Professional Documents
Culture Documents
Abstract
Using conflict narratives reported by adolescents in grade 7 (mean age = 13.4 years),
this study investigated the interactional properties and developmental functions of four
types of aggressive behaviors: social aggression, direct relational aggression, physical aggression, and verbal aggression. A total of 475 participants from the Carolina
Longitudinal Study (Cairns & Cairns, 1994) were included. Results showed that the
majority of conflict interactions involved more than a dyad. The use of social aggression (e.g., concealed social attack) was associated with more individuals involved in
the conflict. Social aggression was primarily reported as an initiating behavior for
interpersonal conflicts, while direct relational aggression was a responding behavior.
Medium to high levels of reciprocity were found for physical, verbal, and direct relational aggression, whereas a low level of reciprocity was reported for social aggression. School authorities were most likely to intervene in physical aggression. The use
of social aggression was associated with higher network centrality among adolescents.
Developmental maladjustment in late adolescence and early adulthood was primarily predicted by physical aggression.
Keywords: social aggression; confrontational aggression; interpersonal conflict; social
networks; developmental prediction
Aggression is not a unitary construct, and there exist different forms. Some aggressive behaviors are prototypic, such as physical aggression and verbal aggression.
Others are less prototypic, but have the potential to produce pain. These include hurtful
acts that are less open, more disguised, indirect, and manipulative. Robert Sears and
his colleagues (Sears, 1961; Sears, Rau, & Alpert, 1965) first identified prosocial
aggression in contrast to antisocial aggression. Later Norma Feshbach (1969) used
indirect aggression to refer to subtly aggressive behaviors such as ignoring and
excluding another person. More recently, different terms have been used to label subtle
Correspondence concerning this article should be addressed to Hongling Xie, Center for Developmental
Science, 100 East Franklin Street, CB#8115, University of North Carolina at Chapel Hill, Chapel Hill,
NC 275998115, USA. Email: hlxie@email.unc.edu
Blackwell Publishers Ltd. 2002. Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street,
Malden, MA 02148, USA.
206
207
damage may be largely delayed in time and transformed in context. As a result, the
chances of revenge by the victim and punishment by an authority may be largely
reduced.
Second, the effectiveness of non-confrontational aggression may depend on the use
of social networks as a vehicle for attack. The damage to the victim is not automatic
and requires a process of mediation through other individuals in the social community. For a social attack to be effective, the perpetrator has to engage the participation
and help of other individuals within the social network. In contrast, the effectiveness
of confrontational aggression is not dependent upon the mediation of other individuals in the social network.
Third, the use of non-confrontational aggression may require the processing and
utilization of more complicated social information than confrontational strategies. In
order for a social attack to be effective, a person needs accurate knowledge of the
networks of interpersonal relationships and subtle skills of manipulation. It was found
that high levels of indirect aggression were associated with high levels of social intelligence, while confrontational aggression (i.e., physical and verbal aggression) was
not (Kaukiainen, Bjorkqvist, Lagerspetz, et al., 1999).
The Current Study
It is important to understand how these distinctive forms of interpersonal aggression
serve different social functions (Xie, Cairns, & Cairns, in press). In this study, we differentiate two types of subtle aggression: social aggression that employs nonconfrontational strategies and direct relational aggression that uses confrontational
strategies. Direct relational aggression in this study refers to actions whereby interpersonal damage is achieved by confrontational strategies that involve the relationship between two persons. It includes not talking to or hanging around with someone,
deliberately ignoring someone, threatening to withdraw friendship or affection, and
excluding someone from a group activity by telling him/her directly that he or she is
not welcome. Social aggression includes gossiping, social exclusion, social isolation,
social alienation, writing notes to a third party, talking about someone behind his/her
back, stealing friends or romantic partners, the triangulation of friendship or romantic relationship, and telling secrets/betrayal of trust. These two subtle forms of aggression are studied along with two prototypic forms: physical aggression and verbal
aggression.
A general issue that concerned this study is whether different forms of aggression
serve different functions in development and in social interactions. Recent investigations (e.g., Crick, 1997; Kaukiainen et al., 1999) have focused on individual differences of aggression and the association between different forms of aggression and
concurrent social-psychological adjustment (e.g., peer rejection, externalizing and
internalizing problems, and social intelligence). Little is known about the differential
properties of different aggressive behaviors in social interactions and their long-term
developmental consequences (see also Underwood, Galen, & Paquette, 2001b). In this
study we employ a dual approach. On the basis of narrative reports of interpersonal
conflicts, we conduct a micro-analysis to examine the interactive properties of different aggressive behaviors in social interactions. On the macro-level, we assess individual differences to identify (1) developmental correlates in other domains that are
related to different aggressive behaviors (e.g., academic competence and popularity),
and (2) the linkage between aggressive behaviors and developmental maladjustment
Blackwell Publishers Ltd. 2002
208
in late adolescence and early adulthood (i.e., school dropout, teen parenthood, and
criminal arrest).
Given the non-confrontational nature of social aggression, we expect a low level of
reciprocity (i.e., social aggression begets social aggression). In contrast, medium to
high levels of reciprocity are expected for physical aggression, verbal aggression, and
direct relational aggression. Further, conflicts involving social aggression would have
more participants than conflicts involving only confrontational aggression (e.g., physical aggression, verbal aggression, and direct relational aggression). Given the necessity for an individual to draw other persons participation in order to complete an
effective social attack, we expect that the use of social aggression is associated with
higher levels of social network centrality. On the basis of previous research (e.g.,
Kaukiainen et al., 1999; Pulkkinen, 1992; Underwood, Galen, & Paquette, 2001a), we
expect that different forms of aggression would have different developmental correlates (e.g., popularity and academic competence) and show different levels of robustness in predicting subsequent developmental outcomes such as school dropout.
Specifically, reliable prediction will be primarily yielded by physical aggression.
Narrative reports of interpersonal conflicts with peers are utilized to assess the interactional and developmental properties of different aggressive behaviors. These reports
represent an individuals accounts or perceptions of specific conflicts and aggressive
behaviors that are not directly observed by an investigator. We speculate that a primary
bias in these accounts may be the self-enhancement effect (e.g., Alicke, Koltz,
Breitenbecher et al.,1995; Brown & Dutton, 1995; Cairns & Cairns, 1988; Xie,
Mahoney, & Cairns, 1999). Participants may show a tendency to blame others and to
report negative behaviors of other individuals rather than their own. Given this consideration, we first examine the characteristics of narrative accounts, and then determine whether similar interactional patterns were reported for the self and others.
Method
Participants and Design
This study involves 475 seventh-grade participants (248 girls and 227 boys; Mean age
= 13.4 years, SD = .58) from the Carolina Longitudinal Study (Cairns & Cairns, 1994).
The mean socioeconomic status on the Duncan Scale (Stevens & Featherman, 1981)
is 31.6 (SD = 17.8), and the full range of occupations is represented in the sample.
The proportion of African-American participants in this study is 30% (141/475). Participants were recruited from three middle schools from the suburban and rural areas
in the mid-Atlantic United States. All students in the designated grade in each school
were invited to participate in the study. A signed consent was required from both parent
and student with an average participation rate of 68% (475/698). Comparison between
participants of this study and students of the entire grades revealed no participation
bias in terms of sex, race, and extreme cases of aggressive behaviors.
Semi-structured individual interviews were conducted annually through grade 12.
An effort was made to individually interview all original participants each year, regardless of whether they were in school or had dropped out (see Cairns & Cairns, 1994).
More than 98% of the participants were interviewed by the end of high school. For
the purpose of this investigation, aggressive behaviors were identified in interpersonal
conflicts reported by participants when they were in the first year of the study (i.e.,
the 7th grade).
Blackwell Publishers Ltd. 2002
209
Measures
Interpersonal Conflicts and Aggressive Strategies. In the course of a semi-structured
individual interview, participants were asked to (1) identify both the same-sex peer(s)
and the opposite-sex peer(s) who bothered them or caused them trouble, and (2)
describe the conflicts in detail. The following probes were used to help clarify the
narratives: How did it start?, What happened?, What did you do?, How did it
end?, Did anything else come of it?, How did you feel about it?, and How did
the other person(s) feel about it? About 10% (48/475) of participants denied any conflict with other person(s), 36% (172/475) reported only one conflict. The audio-taped
interviews were first transcribed verbatim by the research staff in the CLS project.
These transcripts were then coded during the course of this study.
The narrative accounts of each conflict were coded in terms of (1) the identity
of the other persons involved; (2) the participants attribution of responsibility for
conflict initiation (self, other, or both); (3) the total number of individuals involved;
(4) the participants aggressive behaviors in the conflict (social aggression, hostile
physical aggression, hostile verbal aggression, direct relational aggression, and
other negative behaviors); (5) the other persons aggressive behaviors in the conflict
(See Appendix A for the operationalization of these aggressive behaviors); and
(6) intervention and/or punishment by the school authorities (1intervened, 0not
intervened). All reported aggressive behaviors were coded. High levels of reliability
were established. A total of 50 interviews were independently coded by two research
assistants. Inter-coder agreement ranged from 89% to 99% (Kappa: .56 to .91 with
median = .84).
Aggressive Exchange in Conflict Interactions. In addition to the aggressive strategies,
we also coded the interactive sequence and responsive patterns of behaviors. For each
aggressive behavior, we coded the perpetrator, the victim, whether it is an initiating
behavior or a response to a previous attack, and the reactive behavior from the victim.
For socially aggressive behavior, we also coded (1) whether the identity of the perpetrator was revealed to the victim, and (2) the types of aggression generated between
the victim and a third party. In each conflict, up to two actions were included for each
type of aggression (social, physical, verbal, and direct relational). This operation
accounted for 96% of all aggressive acts. Another 50 interviews were independently
coded by two research assistants. Inter-coder agreement ranged from 82% to 96%
(Kappa: .45 to .87 with median = .75). Since parallel patterns were obtained across
the two conflicts, they were combined for analyses.
Peer-narrative Measures of Aggression. Scores on the peer-narrative measure of social
aggression were computed according to the number of times a child was identified
using social aggression strategies in the conflict narratives reported by their peers in
the same school. The measure yielded highly skewed distributions. About 84%
(397/475) of the participants scored 0, and only 4% (19/475) of the participants
received scores greater than 1. Such highly skewed distributions presented problems
in subsequent parametric analyses. A decision was made to treat the peer-narrative
measure of social aggression as a dichotomous variable (0receiving no report on
socially aggressive behaviors, 1receiving one or more reports).
Treating the peer-narrative measure as a dichotomous variable also permitted the
avoidance of standardizing scores on the peer-narrative measure in order to adjust for
Blackwell Publishers Ltd. 2002
210
211
212
tion on teen parenthood status was available for 460 participants by the time of the
most recent data collection. Five deaths occurred before age 20. Ten participants (3
females and 7 males) were not interviewed after high school, resulting in a recovery
rate of 97.9% (460/470). These original participants were not teen parents when we
last saw them. More teen mothers (21%, 50/243) were found than teen fathers (9%,
19/217; c2(1,460) = 12.56, p < .001).
Criminal Arrest. When the participants were in their early 20s, information on
criminal arrests was obtained from the State Bureau of Investigation (SBI) for 95%
(453/475) of the participants who stayed in the original state. No arrest had occurred
for 89% of the participants, and 8.4% had been arrested more than once. The measure
for criminal arrest was then transformed into a dichotomous variable (0not being
arrested; 1being arrested at least once). A higher proportion of criminal arrests
occurred among males (16%, 35/213) than among females (6%, 15/240; c2(1,453) =
11.91, p < .001).
Results
Characteristics of Narrative Accounts
Whos to Blame? Not Me! There was a strong tendency for the participants to blame
others for the conflict. Among all the conflicts in which attributions were made, the
vast majority (93.6%, 627/670) was blamed on others (i.e., the opponent and/or the
third party). Only 1.5% (10/670) was solely blamed on the self. Joint responsibility
was claimed for a small proportion (4.9%, 33/670) of the conflicts.
Differential Reports of Aggressive Behaviors. The likelihood of a participant reporting others aggressive behaviors was compared with the likelihood of reporting his/her
own in the conflicts. The unit of analysis was the individual participant. As shown in
Table 1, 32% of participants solely reported others verbal aggression, and 5% solely
reported their own. The difference was highly significant (McNemars test Qm = 84.78,
p < .0001).1 About a quarter of the participants (23%) reported both their own and
Aggressive behaviors
Physical aggression
Verbal aggression
Social aggression
Direct relational aggression
Self
only
Other
only
Both self
and other
No
report
.08 (35)
.05 (20)
.01 (6)
.12 (52)
.10 (43)
.32 (136)
.24 (102)
.06 (25)
.27 (114)
.23 (98)
.06 (26)
.09 (40)
.55 (235)
.41 (173)
.69 (293)
.73 (310)
Note: Forty-eight participants denied any conflict, resulting in a total n of 427 for each type of
aggression.
Blackwell Publishers Ltd. 2002
213
other persons verbal aggression. A similar pattern of differential reports was observed
for social aggression (24% other vs. 1% self; McNemars test Qm = 83.56, p < .0001).
In contrast, participants were more likely to report direct relational aggression of their
own (12%) than to report others (6%; McNemars test Qm = 8.78, p < .01). No differential reports were found for physical aggression (McNemars test Qm = .63, n.s.).
Differential Properties of Aggressive Behaviors in Social Interactions
Beyond a Dyad. On average, three persons (mean = 3.03) were mentioned as being
actively involved in the initiation, escalation, or termination of an interpersonal conflict. Traditional models of dyadic social interaction seem to have limited generalization. Only 35% (244/693) of these reported conflicts involved a dyad, 41% (283/693)
involved three individuals, and 24% involved four or more individuals. In addition,
the use of non-confrontational behavior of social aggression was associated with more
individuals being involved in the conflict [Mantel-Haenszel test c2(1,691) = 86.94, p <
.001]. Specifically, four or more persons were reported in 53% (74/140) of conflicts
where social aggression was used (SA-conflicts), while only in 17% (91/551) of conflicts where social aggression was not used (non-SA-conflicts). The majority of nonSA-conflicts involved two (42%, 229/551) or three individuals (42%, 231/551). In
contrast, only 11% (15/140) of SA-conflicts involved two individuals as reported by
the participants.2
Reciprocity of Aggressive Interactions. The proportions of different reactive behaviors
towards a specific aggressive action were summarized in Table 2. When insufficient
information was available for coding a specific response behavior, the behavioral
exchange was excluded from the following analyses. In approximately 9.3% of socially
Social
Physical
Verbal
Direct relational
.30
.05
.05
.24
.09
.04
.03
.20
.01
132
.25
.00
.48
.03
.01
.06
.01
.14
.02
289
.15
.01
.13
.34
.08
.10
.01
.16
.02
358
.43
.02
.02
.06
.32
.01
.03
.08
.03
142)
214
aggressive behaviors, the victim was not able to identify the perpetrator. These behavioral exchanges were also excluded from the analysis. It was reasoned that when a
victim did not know who the perpetrator was, no reactive action could have been taken
against the perpetrator.
The hypothesis of aggression begets aggression was tested (Table 1). The probability that social aggression begets social aggression was 5%. In contrast, the probability that physical aggression begets physical aggression was 48%. Medium levels
of reciprocity were found for verbal aggression (34%) and direct relational aggression
(32%). No significant difference was found between verbal and direct relational
aggression [c2(1,500) = .18, p > .60]. Significant differences were found for all other
pairs of comparisons on reciprocity (c2s > 9.58, ps < .01).
It should be noted that low reciprocity for socially aggressive behavior did not imply
that social aggression failed to elicit hostile aggressive responses. Verbal aggression
and physical aggression accounted for 29% of the reactions of the victim towards the
perpetrator. In addition, 22% (38/172) of socially aggressive acts led to physical and/or
verbal aggression between the victim and a third party not including the perpetrator,
a unique feature of social aggression.
Differential Reciprocity. Since children reported different levels of aggressive behavior for themselves and others, it is important to examine the differences in the
reported reciprocity as a function of self-initiated attack and other-initiated attack.
No systematic self-other bias on the reciprocity was obtained across different forms
of aggression. Both within-subject analyses and between-subject analyses revealed
no significant self-other differences for social and direct relational aggression. The
reported mean reciprocity of social aggression was 4% (SD = .20, n = 93) for the participants and 10% (SD = .31, n = 30) for others (within-subject F(1,14) = .32, p = .58;
between-subject F(1,92) = 1.73, p = .19). On average, participants reciprocated 43%
(SD = .49, n = 42) of direct relational attacks and others reciprocated 31% (SD = .45,
n = 80; within-subject F(1,11) = .02, p = .88; between-subject F(1,97) = 1.67, p = .20).
When participants reported both their own and others use of physical aggression,
a higher level of reciprocity was found for the self (.85 vs. .37; within-subject F(1,69)
= 38.24, p < .001). When participants only reported either their own or others use of
physical aggression, no significant self-other difference was found on the reciprocity
(.34 vs. .40; between-subject F(1,106) = .40, p = .53). In contrast, no significant difference in reciprocity was found when participants reported both their own and others
use of verbal aggression (.55 vs. .41; within-subject F(1,57) = 1.73, p = .19), while a
higher level of reciprocity was reported for others in between-subject tests (.18 vs.
.50; F(1,170) = 20.27, p < .001).
Whether it was aggressive behaviors of the self or of others, a consistent pattern of
difference was observed across different forms. Very low levels of reciprocity were
found for non-confrontational behaviors of social aggression, while moderate to high
levels of reciprocity were found for confrontational behaviors of physical, verbal, and
direct relational aggression.
Features of Provocation and Response. The likelihood of aggressive actions being
reported as an initiating behavior or a response to an attack was examined. Since the
four types of aggressive behavior differ in their reciprocity, only the first action of
each type in a conflict was included in the analyses. Eighty percent (80%, 78/98) of
the time, socially aggressive behaviors were reported as an initiating act for interper Blackwell Publishers Ltd. 2002
215
sonal conflicts, and only 20% of the time were they reported as a response to a provocation. In contrast, direct relational aggression was rarely reported as an initiating
behavior (14%, 11/81) but primarily reported as a response to negative interchanges
(86%). The difference was highly significant (c2(1,179) = 77.30, p < .0001). A similar
pattern was observed for physical and verbal aggression. Sixty-one percent (109/178)
of physical aggression and 54% (113/210) of verbal aggression were reported at the
beginning of a conflict (c2(1,388) = 2.17, p = .14). Both social and direct relational
aggression differ from physical and verbal aggression (c2s > 9.74, ps < .01).
Intervention and/or Punishment by School Authorities. School authorities were most
likely to intervene in a conflict where physical aggression was used (42%, 93/223).
Intervention or punishment by school authorities was reported in about one-third
(100/312) of conflicts where verbal aggression was used. The lowest likelihood of
intervention by authorities occurred in conflicts where social aggression and/or direct
relational aggression were used [23% (35/151) for social aggression, and 20%
(25/128) for direct relational aggression]. Statistical comparisons indicated significant
differences between physical aggression and all other types of aggression (c2s > 7.55,
ps < .01).3 School authorities were more likely to intervene in verbal aggression than
direct relational aggression (c2(1,501) = 9.21, p < .01). No significant difference was
found between social and direct relational aggression (c2(1,192) = 0.39, p > .50), and
between social and verbal aggression (c2(1,292) = 3.50, p = .06).
Aggressive Behaviors and Social Network Centrality
Gender Differences in Aggressive Behaviors. Individual differences in the use of
aggressive behaviors were assessed by peer-narrative measures. More girls used social
aggression and direct relational aggression than boys (.23 Vs .09 for social aggression, and .17 Vs .04 for direct relational aggression, F(1,473)s > 6.79, ps < .001). In contrast, more boys used physical aggression than girls (.19 Vs .09, F(1,473) = 10.17, p <
.01). No gender difference was found for verbal aggression (.24 for boys and .27 for
girls, F(1,473) = .82, p > .30).
Social Network Centrality. The relations between a persons social network centrality
and his/her use of different aggressive behaviors were depicted in Figure 1. Analysis
of variance was performed where the peer-narrative measure of aggression was the
dependent variable and network centrality and gender were independent variables.
Social aggression was associated with higher levels of network centrality (F(2,469) =
4.09, p < .05). This association did not differ by gender (centrality X gender interaction: F(2,469) = .023, p > .80). None of the other aggressive behaviors (i.e., verbal, physical, and direct relational aggression) was significantly related to network centrality
(F(2,469)s < 1.97, ps > .10).
Different Types of Aggression and Developmental Adjustment
Relations Among Different Aggressive Behaviors. The correlations among peernarrative measures of different aggressive behaviors are summarized in Table 3.
Moderate levels of correlation were found between physical aggression and verbal
aggression (r = .35), and between verbal aggression and social aggression (r = .35).
Low levels of correlation were obtained between physical aggression and direct rela Blackwell Publishers Ltd. 2002
216
.40
Physical aggression
Verbal aggression
.30
.30
.20
.20
.10
.10
.40
.40
Social aggression
.30
.30
.20
.20
.10
.10
tional aggression (r = .10) and social aggression (r = .15). All other correlations were
modest. It should be noted that such patterns of correlation, to some extent, reflect the
co-occurrence of aggressive behaviors within an interpersonal conflict, since peernarrative measures of aggression were derived from narrative reports and often multiple aggressive behaviors were reported.
Aggression and Individual Adjustment. The relations between different aggressive
behaviors and teacher-ratings of concurrent adjustment were summarized in Table 4.
Given the inter-correlations among different aggressive behaviors, partial correlation
coefficients were calculated after controlling for gender and the other aggressive
behaviors. The peer-narrative measure of physical aggression was significantly correlated with most of the adjustment scores of teacher-ratings: high aggression, low
Blackwell Publishers Ltd. 2002
217
Table 3. Intercorrelations Among Different Aggressive Behaviors after Controlling for Gender
Aggressive behaviors
Physical aggression
Verbal aggression
Social aggression
Direct relational aggression
Physical
Verbal
Social
Direct relational
1.00
.35***
.15**
.10*
1.00
.35***
.25***
1.00
.27***
1.00
Physical
Verbal
Social
Direct relational
Concurrent adjustment
ICST-Aggression
ICST-Academic competence
ICST-Popularity
ICST-Olympian
ICST-Affiliation
.16***
-.19***
-.14**
-.06
-.17**
.21***
-.11*
-.05
-.11*
-.06
.03
.01
.02
.06
-.02
-.02
.00
.10*
.11*
.13**
.20***
.03
.20***
.08
.11*
.00
.06
.05
.08
Future adjustment
School dropout
Teen parenthood
Criminal arrest
-.09
-.09
-.03
academic competence, low popularity, and low affiliation. The use of verbal aggression was related to higher teacher-ratings of aggression, lower scores of academic
competence, and lower scores of Olympian. After controlling for the effects of gender
and the other aggressive behaviors, social aggression alone was not related to any of
the teacher-ratings. In contrast, the use of direct relational aggression was associated
with higher scores of popularity, Olympian, and affiliation.
Three indices were employed for subsequent maladjustment: school dropout, teen
parenthood, and criminal arrest. Partial correlation coefficients were calculated after
controlling for gender and other aggressive behaviors. The use of physical aggression
significantly increased the likelihood of school dropout and criminal arrest (Table 4).
Verbal aggression was associated with a higher risk for teen parenthood. The use of
Blackwell Publishers Ltd. 2002
218
social aggression or direct relational aggression, however, did not increase the risk of
maladjustment in late adolescence and early adulthood.
Discussion
In this study we found that negative social interactions among adolescents often
involved more than a dyad. This finding is consistent with observational studies of
aggressive behaviors in natural settings (e.g., Pepler & Craig, 1995). Even more individuals were involved in negative exchanges when disguised social aggression was
used. If it takes two to fight, it takes at least three to gossip. The findings suggest that
laboratory observations of dyadic interactions may have limited generalizations to
natural settings. Furthermore, they suggest that broader social networks beyond dyadic
friendships should be assessed in order to understand the operation and social functions of aggressive behaviors.
This is not limited to social aggression, however. Once the attack is detected and
the identity was revealed, social aggression often resulted in open confrontations of
physical and/or verbal aggression between the perpetrator and the victim (29%). It
also led to physical and/or verbal aggression between the victim and a third party not
involving the perpetrator (22%). Some of the physical fights between two adolescents
were actually set up by other adolescents using social aggression (see also Artz, in
press). One seventh-grade male described how he was set up by other students in the
school to fight with another male:
. . . See people, they like to see a good fight and you know, they just like to see a fight
and theyll say like, he talked about your mother and all stuff like that . . .
Despite the fact that social aggression had a good chance of leading to physical
aggression, school authorities primarily focused their effort on intervening when physical aggression occurred. No doubt, it is hard for the authorities to identify the subtle
exchanges of social aggression, and physical aggression is more destructive and disruptive. Recognizing the early signs of social aggression among students, however, may
enable the authorities to intervene at an earlier stage and prevent disruptive confrontations later on.
Over 9% of the time, victims of socially aggressive attacks were not able to identify the perpetrator. Although this proportion may not appear high, it should be remembered that there were likely incidents of social aggression in which the victim did
not even know of the attack and hence failed to report it. In contrast, the perpetrators
identity was revealed in every episode of confrontational aggression (i.e., direct relational, physical, and verbal). Social aggression also generated low levels of immediate reciprocity even when the victim knew who the perpetrator was. These findings
confirmed previous speculations about the adaptive features of socially aggressive
behavior (Bjrkqvist, Lagerspetz, & Kaukiainen, 1992; Cairns & Cairns, 1994).
Namely, the use of social aggression enables the perpetrator to obscure his/her identity and avoid immediate escalation and revenge.
Cara found a rumor was going around about her stealing money from her mom; when
asked about the person who did it, Cara replied: They never did find out, . . they said
that Jenny said it, and Jenny said that Diane said it, but then um then, Jenny said that
Amy said it, then Amy said that Angie said, and Angie said that Amy said, and Amy said
that Jenny said it, so we never did find out who said it . . .
The use of aggression did not exclude adolescents from peer social networks. On
the contrary, high centrality in peer networks may enable the adolescents to use more
Blackwell Publishers Ltd. 2002
219
220
221
222
223
Xie, H., Cairns, B. D., & Cairns, R. B. (in press). The development of aggressive behaviors
among girls: Measurement issues, social functions, and differential trajectories. In D. Pepler,
K. Madsen, C. Webster, & K. Levene (Eds.), Development and Treatment of Girlhood Aggression. Mahwah, NJ: Lawrence Erlbaum Associates.
Xie, H., Cairns, R. B., & Cairns, B. D. (1999). Social networks and configurations in innercity schools: Aggression, popularity, and implications for students with EBD. Journal of
Emotional and Behavioral Disorders, 7, 147155.
Xie, H., Mahoney, J. L., & Cairns, R. B. (1999). Through a looking glass or a hall of mirrors?
Self ratings and teacher ratings of academic competence over development. International
Journal of Behavioral Development, 23, 163183.
Notes
1. The null hypothesis tested was that there was an equal probability to report others aggressive
behavior as to report ones own (r0 = .50). The McNemar statistics were calculated, which had a chisquare distribution with 1 degree of freedom (see Koch & Edwards, 1988; also McNemar, 1962).
2. It should be noted that the use of social aggression implies the participation of a third person in completing the attack. In this analysis, we calculated the number of individuals who were explicitly reported
by the participants. In some of the conflicts involving social aggression, participants did not explicitly
mention a third person, although it was implied.
3. Pairwise comparisons were made excluding the conflicts containing both types of aggression. For
instance, the comparison between social and physical aggression on their likelihood of intervention by
authorities was conducted excluding conflicts containing both social and physical aggressive behaviors.
Acknowledgments
Support for this research was provided by Spencer Foundation and NIMH (MH61293,
MH45532, and MH52429). We thank Thomas W. Farmer and Kathryn E. Hood for their
constructive feedback, and Laura L. Bobis and Anne B. Matthews for their help in data coding.
We also appreciate the comments and suggestions made by the anonymous reviewers and
co-editor Hildy Ross.
224
Physical Aggression
Confrontational, use of physical force or weapon as a way to attack, non-playful acts,
and the identity of the perpetrator is directly present.
Examples. Beating up someone, a bad fight, throwing a chair at someone, jumping
on someone, getting in a fight, slap, swing, hit, push, and threatening to fight.
Verbal Aggression
Confrontational, use of negative words as a way to attack, non-playful verbal
exchange, and the identity of the perpetrator is directly present.
Examples. Verbal insult, hostile threat, verbal humiliation, curse, yell, argument, fuss,
name calling, questioning, blaming, and hostile teasing.
This document is a scanned copy of a printed document. No warranty is given about the accuracy of the copy.
Users should refer to the original published version of the material.