You are on page 1of 9

Developmental Psychology Copyright 1988 by the American Psychological Association, Inc.

1988, Vol. 24, No. 4, 580-588 0012-1649/88/S00.75

Cognitive Mediators of Aggression in Adolescent Offenders:


1. Assessment
Ronald G. Slaby and Nancy G. Guerra
Harvard University
Department of Human Development

The role of cognitive mediators in identifying differences in aggression was examined. Male and
female adolescents incarcerated for antisocial aggression offenses and high-school students rated as
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

either high or low in aggression were compared in terms of two sets of cognitive mediators: skills in
This document is copyrighted by the American Psychological Association or one of its allied publishers.

solving social problems and beliefs supporting aggression. Antisocial-aggressive individuals were
most likely (and low-aggressive individuals were least likely) to solve social problems by: denning
problems in hostile ways; adopting hostile goals; seeking few additional facts; generating few alterna-
tive solutions; anticipating few consequences for aggression; and choosing few "best" and "second
best" solutions that were rated as "effective." Antisocial-aggressive individuals were also most likely
to hold a set of beliefs supporting the use of aggression, including beliefs that aggression: is a legiti-
mate response; increases self-esteem; helps avoid a negative image; and does not lead to suffering by
the victim. The ways in which these findings further elaborate a model of social-cognitive develop-
ment and extend it to antisocial-aggressive adolescents are discussed.

As children grow older, their individualized patterns of ag- aggression has been conducted with elementary-school children
gression have been found to become increasingly stable and and has not been extended to adolescents. This research has
differentiated from those of their age-mates (Olweus, 1984). To also been limited to the range of typical aggressive behaviors
account for this finding, it has been hypothesized that, as they found in school children. Yet, for some adolescents, patterns of
develop, individuals* patterns of aggression are governed to an aggression become extreme and antisocial, leading eventually
increasing extent by internalized and habitual patterns of cog- to the criminal acts ofjuvenile offenders. By comparing the cog-
nition (Eron, 1987; Slaby & Roedell, 1982). For example, a va- nitive mediational patterns of antisocial aggressive juvenile
riety of studies have focused on identifying those social infor- offenders with those of high-aggressive and low-aggressive high-
mation-processing skills that differentiate aggressive and non- school students, the present study was designed to extend previ-
aggressive children (e.g., Dodge, 1980; Richard & Dodge, 1982; ous research on social-cognitive development beyond the age
Spivack & Shure, 1974). Somewhat separate from the research of elementary-school children and beyond the range of typical
on cognitive skills involved in processing information, a second aggressive behaviors found among such children.
body of research has focused on cognitive content in the form
The specific cognitive skills assessed represent an elaboration
of generalized beliefs that represent adopted standards of con-
of recent research on the social problem-solving process and its
duct or response-outcome expectancies supporting the use of
aggression (e.g., Bandura, 1986; Perry, Perry, & Rasmussen, relation to aggression. Whereas early studies in this area em-
1986). Little attention has been given to developing an integra- phasized the relation between aggressive behavior and deficits
tive model of social-cognitive development that explores both in the child's ability to generate alternative solutions to social
the processes and the content of cognition that may serve to problems (Shure & Spivack, 1976; Spivack & Shure, 1974),
mediate aggressive behavior. subsequent efforts have focused on a broader range of cognitive
Furthermore, much of the research on cognitive mediation of processes. The work of Dodge and his associates (e.g., Dodge,
1980; Dodge & Frame, 1982) has been notable in this area. On
the basis of research with school-age children, Dodge (1986) has
formulated afive-stepsequential model and has proposed that
This article is based on a dissertation by Nancy G. Guerra, submitted
aggressive behavior may result from deficits and biases in pro-
in partial fulfillment of the requirements for the doctoral degree, Har-
vard University, 1986. cessing social information at any or all of the steps. The five
This research was supported by a grant from the State of California, steps he proposed are: (1) encoding of social cues, (2) represen-
Office of Criminal Justice Planning. Points of view or opinions in this tation and interpretation of cues, (3) response search, (4) re-
article are those of the authors and do not necessarily represent the sponse decision, and (5) enactment. Relating to the first four
official position or policies of the Office of Criminal Justice Planning. steps of this model, six information-processing components of
We would like to thank the staff of the Santa Barbara City and County social problem solving were investigated in the present study:
School Districts and the staff of the California Youth Authority for their seeking information, defining the problem, selecting a goal,
cooperation. We also thank Ed DeVos and Diana Arezzo for their help generating alternative solutions, anticipating consequences, and
in preparing the figures.
prioritizing responses.
Correspondence concerning this article should be addressed to Nancy
G. Guerra, who is now at the University of Illinois at Chicago, Depart- Information seeking (i.e., requesting additional facts relevant
ment of Criminal Justice, Box 4348, Chicago, Illinois 60680. to the situation) was seen as part of the individuaTs encoding
580
COGNITIVE MEDIATORS OF AGGRESSION 581

process (Step 1) whereby he or she selectively attends to relevant solution thinking is the likelihood that one's "best" and "second
cues and generates needed information. On the basis of related best" solutions are "effective," as the term is defined by Richard
research that has found that aggressive children attend to fewer & Dodge (1982): a solution judged by others to be nonviolent
cues than their less aggressive peers (e.g., Dodge & Newman, and goal-directed. A recent study of elementary-school children
1981), we hypothesized that, when compared with low-aggres- found some evidence that, although high- and low-aggressive
sive students, antisocial-aggressive offenders and high-aggres- children's best solutions were equally likely to be effective, the
sive students would seek less information. second best solutions of high-aggressive children were less likely
As part of the process of representation and interpretation of to be effective than those of low-aggressive children (Richard &
cues (Step 2), an individual must define the problem and select Dodge, 1982). Thus, it was expected that when compared with
a goal. In the present study, subjects were asked to choose be- low-aggressive students, antisocial-aggressive offenders and
tween responses indicating either a hostile or a nonhostile per- high-aggressive students would prioritize their solutions so that
ception of the situation. We hypothesized that the hostile attri- their second best (if not their best) solution would be less likely
butional bias that has been previously found to characterize ag- to be rated as effective.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

gressive school children (e.g., Dodge & Frame, 1982; Nasby, Beyond the cognitive skills involved in processing informa-
Hayden, & dePaulo, 1980) would manifest itself in the way that tion emphasized in Dodge's (1986) model, the present study
antisocial-aggressive and high-aggressive adolescents defined was designed to assess cognitive content in the form of general-
social problems (on the basis of the perception that others were ized beliefs that may support the use of aggressive behavior. Re-
hostilely motivated adversaries) and selected problem-solving cent applications of social learning theory to the study of aggres-
goals (on the basis of attack or defense against the person per- sion have stressed both the motivational and the regulatory
ceived as hostile). functions of an individual's generalized beliefs concerning ag-
One aspect of the response search process (Step 3) involves gression (Bandura, 1986). We assessed three categories of gen-
the subject's skill in generating alternative solutions to a social eral social beliefs regarding aggression: (a) the legitimacy of ag-
problem. Past studies of alternative solution thinking have been gression, (b) the expected outcomes for the aggressor, and (c)
inconclusive; whereas some studies found that aggressive chil- the expected outcomes for the victim. Much of the socialization
dren generated fewer solutions than their less aggressive peers literature points to the multitude of opportunities for children
(Richard & Dodge, 1982; Shure, Newman, & Silver, 1973), to adopt as standards of conduct for themselves the belief that
other studies have failed to support this relation (Sharp, 1981). aggression is often an acceptable response in a variety of situa-
These mixed findings may be partially due to the fact that, in tions (e.g.. "It's O.K. to hit someone if you don't like him or
most previous research using hypothetical situations, subjects her") (Bandura. 1973; Parke & Slaby, 1983). We hypothesized
were presented with a predefined problem and asked to gener- that antisocial-aggressive and high-aggressive subjects would be
ate solutions focused on a specific goal. If the specific preset goal more likely than low-aggressive subjects to believe that aggres-
were simply less relevant for an aggressive individual, he or she sion is a legitimate response in a variety of situations.
might be expected to generate fewer solutions to meet that goal. In addition to providing an individual with standards of con-
Thus, we were interested in assessing the generation of solutions duct, beliefs can represent generalized response-outcome ex-
when subjects were first allowed to select both their problem pectancies concerning the aggressor or the victim that support
and goal. the use of aggression. For example, aggressive school children
Within the social information-processing framework, rela- have been found to be more likely than nonaggressive children
tively little attention has been directed toward the response- to endorse the belief that aggression will both reduce aversive
decision step (Step 4), wherein an individual considers the con- treatment by others and produce tangible rewards (Perry, Perry,
sequences of various responses and prioritizes the set of possi- & Rasmussen, 1986). In addition, aggressive individuals have
ble responses. Several early studies with preschool children been found to expect aggression to enhance self-esteem and to
failed to find significant differences in the number of conse- maintain status among peers (Bandura, 1973; Bornstein, Bel-
quences generated by high- and low-aggressive children (Shure, lack, & Hersen, 1980; Short, 1968). Thus, we hypothesized that
Newman, & Silver, 1973; Spivack & Shure, 1974). The failure antisocial-aggressive and high-aggressive youth would be more
to demonstrate differences between high- and low-aggressive likely to hold the beliefs that aggression increases self-esteem
preschool children on this measure may have resulted, in part, and helps to avoid a negative image with peers.
from a floor effect related to the extremely low number of conse- The disengagement of self-sanctions for behavior that harms
quences generated by virtually all young children (Marsh, others has also been linked with a variety of beliefs, including
1982). On the other hand, in research with adolescents, delin- denial of injury to the victim (Bandura, 1973; Sykes & Matza,
quents have been found to consider fewer consequences than 1957). Accordingly, we hypothesized that antisocial-aggressive
their nondelinquent peers when solving social problems (Hains and high-aggressive youth would be more likely to believe that
&Ryan, 1983). Thus, we hypothesized that antisocial aggressive others deserve to be victimized and that victims of aggression
offenders would anticipate fewer consequences for responding don't suffer.
aggressively than their nonoftender peers in high school who
were either high or low in aggression.
Another aspect of the response-decision process (Step 4) is Method
the prioritization of solutions. Although the ability to generate
Subjects
a variety of solutions may provide an important repertoire from
which to choose one's preferred solutions, a qualitative evalua- Subjects were 144 adolescents, equally divided by sex and ranging in
tion is also involved. One qualitative dimension of alternative age from 15 to 18 years, with a mean age of 17 years, 1 month. They
582 RONALD G. SLABY AND NANCY G. GUERRA

were selected to form three comparable groups of equal size: antisocial ized delinquents. 10-week test-retest reliability (Kendall's T) was estab-
aggressive: high aggressive; and low aggressive. lished for each of the seven measures and ranged from .75 to .93. The
The subjects in the antisocial-aggressive group were drawn from a questionnaire consisted of two hypothetical stories wherein subjects
sample of adolescents incarcerated in a state juvenile correctional facil- were asked to imagine that they were involved in a situation in which a
ity for having committed one or more violent criminal acts (i.e.. assault same-sex, unknown peer is interfering with a personal-instrumental
and battery, robbery, rape, attempted murder, and murder). The institu- goal for reasons that are ambiguous, such as in the following story:
tion was chosen because it was the only maximum security facility in the
state (and hence likely to house the most violent offenders) that accepted You're playing on your school's Softball team and have a big game
female as well as male offenders. The institution housed an average of coming up. \bu've been trying to practice as much as you can after
600 adolescent offenders between the ages of 14 and 21 years; about 60% school. One day you go outside to practice with a friend but can't
were male; about 60% were members of minority groups, predomi- find any bats. You see a guy (girl) sitting on the bench just twirling
nantly Blacks and Hispanics; and about 80% had a history of at least a bat around. "Hey, let me borrow that for awhile," you say. "No
one antisocial-aggressive offense. All inmates1 files were screened to de- way" says the guy (girl).
termine their eligibility on the basis of age (15-18 years), reading level
Following presentation of each story, subjects were asked seven ques-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

(Grade 6 or better), intellectual functioning (not learning handicapped),


tions: (a) Problem definition. What is the problem?; (b) Goal selection,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

offense (antisocial aggression), parole date (not less than 6 months), and
written consent indicating their willingness to participate in a future If you were to solve this problem, what would be your goal?; (c) Number
training program. of facts. Would you need any more information? If so, what?; (d) Num-
ber of solutions. What are all the ways you think you can solve this
The subjects for the high-aggressive and low-aggressive groups were problem?; (e) Best solution, What do you think is the very best solu-
drawn from a sample of students attending the I Ith or 12th grade of a tion?; (f) Second best solution. What do you think is the second best
public high school chosen for its comparability with the correclional solution?; and (g) Number ofconsequences. What are all the things that
facility in terms of geographic location, predominance of lower- and might happen if you did this? (A standard aggressive response was given;
middle-class youth, and high percentage of minority youth (about e.g.. "Grab the bat and hit him/her")
55%). Because teacher ratings have been found to correlate highly with
For problem definition, subjects chose between a hostile (e.g., "This
peer nominations and other indices of aggression (e.g., Eron, Walder, &
guy/girl won't let you use his/her bat, even though he's/she's not using
Lefkowitz, 1971), physical education teachers were asked to fill out a
it") and nonhostile (e.g., "You don't have a bat and want to practice")
10-item scale for all students (/V = 267). For each item (e.g., "uses physi-
definition. On the basis of their choice indicating perceived hostility for
cal force to get his/her way"), subjects were rated on a 5-point scale.
neither, one, or both of the two stories, each subject received a percent-
Level of aggression was determined by summing the ratings for all 10
age score of 0, 50, or 100. Likewise, for goal selection, subjects chose
items. In previous research, this rating scale has been found to show
between a hostile (e.g., "Show the guy/girl not to mess with me") or
high 10-week test-retest reliability (Kendall's T = .96)(Guerra, 1983).
nonhostile (e.g., "Get a bat so that you can practice") response, with
A total of 223 teacher ratings were returned.
responses scored as in problem definition. For number of facts, number
Because of differences typically found between males and females on of solutions, and number ofconsequences, the total number of nondu-
indices of aggression, subjects were classified as high or low aggressive plicative responses was scored by raters on the basis of scoring guidelines
within their own sex group. Subjects in the bottom 10% of the ratings in which only identical responses were to be considered duplicative. Re-
were dropped in order to eliminate extremely nonaggressive subjects. sponses to best and second-best solutions were recorded and later rated
Then, those males and females who scored in the top and bottom one- by an experimenter (who was blind to the status of subjects) as effective
third of the rankings for their sex were classified as high aggressive and if they were nonviolent and nonhostile. Interrater reliability (.94) was
low aggressive, respectively. established for these two measures by having a second coder rate a ran-
For the purpose of matching subjects assigned to the three status dom sample of 25 responses.
groups, subjects from the available pool for each group (156 antisocial Beliefs measures. The beliefs questionnaire consisted of a list of 18
aggressive, 68 high aggressive, 70 low aggressive), subjects were catego- statements (presented in a standard order) depicting beliefs supporting
rized by sex, age (15-16 or 17-18 years), and minority group status aggression. Because this measure was developed specifically for the pres-
(minority or nonminority). Elimination of qualified subjects to achieve ent study, 10-week test-retest reliability was established through a pilot
equal group sizes (24 males and 24 females) was accomplished by ran- study using a sample of 66 institutionalized delinquents and was found
dom selection. Given that IQ data were unavailable, IQ was not used in to be high (Kendall's T - .86). Five beliefs supporting aggression were
the matching procedure. However, a minimal level of intellectual ability measured: (a) Legitimacy of aggression (6 items, e.g., "It's O.K. to hit
for the incarcerated group was established by screening for the criteria someone if you just go crazy with anger"); (b) Aggression increases self-
of reading level and intellectual functioning. esteem (3 items, e.g.. "It's important to show everyone how tough you
are by being a good fighter"); (c) Aggression helps to avoid a negative
Procedure image O items, e.g., "If you back down from a fight, everyone will think
you're a coward"); (d) Victims deserve aggression (3 items, e.g., "If
Procedures for the administration of both assessment measures fol- someone gets beat up, it's usually his or her own fault"); and (e) Victims
lowed those outlined by Marsh (1982). The social problem-solving don't suffer (3 items, e.g., "People who get beat up badly probably don't
questionnaire, followed by the beliefs questionnaire, was administered suffer a lot"). A belief supporting aggression was indicated by a "true"
to subjects in a small group setting. One of two experimenters read each response in half of the items and a "false" response for the other half of
measure in a standardized manner, making sure subjects had adequate the items. The percentage of responses indicating support for aggression
time to answer; no difficulties in understanding the materials were was calculated for each measure. Alpha coefficients, indicating the in-
noted. After testing, subjects were thanked and asked to refrain from ternal consistency of items within each of the five belief categories, were
having discussions with their peers about their participation in the .67, .53. .68. .72, and .37. respectively.
study. The same standardized procedures were followed at both the high
school and the juvenile correctional institution.
Social problem-solving measures. The measures of social problem- Results
solving skills were adapted from the Interpersonal Problem Solving
Analysis (IPA) developed by Marsh (1982). On the basis of an earlier For each of the 12 dependent measures, a 3 X 2 analysis of
pilot study of the adapted measure using a sample of 66 institutional- variance (ANOVA) was performed, with Group Status (antiso-
COGNITIVE MEDIATORS OF AGGRESSION 583
100 PROBLEM DEFINITION 100 - GOAL ORIENTATION
90 - (perceived hostility) (perceived hostility)
90 1
80 - Females 80 1
70 - 70 -
60 - 60 -
0% — —i H 50-
50 - / 0% 1
40 - 40 -
11II I—,
/
30 -
20- /
/
In 30
20
-
- /
/
/

1 Bn
10 - 10 -
0- 1 ^ 0 -
f \ r It •
Low High Antisocial Low i High i Antisocial

AGGRESSION GROUP AGGRESSION GROUP


This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

4.0 - NUMBER OF FACTS 4.0 - NUMBER OF SOLUTIONS


3.5 - 3.5-
3.0- 3.0 -
2.5 -
s 2.5 -

x
2.0 - la 2.0 -
1.5 - 1.5-
1.0 - 1.0 -
0.5- 0.5-
0.0 - 0.0 -
Low High Antisocial Low High Antisocial
AGGRESSION GROUP AGGRESSION GROUP

EFFECTIVE SOLUTIONS fist & 2nd)


"1

• -1 11 (
4.0 - NUMBER OF CONSEQUENCES 90 -
3.5- B0 - 1
3.0- 70-

I• 1
I• ••
B


2.5- 60-
50-
2.0 -
1.5 -
I I
40-
30-

l •1
• I
1.0 -
0.5 - I 20-

0.0 - I • 1
10-
n n i
IAntisocial
B
_•_!
Low High Antisocial Low High

AGGRESSION GROUP AGGRESSION GROUP

Figure 1. Social problem-solving skills for each group.

rial aggressive, high aggressive, and low aggressive) and Sex of sures: for problem definition, F(2, 138) = 35.25, p < .001; for
Subject as the main factors. The data were so strongly differen- goal selection, F(2, 138) = 10.93, p < .001; for number of facts
tiated in support of the hypotheses that 7 of the 12 ANOVAS requested. f(2, 138) = 35.58, p < .001; for number of alterna-
were compromised by the fact that they involved cells with zero tive solutions, F(2, 138) = 29.78, p< .001; for number of conse-
variation (i.e., cells with means of 0% or IOO°b). This result quences. ,R2, 138) = 35.22, p < .001; for effectiveness of best
poses an unusual statistical situation because there is no way solution. F(2. 138) = 14.22. p < .001; and for effectiveness of
statistically to create variation where there is none. In light of second-best solution, F(2,138) = 3 7 . 1 5 , p < . 0 0 l .
both the robustness of ANOVA to violations of its assumptions Post hoc comparisons were then conducted for each measure.
and the strength of the data, ANOVA was used as the common For problem definition, significant differences were revealed be-
basis for analyzing all 12 measures. Post hoc comparisons were tween all status groups. Antisocial-aggressive subjects were sig-
conducted using the Tukey procedure; significant differences nificantly more likely to choose a hostile definition than were
using this procedure reflect at least a .05 level of significance. high- and low-aggressive subjects. High-aggressive subjects were
Mean scores for all measures are illustrated in Figures I and 2.1 significantly more likely to choose a hostile definition than were
low-aggressive subjects. For goal selection, no significant
differences were revealed between antisocial-aggressive and
Group Status
1
ANOVA for the social problem-solving measures revealed a Group means and standard deviations are available from the au-
significant main effect for Group Status for each of the 7 mea- thors on request.
584 RONALD G. SLABY AND NANCY G. GUERRA
LEGITIMACY OF AGGRESSION
100 -.
90 -
80 J Males
70 -
60 - Females
2 50 -
u 40-
30-
20-
10 J
0
Low High Antisocial

AGGRESSION GROUP
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

INCREASES SELF-ESTEEM AVOIDS A NEGATIVE IMAGE


This document is copyrighted by the American Psychological Association or one of its allied publishers.

100 -,
90 -
80 -
70 -
60 -
50 -
5
40 - 0%
30 -
20 -
10 -
0
Low High Antisocial High Antisocial

AGGRESSION GROUP AGGRESSION GROUP

VICTIMS DESERVE AGGRESSION VICTIMS DONT SUFFER

Low High Low High Antisocial

AGGRESSION GROUP AGGRESSION GROUP

Figure 2. Social beliefs supporting the use of aggression for each group.

high-aggressive subjects, although both groups were signifi- Post hoc comparisons were then conducted for each measure.
cantly more likely than the low-aggressive subjects to choose a Significant differences between each of the three status groups
hostile goal. For number of facts, number of solutions, and were found for legitimacy of aggression and for aggression helps
number of consequences, significant differences were revealed to avoid a negative image, with antisocial-aggressive subjects be-
between all groups, with low-aggressive subjects generating ing more likely than both high- and low-aggressive subjects to
more items than both high- and antisocial-aggressive subjects, agree with these beliefs, and high-aggressive subjects being sig-
and with high-aggressive subjects generating more items than nificantly more likely than low-aggressive subjects to agree with
antisocial-aggressive subjects. For best and second-best solu- these beliefs. For the belief that aggression increases self-esteem,
tions, no significant differences were revealed between low- and no significant differences were revealed between high- and low-
high-aggressive subjects, although both were significantly more aggressive subjects, although antisocial-aggressive subjects were
likely than antisocial-aggressive subjects to choose a solution significantly more likely to agree with this belief than subjects
that was rated as effective. in both of these groups. For the belief that victims don't suffer,
ANOVAS for the beliefs measures revealed a significant main no significant differences were revealed between antisocial- and
effect for Group Status for each of the five measures: for legiti- high-aggressive subjects, although both groups were signifi-
macy of aggression, F(2, 138) = 26.42,/? < .001; for aggression cantly more likely than low-aggressive subjects to hold this be-
increases self-esteem, F{2, 138) = 26.60, p < .001: for aggres- lief. Post hoc comparisons for victims deserve aggression re-
sion helps to avoid a negative image, F(2, 138) = 36.10, p < vealed the unexpected finding that antisocial-aggressive sub-
.001; for victims deserve aggression, F(2, 138) = 9.57, p < .001; jects were significantly less likely to hold this belief than subjects
and for victims don't suffer, F(2, 138) = 9.06, p < .001. in both the high- and low-aggressive groups.
COGNITIVE MEDIATORS OF AGGRESSION 585

Sex of Subject Pearson product-moment coefficients were computed. Gener-


ally, all measures except the belief that victims of aggression
ANOVAS for the social problem-solving measures revealed a
deserve to be victimized were consistently intercorrelated with
significant main effect for Sex of Subject for five of the seven
one another. Correlation coefficients among all measures are
measures: for problem definition, F\\, 138) = 25.61,/? < .001:
presented in Table I.
for goal selection, F\ I, 138) = 6.12, p< .05; for number of facts
requested, F\\, 138) = 7.12, p < .01; for effectiveness of best
solution, F(\, 138) = 10.63, p < .01; and for effectiveness of Predictions of Aggression Group Status
second-best solution, F{ 1, 138) = 21.29, p < .001. For problem From the Cognitive Measures
definition and goal selection, males were significantly more
likely than females to focus on perceived hostility. Females Although significant F values for Group Status were reported
asked for significantly more facts than males. For solution in the ANOVAS for all 12 measures, these values offer a poor
effectiveness, females were significantly more likely than males barometer of each measure's relative ability to discriminate be-
tween the three aggression status groups, given that they are
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

to choose best and second-best solutions that were rated as


confounded by intercorrelations among the measures that con-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

effective.
tribute redundantly to overall discrimination, indeed, the mea-
ANOVAS for the beliefs measures revealed a significant main sures of social problem-solving skills were designed to build on
effect for Sex of Subject for three of the five measures: for ag- each other in some ways (e.g., the solutions generated would be
gression increases self-esteem, F(\, 138) = 15.52. p < .001; for expected to be a function of the problem definition selected).
victims deserve aggression. F(1, 138) = 4.06./? < .05: and for Thus, three stepwise discriminant analyses were performed in
victims don't suffer, F( 1, 138) = 7.36, p < .01. Males were sig- an attempt to assess the measures' relative discriminatory
nificantly more likely than females to agree with the beliefs that power according to the Wilks's lambda likelihood ratio crite-
aggression increases self-esteem and that victims don't suffer, rion for maximizing the ratio of between-group to within-group
whereas females were significantly more likely than males to variability. A forward selection procedure was used in which,
agree with the belief that victims of aggression deserve to be at each step, the measure was entered that contributed most
victimized. discriminatory power after partitioning out the variance for
which any previously entered measure had already accounted.
Interactions Between Variables Selection of measures continued until none of the remaining
A significant Group Status X Sex of Subject interaction was variables met the entry criterion for adding significantly (.05)
found for problem definition, F(2, 138) = 7.33, p < .001, with to the discriminatory power of the measures already entered.
the pattern of findings being somewhat different for males and An analysis of the seven measures of social problem-solving
females. For males, antisocial-aggressive and high-aggressive skills revealed the following hierarchy of measures with signifi-
subjects did not differ, but both groups showed significantly cant discriminatory power: for information seeking, r2 = .33;
higher levels of perceived hostility than low-aggressive subjects. for problem definition, partial r2 = .20; for number of conse-
For females, low-aggressive and high-aggressive subjects did not quences, partial r2 = .11; and for effectiveness of second-best
differ, but antisocial-aggressive subjects showed significantly solution, partial r2 = .07. The total variance accounted for by
higher levels of perceived hostility than subjects in both of these these four measures was 71%.
groups. High-aggressive males showed higher levels of perceived An analysis of the five measures of beliefs revealed the follow-
hostility than high-aggressive females. ing hierarchy of measures with significant discriminatory
A significant Group Status x Sex of Subject interaction was power: for aggression helps to avoid a negative image, r2 = .42;
found for effectiveness of best solution, F(2, 138) = 4.39, p < for aggression increases self-esteem, partial r 2 = .13; for aggres-
.05, and for effectiveness of second-best solution, F(2, 138) = sion is a legitimate response, partial r2 = .12; and for victims
3.29, p < .05. For males, antisocial-aggressive subjects were sig- deserve to be victimized, partial r2 = .10. The total variance
nificantly less likely than both high-aggressive and low-aggres- accounted for by these four measures was 77%.
sive subjects to choose effective best and second-best solutions; In order to analyze the relative discriminatory power of the
high-aggressive subjects were also less likely than low-aggressive beliefs measures among the measures of social problem-solving
subjects to choose an effective second-best solution, although skills, the measures of beliefs were first aggregated in terms of
these two groups did not differ in the effectiveness of their best their implied support for the use of aggression. Because the
solution. For females, antisocial-aggressive subjects were sig- measure of subjects' belief that victims deserve to be victimized
nificantly less likely than both high-aggressive and low-aggres- was the only measure that failed to discriminate among the
sive subjects (who did not differ significantly from each other) three aggression groups in the predicted direction, and because
to choose effective best and second-best solutions. Among low- it was the only measure of beliefs that failed to intercorrelate
aggressive subjects, males and females did not differ in the with the other measures of beliefs, this measure was not in-
effectiveness of their best and second-best solutions; however, cluded in the aggregated score.
among both the high-aggressive and the antisocial-aggressive
The analysis of the seven measures of problem-solving skills,
groups, males were significantly less likely than females to
together with the aggregated measure of beliefs supporting ag-
choose effective best and second-best solutions.
gression revealed the following hierarchy of measures with sig-
nificant discriminatory power: for information seeking, r2 =
Correlations Among Measures .33; for aggregated beliefs supporting aggression, partial r2 =
To assess the relation among the dependent measures of so- .21; for problem definition, partial r2 = . 16; for number of con-
cial problem-solving skills and beliefs regarding aggression, sequences, partial r2 = .11; and for effectiveness of second-best
586 RONALD G. SLABY AND NANCY G. GUERRA

Table 1
Imercorrelations Among Measures
Measure 1 10 12

Problem-solving skills
1. Problem definition
(perceived hostility) — .548 • • • _-.333*
Til*** _
.242*** .300*** -.459*** -.547*** .458*** .403*** .580*** -.038 (ns) .371*
2. Goal selection
(perceived hostility) — -.271*** -.241** -.244** -.376*** -.372*** .387*** .254** .372*** .060 (ns) .292***
3. N facts — .561*** .561*** .378*** .424*** -.386*** •.374*** -.413*** .014 (ru) -.218**
4. Absolutions — .714*** .383*** .531*** -.354*** .312*** -.409*** M6(ns) .274**
5. -Vconsequences .326*** -.360*** .046 (ns) -.231**
6. Effectiveness of 348*** 441*** —350***
"best" solution .097 (ns)
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

.380*** -.365*** .065 (ns)


7. Effectiveness of — .617*** -.203*
This document is copyrighted by the American Psychological Association or one of its allied publishers.

"2nd best" solution .361 •** -.485*** .036 (ns) .340***


_ -.446***
Beliefs supporting aggression
8. Legitimacy of
aggression .406*** .447*** .048 (ns) .301***
9. Increases self-
esteem — .408*** -.131 (ns) .116(ns)
10. Avoids a negative
image — -.038 (ns) .312***
11. Victimsdeserve
aggression .005 (ns)
12. Victims don't suffer

* / ; < . 0 5 . **/><.01. ***/>< .001.

solution, partial r2 = .06. The total variance accounted for by from both the high-aggressive and the low-aggressive high-
these five measures was 87°c. school students on the belief that victims deserve aggression,
the findings were in an unexpected direction: Violent juvenile
Discussion offenders were least likely to hold this belief.
The significance to be attached to this belief is unclear be-
Two sets of cognitive mediators were found to differentiate
cause it was the only measure that failed to intercorrelate with
among antisocial-aggressive, high-aggressive, and low-aggres-
any of the other measures of beliefs or skills. One possible expla-
sive adolescents: social problem-solving skills (all seven mea-
nation for the high level of agreement among high-school stu-
sures) and beliefs supporting aggression (all five measures). Al-
dents concerning the belief that victims deserve aggression de-
though almost all measures were intercorrelated, discriminant
rives from Lerner's (1970) "just world" hypothesis, whereby
function analyses revealed that many of these measures pro-
most people typically believe that good things happen to good
vided unique incremental sources of variation for differentiat-
people and bad things happen to bad people. It may be that
ing among the three aggression status groups. In general, high
violent juvenile offenders deviate from this normative belief.
levels of aggression were associated with low display of prob-
Furthermore, it may be that violent juvenile offenders have an
lem-solving skills as well as with high endorsement of beliefs
experiential basis for believing that individuals may become
supporting aggression. Although these findings are consistent
victims through arbitrary or uncontrollable circumstances.
with previous research in identifying social-cognitive factors
that differentiate aggressive and nonaggressive children (e.g., To the extent that endorsement of this belief can be viewed as
Bandura, 1986; Dodge, 1980; Perry et al., 1986; Richard & deviating from the norm, our findings can be taken as confirm-
Dodge, 1982;Spivack&Shure, 1974), this investigation further ing a more general form of our hypotheses. That is, low-aggres-
elaborates the social-cognitive development model in a number sive, high-aggressive, and antisocial-aggressive groups represent
of important ways. increasing levels of aggression that were consistently related to
First, this study extends the model based on the typical range both a decreasingly extensive display of social problem-solving
of aggressive behaviors shown by school children to that of vio- skills and an increasing endorsement of nonnormative beliefs
lent juvenile offenders who have committed such antisocial ag- concerning aggression. Although some may view juvenile
gressive acts as murder, rape, and assault. Violent juvenile offenders as a qualitatively distinct category of subjects on the
offenders differed from low-aggressive high-school students on basis of both the nature and extent of their antisocial aggression,
each of the cognitive skills and beliefs measures. In addition, these findings alternatively suggest that violent juvenile offend-
they differed from the high-aggressive high-school students on ers are linked to their nonoffender peers by a common set of
each of the skills measures except goal selection and on each of cognitive mediational variables that can be described as con-
the beliefs measures except the belief that victims of aggression tinua along which violent juvenile offenders score at one end
don't suffer. Whereas the violent juvenile offenders differed and low-aggressive high-school students score at the other end.
COGNITIVE MEDIATORS OF AGGRESSION 587

Investigation is warranted of both quantitative differences in ular categories of individuals, indicating prejudgments on the
cognitive mediational measures and qualitative differences in basis of such variables as gender, age, race, or physical size.
patterns of interrelationship among these and other measures Through investigation of the source of hostile perceptions of
in further comparing these groups to others such as youth en- others, the link between the development of aggression and the
gaged in various types of delinquency (see Hains& Ryan, 1983) development of prejudicial behavior might be more clearly es-
and individuals who are repeatedly victimized (see Olweus, tablished.
1978). Finally, our results provide preliminary support for the appli-
Second, by focusing on adolescents, the present study extends cation to the study of aggressive behavior of a more comprehen-
the developmental scope of previous aggression research that sive model of social-cognitive development that combines skills
has been guided by the emerging paradigm of social-cognitive in social in formation-processing with beliefs whose content
development. This previous research has focused almost exclu- offers general support for the use of aggression. Our results indi-
sively on preschool and elementary-school children. It has gen- cated that both the skills measures and the beliefs measures dis-
erally demonstrated more extensive differences in cognitive me- criminated significantly between the three aggression groups,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

diational patterns between aggressive and nonaggressive chil- and. together, these two sets of measures offered the best basis
This document is copyrighted by the American Psychological Association or one of its allied publishers.

dren in the later elementary years as compared with differences for discrimination among these groups. Although cognitive
found for younger children (e.g.. Perry et al., 1986; Richard & measures of skills and beliefs arise from different research tradi-
Dodge, 1982). Our findings with adolescents reveal an even tions (information processing vs. social learning), imply differ-
greater margin of difference for a wider range of cognitive medi- ent forms of cognitive mediation (process based vs. content
ators than has been previously reported, suggesting that, with based), and emphasize different levels of response to situational
development, individual patterns of cognition may become in- cues (specific vs. general), their separate conceptualization and
creasingly differentiated. assessment within a common theoretical framework permits
Third, our findings reveal a number of sex differences in the important questions to be raised concerning their separate and
cognitive factors associated with aggression. Across all aggres- interactive effects.
sion groups, males displayed less extensive skills in solving so- Indeed, the distinctions between process and content of cog-
cial problems (infiveof the seven measures) and they were more nitive mediation and between specific and general bases for re-
inclined to hold beliefs supporting the use of aggression (in sponding to social cues have not been clearly maintained in the
three of the five measures) than were females. On the other social-cognitive literature. For example, although Dodge's
hand, females were more likely than males to hold the belief that (1986) model focuses on social information-processing skills
victims deserve to be victimized, possibly indicating a greater applied to specific situational cues, the repeated finding that
adherence to the normative belief in a "just world" (Lerner, highly-aggressive children favor a hostile perception of others
1970). as was discussed earlier. There were no sex differences when processing ambiguous information raises questions re-
for two measures involving the quantity of responses generated garding the role of specific processes and general content. That
(number of solutions and number of consequences) nor for the is, it is unclear whether this hostile bias represents a deficit in
belief that aggression is a legitimate response. processing information whereby aggressive children merely as-
If males and females display equal facility in generating re- sume that others' intentions would be similar to their own, pre-
sponses and are equally likely to endorse the belief that aggres- sumably hostile, intentions, or whether this hostile bias reflects
sion is a legitimate response, why might males be more likely a more general set of "paranoid" beliefs about others' motiva-
than females to display aggression, as has been reliably found tions that might affect processing and interpretation of cues.
in previous research (see Parke & Slaby, 1983)? Our findings Although content and process may thus be conceptualized as
suggest that one important difference may be in the way males somewhat distinct components of cognition, they clearly inter-
and females formulate social problems and anticipate out- act. Information processing implies that certain content is be-
comes. Males were more likely than females to base their prob- ing processed. For example, in generating consequences, an in-
lem definition and goal selection on the perception that the dividual must generate or draw from an existing repertoire of
other was a hostilely motivated adversary, to seek less informa- available responses. If he or she holds the belief, for instance,
tion that might permit disconfirmation of this perception, and that heightened self-esteem is a consequence of aggression, this
to choose best and second-best solutions that were not effective. response might be available for consideration during the re-
Thus, males and females may have been solving qualitatively sponse-evaluation step of information processing. However, this
different problems. particular belief might also directly lead the individual to seek
This sex difference in hostile perception is in need of further out opportunities to aggress. Beliefs may serve either as guides
investigation, because neither its source nor its range of target to the specific information to be processed or as direct guides
individuals was addressed in this study. Because subjects were to action (e.g., Sigel, 1986).
presented with stories involving only target individuals of the Because the skills and beliefs measures of the present study
same sex as themselves, it is unclear to what extent the observed provide only a loosely formulated way of addressing these con-
sex differences in problem-solving skills may have been due to ceptual distinctions within a common framework, more work is
the sex of the target individual, with male targets generally per- needed in the ways in which these cognitive mediational factors
ceived as more likely to be hostile adversaries than female tar- interact. Rather than limiting the assessment of beliefs to their
gets, as previous research has suggested (Barrett, 1979; Perry et endorsement or nonendorsement, future studies might assess
al., 1986; Smith & Green, 1974). Research is warranted on the both the strength of various beliefs and the salience of those
extent to which individuals' social perceptions of hostile adver- beliefs in the problem-solving process. For example, do individ-
saries are applied generally to all others or specifically to partic- uals spontaneously generate or fail to generate particular conse-
588 RONALD G. SLABY AND NANCY G. GUERRA

quences as a function of the strength of their beliefs about those processes among juvenile delinquents and nondelinquent peers.
consequences? Do individuals with more strongly held beliefs Child Development, 54, 1536-1544.
process information less extensively or even circumvent the Huesmann. L. R.( Eron, L. D., Lefkowitz, M. M., & Walder, L. O.
processing of information, particularly in highly arousing situa- (1984). Stability of aggression over time and generations. Develop-
tions? Do individuals with an impulsive cognitive style default mental Psychology. 20. J120-1134.
more frequently in the processing of specific information when Lerner. M. J. (1970). The desire for justice and reaction to victims. In
J. Macaulay & L. Berkowitz (Eds.), Altruism and helping behavior
confronting social problems in favor of using more general so-
(pp. 205-229). New \brk: Academic Press.
cial beliefs as guides to their actions?
Marsh, D. T. (1982). The development of interpersonal problem-solving
Further research is also needed on the relative stability and among elementary-school children. Journal of Genetic Psychology,
changeability of cognitive mediational factors associated with M0. 107-118.
aggressive and violent behavior. The fact that aggression in Nasby, W, Hayden, B., & dePaulo, B. M. (1980). Attribulional bias
childhood is a strong predictor of aggressive and antisocial be- among aggressive boys to interpret unambiguous social stimuli as dis-
haviors in adulthood, including violent criminal offenses, traffic plays of hostility. Journal of Abnormal Psychology, 89, 459-468.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Olweus, D. (1978). Aggression in the schools. New York: Wiley.


This document is copyrighted by the American Psychological Association or one of its allied publishers.

violations, spouse abuse, and a tendency toward severe punish-


ment of one's own children (Huesmann. Eron, Lefkowitz, & Olweus, D. (1984). Development of stable aggressive reaction patterns
in males. In R. Blanchard & C. Blanchard (Eds.). Advances in the
Walder, 1984), suggests the possibility that central organizing
study ofaggression (Vol. I.pp. 103-137). New \brk: Academic Press.
factors, if left unaddressed, may play a role in the maintenance
Parke, R. D.. & Slaby. R. G. (1983). The development of aggression. In
and escalation of aggression over time. Our data, although lim- P. H. Mussen (Series Ed.) and E. M. Hetherington (Vol. Ed.), Hand-
ited to a one-time assessment of adolescents, raise the possibil- hook of child psychology. Vol. 4. Socialization, Personality, andSocial
ity that cognitive factors, representing habitual patterns of cog- Development (pp. 547-641). New York: Wiley.
nitive mediation that underly aggression, may serve to differen- Perry, D. G.. Perry, L. C , & Rasmussen, P. (1986). Cognitive social
tiate and stabilize an individual's use of aggression in particular learning mediators of aggression. Child Development. 57, 700-71 1.
situations. Nevertheless, these emerging patterns of cognitive Richard, B.. & Dodge, K. A. (1982). Social maladjustment and prob-
mediation are seen as potentially changeable through interven- lem-solving in school -aged children. Journal of Consulting and Clini-
tions designed to build social problem-solving skills and to cal Psychology. 42, 787-793.
modify beliefs related to the use of aggression. Sharp, K. C. (1981). Impact of interpersonal problem-solving training
on preschoolers' social competency. Journal of Applied Developmen-
tal Psychology, 2, 175-187.
References Short, J. R, Jr. (Ed.). (1968). Gang delinquency and delinquent subcul-
tures. New York: Harper & Row.
Band lira, A. (1973). Aggression: A social learning analysis. Englewood Shure, M. B.. Newman, S.. & Silver, S. (1973, March). Problem-solving
Cliffs, NJ: Prentice-HaU. thinking among adjusted, impulsive, and inhibited Head Start chil-
Bandura, A. (1986). Social foundations of thought and action. Engle- dren. Paper presented at a meeting of the Eastern Psychological Asso-
wood Cliffs, NJ: Prentice-Hall. ciation, Washington, DC.
Barrett, D. E. (1979). A naturalistic study of sex differences in chil- Shure, M. B., & Spivack, G. (1976). Problem-solving techniques in
dren's aggression. Merrill-Palmer Quarterly, 25. 193-203. childrearing. San Francisco, CA: Jossey-Bass.
Bornstein, M.. Bellack, A. S.. & Hersen, M. (1980). Social-skills training Sigel, I. E. (1986). Reflections on the belief-behavior connection: Les-
for highly aggressive children. Behavior Modification, 4, 173-186. sons learned from a research program on parental belief systems and
Dodge, K. A. (1980). Social cognition and children's aggressive behav- teaching strategies. In R. D. Ashmore & D. M. Brozdinsky (Eds.),
ior. Child Development, 51, 162-170. Thinking about the family: Views of parents and children (pp. 3 5 -
Dodge, K. A. (1986). A social information-processing model of social 65). Hillsdale. NJ: Erlbaum.
competence in children. In M. Perlmutter (Ed.), Minnesota sympo- Slaby, R. G., & Roedell, W. C (1982). Development and regulation of
sium on child psychology (Vol. 18, pp. 77-125). Hillsdale, NJ: Erl- aggression in young children. In J. Worell (Ed.), Psychological devel-
baum. opment in the elementary years (pp. 97-149). New York: Academic
Dodge, K. A., & Frame, C. L. (1982). Social cognitive biases and deficits Press.
in aggressive boys. Child Development. 53. 629-635. Smith, P. K., & Green, M. (1974). Aggressive behavior in English nurs-
Dodge. K. A.. & Newman, J. P. (1981). Biased decision-making pro- eries and playgroups. Sex differences and responses of adults. Child
cesses in aggressive boys. Journal of Abnormal Psychology. 90, 375- Development, 45, 211-214.
379. Spivack, G.. & Shure. M. B. (1974). Social adjustment of young chil-
Eron. L. (1987). The development of aggressive behavior from the per- dren: A cognitive approach to solving real-life problems. Washington,
spective of a developing behaviorism. American Psirhologist. 42, DC: Jossey-Bass.
435-442. Sykes, G., & Matza. D. (1957). Techniques of neutralization: A theory
Eron, L. D., Walder, L. Q , & Lefkowitz, M. M. (1971). The learning of of delinquency. American Sociological Review, 22, 664-670.
aggression in children. Boston: Little, Brown.
Guerra, N. (1983). Social-cognitive deficits in aggressive boys: A prelim- Received June 12, 1987
inary study. Unpublished manuscript. Harvard University. Revision received January 15, 1988
Hains, A. A., &Ryan, E. B. (1983). The development of social cognitive Accepted January 16, 1988 •

You might also like