<%BANNER%>

Explaining Discrepancies between Self and Peer Reports of Aggression in Adolescence

Permanent Link: http://ufdc.ufl.edu/UFE0041879/00001

Material Information

Title: Explaining Discrepancies between Self and Peer Reports of Aggression in Adolescence
Physical Description: 1 online resource (93 p.)
Language: english
Creator: Clemans, Katherine
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: Psychology -- Dissertations, Academic -- UF
Genre: Psychology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Though most studies of aggressive behavior in early adolescence employ self- or peer-report methods to collect aggression data, these two measurement methods demonstrate weak correlations with one another in the literature. Social desirability has been identified as a source of bias in self-reported aggression and is often controlled in analyses using self-reported measures. Similarly, aggression-related social schemas could be a source of bias in peer-reported aggression; however, no control measure for this currently exists. The present study investigated potential differences in the psychosocial correlates of self- and peer-identified early adolescent direct and indirect aggressors. In addition, scales assessing gender, race, and popularity-based aggression stereotyping bias were created for the study as a way to tap participants social schemas, and their relationships to nominations of aggressive peers were examined. Participants (314 middle school students; M age = 12.83; SD = .96) were categorized into groups based on self-reported and peer-nominated aggression scores and compared across a number of demographic and psychosocial factors. After controlling for social desirability bias, self-identified aggressors were characterized by higher levels of manipulative behavior, whereas peer-identified aggressors were characterized by particular race, gender, and sociometric patterns. Specifically, peer-identified direct aggressors were more likely than self-identified aggressors to be African American, and peer-identified direct and indirect aggressors were less well-liked but more socially visible and popular than self-identified aggressors. Overall, results suggested that self- and peer-report methods identify qualitatively different groups of aggressive adolescents. Furthermore, participants endorsed expected gender and popularity-based stereotypes of aggressive adolescents. Endorsement of gender stereotypes of direct and indirect aggressive adolescents was related to the gender of nominated aggressive peers, but not related to peer-reported aggression levels. The wide use of peer report methods in the present literature on adolescent aggression suggests the need for better understanding of factors that influence those reports, as bias in peer reports is often not considered in interpreting findings. The potential influence of stereotyping in peer-identified aggression and the need for further investigation of procedures that control for bias in peer-report measures are discussed.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Katherine Clemans.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Graber, Julia A.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041879:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041879/00001

Material Information

Title: Explaining Discrepancies between Self and Peer Reports of Aggression in Adolescence
Physical Description: 1 online resource (93 p.)
Language: english
Creator: Clemans, Katherine
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: Psychology -- Dissertations, Academic -- UF
Genre: Psychology thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Though most studies of aggressive behavior in early adolescence employ self- or peer-report methods to collect aggression data, these two measurement methods demonstrate weak correlations with one another in the literature. Social desirability has been identified as a source of bias in self-reported aggression and is often controlled in analyses using self-reported measures. Similarly, aggression-related social schemas could be a source of bias in peer-reported aggression; however, no control measure for this currently exists. The present study investigated potential differences in the psychosocial correlates of self- and peer-identified early adolescent direct and indirect aggressors. In addition, scales assessing gender, race, and popularity-based aggression stereotyping bias were created for the study as a way to tap participants social schemas, and their relationships to nominations of aggressive peers were examined. Participants (314 middle school students; M age = 12.83; SD = .96) were categorized into groups based on self-reported and peer-nominated aggression scores and compared across a number of demographic and psychosocial factors. After controlling for social desirability bias, self-identified aggressors were characterized by higher levels of manipulative behavior, whereas peer-identified aggressors were characterized by particular race, gender, and sociometric patterns. Specifically, peer-identified direct aggressors were more likely than self-identified aggressors to be African American, and peer-identified direct and indirect aggressors were less well-liked but more socially visible and popular than self-identified aggressors. Overall, results suggested that self- and peer-report methods identify qualitatively different groups of aggressive adolescents. Furthermore, participants endorsed expected gender and popularity-based stereotypes of aggressive adolescents. Endorsement of gender stereotypes of direct and indirect aggressive adolescents was related to the gender of nominated aggressive peers, but not related to peer-reported aggression levels. The wide use of peer report methods in the present literature on adolescent aggression suggests the need for better understanding of factors that influence those reports, as bias in peer reports is often not considered in interpreting findings. The potential influence of stereotyping in peer-identified aggression and the need for further investigation of procedures that control for bias in peer-report measures are discussed.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Katherine Clemans.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Graber, Julia A.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041879:00001


This item has the following downloads:


Full Text





EXPLAINING DISCREPANCIES BETWEEN SELF AND PEER REPORTS OF
AGGRESSION IN ADOLESCENCE





















By

KATHERINE HALE CLEMANS


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2010


































2010 Katherine Hale Clemans









ACKNOWLEDGEMENTS

To my advisor, Dr. Julia Graber, thank you for being such a wonderful mentor and role

model. Thank you for all the guidance and professional support, for your flexibility and

understanding, and for your unfailing encouragement of all my varied research interests over the

years. It has been a true pleasure to work with you, and I am grateful to have had the opportunity

to be your student.

To my committee, Drs. Susan Bluck, Catherine Cottrell, and Patricia Ashton, thank you for

the insightful comments, brainstorming meetings, and individual guidance you have provided

me. Many thanks for your valuable assistance and advice.

To my research assistants, Kathleen Endorf, Joe Orovecz, Jordan Powers, Kimberly Papa,

Tarah Parrino, and David Alexander, thank you for your assistance in collecting and entering

data. I couldn't have accomplished this project in the time that it took without such a terrific and

competent team.

To Dr. Russ Froman and the faculty, staff, and students ofP.K. Yonge Developmental

Research School, thank you for welcoming me into your school and for allowing me to collect

this data.

And to my family, for their support and love, thank you.









TABLE OF CONTENTS

page

A CK N OW LED GEM EN TS ......... ......... ............... ........................... .................................. 3

L IST O F T A B L E S ...................... ............... ....................................................... 7

LIST O F FIGU RE S ................................................................. 8

ABSTRAC T ...........................................................................................

CHAPTER

1 IN TRODUCTION ................. ..................... ..... ........... ...... ........ ........ .. 11

M measuring Aggression in Adolescence ............................................................. ... ............ 11
Comparisons of Self- and Peer-Identified Aggressive Adolescents....................................15
Stereotyping Bias in Peer Nominations of Aggressive Individuals ..................................17
Psychosocial Factors Related to Aggression in Adolescence.....................................19
E m otion and Personality Factors......................................................................... ...... 20
S o cial F acto rs ................................................................2 1
D em graphic C haracteristics................................................. .............................. 22
T h e P re sen t S tu d y .......................................................................................................2 3
S p e c ific A im s .................................................................................................................... 2 3

2 METHODS .........................................26

P articip an ts .........................................................................2 6
P ro c e d u re .............. .... ...............................................................2 6
M easu res .. ...................... ................................................................2 7
P eer-R reported A aggression ......................................................................................... 27
Self-R reported A aggression ......................................................................................... 2 8
Direct self-reported aggression. ....................................................... 28
Indirect self-reported aggression. ........................................ .... ........ 28
A aggression Stereotyping ........................................................................................... 2 9
S o c ia l D e sirab ility ..................................................................................................... 3 1
Sociometric Categorizations.................................. 31
Social preference and impact. .................................................31
P erceiv ed p opu clarity .................................................................................. 3 1
Em otion and Personality Indicators .............................................................. ... 32
M anipulative behavior................................................... 32
R em orselessness .............................................................................................. 32
E m p a th y ............................................................................................................. 3 2
A n g e r ........................................................................................................ 3 2
A n g er re g u latio n ................................................................................................. 3 3




4










3 R E SU L T S ....................................................... 34

C reaction of A aggressor G groups ............................................................................... ........ 34
P relim inary A nalyses........... ................................................................ .......... ....... 34
Criteria for Group M em bership......................................................... .............. 36
D descriptive Statistics ....................................................... ........................ ............ .37
A g g re ssio n ................................................................................................................. 3 7
S te re o ty p in g ............................ ........ ........................ .. ................................................3 7
Behavioral, Emotional, and Sociometric Characteristics..............................................39
Potential Covariates ............... ................. ............ ......................... ... 40
S o cial D e sirab ility ..................................................................................................... 4 0
G e n d e r .........................................................................4 0
Race .......................................40
G ra d e ..........................................................................4 1
F o rm o rd e r ................................................................................................................. 4 1
S u m m ary ..................................................................................................... ............... 4 2
Analyses A dressing Question 1 .................................................. ......... 42
Behavioral & Emotional Characteristics .............................................................. 42
Direct Aggression .............................................................................. 43
Indirect A aggression ......................... .......................................................... ............... 44
D em graphic and Sociom etric Characteristics ........................................ .............. 45
D em o g rap h ic s ....................................................................................4 5
Sociometric Characteristics...................... ......... .......... 47
Analyses A dressing Question 2 ................ .... .... ........ ....... ..... ... .. ..................... 49
Stereotyping Scores and Nominated Peers' Gender, Race, and Popularity .................49
Direct aggression................... ...................50
In direct aggression .......................................................................................5 1
S u m m ary ............... ... ...................................................................................... 5 1
Stereotyping Scores and Aggression Group Membership .................................... ..52
Nominators' Average Stereotyping Scores and Levels of Peer-Reported Aggression ...54

4 DISCUSSION ................................................................ 64

Do Self- and Peer-Report Aggression Measures Identify Students with Different
P sychosocial P profiles? ................................................................................... ............ 64
High Self vs. High Peer: Antisocial Indicators ....................... .......................65
High Self vs. High Peer: Sociometric and Demographic Characteristics .......................66
The High Multiple Group: "True" Aggressors? ....................................... .............67
Are Stereotyping Scores Related to the Demographic and Sociometric Characteristics of
Peers N om inated as A aggressive? ............................................................. ............. 69
Strength s an d L im station s ...................................................................... ...................... ..... 72
D evelopm mental Considerations................................ ............................ ............... 74
Conclusions........................ ................... ................ 75









APPENDIX

A SU R V E Y M E A SU R E S ............................................................................... .....................77

B CORRELATIONS BETWEEN STUDY VARIABLES .............. ............ .....................79

C STUDY VARIABLE MEAN DIFFERENCES FOR GENDER, RACE, AND GRADE
L E V E L ............................. .......................... .................................. 82

D SUMMARY OF QUESTION 1 ANALYSES WITH MORE STRINGENT GROUP
C R IT E R IA ........................ ........................................85

L IST O F R E F E R E N C E S ............................................... ......... .................. ...............................87

B IO G R A PH IC A L SK E T C H ............................................ ......... ................... ...........................93








































6









LIST OF TABLES


Table page

3-1 Group Ns and Aggression Means and Standard Deviations by Aggressor Group ............57

3-2 Membership Agreement for Direct and Indirect Aggression Groups...........................5..58

3-3 Means and Standard Deviations for Aggression and Aggression Stereotyping ...............59

3-4 Observed and Expected Gender and Race Distributions across Aggression Groups........60

3-5 Summary of Hierarchical Regression Analyses for Particpants' Gender Stereotyping
Scales Predicting Gender Percentages of Nominated Peers ...........................................61









LIST OF FIGURES

Figure page

1-1 Conceptual model of factors influencing self- and peer-based measures of aggressive
b eh av ior. ......... ......... .. .. ......... .. .. ......................................................2 5

3-1 Graphical representation of group means by aggression type for behavioral and
em optional characteristics.......................................................................... ....................62

3-2 Graphical representation of group means by aggression type for sociometric
characteristics ..................................... .................. ................ ......... 63









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

EXPLAINING DISCREPANCIES IN SELF AND PEER REPORTS OF AGGRESSION IN
ADOLESCENCE

By

Katherine Hale Clemans

August 2010

Chair: Julia A. Graber
Major: Psychology

Though most studies of aggressive behavior in early adolescence employ self- or peer-

report methods to collect aggression data, these two measurement methods demonstrate weak

correlations with one another in the literature. Social desirability has been identified as a source

of bias in self-reported aggression and is often controlled in analyses using self-reported

measures. Similarly, aggression-related social schemas could be a source of bias in peer-

reported aggression; however, no control measure for this currently exists. The present study

investigated potential differences in the psychosocial correlates of self- and peer-identified early

adolescent direct and indirect aggressors. In addition, scales assessing gender, race, and

popularity-based aggression stereotyping bias were created for the study as a way to tap

participants' social schemas, and their relationships to nominations of aggressive peers were

examined. Participants (314 middle school students; M age = 12.83; SD = .96) were categorized

into groups based on self-reported and peer-nominated aggression scores and compared across a

number of demographic and psychosocial factors. After controlling for social desirability bias,

self-identified aggressors were characterized by higher levels of manipulative behavior, whereas

peer-identified aggressors were characterized by particular race, gender, and sociometric

patterns. Specifically, peer-identified direct aggressors were more likely than self-identified









aggressors to be African American, and peer-identified direct and indirect aggressors were less

well-liked but more socially visible and popular than self-identified aggressors. Overall, results

suggested that self- and peer-report methods identify qualitatively different groups of aggressive

adolescents. Furthermore, participants endorsed expected gender and popularity-based

stereotypes of aggressive adolescents. Endorsement of gender stereotypes of direct and indirect

aggressive adolescents was related to the gender of nominated aggressive peers, but not related

to peer-reported aggression levels. The wide use of peer report methods in the present literature

on adolescent aggression suggests the need for better understanding of factors that influence

those reports, as bias in peer reports is often not considered in interpreting findings. The potential

influence of stereotyping in peer-identified aggression and the need for further investigation of

procedures that control for bias in peer-report measures are discussed.









CHAPTER 1
INTRODUCTION

Adolescent aggression has long been of interest to researchers, educators, clinicians,

parents, and policy makers. Aggressive behavior during this time can be direct (e.g., physical

fighting, verbal threats or insults) or indirect (e.g., rumor spreading, social exclusion,

withholding of friendship). Much literature on such aggression and antisocial behavior has

focused on understanding its development and on identifying factors which place individuals at

risk. Despite the ample literature on the development of aggressive behavior, however, questions

remain regarding the best method of assessment or source of information on aggression in

adolescence. The present study investigates whether characteristics of early adolescents

identified as aggressive differ as a function of self- and peer-report methods of assessment, as

well as potential factors which may contribute to bias in these methods.

Measuring Aggression in Adolescence

Assessment issues become increasingly important during early adolescence for several

reasons. First, time spent with friends and participating in activities with peers increases (Larson

& Richards, 1991), and secondly, adolescents begin to distance themselves from their parents

and engage in more activities that may be unmonitored by parents and teachers (Marshall,

Tilton-Weaver, & Bosdet, 2005). For this reason, the teacher and parent reports that are

frequently used to obtain information on aggressive behavior in childhood may miss important

behaviors, particularly concerning indirect forms of aggression that are less easily observed by

an outsider to the peer group. For instance, with the advent of widespread text messaging and

online social network usage among early adolescents (Hinduja & Patchin, 2008), much indirect

aggression may be taking place increasingly in private venues to which adult observers have

little to no access. Consequently, researchers may be better served by focusing on self-reported









information or on information given by peers, whom many researchers expect to have the most

accurate information about an adolescent's social behavior-not least because peers are usually

the objects at whom the adolescent directs his or her aggression (Peets & Kikas, 2006). Thus,

while studies using preschool and elementary school age groups frequently employ parent and

teacher ratings of aggression, almost all current studies of aggression in middle school and high

school age groups use at least one self- or peer-reported measurement method.

Longitudinal evidence suggests that agreement between self-reported and peer-reported

aggression peaks in early adolescence (Pakaslahti & Kelitkangas-Jarvinen, 2000). However,

although peer-report and self-report measures of aggression are frequently intended to be

measurements of the same construct, the literature to date reveals a consistent trend of weak

correlations between the two, usually falling within the range of r = .10 to r = .35, which

suggests a low level of shared variance between the two measures (Achenbach, McConaughy, &

Howell, 1987; Card, Stucky, Sawalani, & Little, 2008; Epkins & Myers, 1994; Henry et al.,

2006; Pakaslahti & Kelitkangas-Jarvinen, 2000; Pellegrini & Bartini, 2000; Xie, Cairns, &

Cairns, 2002). This is problematic for researchers seeking to investigate a single construct of

aggressive behavior.

Figure 1-1 portrays a conceptual model of the factors which can influence self- and peer-

based measures of aggressive behavior. At the core of the model lie the multitude of factors

which contribute to aggression at various points in the life span. During adolescence, numerous

biological, psychological, and social processes take place which shape an adolescent's social

behavior. As Bronfenbrenner's Ecological Model of Development (Bronfenbrenner, 1979)

indicates, adolescents develop via interactions within many different contexts, including

immediate familial and peer environments, along with broader cultural and societal contexts.









Much research has been conducted on the role of these factors in the development and

manifestation of aggression in adolescence. These include, but are not limited to, individual

factors such as cognitive deficits, hyperactivity, emotion levels and emotion regulation abilities;

familial factors such as parental punishment style and mother's IQ, social factors such as peer

rejection, school involvement, and early pubertal development in relation to one's peers, broader

environmental factors such as socioeconomic status and characteristics of one's neighborhood,

and cultural factors such as gendered expectations for behavior (see Dodge, Coie, & Lynam,

2006, for a review). These factors are thought to directly influence or be indicative of an

individual's overarching tendency to engage in aggressive behavior, and thus, barring

measurement error, their influence is generally expected to be captured by self- and peer-report

measurement methods in similar fashions.

If psychosocial influences were the only cause of variability in measurements of

aggressive behavior, we should see considerably greater overlap in ratings of aggression

originating from different sources. As it is, that is not the case. Several factors may play a role in

the relatively weak overlap between different informant reports of aggression. Achenbach and

colleagues (1987) suggest that, rather than calling into question the validity of any one

measurement type, low correlations between aggression scales may represent variations in the

same underlying behavioral construct; these variations manifest themselves differently according

to the context or situations in which they are experienced. For instance, an individual rating

herself/himself on aggressive behavior may recall instances in which behavior was directed

toward siblings at home, which classmates would not have occasion to observe. Similarly,

certain circumstances may arise in which an individual is effectively able to conceal behavior

directed at peers, as in the anonymous generation of a rumor. In both of these cases, individuals









would have access to information about their own aggression which would not be shared by

peers. Alternatively, individuals may be sometimes unaware of the effect of their actions on

others' feelings or social reputations; leading peers to categorize as aggressive certain behaviors

that the individual would not. Circumstances such as these may lead to variability in the

influence of situational factors on measurements of aggressive behavior and contribute to the

divergence of peer and self reports.

Low correlations between aggression measures, however, may also point to biases related

to the methods in which the data are collected. Aggression is generally considered to be a

socially unacceptable behavior; thus, in the case of self-reported data, individuals may

underreport aggressive behavior to cast themselves in a better light to both themselves and to

researchers (Peets & Kikas, 2006; Nederhof, 1985). As a result, social desirability scales, which

measure an individual's tendency to report that they engage in culturally approved, though

unlikely, behaviors (Crowne & Marlowe, 1960; Reynolds, 1982), have been developed, and

these scales are often used as control variables to reduce the influence of social desirability bias

in analyses of self-reported behavior.

Less attention has been paid to biases in peer reports of aggression, as many researchers

are of the opinion that peers are more valid sources of information about an individual's behavior

due to the fact that, since they are not reporting on their own behavior, issues of social

desirability bias are not thought to be as applicable (e.g., Peets & Kikas, 2006). Peer reports,

however, can be affected by adolescents' social schemas. A social schema is a cognitive

structure that represents a person's knowledge about the traits and goals of particular individuals

who fall into specific social categories (Fiske & Taylor, 1991). Social schemas help people

organize and interpret information about their social world and provide expectations for others'









behavior, and thus can lead to stereotypes of particular social groups. Schemas can also create

biases in the encoding and recall of information about a social event (Hamilton, Stroessner, &

Driscoll, 1994; Younger, Schneider, & Daniels, 1991). An individual's gender-role stereotypes,

for instance, can influence the perception and recollection of others' behavior, including a

tendency to recall gender-consistent information more often than gender-inconsistent

information (Fiske & Taylor, 1991; Cantor & Mischel, 1977). Alternatively, if a person

possesses a schema for a certain social group which includes the traits "hostile" or "aggressive,"

he/she may selectively notice hostile cues over non-hostile cues in ambiguous social situations

with a member of that group (Sagar & Schofield, 1980). Social schemas also become more

salient when individuals think about members of social outgroups (i.e., social categories, such as

race, gender, or nationality, of which the individual is not a member; Hamilton et al., 1990).

Comparisons of Self- and Peer-Identified Aggressive Adolescents

Regardless of whether or not low levels of correlation between self- and peer-reported

aggression are the result of participant biases, the relative lack of agreement between these

measurement methods suggests that groups of students identified as aggressive may differ in

important ways depending on the method by which aggression is assessed. Surprisingly little

research, however, has addressed this question directly. One exception is Card et al. (2008), a

meta-analysis which found that gender differences in physical aggression (i.e., boys were more

physically aggressive than girls) were significantly larger for studies using peer nomination

methods than for studies using self-report methods. Card and colleagues also found that peer

nomination and self-report studies did not differ in the strength of relationships between

aggression and emotional dysregulation. In addition, a study which examined differences in self-

versus peer-reports of victimization (Graham, Bellmore, & Juvonen, 2003) found that self-

identified victims experienced increased psychological maladjustment, whereas peer-identified









victims did not differ from non-victims in psychological maladjustment but did perform more

poorly in school. In addition, the peer-identified victims were more likely than the self-identified

victims to be African American and male. These results suggest that the profile of students

identified as victims of peer aggression can differ substantially depending on the measurement

method used. However, the study did not investigate group differences in peer- and self-

identified perpetrators of aggression, but instead focused on victims only.

In light of the relative lack of research addressing these questions, Clemans and Sontag

(2009, April) conducted a group-based analysis to investigate potential differences in the

psychosocial correlates of self-and peer-identified early adolescent aggressors. Several

demographic, sociometric, behavioral, and emotion/personality factors were selected on which to

compare peer- and self-identified aggressors. In addition, a measure of social desirability was

included in order to reduce underreporting bias and to increase the validity of the self-report

aggression measures and other self-reported variables in the study (Nederhof, 1985). Similar to

previous studies utilizing multiple informant reports, significant but weak correlations between

self- and peer-reported aggression were found (r = .15, p < .001 for both direct and indirect

aggression). The low correlations suggested that substantially different groups of aggressive

students were being identified by each method.

Results indicated that self- and peer-identified aggressors were indeed characterized by

relatively different psychosocial profiles. Direct self-identified aggressors showed significantly

higher levels of socially manipulative behavior, remorseless/unemotional affect, and delinquency

than other groups, and indirect self-identified aggressors showed high levels of socially

manipulative behavior and delinquency. In contrast, peer-identified aggressors were

characterized by specific demographic and sociometric patterns. For instance, although self-









identified directly aggressive participants did not differ from nonaggressors in their racial/ethnic

distributions, direct aggressors who were nominated by peers were significantly more likely to be

African American and significantly less likely to be European American than were

nonaggressors. In addition, peer-identified indirect aggressors were significantly more likely than

self-identified indirect aggressors to be female, and peer-identified indirect aggressors were also

more socially visible, or well-known within the social group, than either self-identified indirect

aggressors or nonaggressors.

The results from Clemans and Sontag that peer-nominated direct aggressors tended to be

African American, while peer-nominated indirect aggressors tended to be socially visible

females, are patterns clearly reminiscent of social stereotypes of directly and indirectly

aggressive adolescents that may be driven by popular media (e.g., "mean girls," "gangsta"

culture). These findings suggested that early adolescents may be using social schemas to inform

their nominations of aggressive classmates, resulting in a bias due to stereotyping (Fiske &

Taylor, 1991; Giles & Heyman, 2005).

Stereotyping Bias in Peer Nominations of Aggressive Individuals

Stereotyping bias can come into play when beliefs about the nature of a group of people

influence individuals' interpretations and recollections of their behavior. Several studies have

employed survey-based measures of stereotyping as it specifically relates to aggression or to

group comparisons. A classic example of this is the Katz & Braly (1933) checklist, in which

participants mark on a list the traits which they think describe a particular social or demographic

group. However, this measure may be particularly susceptible to bias due to social desirability

(Whitley & Kite, 2006). Sagar and Schofield (1980), employing a more subtle approach,

measured the influence of racial stereotypes on the interpretation of ambiguous social behavior

using a series of pictorial cues and verbal descriptions of ambiguously aggressive social









interactions in which the race of the actor or target was systematically varied. Subjects were then

asked to rate the actor on several traits (eg., friendly, threatening). There are drawbacks

associated with this measure, however, including the time-consuming nature of the task and its

dependence on the perceived ambiguity of the social interaction in question, which can vary

among scenarios as well as from person to person.

Alternatively, Ryan, Judd, and Park (1996) used a mean range and estimation task to

measure racial stereotypes. This procedure consisted of a series of behavioral dimensions with

opposing behaviors as endpoints (e.g., well-dressed poorly-dressed), on which participants

were asked to mark their perceived average of the racial group in question, as well as where they

believed the highest and lowest members of the group would be. While this procedure is

appealing because it employs a positive/negative comparison and may invoke more implicit

reactions than free responses would invoke, the use of statistical terms such as "average" and

"range" may be too sophisticated an instructional procedure for early adolescents. An assessment

method similar to the group comparison used by Ryan et al. (1996), yet with simpler instructions,

is that used by Otten & Stapel (2007). Participants indicated on a rating spectrum whether a

particular behavioral or emotional trait applied more to one or another particular ethnic group,

which comprised the two endpoints on the spectrum. Participants could indicate that it applied

equally to both groups by selecting the midpoint of the spectrum.

The absence of an aggression stereotyping measure was a major limitation of Clemans

and Sontag (2009, April), because we were not able to directly test whether aggression

stereotyping tendencies impacted ratings of peer nominations. However, this study did

investigate whether the tendency to nominate peers along certain demographic lines appeared to

be shared among all peer nominators, or whether this tendency appeared to be particularly strong









for members of the corresponding demographic outgroup. For instance, are boys more likely

than girls to nominate other-gender classmates as indirect aggressors? Because genders are

highly segregated in their social interactions during early adolescence, a preference for same-sex

nominations on peer-nomination measures of social behavior is expected (Coie, Dodge, &

Kuperschmidt, 1990; Maccoby, 1998); thus, high levels of other-sex nominations may suggest

that stereotyping is taking place.

For gender, results showed that boys were much more likely than girls to nominate

members of the other sex as indirect aggressors, resulting in a greater number of peer-identified

indirectly aggressive girls. Because this result was not supported by a greater number of self-

identified indirectly aggressive females in the current study, nor by results of previous meta-

analytic findings of trivial differences in levels of indirect aggression between boys and girls

(Card et al., 2008), it lends further support to the suggestion that some students in this study were

in fact influenced by current social stereotypes of aggressive behavior in adolescence. Similar

results were also found by Card, Hodges, Little and Hawley (2005): Sixth-grade males

nominated a larger proportion of other-sex members as indirect aggressors than did their female

classmates. Beyond Card et al., I am aware of no other research to date which directly

investigates the influence of racial or gender stereotyping bias on peer nominations of aggressive

individuals.

Psychosocial Factors Related to Aggression in Adolescence

One purpose of the present study is to replicate and extend the findings from Clemans

and Sontag, which suggest that substantially different groups of adolescents who possess

dissimilar psychosocial profiles are being identified by self- and peer-report measurement

methods. Although many factors can be investigated in relation to aggressive behavior, for the

purposes of parsimony and specificity, the present study focuses on investigating differences in a









constellation of psychosocial characteristics which represent both individual and social

influences on behavior. These include several emotional and personality traits known to be

indicative of aggressive behavior as well as influential in its development, peer-driven indicators

of aggressive behavior relating to social status within one's peer group, and demographic

indicators linked to differences in aggressive behavior, including gender, race, and

socioeconomic status.

Emotion and Personality Factors

Empathy and remorse. Empathy refers to the personal experience of another's affective

state after observing or learning of that state (Eisenberg, Spinrad, & Sadovsky, 2006); remorse,

or guilt, is a feeling of discomfort following a transgression (Eisenberg, 2000). Low levels of

empathy and remorse are emotional indicators of psychopathic personality and behavioral

tendencies (Andershed, Kerr, Stattin, & Levander, 2002; Lynam, 1996). Children and

adolescents who have low levels of these emotions engage in more frequent and severe forms of

aggressive behavior (Eisenberg et al., 2002; Saltaris, 2002) and rate aggressive behavior as more

morally permissible than do their peers (Eisenberg, Miller, Shell, McNalley, & Shea, 1991).

Manipulative behavior. Social manipulation, including the telling of lies and the use of

dishonest charm to achieve social goals, is a behavioral indicator of psychopathic personality

(Andershed et al., 2002) and thus is similarly linked to elevated levels of aggression in

adolescence (Saltaris, 2002).

Anger and anger regulation. The emotional experience of anger is characterized by

"physiological arousal and cognitions of antagonism" (Novaco, 1994, p. 32); trait anger is

defined as an enduring propensity to become angry (Spielberger, Jacobs, Russell, & Crane,

1983). High anger levels contribute to aggressive behavior during adolescence (Cornell,

Peterson, & Richards, 1999; Nichols, Graber, Brooks-Gunn, & Botvin, 2006), due, in part, to the









fact that feelings of anger increase reactive behaviors to real or perceived hostile situations

(Muris, van der Pennen, Sigmond, & Mayer, 2008).

Anger and the ability to effectively regulate anger (i.e., to consciously reduce the

intensity of anger through mental or behavioral exercises) are related, since an individual with a

higher baseline anger level will have a relatively tougher job of controlling his/her temper.

Though the influence of emotion regulation factors on behavior has received much attention in

recent years (Eisenberg, Morris, & Spinrad, 2004), research on anger regulation specifically has

been scarce (Zeman, Shipman, & Suveg, 2002). Some studies find no direct link between anger

regulation and aggression in younger children (Dearing et al., 2002). Clemans, Graber, Nichols,

Brooks-Gunn, and Botvin (2007, March), however, found that a reduced ability to consciously

regulate experiences of anger in early adolescence was uniquely associated with increased

aggressive behavior even after accounting for levels of trait-based anger. As such, it was of

interest in the present study.

Social Factors

Social preference and social impact. Social preference refers to how well-liked one is

by one's peers, while social impact refers to how visible one is within one's peer group. Social

preference and social impact have somewhat different relationships with aggressive behavior in

adolescence. Specifically, high social impact is significantly related to high levels of aggression,

and children and adolescents who have high social impact scores are more likely to be aggressive

than children with high social preference scores (Newcomb, Bukowski, & Pattee, 1993).

Perceived popularity. Perceived popularity differs from peer acceptance and social

visibility in that it incorporates levels of dominance within the peer group. Perceived popularity

is usually correlated with social impact scores at around r = .50, and, like social impact, differs

from peer acceptance in its relationships with aggressive behavior. This is particularly true for









indirect forms of aggression. Cillessen & Borch (2006) found that while relationally aggressive

and non-relationally aggressive middle school students tended to have similar levels of peer

acceptance, relationally aggressive students were rated a full standard deviation higher on

perceived popularity measures. In contrast, physically aggressive middle school students were

somewhat less well-liked than non-physically aggressive students, but the two groups had similar

levels of perceived popularity.

Demographic Characteristics

Gender. Gender has an established relationship with physical aggression in the literature.

At all ages, males are more likely than females to be physically aggressive (Archer, 2004; Card

et al., 2008), to engage in direct verbal aggression (Archer, 2004; Card et al., 2008), and to

commit violent crime (Moffitt, Caspi, Rutter, & Silva, 2001). During high school, 44% of boys,

but only 27% of girls, reported having engaged in a physical fight in the last year (Center for

Disease Control and Prevention, 2008). When effect sizes from studies using self-report, peer

and teacher reports, and observational measurement methods are averaged together across the

lifespan, males tend to be higher than females in physical and verbal aggression by about .5

standard deviations (Hyde, 1984).

Indirect aggression does not show similar patterns of gender differences. Based on early

studies in which peers perceived girls as higher than boys in these behaviors, some researchers

have argued that indirectly aggressive behavior is a female-normative form of aggression (e.g.,

Crick, 1997). Evidence suggests that peers also endorse the view that females are the primary

perpetrators of indirect aggression, rating indirect aggression by a female as more serious than

the same behavior by male (Basow, Cahill, Phelan, Longshore, & McGillicuddy-DeLisi, 2007;

Coyne, Archer, Eslea, & Liechty, 2008). Subsequent studies have shown mixed results when

assessing actual levels of indirect aggression, however, and meta-analytic reviews have failed to









demonstrate significant gender differences in overall levels of indirectly aggressive behavior

across childhood and adolescence (Archer, 2004; Card et al., 2008).

The Present Study

The present study had two main parts. The first was a replication and extension of

Clemans et al. which assessed potential differences in patterns of psychosocial characteristics

among self- and peer-identified aggressive adolescents. To do so, participants were divided into

groups based on their levels of self-reported and peer-identified aggression, and groups were

compared with one another on a range of variables. The second part of the study directly

examined the relationship of stereotyping bias to nominations of aggressive peers.

Specific Aims

The specific aims of the present study were as follows:

Specific aim 1: To replicate previous findings which showed that the psychosocial

profiles of self-identified and peer-identified aggressors follow significantly different patterns,

and furthermore, that these patterns are reliant on the type of aggression being assessed (i.e.,

direct or indirect).

Based on findings in Clemans and Sontag (2009, April) and in previous literature, I

expected self-identified aggressors to differ from peer-identified or nonaggressors on emotion

and personality indicators of aggressive tendencies; self-identified aggressors should have higher

levels of remorselessness, manipulative behavior, and anger, and lower levels of empathy and

anger regulation, and that the patterns of these differences will similar for direct aggression and

indirect aggression. Furthermore, I expected these relationships to be significant even after

controlling for potential bias due to social desirability.

Similarly, it was expected that peer-identified aggressors would have higher social impact

and perceived popularity scores than self-identified aggressors, that peer-identified indirect









aggressors would be disproportionately female, and that direct aggressors would be

disproportionately African American. I also expected, based on previous research linking

physical aggression to male gender stereotypes, that peer-identified direct aggressors will be

disproportionately male (Giles & Heyman, 2005; Loy & Norland, 1981).

Specific Aim 2: To determine heitier variance in peer nominations of aqi e \\i'

adolescents can be explained by nominators' stereotypes of a',g ei .iv'e peers.

The second aim of the study was to address the lack of literature investigating potential

bias in peer nominations of aggression (in Figure 1-1, the farthest-right box). To assess the use of

social schemas in adolescents' nominations of aggressive peers, an aggression stereotyping

measure was created for this study based on procedures used in Otten and Stappel (2007). It

included a list of aggressive and prosocial behaviors on which participants compared gender

(girls vs. boys), racial groups (Black vs. White), and sociometric groups (popular vs. unpopular).

Stereotyping scales were expected to be correlated with their corresponding demographic or

sociometric category of peer nominations (i.e., participants who tend to associate females with

indirect aggression were expected to be more likely to nominate female classmates as indirect

aggressors on the peer nomination portion of the survey). Furthermore, it was expected that the

stereotyping scales would account for some of the variance in peer-reported aggression that was

not explained by levels of self-reported aggression, thus recommending their use as control

variables in future studies of aggression utilizing peer-reported measurement methods.



































Figure 1-1. Conceptual model of factors influencing self- and peer-based measures of aggressive
behavior.









CHAPTER 2
METHODS

Participants

Participants were 315 6t, 7th, and 8th grade students (Mage = 12.83; SD = .96) from a

single middle school in a small Southeastern city. Ninety-six percent of the student population

participated in the study. The student body of the school was representative of racial and

socioeconomic distributions of the county in which it was located. Approximately 35% of

students attending the school are eligible for the free/reduced lunch program (an indirect measure

of socioeconomic status). Participants were 50.8% European American, 22.6% African

American, 15.0% Hispanic/Latino, 2.8% Asian, and 8.9% other ethnicities. Girls comprised

49.5% of the sample.

Procedure

Consent procedures were approved by the Internal Review Board of the University of

Florida. The review board approved a waiver of active parental consent due to the fact that

active consent procedures may exclude students with the highest levels of problematic behavior,

reducing the generalizability of all findings (Tigges, 2003). Letters were sent to the home

addresses of parents/guardians which contained a letter explaining the purpose of the study,

details about the anonymity and confidentiality measures in place to protect students' privacy,

and a prepaid, self-addressed postcard which parents could use to decline consent for their

child's participation if they so desired. Phone and email contact information for the study office

were also provided for this purpose.

All measures were presented in self-report surveys administered by a trained research

team during school periods. Prior to administration of the surveys, students were given an assent

cover sheet, which briefly introduced the study explained the anonymity and confidentiality









procedures. Questionnaires were identifiable by a unique ID number only. Students were

verbally asked for their consent to participate prior to beginning the survey, and were informed

that they could stop at any time once they started with no negative consequences. Students who

declined to participate or who were withdrawn by their parents were given a free period until

their class completed the survey.

To address potential priming effects of the aggression stereotyping measures on peer

nominations, the survey was divided into two shorter counterbalanced sections, one containing

the aggression stereotyping measure and one containing the peer nomination measure.

Participants completed the sections in two separate class periods at least one week apart.

Researchers also returned to the school for a final day of data collection so that students who had

been absent on a regular administration day could complete their missed surveys.

Measures

Peer-Reported Aggression

Peer-reported aggression measures were based on procedures used in Putallaz et al.

(2007). The measure comprised ten items, five of which assessed directly aggressive behaviors

and five of which assessed indirectly aggressive behaviors. Because a specific aim of this study

was to make comparisons between peer- and self-reported aggression, the wording peer-report

aggression items mirrored as closely as possible a subset of items from the self-reported

aggression measures (described below). Example items include "gossip or say mean things about

other kids behind their backs" (indirect aggression) and "get in fights a lot" (direct aggression).

A complete list of items is included in Appendix A.

For each item, students were instructed to write the first and last names of kids in their

grade who best fit the item description. Students could nominate an unlimited number of names

if they wished, although the vast majority nominated 0 to 3 names for each item. Rosters listing









the first and last names of same-grade classmates were provided during this portion of the

questionnaire in order to assist in identification and spelling.

Total nominations for each item were standardized by grade and summed to create two

individual scales reflecting peer-report direct aggression (a = .88) and indirect aggression (a =

.77).

Self-Reported Aggression

Direct self-reported aggression. Direct self-reported aggression was assessed using

items from the Aggression Scale (Orpinas & Frankowski, 2001). This scale has demonstrated

good reliability in previous research by its authors. Students were asked to indicate how often

they had engaged in a range of behaviors during the past year. Example items included "I got

into a physical fight" and "I threatened to hurt or hit someone." Response options utilized a 1-4

Likert Scale in which 1=Never, 2=Once or twice, 3=A few times, and 4=Often. Responses were

averaged to create two individual scales reflecting self-reported direct and indirect aggression;

higher scores indicated higher levels of aggression.

The full direct aggression scale contained 8 items (a = .85). However, in order to keep

the construction of the self-report and peer-report aggression measures as similar as possible, a

reduced scale, which contained 5 items with similar wording to those assessed in the direct peer-

report measure, was utilized in subsequent analyses. The reduced scale was highly correlated

with the full direct aggression scale (r = .96) and demonstrated acceptable reliability (a = .78).

Indirect self-reported aggression. Indirect self-reported aggression was assessed using

items from the Revised Peer Experiences Questionnaire (Prinstein, Boergers, & Vernberg, 2001).

The RPEQ indirect aggression subscale has demonstrated good reliability with youth in this age

range and demographic background. Students were asked to indicate how often they had engaged

in a range of behaviors during the past year. Example items include "I said mean things about









someone behind his/her back" and "I left someone out on an activity or conversation that he/she

really wanted to be included in." Response options utilized a 1-4 Likert Scale in which l=Never,

2=Once or twice, 3=A few times, and 4=Often. Responses were averaged to create two

individual scales reflecting self-reported direct and indirect aggression; higher scores will

indicated higher levels of aggression.

The full indirect aggression scale contained 12 items (a = .77). A reduced scale which

contained 5 items with similar wording to those assessed in the direct peer-report measure was

also created. The reduced scale was highly correlated with the full indirect aggression scale (r =

.90) but demonstrated low, though not unacceptable, reliability (a = .56). The lower alpha levels

for indirect aggression are likely due to the fact that several different types of behavior are

assessed (including social exclusion, rumor spreading, and withdrawal of friendship), whereas

direct aggression items assess only physical and verbal forms and therefore overlap with each

other to a greater extent. Other often-used assessments of indirect aggression utilizing low item

counts have shown similarly low reliability levels. In order to keep the construction of the self-

report and peer-report aggression measures as similar as possible, the reduced scale was

employed in all subsequent analyses of indirect aggression.

Aggression Stereotyping

The aggression stereotyping scales for this study used a format adapted from Otten and

Stapel (2007) and were comprised of 12 items listing aggressive and prosocial behaviors. For

each item, students were asked to indicate on a 7-point spectrum scale whether they thought the

behavior applied more to one particular group of people, more to another group of people, or

equally to both groups. The set of 12 items was repeated for three group comparisons: gender

("boys vs. girls"), race ("Black kids vs. White kids"), and popularity ("popular kids vs.









unpopular kids"). These three dyads were chosen based on observed patterns in the demographic

and sociometric characteristics of peer-nominated aggressors in previous research.

Of the 12 items in each scale, 10 assessed aggression and tapped similar behaviors as the

5 direct and 5 indirect items from the peer-reported aggression scale. The two remaining items,

which assessed prosocial behavior, were included as filler items. A complete list of items is

presented in Appendix A. Scores on the aggressive behavior items were averaged to produce

overall direct and indirect aggression scores for each scale, and centered so that negative values

indicated greater bias toward the left-hand listed group and positive values indicated greater bias

toward the right-hand listed group. Reliability scores for the aggression stereotyping scales were

as follows: direct gender (a = .68); direct race (a = .78); direct popularity (a = .80); indirect

gender (a = .61); indirect race (a = .66; indirect popularity (a = .71).

Nominators' average stereotyping scores. In some analyses, a variable representing

the average stereotyping scores of the peers nominating a particular individual was used. This

was initially calculated in two ways: (1) For each time a participant's name was nominated on a

peer-reported aggression item, the corresponding score of the peer who made that nomination

was substituted and these scores were averaged together; (2) The same procedure, except before

averaging, scores were checked for redundancy nominations (e.g., a peer nominated the same

participant on multiple aggression items), and redundancies were removed before scores were

averaged so that each nominator's score was only counted once. This was done for all three

stereotyping scales and conducted separately for direct and indirect aggression, resulting in six

average scores.

The two different methods of calculation did not significantly effect scores. Correlations

between the two calculation methods on any one particular scale were very high (r = .98 to .99, p









< .001), indicating that the method of calculation did not significantly affect scores. Thus, only

the scores with redundancies removed (method 2) were used in subsequent analyses.

Social Desirability

Social desirability bias was assessed using a 10-item version of the Marlowe-Crowne

Social Desirability Scale (Strahan & Gerbasi, 1972), which consists of a series of statements

about socially desirable or undesirable behavior (e.g., "I am always polite, even to people who

are disagreeable"). Participants were asked to indicate whether the statements were true or false

as they pertained to themselves. Items were assigned a 1 if the participant selected the more

socially desirable response and a 0 if the participant selected the less socially desirable response.

Items were then averaged to produce an overall score with a range of 0-1, with higher scores

indicating greater social desirability bias.

Sociometric Categorizations

Social preference and impact. Four additional items were included in the peer-

nomination portion of the questionnaire. The first two, "write the names of kids in your grade

whom you like the most" and "write the names of kids in your grade whom you like the least,"

were used to mathematically compute social preference and social impact scores (Coie, Dodge,

& Coppotelli, 1982). Social preferences scores were created by subtracting standardized "liked

least" scores from standardized "liked most" scores. Social impact scores were created by

summing standardized "liked most" and "liked least" scores.

Perceived popularity. The final two items, "write the names of kids who are the most

popular in your grade" and "write the names of kids who are the least popular in your grade,"

assessed perceived popularity, or how popular one is within one's group. A popularity spectrum

score was created by subtracting standardized "least popular" nominations from standardized

"most popular" nominations.









Emotion and Personality Indicators

Manipulative behavior. Socially manipulative behavior (a = .88) was measured using

15 items from the Youth Psychopathic Traits Inventory (YPI; Andershed et al., 2002) which

assessed manipulativeness, dishonest charm, and lying behaviors. Participants were asked to

indicate how well each item applied to them. Example items include "I am good at getting

people to believe me when I make something up;" "It's easy for me to charm others to get what I

want from them." The response scale ranged from 1 ("almost always untrue") to 5 ("almost

always true"). Responses were averaged such that higher scores indicated greater levels of

manipulative behavior.

Remorselessness. Remorselessness (a = .69) was measured using 5 items from the YPI

which assessed a lack of guilty feelings in relation to one's behaviors. Participants were asked to

indicate how well each item applied to them. Example items included "I seldom regret the things

I do, even if other people feel that they are wrong." The response scale ranged from 1 ("almost

always untrue") to 5 ("almost always true"). Responses were averaged such that higher scores

indicated greater remorselessness.

Empathy. Empathetic responding (a = .82) was assessed using the 20-item Basic

Empathy Scale (BES; Joliffe & Farrington, 2006). Participants are asked to indicate how much

they agree or disagree with a series of statements. Example items include "I can usually work out

when my friends are scared" and "I don't become sad when I see other people crying" (reverse

coded). The response scale ranged from 1 ("strongly disagree") to 5 ("strongly agree").

Responses were averaged such that higher scores indicated greater levels of empathy.

Anger. The seven-item anger subscale from the Buss and Perry (1992) Aggression

Questionnaire was used to assess trait levels of anger (a = .72). Students were asked to rate how

well a series of statements fit them. Items included "I sometimes feel like a powder keg ready to









explode" and "Some of my friends think I'm a hothead." Response categories ranged from 1

(Really Not True for Me) to 5 (Really True for Me). Items were averaged such that higher scores

indicated greater anger.

Anger regulation. Anger reduction skills (a = .68) were assessed with a six-item scale

created for the Life Skills Training Program (Epstein, Botvin, Diaz, Baker, & Botvin, 1997).

Participants were asked how often they engaged in a series of activities when they felt really

angry. Items included "Count to ten," "Take a few deep breaths," and "Tell myself this isn't

worth fighting over (it's no big deal.)" Response categories ranged from 1 (Never) to 5

(Always). Items were averaged such that higher scores indicated greater skill at conscious anger

reduction.









CHAPTER 3
RESULTS

Creation of Aggressor Groups


Following procedures utilized in Cillessen and Borch (2006), groups of students were

created based on self-reported (reduced scale) and peer-reported aggression scores. Two sets of

groups, one for direct aggression and one for indirect aggression, were created. Within each type

of aggression, students could fall into one of three categories: (1) low aggression (henceforth

referred to in results and discussion as "low"); (2) high self-reported aggression only ("high

self"); (3) high peer-reported aggression only ("high peer"); (4) high in both self and peer

aggression ("high multiple"). Group membership within aggression type (i.e., direct versus

indirect) was mutually exclusive, but students could be members of both a direct and an indirect

high-aggression group.

Preliminary Analyses

What constituted "high" aggression for self- and peer-reported scores was determined by

cutoff scores based on methods utilized in past literature as well as on conceptual considerations.

For instance, Cillessen and Borch (2006) selected, as their criterion for membership in a high

sociometric popularity or perceived popularity group, the cutoff of .5 SD above the mean on one

of these measures. This resulted in 39.4% of their sample achieving membership in at least one

of the two "high popularity" groups. Clemans and Sontag (2009, April) also utilized the .5 SD

cutoff for membership in either high self-reported or high peer-reported aggression groups,

resulting in 23-28% of the sample meeting the criteria for a high self-reported aggression, 15-

17% of the sample meeting the criteria for high peer-reported aggression, and 45.5% of the

sample achieving membership in at least one "high" direct or indirect aggression group;

individual group Ns for high self, high peer, and high multiple report groups ranged from 18 to









57. In that study, the .5 SD cutoff score was sufficient to produce diverging patterns of

behavioral, emotional, and sociometric characteristics for high self and high peer groups.

In the present study, as in prior work, the .5 SD criterion resulted in about 40% of the

sample meeting criteria for at least one "high" group. One concern is whether the .5 SD cutoff is

sufficiently stringent for identifying truly aggressive adolescents. Notably, measures of self-

reported aggression almost always result in positively skewed data, since the majority of any

normative adolescent population engages in either no involvement or only sporadic involvement

in direct and indirect aggressive behavior. This results in a smaller subset of students who score

.5 SD or higher above the mean on any one measure than would be expected with normally

distributed data. The relatively large percentage of students meeting the criteria for at least one

"high" group in previous research may be interpreted as an indicator of the varied forms that

aggression takes during adolescence, as well as the range of opinions between individuals and

their peers as to who engages in "high" aggression, rather than as evidence of too lenient a cutoff

score.

In addition, preliminary analyses found a pronounced positive skewness in the peer-

reported aggression measures. Although the majority of students did not receive any

nominations on peer-reported aggression items, the range of scores was quite large, and there

were several outliers with exceptionally high nomination tallies (one student received 108 total

nominations on individual aggression items), resulting in substantially fewer numbers of students

who were above the .5 SD cutoff for peer-reported aggression (12.5% to 15.8% of total sample)

than were above the cutoff for self-reported aggression (21.9% to 24.3% of total sample). In this

case, it was determined that the .5 SD cutoff might be non-representative for peer-reported

aggression and that selecting a "high" group based on a nomination tally cutoff score would be









more appropriate for creating "high" peer-reported aggression groups. A cutoff of 4 nominations

was selected for the reason that it included relatively the same numbers of students (23.7% to

24.9% of total sample) as the .5 SD cutoff for self-reported aggression, which increased the

equality of high aggression group Ns.

To explore the effect of using more stringent cutoff criteria, all group-based analyses in

the present investigation were also conducted with 1 SD cutoff criteria for self-reported scales

and a cutoff of 5 nominations for the peer-reported aggression. This approach produced smaller

cell sizes for the aggression groups, but for the most part, patterns of differences between groups

did not change. Findings using the more stringent criteria which differed from the reported

analyses are presented and discussed in Appendix D.

Criteria for Group Membership

As such, the criteria for group membership within aggression groups were as follows:

"High self' aggressors were students whose self-reported aggression score was .5 SD or higher

above the mean, but who had fewer than 4 total nominations for peer-reported aggression. "High

peer" aggressors were students who had 4 or more total nominations for peer-reported

aggression, but who scored lower than .5 SD above the mean on self-reported aggression. "High

multiple" aggressors had scores .5 SD or higher above the mean or higher on self-reported

aggression and 4 or more total nominations for peer-reported aggression. The "low" group was

comprised of the remaining students. High self, high peer, high multiple, and low aggression

groups were created for both direct and indirect aggression separately, resulting in a total of eight

groups. Table 3-1 presents group Ns, mean levels of aggression for each group. Table 3-2

presents membership agreement across direct and indirect aggression. Seventy-five students

(23.8% of the sample) were members of both a direct and an indirect high-aggression group.









Descriptive Statistics


Aggression

Means and standard deviations for all self-reported and peer-reported aggression

variables are presented in Table 3-3. Consistent with data on aggressive behavior in normative

populations, all were positively skewed. Thus, in subsequent analyses, aggression variables have

been transformed via the square root function to increase the normality of their distributions (this

excludes group-based analyses using previously created "high aggression" groups).

A full table of correlations between all study variables is presented in Appendix B. As

expected, Pearson correlations between the self- and peer-reported forms of aggression were

significant but somewhat small, r = .27, p < .001 for self- and peer-reported direct forms; r = .23;

p < .001 for self- and peer-reported indirect forms. Controlling for social desirability did not

affect the strength of either relationship. The low amount of shared variance indicated by these

correlations suggested that substantially different groups of individuals were being identified as

high in aggression according to each method.

Agreement within reporting method was better, r = .42, p < .001 for direct and indirect

self reports; r = .68, p < .001 for direct and indirect peer reports. Students appeared to make

more of a distinction between direct and indirect forms of aggression in their own behavior than

they did in their peers' behavior. Since more than 50% of the variance in even the peer-reported

measures was unshared, however, direct and indirect aggression were treated as distinct forms

and were evaluated separately in subsequent analyses.

Stereotyping

Means and standard deviations for individual stereotyping scales are presented in Table

3-3. All scores had good distribution. The shift of each scale in one direction or the other from

the center reflected common social stereotypes relating to the gender and popularity of









aggressive adolescents: Students endorsed the view that boys were more likely to engage in

direct aggression (shift to boys' side = 1.05 SD), whereas girls were more likely to engage in

indirect aggression (shift to girls' side = .96 SD); students also believed that popular students

were more likely than unpopular students to engage in both forms of aggression, but this was

particularly true for indirect forms (shift to popular side = 1.15 SD for direct and 1.49 SD for

indirect).

Students showed the least evidence of aggression stereotyping for racial groups. Students

endorsed the view that Black students were more likely than White students to engage in direct

aggression (shift to Black side = .72 SD), but this was the smallest significant shift out of all the

scales. No significant shift was found for indirect aggression. Three percent of the sample (N

11) completed the stereotyping scales for gender and popularity but left the racial group

stereotyping scales blank, presumably because they were uncomfortable answering these

particular questions. These students were compared to the rest of the sample on all relevant

study variables (including aggression, aggression stereotyping, peer nomination demographic

percentages, and behavioral and emotional characteristics) on a series of t-tests of independent

samples; no significant differences atp < .05 were present.

Direct and indirect aggression stereotyping scales comparing popular and unpopular

students showed the most agreement (r = .61, p < .001); students believed that popular students

were more likely than unpopular students to engage in all types of aggression. Students who

endorsed the view that popular students were more indirectly aggressive than unpopular students

were also slightly more likely than other students to believe that girls were more indirectly

aggressive than boys (r = .20, p < .001). Although other correlations between stereotyping scales









were significant atp < .05, all were weak (r < .20); as such, they were not considered particularly

notable.

Behavioral, Emotional, and Sociometric Characteristics

The behavioral and emotional characteristics measured in the current study included

manipulative behavior (M= 2.16, SD = .69), remorselessness (M= 2.14, SD = .78), empathy (M

= 3.52, SD= .50), anger (M= 2.55, SD= .79), and anger regulation skills (M= 2.39, SD= .80).

Each had good distribution, with minimal positive skew for manipulative behavior and

remorselessness. The sociometric characteristics measured in the current study included social

preference, social impact, and perceived popularity; these were standardized into z-scores or

calculated from standardized scores. All had normal distribution.

Out of 10 possible correlations between behavioral and emotional characteristics, 7 were

significant at thep < .01 level, and all variables were significantly related to at least two others

(see Appendix B for correlation values). Furthermore, all significant relationships made

conceptual sense (e.g., lower levels of empathy were associated with higher levels of

remorselessness and manipulative behavior). These results suggested that together, the variables

represented a psychosocial profile which, reversing empathy, could be considered prosocial at

lower levels and antisocial at higher levels.

Previous literature (e.g., Newcomb et al., 1983) has suggested that social preference and

social impact tend to be distinct domains of sociometric status. The present study corroborated

these findings: social preference and social impact were weakly correlated (r = -. 14, p < .05).

Both, however, were related to perceived popularity (r = .37, p < .001 for social preference; r =

.40, p < .001 for social impact). This supported the conceptual distinction that social preference

is most related to likeability, social impact is most related to notice and influence within the peer









group, and students who are universally considered to be the most popular in their grade tend to

be high in both of these qualities.

Potential Covariates

Social Desirability

The Marlow-Crowne social desirability scale (M= .50, SD = .24) was employed in this

study as a measure of reluctance to self-report socially undesirable behavior. Accordingly, social

desirability scores were significantly correlated atp < .01 with self-reported direct aggression (r

= -.32), self-reported indirect aggression (r = -.43), anger (r = -.43), manipulative behavior (r = -

.38), and remorselessness (r = -.25). Social desirability was not related to any other self-reported

variables in the study, including anger regulation, empathy, and stereotyping scores, and did not

significantly differ by race, gender, or grade level.

Gender

Mean differences between boys and girls were significant at thep < .05 level for several

behavioral, emotional, and sociometric characteristics as well as a few stereotyping scales. All

significant findings are presented in Appendix C. In general, girls reported lower levels of

antisocial characteristics and higher levels of prosocial characteristics, and were considered by

peers to be more well-liked and popular than boys. Girls were also more likely than boys to

endorse the stereotypes of girls and popular students as indirect aggressors. Finally, boys showed

greater levels of both self- and peer-reported direct aggression than girls, while girls had higher

levels of self-reported indirect aggression than boys. (The relationship of gender to aggression is

further addressed in group-based analyses below.)

Race

Fewer significant differences within study variables existed for racial groups. All

significant findings are presented in Appendix C. African American students showed slightly









lower levels of empathy than European American students and had slightly lower social

preference scores. Differences that did exist tended to be within stereotyping scales: African

American students were less likely than other students to endorse stereotypes of African

Americans as direct and indirect aggressors as well as less likely to endorse the stereotype of

girls as indirect aggressors. African American students also had higher self-reported direct

aggression scores than did other students, and were more likely than other students to be

nominated by peers as both directly and indirectly aggressive. (The relationship of race to

aggression is further addressed in group-based analyses below).

Grade

One-way analyses of variance revealed several grade-level differences on relevant study

variables, including stereotyping scales and behavioral and emotional characteristics. All

significant findings are presented in Appendix C. The majority of findings seemed due to

differences between the 6th graders and the other two grades. For instance, 6th graders tended to

report lower levels of antisocial characteristics than did students in higher grades. No significant

differences by grade level emerged for self-reported direct or indirect aggression (peer reported

aggression variables were standardized by grade).

Form order

The presentation of measures was counterbalanced during data collection, with half of

each grade completing stereotyping measures the first week and aggression measures the second

week (and vice versa). To test for possible form order effects, a series of t-tests of independent

means was conducted for all relevant study variables. Only two significant effects emerged:

students who completed stereotyping measures last were slightly more likely to endorse the

stereotype of girls as indirect aggressors (M difference = .25, t = 2.57, p < .05) and students who









completed aggression measures last nominated a higher percentage of female classmates as

indirect aggressors (M difference = 14%, t = 2.23, p < .05).

Summary

Social desirability, gender, race, and grade were identified as necessary covariates for

analyses addressing Question 1 (below), due to significant relationships with aggression

variables used to assign group membership and/or at least one of the behavioral, emotional, or

sociometric outcome variables. Gender, race, grade, and form order were also related to

stereotyping scores and peer nomination demographic percentages, and thus were included

where relevant in analyses addressing Question 2 (below).

Analyses Addressing Question 1: Do Self-Identified and Peer-Identified Aggressive
Adolescents Show Distinct Patterns of Psychosocial and Demographic Characteristics?

Behavioral & Emotional Characteristics

As indicated, it was hypothesized that higher levels of antisocial emotional and

behavioral characteristics would characterize self-identified aggressive adolescents, whereas

peer-identified aggressive adolescents would resemble nonaggressors in these characteristics.

The significant relationships between the five behavioral and emotional characteristics suggested

the need for a multivariate approach to examine group mean differences. For both direct and

indirect aggression, these variables were analyzed in multivariate analyses of covariance

(MANCOVAs) as well as planned univariate follow-up tests for specific group differences on

individual variables. All analyses included social desirability, gender, race (coded as African

American vs. other) and grade level as covariates due to the significant relationships found in

preliminary analyses. Estimated marginal group means and 95% confidence intervals for

individual behavioral and emotional characteristics are graphically represented in Figure 3-1 for

both direct and indirect aggressor groups.









Direct Aggression

The multivariate effect for direct aggression was significant, Wilkes' A = .81,

F(15,731.95) = 3.94, p > .001. Univariate analyses revealed that the multivariate effect was

primarily driven by significant group differences for manipulative behavior (F = 11.31, p <

.001), remorselessness (F = 6.70, p < .001), anger (F = 6.43, p < .001), and, to a lesser extent,

anger regulation (F = 2.71, p < .05). Empathy did not have a significant univariate effect.

Estimated marginal means comparisons (Figure 3-1) indicated that self-identified

aggressive adolescents (the high self and high multiple groups) had significantly higher levels of

manipulative behavior than either the high peer group or the low aggression group (M

differences = .43 to .54; p < .01). The high peer group and the low aggression group did not

significantly differ from one another. For remorselessness, self-identified aggressive adolescents

(the high self and high multiple groups) had significantly higher levels of remorselessness than

did the low aggression group (M differences = .45 to .50; p < .01), whereas solely peer-identified

aggressors (the high peer group) did not significantly differ in remorselessness from the low

aggression group. Similarly for anger, the high self and high multiple groups had higher levels of

anger than did the low aggression group (M differences = .44 to .48; p < .01), whereas the high

peer and low aggression groups did not significantly differ from one another. Finally, for anger

regulation, the significant univariate effect for anger regulation was primarily driven by lower

levels in the high multiple group compared to all other groups (M differences = .37 to .41; p <

.01); no other significant group differences were present.

Summary. The overall pattern of findings for direct aggression groups suggested that

self-identified aggressors (the high self and high multiple groups) had the highest levels of

antisocial behavioral and emotional characteristics. In addition, solely self-identified aggressors

had higher levels of manipulative behavior than solely peer-identified aggressors. Although the









high self group differed from low aggressors on three of the five variables, no significant

differences between the high peer and low direct aggression groups were present.

Indirect Aggression

The multivariate effect for indirect aggression was significant, Wilkes' A = .84,

F(15,731.95) = 3.15, p > .001. In a similar pattern to that of direct aggression, univariate

analyses revealed that the multivariate effect was primarily driven by significant group

differences for manipulative behavior (F[3,281] = 11.53, p < .001), remorselessness (F[3,281] =

6.21,p < .001) and anger (F[3,281] =3.15,p <.05). No significant univariate effects existed for

anger regulation or empathy. Estimated marginal means comparisons (Figure 3-1) indicated that

self-identified aggressive students (high self and high multiple groups) had significantly higher

levels of manipulative behavior than both the high peer group or the low aggression group (M

differences = .44 to .68; p < .01). Solely peer-identified aggressive students (the high peer group)

did not differ from the low aggression group.

The significant univariate effect for remorselessness was primarily driven by significantly

elevated levels in the high multiple group compared to all other groups (M differences = .50 to

.69; p <.01). Similar to remorselessness, the high multiple group had significantly elevated

levels of anger compared to both the low aggression and high peer groups. However, the high

multiple group did not differ from the high self group in anger levels.

Summary. The overall pattern of findings for direct aggression groups suggested that

indirect aggressors identified both by themselves and by their peers had the highest levels of

antisocial behavioral and emotional characteristics. In addition, solely self-identified aggressors

had higher levels of manipulative behavior than solely peer-identified aggressors. Again, no

significant differences between the high peer and low aggression indirect groups were present.









Demographic and Sociometric Characteristics

It was hypothesized that peer-identified aggressive adolescents would be distinguished

from self-identified or nonaggressive adolescents by demographic and sociometric patterns

which evoke common social stereotypes of direct and indirect aggression. Specifically, the high

peer group for direct aggression was expected to contain greater percentages of males and

African Americans than the high self group for direct aggression. In addition for indirect

aggression, the high peer group was expected to contain a greater percentage of females and to

have higher social visibility than the high self group.

Demographics

Group differences in gender and racial distributions were each assessed via Pearson 2

analyses. These were performed in three ways. First, an overall X2 tested for differences in race

or gender in the full sample using all four possible groups, including the low aggression group.

Pending a significant effect in the overall test, comparisons between groups were tested first by

comparing the low aggression group to all other students and then by comparing only high

aggression groups while excluding the low aggression group. These follow-up tests provided

information as to whether significant effects were a result of differences between the specific

methods (self or peer report) used to measure aggression, or whether they were primarily driven

by differences between nonaggressors and aggressive students identified by any measurement

type.

Direct aggression: gender. The overall analysis for gender was significant, 2(3) =

20.18, p < .001. This indicated that a significant relationship existed between gender and direct

aggression group membership. Observed and expected cell counts for the overall test are

provided in Table 3-4. Follow-up analyses revealed that the significant overall effect was

primarily driven by differences between students low in aggression and students identified as









high in aggression by either method (2(1) = 18.14, p < .001), rather than by differences between

high aggression groups (O2(2) = 2.28, p > .05). This suggests that boys were especially likely to

be identified as direct aggressors, regardless of the reporting type used.

Direct aggression: race. The overall analysis for race/ethnicity was significant, /2(6) =

30.08, p < .001. This indicated that a significant relationship existed between race/ethnicity and

aggression group membership. Observed and expected cell counts for the overall test are

provided in Table 3-4. Follow-up chi-squares revealed that, similar to gender, the significant

overall effect was primarily driven by differences between students low in aggression and

students identified as high in aggression by either method Q(2(2) = 17.56, p < .001). However, a

non-random racial group distribution also existed among high aggression groups (2(4) = 10.04,

p > .05). Distribution across aggression groups for African American students showed the most

deviation from expected cell counts: African American students accounted for 15.8% of the low

aggression group and 18.9% of the high self group, but accounted for 39.0% of the high peer

group and 51.4% of the high multiple group.

Indirect aggression: gender. The overall analysis for gender was significant, /2(3) =

11.66, p < .01. This indicated that a significant relationship existed between gender and direct

aggression group membership. Observed and expected cell counts for the overall test are

provided in Table 3-4.Similar to direct aggression, follow-up chi-squares revealed that the

significant overall effect was primarily driven by differences between students low in aggression

and students identified as high in aggression by either method (2[1] = 10.14, p < .01), rather

than by differences between high aggression groups Q(2[2] = .92, p > .05). This suggests that

girls were especially likely to be identified as indirect aggressors, regardless of the reporting type

used.









Indirect aggression: race. The overall analysis for race was significant, /2(6) = 14.94, p

< .05. This indicated that a slight but significant relationship existed between race/ethnicity and

indirect aggression group membership; however, both follow-up /2 tests failed to reach

significance at thep < .05 level. This suggests that the significant overall /2 effect may have

resulted from unexpected distributions across the low aggressors and potentially one other high

aggression group. Observed and expected cell counts for the overall test are provided in Table 3-

4; the largest disagreements between observed and expected cell counts are as follows: Fewer

African American students than expected were members of the low aggression group (18.7%),

whereas more African American students than expected were members of the high peer group

(33.3%) and high multiple group (33.3.%), and fewer "Other" students than expected (9.7% ,

compared to 27.9-33.3% in other groups) were members of the high peer group.

Summary. Although nonrandom gender and racial distributions were present across both

direct and indirect aggression groups, they were mainly driven by differences between students

who had low aggression levels and students who were identified as aggressive by at least one

type of report method. However, nonrandom distributions of racial groups were evident for

direct aggression: the percentage of African American students within the high peer and high

multiple groups was substantially larger than within the low aggression or high self groups.

Although hypothesized to be present, no gender differences existed between high self and high

peer groups for indirect aggression.

Sociometric Characteristics

Significant relationships between perceived popularity, social preference, and social

impact suggested the need for a multivariate approach to examine group mean differences.

MANCOVAs were run for both direct and indirect aggression, as well as univariate planned

follow-up tests for specific group differences on individual variables. All analyses included









gender and race as covariates due to their significant relationships with the distribution of

aggressor groups and at least one of the outcome variables. Since all outcome variables were

standardized by grade before being analyzed, grade level was not included as a covariate. For

both direct and indirect aggressor groups, estimated marginal group means and 95% confidence

intervals for individual sociometric characteristics are graphically represented in Figure 3-2.

Direct aggression. The multivariate effect for direct aggression was significant, Wilkes'

A = .70, F(9,737.57) = 13.03, p > .001). Univariate tests revealed that social preference

(F[3,305] = 12.18, p < .001), social impact (F[3,305] = 26.12, p < .001), and perceived

popularity (F[3,305] = 4.37, p < .01) all showed evidence of mean differences across aggression

groups. Furthermore, estimated marginal means comparisons revealed the same pattern of group

differences for all three variables: students in the high self and high multiple groups had lower

levels of social preference, higher levels of social impact, and higher levels of perceived

popularity than students in the low aggression and high self groups (M differences = .69 to 1.19,

p < .05). No significant differences existed on any sociometric characteristic between the high

self and high multiple groups, nor between the low aggression and high self groups.

Indirect aggression. The multivariate effect for indirect aggression was significant

(Wilkes' A = .66, F(9,737.57) = 15.34, p > .001), and univariate tests revealed that social

preference (F[3,305] = 11.86,p < .001), social impact (F[3,305] = 27.97,p < .001), and

perceived popularity (F[3,305] = 8.07, p < .01) all showed evidence of mean differences across

aggression groups. The high peer group had higher social impact and lower social preference

scores than both the high self or low aggression groups, and the high multiple group had higher

social impact and lower social preference scores than all other groups. Differences for perceived

popularity followed the same pattern as for direct aggression: The high peer and high multiple









groups were perceived as significantly more popular than either the high self or low aggression

groups. Again, no significant differences existed between the high self and low aggression

groups for any sociometric characteristic.

Summary. Significant differences between exclusively self-identified and exclusively

peer-identified aggressive adolescents existed for all sociometric characteristics. For both direct

and indirect aggression, peer-identified aggressors were more popular, more socially visible, and

less well-liked than those identified solely through self-report. In addition, solely self-identified

aggressors did not differ from the low aggression group on any sociometric characteristic,

regardless of aggression type.

Analyses Addressing Question 2: Is the Tendency to Endorse Particular Aggression
Stereotypes Related to Nominations of Aggressive Peers?

Stereotyping Scores and Nominated Peers' Gender, Race, and Popularity

It was hypothesized that participants' stereotyping scores would be related to

characteristics of peers nominated as directly or indirectly aggressive. To test this, a series of

hierarchical regression analyses were conducted; direct and indirect aggression were examined

separately for each of the three stereotyping categories, resulting in a total of six analyses. For

direct and indirect stereotyping scales relating to gender, the outcome variable in question was

the percentage of males that an individual nominated on direct or indirect peer-reported

aggression items; for scales relating to race, it was the percentage of African American students .

For scales relating to popularity, the outcome variable was the mean perceived popularity score

of nominated peers.




1 The "African-American vs. other" dichotomous distinction was chosen for analysis due to the fact that preliminary
analyses suggested that the European American and "other" racial categories tended to resemble one another on
aggression and stereotyping measures.









For each analysis, covariates including gender, race (dummy coded as African American

vs. all others), grade level, and form order were entered in Step 1; however, only those covariates

which demonstrated a significant effect in the full model were retained in the final analysis. The

appropriate stereotyping scale, which corresponded to the same aggression type and

demographic or sociometric category as the outcome variable, was entered in Step 2. In step 3,

an interaction term was added to determine whether the strength of these relationships varied

across gender and racial categories or as a participant's own perceived popularity increased.

Direct aggression. Table 3-5 presents the results of the analysis for gender, which

compared participants' scores on the girls-boys direct aggression stereotyping scale to the

gender distribution of the peers whom they nominated as directly aggressive. Gender

distribution was entered as the percentage of nominations that were of male peers. No effect of

race, grade level, or form order was present, so these covariates were dropped from the final

model. After controlling for gender, girls-boys stereotyping scores explained unique variance

in the gender distribution of nominated peers, = .26, p < .001; AR2 = .07. Students who were

more likely to endorse the view that boys were direct aggressors were more likely to nominate

male peers on the direct aggression items, and students who were more likely to endorse the

view that girls were direct aggressors were more likely to nominate female peers on the direct

aggression items. Furthermore, girls and boys were equally likely to demonstrate this

relationship: there was no interaction effect between respondent's gender and stereotyping score,

f = .04,p > .05, A R2 < .01.

Gender was the only category in which there existed a relationship between direct

aggression stereotyping and corresponding characteristics of nominated peers. After controlling

for covariates, analyses failed to demonstrate significant unique effects or interaction effects for









black-white (f = .01 to .04,p > .05, A R2 < .01) or popular-unpopular (f = -.02,p > .05, A R2

< .01) stereotyping scores.

Indirect aggression. Table 3-5 also presents the results of the analysis of gender for

indirect aggression, which compared participants' scores on the girls-boys indirect aggression

stereotyping scale to the gender distribution of the peers whom they nominated as indirectly

aggressive. The indirect aggression analysis followed a similar pattern as that for direct

aggression. No effect of race, grade level, or form order was present, so these covariates were

dropped from the final model. After controlling for gender, girls-boys stereotyping scores

explained unique variance in the gender distribution of nominated peers, f = .21, p < .01, A R2

.05. Students who were more likely to endorse the view that boys were indirect aggressors were

also more likely to nominate male peers on the indirect aggression items, and students who were

more likely to endorse the view that girls were indirect aggressors were more likely to nominate

female peers on the indirect aggression items. Overall, students tended to endorse the stereotype

that girls engaged in indirect aggression rather than boys. Also similar to direct aggression, girls

and boys were equally likely to demonstrate this relationship: there was no interaction effect

between gender and stereotyping score, f = -.07, p > .05, A R2 < .01.

Gender was the only category in which there existed a significant relationship between

indirect aggression stereotyping and corresponding characteristics of nominated peers. After

controlling for covariates, analyses failed to demonstrate significant unique effects for black-

white (f = -.07, p > .05, A R2 < .03) or popular-unpopular (f =-.02, p > .05, A R2 < .01)

stereotyping scores.

Summary. Students' scores on aggression stereotyping scales appeared to be related to

the gender of peers nominated as aggressive, and this was true for both direct and indirect









aggression. Furthermore, this relationship did not vary by the gender of the nominating student.

Gender was the only stereotyping category for which this relationship was present. No

significant relationships were found between aggression stereotyping and either the race or

popularity of nominated peers.

Stereotyping Scores and Aggression Group Membership

Prior analyses tested whether social schemas influenced the peers whom students

nominated as aggressive. Additionally, it was hypothesized that scales measuring aggression

stereotyping might prove to be useful as a control variable for future studies employing peer-

reported aggression measures, similar to the way in which social desirability bias measures are

currently used to explain variance in self-reported aggression.

The efficacy of social desirability bias in accounting for discrepancies between

aggression measurement methods was supported in the current sample. A one-way ANOVA

comparing aggression groups on mean social desirability scores indicated that, for both direct

and indirect aggression, the high self group had significantly lower levels of social desirability

bias (M= .36, SD = .04) than all other groups (M= .49 to .53, SD = .02 to .04; F(3,292) = 5.37,

p <.01). Thus, students in the high self group, who had been identified as aggressive by

themselves only, could be distinguished from students who had been identified as aggressive by

multiple sources by an increased willingness to report socially undesirable behavior.

It was similarly hypothesized that students who were identified as aggressive solely by

peers might be distinguished from other students by the degree of aggression stereotyping

present in the peers who had nominated them in other words, whether students in the high peer

group were more likely than those in other groups to be the targets of aggression stereotyping

bias. If a student was nominated as aggressive, it was possible to calculate the average

stereotyping scores on each scale of the peers who nominated that student. A series of two-way









ANOVAs and planned follow-up comparison tests of estimated marginal means investigated

group differences in nominators' average stereotyping scores. Analyses were run separately for

direct and indirect aggression. Because nominators' average stereotyping scores were expected

to skew in opposing directions for each opposing pair of social groups which comprised the

endpoints of the scales, the interaction between group membership and the nominees'

corresponding gender, racial, or sociometric category was also assessed. As an example, for the

analysis of aggression group differences in nominators' average scores on the girls-boys direct

aggression stereotyping scale, group membership and gender were both entered into the model as

predictors, and the main effects of each were investigated along with the effect of the interaction

term.

Due to the fact that not all students received nominations on peer-report aggression items

(and thus could not be assigned average nominators' stereotyping scores), these analyses

included only those students who were nominated at least once on the peer-report aggression

measure. One hundred sixty-three students were nominated at least once on direct items and thus

included in analyses of direct aggression, and 184 students were nominated at least once on

indirect items and thus included in analyses of indirect aggression. Note that the high peer report

and combined (peer and self) report groups had to have 4 or more nominations in order to be

classified as "high" on peer report. Thus, students who received at least one but fewer than four

nominations comprised a nominated group that was compared to the high peer report and

combined groups (N 85 for direct aggression and 103 for indirect aggression).

Findings. Of the race, gender, and popularity variables, only gender showed significant

main effects for differences in nominators' average scores on the girls-boys stereotyping scales;

this was true for both direct (F[1,155] = 8.36,p < .01) and indirect aggression (F[1,174] = 7.95,









p < .01). While the nominators of both boys and girls tended to have scores that skewed farther

toward the boys' side for direct aggression, this was more true for the nominators of boys (M=

1.20, SD = .07) than for girls (M= .86, SD = .10). Similarly, while the nominators of both boys

and girls tended to have scores that skewed farther toward the girls' side for indirect aggression,

this was more true for the nominators of girls (M= 1.08, SD = .08) than for boys (M= .73, SD =

.10). No main effects were found for race or for perceived popularity.

Importantly, there were no significant aggression group differences in any of the

nominators' average stereotyping scores, nor were any significant interaction effects present

across analyses. (Because of the absence of significant findings, results from individual analyses

are not reported here for the sake of brevity.) The absence of increased stereotyping bias in the

nominators of students in the high peer group compared to other groups suggests that, unlike

social desirability bias, the aggression stereotyping scales cannot be used to explain

discrepancies between self- and peer-reported aggression scores.

Nominators' Average Stereotyping Scores and Levels of Peer-Reported Aggression

A final set of analyses examined whether nominators' stereotyping scores explained

variance in peer-reported aggression over and above what was already explained by self-reported

aggression scores. Although the previous set of analyses suggested that the use of stereotyping

scales in this role may be limited, a main impetus of the current study was the possibility that

stereotyping scales could be utilized as a control variable to explain variance in peer-reported

aggression. Thus, I re-examined this relationship using continuous forms of self- and peer-

reported aggression, as most studies of aggression in adolescence do not employ a group-based

approach.

A series of hierarchical linear regressions were conducted to examine whether

nominators' stereotyping scores explained variance in peer-reported aggression over and above









what was already explained by self-reported aggression scores. Again, these analyses included

only those students who were nominated at least once on the peer-report aggression measure.

Direct and indirect aggression were investigated separately.

In each analysis, peer-reported aggression was the dependent variable. Gender, race

(entered as African American vs. all others), and perceived popularity were entered in Step, 1,

and self-reported aggression was entered in Step 2. Together, gender, race, and perceived

popularity explained 18.7% of the variance in peer-reported direct aggression and 18.0% of the

variance in peer-reported indirect aggression. After accounting for these variables, self-reported

direct aggression predicted an additional 2.3% of the variance in peer-reported direct aggression

(/8= .16, F[1,153] = 11.81,p < .001); and self-reported indirect aggression predicted an

additional 2.0% of the variance in peer-reported indirect aggression (8 = .15, F[1,172] = 4.24, p

< .01).

In step 3 of each analysis, nominators' average stereotyping scores on a particular scale

(e.g., girls-boys direct aggression) were entered along with the interaction between the scale

and its corresponding demographic or sociometric category (gender, race, or perceived

popularity). This was run six times in order to investigate the impact of each stereotyping scale

separately.

One marginally significant effect emerged. Nominators' average scores on the girls-

boys indirect aggression stereotyping scale was marginally related to peer-reported indirect

aggression (/ = .13, p = .067). The interaction between gender and nominators' stereotyping

scores was also marginally related to peer-reported indirect aggression (f = -.12, p = .069),

indicating that the relationship between nominators' scores and peer-reported indirect aggression

was slightly stronger for boys than it was for girls. Together, they accounted for an additional









3% of variance in peer-reported aggression (F[2,170] = 3.44, p < .001). No effects were

significant for any other category of nominators' average stereotyping scores.











Table 3-1. Group Ns and Aggression Means and Standard Deviations by Aggressor Group
Aggression Type

Self-Reported Peer-Reported Self-Reported Peer-Reported
Group N Direct Direct Indirect Indirect
Direct aggression
Low 200 1.35 (.28) -.33 (.12) 1.54 (.42) -.27 (.50)
High self 37 2.50 (.48) -.29 (.11) 1.96 (.55) -.25 (.49)
High peer 41 1.47 (.30) .93 (1.67) 1.68 (.44) .94 (1.67)
High multiple 35 2.61 (.57) 1.17 (1.68) 1.86 (.38) .89 (1.43)
Indirect
aggression
Low 179 1.48 (.49) -.29 (.20) 1.40 (.27) -.43 (.19)
High self 53 1.97 (.76) -.15 (.62) 2.24 (.29) -.35 (.23)
High peer 54 1.66 (.54) .63 (1.55) 1.51 (.24) 1.05 (1.16)
High multiple 27 2.04 (.71) 1.04(1.94) 2.32 (.36) 1.64(1.61)
Note. Self-reported values represent mean scores on 1-5 Likert scale. Peer reported values have been standardized
so that the full sample mean for each variable = 0; SD = 1.










Table 3-2. Membership agreement for direct and indirect aggression
groups
Indirect Groups
High
Direct Groups Low High self High peer multiple
Low 141 29 22 8
High self 16 16 3 2
High peer 13 4 16 8
High multiple 9 4 13 9
Note. Values represent the number of participants falling into each category.










Table 3-3. Means and Standard Deviations for Aggression and Aggression
Stereotyping
Measure/Variable Direct Indirect
Aggression
self-reported 1.64 (.61) 1.64 (.46)
peer-reported (unstandardized) 4.1 (10.78) 2.7 (5.20)
Aggression stereotyping
boys-girls .96 (.91) -.81 (.84)
black-white -.72 (.89) .17 (.71)
popular-unpopular -1.15 (1.01) -1.43 (.96)
Note. Standard deviations in parentheses. Values for aggression stereotyping scales have
been centered so that 0 represents a neutral view, negative scores represent a shift in the
direction of the left-hand (first-listed) group, and positive scores represent a shift in the
direction of the right-hand (second-listed) group.










Table 3-4. Observed and Expected Gender and Race Distributions across Aggression Groups
Direct Aggression Indirect Aggression

High High High High High High
Category/Level Low Self Peer Multiple Low Self Peer Multiple


Gender
Boys
observed (expected)
% within aggression group
Girls
observed (expected)
% within aggression group
Race/Ethnicity
European American
observed (expected)
% within aggression group
African American
observed (expected)
% within aggression group
Other
observed (expected)
% within aggression group


84 (102)
41.4%


119(101)
58.6%




114(102)
56.2%


32 (47)
15.8%


57 (54)
28.1%


21(19)
56.8%


16(18)
43.2%




18(18)
48.6%


7(9)
18.9%


12 (10)
32.4%


29 (21)
70.7%


12 (20)
29.3%




14 (21)
34.1%


16(10)
39.0%


11 (11)
26.8%


25 (18)
71.4%


10(17)
28.6%




13 (18)
50.3%


18(8)
51.4%


4(10)
11.4%


106 (92)
58.2%


76 (90)
41.8%




91(92)
51.6%


34 (42)
18.7%


54 (48)
29.7%


19 (27)
35.8%


34 (26)
64.2%




25(27)
47.2%


12 (12)
22.6%


16(14)
30.2%


24 (27)
44.4%


30 (27)
55.6%




31(27)
57.4%


18(13)
33.3%


5(14)
9.3%


10(14)
37.0%


17(13)
63.0%




9(14)
33.3%


9 (6)
33.3%


9 (7)
33.3%










Table 3-5. Summary of hierarchical regression analyses for participants' gender
stereotyping scales predicting gender percentages of nominated peers
Direct Indirect
Aggression Aggression
Variable AR2 f2
Step 1 .10*** .28***
gender (female vs. other) -.35*** -.50***
grade level -.05 -.24***
Step 2 .06*** .04**
gender (female vs. other) -.32*** -.47***
grade level -.01 -.23***
girls-boys stereotyping score .23** .20**
Step 3 <.01 <.01
gender (female vs. other) -.34*** -.47***
grade level <.01 -.24***
girls-boys stereotyping score .23** .21**
gender X girls-boys stereotyping score .10 -.07
***p <.001.**p < .01.
Note. Dependent variables = % males out of the total number of nominated peers for direct and indirect
aggression items, respectively. Only covariates which remained significant in at least one of the final models
have been reported here.












Manipulative behavior


Direct Indirect


Anger


Direct


Indirect


Remorselessness


Direct Indirect


Anger regulation


Direct


Indirect


Empathy


Group
35
3 Low
25 High self
2 High peer
15 D High multiple

Direct Indirect


Figure 3-1. Graphical representation of group means by aggression type for behavioral and
emotional characteristics. Direct and indirect aggression groups were examined in
separate analyses. Values are mean scores on 1-5 Likert Scales. Brackets around
each mean represent 95% confidence intervals. Significant group differences atp <
.05 are represented by pairs of confidence interval brackets which overlap by <
-30%; significant group differences at p<.01 are represented by pairs of confidence
interval brackets which show no overlap (Cumming & Finch, 2005).


40










Social preference


Direct


Perceived popularity


Group
] Low
SHigh self
High peer
D High multiple


Figure 3-2. Graphical representation of group means by aggression type for sociometric
characteristics. Direct and indirect aggression groups were examined in separate
analyses. Variables are standardized so that the sample mean = 0; SD = 1. Brackets
around each mean represent 95% confidence intervals. Significant group differences
atp < .05 are represented by pairs of confidence interval brackets which overlap by <
-30%; significant group differences at p<.01 are represented by pairs of confidence
interval brackets which show no overlap (Cumming & Finch, 2005).


Direct


Indirect


Indirect


+4


Direct


Indirect


Social visibility


==


=ME


-0









CHAPTER 4
DISCUSSION

The present study addressed two issues concerning the identification of aggression in early

adolescence. First, to determine whether self-report and peer-report measurement methods

identify students with differing psychosocial profiles, groups of self-identified, peer-identified,

self and peer-identified, and low aggressors were compared on several emotional, behavioral,

sociometric, and demographic characteristics. Second, the incidence of aggression stereotyping

bias among early adolescents was examined, along with its relationship to the demographic and

sociometric characteristics of adolescents identified as aggressive by their peers, and its potential

use as a control variable to explain variance in peer-reported aggression.

Do Self- and Peer-Report Aggression Measures Identify Students with Different
Psychosocial Profiles?

Although self- and peer-reported measures of aggression are usually intended to be

measurements of the same underlying construct, research on aggressive behavior in childhood

and adolescence has found relatively low agreement among the two types of measurement

methods. This suggests that self- and peer-report methods of assessing aggression are

identifying different groups of students. One goal of the present study was to determine whether

students identified as aggressive by self- and peer-report methods differed from one another on a

range of key psychosocial characteristics.

A group-based analytic approach was used to identify students who were high in self-

reported aggression ("high self'), high in peer-reported aggression ("high peer"), high in both

self- and peer-reported aggression ("high multiple"), or high in neither ("low"). Groups were

created for both direct and indirect forms of aggression and compared with one another on

several behavioral, emotional, sociometric, and demographic indicators. It was hypothesized

that self-identified aggressive students would have elevated levels of antisocial emotional and









behavioral traits, while peer-identified aggressors would be characterized by patterns of

demographic and sociometric characteristics which resembled common social stereotypes of

aggressive adolescents (e.g., "mean girls"). Specifically, peer-identified direct aggressors were

expected to be characterized by higher percentages of male and African American students,

while peer-identified indirect aggressors were expected to be characterized by higher percentages

of female students, lower levels of social preference, and higher levels of social impact and

perceived popularity.

High Self vs. High Peer: Antisocial Indicators

Of particular interest in the analyses were comparisons of the high self and high peer groups -

that is, students who were identified solely by self report or solely by peer report. Results

suggested that, after controlling for social desirability and relevant demographic indicators, such

as gender, race, and grade level, self-identified and peer-identified aggressive adolescents

differed somewhat from one another on emotional and behavioral indicators of antisocial

personality. For both direct and indirect aggression, the high self group showed significantly

higher levels of manipulative behavior than the high peer group. In addition, for direct

aggression, the high self group had significantly elevated levels of both anger and

remorselessness over the low aggression group, while the high peer group was indistinguishable

from low aggressors on these variables.

These results may suggest that direct aggression is a better indicator of antisocial

tendencies than indirect aggression. However, on the majority of antisocial indicators, although

the high self group's levels were higher than those of the high peer group, the high self and high

peer groups did not significantly differ from one another. It is possible that this was due to the

fact that the high aggression group sizes were somewhat small, which may have affected the

power of the analyses. Nevertheless, Clemans and Sontag (2009, April) found significant









differences between high self- and high peer-report groups with similar group sizes. As such, the

hypothesis that self- and peer-identified aggressors would show distinct differences on indicators

of antisocial personality was partially supported by the present study.

High Self vs. High Peer: Sociometric and Demographic Characteristics

One area in which hypotheses of group differences were fully supported, however, was

sociometric characteristics. Social standing within the peer group was assessed by social

preference (how well liked one is), social impact (how socially visible one is) and perceived

popularity (how "popular" one is considered by his/her classmates). For both direct and indirect

aggression, all three variables showed similar, pronounced patterns of results, with the high peer

group being less well-liked, more socially visible, and higher in perceived popularity than the

high self group or the low aggressors.

Furthermore, it was expected that group differences in demographic characteristics would

vary between direct and indirect aggression. For gender, this proved to be the case: Regardless

of measurement type, boys were more likely than girls to be identified as direct aggressors, and

girls were more likely than boys to be identified as indirect aggressors. Surprisingly, no

significant gender differences across the high aggression groups were observed for either type of

aggression, although it had been hypothesized that these gender patterns would be significantly

more pronounced in the high peer groups than in the high self groups.

There were, however, differences between the high self and high peer groups in racial

categorization. Specifically, African Americans made up only 19% of the high self group (which

was similar to the percentage of African Americans in the full sample), but comprised nearly

40% of the high peer group. An elevated percentage of African Americans was also seen in the

high peer group for indirect aggression (33%), whereas the high self group again was

comparable to the racial distribution of the full sample.









In summary, it appeared that students identified solely by self-reported measures

(especially those who identified themselves as being high in direct aggression) tended to have

elevated levels of antisocial behavioral and emotional characteristics over students low in

aggression, but resembled low aggressors in demographic and sociometric characteristics. In

contrast, students identified solely by peer-reported measures mimicked students with low

aggression in emotional and behavioral traits, but were distinguished from low aggressors by

lower social preference, higher social impact, higher perceived popularity, and greater

percentages of African American students. Gender, surprisingly, was the only variable category

which did not demonstrate different patterns of findings for self- and peer-reported aggressors.

The High Multiple Group: "True" Aggressors?

There was quite a bit of evidence that, for both direct and indirect aggression, the high

multiple report group was distinct from either of the other two high aggression groups and may

be the most valid representation of aggressive adolescents. This group, for instance, had the

most pronounced levels of antisocial indicators: students in the high multiple group were

significantly different than low aggressors on almost every indicator which demonstrated

differences between groups. For direct aggression groups, this included higher levels of

manipulative behavior, remorselessness, anger, and lower levels of anger regulation; for indirect

aggression groups, higher levels were seen for remorselessness and anger. In some cases, the

high multiple group also showed more elevated levels of antisocial indicators than both of the

other high aggression groups, although patterns suggested that on these indicators, the high

multiple group tended to have levels that more closely resembled those of solely self-identified

aggressors than those of solely peer-identified aggressors.

In sociometric characteristics, however, the high multiple group was indistinguishable

from the high peer group. For both direct and indirect aggression, both the high multiple and









high peer groups were less well-liked, more socially visible, and more popular than the high self

or low aggressors. Additionally, both the high peer and high multiple groups had elevated

percentages of African American students (this was particularly true for direct aggression).

In short, the high multiple group conformed to nearly all characteristics measured in this

study that were found in previous research to be characteristic of aggressive adolescents, while

the high self group tended to be distinguished only on antisocial tendencies and the high peer

group tended to be distinguished only on demographic and sociometric characteristics.

Achenbach (1987) suggested that both self- and peer-report methods may provide unique and

important information about adolescent problem behavior that would not be captured by other

informant methods. As such, the use of multiple informant methods as a general rule for

research on aggressive behavior may be advised. In addition, while several studies have

included peer- and self-reported behaviors, often the decision is made to use only one reporting

type in analyses. The present investigation suggests that this is not an arbitrary decision and will

likely impact the results. Ideally, future studies should report how findings may have varied by

reporter.

Of course, not all research endeavors have the time and monetary resources to collect

aggression data from multiple informants. For those studies utilizing a single measure of

aggression, potential sources of bias become a more salient issue. We know that social

desirability accounts for variance in self-reported aggression, and for this reason it is frequently

utilized as a control variable in studies that employ self-reported aggression measures. On the

other hand, no similar ubiquitous assessment of bias exists for peer-reported aggression

measures. Therefore, the second goal of the current study was to investigate the utility of a

measure designed to assess students' tendency to endorse demographic and sociometric









stereotypes of aggressive behavior, as it was hypothesized that this tendency might result in bias

in nominations of aggressive peers.

Are Stereotyping Scores Related to the Demographic and Sociometric Characteristics of
Peers Nominated as Aggressive?

A series of scales designed to measure direct and indirect aggression stereotyping bias

were created for the purpose of this study. Three scales, each item of which comprised a

spectrum with a dichotomous pair of gender, racial, or popularity-based groups as its endpoints,

described an aggressive behavior and asked students to rate whether each item applied more to

the first group, more to the second group, or equally to both groups. Each scale demonstrated

good distribution and skewed away from center in the direction one would expect according to

common social stereotypes of aggressive adolescents. Students tended to endorse the view that

boys were more directly aggressive than girls, that girls were more indirectly aggressive than

boys, that African American students were more directly aggressive than European American

students, and that popular students were more directly and indirectly aggressive than unpopular

students. The association between popularity and indirect aggression was the strongest of all the

scales. Internal reliability was acceptable for racial and popularity scales, and on the low side

(though not in the unacceptable range) for gender scales.

The relationship of the aggression stereotyping scales to students' tendencies to nominate

particular demographic and sociometric categories of aggressive peers was examined. Effects

were found mostly for those scales which tapped gender stereotyping. Linear regression

analyses indicated that, after controlling for gender, a student's score on the direct gender

stereotyping scale was significantly associated with the gender distribution of the peers he or she

nominated as directly aggressive, and the same was true for indirect aggression. For each scale,

the more one skewed toward the boys' side of the stereotyping spectrum, the more likely one









was to nominate male peers on peer-report aggression items (and vice versa for females). Thus,

one's tendency to endorse a gender-typed view of aggression does appear to explain variance in

the gender of the peers one identifies as being aggressive. Similar results, however, were not

found for race or for popularity.

Although stereotyping scores partially accounted for gender differences in nominated

peers, this relationship did not appear to be helpful in explaining variance in overall levels of

peer-reported aggression. A comparison of students who were solely-peer identified high

aggressors and students who were identified as aggressive by multiple sources revealed no

significant differences in stereotyping scores between the two; furthermore, no group differences

existed when the interactions between stereotyping scores and demographic/sociometric

characteristics were examined. Finally, with the exception of one marginal effect for the indirect

aggression gender scale, stereotyping scores did not explain any significant variance in peer

reports of aggression over and above that already explained by participants' self-reported

aggression scores. As such, although the stereotyping scales used in this study do seem to

predict some characteristics of aggressive peer nominations, as a whole they do not appear to be

particularly useful as control variables for peer-reported aggression.

The ineffectiveness of the stereotyping scales to explain variance in levels of peer-

reported aggression might be interpreted in several ways. It was hypothesized that bias due to

social schemas of aggressive adolescents would be associated with the type of peers one

nominated as aggressive, and thus with overall peer-reported aggression levels for particular

demographic and sociometric categories of adolescents. Although the first part appeared to be

true in the case of gender, the study found no significant relationships between levels of peer-

reported aggression and scores on any of the stereotyping scales. These findings make sense in









that among those who stereotype, nominations are influenced by these stereotypes. Yet across

young adolescents, nominations demonstrated expected gender distributions as found in other

studies: that is, boys have higher rates of direct aggression and girls and boys had similar rates of

indirect aggression (e.g., similar numbers of nominations).

Of course, it may also be the case that adolescents' social schemas about aggressive peers

are having a greater influence on peer reports of aggression than found in the present study, but

that the stereotyping scales created for this study are simply not doing a very good job of

accurately describing those schemas. For instance, a scale which allowed one to rate social

groups individually, rather than through comparison format, may have produced a more nuanced

picture of adolescents' views on the relative levels of aggression among social groups. In

addition, the scales only assessed social schemas of gender, race, and popularity; it is possible

that adolescents' social schemas of aggression involve demographic and/or social criteria that

were not assessed, such as socioeconomic status, physical appearance, or membership in specific

social crowds.

Although social desirability bias was not related to the patterns of answers observed for

the stereotyping scales, there was evidence that some students were uncomfortable categorizing

social groups in this way, particularly for the scale which compared racial groups. It is also

possible, then, that the ineffectiveness of stereotyping scales to explain variance in peer-reported

aggression may be due to their relatively explicit format. Since explicit and implicit stereotypes

can differ from one another in direction (Whitley & Kite, 2006), a measure designed to assess

implicit aggression stereotyping may produce different patterns of results and affect overall

conclusions about the role of stereotyping in peer-reported aggression.









Strengths and Limitations

A strength of the present study is that it utilized a passive consent procedure, which is

crucial for research investigating aggression and other socially undesirable behavior in the school

setting (Tigges, 2003). Ninety-six percent of all middle school students at the school participated

in the study. Thus, the peer-report data for this school is considered to be highly reliable (Crick

& Ladd, 1990). However, the investigation of these relationships in only one setting is a

potential limitation and may affect the generalizability of the results. The school in question is

not a traditional public middle school: Although there are no tuition costs and it strives to enroll

a student body that is demographically representative of the larger county, students and their

families have to complete an application process, which necessitates some level of parental

involvement and selection. In addition, due to the fact that it is a research school affiliated with a

large state university, students may have had more experience with educational and

psychological research than students at other schools in the county. Thus, the student body may

differ from other schools in ways which affected the results of the study. Replication of these

findings in other middle school settings would be required to fully address this question.

Additionally, it should be observed that self- and peer-reported aggressors differed from

low aggressors mostly on scales which employed the same measurement type as the aggression

scale in question; that is, self-reported aggressors had elevated levels of self-reported, socially

undesirable behavioral and emotional characteristics, whereas peer-reported aggressors differed

from low aggressors on sociometric variables which were derived from peer nominations. The

possibility of measurement bias is thus a concern. To address the issue of measurement bias

within self-reports, a measure of social desirability bias was incorporated into the survey, which

assessed participants' unwillingness to admit to common but social undesirable thoughts,

feelings and behaviors. Accordingly, social desirability bias was negatively correlated with









scores for manipulative behavior, remorselessness, anger, and both direct and indirect self-

reported aggression. After factoring out variability due to social desirability bias, however,

relationships between self-reported aggression and manipulative behavior, remorselessness, and

anger remained significant. Furthermore, self-reported direct aggression showed significant

negative associations with anger regulation, which was not itself related to social desirability bias

but fit conceptually into a constellation of antisocial characteristics. Together, these findings

increase the likelihood that self-reported results are representative of true findings and not

artifacts of measurement error. At the present time, there is no established construct that can be

controlled for in order to reduce bias in peer reports. Rather, additional research in needed on

factors that influence reporter bias (see below). It should be noted, however, that peer-reported

sociometric variables actually incorporate peers' opinions into their definition (e.g., whom do

you like? Whom do you think is popular?) and thus may be more robust to potential bias than

peer-reported variables which measure behavior.

The present study sought to provide explanations for discrepancies between self-reported

and peer-reported measurement of aggression by assessing the roles of social desirability bias

and aggression stereotyping bias. It should be noted that I left unexamined many other potential

reasons as to why self and peer reports of aggression may differ in adolescence. For instance,

the same behavior may appear aggressive when coming from a member of one particular social

group but not when coming from another. Similarly, students may rate a behavior as aggressive

in an unknown peer, but may not consider the same behavior to be aggressive if performed by a

friend. Discrepancies between self- and peer-reported aggression may also arise from a lack of

awareness on the part of some self-raters to recognize their behavior as aggressive in the eyes of

other students. In addition to examining the effects of social desirability and social stereotyping,









a thorough investigation of the model in Figure 1-1 should also take these factors into account.

Finally, the direction of effects in the relationship between stereotypes and reports of peer

behavior should also be investigated in order to shed light on the origin and maintenance of

social schemas of aggression.

Developmental Considerations

The present study focused on the measurement of aggressive behavior during the early

adolescent period, using a sample which was approximately 11 to 14 years old. This age group

was selected in part because longitudinal investigation of self and peer reports of aggression

indicated that agreement between the two is higher during this particular stage than at any other

time in childhood or adolescence. Thus, estimations of discrepancies between self and peer

reports were expected to be at their most conservative during this time, increasing the likelihood

that the differences found would apply to other stages of childhood and adolescence as well. In

addition, both self and peer reports of behavior are often used with early adolescents due to the

fact they are mature enough to complete lengthy survey measures, yet still attend schools in

which they are familiar with the majority of same-grade peers.

At the same time, early adolescence is a unique period of the life span. For instance, early

to midadolescence is the life stage during which the influence of the peer group on physical

appearance, likes/dislikes, and social behavior is at is strongest (Brown, 2004). This is also a

period when gender intensification, which refers to strengthening in the adherence to traditional

gender roles, is observed in many individuals (Galambos, Almeida, & Petersen, 1990; Hill &

Lynch, 1983; see Clemans, DeRose, Graber, & Brooks-Gunn, 2010, for a review of gender

development in adolescence). Finally, early adolescence marks the end of the period of

pervasive gender self-segregation which begins in middle childhood (Maccoby, 1998). As such,

it is a unique developmental time in a young person's life, and it remains unclear as to whether









relationships between stereotyping and peer reports of aggression undergo systematic changes

across longer age ranges. For instance, the tendency of some early adolescents to intensify in

their endorsement of traditional gender role stereotypes may make gender stereotypes of

aggression more salient at this period than they would be at other developmental periods. Future

studies which examine this relationship should ideally do so at multiple points in childhood and

adolescence to assess possible developmental trends.

Conclusions

The results of the present study suggest that the identification of aggressive individuals is

a multifaceted issue and that adolescents with differing demographic and psychosocial

characteristics may be identified depending on the method of assessment (Graham et al., 2003).

It is important that researchers and educators seeking to identify and/or curb aggressive behavior

in school settings remain aware that assessment methods may be providing different perspectives

on aggressive behavior. The present study investigated discrepancies between self and peer

reports of aggression; however, parent and teacher reports are also commonly utilized in

childhood and early adolescence, and should be similarly examined in regard to their

relationships with other reporting methods. The use of an aggregate measure of aggression

derived from multiple report methods, rather than the reliance on a single measurement type, is

advised for future studies.

The present study also tested the hypothesis that bias due to social schemas of aggressive

adolescence would account for variance in peer reports of aggression. Although the applicability

of the social stereotyping scales used here was limited, research in this area has been extremely

scarce, and future studies which investigate this relationship and other potential sources of peer

report bias are warranted. The endorsement of expected aggression stereotypes by participants in

the present study indicates the continued salience of demographic and social stereotypes in our









society. It is important to understand how these and other biases may influence peer relationships

and promote labeling in the peer group not merited by behavior.









APPENDIX A
SURVEY MEASURES

Aggression Stereotyping

Instructions: The following three sections will ask you about things other kids do. Please make
sure to read the instructions carefully.
For each item, indicate how much you think the statement applies to (group 1) or (group
2). (This instruction was repeated three times at the beginning of each scale; group 1 and
group 2 were replaced with girls or boys, black kids or white kids, and popular kids or
unpopular kids.
Scale: 7-point spectrum; 1 = Mostly true of (group 1); 4 = Equally true of both groups; 7 =
Mostly true of (group 2)

Indirect Aggression Stereotyping
1. Leave other kids out on purpose.
2. Gossip or say mean things about other kids behind their backs.
3. Spread rumors about other kids.
4. Give other kids the silent treatment.
5. Tell other kids they won't be their friend anymore in order to get something they want.

Direct Aggression Stereotyping
1. Tease other kids.
2. Call other kids mean names to their face.
3. Push or shove other kids.
4. Get in fights a lot.
5. Threaten other kids.

Prosocial Filler Items
1. Are nice and friendly to people when they need help.
2. Stick up for kids who are being picked on or excluded.


Peer-Reported Aggression

Instructions: the following section is about other kids in your grade. For each question, write the
first and last names of the kids to whom you think the statement best applies. Do not
write anything else about any person except his or her first and last name. If you are
unsure of how to spell a name, please look it up on the roster provided. You can
nominate the same person for more than one item if you want.
Write the names of kids who...

Peer-Reported Indirect Aggression
1. ... say mean things about other kids behind their backs.
2. ...give other kids the silent treatment.
3. ... spread rumors about other kids to damage their social reputation.
4. ...leave other kids out of activities that those other kids really want to be included in.









5. ...tell other kids they won't be their friend anymore in order to get back at them for
something.

Peer-Reported Indirect Aggression
1....tease other kids to make them angry.
2. ...get into physical fights.
3. ...push or shove other kids.
4. ...threaten to hurt or hit other kids.
5. ... call other kids bad names to their face.









APPENDIX B
CORRELATIONS BETWEEN STUDY VARIABLES











Variable 1 2 3 4 5 6 7 8 9 10


1. Self-reported direct aggression
2. Self-reported indirect aggression
3. Peer-reported direct aggression
4. Peer-reported indirect aggression
5. Direct aggression stereotyping
(gender)
6. Direct aggression stereotyping
(race)
7. Direct aggression stereotyping
(popularity)
8. Indirect aggression stereotyping
(gender)
9. Indirect aggression stereotyping
(race)
10. Indirect aggression stereotyping
(popularity)
11. Manipulative behavior
12. Remorselessness
13. Empathy
14. Anger
15. Anger regulation
16. Social desirability
17. Social preference
18. Social impact
19. Perceived popularity
*p<.05. **p<.01.


.45**
.33** .19**
.19** .24** .70**


.02 -.02


.07 -.03 .10 -.02 -.12*


.02 -.14*


.01 -.07


.03 .06 .03 .00 .09 .16**


.13*


-.10 -.03


.02 -.11 .01 -.05 -.17** .18** .61** .20** .01

.46** .49** .16** .12* -.05 .03 -.02 -.03 .09
.34** .26** .17** .16** -.01 .09 -.09 .04 .04
-.21** -.02 -.15** .04 .12* -.10 .02 .05 .02
.36** .35** .09 .06 -.01 .05 -.10 -.03 .05
-.22** .07 -.16** -.04 .06 -.13* .05 -.08 -.08
-.34** -.44** -.02 -.04 .00 -.02 .04 -.02 -.03
-.14* -.04 -.52** -.38** -.10 -.08 .18** -.06 .00
.14* .17** .58** .60** -.03 -.02 .01 .03 -.05
.06 .17** .19** .39** -.12* -.09 .05 -.04 .03


-.01
.01
-.06
-.14*
-.06
.09
.12*
.07
.11












Variable 11 12 13 14 15 16 17 18
1. Self-reported direct aggression
2. Self-reported indirect aggression
3. Peer-reported direct aggression
4. Peer-reported indirect aggression
5. Direct aggression stereotyping
(gender)
6. Direct aggression stereotyping
(race)
7. Direct aggression stereotyping
(popularity)
8. Indirect aggression stereotyping
(gender)
9. Indirect aggression stereotyping
(race)
10. Indirect aggression stereotyping
(popularity)
11. Manipulative behavior
12. Remorselessness .51**
13. Empathy -.19** -.38**
14. Anger .29** .32** .03
15. Anger regulation -.09 -.19** .36** .04
16. Social desirability -.38** -.25** .01 -.43** .11
17. Social preference -.05 -.20** .18** -.10 .19** .09
18. Social impact .09 .06 .13* .05 .01 .02 -.14
19. Perceived popularity .08 -.09 .10 -.07 .04 -.03 .39** .44**
*p<.05. **p<.01.










APPENDIX C
STUDY VARIABLE MEAN DIFFERENCES FOR GENDER, RACE,


AND GRADE LEVEL


Significant differences within selected study variables GENDER
Measure/Variable Boys M(SD) Girls M(SD) t df p
Aggression
self-reported direct aggression 1.32 (.23) 1.20(.19) 4.64 (311) <.001
self-reported indirect aggression 1.24 (.16) 1.30 (.18) -2.74 (311) .006
peer-reported indirect aggression 1.01 (.40) .87 (.20) 3.98 (315) <.001
Aggression stereotyping
boys-girls (indirect) -.69 (.81) -.93 (.86) 2.44 (297) .015
popular-unpopular (indirect) -1.29 (1.01) -1.56 (.89) 2.50 (297) .013
Peer nominations on aggression items
direct aggression -% male .88 (.23) .70 (.34) 4.62 (234) <.001
direct aggression % African American .42 (.33) .55 (.33) -3.04 (234) .003
indirect aggression -% male .69 (.40) .29 (.36) 7.48 (189) <.001
Behavioral/Emotional Characteristics
manipulative behavior 2.28 (.71) 2.04 (.65) 3.02 (306) .003
remorselessness 2.26 (.76) 2.01(77) 2.74 t'114 .006
empathy 3.32 (.49) 3.73 (.43) -7.70 (299) <.001
anger reduction 2.16 (.76) 2.63 (.76) -5.50 (308) <.001
Sociometric Characteristics
social preference -.18(1.14) .18 (.77) -3.23 (309) .001
Note. Analyses performed on all relevant study variables. Only those significant atp < .05 are listed.











Significant gender differences within selected study variables RACE
African
White American Other
Measure/Variable M(SD) M(SD) M(SD) F df p

Aggression

self-reported direct aggression 1.24 (.20)a 1.33 (.25),b 1.24 (.21)b 4.87 (2,310) .008

peer-reported direct aggression .90 (.28)a 1.13 (.44),b .87 (.21)b 17.10 (2,313) <.001

peer-reported indirect aggression .90 (.33)a 1.05 (.44),b .86 (.38)b 5.67 (2,313) .004
Aggression stereotyping
girls-boys (direct) 1.04 (.93)a .70 (.88),b 1.01 (.84)b 3.78 (2,300) .024
black-white (direct) -.79 (.91), -.36 (.78) ,b -.88 (.88)b 6.59 (2,285) .002
black-white (indirect) .06 (.67)a .37 (.67)a .20 (.77) 4.68 (2,286) .010
Peer nominations on aggression
items
indirect aggression % White .48 (.38)a .30 (.36)a .40 (.38) 3.40 (2,188) .036
Behavioral/Emotional Characteristics
empathy 3.58 (.51)a 3.35 (.47),b 3.57 (.49)b 5.74 (2,298) .004
Sociometric Characteristics
social preference .03 (.97) -.29 (1.10)a .20 (.89)b 4.91 (2,308) .008

Note. Analyses performed on all relevant study variables with post-hoc Games-Howell correction due to unequal group sizes; only
variables with significant differences are listed. Same-subscript pairs within a single variable indicate significant mean
differences atp< .05.










Significant gender differences within selected study variables GRADE LEVEL
Measure/Variable 6thM(SD) 7thM(SD) 8thM(SD) F df p
Aggression stereotyping
girls-boys (direct) 1.09 (.97)a 1.08 (.90)b .70 (.79)a,b 6.19 (2,300) .002
black-white (direct) -.96 (.98)a,b -.59 (.85)a -.62 (.79)b 5.25 (2,285) .006
black-white (indirect) -.06 (.81), .36 (.65)a .18 (.59) 8.69 (2,296) <.001
Peer nominations on aggression items
direct aggression- % African American .51 (.32)a .35 (.33),b .58 (.31)b 10.76 (2,229) <.001
indirect aggression % African American .43 (.40)a .20 (.32)a,b .41 (.38)b 8.28 (2,184) <.001
Behavioral/Emotional Characteristics
manipulative behavior 2.03 (.65)a 2.12 (.64) 2.33 (.75)a 5.31 (2,302) .005
remorselessness 2.02 (.74)a 2.14 (.78) 2.29 (.80)a 3.06 (2,300) .048
anger reduction 2.57 (.85)a 2.23 (.72)a 2.37 (.78) 4.69 (2,307) .010
Note. Analyses performed on all relevant study variables with post-hoc Bonferroni correction; only variables with significant
differences are listed. Same-subscript pairs within a single variable indicate significant mean differences atp< .05.









APPENDIX D
SUMMARY OF QUESTION 1 ANALYSES WITH MORE STRINGENT GROUP CRITERIA

All analysis addressing Question 1 were re-run using group membership cutoff criteria of

one SD for self-reported aggression and 5 nominations for peer-reported aggression (referred to

below as "more stringent"). In general, patterns of results mirrored those of the .5 standard

deviation/4 nomination analyses (referred to below as "less stringent").

Differences in the patterns of results were observed for emotional and behavioral

characteristics only; these are reported below. Patterns of results for demographic and

sociometric characteristics did not change.

Emotional and Behavioral Characteristics

Manipulative behavior. Using the more stringent criteria for direct aggression,

estimated marginal means comparisons indicated that the high multiple group had significantly

higher levels of manipulative behavior than the low aggression group only. With the less

significant criteria, it was higher than both the low aggression and high peer groups. This was

because the estimated marginal mean of the high peer group increased slightly, from 2.09 to

2.23. All other differences remained unchanged.

For indirect aggression, the high multiple group increased slightly, from 2.76 to 2.94,

causing the difference between the high self and high multiple groups to become significant atp

< .05. All other differences remained unchanged.

Remorselessness. Estimated marginal means of the direct high self and high peer groups

increased slightly (from 2.54 to 2.69 for high self; from 2.20 to 2.35 for high peer). This caused

the difference between the high peer and low aggression groups to become significant atp < .05;

other significance patterns remained unchanged.









For indirect aggression, a slight increase in the high peer group resulted in a significant

difference between the high peer and high self groups atp < .05. All other differences remained

unchanged.

Anger. For direct aggression, estimated marginal means of the high self and high peer

groups increased slightly (from 2.88 to 3.10 for high self; from 2.60 to 2.72 for high peer). The

difference between the high self and high peer groups became significant atp < .05; other

patterns remained unchanged. Patterns for indirect aggression remained unchanged.

Anger regulation. The estimated marginal mean of the direct high self group decreased

slightly (from 2.42 to 2.31) and was no longer significantly different than the high multiple

group. Other patterns remained unchanged.

Summary. Use of the more stringent criteria resulted in slight increases in antisocial

behavior for the high self and high peer groups. However, the overall patterns of findings

remained unchanged from the previous analyses in that the high self and high multiple groups

had the highest levels of manipulative behavior, remorselessness, and anger, and the lowest

levels of anger regulation.









LIST OF REFERENCES


Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/adolescent behavioral
and emotional problems: Implications of cross-informant correlations for situational
specificity. Psychological Bulletin, 101, 213-232.

Andershed, H., Kerr, M., Stattin, H., & Levander, S. (2002). Psychopathic traits in non-referred
youths: A new assessment tool. In E. Blauuw & L. Sheridan (Eds.), Psychopaths:
Current internationalperspectives (pp. 131-158). Elsevier: The Hague.

Archer, J. (2004). Sex differences in aggression in real-world settings: A meta-analytic review.
Review of General Psychology, 8, 291-322.

Basow, S. A., Cahill, K. F., Phelan, J. E., Longshore, K., & McGillicuddy-DeLisi, A. (2007).
Perceptions of relational and physical aggression among college students: Effects of
gender of perpetrator, target, and perceiver. Psychology of Women Quarterly, 31, 85-95.

Bronfenbrenner, U. (1979). The ecology of human development. Cambridge, MA: Harvard
University Press.

Brown, B. B. (2004). Adolescents' relationships with peers. In R. M. Lerner & L. Steinberg
(Eds.), Handbook of adolescent psychology (2 ed., pp. 363-394). New York: Wiley.

Buss, A. H., & Perry, M. (1992). The aggression questionnaire. Journal ofPersonality and
Social Psychology, 63, 452-459.

Cantor, N., & Mischel, W. (1977). Traits as prototypes: Effects on recognition memory. Journal
of Personality and Social Psychology, 35, 38-48.

Card, N. A., Hodges, E. V. E., Little, T. D., & Hawley, P. H. (2005). Gender effects in peer
nominations for aggression and social status. International Journal ofBehavioral
Development, 29, 146-155.

Card, N. A., Stucky, B. D., Sawalani, G. M., & Little, T. D. (2008). Direct and indirect
aggression during childhood and adolescence: A meta-analytic review of gender
differences, intercorrelations, and relations to maladjustment. Child Development, 79,
1185-1229.

Center for Disease Control and Prevention. (2008). Health risk behaviors by sex: National
YRBS, 2007. Retrieved April 11, 2009, from
http://www.cdc.gov/HealthyYouth/yrbs/pdf/yrbs07 us disparity_sex.pdf

Cillessen, A. H. N., & Borch, C. (2006). Developmental trajectories of adolescent popularity: A
growth curve modelling analysis. Journal ofAdolescence, 29, 935-959.

Clemans, K. H., DeRose, L., Graber, J. A., & Brooks-Gunn, J. (2010). Gender in adolescence: A
person-in-context approach to gender identity and gender roles. In J. Christler & D.









McCreary (Eds.), Handbook of Gender Research in Psychology. New York: Springer
Publishing Company.

Clemans, K. H., Graber, J. A., Nichols, T. R., Brooks-Gunn, J., & Botvin, G. J. (2007, March).
Predicting aggression from ambiguous scenarios: The contribution of attributions and
emotion processes. Poster presented at the Society for Research in Child Development
Biennial Convention, Boston, MA.

Clemans, K. H., & Sontag, L. M (2009, April). Characteristics of early adolescent aggressors:
Peer vs. individual approaches to measuring aggression. Paper presented at the biennial
meeting of the Society for Research in Child Development, Denver, CO.

Coie, J. D., Dodge, K. A., & Coppotelli, H. (1982). Dimensions and types of social status: A
cross-age perspective. Developmental Psychology, 18, 557-570.

Coie, J. D., Dodge, K. A., & Kupersmidt, J. B. (1990). Peer group behavior and social status. In
S. R. Asher & J. D. Coie (Eds.), Peer rejection in childhood (pp. 17-59). New York, NY:
Cambridge University Press.

Coyne, S.M., Archer, J., Eslea, M. & Liechty, T. (2008). Adolescent perceptions of indirect
forms of relational aggression: Sex of perpetrator effects. Aggressive Behavior, 34, 577-
583.

Cornell, D. G., Peterson, C. S., & Richards, H. (1999). Anger as a predictor of aggression among
incarcerated adolescents Journal of Consulting and Clinical Psychology, 67, 108-115.

Crick, N. R. (1997). Engagement in gender normative versus nonnormative forms of aggression:
Links to social-psychological adjustment. Developmental Psychology, 33, 610-617.

Crick, N. R., & Ladd, G. W. (1990). Children's perceptions of the outcomes of social strategies:
Do the ends justify being mean? Developmental Psychology, 26, 612-620.

Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of
psychopathology. Journal of consulting Psychology, 24, 349-354.

Cumming, G. & Finch, S. (2005). Confidence intervals and how to read pictures of data.
American Psychologist, 60, 170-180.

Dearing, K. F., Hubbard, J. A., Ramsden, S. R., Parker, E. H., Relyea, N., Smithmyer, C. M., et
al. (2002). Children's self-reports about anger regulation: Direct and indirect links to
social preference and aggression. Merrill-Palmer Quarterly, 48, 308-336.

Dodge, K. A., Coie, J. D., & Lynam, D. (2006). Aggression and antisocial behavior in youth. In
N. Eisenberg, W. Damon & R. M. Lerner (Eds.), Handbook of child psychology (6 ed.,
Vol. 3, pp. 719-788). New York: John Wiley & Sons, Inc.

Eisenberg, N. (2000). Emotion, regulation, and moral development. Annual Review of
Psychology, 51, 665-697.









Eisenberg, N., Guthrie, I. K., Cumberland, A., Murphy, B. C., Shepard, S. A., Zhou, Q., et al.
(2002). Prosocial development in early adulthood: A longitudinal study. Journal of
Personality and Social Psychology, 82, 993-1006.

Eisenberg, N., Miller, P. A., Shell, R., McNalley, S., & Shea, C. (1991). Prosocial development
in adolescence: A longitudinal study. Developmental Psychology, 27, 849-857.

Eisenberg, N., Morris, A. S., & Spinrad, T. L. (2004). Emotion-related regulation: The construct
and its measurement In D. Teti (Ed.), Handbook of Research Methods in Developmental
Psychology (pp. 423-442). Oxford: Blackwell

Eisenberg, N., Spinrad, T. L., & Sadovsky, A. (2006). Empathy-related responding in children.
In M. Killen & J. Smetana (Eds.), Handbook of moral development (pp. 517-549).
Mahwah, NJ: Lawrence Erlbaum Associates.

Epkins, C. C., & Myers, A. W. (1994). Assessment of childhood depression, anxiety, and
aggression: Convergent and discriminant validity of self-, parent-, teacher-, and peer-
report measures. Journal ofPersonality Assessment, 62, 364-381.

Epstein, J. A., Botvin, G. J., Diaz, T., Baker, E., & Botvin, E. M. (1997). Reliability of social and
personal competence measures for adolescents. Psychological Report, 81, 449-450.

Fiske, S. T., & Taylor, S. E. (1991). Social categories and schemas. In Social Cognition (2nd ed.,
pp. 96-141). New York: McGraw-Hill, Inc.

Galambos, N. L., Almeida, D. M., & Petersen, A. C. (1990). Masculinity, femininity, and sex
role attitudes in early adolescence: Exploring gender intensification. Child Development,
61, 1905-1914.

Giles, J. W., & Heyman, G. D. (2005). Young children's beliefs about the relationship between
gender and aggressive behavior. Child Development, 76, 107-121.

Graham, S., Bellmore, A., & Juvonen, J. (2003). Peer victimization in middle school: When self-
and peer views diverge. Journal ofApplied School Psychology, 19, 117-137.

Hamilton, D. L., Stroessner, S. J., & Driscoll, D. M. (1994). Social cognition and the study of
stereotyping. In P. G. Devine, D. L. Hamilton & T. M. Ostrom (Eds.), Social cognition:
Impact on social psychology (pp. 292-323). San Diego, CA: Academic Press, Inc.

Henry, D. B., & The Metropolitan Area Child Study Research Group (2006). Association
between peer nominations, teacher ratings, self-reports, and observations of malicious
and disruptive behavior. Assessment, 13, 241-252.

Hill, J. P., & Lynch, M. E. (1983). The intensification of gender-related role expectations during
early adolescence. In J. Brooks-Gunn & A. C. Petersen (Eds.), Girls at puberty:
Biological and psychosocial perspectives (pp. 201-228). New York: Plenum.









Hinduja, S., & Patchin, J. W. (2008). Cyberbullying: An exploratory analysis of factors related to
offending and victimization. Deviant Behavior, 29, 129-156.

Hyde, J. S. (1984). How large are gender differences in aggression? A developmental meta-
analysis. Developmental Psychology, 20, 722-736.

Jolliffe, D., & Farrington, D. P. (2006). Development and validation of the Basic Empathy Scale.
Journal ofAdolescence, 29, 589-611.

Katz, D., & Braly, K. (1933). Racial stereotypes of one hundred college students. Journal of
Abnormal and Social Psychology, 28, 280-290.

Larson, R., & Richards, M. H. (1991). Daily companionship in late childhood and early
adolescence: Changing developmental contexts. Child Development, 62, 284-300.

Loy, P., & Norland, S. (1981). Gender convergence and delinquency. The Sociological
Quarterly, 22, 275-283.

Lynam, D. R. (1996). Early identification of chronic offenders: Who is the fledgling psychopath?
Psychological Bulletin, 120, 209-234.

Maccoby, E. E. (1998). The two sexes: Growing up apart, coming together. Cambridge, MA:
Belknap Press.

Marshall, S. K., Tilton-Weaver, L. C., & Bosdet, L. (2005). Information management:
Considering adolescents' regulation of parental knowledge. Journal ofAdolescence, 28,
633-647.

Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. (2001). Sex differences in antisocial behavior:
Conduct disorder, delinquency, and violence in the Dunedin Longitudinal Study.
Cambridge: Cambridge University Press.

Muris, P., van der Pennen, E., Sigmond, R., & Mayer, B. (2008). Symptoms of anxiety,
depression, and aggression in non-clinical children: Relationships with self-report and
performance-based measures of attention and effortful control. ChildPsychiatry and
Human Development, 39, 455-467.

Nederhof, A. J. (1985). Methods of coping with social desirability bias: A review. European
Journal of Social Psychology, 15, 263-280.

Newcomb, A. F., Bukowski, W. M., & Pattee, L. (1993). Children's peer relations: a meta-
analytic review of popular, rejected, neglected, controversial, and average sociometric
status Psychological Bulletin, 113, 99-128.

Nichols, T. R., Graber, J. A., Brooks-Gunn, J., & Botvin, G. J. (2006). Sex differences in overt
aggression and delinquency among urban minority middle school students. Journal of
Applied Developmental Psychology, 27, 78-91.









Novaco, R. B. (1994). Anger as a risk factor for violence among the mentally disordered. In J.
Monahan & H. H. Steadman (Eds.), Violence and mental disorder: Developments in risk
assessment (pp. 21-59). Chicago: University of Chicago Press.

Orpinas, P., & Frankowski, R. (2001). The aggression scale: A self-report measure of aggressive
behavior for young adolescents. Journal of Early Adolescence, 21, 50-67.

Otten, S., & Stapel, D. A. (2007). Who is this Donald? How social categorization affects
aggression-priming effects. European Journal of Social Psychology, 37, 1000-1015.

Pakaslahti, L., & Keltikangas-Jarvinen, L. (1997). The relationships between moral approval of
aggression, aggressive problem-solving strategies, and aggressive behavior in 14-year-old
adolescents Journal of Social Behavior and Personality 12, 905-924.

Peets, K., & Kikas, E. (2006). Aggressive strategies and victimization during adolescence: Grade
and gender differences, and cross-informant agreement. Aggressive Behavior, 32, 68-79.

Pellegrini, A. D., & Bartini, M. (2000). An empirical comparison of methods of sampling
aggression and victimization in school settings. Journal of Educational Psychology, 92,
360-366.

Prinstein, M. J., Boergers, J., & Vernberg, E. M. (2001). Overt and relational aggression in
adolescents: Social-psychological adjustment of aggressors and victims. Journal of
Clinical Child Psychology, 30, 479-491.

Putallaz, M., Grimes, C. L., Foster, K. J., Kupersmidt, J. B., Coie, J. D., & Dearing, K. (2007).
Overt and relational aggression and victimization: Multiple perspectives within the
school setting. Journal of School Psychology, 45, 523-547.

Reynolds, W. M. (1982). Development of reliable and valid short forms of the Marlowe-Crowne
Scale of Social Desirability. Journal of Clinical Psychology, 38, 119-125.

Ryan, C. S., Judd, C. M., & Park, B. (1996). Effects of racial stereotypes on judgments of
individuals: The moderating role of perceived group variability. Journal ofExperimental
Social Psychology, 32, 71-103.

Sagar, H. A., & Schofield, J. A. (1980). Racial and behavioral cues in Black and White children's
perceptions of ambiguously aggressive acts. Journal ofPersonality and Social
Psychology, 39, 590-598.

Saltaris, C. (2002). Psychopathy in juvenile offenders: Can temperament and attachment be
considered as robust developmental precursors? Clinical Psychology Review, 22, 729-
752.

Spielberger, C. D., Jacobs, G. A., Russell, S. F., & Crane, R. J. (1983). Assessment of anger: The
State-Trait Anger Scale. In J. Butcher & C. D. Spielberger (Eds.), Advances in
personality assessment (2nd ed., pp. 159-187). Hillsdale, NJ: Erlbaum.









Strahan, R., & Gerbasi, K. C. (1972). Short, homogeneous versions of the Marlow-Crowne
Social Desirability Scale. Journal of Clinical Psychology, 28, 191-193.

Tigges, B. B. (2003). Parental consent and adolescent risk behavior research. Journal of Nursing
Scholarship, 35, 283-289.

Whitley, B. E., & Kite, M. E. (2006). The Psychology of Prejudice and Discrimination. Belmont,
CA: Thomson-Wadsworth.

Xie, H., Cairns, R. B., & Cairns, B. D. (2002). The development of social aggression and
physical aggression: A narrative analysis of interpersonal conflicts. Aggressive Behavior,
28, 341-355.

Younger, A. J., Schneider, B. H., & Daniels, T. (1991). Aggression and social withdrawal as
viewed by children's peers: Conceptual issues in assessment and implications for
prevention. Journal of Psychiatry and Neuroscience, 16, 139-145.

Zeman, J., Shipman, K., & Suveg, C. (2002). Anger and sadness regulation: Predictions to
internalizing and externalizing symptoms in children. Journal of Clinical Child and
Adolescent Psychology, 31, 393-398.









BIOGRAPHICAL SKETCH

Katherine Clemans received a B.A. in psychology from Duke University in 2002 and an

M.S. in psychology from the University of Florida in 2007. She is a recipient of the University

of Florida's J. Hillis Miller Presidential Fellowship. Her research interests include the

development of aggression and moral judgment during adolescence.





PAGE 1

1 EXPLAINING DISCREPAN CIES BETWEEN SELF AN D PEER REPORTS OF AGGRESSION IN ADOLES CENCE By KATHERINE HALE CLEMANS A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUI REMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Katherine Hale Clemans

PAGE 3

3 ACKNOWLEDGEMENTS To my advisor, Dr. Julia Graber, thank you for being such a wonderful mentor and role model. T hank you f or all the guidance and professional support, for your flexibility and understanding, and for your unfailing encouragement of all my varied research interests over the years. It has been a true pleasure to work with you and I am grateful to have had the opportunity to be y our student To my committee, Drs. Susan Bluck, Catherine Cottrell, and Patricia Ashton, thank you for the insightful comments, brainstorming meetings, and individual guidance you have provided me Many thanks for your valuable assistance and advice. To m y research assistants, Kathleen Endorf, Joe Orovecz, Jordan Powers, Kimberly Papa, Tarah Parrino, and David Alexander thank you for your assistance in collecting and entering data. I couldnt have accomplished this project in the time that it took withou t such a terrific and competent team. To Dr. Russ Froman and the faculty, staff and students of P.K. Yonge Developmental Research School, thank you for welcoming me into your school and for allowing me to collect this data. And to my family, for their support and love t hank you.

PAGE 4

4 TABLE OF CONTENTS page ACKNOWLEDGEMENTS .............................................................................................................3 LIST OF TABLES ...........................................................................................................................7 LIST OF FIGURES .........................................................................................................................8 ABSTRACT .....................................................................................................................................9 CHAPTER 1 INTRODUCTION ..................................................................................................................11 Measuring Aggres sion in Adolescence ..................................................................................11 Comparisons of Self and Peer Identified Aggressive Adolescents .......................................15 Stereotyping Bias in Peer Nominations of Aggressive Individuals ........................................17 Psychosocial Factors Related to Aggression in Adolescence .................................................19 Emotion and Personality Factors .....................................................................................20 Social Factors ..................................................................................................................21 Demographic Characteristics ...........................................................................................22 The Present Study ...................................................................................................................23 Specific Aims ..........................................................................................................................23 2 METHODS .............................................................................................................................26 Participants .............................................................................................................................26 Procedure ................................................................................................................................26 Measures .................................................................................................................................27 Peer Reported Aggression ...............................................................................................27 Self Reported Aggression ...............................................................................................28 Direct self reported aggression. ...............................................................................28 Indirect self reported aggression. .............................................................................28 Aggression Stereotyping .................................................................................................29 Social Desirability ...........................................................................................................31 So ciometric Categorizations ............................................................................................31 Social preference and impact. ..................................................................................31 Perceived popularity. ................................................................................................31 Emotion and Personality Indicators .................................................................................32 Manipulative behavior ..............................................................................................32 Remorselessness .......................................................................................................32 Empathy ...................................................................................................................32 Anger ........................................................................................................................32 Anger regulation .......................................................................................................33

PAGE 5

5 3 RESULTS ...............................................................................................................................34 Creation of Aggressor Groups ................................................................................................34 Preliminary Analyses .......................................................................................................34 Criteria for Group Membership .......................................................................................36 Descriptive Statistics ..............................................................................................................37 Aggression .......................................................................................................................37 Stereotyping .....................................................................................................................37 Behavioral, Emotional, and Sociometric Characteristics ................................................39 Potential Covariates ................................................................................................................40 Social Desirability ...........................................................................................................40 Gender .............................................................................................................................40 Race .................................................................................................................................40 Grade ...............................................................................................................................41 Form order .......................................................................................................................41 Summary ..........................................................................................................................42 Analyses A ddressing Question 1 ............................................................................................42 Behavioral & Emotional Characteristics .........................................................................42 Direct A ggression ............................................................................................................43 Indirect A ggression .........................................................................................................44 Demographic and Sociometric Characteristics ...............................................................45 Demographics ...........................................................................................................45 Sociometric C haracteristics ......................................................................................47 Analyses A ddressing Question 2 ............................................................................................49 Stereotypin g Scores and Nominated Peers Gender, Race, and P opularity ....................49 Direct aggression ......................................................................................................50 Indirect aggression ...................................................................................................51 Summary ..................................................................................................................51 Stereotyping Scores and Aggression Group M embership ...............................................52 Nomin ators A verage Stereotyping Scores and Levels of Peer Reported A ggression ...54 4 DISCUSSION .........................................................................................................................64 Do Self and Peer Report Aggression Measures I dent ify Students with Different Psychosocial P rofiles? .........................................................................................................64 High Self vs. High Peer: Antisocial I ndicators ...............................................................65 High Self vs. High P eer: Sociometric and Demographic C haracteristics .......................66 The High Multiple G roup: True A ggressors? ..............................................................67 Are Stereotyping S cores R elated t o the Demographic and Sociometric Characteristics of Peers Nominated as A ggressive? ........................................................................................69 Strengths and L imitations .......................................................................................................72 Developmenta l C onsiderations ...............................................................................................74 Conclusions .............................................................................................................................75

PAGE 6

6 APPENDIX A SURVEY MEASURES ..........................................................................................................77 B CORRELATIONS BETWEEN STUDY VARIABLES ........................................................79 C STUDY VARIABLE MEAN DIFFERENCES FOR GENDER, RACE, AND GRADE LEVEL ....................................................................................................................................82 D SUMMARY OF QUESTION 1 ANALYSES WITH MORE STRINGENT GROUP CRITERIA ..............................................................................................................................85 LIST OF REFERENCES ...............................................................................................................87 BIOGRAPHICAL SKETCH .........................................................................................................93

PAGE 7

7 LIST OF TABLES Table page 31 Group N s and Aggression Means and Standard Deviations by Aggressor Group ............57 32 Membership Agreement for Direct and Indirect Aggression G roups ................................58 33 Means and Standard Deviations for Aggression and Aggression Stereotyping ................59 34 Observed and Expected Gender and Race Distributions across Aggression Groups ........60 35 Summary of Hierarchical Regression Analyses for Particpants' Gender St ereotyping Scales Predicting Gender Percentages of Nominated P eers ..............................................61

PAGE 8

8 LIST OF FIGURES Figure page 11 Conceptual model of factors influencing self and peer based measures of aggressive behavior. .............................................................................................................................25 31 Graphical representation of group means by aggression type for behavioral and emotional characteristics ....................................................................................................62 32 Graphical representation of group means by aggression type for sociometric characteristics .....................................................................................................................63

PAGE 9

9 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy EXPLAINING DISCREPANCIES IN SELF AND PEER REPORTS OF AGGRESSION IN ADOLESCENCE By Katherine Hale Clemans August 2010 Chair: Julia A. Graber Major: Psycholog y Though most studies of aggressive behavior in early adolescence employ self or peer report methods to collect aggression data, these two measurement methods demonstrate weak correlations with one another in the literature. Social desirability has bee n identified as a source of bias in self reported aggression and is often controlled in analyses using self reported measures. Similarly, aggressionrelated social schemas could be a source of bias in peer reported aggression; however, no control measure for this currently exists. The present study investigated potential differences in the psychosocial correlates of self and peer identified early adolescent direct and indirect aggressors. In addition, scales assessing gender, race, and popularity based aggression stereotyping bias were created for the study as a way to tap participants social schemas, and their relationships to nominations of aggressive peers were examined. Participants (314 middle school students; M age = 12.83; SD = .96) were categor ized into groups based on self reported and peer nominated aggression scores and compared across a number of demographic and psychosocial factors. After controlling for social desirability bias, self identified aggressors were characterized by higher leve ls of manipulative behavior, whereas peer identified aggressors were characterized by particular race, gender, and sociometric patterns. Specifically, peer identified direct aggressors were more likely than self identified

PAGE 10

10 aggressors to be African American, and peer identified direct and indirect aggressors were less well liked but more socially visible and popular than self identified aggressors. Overall, results suggested that self and peer report methods identify qualitatively different groups of aggr essive adolescents. Furthermore, participants endorsed expected gender and popularity based stereotypes of aggressive adolescents. Endorsement of gender stereotypes of direct and indirect aggressive adolescents was related to the gender of nominated aggr essive peers, but not related to peer reported aggression levels. The wide use of peer report methods in the present literature on adolescent aggression suggests the need for better understanding of factors that influence those reports, as bias in peer re ports is often not considered in interpreting findings. The potential influence of stereotyping in peer identified aggression and the need for further investigation of procedures that control for bias in peer report measures are discussed.

PAGE 11

11 CHAPTER 1 INTR ODUCTION Adolescent aggression has long been of interest to researchers, educators, clinicians, parents, and policy makers. Aggressive behavior during this time can be direct (e.g., physical fighting, verbal threats or insults) or indirect (e.g., rumor spreading, social exclusion, withholding of friendship). Much literature on such aggression and antisocial behavior has focused on understanding its development and on identifying factors which place individuals at risk Despite the ample literature on the d evelopment of aggressive behavior, however, questions remain regarding the best method of assessment or source of information on aggressi on in adolescence. The present study investigates whether characteristics of early adolescents identified as aggressive differ as a function of self and peer report methods of assessment, as well as potential factors which may contribute to bias in these methods. Measuring A ggression in A dolescence Assessment issues become increasingly important during early adolescence for several reasons. F irst, time spent with friends and participating in activities with peers increases (Larson & Richards, 1991), and secondly, adolescents begin to distance themselves from their parents and engage in more activities that may be unmonit ored by parents and teachers (Marshall, Tilton Weaver, & Bosdet, 2005). For this reason, the teacher and parent reports that are frequently used to obtain information on aggressive behavior in childhood may miss important behaviors particularly concerning indirect forms of aggression that are less easily observed by an outsider to the peer group. For instance, with the advent of widespread text messaging and online social network usage among early adolescents (Hinduja & Patchin, 2008), much indirect aggres sion may be taking place increasingly in private venues to which adult observers have little to no access. Consequently researchers may be better served by focusing on self reported

PAGE 12

12 information or on information given by peers, whom many researchers expect to have the most accurate information about an adolescents social behavior not least because peers are usually the objects at whom the adolescent directs his or her aggression (Peets & Kikas, 2006). Thus, while studies using preschool and elementary sch ool age groups frequently employ parent and teacher ratings of aggression, almost all current studies of aggression in middle school and high school age groups use at least one self or peer reported measurement method. Longitudinal evidence suggests that agreement between self reported and peer reported aggression peaks in early adolescence (Pakaslahti & Kelitkangas Jrvinen, 2000). However, although peer report and self report measures of aggression are frequently intended to be measurements of the same construct, the literature to date reveals a consistent trend of weak correlations between the two, usually falling within the range of r = .10 to r = .35, which suggests a low level of shared variance between the two measures (Achenbach, McConaughy, & Howe ll, 1987; Card, Stucky, Sawalani, & Little, 2008; Epkins & Myers, 1994; Henry et al. 2006; Pakaslahti & Kelitkangas Jrvinen, 2000; Pellegrini & Bartini, 2000; Xie, Cairns, & Cairns, 2002). This is problematic for researchers seeking to investigate a sing le construct of aggressive behavior. Figure 11 portrays a conceptual model of the factors which can influence self and peer based measures of aggressive behavior. At the core of the model lie the multitude of factors which contribute to aggression at va rious points in the life span. During adolescence, numerous biological, psychological, and social processes take place which shape an adolescents social behavior. As Bronfenbrenners Ecological Model of Development ( Bronfenbrenner, 1979) indicates adoles cents develop via interactions with in many different contexts, including immediate familial and peer environments, along with broader cultural and societal contexts.

PAGE 13

13 Much research has been conducted on the role of these factors in the development and manif estation of aggression in adolescence. These include, but are not limited to, individual factors such as cognitive deficits, hyperactivity, emotion levels and emotion regulation abilities; familial factors such as parental punishment style and mothers IQ, social factors such as peer rejection, school involvement, and early pubertal development in relation to ones peers, broader environmental factors such as socioeconomic status and characteristics of ones neighborhood, and cultural factors such as gender ed expectations for behavior (see Dodge, Coie, & Lynam, 2006, for a review) T hese factors are thought to directly influence or be indicative of an individuals overarching tendency to engage in a ggressive behavior, and thus barring measurement error, their influence is generally expected to be captured by self and peer report measurement methods in similar fashions. If psychosocial influences were the only cause of variability in measurements of aggressive behavior, we should see considerably greater overlap in ratings of aggression originating from different sources. As it is, that is not the case. S everal factors may play a role in the relatively weak overlap between different informant reports of aggression. Achenbach and colleagues (1987) suggest that rather than calling into question the validity of any one measurement type, low correlation s between aggression scales may represent variations in the same underlying behavioral construct ; these variations manifest themselves differently according to t he context or situations in which they are experienced. For instance, an individual rating herself/ himself on aggressive behavior may recall instances in which behavior was directed toward siblings at home, which classmates would not have occasion to obser ve. Similarly, certain circumstances may arise in which an individual is effectively able to conceal behavior directed at peers, as in the anonymous generation of a rumor. In both of these cases, individuals

PAGE 14

14 would have access to information about their own aggression which would not be shared by peers. Alternatively, individuals may be sometimes unaware of the effect of their actions on others feelings or social reputations; leading peers to categorize as aggressive certain behavior s that the individual would not. Circumstances such as these may lead to variability in the influence of situational factors on measurements of aggressive behavior and contribute to the divergence of peer and self reports. L ow correlations between aggression measures, however, m ay also point to biases related to the methods in which the data are collected. A ggression is generally considered to be a socially unacceptable behavior; thus, in the case of self reported data, individuals may underreport aggressive behavior to cast them selves in a better light to both themselves and to researchers (Peets & Kikas, 2006; Nederhof, 1985). As a result, social desirability scales, which measure an individuals tendency to report that they engage in culturally approved, though unlikely, behavi ors (Crowne & Marlowe, 1960; Reynolds, 1982), have been developed, and these scales are often used as control variables to reduce the influence of social desirability bias in analyses of self reported behavior. Less attention has been paid to biases in peer reports of aggression, as many researchers are of the opinion that peers are more valid sources of information about an individuals behavior due to the fact that since they are not reporting on their own behavior, issues of social desirability bias a re not thought to be as applicable (e.g., Peets & Kikas, 2006). Peer reports, however, can be affected by adolescents social schemas A social schema is a cognitive structure that represents a persons knowledge about the traits and goals of particular individuals who fall into specific social categories (Fiske & Taylor, 1991). Social schemas help people organize and interpret information about their social world and provide expectations for others

PAGE 15

15 behavior a nd thus can lead to stereotypes of particular social groups Schemas can also create biases in the encoding and recal l of information about a social event ( Hamilton, Stroessner, & Driscoll, 1994; Younger, Schneider, & Daniels, 1991). An individuals gender role stereotypes, for instance, can influence the perception and recollection of others behavior, including a tendency to recall gender consistent information more often than gender inconsistent information ( Fiske & Taylor, 1991; Cantor & Mischel, 1977). Alternatively, if a person possesses a schema for a certain social group which includes the traits hostile or aggressive, he/she may selectively notice hostile cues over non hostile cues in ambiguous social situations with a member of that group (Sagar & Schofield, 1980). Social schemas also beco me more salient when individuals think about members of social outgroups (i.e., social categories, such as race, gender, or nationality, of which the individual is not a member; Hamilton et al., 1990). Comparisons of S elf and P eer I dentified A ggressive A dolescents Regardless of w hether or not low levels of correlation between self and peer reported aggression are the result of participant biases the relative lack of agreement between these measurement methods suggests that groups of students identified as aggressive may differ in important ways depending on the method by which aggression is assessed. Surprisingly little research, however ha s addressed this question directly. One exception is Card et al. (2008), a meta analysis which found that gender differences in physical aggression (i.e., boys were more physically aggressive than girls) were significantly larger for studies using peer nomination methods than for studies using self report methods. Card and colleagues also found that peer nomination and self report studies did not differ in the strength of relationships between aggression and emotional dysregulation. In addition, a study which examined differences in self versus peer reports of victimization (Graham, Bellmore, & Juvonen, 2003) found that self identified victims experienced increased psychological maladjustment, wh ereas peer identified

PAGE 16

16 victims did not differ from nonvictims in psychological maladjustment but did perform more poorly in school. In addition, the peer identified victims wer e more likely than the self identified victims to be African American and male. These results suggest that the profile of students identified as victims of peer aggression can differ substantially depending on the measurement method used. However, the stud y did not investigate group differences in peer and self identified perpetrators of aggression, but instead focused on victims only. In light of the relative lack of research addressing these questions, Clemans and Sontag ( 2009, April ) conducted a groupbased analysis to investigate potential differences in the psychosocial correlates of self and peer identified early adolescent aggressors. S everal demographic, sociometric, behavioral, and emotion/personality factors were selected on which to compare peer and sel f identified aggressors. In addition, a measure of social desirability was included in order to reduce underreporting bias and to increase the validity of the self report aggression measures and other self reported variables in the study (Nederhof 1985). Similar to previous studies utilizing multiple informant reports, significant but weak correlations between self and peer reported aggression were found ( r = .15, p < .001 for both direct and indirect aggression). The low correlations suggested t hat substantially different groups of aggressive students were being identified by each method. Results indicated that self and peer identified aggressors were indeed characterized by relatively different psychoso cial profiles. Direct self identified agg ressors showed significantly higher levels of socially manipulative behavior, remorseless/unemotional affect, and delinquency than other groups and indirect self identified aggressors showed high levels of socially manipulative behavior and delinquency I n contrast, peer identified aggressors were characterized by specific demographic and sociometric patterns. For instance, although self -

PAGE 17

17 identified directly aggressive participants did not differ from nonaggressors in their racial/ethnic distributions, dire ct aggressors who were nominated by peers were significantly more likely to be African American and significantly less likely to be European American than were nonaggressors In addition, peer identified indirect aggressors were significantly more likely t han self identified indirect aggressors to be female, and peer identified indirect aggressors were also more socially visible or well known within the social group, than either self identified indirect aggressors or nonaggressors. The results from Cleman s and Sontag that peer nominated direct aggressors tended to be African American, while peer nominated indirect aggressors tended to be socially visible females, are patterns clearly reminiscent of social stereotypes of directly and indirectly aggressive adolescents that may be driven by popular media (e.g., mean girls, gangsta culture). These findings suggested that early adolescents may be using social schemas to inform their nominations of aggressive classmates resulting in a bias due to stereotypin g (Fiske & Taylor, 1991; Giles & Heyman, 2005). Stereotyping Bias in Peer Nominations of Aggressive Individuals Stereotyping bias can come into play when beliefs about the nature of a group of people influence individuals interpretations and recollections of their behavior. Several studies have employed survey based measures of stereotyping as it specifically relates to aggression or to group comparisons. A classic example of this is the Katz & Braly (1933) checklist, in which participants mark on a list the traits which they think describe a particular social or demographic group. However, this measure may be particularly susceptible to bias due to social des irability (Whitley & Kite, 2006 ). Sagar and Schofield (1980), employing a more subtle approach, m easured the influence of racial stereotypes on the interpretation of ambiguous social behavior using a series of pictorial cues and verbal descriptions of ambiguously aggressive social

PAGE 18

18 interactions in which the race of the actor or target was systematicall y varied. Subjects were then asked to rate the actor on several traits (eg., friendly, threatening). There are drawbacks associated with this measure, however, including the time consuming nature of the task and its dependence on the perceived ambiguity of the social interaction in question, which can vary among scenarios as well as from person to person. Alternatively, Ryan, Judd, and Park (1996) used a mean range and estimation task to measure racial stereotypes. This procedure consisted of a series of b ehavioral dimensions with opposing behaviors as endpoints (e.g., well dressed poorly dressed), on which participants were asked to mark their perceived average of the racial group in question, as well as where they believed the highest and lowest members of the group would be. While this procedure is appealing because it employs a positive/negative comparison and may invoke more implicit reactions than free responses would invoke the use of statistical terms such as average and range may be too sophi sticated an instructional procedure for early adolescents. An assessment method similar to the group comparison used by Ryan et al. (1996), yet with simpler instructions is that used by Otten & Stapel (2007). P articipants indicated on a rating spectrum wh ether a particular behavioral or emotional trait applied more to one or another particular ethnic group, which comprised the two endpoints on the spectrum. Participants could indicate that it applied equally to both groups by selecting the midpoint of the spectrum. The absence of an aggression stereotyping measure was a major limitation of Clemans and Sontag (2009, April) because we were not able to directly test whether aggression stereotyping tendencies impacted ratings of peer nominations. However, th is study did investigate whether the tendency to nominate peers along certain demographic lines appeared to be shared among all peer nominators, or whether this tendency appeared to be particularly strong

PAGE 19

19 for members of the corresponding demographic outgro up. For instance, are boys more likely than girls to nominate other gender classmates as indirect aggressors? Because genders are highly segregated in their social interactions during early adolescence, a preference for samesex nominations on peer nominat ion measures of social behavior is expected ( Coie, Dodge, & Kuperschmidt, 1990; Mac c oby, 1998) ; thus, high levels of other sex nominations may suggest that stereotyping is taking place. For gender, results showed that boys were much more likely than girls to nominate members of the other sex as indirect aggressors, resulting in a greater number of peer identified indirectly aggressive girls. Because this result was not supported by a greater number of self identified indirectly aggressive females in the cu rrent study, nor by results of previous meta analytic findings of trivial differences in levels of indirect aggression between boys and girls (Card et al., 2008), it lends further support to the suggestion that some students in this study were in fact influence d by current social stereotypes of aggressive behavior in adolescence. Similar results were also found by Card, Hodges, Little and Hawley (2005): Sixthgrade males nominated a larger proportion of other sex members as indirect aggressors than did thei r female classmates. Beyond Card et al., I am aware of no other research to date which directly investigates the influence of racial or gender stereotyping bias on peer nominat ions of aggressive individuals. Psychosocial Factors Related to Aggression in Ad olescence One purpose of the present study is to replicate and extend the findings from Clemans and Sontag which suggest that substantially different groups of adolescents who possess dissimilar psychosocial profiles are being identified by self and peer report measurement methods. Although many factors can be investigated in relation to aggressive behavior, for the purposes of parsimony and specificity the present study focuses on investigating differences in a

PAGE 20

20 constellation of psychosocial characteris tics which represent both individual and social influences on behavior. These include several emotional and personality traits known to be indicative of aggressive behavior as well as influential in its development, peer driven indicators of aggressive beh avior relating to social status within ones peer group, and demographic indicators linked to differences in aggressive behavior, including gender, race, and socioeconomic status. Emotion and Personality Factors Empathy and r emorse. Empathy refers to the personal experience of anothers affective state after observing or learning of that state ( Eisenberg, Spinrad, & Sadovsky, 2006); remorse or guilt, is a feeling of discomfort following a transgression (Eisenberg, 2000) Low levels of empathy and remorse are emotional indicators of psychopathic personality and behavioral tendencies (Andershed, Kerr, Stattin, & Levander, 2002; Lynam, 1996). Children and adolescents who have low levels of these emotions engage in more frequent and severe forms of aggressive behavior (Eisenberg et al., 2002; Saltaris, 2002) and rate aggressive behavior as more morally permissible than do their peers (Eisenberg, Miller, Shell, McNalley, & Shea, 1991). Manipulative behavior. Social manipulation, including the telling of lies an d the use of dishonest charm to achieve social goals, is a behavioral indicator of psychopathic personality (A ndershed et al., 2002) and thus is similarly linked to elevated levels of aggression in adolescence (Saltaris, 2002). Anger and anger regulation. The emotional experience of anger is characterized by physiological arousal and cognitions of antagonism ( Novaco, 1994, p. 32); trait anger is defined as an enduring propensity to become angry (Spielberger, Jacobs, Russell, & Crane, 1983). High anger le vels contribute to aggressive behavior during adolescence ( Cornell Peterson, & Richards, 1999; Nichols, Graber, Brooks Gunn, & Botvin, 2006), due in part to the

PAGE 21

21 fact that feelings of anger increase reactive behaviors to real or perceived hostile situati ons (Muris, van der Pennen, Sigmond, & Mayer, 2008). Anger and the ability to effectively regulate anger (i.e., to consciously reduce the intensity of anger through mental or behavioral exercises) are related, since an individual with a higher baseline ang er level will have a relatively tougher job of controlling his/her temper Though the influence of emotion regulation factors on behavior has received much attention in recen t years (Eisenberg, Morris & Spinrad, 2004), research on anger regulation specifi cally has been scarce (Zeman, Shipman, & Suveg, 2002). S ome studies find no direct link between anger regulation and aggression in younger children (Dearing et al., 2002). Clemans, Graber, Nichols, Brooks Gunn, and Botvin (2007, March), however, found that a reduced ability to consciously regulate experiences of anger in early adolescence was uniquely associated with increased aggressive behavior even after accounting for levels of trait based anger. As such, it was of interest in the present study. Social F actors Social preference and social impact Social preference refers to how well liked one is by ones peers, while social impact refers to how visible one is within ones peer group. Social preference and social impact have somewhat different relations hips with aggressive behavior in adolescence. Specifically, high social impact is significantly related to high levels of aggression, and children and adolescents who have high social impact scores are more likely to be aggressive than children with high s ocial preference scores (Newcomb, Bukowski, & Pattee, 1993). Perceived popularity. Perceived popularity differs from peer acceptance and social visibility in that it incorporates levels of dominance within the peer group. Perceived popularity is usually correlated with social impact scores at around r = .50, and, like social impact differs from peer acceptance in its relationships with aggressive behavior. This is particularly true for

PAGE 22

22 indirect forms of aggression. Cillessen & Borch (2006) found that whi le relationally aggressive and nonrelationally aggressive middle school students tended to have similar levels of peer acceptance, relationally aggressive students were rated a full standard deviation higher on perceived popularity measures. In contrast, physically aggressive middle school students were somewhat less well liked than nonphysically aggressive students, but the two groups had similar levels of perceived popularity. Demographic C haracteristics Gender. Gender has an established relationship w ith physical aggression in the literature. At all ages, males are more likely than females to be physically aggressive (Archer, 2004; Card et al. 2008), to engage in direct verbal aggression (Archer, 2004; Card et al., 2008), and to commit violent crime ( Moffitt, Caspi, Rutter, & Silva, 2001). During high school, 44% of boys, but only 27% of girls, reported having engaged in a physical f ight in the last year (Center for Disease Control and Prevention, 2008). When effect sizes from studies using self repor t, peer and teacher reports, and observational measurement methods are averaged together across the lifespan, males tend to be higher than females in physical and verbal aggression by about .5 standard deviations (Hyde, 1984). Indirect aggression does not show similar patterns of gender differences. Based on early studies in which peers perceived girls as higher than boys in these behaviors, some researchers have argued that indirectly aggressive behavior is a female normative form of aggression (e.g., Cri ck, 1997). Evidence suggests that peers also endorse the view that females are the primary perpetrators of indirect aggression, rating indirect aggression by a female as more serious than the same behavior by male (Basow, Cahill, Phelan, Longshore, & McGi llicuddy DeLisi, 2007; Coyne, Archer, Eslea, & Liechty, 2008) Subsequent studies have shown mixed results when assessing actual levels of indirect aggression however, and metaanalytic reviews have failed to

PAGE 23

23 demonstrate significant gender differences in overall levels of indirectly aggressive behavior across childhood and adolescence (Archer, 2004; Card et al., 2008). The Present S tudy The present study had two main parts. The first was a replication and extension of Clemans et al. which assessed pot ential differences in patterns of psychosocial characteristics among self and peer identified aggressive adolescents To do so, participants were divided into groups based on their levels of self reported and peer identified aggression, and groups were c ompared with one another on a range of variables. The second part of the study directly examine d the relationship of stereotyp ing bias to nominations of aggress ive peers Specific Aims The specific aims of the present study were as follows: Specific aim 1: To replicate previous findings which showed that the psychosocial profiles of self identified and peer identified aggressors follow significantly different patterns, and furthermore, that these patterns are reliant on the type of aggression being asses sed (i.e., direct or indirect). Bas ed on findings in Clemans and Sontag (2009, April) and in previous literature I expect ed self identified aggressors to differ from peeridentified or nonaggressors on emotion and personality indicators of aggressive tendencies; self identified aggressors should have higher levels of remorselessness, manipulative behavior, and anger, and lower levels of empathy and anger regulation, and that the patterns of these differences will similar for direct aggression and indirect aggression. Furthermore, I expect ed these relationships to be significant even after controlling for potential bias due to social desirability Similarly, it was expect ed that peer identified aggressors w ould have higher social impact and perceived popul arity scores than self identified aggressors, that peer identified indirect

PAGE 24

24 aggressors w ould be disproportionately female, and that direct aggressors w ould be disproportionately African American. I also expect ed based on previous research linking physical aggression to male gender stereotypes, that peer identified direct aggressors will be disproportionately male ( Giles & Heyman, 2005; Loy & Norland, 1981). Specific A im 2: To determine whether variance in peer nominations of aggressive adolescents can be explained by nominators stereotypes of aggressive peers. The second aim of the study was to address the lack of literature investigating potential bias in peer nominations of aggression (in Figure 11, the farthest right box). T o assess the use of social schemas in adolescents nominations of aggressive peers, an aggression stereotyping measure was created for this study based on procedures used in Otten and Stappel (2007). It include d a list of aggressive and prosocial behaviors on which participants com pared gender (girls vs. boys), racial groups (Black vs. White), and sociometric groups (popular vs. unpopular). Stereotyping scales we re expected to be correlated with their corresponding demographic or sociometric category of peer nominations (i.e., parti cipants who tend to associate females with indirect aggression w ere expected to be more likely to nominate female classmates as indirect aggressors on the peer nomination portion of the survey). Furthermore, it was expected that the stereotyping scales w ould account for some of the variance in peer report ed aggression that was not explained by levels of self reported aggression, thus recommending their use as control variables in future studies of aggression utilizing peer reported measurement methods

PAGE 25

25 Underlying biological, psychological, and social processes Bias due to social desirability Bias due to peers social schemas Situationspecific influences on behavior Situationspecific influences on behavior Self -reported aggression Peer -reported aggression Figure 1 1. Conceptual model of factors influencing self and peer based measures of aggressive behavior.

PAGE 26

26 CHAPTER 2 METHODS Participants Participants were 315 6th, 7th, and 8th grade students ( M age = 12.83; SD = .96) from a single middle school in a small Southeastern city. Ninety six percent of the student population participated in the study. The student body of the school was representative of racial and socioeconomic distributions of the county in which it was located. Approximately 35% of s tudents attending the school are eligible for the free/reduced lunch program (an indirect measure of socioeconomic status ). Participants were 50.8 % European American, 2 2.6% African American, 15.0% Hispanic/Latino, 2.8% Asian, and 8.9% other ethnicities. G irls comprised 49.5% of the sample. Procedure Consent procedures were approved by the Internal Review Board of the University of Florida. The review board approved a waiver of active parental consent due to the fact that active consent procedures may ex clude students with the highest levels of problematic behavior, reducing the generalizability of all findings (Tigges, 2003) Letters were sent to the home addresses of parents/guardians which contained a letter explaining the purpose of the study, detail s about the anonymity and confidentiality measures in place to protect students privacy, and a prepaid, self addressed postcard which parents could use to decline consent for their child s participation if they so desire d. P hone and email contact informat ion for the study office were also provided for this purpose All measures were presented in self report surveys administered by a trained research team during school periods. Prior to administration of the surveys students were given an assent cover sh eet which briefly introduce d the study explain ed the anonymity and confidentiality

PAGE 27

27 procedures Questionnaires w ere identifiable by a unique ID number only. Students were verbally asked for their consent to participate prior to beginning the survey and we re i nformed that they could stop at any time once they started with no negative consequences. Students who decline d to participate or who were withdrawn by their parents were given a free period until their class completed the survey To address potenti al priming effects of the aggression stereotyping measures on peer nominations, the survey w as divided into two shorter counterbalanced sections, one containing the aggression stereotyping measure and one containing the peer nomination measure. Participant s completed the sections in two separate class periods at least one week apart Researchers also returned to the school for a final day of data collection so that students who had been absent on a regular administration day could complete their missed sur veys. Measures Peer Reported A ggression P eer reported aggression measures were based on procedures used in Putallaz et al. (2007). The measure comprised ten items, five of which assessed directly aggressive behaviors and five of which assessed indirectly aggressive behaviors. Because a specific aim of this study was to make comparisons between peer and self reported aggression, the wording p eer report aggress ion items mirror ed as closely as possible a subset of i tems from the self reported aggression me asures (described below). Example items include gossip or say mean things about other kids behind their backs (indirect aggression) and get in fights a lot (direct aggression). A complete list of items is included in Appendix A. For each item, stude nts were instructed to write the first and last names of kids in their grade who best fit the item description. Students could nominate an unlimited number of names if they wished, although the vast majority nominated 0 to 3 names for each item. Rosters listing

PAGE 28

28 the first and last names of same grade classmates were provided during this portion of the questionnaire in order to assist in identification and spelling. Total n ominations for each item were standardized by grade and summed to create two individual scales reflecting peer report direct aggression ( = .88) and indirect aggression ( = .77) Self R ep orted A ggression Direct self reported aggression. Direct self reported aggression w as assessed using items from the Aggression Scale (Orpinas & Franko wski, 2001). This scale has demonstrated good reliability in previous research by its authors. Students were asked to indicate how often they had engaged in a range of behaviors during the past year. Example items include d I got into a physical fight an d I threatened to hurt or hit someone. Response o ptions utilize d a 1 4 Likert Scale in which 1=Never, 2=Once or twice, 3=A few times, and 4=Often. Responses w ere averaged to create two individual scales reflecting self report ed direct and indirect aggres sion; higher scores indicate d higher levels of aggression. The full direct aggression sc ale contained 8 items ( = .85). However, in order to keep the construction of the self report and peer report aggression measures as similar as possible, a reduced s cale, which contained 5 items with similar wording to those assessed in the direct peer report measure, was utilized in subsequent analyses. The reduced scale was h ighly correlated with the full direct a ggre ssion s cale ( r = .96) and demonstrated acceptabl e reliability ( = .78). Indirect self reported aggression. Indirect self reported aggression w as assessed using items from the Revised Peer Experiences Questionnaire (Prinstein, Boergers, & Vernberg, 2001) The RPEQ indirect aggression subscale has demonstrated good re liability with youth in this age range and demographic background Students were asked to indicate how often they had engaged in a range of behaviors during the past year. Example items include I said mean things about

PAGE 29

29 someone behind his/her back and I left someone out on an activity or conversation that he/she really wanted to be included in. Response options utilize d a 1 4 Likert Scale in which 1=Never, 2=Once or twice, 3=A few times, and 4=Often. Responses were averaged to create two individual sca les reflecting self report ed direct and indirect aggression; higher scores will indicate d higher levels of aggression. The full indirect aggression scale contained 12 items ( = .77). A reduced scale which contained 5 items with similar wording to those assessed in the direct peer report measure was also created The reduced scale was highly correlated with the full indirect aggression scale ( r = .90) but demonstrated low, though not unacceptable, reliability ( = .56). The lower alpha levels for indire ct aggression are likely due to the fact that several different types of behavior are assessed (including social exclusion, rumor spreading, and withdrawal of friendship), whereas direct aggression items assess only physical and verbal forms and therefore overlap with each other to a greater extent. Other often used assessments of indirect aggression utilizing low item counts have shown s imilarly low reliability levels In order to keep the construction of the self report and peer report aggression measur es as similar as possible, the reduced scale was employed in all subsequent analyses of indirect aggression. Aggression S tereotyping The a ggression stereotyping scales for this study used a format adapted from Otten and Stapel (2007) and were comprised of 12 items listing aggressive and prosocial behaviors. For each item, students were asked to indicate on a 7 point spectrum scale whether they thought the behavior applied more to one particular group of people, more to another group of people, or equally to both groups. The set of 12 items was repeated for three group comparisons: gender ( boys vs. girls ) race ( Black kids vs. White kids ) and popularity ( p opular kids vs.

PAGE 30

30 unpopular kids). These three dyads were chosen based on observed patterns in the demographic and sociometric characteristics of peer nominated aggressors in previous research. Of the 12 items in each scale, 10 assessed aggression and tapped similar behaviors as the 5 direct and 5 indirect items from the peer reported aggression sc ale. The two remaining items, which assessed prosocial behavior, were included as filler items A complete list of items is presented in Appendix A Scores on the aggressive behavior items were averaged to produce overall direct and indirect aggression sc ores for each scale, and centered so that negative values indicated greater bias toward the left hand listed group and positive values indicated greater bias towar d the right hand listed group. Reliability scores for the aggression stereotyping scales wer e as follows: direct gender ( = .68); direct race ( = .78); direct popularity ( = .80); indirect gender ( = .61); indirect race ( = .66; indirect popularity ( = .71). Nominators average stereotyping scores. In some analyses, a variable representin g the average stereotyping scores of the peers nominating a particular individual was used. This was initially calculated in two ways: (1) For each time a participant's name was nominated on a peer reported aggression item, the corresponding score of the peer who made that nomination was substituted and these scores were averaged together; (2) The same procedure, except before averaging, scores were checked for redundancy nominations (e.g., a peer nominated the same participant on multiple aggression items ), and redundancies were removed before scores were averaged so that each nominators score was only counted once. This was done for all three stereotyping scales and conducted separately for direct and indirect aggression, resulting in six average scores The two different methods of calculation did not significantly effect scores. Correlations between the two calculation methods on any one particular scale were very high ( r = .98 to .99, p

PAGE 31

31 < .001), indicating that the method of calculation did not signi ficantly affect scores. Thus, only the scores with redundancies removed (method 2) were used in subsequent analyses. Social D esirability Social desirability bias was assessed using a 10 item version of the Marlowe Crowne Social Desirability Scale (Straha n & Gerbasi, 1972), which consists of a series of statements about socially desirable or undesirable behavior (e.g., I am always polite, even to people who are disagreeable). Participants we re asked to indicate whether the statements were true or false a s they pertained to themselves. Items we re assigned a 1 if the participant selected the more socially desirable response and a 0 if the participant selected the less socially desirable response. Items were then averaged to produce an overall score with a range of 0 1, with higher scores indicating greater social desirability bias. Sociometric C ategorizations Social preference and impact Four additional items were included in the peer nomination portion of the questionnaire. The first two, write the names of kids in your grade whom you like the most and write the names of kids in your grade whom you like the least, w ere used to mathematically compute social preference and social impact scores (Coie, Dodge, & Coppotelli, 1982). Social preferences scores we re created by subtracting standardized liked least scores from standardized liked most scores. Social impact scores we re created by summing standardized liked most and liked least scores. Perceived popularity. The final two items, write the n ames of kids who are the most popular in your grade and write the names of kids who are the least popular in your grade assessed perceived popularity, or how popular one is within ones group. A popularity spectrum score was created by subtracting stan dardized least popular nominations from standardized most popular nominations.

PAGE 32

32 Emotion and Personality Indicators Manipulative behavior Socially manipulative behavior ( = .88) w as measured using 15 items from the Y outh Psychopathic Traits Inventory (YPI ; Andershed et al., 2002) which assessed manipulativeness, dishonest charm, and lying behaviors. Participants we re asked to indicate how well each item applied to them. E xample items include I am good at getting people to believe me when I make something up; Its easy for me to charm others to get what I want from them. The response scale ranged from 1 (almost always untrue) to 5 (almost always true). Responses wer e averaged such that higher scores indicated greater levels of manipulative behavior Remorselessness Remorselessness ( = .69) was measured using 5 items from the YPI which assess ed a lack of guilty feelings in relation to ones behaviors. Participants we re asked to indicate how well each item applied to them. Example items include d I seldom regret the things I do, even if other people feel that they are wrong. The response scale ranged from 1 (almost always untrue) to 5 (almost always true). Responses w ere averaged such that higher scores indicate d greater remorselessness Empathy Empathetic responding ( = .82) w as assessed using the 20 item Basic Empathy Scale (BES; Joliffe & Farrington, 2006). Participants are asked to indicate how much they agree or disagree with a series of statement s Example items include I can usually work out when my friends are scared and I dont become sad when I see other people crying (reverse coded). The response scale ranged from 1 (strongly disagree) to 5 (strongly agree). Responses w ere averaged such that higher scores indicate d greater levels of empathy. Anger. The seven i tem anger subscale from the Buss and Perry (1992) Aggression Questionnaire was used to assess trait levels of anger ( = .72) Students we re asked to rate how well a series of statements fit them. Items included I sometimes feel like a powder keg ready to

PAGE 33

33 explode and Some of my friends think Im a hothead. Response categories range d from 1 (Really Not True for Me) to 5 (Really True for Me). Items were averaged such that higher scores indicate d greater anger Anger regulation Anger reduction skills ( = .68) were assessed with a six item scale created for the Life Skills Training Program (Epstein, Botvin, Diaz, Baker, & Botvin, 1997). Participants were asked how often they engaged in a series of activities when they felt really angry. Items include d C ount to ten, Take a few deep breaths, and Tell myself this isnt worth fighting over (its no big deal.) Response categories ranged from 1 (Never) to 5 (Always). Items w ere averaged such that higher scores indicated greater skill at conscious anger re duction.

PAGE 34

34 CHAPTER 3 RESULTS Creation of Aggressor G roups Following procedures utilized in Cillessen and Borch (2006), groups of students were created based on self reported (reduced scale) and peer reported aggression scores. Two sets of groups, one f or direct aggression and one for indirect aggression, were created. Within each type of aggression, students could fall into one of three categories: (1) low aggression (henceforth referred to in results and discussion as low); (2) high self reported ag gression only (high self); (3) high peer reported aggression only (high peer); (4) high in both self and peer aggression (high multiple). Group membership within aggression type (i.e., direct versus indirect) was mutually exclusive, but students could be members of both a direct and an indirect high aggression group. Preliminary A nalyses What constituted high aggression for self and peer reported scores was determined by cutoff scores based on methods utilized in past literature as well as on conc eptual considerations. For instance, Cillessen and Borch (2006) selected, as their criterion for membership in a high sociometric popularity or perceived popularity group, the cutoff of .5 SD above the mean on one of these measures. This resulted in 39.4% of their sample achieving membership in at least one of the two high popularity groups. Clemans and Sontag (2009, April) also utilized the .5 SD cutoff for membership in either high self reported or high peer reported aggression groups, resulting in 2 328% of the sample meeting the criteria for a high self reported aggression, 1517% of the sample meeting the criteria for high peer reported aggression, and 45.5% of the sample achieving membership in at least one high direct or indirect aggression group; individual group N s for high self, high peer, and high multiple report groups ranged from 18 to

PAGE 35

35 57. In that study, the .5 SD cutoff score was sufficient to produce diverging patterns of behavioral, emotional, and sociometric characteristics for high s elf and high peer groups. In the present study as in prior work, the .5 SD criterion resulted in about 40% of the sample meeting criteria for at least one high group. One concern is whether the .5 SD cutoff is sufficiently stringent for identifying tr uly aggressive adolescents. Notably, measures of self reported aggression almost always result in positively skewed data, since the majority of any normative adolescent population engages in either no involvement or only sporadic involvement in direct and indirect aggressive behavior. This results in a smaller subset of students who score .5 SD or higher above the mean on any one measure than would be expected with normally distributed data. The relatively large percentage of students meeting the criteria for at least one high group in previous research may be interpreted as an indicator of the varied forms that aggression takes during adolescence, as well as the range of opinions between individuals and their peers as to who engages in high aggressio n, rather than as evidence of too lenient a cutoff score. In addition, preliminary analyses found a pronounced positive skewness in the peer reported aggression measures. Although the majority of students did not receive any nominations on peer reported aggression items, the range of scores was quite large, and there were several outliers with exceptionally high nomination tallies (one student received 108 total nominations on individual aggression items), resulting in substantially fewer numbers of stude nts who were above the .5 SD cutoff for peerreported aggression (12.5% to 15.8% of total sample) than were above the cutoff for self reported aggression (21.9% to 24.3% of total sample). In this case, it was determined that the .5 SD cutoff might be nonrepresentative for peer reported aggression and that selecting a high group based on a nomination tally cutoff score would be

PAGE 36

36 more appropriate for creating high peer reported aggression groups. A cutoff of 4 nominations was selected for the reason tha t it included relatively the same numbers of students (23.7% to 24.9% of total sample) as the .5 SD cutoff for self reported aggression, which increased the equality of high aggression group N s. To explore the effect of using more stringent cutoff crite ria, all groupbased analyses in the present investigation were also conducted with 1 SD cutoff criteria for selfreported scales and a cutoff of 5 nominations for the peer reported aggression. T his approach produce d smaller cell sizes for the aggression groups, but for the most part, patterns of differences between groups did not change F indings using the more stringent criteria which differed from the reported analyses are presented and discussed in Appendix D Criteria for G roup M embership As such, the criteria for group membership within aggression groups were as follows: High self aggressors were students whose self reported aggression score was .5 SD or higher above the mean, but who had fewer than 4 total nominations for peer reported aggression. High peer aggressors were students who had 4 or more total nominations for peer reported aggression, but who scored lower than .5 SD above the mean on self reported aggression. High multiple aggressors had scores .5 SD or higher above the mean or high er on self reported aggression and 4 or more total nominations for peer reported aggression. The low group was comprised of the remaining students. High self, high peer, high multiple, and low aggression groups were created for both direct and indirect aggression separately, resulting in a total of eight groups. Table 31 presents group N s, mean levels of aggression for each group. Table 32 presents membership agreement across direct and indirect aggression. Seventy five students (2 3.8% of the sample ) were members of both a direct and an indirect highaggression group.

PAGE 37

37 Descriptive S tatistics Aggression Means and standard deviations for all self reported and peer reported aggression variables are presented in Table 33. Consistent with data on aggressive behavior in normative populations, all were positively skewed. Thus, in subsequent analyses, aggression variables have been transformed via the square root function to increase the normality of their distributions (this excludes group based analy ses using previously created high aggression groups). A full table of correlations between all study variables is presented in Appendix B As expected, Pearson correlations between the self and peer reported forms of aggression were significant but som ewhat small, r = .27, p < .001 for self and peer reported direct forms; r = .23; p < .001 for self and peer reported indirect forms. Controlling for social desirability did not affect the strength of either relationship. The low amount of shared varian ce indicated by these correlations suggested that substantially different groups of individuals were being identified as high in aggression according to each method. Agreement within reporting method was better, r = .42, p < .001 for direct and indirect se lf reports; r = .68, p < .001 for direct and indirect peer reports. Students appeared to make more of a distinction between direct and indirect forms of aggression in their own behavior than they did in their peers behavior. Since more than 50% of the variance in even the peer reported measures was unshared, however, direct and indirect aggression were treated as distinct forms and were evaluated separately in subsequent analyses. Stereotyping Means and standard deviations for individual stereotyping sc ales are presented in Table 33. All scores had good distribution. The shift of each scale in one direction or the other from the center reflected common social stereotypes relating to the gender and popularity of

PAGE 38

38 aggressive adolescents: Students endorse d the view that boys were more likely to engage in direct aggression (shift to boys side = 1.05 SD ), whereas girls were more likely to engage in indirect aggression (shift to girls side = .96 SD ); students also believed that popular students were more li kely than unpopular students to engage in both forms of aggression, but this was particularly true for indirect forms (shift to popular side = 1.15 SD for direct and 1.49 SD for indirect). Students showed the least evidence of aggression stereotyping for racial groups. Students endorsed the view that Black students were more likely than White students to engage in direct aggression (shift to Black side = .72 SD ) but this was the smallest significant shift out of all the scales. No significant shift was f ound for indirect aggression. Three percent of the sample ( N = 11) completed the stereotyping scales for gender and popularity but left the racial group stereotyping scales blank, presumably because they were uncomfortable answering these particular quest ions. These students were compared to the rest of the sample on all relevant study variables (including aggression, aggression stereotyping, peer nomination demographic percentages, and behavioral and emotional characteristics) on a series of t tests of i ndependent samples; no significant differences at p < .05 were present. Direct and indirect aggression stereotyping scales comparing popular and unpopular students showed the most agreement ( r = .61, p < .001); students believed that popular students we re more likely than unpopular students to engage in all types of aggression. Students who endorsed the view that popular students were more indirectly aggressive than unpopular students were also slightly more likely than other students to believe that girls were more indirectly aggressive than boys (r = .20, p < .001). Although other correlations between stereotyping scales

PAGE 39

39 were significant at p < .05, all were weak ( r < .20); as such, they were not considered particular ly notable Behavioral, E motional, and Sociometric C haracteristics The behavioral and emotional characteristics measured in the current study included manipulative behavior ( M = 2.16, SD = .69) remorselessness ( M = 2.14, SD = .78) empathy ( M = 3.52, SD = .50) anger ( M = 2.55, SD = .79) and anger regulation skills ( M = 2.39, SD = .80) Each had good distribution, with minimal positive skew for manipulative behavior and remorselessness. The sociometric characteristics measured in the current study included social preference, social imp act and perceived popularity ; these were standardized into z scores or calculated from standardized scores All had normal distribution. Out of 10 possible correlations between behavioral and emotional characteristics 7 were significant at the p < .01 level, and all variables were significantly related to at least two others (see Appendix B for correlation values) Furthermore, all significant relationships made conceptual sense (e.g., lower levels of empathy were associated with higher levels of remor selessness and manipulative behavior). These results suggested that together, the variables represented a psychosocial profile which, reversing empathy, could be considered prosocial at lower levels and antisocial at higher levels. Previous literature (e .g., Newcom b et al., 1983) has suggested that social preference and social impact tend to be distinct domains of sociometric status. The present study corroborated these findings: social preference and social impact were weakly correlated ( r = .14, p < 05). Both, however, were related to perceived popularity ( r = .37, p < .001 for social preference; r = .40, p < .001 for social impact). This supported the conceptual distinction that social preference is most related to likeability, social impact is mos t related to notice and influence within the peer

PAGE 40

40 group, and students who are universally considered to be the most popular in their grade tend to be high in both of these qualities. Potential C ovariates Social D esirability The Marlow Crowne social desirability scale ( M = .50, SD = .24) was employed in this study as a measure of reluctance to self report socially undesirable behavior. Accordingly, social desirability scores were significantly correlated at p < .01 with self reported direct aggression ( r = .32), self reported indirect aggression ( r = .43), anger ( r = .43), manipulative behavior ( r = .38), and remorselessness ( r = .25). Social desirability was not related to any other self reported variables in the study, including anger regulation em pathy and stereotyping scores, and did not significantly differ by race, gender, or grade level. Gender Mean differences between boys and girls were significant at the p < .05 level for several behavioral, emotional, and sociometric characteristics as well as a few stereotyping scales. All significant findings are presented in Appendix C In general, girls reported lower levels of antisocial characteristics and higher levels of prosocial characteristics, and were considered by peers to be more well liked and popular than boys. Girls were also more likely than boys to endorse the stereotypes of girls and popular students as indirect aggressors Finally, boys showed greater levels of both self and peer reported direct aggression than girls, while girls had higher levels of self reported indirect aggression than boys. (The relationship of gender to aggression is further addressed in groupbased analyses below .) Race Fewer significant differences within study variables existed for racial groups. All significant findings are presented in Appendix C African American students showed slightly

PAGE 41

41 lower levels of empathy than European American students and had slightly lower social preference scores Differences that did exist tended to be within stereotyping scale s: African American students were less likely than other stude nts to endorse stereotypes of African Americans as direct and indirect aggressors as well as less likely to endorse the stereotype of girls as indirect aggressors African American students also had higher self reported direct aggression scores than did other students, and were more likely than other students to be nominated by peers as both directly and indirectly aggressive (The relationship of race to aggression is further addressed in groupbased analyses below). Grade Oneway analyses of variance revealed several grade level differences on relevant study variables, including stereotyping scales and behavioral and emotional characteristics. All significant findings are presented in Appendix C The majority of findings seemed due to differences between the 6th graders and the other two grades. For instance, 6th graders tended to report lower levels of antisocial characteristics than did students in higher grades. No significant differences by grade level emerged for self reported direct or indirect aggression (peer reported aggression variables were standardized by grade). Form order The presentation of measures was counterbalanced during data collection, with half of each grade completing stereotyping measures the first week and aggression measures the second week (and vice versa). To test for possible form order effects, a series of t tests of independent means was conducted for all relevant study variables. Only two significant effects e merged: students who completed stereotyping measures last were slightly more likely to endorse the stereotype of girls as indirect aggressors ( M difference = .25, t = 2.57, p < .05) and students who

PAGE 42

42 completed aggression measures last nominated a higher per centage of female classmates as indirect aggressors ( M difference = 14%, t = 2.23, p < .05). Summary Social desirability, gender, race, and grade were identified as necessary covariates for analyses addressing Question 1 (below), due to significant relati onships with aggression variables used to assign group membership and/or at least one of the behavioral, emotional, or sociometric outcome variables. Gender, race, grade, and form order were also related to stereotyping scores and peer nomination demographic percentages, and thus were included where relevant in analyses addressing Question 2 (below). Analyses A ddressing Question 1 : Do Self Identified and P eer Identified Aggressive Adolescents Show Distinct Patterns of Psychosocial and Demographic C haracteristics? Behavioral & Emotional Characteristics As indicated, it was hypothesized that higher levels of antisocial emotional and behavioral characteristics would characterize self identified aggressive adolescents, whereas peer identified aggressive adoles cents would resemble nonaggressors in these characteristics. The significant relationships between the five behavioral and emotional characteristics suggested the need for a multivariate approach to examine group mean differences. For both direct and indir ect aggression, these variables were analyzed in multivariate analyses of co variance (MAN C OVAs) as well as planned univariate follow up tests for specific group differences on individual variables All analyses included social desirability, gender, race ( coded as African American vs. other) and grade level as covariates due to the significant relationships found in preliminary analyses. E stimated marginal group means and 95% confidence intervals for individual behavioral and emotional characteristics are graphically represented in Figure 3 1 for both direct and indirect aggressor groups.

PAGE 43

43 Direct A ggression The multivariate effect for direct aggression was significant, Wilkes = .81 F (15,731.95) = 3.94, p > .001. Univariate analyses revealed that the m ultivariate effect was primarily driven by significant group differences for manipulative behavior ( F = 11.31, p < .001), remorselessness ( F = 6.70, p < .001), anger ( F = 6.43, p < .001), and, to a lesser extent, anger regulation ( F = 2.71, p < .05). Empa thy did not have a significant univariate effect. Estimated marginal means comparisons (Figure 31) indicated that self identified aggressive adolescents (the high self and high multiple groups) had significantly higher levels of manipulative behavior tha n either the high peer group or the low aggression group ( M differences = .43 to .54; p < .01). The high peer group and the low aggression group did not significantly differ from one another. For remorselessness, self identified aggressive adolescents (th e high self and high multiple groups) had significantly higher levels of remorselessness than did the low aggression group ( M differences = .45 to .50; p < .01), whereas solely peer identified aggressors (the high peer group) did not significantly differ i n remorselessness from the low aggression group. Similarly for anger, the high self and high multiple groups had higher levels of anger than did the low aggression group ( M differences = .44 to .48; p < .01), whereas the high peer and low aggression groups did not significantly differ from one another. Finally, for anger regulation, t he significant univariate effect for anger regulation was primarily driven by lower levels in the high multiple group compared to all other groups ( M differences = .37 to .41; p < .01); no other significant group differences were present. Summary The overall pattern of findings for direct aggression groups suggested that self identified aggressors (the high self and high multiple groups) had the highest levels of antisocial b ehavioral and emotional characteristics. In addition, solely self identified aggressors had higher levels of manipulative behavior than solely peer identified aggressors. Although the

PAGE 44

44 high self group differed from low aggressors on three of the five vari ables, n o significant differences between the high peer and low direct aggression groups were present. Indirect A ggression The multivariate effect for indirect aggression was significant, Wilkes = .84, F (15,731.95) = 3.15, p > .001. In a similar patt ern to that of direct aggression, univariate analyses revealed that the multivariate effect was primarily driven by significant group differences for manipulative behavior ( F [3,281] = 11.53, p < .001), remorselessness ( F [3,281] = 6.21, p < .001) and anger ( F [3,281] = 3.15, p < .05). No significant univariate effects existed for anger regulation or empathy. Estimated marginal means comparisons (Figure 31) indicated that self identified aggressive students (high self and high multiple groups) had signific antly higher levels of manipulative behavior than both the high peer group or the low aggression group ( M differences = .44 to .68; p < .01). Solely peer identified aggressive students (the high peer group) did not differ from the low aggression group. The significant univariate effect for remorselessness was primarily driven by significantly elevated levels in the high multiple group compared to all other groups ( M differences = .50 to .69; p < .01). Similar to remorselessness, the high multiple group had significantly elevated levels of anger compared to both the low aggression and high peer groups. However, the high multiple group did not differ from the high self group in anger levels. Summary The overall pattern of findings for direct aggression groups suggested that indirect aggressors identified both by themselves and by their peers had the highest levels of antisocial behavioral and emotional characteristics. In addition, solely self identified aggressors had higher levels of manipulative behavior than solely peer identified aggressors. Again, no significant differences between the high peer and low aggression indirect groups were present.

PAGE 45

45 Demographic and Sociometric Characteristics It was hypothesized that peer identified aggressive adolesc ents would be distinguished from selfidentified or nonaggressive adolescents by demographic and sociometric patterns which evoke common social stereotypes of direct and indirect aggression. Specifically, the high peer group for direct aggression was expe cted to contain greater percentages of males and African Americans than the high self group for direct aggression. In addition for indirect aggression, the high peer group was expected to contain a greater percentage of females and to have higher social v isibility than the high self group. Demographics Group differences in gender and racial distributions were each assessed via Pearson 2 analyses. These were performed in three ways. First, an overall 2 tested for differences in race or gender in the full sample using all four possible groups, including the low aggression group. Pending a significant effect in the overall test, compa risons between groups were tested first by comparing the low aggression group to all other students and then by comparing only high aggression groups while excluding the low aggression group. These follow up tests provided information as to whether signif icant effects were a result of differences between the specific methods (self or peer report) used to measure aggression, or whether they were primarily driven by differences between nonaggressors and aggressive students identified by any measurement type. Direct aggression: gender. The overall analysis for gender was significant, 2(3) = 20.18, p < .001. This indicated that a significant relationship existed between gender and direct aggression group membership. Observed and expected cell counts for the overall test are provided in Table 34. Follow up analyses revealed that the significant overall effect was primarily driven by differences between students low in aggression and students identified as

PAGE 46

46 high in aggression by either method ( 2(1) = 18.14, p < .001), rather than by differences between high aggression groups ( 2(2) = 2.28, p > .05). This suggests that boys were especially likely to be identified as direct aggressors, regardless of the reporting type used. Direct aggression: race. The overa ll analysis for race/ethnicity was significant, 2(6) = 30.08, p < .001. This indicated that a significant relationship existed between race/ethnicity and aggression group membership. Observed and expected cell counts for the overall test are provided in Table 34. Follow up chi squares revealed that, similar to gender, the significant overall effect was primarily driven by differences between students low in aggression and students identified as high in aggression by either method ( 2(2) = 17.56, p < .001). However, a nonrandom racial group distribution also existed among high aggression groups ( 2(4) = 10.04, p > .05). Distribution across aggression groups for African American students showed the most deviation from expected cell counts: African Ameri can students accounted for 15.8% of the low aggression group and 18.9% of the high self group, but accounted for 39.0% of the high peer group and 51.4% of the high multiple group. Indirect aggression: gender. The overall analysis for gender was signific ant, 2(3) = 11.66, p < .01. This indicated that a significant relationship existed between gender and direct aggression group membership. Observed and expected cell counts for the overall test are provided in Table 34.Similar to direct aggression, follow up chi squares revealed that the significant overall effect was primarily driven by differences between students low in aggression and students identified as high in aggression by either method ( 2[1] = 10.14, p < .01), rather than by differences between hig h aggression groups ( 2[2] = .92, p > .05). This suggests that girls were especially likely to be identified as indirect aggressors, regardless of the reporting type used.

PAGE 47

47 Indirect aggression: rac e The overall analysis for race was significant, 2(6) = 14.94, p < .05. This indicated that a slight but significant relationship existed between race/ethnicity and indirect aggression group membership; however, both follow up 2 tests failed to reach significance at the p < .05 level. This suggests that the significant overall 2 effect may have resulted from unexpected distributions across the low aggressors and potentially one other high aggressi on group. Observed and expected cell counts for the overall test are provided in Table 34; the largest disagre ements between observed and expected cell counts are as follows: Fewer African American students than expected were members of the low aggression group (18.7%), whereas more African American students than expected were members of the high peer group (33.3% ) and high multiple group (33.3.%), and fewer Other students than expected (9.7% compared to 27.933.3% in other groups) were members of the high peer group. Summary. Although nonrandom gender and racial distributions were present across both direct and indirect aggression groups, they were mainly driven by differences between students who had low aggression levels and students who were identified as aggressive by at least one type of report method. However, nonrandom distributions of racial groups we re evident for direct aggression: the percentage of African American students within the high peer and high multiple groups was substantially larger than within the low aggression or high self groups. Although hypothesized to be present, no gender differe nces existed between high self and high peer groups for indirect aggression. Sociometric C haracteristics Significant relationships between perceived popularity, social preference, and social impact suggested the need for a multivariate approach to examin e group mean differences. MANCOVAs were run for both direct and indirect aggression, as well as univariate planned follow up tests for specific group differences on individual variables. All analyses included

PAGE 48

48 gender and race as covariates due to their si gnificant relationships with the distribution of aggressor groups and at least one of the outcome variables. Since all outcome variables were standardized by grade before being analyzed, grade level was not included as a covariate. F or both direct and indirect aggressor groups e stimated marginal group means and 95% confidence intervals for individual sociometric characteristics are graphically represented in Figure 32. Direct aggression. The multivariate effect for direct aggression was significant, Wilkes = .70, F (9,737.57) = 13.03, p > .001). Univariate tests revealed that social preference ( F [3,305] = 12.18, p < .001), social impact ( F [3,305] = 26.12, p < .001), and perceived popularity ( F [3,305] = 4.37, p < .01) all showed evidence of mean differ ences across aggression groups. Furthermore, estimated marginal means comparisons revealed the same pattern of group differences for all three variables: students in the high self and high multiple groups had lower levels of social preference, higher leve ls of social impact, and higher levels of perceived popularity than students in the low aggression and high self groups ( M differences = .69 to 1.19, p < .05). No significant differences existed on any sociometric characteristic between the high self and high multiple groups, nor between the low aggression and high self groups. Indirect aggression. The multivariate effect for indirect aggression was significant ( Wilkes = .66, F (9,737.57) = 15.34, p > .001), and univariate tests revealed that social pr eference ( F [3,305] = 11.86, p < .001), social impact ( F [3,305] = 27.97, p < .001), and perceived popularity ( F [3,305] = 8.07, p < .01) all showed evidence of mean differences across aggression groups. The high peer group had higher social impact and lower social preference scores than both the high self or low aggression groups, and the high multiple group had higher social impact and lower social preference scores than all other groups. Differences for perceived popularity followed the same pattern as for direct aggression: The high peer and high multiple

PAGE 49

49 groups were perceived as significantly more popular than either the high self or low aggression groups. Again, no significant differences existed between the high self and low aggression groups for any sociometric characteristic. Summary. Significant differences between exclusively self identified and exclusively peer identified aggressive adolescents existed for all sociometric characteristics. For both direct and indirect aggression, peer identified aggressors were more popular, more socially visible, and less well liked than those identified solely through self report. In addition, solely self identified aggressors did not differ from the low aggression group on any sociometric characteristic, regar dless of aggression type. Analyses Addressing Question 2: Is the Tendency to Endorse Particular Aggression Stereotypes Related to Nominations of Aggressive P eers? Stereotyping Scores and Nominated Peers Gender, Race, and P opularity It was hypothesized that participants stereotyping scores would be related to characteristics of peers nominated as directly or indirectly aggressive. To test this, a series of hierarchical regression analyses were conducted; direct and indirect aggression were examined separ ately for each of the three stereotyping categories, resulting in a total of six analyses. For direct and indirect stereotyping scales relating to gender, the outcome variable in question was the percentage of males that an individual nominated on direct or indirect peer reported aggression items; for scales relating to race, it was the percentage of African American students1. For scales relating to popularity, the outcome variable was the mean perceived popularity score of nominated peers. 1 The African American vs. other dichotomous distinction was chosen for analysis due to the fact that preliminary analyses suggested that the European American and other racial categ ories tended to resemble one another on aggression and stereotyping measures.

PAGE 50

50 For each a nalysis, covariates including gender, race (dummy coded as African American vs. all others), grade level, and form order were entered in Step 1; however, only those covariates which demonstrated a significant effect in the full model were retained in the f inal analysis. The appropriate stereotyping scale, which corresponded to the same aggression type and demographic or sociometric category as the outcome variable, was entered in Step 2. In step 3, an interaction term was added to determine whether the st rength of these relationships varied across gender and racial categories or as a participants own perceived popularity increased. Direct aggression Table 35 presents the results of the analysis for gender, which compared participants scores on the gir ls boys direct aggression stereotyping scale to the gender distribution of the peers whom they nominated as directly aggressive. Gender distribution was entered as the percentage of nominations that were of male peers. No effect of race, grade level, or form order was present, so these covariates were dropped from the final model. After controlling for gender, girls boys stereotyping scores explained unique variance in the gender distribution of nominated peers, = .26, p < .001; R2 = .07. Students who were more likely to endorse the view that boys were direct aggressors were more likely to nominate male peers on the direct aggression items, and students who were more likely to endorse the view that girls were direct aggressors were more likely to nominate female peers on the direct aggression items. Furthermore, girls and boys were equally likely to demonstrate this relationship: there was no interaction effect between respondents gender and stereotyping score, = .04, p R2 < .01. Gender was the only category in which there existed a relationship between direct aggression stereotyping and corresponding characteristics of nominated peers. After controlling for covariates, analyses failed to demonstrate significant unique effects or interaction effects for

PAGE 51

51 black white ( = .01 to .04, p R2 < .01) or popular unpopular ( = .02, p R2 < .01) stereotyping scores. Indirect aggression Table 35 also presents the results of the analysis of gender for indirect aggression, which compared participants scores on the girls boys indirect aggression stereotyping scale to the gender distribution of the peers whom they nominated as indirectly aggressive. The indirect aggression analysis followed a similar pattern as that for direct aggression. No effect of race, grade level, or form order was present, so these covariates were dropped from the final model. After controlling for gender, girls boys stereotyping scores explained unique variance in the gender distribution of nominated peers, = .2 1, p < .0 R2 = .05. Students who were more likely to endorse the view that boys were indirect aggressors were also more likely to nominate male peers on the indirect aggression items, and students who wer e more likely to endorse the view that girls were indirect aggressors were more likely to nominate female peers on the indir ect aggression items. O verall students tended to endorse the stereotype that girls engaged in indirect aggression rather than boys Also similar to direct aggression, girls and boys were equally likely to demonstrate this relationship: there was no interaction effect between gender and stereotyping score, = .07, p R2 < .01. Gender was the only category in which there existed a significant relationship between indirect aggression stereotyping and corresponding characteristics of nominated peers. After controlling for covariates, analyses failed to demonstrate significant unique effects for black white ( = .07, p > .05 R2 < .03) or popular unpopular ( = .02, p R2 < .01) stereotyping scores. Summary Students scores on aggression stereotyping scales appeared to be related to the gender of peers nominated as aggressive, and this was true for both direct and indirect

PAGE 52

52 aggression. Furthermore, this relationship did not vary by the gender of the nominating student. Gender was the only stereotyping category for which this relationship was present. No significant relationships were found between aggression stereotyping and either the race or popularity of nominated peers Stereotyping Scores and Aggression Group M embership Prior analyses tested whether social schemas influenced the peers whom students nominated as aggressive. Additionally, it was hypothesized that scales measuring aggression stereotyping might prove to be useful as a control variable for future studies employing peer reported aggression measures, similar to the way in which social desirability bias measures are currently used to explain variance in self reported aggression. The efficacy of social desirability bias in accounting for discrepancies between aggression measurement methods was supported in the current sample A one way ANOVA comparing aggression groups on mean social desirability scores indicated that, for both direct and indirect aggression, the high self group had significantly lower levels of social desirability bias ( M = .36, SD = .04) than all other groups ( M = .49 to .53, SD = .02 to .04; F (3,292) = 5.37, p < .01). Thus, students in the high self group, who had been identified as aggressive by themselves only, could be distinguished from students who had been identified as aggressive by multiple source s by an increased willingness to report socially undesirable behavior. It was similarly hypothesized that students who were identified as aggressive solely by peers might be distinguished from other students by the degree of aggression stereotyping pres ent in the peers who had nominated them in other words, whether students in the high peer group were more likely than those in other groups to be the targets of aggression stereotyping bias. If a student was nominated as aggressive, it was possible to c alculate the average stereotyping sco res on each scale of the peers who nominated that student. A series of two way

PAGE 53

53 ANOVAs and planned follow up comparison tests of estimated marginal means investigated group differences in nominators average stereotyping scores. Analyses were run separately for direct and indirect aggression. Because nominators average stereotyping scores were expected to skew in opposing directions for each opposing pair of social groups which comprised the endpoints of the scales, t he interaction between group membership and the nominees corresponding gender, racial, or sociometric category was also assessed. As an example, for the analysis of aggression group differences in nominators average scores on the girls boys direct aggre ssion stereotyping scale, group membership and gender were both entered into the model as predictors, and the main effects of each were investigated along with the effect of the interaction term. Due to the fact that not all students received nominations on peer report aggression items (and thus could not be assigned average nominators stereotyping scores), these analyses included only those students who were nominated at least once on the peer report aggression measure. One hundred sixty three students were nominated at least once on direct items and thus included in analyses of direct aggression, and 184 students were nominated at least once on indirect items and thus included in analyses of indirect aggression. Note that the high peer report and combi ned (peer and self) report groups had to have 4 or more nominations in order to be classified as high on peer report. Thus, students who received at least one but fewer than four nominations comprised a nominated group that was compared to the high peer report and combined groups ( N = 85 for direct aggression and 103 for indirect aggression). Findings Of the race, gender, and popularity variables, only gender showed significant main effects for differences in nominators average scores on the girls boys stereotyping scales; this was true for both direct ( F [ 1,155] = 8.36, p < .01) and indirect aggression ( F [ 1,174] = 7.95,

PAGE 54

54 p < .01). While the nominators of both boys and girls tended to have scores that skewed farther toward the boys side for direct aggression, this was more true for the nominators of boys ( M = 1.20, SD = .07) than for girls (M = .86, SD = .10). Similarly, while the nominators of both boys and girls tended to have scores that skewed farther toward the girls side for indirect aggress ion, this was more true for the nominators of girls ( M = 1.08, SD = .08) than for boys (M = .73, SD = .10) No main effects were found for race or for perceived popularity. Importantly, there were no significant aggression group differences in any of the nominators average stereotyping scores, nor were any significant interaction effects present across analyses. (Because of the absence of significant findings, results from individual analyses are not reported here for the sake of brevity.) The absence of increased stereotyping bias in the nominators of students in the high peer group compared to other groups suggests that, unlike social desirability bias, the aggression stereotyping scales cannot be used to explain discrepancies between self and peer r eported aggression scores. Nominators Average Stereotyping Scores and Levels of PeerReported A ggression A final set of analyses examined whether nominators stereotyping scores explained variance in peer reported aggression over and above what was alrea dy explained by self reported aggression scores. Although the previous set of analyses suggested that the use of stereotyping scales in this role may be limited, a main impetus of the current study was the possibility that stereotyping scales could be uti lized as a control variable to explain variance in peer reported aggression. Thus, I re examined this relationship using continuous forms of self and peer reported aggression, as most studies of aggression in adolescence do not employ a groupbased approa ch. A series of hierarchical linear regressions were conducted to examine whether nominators stereotyping scores explained variance in peer reported aggression over and above

PAGE 55

55 what was already explained by self reported aggression scores. Again, these analyses included only those students who were nominated at least once on the peer report aggression measure. Direct and indirect aggression were investigated separately. In each analysis, peer reported aggression was the dependent variable. Gender, race (entered as African American vs. all others), and perceived popularity were entered in Step ,1, and self reported aggression was entered in Step 2. Together, gender, race, and perceived popularity explained 18.7% of the variance in peer reported direct ag gression and 18.0% of the variance in peer reported indirect aggression. After accounting for these variables, self reported direct aggression predicted an additional 2.3% of the variance in peer reported direct aggression ( = .16, F [1,153] = 11.81, p < .001) ; and self reported indirect aggression predicted an additional 2.0% of the variance in peer reported indirect aggression ( = 15 F [1,172] = 4.24, p < .01). In step 3 of each analysis, nominators average stereotyping scores on a particular scale (e.g., girls boys direct aggression) were entered along with the interaction between the scale and its corresponding demographic or sociometric category (gender, race, or perceived popularity). This was run six times in order to investigate the impact of each stereotyping scale separately. One marginally significant effect emerged. Nominators average scores on the girls boys indirect aggression stereotyping scale was marginally related to peer reported indirect aggression ( = .13, p = .067). The interaction between gender and nominators stereotyping scores was also marginally related to peer reported indirect aggression ( = .12, p = .069), indicating that the relationship between nominators scores and peer reported indirect aggression was slightly stronger for boys than it was for girls. Together, they accounted for an additional

PAGE 56

56 3% of variance in peer reported aggression ( F [2,170] = 3.44, p < .001). No effects were significant for any other category of nominators average stereotyping scores.

PAGE 57

57 Table 3 1. Group N s and Aggression Means and Standard Deviations by Aggressor Group Aggression Type Group N Self Reported Direct Peer Reported Direct Self Reported Indirect Peer Reported Indirect Direct aggression Low 200 1.35 (.28) .33 (.12) 1.54 (.42) .27 (.50) High self 37 2.50 (.48) .29 (.11) 1.96 (.55) .25 (.49) High peer 41 1.47 (.30) .93 (1.67) 1.68 (.44) .94 (1.67) High multiple 35 2.61 (.57) 1.17 (1.68) 1.86 (.38) .89 (1.43) Indir ect aggression Low 179 1.48 (.49) .29 (.20) 1.40 (.27) .43 (.19) High self 53 1.97 (.76) .15 (.62) 2.24 (.29) .35 (.23) High peer 54 1.66 (.54) .63 (1.55) 1.51 (.24) 1.05 (1.16) High multiple 27 2.04 (.71) 1.04 (1.94) 2.32 (.36) 1.64 (1.61) Note. Self rep orted values represent mean scores on 1 5 Likert scale. Peer reported values have been standardized so that the full sample mean for each variable = 0; SD = 1.

PAGE 58

58 Table 3 2. Membership agreement for direct and indirect a ggression groups Indirect Groups Direct Groups Low High self High peer High multiple Low 141 29 22 8 High self 16 16 3 2 High peer 13 4 16 8 High multiple 9 4 13 9 Note. Values represent the number of participants falling into e ach category.

PAGE 59

59 Table 3 3. Means and Standard Deviations for Aggression and Aggression Stereotyping Measure/Variable Direct Indirect Aggression self reported 1.64 (.61) 1.64 (.46) peer reported (unstandardized) 4.1 (10.78) 2.7 (5.20) Aggr ession stereotyping boys girls .96 (.91) .81 (.84) black white .72 (.89) .17 (.71) popular unpopular 1.15 (1.01) 1.43 (.96) Note. Standard deviations in parentheses. Values for aggression stereotyping scales have been centered so that 0 represents a neutral view, negative scores represent a shift in the direction of the left hand (first listed) group, and positive scores represent a shift in the direction of the right hand (second listed) group.

PAGE 60

60 Table 3 4. Observed and Expected G ender and Race Distributions across Aggression Groups Direct Aggression Indirect Aggression Category/Level Low High Self High Peer High Multiple Low High Self High Peer High Multiple Gender Boys observed (expected) 84 (102) 21 (19) 29 (21) 25 (18) 106 (92) 19 (27) 24 (27) 10 (14) % within aggression group 41.4% 56.8% 70.7% 71.4% 58.2% 35.8% 44.4% 37.0% Girls observed (expected) 119 (101) 16 (18) 12 (20) 10 (17) 76 (90) 34 (26) 30 (27) 17(13) % within aggression group 58.6% 43.2% 29.3% 28.6% 41.8% 64.2% 55.6% 63.0% Race/Ethnicity European American observed (expected) 114 (102) 18 (18) 14 (21) 13 (18) 91 (92) 25(27) 31 (27) 9 (14) % wi thin aggression group 56.2% 48.6% 34.1% 50.3% 51.6% 47.2% 57.4% 33.3% African American observed (expected) 32 (47) 7 (9) 16 (10) 18 (8) 34 (42) 12 (12) 18 (13) 9 (6) % within aggression group 15.8% 18.9% 39.0% 51.4% 18.7% 22.6% 33.3% 33.3% Other observed (expected) 57 (54) 12 (10) 11 (11) 4 (10) 54 (48) 16 (14) 5 (14) 9 (7) % within aggression group 28.1% 32.4% 26.8% 11.4% 29.7% 30.2% 9.3% 33.3%

PAGE 61

61 Table 3 5. Summary of hierarchical reg ression analyses for partic i pants gender stereotyping scales predicting gender percentages of nominated peers Direct Aggression Indirect Aggression Variable R 2 R 2 Step 1 .10*** .28*** gender (female vs. other) .35*** .50*** grade level .05 .24*** Step 2 .06*** .04** gender (female vs. other) .32*** .47*** grade level .01 .23*** girls boys stereotyping score .23** .20** Step 3 <.01 <.01 gender (female vs. other) .34*** .47*** grade level < .01 .24*** girls boys stereotyping score .23** .21** gender X girls boys stereotyping score .10 .07 ***p < .001. ** p < .01. Note. Dependent variables = % males out of the total number of nominated peers for direct and indirect aggression items, respectively. Only covariates which remained significant in at least one of the final models have been reported here.

PAGE 62

62 Figure 3 1. Graphical representation of group means by aggression type f or behavioral and emotional characteristics Direct and indirect aggression groups were examined in separate analyses. Values are mean scores on 1 5 Likert Scales. Brackets around each mean represent 95% confidence intervals. Significant group differences at p < .05 are represented by pairs of confidence interval brackets which overlap by < ~30%; significant group differences at p<.01 are represented by pairs of confidence interval brackets which show no overlap (Cumming & Finch, 2005). Manipulative behavior 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Direct Indirect Remorselessness 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Direct Indirect Anger 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Direct Indirect Anger regulation 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Direct Indirect Empathy 1 1.5 2 2.5 3 3.5 4 Direct Indirect Group High self High peer High multiple Low

PAGE 63

63 Figure 3 2. Graphical representation of group means by aggression type for sociometric characteristics Direct and indirect aggression groups were examined in separate analyses. Variables are standardized so that the sample mean = 0; SD = 1. Brackets around each mean represent 95% confidence intervals. Significant group differences at p < .05 are represented by pairs of confidence interval brackets which overlap by < ~30%; significant group differences at p<.01 are represented by pairs of confidence interval brackets which show no overlap (Cumming & Finch, 2005). Social preference -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Direct Indirect Social visibility -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Direct Indirect Perceived popularity -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Direct Indirect Group High self High peer High multiple Low

PAGE 64

64 CHAPTER 4 DISCUSSION The present study addressed two issues concerning the identification of aggression in early adolescence. First, to determine whether self report and peer repor t measurement methods identify students with differing psychosocial profiles, groups of self identified, peer identified, self and peer identified, and low aggressors were compared on several emotional, behavioral, sociometric, and demographic characterist ics. Second, the incidence of aggression stereotyping bias among early adolescents was examined, along with its relationship to the demographic and sociometric characteristics of adolescents identified as aggressive by their peers, and its potential use as a control variable to explain variance in peer reported aggression. Do Self and PeerR eport Aggression Measures Identify Students with Different Psychosocial P rofiles? Although self and peer reported measures of aggression are usually intended to be measurements of the same underlying construct, research on aggressive behavior in childhood and adolescence has found relatively low agreement among the two types of measurement methods. This suggests that self and peer report methods of assessing aggre ssion are identifying different groups of students. One goal of the present study was to determine whether students identified as aggressive by self and peer report methods differed from one another on a range of key psychosocial characteristics. A gr oupbased analytic approach was used to identify students who were high in self reported aggression (high self), high in peer reported aggression (high peer), high in both self and peer reported aggression (high multiple), or high in neither (low) Groups were created for both direct and indirect forms of aggression and compared with one another on several behavioral, emotional, sociometric, and demographic indicators. It was hypothesized that self identified aggressive students would have eleva ted levels of antisocial emotional and

PAGE 65

65 behavioral traits, while peer identified aggressors would be characterized by patterns of demographic and sociometric characteristics which resembled common social stereotypes of aggressive adolescents (e.g., mean gi rls). Specifically, peer identified direct aggressors were expected to be characterized by higher percentages of male and African American students, while peer identified indirect aggressors were expected to be characterized by higher percentages of fema le students, lower levels of social preference, and higher levels of social impact and perceived popularity. High Self vs. H igh P eer: Antisocial I ndicators Of particular interest in the analyses were comparisons of the high self and high peer groups that is, students who were identified solely by self report or solely by peer report. Results suggested that, after controlling for social desirability and relevant demographic indicators, such as gender, race, and grade level, self identified and peer identified aggressive adolescents differed somewhat from one another on emotional and behavioral indicators of antisocial personality. For both direct and indirect aggression, the high self group showed significantly higher levels of manipulative behavior than the high peer group. In addition, for direct aggression, the high self group had significantly elevated levels of both anger and remorselessness over the low aggression group, while the high peer group was indistinguishable from low aggressors on these va riables. These results may suggest that direct aggression is a better indicator of antisocial tendencies than indirect aggression. However, on the majority of antisocial indicators, although the high self groups levels were higher than those of the hi gh peer group, the high self and high peer groups did not significantly differ from one another. It is possible that this was due to the fact that the high aggression group sizes were somewhat small, which may have affected the power of the analyses. Nev ertheless, Clemans and Sontag (2009, April) found significant

PAGE 66

66 differences between high self and high peer report groups with similar group sizes. As such, the hypothesis that self and peer identified aggressors would show distinct differences on indicat ors of antisocial personality was partially supported by the present study. High Self vs. High P eer: Sociometric and Demographic C haracteristics One area in which hypotheses of group differences were fully supported, however, was sociometric characteris tics. Social standing within the peer group was assessed by social preference (how well liked one is ), social impact (how socially visible one is ) and perceived popularity (how popular one is considered by his/her classmates). For both direct and indir ect aggression, all three variables showed similar, pronounced patterns of results, with the high peer group being less well liked, more socially visible, and higher in perceived popularity than the high self group or the low aggressors. Furthermore, it w as expected that group differences in demographic characteristics would vary between direct and indirect aggression. For gender, this proved to be the case: Regardless of measurement type, boys were more likely than girls to be identified as direct aggres sors, and girls were more likely than boys to be identified as indirect aggressors. Su r prisingly, no significant gender differences across the high aggression groups were observed for either type of aggression, although it had been hypothesized that these gender patterns would be significantly more pronounced in the high peer groups than in the high self groups. There were, however, differences between the high self and high peer groups in racial categorization. Specifically, African Americans made up o nly 19% of the high self group (which was similar to the percentage of African Americans in the full sample), but comprised nearly 40% of the high peer group. An elevated percentage of African Americans was also seen in the high peer group for indirect aggression (33%), whereas the high self group again was comparable to the racial distribution of the full sample.

PAGE 67

67 In summary, it appeared that students identified solely by self reported measures (especially those who identified themselves as being high in direct aggression) tended to have elevated levels of antisocial behavioral and emotional characteristics over students low in aggression, but resembled low aggressors in demographic and sociometric characteristics. In contrast, students identified solely by peer reported measures mimicked students with low aggression in emotional and behavioral traits, but were distinguished from low aggressors by lower social preference, higher social impact higher perceived popularity and greater percentages of African American students Gender, surprisingly, was the only variable category which did not demonstrate different patterns of findings for self and peer reported aggressors. The High M ultiple G roup : True A ggressors? There was quite a bit of evidence that, f or both direct and indirect aggression, the high multiple report group was distinct from either of the other two high aggression groups and may be the most valid representation of aggressive adolescents This group, for instance, had the most pronounced l evels of antisocial indicators: students in the high multiple group were significantly different than low aggressors on almost every indicator which demonstrated differences between groups. For direct aggression groups, this included higher levels of mani pulative behavior, remorselessness, anger, and lower levels of anger regulation; for indirect aggression groups, higher levels were seen for remorselessness and anger. In some cases, the high multiple group also showed more elevated levels of antisocial i ndicators than both of the other high aggression groups, although patterns suggested that on these indicators, the high multiple group tended to have levels that more closely resembled those of solely self identified aggressors than those of solely peer id entified aggressors. In sociometric characteristics, however, the high multiple group was indistinguishable from the high peer group. For both direct and indirect aggression, both the high multiple and

PAGE 68

68 high peer groups were less well liked, more socially visible, and more popular than the high self or low aggressors. Additionally both the high peer and high multiple groups had elevated percentages of African American students (this was particularly true for direct aggression). In short, the high multiple group conformed to nearly all characteristics measured in this study that were found in previous research to be characteristic of aggressive adolescents, while the high self group tended to be distinguished only on antisocial tendencies and the high pee r group tended to be distinguished only on demographic and sociometric characteristics. Achenbach (1987) suggested that both self and peer report methods may provide unique and important information about adolescent problem behavior that would not be captured by other informant methods. As such, the use of multiple informant methods as a general rule for research on aggressive behavior may be advised. In addition, while several studies have included peer and self reported behaviors, often the decision is made to use only one reporting type in analyses. The present investigation suggests that this is not an arbitrary decision and will likely impact the results. Ideally, future studies should report how finding s may have varied by reporter. Of course, not all research endeavors have the time and monetary resources to collect aggression data from multiple informants. For those studies utilizing a single measure of aggression, potential sources of bias become a more salient issue. We know that social de sirability accounts for variance in self reported aggression, and for this reason it is frequently utilized as a control variable in studies that employ self reported aggression measures. On the other hand, no similar ubiquitous assessment of bias exists for peer reported aggression measures. Therefore, the second goal of the current study was to investigate the utility of a measure designed to assess students tendency to endorse demographic and sociometric

PAGE 69

69 stereotypes of aggressive behavior, as it was h ypothesized that this tendency might result in bias in nominations of aggressive peers. Are Stereotyping Scores Related to the Demographic and Sociometric Characteristics of Peers Nominated as A ggressive? A series of scales designed to measure direct and indirect aggression stereotyping bias were created for the purpose of this study. Three scales, each item of which comprised a spectrum with a dichotomous pair of gender, racial, or popularity based groups as its endpoints, described an aggressive behavior and asked students to rate whether each item applied more to the first group, more to the second group, or equally to both groups. Each scale demonstrated good distribution and skewed away from center in the direction one would expect according to commo n social stereotypes of aggressive adolescents. Students tended to endorse the view that boys were more directly aggressive than girls, that girls were more indirectly aggressive than boys, that African American students were more directly aggressive than European American students, and that popular students were more directly and indirectly aggressive than unpopular students. The association between popularity and indirect aggression was the strongest of all the scales. Internal reliability was acceptable for racial and popularity scales, and on the low side (though not in the unacceptable range) for gender scales. The relationship of the aggression stereotyping scales to students tendencies to nominate particular demographic and sociometric categories o f aggressive peers was examined. Effects were found mostly for those scales which tapped gender stereotyping. Linear regression analyses indicated that, after controlling for gender, a students score on the direct gender stereotyping scale was significantly associated with the gender distribution of the peers he or she nominated as directly aggressive, and the same was true for indirect aggression. For each scale, the more one skewed toward the boys side of the stereotyping spectrum, the more likely one

PAGE 70

70 was to nominate male peers on peer report aggression items (and vice versa for females). Thus, ones tendency to endorse a gender typed view of aggression does appear to explain variance in the gender of the peers one identifies as being aggressive. Si milar results, however, were not found for race or for popularity. Although stereotyping scores partially accounted for gender differences in nominated peers, this relationship did not appear to be helpful in explaining variance in overall levels of peer reported aggression. A comparison of students who were solely peer identified high aggressors and students who were identified as aggressive by multiple sources revealed no significant differences in stereotyping scores between the two; furthermore, no gr oup differences existed when the interactions between stereotyping scores and demographic/sociometric characteristics were examined. Finally, with the exception of one marginal effect for the indirect aggression gender scale, stereotyping scores did not e xplain any significant variance in peer reports of aggression over and above that already explained by participants self reported aggression scores. As such, although the stereotyping scales used in this study do seem to predict some characteristics of a ggressive peer nominations, as a whole they do not appear to be particularly useful a s control variable s for peerreported aggression. The ineffectiveness of the stereotyping scales to explain variance in levels of peer reported aggression might be interp reted in several ways. It was hypothesized that bias due to social schemas of aggressive adolescents would be associated with the type of peers one nominated as aggressive, and thus with overall peer reported aggression levels for particular demographic a nd sociometric categories of adolescents. Although the first part appeared to be true in the case of gender, the study found no significant relationships between levels of peer reported aggression and scores on any of the stereotyping scales. These findi ngs make sense in

PAGE 71

71 that among those who stereotype, nominations are influenced by these stereotypes. Yet across young adolescents, nominations demonstrated expected gender distributions as found in other studies: that is, boys have higher rates of direct ag gression and girls and boys had similar rates of indirect aggression (e.g., similar numbers of nominations). Of course, it may also be the case that adolescents social schemas about aggressive peers are having a greater influence on peer reports of aggression than found in the present study, but that the stereotyping scales created for this study are simply not doing a very good job of accurately describing those schemas. For instance, a scale which allowed one to rate social groups individually, rather than through comparison format, may have produced a more nuanced picture of adolescent s views on the relative levels of aggression among social groups. In addition, the scales only assessed social schemas of gender, race, and popularity; it is possible t hat adolescents social schemas of aggression involve demographic and/or social criteria that were not assessed, such as socioeconomic status, physical appearance, or membership in specific social crowds. Although social desirability bias was not related to the patterns of answers observed for the stereotyping scales, there was evidence that some students were uncomfortable categorizing social groups in this way, particularly for the scale which compared racial groups. It is also possible, then, that the ineffectiveness of stereotyping scales to explain variance in peer reported aggression may be due to their relatively explicit format. Since explicit and implicit stereotypes can differ from one another in direction (Whitley & Kite, 2006) a measure desig ned to assess implicit aggression stereotyping may produce different patterns of results and affect overall conclusions about the role of stereotyping in peer reported aggression.

PAGE 72

72 Strengths and L imitations A strength of the present study is that it utiliz ed a passive consent procedure, which is crucial for research investigating aggression and other socially undesirable behavior in the school setting (Tigges, 2003). Ninety six percent of all middle school students at the school participated in the study. Thus, the peer report data for this school is considered to be highly reliable ( Crick & Ladd, 1990 ). However, the investigation of these relationships in only one setting is a potential limitation and may affect the generalizability of the results. The school in question is not a traditional public middle school : A lthough there are no tuition costs and it strives to enroll a student body that is demographically representative of the larger county, students and their families have to complete an applicati on process, which necessitates some level of parental involvement and selection. In addition, due to the fact that it is a research school affiliated with a large state university, students may have had more experience with educational and psychological r esearch than students at other schools in the county. Thus, the student body may differ from other schools in ways which affected the results of the study. Replication of these findings in other middle school settings would be required to fully address t his question. Additionally, it should be observed that self and peer reported aggressors differed from low aggressors mostly on scales which employed the same measurement type as the aggression scale in question ; that is, self reported aggressors had ele vated levels of self reported, socially undesirable behavioral and emotional characteristics, whereas peer reported aggressors differed from low aggressors on sociometric variables which were derived from peer nominations. The possibility of measurement b ias is thus a concern. To address the issue of measurement bias within self reports, a measure of social desirability bias was incorporated into the survey, which assesse d participants unwillingness to admit to common but social undesirable thoughts, feelings and behaviors. Accordingly, social desirability bias was negatively correlated with

PAGE 73

73 scores for manipulative behavior, remorselessness, anger, and both direct and indirect self reported aggression. After factoring out variability due to social desir ability bias, however, relationships between self reported aggression and manipulative behavior, remorselessness, and anger remained significant. Furthermore, self reported direct aggression showed significant negative associations with anger regulation, which was not itself related to social desirability bias but fit conceptually into a constellation of antisocial characteristics. Together, these findings increase the likelihood that self reported results are representative of true findings and not artif acts of measurement error. At the present time, there is no established construct that can be controlled for in order to reduce bias in peer reports Rather, additional research in needed on factors that influence reporter bias (see below). It should be noted, however, that peer reported sociometric variables actually incorporate peers opinions into their definition (e.g., whom do you like? Whom do you think is popular?) and thus may be more robust to potential bias than peer reported variables which measure behavior. The present study sought to provide explanations for discrepancies between self reported and peer reported measurement of aggression by assessing the roles of social desirability bias and aggression stereotyping bias. It should be noted that I left unexamined many other potential reasons as to why self and peer reports of aggression may differ in adolescence. For instance, the same behavior may appear aggressive when coming from a member of one particular social group but not when coming f rom another. Similarly, students may rate a behavior as aggressive in an unknown peer, but may not consider the same behavior to be aggressive if performed by a friend. Discrepancies between self and peer reported aggression may also arise from a lack o f awareness on the part of some self raters to recognize their behavior as aggressive in the eyes of other students. In addition to examining the effects of social desirability and social stereotyping,

PAGE 74

74 a thorough investigation of the model in Figure 1 1 s hould also take these factors into account. Finally, the direction of effects in the relationship between stereotypes and reports of peer behavior should also be investigated in order to shed light on the origin and maintenance of social schemas of aggression. Developmental C onsiderations The present study focused on the measurement of aggressive behavior during the early adolescent period, using a sample which was approximately 11 to 14 years old. This age group was selected in part because longitudi nal investigation of self and peer reports of aggression indicated that agreement between the two is higher during this particular stage than at any other time in childhood or adolescence. Thus, estimations of discrepancies between self and peer r eports w ere expected to be at their most conservative during this time, increasing the likelihood that the differences found would apply to other stages of childhood and adolescence as well In addition, both self and peer reports of behavior are often used with early adolescents due to the fact they are mature enough to complete lengthy survey measures, yet still attend schools in which they are familiar with the majority of same grade peers. At the same time, early adolescence is a unique period of the life spa n. For instance, early to midadolescence is the life stage during which the influence of the peer group on physical appearance, likes/dislikes, and social behavior is at is strongest (Brown, 2004). This is also a period when gender intensification, which refers to strengthening in the adherence to traditional gender roles, is observed in many individuals ( Galambos, Almeida, & Petersen 1990; Hill & Lynch, 1983; see Clemans, DeRose, Graber, & Brooks Gunn, 2010, for a review of gender development in adolesce nce). Finally, early adolescence marks the end of the period of pervasive gender self segregation which begins in middle childhood (Maccoby, 1998) As such, it is a unique developmental time in a young persons life, and it remains unclear as to whether

PAGE 75

75 relationships between stereotyping and peer reports of aggression undergo systematic changes across longer age ranges For instance, the tendency of some early adolescents to intensify in their endorsement of traditional gender role stereotypes may make g ender stereotypes of aggression more salient at this period than they would be at other developmental periods. Future studies which examine this relationship should ideally do so at multiple points in childhood and adolescence to assess possible developmental trends. Conclusions The results of the present study suggest that the identification of aggressive individuals is a multifaceted issue and that adolescents with differing demographic and psychosocial characteristics may be identified depending on the method of assessment (Graham et al., 2003). It is important that researchers and educators seeking to identify and/or curb aggressive behavior in school settings remain aware that assessment methods may be providing different perspectives on aggressive be havior. The present study investigated discrepancies between self and peer reports of aggression; however, parent and teacher reports are also commonly utilized in childhood and early adolescence, and should be similarly examined in regard to their relati onships with other reporting methods. The use of an aggregate measure of aggression derived from multiple report methods, rather than the reliance on a single measurement type, is advised for future studies. The present study also tested the hypothesis t hat bias due to social schemas of aggressive adolescence would account for variance in peer reports of aggression. Although the applicability of the social stereotyping scales used here was limited, research in this area has been extremely scarce, and fut ure studies which investigate this relationship and other potential sources of peer report bias are warranted. The endorsement of expected aggression stereotypes by participants in the present study indicates the continued salience of demographic and soci al stereotypes in our

PAGE 76

76 society. I t is important to understand how these and other biases may influence peer relationships and promote labeling in the peer group not merited by behavior.

PAGE 77

77 APPENDIX A SURVEY MEASURES Aggression Stereotyping Instructions: T he following three sections will ask you about things other kids do. Please make sure to read the instructions carefully. For each item, indicate how much you think the statement applies to (group 1) or (group 2) (This instruction was repeated three tim es at the beginning of each scale; group 1 and group 2 were replaced with girls or boys black kids or white kids and popular kids or unpopular kids Scale: 7 point spectrum; 1 = Mostly true of (group 1); 4 = Equally true of both groups; 7 = Mostly true of (group 2) Indirect Aggression Stereotyping 1. Leave other kids out on purpose. 2. Gossip or say mean things about other kids behind their backs. 3. Spread rumors about other kids. 4. Give other kids the silent treatment. 5. Tell other kids they wont be their friend anymore in order to get something they want. Direct Aggression Stereotyping 1. Tease other kids. 2. Call other kids mean names to their face. 3. Push or shove other kids. 4. Get in fights a lot. 5. Threaten other kids. Prosocial Filler Items 1. Are nice and friendly to people when they need help. 2. Stick up for kids who are being picked on or excluded. Peer Reported Aggression Instructions: the following section is about other kids in your grade. For each question, write the first and la st names of the kids to whom you think the statement best applies. Do not write anything else about any person except his or her first and last name. If you are unsure of how to spell a name, please look it up on the roster provided. You can nominate the same person for more than one item if you want. Wri te the names of kids who Peer Reported Indirect Aggression 1. say mean things about other kids behind their backs. 2. g ive other kids the silent treatment. 3. s pread rumors about other kids to dama ge their social reputation 4. leave other kids out of activities that those other kids really want to be included in.

PAGE 78

78 5. t ell other kids they wont be their friend anymore in order to get back at them for something Peer Reported Indirect Aggression 1. t ease other kids to make them angry 2. get into physical fights. 3. push or shove other kids. 4. threaten to hurt or hit other kids. 5. c all other kids bad names to their face.

PAGE 79

79 APPENDIX B CORRELATIONS BETWEEN STUDY VARIABLES

PAGE 80

80 Variable 1 2 3 4 5 6 7 8 9 10 1. Self reported direct aggression 2. Self reported indirect aggression .45** 3. Peer reported direct aggression .33** .19** 4. Peer reported indirect aggression .19** .24** .70** 5. Direct aggression stereo typing (gender) .02 .02 .02 .05 6. Direct aggression stereotyping (race) .07 .03 .10 .02 .12* 7. Direct aggression stereotyping (popularity) .06 .05 .09 .12* .14* .12* 8. Indirect aggression stereotyping (gender) .02 .14* .01 .07 .14* .17** .13* 9. Indirect aggression stereotyping (race) .03 .06 .03 .00 .09 .16** .10 .03 10. Indirect aggression stereotyping (popularity) .02 .11 .01 .05 .17** .18** .61** .20** .01 11. Manipulative behavior .46** .49** .16** .12* .05 .03 .02 .03 .09 .01 12. Remorselessness .34** .26** .17** .16** .01 .09 .09 .04 .04 .01 13. Empathy .21** .02 .15** .04 .12* .10 .02 .05 .02 .06 14. Anger .36** .35** .09 .06 .01 .05 .10 .03 .05 .14* 15. Anger regulation .22** .07 16** .04 .06 .13* .05 .08 .08 .06 16. Social desirability .34** .44** .02 .04 .00 .02 .04 .02 .03 .09 17. Social preference .14* .04 .52** .38** .10 .08 .18** .06 .00 .12* 18. Social impact .14* .17** .58** .60** .03 .02 .01 .03 .0 5 .07 19. Perceived popularity .06 .17** .19** .39** .12* .09 .05 .04 .03 .11 p < .05. ** p < .01.

PAGE 81

81 Variable 11 12 13 14 15 16 17 18 1. Self reported direct aggression 2. Self reported indirect aggression 3 Peer reported direct aggression 4. Peer reported indirect aggression 5. Direct aggression stereotyping (gender) 6. Direct aggression stereotyping (race) 7. Direct aggression stereotyping (popularity) 8. Indire ct aggression stereotyping (gender) 9. Indirect aggression stereotyping (race) 10. Indirect aggression stereotyping (popularity) 11. Manipulative behavior 12. Remorselessness .51** 13. Empathy .19** .38** 14. Anger .29** .32** .03 15. Anger regulation .09 .19** .36** .04 16. Social desirability .38** .25** .01 .43** .11 17. Social preference .05 .20** .18** .10 .19** .09 18. Social impact .09 .06 .13* .05 .01 .02 .14 19. Perce ived popularity .08 .09 .10 .07 .04 .03 .39** .44** p < .05. ** p < .01.

PAGE 82

82 APPENDIX C STUDY VARIABLE MEAN DIFFERENCES FOR GENDER, RACE, AND GRADE LEVEL Significant differences within selected study variables GENDER Measure/Variab le Boys M (SD) Girls M (SD) t df p Aggression self reported direct aggression 1.32 (.23) 1.20(.19) 4.64 (311) <.001 self reported indirect aggression 1.24 (.16) 1.30 (.18) 2.74 (311) .006 peer reported indirect aggression 1.01 (.40) .87 (.20) 3.98 (315) <.001 Aggression stereotyping boys girls (indirect) .69 (.81) .93 (.86) 2.44 (297) .015 popular unpopular (indirect) 1.29 (1.01) 1.56 (.89) 2.50 (297) .013 Peer nominations on aggression items direct aggressi on % male .88 (.23) .70 (.34) 4.62 (234) <.001 direct aggression % African American .42 (.33) .55 (.33) 3.04 (234) .003 indirect aggression % male .69 (.40) .29 (.36) 7.48 (189) <.001 Behavioral/Emotional Characteristics manipul ative behavior 2.28 (.71) 2.04 (.65) 3.02 (306) .003 remorselessness 2.26 (.76) 2.01 (77) 2.74 (304) .006 empathy 3.32 (.49) 3.73 (.43) 7.70 (299) <.001 anger reduction 2.16 (.76) 2.63 (.76) 5.50 (308) <.001 Sociometric Characteristics social preference .18 (1.14) .18 (.77) 3.23 (309) .001 Note. Analyses performed on all relevant study variables. Only those significant at p < .05 are listed.

PAGE 83

83 Significant gender differences within selected study variables RACE Measure/V ariable White M (SD) African American M (SD) Other M (SD) F df p Aggression self reported direct aggression 1.24 (.20) a 1.33 (.25) a,b 1.24 (.21) b 4.87 (2,310) .008 peer reported direct aggression .90 (.28) a 1.13 (.44) a,b .87 (.21) b 17.10 (2,3 13) <.001 peer reported indirect aggression .90 (.33) a 1.05 (.44) a,b .86 (.38) b 5.67 (2,313) .004 Aggression stereotyping girls boys (direct) 1.04 (.93) a .70 (.88) a,b 1.01 (.84) b 3.78 (2,300) .024 black white (direct) .79 (.91) a .36 (.78) a,b .88 (.88) b 6.59 (2,285) .002 black white (indirect) .06 (.67) a .37 (.67) a .20 (.77) 4.68 (2,286) .010 Peer nominations on aggression items indirect aggression % White .48 (.38) a .30 (.36) a .40 (.38) 3.40 (2,188) .036 Behavi oral/Emotional Characteristics empathy 3.58 (.51) a 3.35 (.47) a,b 3.57 (.49) b 5.74 (2,298) .004 Sociometric Characteristics social preference .03 (.97) .29 (1.10) a .20 (.89) b 4.91 (2,308) .008 Note. Analyses performed on all relevan t study variables with posthoc Games -Howell correction due to unequal group sizes; only variables with significant differences are listed. Same-subscript pairs within a single variable indicate significant mean differences at p < .05.

PAGE 84

84 Significant gender differences within selected study variables GRADE LEVEL Measure/Variable 6th M (SD) 7th M (SD) 8th M (SD) F df p Aggression stereotyping girls boys (direct) 1.09 (.97) a 1.08 (.90) b .70 (.79) a,b 6.19 (2,300) .002 black white (direct) .96 (.98) a,b .59 (.85) a .62 (.79) b 5.25 (2,285) .006 black white (indirect) .06 (.81) a .36 (.65) a .18 (.59) 8.69 (2,296) <.001 Peer nominations on aggression items direct aggression % African American .51 (.32) a .35 (.33) a,b .58 (.3 1) b 10.76 (2,229) <.001 indirect aggression % African American .43 (.40) a .20 (.32) a,b .41 (.38) b 8.28 (2,184) <.001 Behavioral/Emotional Characteristics manipulative behavior 2.03 (.65) a 2.12 (.64) 2.33 (.75) a 5.31 (2,302) .005 rem orselessness 2.02 (.74) a 2.14 (.78) 2.29 (.80) a 3.06 (2,300) .048 anger reduction 2.57 (.85) a 2.23 (.72) a 2.37 (.78) 4.69 (2,307) .010 Note. Analyses performed on all relevant study variables with post hoc Bonferroni correction; only variables with s ignificant differences are listed. Same subscript pairs within a single variable indicate significant mean differences at p < .05.

PAGE 85

85 APPENDIX D SUMMARY OF QUESTION 1 ANALYSES WITH MORE STRINGENT GROUP CRITERIA All analysis addressing Question 1 we re re run using group membership cutoff criteria of one SD for self reported aggression and 5 nominations for peer reported aggression (referred to below as more stringent ) In general, patterns of results mirrored those of the .5 standard deviation/4 nomination analyses (referred to below as less stringent) Differences in the patterns of results were observed for emotional and behavioral characteristics only; these are reported below. Patterns of results for demographic and sociometric characterist ics did not change. Emotional and Behavioral Characteristics Manipulative behavior. Us ing the more stringent criteria for direct aggression, estimated marginal means comparisons indicated that the high multiple group had significantly higher levels of ma nipulative behavior than the low aggression group only. With the less significant criteria, it was higher than both the low aggression and high peer groups. This was because the estimated marginal mean of the high peer group increased slightly, from 2.09 to 2.23. All other differences remained unchanged. For indirect aggression, the high multiple group increased slightly, from 2.76 to 2.94, causing the difference between the high self and high multiple groups to become significant at p < .05. All other differences remained unchanged. Remorselessness. Estimated marginal means of the direct high self and high peer groups increased slightly (from 2.54 to 2.69 for high self; from 2.20 to 2.35 for high peer). This caused the difference between the high peer and low aggression groups to become significant at p < .05; other significance patterns remained unchanged.

PAGE 86

86 For indirect aggression, a slight increase in the high peer group resulted in a significant difference between the high peer and high self gro ups at p < .05. All other differences remained unchanged. Anger. For direct aggression, e stimated marginal means of the high self and high peer groups increased slightly (from 2.88 to 3.10 for high self; from 2.60 to 2.72 for high peer). The difference between the high self and high peer groups became significant at p < .05; other patterns remained unchanged. Patterns for indirect aggression remained unchanged. Anger regulation. The estimated marginal mean of the direct high self group decreased sligh tly (from 2.42 to 2.31) and was no longer significantly different than the high multiple group. Other patterns remained unchanged. Summary. U se of the more stringent criteria resulted in slight increases in antisocial behavior for the high self and high peer groups. However, the overall patterns of findings remained unchanged from the previous analyses in that the high self and high multiple groups had the highest levels of manipulative behavior, remorselessness, and anger, and the lowest levels of ange r regulation.

PAGE 87

87 LIST OF REFERENCES Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/adolescent behavioral and emotional problems: Implications of cross informant correlations for situational specificity. Psychological Bulletin, 101, 213232. Andershed, H., Kerr, M., Stattin, H., & Levander, S. (2002). Psychopathic traits in nonreferred youths: A new assessment tool. In E. Blauuw & L. Sheridan (Eds.), Psychopaths: Current international perspectives (pp. 131 158). Elsevier: The Hague. Arch er, J. (2004). Sex differences in aggression in real world settings: A meta analytic review. Review of General Psychology, 8 291322. Basow, S. A., Cahill, K. F., Phelan, J. E., Longshore, K., & McGillicuddy DeLisi, A. (2007). Perceptions of relational an d physical aggression among college students: Effects of gender of perpetrator, target, and perceiver. Psychology of Women Quarterly, 31, 8595. Bronfenbrenner, U. (1979). The ecology of human development Cambridge, MA: Harvard University Press. Brown, B. B. (2004). Adolescents' relationships with peers. In R. M. Lerner & L. Steinberg (Eds.), Handbook of adolescent psychology (2 ed., pp. 363394). New York: Wiley. Buss, A. H., & Perry, M. (1992). The aggression questionnaire. Journal of Personality and Soc ial Psychology, 63, 452459. Cantor, N., & Mischel, W. (1977). Traits as prototypes: Effects on recognition memory. Journal of Personality and Social Psychology, 35, 3848. Card, N. A., Hodges, E. V. E., Little, T. D., & Hawley, P. H. (2005). Gender effect s in peer nominations for aggression and social status. International Journal of Behavioral Development, 29, 146155. Card, N. A., Stucky, B. D., Sawalani, G. M., & Little, T. D. (2008). Direct and indirect aggression during childhood and adolescence: A me ta analytic review of gender differences, intercorrelations, and relations to maladjustment. Child Development, 79, 11851229. Center for Disease Control and Prevention. (2008). Health risk behaviors by sex: National YRBS, 2007. Retrieved April 11, 2009, from http://www.cdc.gov/HealthyYouth/yrbs/pdf/yrbs07_us_disparity_sex.pdf. Cillessen, A. H. N., & Borch, C. (2006). Developmental trajectories of adolescent popularity: A growth curve modelling analysis. Journal of Adolescence, 29, 935959. Clemans, K. H., DeRose, L., Graber, J. A., & Brooks Gunn, J. ( 2010). Gender in adolescence: A personin context approach to gender identity and gender roles. In J. Christler & D.

PAGE 88

88 McCreary (Eds.), Handbook of Gender Research in Psychology New York: Springer Publishing Company. Clemans, K. H., Graber, J. A., Nichols, T. R., Brooks Gunn, J., & Botvin, G. J. (2007, March). Predicting aggression from ambiguous scenarios: The contribution of attributions and emotion processes. Poster presented at the Society for Research in Ch ild Development Biennial Convention, Boston, MA. Clemans, K. H., & Sontag, L. M (2009, April). Characteristics of early adolescent aggressors: Peer vs. individual approaches to measuring aggression. Paper presented at the biennial meeting of the Society f or Research in Child Development, Denver, CO. Coie, J. D., Dodge, K. A., & Coppotelli, H. (1982). Dimensions and types of social status: A cross age perspective. Developmental Psychology, 18, 557570. Coie, J. D., Dodge, K. A., & Kupersmidt, J. B. (1990). Peer group behavior and social status. In S. R. Asher & J. D. Coie (Eds.), Peer rejection in childhood (pp. 1759). New York, NY: Cambridge University Press. Coyne, S.M., Archer, J., Eslea, M. & Liechty, T. (2008). Adolescent perceptions of indirect form s of relational aggression: Sex of perpetrator effects. Aggressive Behavior, 34, 577583. Cornell, D. G., Peterson, C. S., & Richards, H. (1999). Anger as a predictor of aggression among incarcerated adolescents Journal of Consulting and Clinical Psychology, 67, 108115. Crick, N. R. (1997). Engagement in gender normative versus nonnormative forms of aggression: Links to social psychological adjustment. Developmental Psychology, 33, 610617. Crick, N. R., & Ladd, G. W. (1990). Children's perceptions of the outcomes of social strategies: Do the ends justify being mean? Developmental Psychology, 26, 612620. Crowne, D. P., & Marlowe, D. (1960). A new scale of social desirability independent of psychopathology. Journal of consulting Psychology, 24, 349354. Cumming, G. & Finch, S. (2005). Confidence intervals and how to read pictures of data. American Psychologist, 60, 170180. Dearing, K. F., Hubbard, J. A., Ramsden, S. R., Parker, E. H., Relyea, N., Smithmyer, C. M., et al. (2002). Children's self reports a bout anger regulation: Direct and indirect links to social preference and aggression. Merrill Palmer Quarterly, 48, 308336. Dodge, K. A., Coie, J. D., & Lynam, D. (2006). Aggression and antisocial behavior in youth. In N. Eisenberg, W. Damon & R. M. Lerne r (Eds.), Handbook of child psychology (6 ed., Vol. 3, pp. 719788). New York: John Wiley & Sons, Inc. Eisenberg, N. (2000). Emotion, regulation, and moral development. Annual Review of Psychology, 51, 665697.

PAGE 89

89 Eisenberg, N., Guthrie, I. K., Cumberland, A., Murphy, B. C., Shepard, S. A., Zhou, Q., et al. (2002). Prosocial development in early adulthood: A longitudinal study. Journal of Personality and Social Psychology, 82, 9931006. Eisenberg, N., Miller, P. A., Shell, R., McNalley, S., & Shea, C. (1991). Prosocial development in adolescence: A longitudinal study. Developmental Psychology, 27, 849857. Eisenberg, N., Morris, A. S., & Spinrad, T. L. (2004). Emotionrelated regulation: The construct and its measurement In D. Teti (Ed.), Handbook of Research M ethods in Developmental Psychology (pp. 423 442). Oxford: Blackwell Eisenberg, N., Spinrad, T. L., & Sadovsky, A. (2006). Empathy related responding in children. In M. Killen & J. Smetana (Eds.), Handbook of moral development (pp. 517549). Mahwah, NJ: La wrence Erlbaum Associates. Epkins, C. C., & Myers, A. W. (1994). Assessment of childhood depression, anxiety, and aggression: Convergent and discriminant validity of self parent teacher and peer report measures. Journal of Personality Assessment, 62, 364381. Epstein, J. A., Botvin, G. J., Diaz, T., Baker, E., & Botvin, E. M. (1997). Reliability of social and personal competence measures for adolescents. Psychological Report, 81, 449450. Fiske, S. T., & Taylor, S. E. (1991). Social categories and schemas. In Social Cognition (2nd ed., pp. 96141). New York: McGraw Hill, Inc. Galambos, N. L., Almeida, D. M., & Petersen, A. C. (1990). Masculinity, femininity, and sex role attitudes in early adolescence: Exploring gender intensification. Child Developmen t, 61, 19051914. Giles, J. W., & Heyman, G. D. (2005). Young children's beliefs about the relaionship between gender and aggressive behavior. Child Development, 76, 107121. Graham, S., Bellmore, A., & Juvonen, J. (2003). Peer victimization in middle school: When self and peer views diverge. Journal of Applied School Psychology, 19, 117137. Hamilton, D. L., Stroessner, S. J., & Driscoll, D. M. (1994). Social cognition and the study of stereotyping. In P. G. Devine, D. L. Hamilton & T. M. Ostrom (Eds.), Social cognition: Impact on social psychology (pp. 292323). San Diego, CA: Academic Press, Inc. Henry, D. B., & The Metropolitan Area Child Study Research Group (2006). Association between peer nominations, teacher ratings, self reports, and observations o f malicious and disruptive behavior. Assessment,13, 241252. Hill, J. P., & Lynch, M. E. (1983). The intensification of gender related role expectations during early adolescence. In J. Brooks Gunn & A. C. Petersen (Eds.), Girls at puberty: Biological and psychosocial perspectives (pp. 201 228). New York: Plenum.

PAGE 90

90 Hinduja, S., & Patchin, J. W. (2008). Cyberbullying: An exploratory analysis of factors related to offending and victimization. Deviant Behavior, 29 129156. Hyde, J. S. (1984). How large are gende r differences in aggression? A developmental metaanalysis. Developmental Psychology, 20, 722736. Jolliffe, D., & Farrington, D. P. (2006). Development and validation of the Basic Empathy Scale. Journal of Adolescence, 29, 589611. Katz, D., & Braly, K. ( 1933). Racial stereotypes of one hundred college students. Journal of Abnormal and Social Psychology, 28, 280290. Larson, R., & Richards, M. H. (1991). Daily companionship in late childhood and early adolescence: Changing developmental contexts. Child Dev elopment, 62, 284300. Loy, P., & Norland, S. (1981). Gender convergence and delinquency. The Sociological Quarterly, 22, 275283. Lynam, D. R. (1996). Early identification of chronic offenders: Who is the fledgling psychopath? Psychological Bulletin, 120, 209234. Maccoby, E. E. (1998). The two sexes: Growing up apart, coming together. Cambridge, MA: Belknap Press. Marshall, S. K., TiltonWeaver, L. C., & Bosdet, L. (2005). Information management: Considering adolescents' regulation of parental knowledge. Journal of Adolescence, 28, 633647. Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. (2001). Sex differences in antisocial behavior: Conduct disorder, delinquency, and violence in the Dunedin Longitudinal Study Cambridge: Cambridge University Press. Mu ris, P., van der Pennen, E., Sigmond, R., & Mayer, B. (2008). Symptoms of anxiety, depression, and aggression in nonclinical children: Relationships with self report and performance based measures of attention and effortful control. Child Psychiatry and H uman Development, 39, 455467. Nederhof, A. J. (1985). Methods of coping with social desirability bias: A review. European Journal of Social Psychology, 15, 263280. Newcomb, A. F., Bukowski, W. M., & Pattee, L. (1993). Children's peer relations: a meta an alytic review of popular, rejected, neglected, controversial, and average sociometric status Psychological Bulletin, 113, 99128. Nichols, T. R., Graber, J. A., Brooks Gunn, J., & Botvin, G. J. (2006). Sex differences in overt aggression and delinquency am ong urban minority middle school students. Journal of Applied Developmental Psychology, 27, 7891.

PAGE 91

91 Novaco, R. B. (1994). Anger as a risk factor for violence among the mentally disordered. In J. Monahan & H. H. Steadman (Eds.), Violence and mental disorder: Developments in risk assessment (pp. 2159). Chicago: University of Chicago Press. Orpinas, P., & Frankowski, R. (2001). The aggression scale: A self report measure of aggressive behavior for young adolescents. Journal of Early Adolescence, 21, 5067. Ott en, S., & Stapel, D. A. (2007). Who is this Donald? How social categorization affects aggressionpriming effects. European Journal of Social Psychology, 37, 10001015. Pakaslahti, L., & Keltikangas Jarvinen, L. (1997). The relationships between moral approval of aggression, aggressive problem solving strategies, and aggressive behavior in 14year old adolescents Journal of Social Behavior and Personality 12, 905924. Peets, K., & Kikas, E. (2006). Aggressive strategies and victimization during adolescence: Grade and gender differences, and cross informant agreement. Aggressive Behavior, 32, 6879. Pellegrini, A. D., & Bartini, M. (2000). An empirical comparison of methods of sampling aggression and victimization in school settings. Journal of Educational Psy chology, 92, 360366. Prinstein, M. J., Boergers, J., & Vernberg, E. M. (2001). Overt and relational aggression in adolescents: Social psychological adjustment of aggressors and victims. Journal of Clinical Child Psychology, 30, 479491. Putallaz, M., Grim es, C. L., Foster, K. J., Kupersmidt, J. B., Coie, J. D., & Dearing, K. (2007). Overt and relational aggression and victimization: Multiple perspectives within the school setting. Journal of School Psychology, 45, 523547. Reynolds, W. M. (1982). Developme nt of reliable and valid short forms of the Marlowe Crowne Scale of Social Desirability. Journal of Clinical Psychology, 38, 119125. Ryan, C. S., Judd, C. M., & Park, B. (1996). Effects of racial stereotypes on judgments of individuals: The moderating rol e of perceived group variability. Journal of Experimental Social Psychology, 32, 71103. Sagar, H. A., & Schofield, J. A. (1980). Racial and behavioral cues in Black and White children's perceptions of ambiguously aggressive acts. Journal of Personality and Social Psychology, 39, 590598. Saltaris, C. (2002). Psychopathy in juvenile offenders: Can temperament and attachment be considered as robust developmental precursors? Clinical Psychology Review, 22, 729752. Spielberger, C. D., Jacobs, G. A., Russell, S. F., & Crane, R. J. (1983). Assessment of anger: The State Trait Anger Scale. In J. Butcher & C. D. Spielberger (Eds.), Advances in personality assessment (2nd ed., pp. 159187). Hillsdale, NJ: Erlbaum.

PAGE 92

92 Strahan, R., & Gerbasi, K. C. (1972). Short, homoge neous versions of the Marlow Crowne Social Desirability Scale. Journal of Clinical Psychology, 28, 191193. Tigges, B. B. (2003). Parental consent and adolescent risk behavior research. Journal of Nursing Scholarship, 35, 283289. Whitley, B. E., & Kite, M E. (2006). The Psychology of Prejudice and Discrimination. Belmont, CA: ThomsonWadsworth. Xie, H., Cairns, R. B., & Cairns, B. D. (2002). The development of social aggression and physical aggression: A narrative analysis of interpersonal conflicts. Aggr essive Behavior, 28, 341355. Younger, A. J., Schneider, B. H., & Daniels, T. (1991). Aggression and social withdrawal as viewed by children's peers: Conceptual issues in assessment and implications for prevention. Journal of Psychiatry and Neuroscience, 16, 139145. Zeman, J., Shipman, K., & Suveg, C. (2002). Anger and sadness regulation: Predictions to internalizing and externalizing symptoms in children. Journal of Clinical Child and Adolescent Psychology, 31, 393398.

PAGE 93

93 BIOGRAPHICAL SKETCH Katherine C lemans received a B.A. in psychology from Duke University in 2002 and an M.S. in psychology from the University of Florida in 2007. She is a recipient of the University of Floridas J. Hillis Miller Presidential Fellowship. Her research interests include the development of aggression and moral judgment during adolescence.