypothesis that the two populations’ mean scores were the same on these variables (MANOVA F(9, 188) = 13.3, P<.OOOI). However, contrary to what was expected, the means on the three individual attribution scores are very similar for the two samples and did not contribute to the overall significant difference. One cannot reject the hy- pothesis that the population means on these attribution scores are identical (MANOVA F(4,195) = 1.95, ns) for delinquents and nondelinquents (see Appendix). Rather, the significant overall multivariate F reflects the delinquents’ higher physical aggression

TABLE I. Comparison of Means and Standard Deviations for Delinquent and Nondelinquent Samples*

Behavior Attributions affect Response endorsement

Sample N CON EXT STAB AGG HOS VERB PHYS ASRT

Nondelinquent 119 16.1 13.4 8.6 4.1 4.7 1.35 0.81 2.02 boys (3.0) (3.5) (3.3) (2.1) (2.0) (1.02) ( .85) ( .94)

Delinquent 79 16.2 13.9 8.6 8.3 4.4 1.91 1.41 I .87 boys (4.2) (4.2) (4.6) (4.2) (2.7) (1.34) (1.28) (1.15)

P < ns ns ns .OOOl ns .01 ,001 ns MANOVA F(9,188) = 13.3, P < ,001

*Range of scores for controllability (CON), externality (EXT) and stability (STAB) = 3-27; range of scores for self-reported aggression (AGG) = 0-17; range or scores for hostility (HOS) = 0-9; range of scores for verbally aggressive (VERB), physically aggressive (PHYS), and assertive (ASRT) responses = 0-4.

428 Guerra et al.

scores (F(1,196) = 84.7, P<.OOOl), and the delinquents' greater preference for ver- bally aggressive (F( 1,196) = 10.9, fY.001) and physically aggressive (F(1,196) = 15.5, P<.OOl) responses.

These results add to the validity of the self-reported physical aggression measure and the response endorsement measure, but they are at odds with our hypothesis con- cerning attributions and aggression. While the delinquents are clearly much more ag- gressive than the nondelinquents and endorse more verbally and physically aggressive responses, they do not differ from nondelinquents in the kind of attributions they make about social failure.

In Table I1 the correlations between the key variables are compared for the two sam- ples. As one would expect, within both samples self-reported physical aggression in the recent past was correlated with anticipating hosti!ity after social failure and was correlated with the choice of more aggressive responses after social failure. Similarly, the intercorrelations among the different response endorsements and between them and hostility were about as expected. However, of greatest interest are the correlations be- tween the attributions measures and self-reported aggression, and the correlations be- tween the attributions and endorsement of hostile affect and aggressive responses.

Within the delinquent sample, self-reported aggression was correlated with a ten- dency to attribute social failure to controllable causes, but did not correlate with exter- nality or stability. Attributions of controllability were only marginally significantly correlated with anticipation of hostile affect and did not correlate directly with any of the response endorsement measures. However, anticipation of hostile affect did corre- late significantly positively with endorsement of a physically aggressive response. In addition, blaming social failure on stable causes did not correlate with hostile affect, but did correlate significantly positively with endorsement of an aggressive response and significantly negatively with endorsement of an assertive response.

However, within the nondelinquent population, a different pattern obtained. Self- reported aggression did not correlate with any of the attribution scores. The correla- tions of anticipated hostile affect with controllability, externality, and stability were all

TABLE 11. Correlations for Delinquent Boys (N = 79) Above the Diagonal and Nondelinquent Boys (N = 119) Below the Diagonalt

CON EXT STAB AGG HOS VERB PHYS ASRT

Controllable (CON) .02 -.26*d .28**' .20' -.05 . I 1 . I 1 External (EXT) .09 .03' . I3 .08 -.I0 - .09g -.05 Stable (STAB) - .02d .43***' .03 .08 .I6 30**h - 33**1 Self-reported

aggression (AGG) - .07' .07 . I4 .35*** .19+' .39*** -.23* Hostile affective

response (HOS) .21* .27** .18* .35*** .2Ik .31** -.06 Endorse verbal

agg. resp. (VERB) 0.0 .08 . I 1 &***I .52***k ,71*** - 37*** Endorse physical

agg. resp. (PHYS) .05 .15' .06" .32*** .43*** .64*** - .41*** Endorse assertive

resp. (ASRT) - .06 .02 -.09' - . I 1 -.04 -.17+ -.22**

tCorrelations with the same superscript letter differed significantly at the .I0 level (two-tailed test). Two tailed probabilities: 'P < .01 *P < .05 **P < .O1 ***P < ,001

Attributions and Aggression 429

significantly positive, but the correlation between these attribution measures and re- sponse endorsements were virtually zero. However, hostile affect did correlate signifi- cantly with endorsement of verbally and physically aggressive responses.

The correlations that differed significantly between the samples are noted in Table 11. One can see that the correlation between controllability and self-reported aggres- sion was significantly higher for the delinquents than for the nondelinquents. However, the correlations between endorsement of verbal aggression and both anticipated hostil- ity and self-reported aggression were very significantly higher for the nondelinquents.

To test whether the relations between attributions of controllability, externality, and stability and the other variables include interactive effects beyond the additive individ- ual effects of the scales, a series of multiple regressions were carried out. Each of the variables in Table I1 was predicted from the three individual attribution scores and their interactive composite for each sample. In none of these equations did the interactive composite have a significant effect once the three individual scale scores were in the equation. These results suggest that there is no significant interactive effect of control- lable, external, and stable attributions beyond their additive effects in either the delin- quent or non-delinquent sample.

In order to better assess the differential importance to aggressive behavior of attribu- tions about social failure in delinquent and nondelinquent boys, we constructed two structural models. The structural models are diagramed in Figures 1 and 2. The general models allow for attributions about social failure to affect directly the hostility the boy

DELINQUENT BOYS N - 7 9 Chl-Sq(8)S 10.4 p>.24 RMSE=.059

Fig. I , A structural model to explain the correlations observed within the delinquent sample.

430 Guerru et ul.

HIGH SCHOOL BOYS N=l19 Chi-Sq(8)=7.98

RMSE=.038 p>.43

Fig. 2. A structural model to explain the correlations observed within the nondelinquent sample

anticipates after failure as well as to affect directly the response endorsed. The hostility the boy anticipates is also allowed to affect the response endorsed. Both response en- dorsement and anticipated hostility then are allowed to affect actual self-reported ag- gressive behavior.

The parameters of this structural model were estimated separately from the two sam- ples using the EZPATH program. The resulting model for the delinquent population is shown in Figure 1, and for the nondelinquent population is shown in Figure 2. Both models reproduce the observed correlation matrices very well, i.e., the models explain the observed data. The dotted paths are those for which the coefficients were less than one standard error away from zero (nonsignificant).

According to the model in Figure 1 that fits the delinquent boys data (Chi-square (8) = 10.4, P>.24, RMSE = .059), self-reported physical aggression is predicted by the tendency to feel hostile after social failures (.28**) and by the tendency to select physi- cally aggressive responses to social failure (.37**). Selecting a physically aggressive response to failure is itself predicted by hostile affective reactions (.24**), and by attrib- uting social failure to stable (.32***) and controllable (. 18**) causes. These attribu- tions, particularly attributions of controllability (.24**), also predict hostile affective reactions. On the other hand, the endorsement of an assertive response, which is slightly predictive of less aggressive behavior, is strongly negatively related ( - .33***) to a be- lief that social failure is due to stable causes. Thus, for delinquents, attributing social failure to controllable and stable causes predicts a choice of physically aggressive re- sponses and a rejection of assertive responses. These attributions also predict a hostile

Attributions and Aggression 431

reaction which adds to the preference for aggressive responses. Finally, the hostile re- actions and preference for a physically aggressive response both predict actual level of self-reported aggressive behavior.

The model that fits the nondelinquent high school boys’ data is quite different, as one can see from Figure 2, though the model fits the data very well (Chi-square (8) = 7.98, P>.43, RMSE = .038). Attributing social failure to controllable, external, sta- ble causes has no significant direct effect on any of the response endorsement vari- ables. Furthermore, neither a hostile emotional reaction to social failure nor the choice of a physically aggressive response to a social failure significantly predicts self-reported aggressive behavior in nondelinquent high school boys. What does predict actual re- cent aggressive behavior is the choice of a verbally aggressive response to social fail- ure (.38***). As with delinquents, a hostile reaction to social failure predicts the choice of a verbally aggressive response (.5 1 ***) and a physically aggressive response (.43***), and attributing social failure to controllable causes predicts hostility (. 19**). Unlike delinquents, for high school students, attributing social failure to an external cause also predicts reported hostility (.21**).

DISCUSSION

These results suggest that there are both similarities and differences in cognitive pro- cessing following social failure for delinquent and nondelinquent adolescent boys of the same age. Looking at the similarities, the delinquent boys, on average, were no more likely than the nondelinquent boys to attribute social failure to controllable, exter- nal, and stable causes, even though they were substantially more aggressive and en- dorsed more verbally and physically aggressive responses. This finding suggests that simple assessments of differences in attributional style between delinquents and non- delinquents may not prove fruitful. Rather, our findings suggest that studying the relations between attributions about failure, hostile affective reactions, subsequent behavioral choices, and actual aggressive behavior in delinquent and nondelinquent populations may inform us about differences in the cognition (attribution)-affect-behavior linkages in these two populations.

For the nondelinquent boys, attributing social failure to external and controllable factors only indirectly predicted endorsement of verbally and physically aggressive re- sponses through increased hostile affect. As shown in Figure 2, there are no direct effects between attributions and response endorsement. This finding supports Weiner’s ( 1986) theory that attributions primarily guide behavior indirectly by giving rise to a specific affective reaction which then influences subsequent behavior, as previous stud- ies with normal populations have demonstrated [e.g., Meyer and Mulherin, 1980; Weiner, 19801. Also consistent with Weiner (1982), our results demonstrate a relation between attributing negative self-related outcomes to controllable factors and a subsequent hos- tile afSective reaction (e.g., anger and rage), which, in turn, predicts aggressive re- sponding. Finally, for these nondelinquent boys, actual physically aggressive behavior is related to the tendency to endorse verbally aggressive responses to social failure, but is not related to individual differences in endorsement of physically aggressive responses or to individual differences in anticipated hostility. This finding is somewhat difficult to interpret, although it may simply be that the physical aggression scores of the non- delinquent boys were too low and nonvariable to provide an adequate test.

432 Guerra et al.

In contrast, for delinquent boys, attributions of controllability had both an indirect and direct effect on endorsement of a physically aggressive response, and anticipated hostility and endorsement of a physically aggressive response directly predicted greater actual physical aggression. Thus, the relations between attributions, affect, and behav- ior differed from the relations found in more normative populations, with more support for a direct linkage between attributions and subsequent behavior among delinquents.

Of course, one must be judicious and not read too much into these results. The mod- els fit the data very well, but other models might fit even better. It is unlikely that a more powerful test of the relations (with greater sample sizes) would change the results as the effect sizes are quite small for the hypotheses that could not be rejected. Never- theless, failing to find a relation does not prove that one does not exist. One must also be careful about reading “causality” into these results. While the model posits direc- tional relations, the data were collected at one point in time. Models with other direc- tions for the effects might fit the data equally well.

These findings also call into question the utility of the locus of causality construct vis a vis aggression, particularly in terms of the previously demonstrated relation be- tween externality and aggression [Sadowski and Wenzel, 19821. One possible explana- tion is that measures of intemalityiexternality typically used in many studies may confound locus of causality with control beliefs. A related issue concerns the specific attribu- tional dimensions measured and the conceptualization of these dimensions relevant to aggressive behavior. In the present study, we relied primarily on the attributional di- mensions of locus of causality, stability, and controllability. It is important to under- stand, however, that the focus on these three dimensions in the psychological literature to a great extent followed from an interest in learned helplessness as related to both the mood state of depression and achievement research. Yet, as the ontogeny of aggressive behavior is qualitatively different, we must question whether these or other causal di- mensions should be of most relevance.

For example, while our results suggest that attributions of controllability following social failure are related to delinquent’s aggressive behavior, it is important to consider the most appropriate definition of this and other dimensions with respect to aggressive behavior. The literature on reactance [e.g., Wortman and Brehm, 19751 suggests that the dimension of controllability should be related to hostility and aggression in two ways-first by initial expectations of the controllability of desired outcomes by one- self, and second by the perception that this expected control has been taken away by others. Clearly, a weakness of the present study was the failure specifically to differen- tiate whether “controllability” refers to something one should have been able to con- trol or something which the other person controlled, although the latter is more likely since attributions were made following specific outcomes caused by others. Also, it may be useful to further distinguish one’s beliefs about the specific reasons underlying the other’s behavior. For instance, negative arousal and subsequent aggressive respond- ing may increase when others’ actions are seen as intentional, hostile, and directed towards controlling one’s own behavior.

Within recently proposed cognitive models for aggression [e.g., Berkowitz, 1993; Dodge, 1990; Guerra and Slaby, 1990; Huesmann, 19881, aggression is promoted by the development of cognitive structures that either promote instrumental aggressive be- havior to solve social problems or stimulate anger as an emotional response to social failure. From this theoretical viewpoint the kinds of attributions about the causes of negative outcome events that would seem most likely to stimulate anger or to cue in- strumental aggressive responses would seem most relevant. The current study suggests

Attributions and Aggression 433

that controllability, externality, and stability (as they have been defined in the dysphoric mood research) have relevance to aggression, but that a new conception of the dimen- sions of causality might be beneficial.

On what dimensions should such a taxonomy focus? First, it seems plausible that if a child believes a negative event is caused by another individual, the child is more likely to experience anger and hostility toward that individual. Such attributions would qualify as external attributions, but the externality is not the key. Rather, the key is whether the cause is viewed as due to the self or due to another person (as opposed to a range of external factors including luck and chance, Other-Self). A second dimension that should be of importance in affecting the child’s emotional reaction to a negative event would be the child’s perception of the intentionality of the other’s act. Was the negative event caused deliberately by the other or was it unintentional (Intentional- Unintentional)? A third dimension relevant to anger would be whether the other who caused a negative event for the child is seen as controlling the child or not. This Controlling-Not-Controllitzg distinction is not the same as the “controllable” distinc- tion discussed in earlier attribution literature. The question is whether the act causing the negative event is seen as an attempt to control the child’s behavior or not. The literature on counter-control and reactance suggests that more anger would be gener- ated in a child when an act by another is perceived by the child to be aimed at control- ling the child [Wortman and Brehm, 19751. A fourth dimension of relevance to anger would be a Hostile-Non-hostile dimension. The question is to what extent does the child perceive the negative event is due to another’s hostility toward the child. For example, the effect of a parent’s punishment on a child’s anger may be quite different depending on whether it is viewed as hostile or non-hostile even though it is perceived as other-caused, intentional, and controlling. In fact, Eron et al. [1971] finding that children who identify with their parents tend to respond to parental punishment more positively could be viewed in this light. The hypothesis that aggressive behavior is related to a tendency to attribute negative social outcomes to Others’ Intentional, Hos- tile, Controlling Behaviors would thus seem to be a plausible next step for investigation.

ACKNOWLEDGMENTS

tute of Mental Health to the first author. This research was supported in part by grant MH44768-01 from the National Insti-

REFERENCES Abramson LY, Seligman MEP, Teasdale, JD (1978),

Learned helplessness in humans: Critique and re- formulation. Journal of Abnormal Psychology 87:49-74.

Berkowitz L (1993): “Aggression: Its Causes, Con- sequences, and Control.” New York: McGmHill.

Dodge KA (1980): Social cognition and children’s ag- gressive behavior. Child Development 5 1: 162- 170.

Dodge KA ( 1986): A social-information-processing model of social competence in children. In Perl- mutter M (ed) “Minnesota Symposium on Child Psychology (Vol. 18)” Hillsdale, NJ: Lawrence Erlbaum, pp 77-125.

Dodge KA, Frame CL (1982): Social-cognitive biases

and deficits in aggressive boys. Child Develop- ment 53:629-635.

Dodge KA, Coie JD, Pettit GS, Price JM (1990): Peer status and aggression in boys’ groups: De- velopmental and contextual analyses. Child De- velopment 61:1289-1309. Eron LD ( 1987): Thedevelopmentof aggressive behavior from the perspective of a develop- ing behaviorist. American Psychologist 42:

Eron LD, Walder LO, Lefkowitz MM (1971): “The Learning of Aggression in Children.” Boston: Little, Brown.

Guerra NG, SIaby RG (1990): Cognitive mediators

435-442.

434 Guerra et al.

of aggression in adolescent offenders: Interven- tion. Developmental Psychology 26:269-277.

Guerra NG, Huesmann LR, Zelli A (1990): Attri- butions for social failure and aggression in in- carcerated delinquent youth. Journal of Abnormal Child Psychology 18:347-356.

Hirschi T, Hindelang MJ (1977): Intelligence and delinquency: A revisionist view, American So- ciological Review 42:571-587.

Huesmann LR ( 1988): An information-processing model for the development of aggression. Ag- gressive Behavior 14: 13-24.

Huesmann LR, Eron LD (1984): Cognitive process and the persistence of aggressive behavior. Ag- gressive Behavior 10:243-25 I .

Huesmann LR, Lefkowitz MM, Eron LD (1978): The sum of MMPl F, 4, and 9 as a measure of aggression. Journal of Consulting and Clinical

Huesmann LR, Eron LD, Lefkowitz MM, Walder LO (1984a): The stability of aggression over time and generations. Developmental Psychology 20: 1120-1 134.

Huesmann LR, Lagerspetz K, Eron LD (1984b): In- tervening variables in the TV violence-aggression relation: Evidence from two countries. Develop- mental Psychology 20:746-775.

Kupersmidt JB, Coie JD (1990): Preadolescent peer status, aggression and school adjustment as pre- dictors of externalizing problems in adolescence. Child Development 61:1350-1362.

Lefkowitz MM, Eron LD, Walder LO, Huesmann LR (1977): “Growing Up to be Violent: A Lon- gitudinal Study of the Development of Aggres- sion.” New York: Pergamon.

Larsen K (1984): Me&-cognitive training with learn- ing disabled delinquents. Unpublished manu- script. University of California, Santa Barbara.

Meyer JP, Mulherin A (1980): From attribution to helping: An analysis of the mediating effects of affect and expectancy. Journal of Personality and Social Psychology 39:20 1-2 10.

Olweus D (1984). Development of stable aggressive reaction patterns in males. In Blanchard R, Blan- chard C (eds): “Advances in Aggression Research (Vol I ).” New York: Academic Press, pp 103- 137.

Psychology 46: 1071-1078.

Oyserman D, Markus H (1990): Possible selves and delinquency. Journal of Personality and Social

Richard B, Dodge KA (1982): Social maladjustment and problem-solving in school-aged children. Journal of Consulting and Clinical Psychology

Russell D (1982): The causal dimension scale: A measure of how individuals perceive causes. Jour- nal of Personality and Social Psychology 42: 1137-1 145.

Russell D, McAuley E (1986): Causal attributions, causal dimensions, and affective reactions to suc- cess and failure. Journal of Personality and So- cial Psychology 50: l 174- l 185.

Sadowski CJ, Wenzel DM (1982): The relationship of locus of control dimensions to reported hos- tility and aggression. Journal of Psychology

Seligman MEP ( 1975): “Helplessness.” San Fran- cisco, CA: Freeman.

Shure MB, Spivack G (1976): “Problem-Solving Techniques in Childrearing. San Francisco, CA: Jossey-Bass.

Slaby RG, Guerra NG (1988): Cognitive mediators of aggression in adolescent offenders: I. Assess- ment. Developmental Psychology 24380-588.

Strauss MA, Gelles RJ, Steinmetz SK (1979): “Behind Closed doors: Violence in Families.” New York: Doubleday/Anchor.

Neiner B ( 1 980): A cognitive (attribution)-emotion- action model of motivated behavior: An analy- sis of judgment of help-giving. Journal of Per- sonality and Social Psychology 39: 186-200.

Weiner B (1982): The emotional consequences of causal attributions. In Clark MC, Fiske ST (eds), “Affect and Cognition.” Hillsdale, NJ: Erlbaum.

Weiner B, Perry RP, Magnusson J (1988): An attributional analysis of reactions to stigmas. Journal of Personality and Social Psychology 55:738-748.

Wortman CG, Brehm JW (1975): Responses to un- controllable outcomes: An integration of reac- tance theory and the learned helplessness model. Advances in Experimental Social Psychology

Psychology, 59: 1 12- 125.

421787-793.

112r227-230.

8:277-336.

APPENDIX

This MANOVA included an interactive product term for the three attribution meas- sures as another criterion variable; so it is fair to say that the delinquents and non- delinquents also do not differ in the extent to which they attributed failure to controllable and external and stable causes.