1 CHILDRENS TENDENCY TO DEFEND VICTIMS OF SCHOOL BULLYING: GENDER SOCIAL IDENTITY, AND NORMATIVE PRESSURE By JAMES R. PORTER A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLM ENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009
2 2009 James R. Porter
3 To my wife, Karen W. Porter, who built a life and a family with me while endu ring the challen ges related to my efforts on this project
4 ACKNOWLEDGMENTS I want to thank Dr. Sondra Smith, my committee chair, for her instruction and support. Dr. Smith helped me hone my questions regarding violence and conflict resolution into a research specialty area that seeks out what is best about humanity. Her encouragement and direction have helped me shape my professional goals and my identity as a counselor, researcher, and citizen of the world. Her personal guidance ha s elevated my understanding about what it means to be a husband and a father. I would also like to thank D r. Harry Daniels for showing me a greater potential in myself. My faith in my ability to observe and conceptualize the diverse realities that people encounter was garnered from his w illingness to ask more of me in class and in this dissertation. Dr. Peter Sherrard never gave to me what he thought I could give myself. He taught me personal responsibility and the value of being present for others; and I am gratef ul for the spiritual s trength I gained from his presence Dr. David Miller gave me valuable technical support, and showed me that it was all right to enjoy statistical analysis, and that the conclusions we pronounce and the decisions we make as based on social science inquiry have consequences in peoples lives. I thank these professors for exemplifying concern for a reality that cannot be captured in a research report. I also wish to thank my family. My wife, Karen, struggled with me through my graduate studies, sometimes su pporting me financially, other times keeping close to me when it was difficult, always showing love in some way. My daughter, Christi, reminded me of how little I know and taught me that the way I behaved when I was away from my laptop was most important to the people I love. My parents may not realize that they picked me up through a lifetime of academic and social struggle, from my earliest memories. They did this in countless ways, sometimes by expecting more of me than I was giving, sometimes by sho wing pride in me, but usually by engaging with me as a valid, powerful human. My mother and father, and my
5 brothers, John and Joe, and my sister -in -law, Ronda, asked me about my dissertation when I wanted to be asked. My cousin, Susan Porter, was willing to talk to me about statistics, just for fun. In addition, I also thank Anne Powers Flenner, Odalis Manduley, Melissa Berryman, Mike Gamble, Dr. Russ Froman, Lisa Clemons, and the administration, faculty, staff, and students at the schools who participate d in this research study. I would also like to thank the students at Loga Springs Academy for their insightful feedback. I thank Dr. James Algina for assisting me with power and accuracy tables Tim Baker for sharing his knowledge about sampling schools, and Candy Spires and Patty Bruner for their help with technical and office matters I am grateful to Jaime Jasser, Keely Hope, Joe Munson, Teresa L ei bforth, Kelly Aissen and many other fellow doctoral students and graduates on whom I leaned during the di ssertation process. I thank the faculty of my undergraduate alma mater, the University of South Florida, for making my acceptance to UF a possibility, and my f riends at UFs Counselor Education Department and at the Gainesville Monthly Meeting of the Reli gious Society of Friends for their interest in my doctoral pursuits. Finally, I would like to acknowledge my school, the University of Florida, for its high academic standards and my country, t he United States of America, for instilling in me a passion fo r equality and social responsibility : the values upon which I am shaping my career.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................................... 4 LIST OF TABLES ................................................................................................................................ 9 LIST OF FIGURES ............................................................................................................................ 10 ABSTRACT ........................................................................................................................................ 11 CHAPTER 1 INTRODUCTION ....................................................................................................................... 13 Scope of the Study ...................................................................................................................... 15 Theoretical Framework ............................................................................................................... 18 Statement of the Problem ............................................................................................................ 19 Need for the Study ...................................................................................................................... 22 Purpose of the Study ................................................................................................................... 24 Rationale for the Methodology................................................................................................... 25 Research Questions ..................................................................................................................... 27 Definition of Terms ..................................................................................................................... 28 Overview of t he Remainder of the Study .................................................................................. 28 2 REVIEW OF RELATED LITERATURE ................................................................................. 29 Further delineation of the Problem ............................................................................................ 29 The State of Research on Defending .......................................................................................... 32 Individual Tendencies and the Defender Role ................................................................... 33 Social -Contextual Variables and the Defender Role ......................................................... 37 Gender and Defending......................................................................................................... 43 Social Identity Theory and Defending ....................................................................................... 49 Support for the Need for the Study ............................................................................................ 54 Summary of Major Points ........................................................................................................... 55 3 METHODOLOGY ...................................................................................................................... 57 Participants .................................................................................................................................. 57 Sampling Procedures ................................................................................................................... 59 Design .......................................................................................................................................... 60 Measures ...................................................................................................................................... 60 The Self Report Defender Scale ......................................................................................... 61 Gender Identity .................................................................................................................... 65 Normative Pressure to Help Victims .................................................................................. 67 Age ........................................................................................................................................ 70 Analyses ....................................................................................................................................... 71 Methodological Limitations ....................................................................................................... 72
7 4 RESULTS .................................................................................................................................... 74 Sample Demographics ................................................................................................................ 74 Reliability and Validity of Instruments ..................................................................................... 75 Defender Scale ..................................................................................................................... 76 Childrens Personal A ttributes Questionnaire Feminine -Masculine Scale ...................... 76 Normative Pressure to Help Victims .................................................................................. 79 Descriptive Statistics ................................................................................................................... 80 Relationships of Defending Scores to Controls and Predictors ............................................... 81 Defending Score Relationships to Control Variables ........................................................ 81 Defending Score Relationships to Predictors, Mediators, and Moderators ..................... 82 Mediation Analyses ............................................................................................................. 83 Moderation Analyses ........................................................................................................... 84 Combining Variables in Predicting Defending .................................................................. 86 Summary ...................................................................................................................................... 87 5 DISCUSSION .............................................................................................................................. 95 Overview of Study and Discussion of Findings ........................................................................ 95 Gender .................................................................................................................................. 97 Gender Identity .................................................................................................................... 98 Normative Pressure to Help Victims .................................................................................. 99 Implications for Theory ............................................................................................................ 101 Implications for Practice ........................................................................................................... 103 Limitations of the Study and Implications for Research ........................................................ 105 Limitations ......................................................................................................................... 105 Implications for Research .................................................................................................. 108 Summary .................................................................................................................................... 111 APPENDIX A PARENTAL CONSENT .......................................................................................................... 112 B STUDENT ASSENT ................................................................................................................ 114 C THE SELF REPORT DEFENDER SCALE MALE CONDITION ...................................... 115 D THE SELF REPORT DEFENDER SCALE FEMALE CONDITION ................................. 116 E NORMATIVE PRESSURE TO HELP VICTIMS ................................................................. 117 F CHILDRENS PERSONAL ATTRIBUTES QUESTIONNAIRE FEMININE MASCULINE SCALE .............................................................................................................. 119 G DEMOGRAPHICS ................................................................................................................... 120
8 LIST OF REF ERENCES ................................................................................................................. 121 BIOGRAPHICAL SKETCH ........................................................................................................... 127
9 LIST OF TABLES Table page 4 1 Descriptive statis tics for dependent and independent variables. ......................................... 88 4 2 Frequencies for best friends age, best friends gender, favorite teachers gender, and assignment of bullying gender scenario. ............................................................................... 89 4 3 Percents of students reporting parents, best friends, and favorite teachers as expecting them to help victims. ............................................................................................. 90 4 4 Pearson produc t -moment correlations between predictors and defending scores. ............. 90 4 5 Semi partial correlations to defending behavior for variables in best initial model by R-square selection, followed by sub sequent models. .......................................................... 91
10 LIST OF FIGURES Figure page 4 1 Mediation diagram showing zero -order correlations (and semi -partial correlations) with gender as the predictor, gender identity as the mediator, and defend ing as the dependent variable. ................................................................................................................ 92 4 2 Mediation diagrams showing zero -order correlations (and semi -partial correlations) for normative pressure ........................................................................................................... 92 4 3 Interaction s between participant gender and normative pressure. ...................................... 93 4 4 Interaction in predicting defending sco res between age and normative pressure from friends.. ................................................................................................................................... 94
11 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 CHILDRENS TENDENCY TO DEFEND VICTIMS OF SCHOOL BULLYING: GENDER SOCIAL IDENTITY, AND NORMATIVE PRESSURE By James R. Porter May 2009 Chair: Sondra Smith -Adcock Major: Mental Health Counseling Bullying is prevalent i n schools in the United States and across the world. Increasing attention has been paid to the phenomenon of bullyin g by researchers and practitioners, because of its negative consequences on children who bully and who are bullied, consequences which may sometimes be let hal. Once studied as a problem of individual bullies and victims, researchers are now recommending that bullying be studied as a group phenomenon. Many bystanders support school bullying, but some appear to work in support of its victims. Labeled defen ding by some authors, childrens tendency to help victim has been studied much less than their tendency to bully. A somewhat persistent finding in studies about defending is that girls seem more likely to defend victims than do boys. Social i dentity t h eory ( Tajfel & Turner, 1986) was used in the study to frame questions about whether gender -based social expectations may relate to childrens tendency to help victims. The current study explore d two potential mediators in the relationship between gender a nd defending behavior. Those mediators were social gender identity and normative pressure from significant others to help victims. Regression analyses suggested that gender identity predict ed defending behavior as well as biological sex and that normati ve pressure from friends may have mediate d the relationship between gender and defending.
12 Limitations of the study include d the lack of causal inference possible in a correlational study, questionable scale psychometrics, weaknesses inherent in self -repo rted observations, and the underrepresentation of males in the study. Implications for research concentrated on improving upon the studys limitations and enlarging the body of research that focuses on defending behavior. Implications for counseling invo lved combining the efforts of pro defending parents, teachers, and students into programs that restructure social groups, reframe gender expectations, and support effective defending behavior in children.
13 CHAPTER 1 INTRODUCTION School bullying, a pattern of aggression between students, is prevalent in the United States. According to a recent study by the World Health Organization, 29.9% of students in the U.S. have bullied other children, been bullied, or both (Nansel et al., 2001). Both students who bully and those who are victimized are at risk for negative mental, behavioral, and physical consequences (Kim, Leventhal, Koh, Hubbard, & Boyce, 2006; Laflamme, Engstrm, Mller, Alldahl, & Hallqvist, 2002), with pathology frequently lasting into adulthood (e.g., Dempsey & Storch, 2008). Furthermore, bullying is associated with lethality. A study of all school related violent deaths in the United States between 1994 and 1999 found that victims of bullying comprised 20% of school related homicide perpetrato rs and 12% of homicide casualties (Anderson et al., 2001). Another study conducted jointly by the U.S. Secret Service and the Department of Education, also concluded that a majority of school shootings in the United States were committed by victims of bul lying (Vossekuil, Fein, Reddy, Borum, & Modzeleski, 2002). Once conceived of as an individual behavioral phenomenon involving only bullies and victims, bullying is now viewed more as a process in which social context factors influence the behavior (Furlo ng, Morrison, & Greif, 2003). When bullying situations occur, children adopt a number of social roles, including roles in which peers ignore, allow, encourage, or physically assist acts of bullying (Salmivalli, Lagerspetz, Bjrkqvist, sterman, Kaukiainen, 1996). However, not everyone promotes or allows bullying. Increased interest in studying the socioecological contexts of bullying has led to the identification of children who tend to intervene to help victims of bullying (Salmivalli et al., 1996). C hildren who act in some way to prevent or intervene in bullying have been referred to in the literature as defenders. Defending represents the only identified prosocial bystander role behavior in bullying.
14 Defenders, because they tend to intervene in bullying, may potentially help improve school safety. Therefore, it seems desirable for caregivers, professionals, and institutions to work toward supporting these children. There is a paucity of evidence that this effort to support defending is happening. Furthermore, there is little or no research specifically focusing on children who defend the victims of bullying. Compared to the rapidly growing body of research dedicated to children who bully or are bullied, a small but increasing number of studies e xplore the phenomenon of peers intervening in bullying. Understanding children who defend against bullying may reveal a new avenue toward preventing school -related violence. The relationship between peer context and defending behavior has been studied s omewhat marginally. In addition, much less has been done to explore a broader socio-ecological context, in which influential adults help establish norms related to defending. Because teachers and parents could conceivably influence bullying and defending behavior, it seems important to explore their possible normative social influence in regard to defending. Support for defending behavior from mothers, fathers, and teachers may be helpful in establishing consistent patterns of defending in students over time. Researchers might be able to facilitate this process of adult support for defending by bringing more institutional attention to defending behavior. Increased institutional attention to defending could create awareness, in parents and teachers, tha t children who adopt a defending role may be encouraged by adult support. Thus, it is important to begin the process of understanding how adults in the social environment are involved in facilitating social norms related to defending. Evidence suggests th at gender norms create a higher expectation for girls to defend than for boys. Defending and bullying tend to vary by gender. Boys appear more likely to bully and
15 girls appear more likely to defend (e.g., Salmivalli, Kaukiainen, & Voeten, 2005). When boys adopt a defending role, they also are less likely than girls to maintain that defending role in the future, or among different friends (Salmivalli, Lappalainen, & Lagerspetz, 1998). A more thorough examination of the apparent defending gap between girl s and boys may aid caregivers, practitioners, and educators in reinforcing defending in both girls and boys. Because a growing body of research considers bullying to be a group process, it is helpful to investigate what social forces are involved in gender differences in childrens defending behavior. A more intricate picture of these seeming gender effects will help determine what concerned adults can do to improve behavior related safety conditions for children at school. The present study focus e d on de fending behavior in children, specifically examining how social normative influences on defending we re associated with gender. Scope of the Study Though bullying is traditionally studied as a problem among individual bullies and victims, a growing number o f researchers assert that bullying only happens as a group process (Sutton & Smith, 1999) and that it is most appropriate to study bullying as emerging from conducive socioecological contexts (Furlong, Morrison, & Greif, 2003). Some research on these soc io ecological contexts has led to the identification of several roles that children play in bullying scenarios. Among the identified bullying related roles, only one is defined by a pattern of benevolent behaviors intended to deter bullying. This benevol ent role that prosocial children often play in bullying scenarios is sometimes called defender. The rest of the bullying -related roles are roles in which children ignore, allow, or even facilitate bullying (Olweus, 2001). Six roles have been empiricall y verified: defender, bully, victim, outsider (i.e., uninvolved children), assistant
16 (i.e., children who physically help the bully) and reinforcer (i.e., children who cheer the bully) (Salmivalli, Lagerspetz, Bjrkqvist, sterman, Kaukiainen, 1996). The notion that children create a context conducive to bullying that the defender role is the only prosocial bullying related role yet identified may be counterintuitive to many. However, Ojala and Nesdale (2004) conclude that societal groups often seem to encourage bullying as a tool for punishing and excluding persons who act outside the expectations for their own groups, or who try to join groups that do not want them. Such processes, say Ojala and Nesdale (2004), conform to components of social identity t heory. Defenders are often identified using a scale developed by Salmivalli and colleagues (1996). According to literature reviewed in Chapter 2 of this dissertation use of this scale shows that many children appear to assume the defending role (e.g., Sa lmivalli et al., 1996). However, it seems that many of those children identified as defenders trade off their defending role to other children from one year to the next, such that less than half of children identifiable as defenders one year will be ide ntifiable as defenders the following year (Menesini, Codecasa, Benelli, & Cowie, 2003). This type of role instability is not reported for the other bullyingrelated roles. Furthermore, there appears to be correlational evidence suggesting that children in the defender role may require more peer support to acquire or maintain their defender role than do children in other bullying related roles. Being identified in other bullying related roles predicts future behavior in keeping with those roles (Salmival li et al., 1998). This finding does not hold for the defending role, but being currently identified as a defender strongly correlates with having defender friends (Salmivalli et al., 1998). This possible reliance of defending behavior on peer influence s eems unfortunate considering no role has been identified in which peers
17 support children who defend, though children who bully have followers who consistently assist and reinforce them. The absence of identified peer roles to support children who defend m ay be partly attributable to the fact that there is little research done on children known to defend. Nevertheless, enough research exists to suggest at least one other interesting pattern related to childrens defending behavior. The most common finding in studies examining defending behavior is that girls appear to be more likely to defend than are boys (Camodeca & Goossens, 2005; Goossens, Olthof, & Dekker, 2006; Menesini et al., 2003; Salmivalli et al., 1996; Salmivalli et al., 1998; Salmivalli Kaukiai nen, Kaistaniemi, Lagerspetz, 1999; Salmivalli, Kaukiainen, & Voeten, 2005; Salmivalli & Voeten, 2004; Sutton & Smith, 1999). That is, nine out of the thirteen studies that directly examine defending behavior find boys to be markedly underrepresented amon g children who defend, although results are mixed in one study (Sutton & Smith, 1999). In addition, boys who defend appear likely to change their defending role more so than girls who defend (Salmivalli, Lappalainen, & Lagerspetz, 1998). The lack of boys in the defender role is troubling because boys appear most likely to be bullies and victims, and may benefit from having defenders within their male peer groups. Research findings to date do not investigate this gender discrepancy in defending behavior further than reporting its occurrence. This gender discrepancy may come from an innate female tendency to be helpful toward others; but it also could be that social and gender norms encourage girls to defend, and discourage boys from defending. Social id entity theory has the potential to explain how social norms may influence the relationships among bullying, defending, and gender.
18 Theoretical Framework Social identity theory (SIT; Tajfel & Turner, 1986) is thought to explain group processes that lead to bullying (Ojala & Nesdale, 2004). Because defending behavior emerges within the same group settings as bullying (e.g., Salmivalli et al., 1996), social identity theory, as a parsimonious explanation of a broad array of group processes, ought to sufficient ly frame research questions about defending. According to SIT, ones self -esteem relies on membership within an in -group that collectively acts in ways that make the in group favorably distinct from a specified out -group (Tajfel & Turner, 1986). In addit ion, group membership guides individual members behavior through various individual and group mechanisms of conformity (Abrams & Hogg, 1990). SIT may be useful in explaining gender differences in defending behavior if genders are considered social group s in addition to being biological groups. Wilson and Liu (2003) and Schmitt, Branscombe, and Kappen (2003) consider gender identity to be a group identity as described by SIT, an interpretation which is endorsed by one of the creators of SIT (Turner & Rey nolds, 2003). Based on assumptions of SIT, it might be predicted that women and men will behave in bullying situations in various ways, depending on social gender expectations and depending on how much they identify with being a female or male. If female s, as a group, are stereotyped as performing a particular behavior (i.e., defending victims), individual females will perform that behavior according to how much they identify with being female and with being distinct from males. Similarly, if a particula r behavior is proscribed for males, individual males will avoid that behavior only insofar as they glean their self -esteem from their maleness as it differs from femaleness. If ones gender is not considered an important in -group and the opposite gender i s not considered the important referential out group, then
19 ones behavior should not follow social gender expectations as stringently as when gender is considered important. Furthermore, group differences in a particular behavior should exist only as far as it is considered important to a groups favorable distinctiveness. If the behavior in question (i.e., defending) is not highly prescribed for one gender and proscribed for the other, then gender differences in that behavior should be small, compared t o how they would be if the behaviors were highly and differentially mandated for each gender. Throughout this dissertation ones identification with being male or female has been termed gender identity Operationally, g ender identity was determined using a measure of gender typing (Hall & Halberstadt, 1980). Social expectations of defending behavior according to gender w ere operationalized using a measure of normative pressure (i.e., from parents, friends, and teachers) to help victims (Rigby & Johnson, 2006). If either gender identity or normative pressure strongly qualif ied the relationship between gender and defending, results were to be interpreted as supporting a social identity theory explanation of defending behavior. Finding support for social identity theory was seen as lend ing credibility to interventions that attempt to alter childrens social environments to encourage group and self -identities that are commensurate with safe and fair play. Failure to find support was construed as pointing to alternative directions in research and intervention in bullying. Statement of the Problem In order to understand the importance of defending, it is important to understand the prevalence and aftermath of bullying. Bullying is verbal or physical, direct or indirect, group or individual aggression that is repeated, and in which there is a power imbalance, consistent in direction, between the victim and the aggressor (Gini & Pozzoli, 2006). It is not a conflict of equally matched opponents. Estimates var y regarding the prevalence of bullying in schools in
20 the United States. For example, Nansel et al. (2001) report that 29.9% of sixth through tenth grade children say they have been involved in bullying either as a perpetrator, victim, or both. Harris In teractive and GLSEN (2005) find that 65% of teens have been harassed or assaulted over the course of one year. Twenty-five percent of children in sixth through tenth grade report bullying others once or twice during the spring term of 1998 and 24.2% rep ort being bullied once or twice during the spring term of 1998 (Nansel et al., 2001). A full 8.8% of middle and early high school students report bullying others every week, and 8.4% report being bullied every week. Furthermore, bullying appears to be mu ch more prevalent in middle school than in high school. About 10% of middle school students report bullying others weekly, and 7.6 to 13.3% report being bullied weekly (Nansel et al., 2001). Teachers are just as, or more, likely than students to say that bullying is a serious problem in their schools (Harris Interactive & GLSEN, 2005). The pattern of aggression that constitutes bullying has particularly negative consequences when compared to less -patterned, incidental aggression (Furlong, Morrison, & Grei f, 2003). Victims often suffer deteriorated school performance and sometimes drop out; and they also complain of somatic symptoms (Gillies Rezo & Bosacki, 2003). Children who bully experience an increased risk of committing criminal acts and bullying oth ers in adulthood (Gillies Rezo & Bosacki, 2003). Some evidence points to bullying being a cause, and not just a consequence, of socially immature behavior in children who are victimized; of increased aggression (i.e., defiance, cruelty, fighting, tendency to argue) in children who bully; and of aggression, conduct problems, lack of behavioral control, and psychopathological externalizing of behaviors in children who are considered both bullies and victims (Kim, Leventhal, Koh, Hubbard, & Boyce, 2006). Eve n relational bullying, in which children are teased, ostracized, and socially sabotaged,
21 has been related to symptoms of depression and social anxiety in college students aged 1825 who were bullied as adolescents (Dempsey & Storch, 2008). Furthermore, bul lying appears physically hazardous, even lethal. Ten percent of all physical injuries in 10to 15 -year -old children are evidently the result of bullying (Laflamme, Engstrm, Mller, Alldahl, & Hallqvist, 2002). In addition, bullying seems to be related to suicide in some victims (Carney 2000; Rigby & Slee, 1999). Homicide is also a prospect for victims. At least two studies find that the majority of school shooters have been victims of bullying (Leary, Kowalski, Smith, & Philips, 2003; Vossekuil et al. 2002). More generally, victims of bullying appear to comprise 12% of victims of various forms of school related homicide, and they comprise 20% of perpetrators of school -related homicide (Anderson et al., 2001). It is possible that adults are helping cr eate environments that are conducive to bullying. For example, 27.8 to 34.6% of children say their parents expect them to support or avoid interfering with bullying (Rigby, 2005). Furthermore, s ome research shows that adults may not be setting a helpful example. Approximately 33% of a nationwide sample of U.S. teachers admit to bullying students a few times or frequently (Twemlow, Fonagy, Sacco, & Brethour Jr., 2006). Of particular theoretical relevance to the present study is the evidence that stud ents acting outside of expectations for their respective genders have reason to fear reprisal from fellow students. Among the many reasons for being harassed or assaulted, as indicated by children in a national U.S. study (Harris Interactive & GLSEN, 2005), were gender expression (26 to 37% of respondents) and sexual orientation (32 to 43%), with more boys than girls reporting reprisal for their gender expression. Twenty -four percent of teens in the same survey say they have
22 heard someone criticize a fellow student for acting too much like the opposite sex; and 33% say that students are harassed about their sexual orientation. While 90% of teens perceived to be lesbian, gay, bisexual, or transgendered are harassed or assaulted, the number is 62% for stu dents not so perceived (Harris Interactive & GLSEN, 2005). While the present study does not address issues related to LGBT status, the implications of the Harris Interactive and GLSEN (2005) study suggest that gender expectations may sometimes be promoted with harassment or physical force. If aggression enforces gender typed behavior, it is easy to imagine that boys might feel safer bullying while girls might feel safer defending. Such force imposed behavior norms may stifle childrens attempts to make p ositive changes when those changes do not conform to gender expectations. In general, bullying appears to be widespread and to cause considerable harm, and linked to the enforcement of gender expectations. In cyclical fashion, these gender expectations m ay proscribe defending in boys while allowing it in girls, leaving boys with the greater likelihood of bullying those who exhibit non-normative behavior. Need for the Study This review contains no claim that the phenomenon of bullying has been under resear ched. This dissertation does, however, contain such a claim regarding research about defending against bullying. Bullying research overwhelms research related to defending. On March 11, 2008, a search of the PsycINFO database using the keyword combinati on bully* OR bulli* (no quotes in search) resulted in 2,518 listed articles, 1,795 coming from peer reviewed journals. Of the 2,518 listings, only 40 included the keyword defend* and those listings did not all discuss persons who defend the victims of bullying. This author found that about thirteen studies directly assessed defending with a validated measure. Even these few studies examining defending discussed children who defend tangentially, focusing interest largely on children who bully or are bullied.
23 Furthermore, there appears to be no indication that researchers, teachers, counselors, parents, or policymakers are doing anything to specifically provide children who defend with targeted support for their behavior. This seems unfortunate conside ring the potential improvements in school safety that might come from well -guided defending behavior. Considering that defenders do not often remain defenders and that they may lack peer support for their behavior, these children may need as much adult attention and support as do children who bully or who are bullied. However, research on the social contexts of children who defend is currently limited to their friendship groups, their popularity, and their classroom behavioral norms. These classroom norms are not gender -specific and do not address norm s outside of the classroom. Gender norms are presumably created and communicated not only by peers and classroom environments, but also by caregivers, cultures, communities, and nations. Gender specific norms have been investigated in regard to delinquent behavior (Ford, Stevenson, Wienir, & Wait, 2002), but they have not yet been applied to the study of defending behavior. Furthermore, d espite the oftenreplicated finding that girls are more likely to d efend than boys, the reasons behind this apparent gender effect have not been studied. Considering the widespread harm attributable to bullying, it seems advisable try new directions in bullying research by studying the relationship between gender and def ending behavior as a focal topic instead of as a marginal one. Children who defend victims of bullying have the potential to impact the outcome of bullying in schools. The actions of these children could help decrease bullying -related violence and its con comitant physical and emotional injuries. Currently, research assessing school bullying reduction programs does not seem to indicate that the programs make any use of children who are thought to take specific initiative in deterring bullying. Program dev elopers
24 who become aware that defenders exist will likely incorporate this knowledge into their programs. In addition, defender aware counselor educators will be able to provide counselors in training with a more complex picture of the response some child ren have to bullying. Therefore, broader knowledge about children who defend and the contexts that influence them seems desirable for counselor educators, counselors -in -training, and practitioners. So far, the relationship between gender and defending has been treated as a demographic detail. A more direct look at the relationship between gender and defending may serve to inform caregivers, educators, and policymakers about the potential for changing social environments to promote prosocial behavior in bo th sexes. Investigating the potentially gender -based disincentives or incentives children perceive in helping the victims of bullying may lead to information that is useful to mental health practitioners as well. Purpose of the Study The present study h a d a long term goal of stimulating research interest about children who defend. Another long -term goal wa s to increase attention and support from teachers, counselors, parents, and other adults for children who defend to the level of support provided for children who bully or are bullied. Over time, increased support and attention may improve the consistency of defending behavior in children. Support for defending may also increase the chance of identifying and acquiring for all children the potential sc hool safety benefits that could result from well -guided defending behavior. For example, awareness on the part of researchers and practitioners regarding the role of defending may result in increased incorporation of children who defend into programs inte nded to ameliorate the bullying problem in schools. The present study wa s intended to begin the process that could lead to the attainment of these long term goals.
25 More immediately, the goal of the present study wa s to examine the reported gender patterns in defending behavior. The study investigate d the level to which the reported gender discrepancy defending behavior wa s attributable to socially derived differences between boys and girls and to social expectations for children to engage in defending beh aviors. Much of the research describing the social context of defending describes only peer related influences. The present study examine d gender norms, which presumably arise from other environments than just those formed by peers, and so the study also inquire d into adult expectations for children to engage in, or refrain from, defending behavior. Rationale for the Methodology This study investigate d whether a relationship exist ed between socialized gender identity and defending behavior, and whether th is relationship influence d the apparent relationship between biological gender and defending behavior. The study employ ed quantitative data gathering and analysis, and used survey instruments to measure the variables of interest, which we re defending be havior, gender, gender identity, normative pressure to defend, and age. All variables we re continuous in order to examine relationships and interactions among variables. A persistent gender difference in defending behavior has been found using various ver sions of the defender scale (e.g., Salmivalli et al., 1996). However, well -described method of direct observation (Pepler & Craig, 1995) of defending behavior did not substantiate this gender gap in defending. Limitations of this direct observational met hod are discussed in Chapter 2. Except for that direct observational method, this author found no other validated method of measuring defending behavior beside the defender scale. Furthermore, because of a potential confound between childrens popularity and their peers tendency to describe them as defenders, also discussed in Chapter 2, peer reports of defending behavior we re excluded from the study. Therefore, self reports on the defender scale we re used to estimate defending behavior.
26 Defending beha vior appears strongly tied to gender, as stated earlier. It was thought meaningful to determine whether there we re connections between ones gender and ones gender identity in predicting defending behavior. Therefore, in addition to asking children to r eport their gender, the present study use d a scale (Hall & Halberstadt, 1980) to measure gender identity in subjects. As explained earlier, social identity theory asserts that gender identity and associated behaviors depend partly on the expectations of others. In order to allow the researcher to examine expectations related to childrens defending behaviors, participants were asked about normative social pressure to defend. Normative pressure to help victims wa s a term used by Rigby and Johnson (2006) to describe the expectations of specific people (i.e., father, mother, teacher, and friends) for a child to defend victims of bullying. Normative pressure to help victims was measured by Rigby (2005) and by Rigby and Johnson (2006) separately for each p ressure source (i.e., mother, father, friends, teacher). Their method of measurement wa s partially replicated in the present study. The research questions presented below represent an exploration of the utility of SIT for explaining gender differences i n defending behavior. A relationship between gender and defending behavior wa s seen as negating support for SIT, unless other variables we re f ound to mediate t his relationship. Therefore, e xcept as regard ed participant gender, each relationship described in each research question required independent support in order for the researcher to claim full support for the use of SIT in explaining gender differences in defending behavior. As wa s applied per comparis on and not at a familywise level, except in the last research que stion which was an attempt to build a model for predicting defending
27 Research Questions The questions that follow represent an itemization of the more general question regarding whether the apparent gender gap in defending is related to the socialization of gender identity. The conceptual framework for the variables chosen for study c ame from social identity theory. Social identity theory suggests the potential for encouraging, in boys as much as in girls, th e assumedly prosocial behavior called defendi ng, and that such encouragement may help to elevate levels of defending in both sexes. If social identity theory wa s to be supported as explaining gender differences in defending behavior, adult and peer normative pressure to help victims, and ones own gender identity should have contribute d to the prediction of defender effects, and should have appear ed to m odify the gender differences in defending scores. The following questions guide d data gathering and analysis. Results of analyses are presented in Chapter 4, referring specifically to each research question by number. 1 Does gender, gender identity, or normative pressure to help victims (i.e., from mother, father, best friends, or favorite teacher) predict defending behavior, when controlling for the three anticipated control variables ? 2 Does gender identity mediate the relationship between gender and defending behavior, when controlling for the three anticipated control variables? 3 Does normative pressure to help victims (i.e., from mother, father, be st friends, or favorite teacher s ) mediate the relationship between gender and defending behavior, when controlling for the three anticipated control variables ? 4 Does gender in-group pressure differ from gender out -group pressure in its relationship to defe nding behavior? That is, is the relationship between normative pressure to help victims and defending behavior different as a function of whether or not the participant and the person exerting pressure are the same gender? 5 Does the presence of adults in a friendship group moderate the relationship between normative pressure and defending? 6 When all independent variables are combined to predict defending behavior, is the predictive value of participant gender changed?
28 Definition of Terms D EFENDING. A ny behavior intended to stop an incident of bullying or to otherwise support a victim of bullying. In the present study a childs score on the six-item, self report Defender Scale (Salmivalli, 1998; Salmivalli, Kaukiainen, Kaistaniemi, & Lagerspetz, 1999; Salmivalli, Lappalainen, & Lagerspetz, 1998) represent ed his or her level of defending behavior. G ENDER I DENTITY. An unarticulated, global sense of ones maleness or femaleness that is acquired early in life and is considered to be relatively impermeabl e to change (Lorenzi Cioldi & Doise, 1990; p77). In the present study the level of feminine versus masculine gender identity wa s measured using a single unidimensional scale (Hall & Halberstadt, 1980) in which high scores denote d high masculinity and lo w scores denote d high femininity. N ORMATIVE P RESSURE TO H ELP V ICTIMS. T he extent to which a childs mother, father, best friends, and favorite teachers would expect him or her to support a victim of bullying versus support a bully. Operationalization of the term wa s accomplished via the adaptation of four single item scales (Rigby & Johnson, 2006) in which a child c ould indicate his or her perception of the expectations of these important others. Higher scores indicate d expectations to support the victi m and lower scores indicated expectations to support bullying. B ULLYING. V erbal or physical, direct or indirect, group or individual aggression that is repeated, and in which there is a power imbalance, consistent in direction, between the victim and the aggressor (Gini & Pozzoli, 2006). This excludes aggression between persons or groups that are equally matched, and in which altercations are anomalous or spontaneous. Bullying wa s not investigated in the present study but its definition wa s important t o understanding what defending is. Overview of the Remainder of the Study Chapter 1 of this dissertation served to introduce both the topic of defending and the objectives of the present study Chapter 2 provides a review of literature relevant to the pr esent study Chapter 3 explains in detail the methodology used in the study Chapter 4 describes results of planned analyses and Chapter 5 includes a discussion of conclusions and limitations of the study, as well as implications for research and practic e.
29 CHAPTER 2 REVIEW OF RELATED LI TERATURE This chapter describes the prevalence and outcomes of school bullying, and how developments in bullying research have led to the identification of several roles children assume in bullying scenarios. One role, t he defender role, involves acting to intervene in bullying. The other roles (i.e., bully, assistant, reinforcer, victim, outsider) describe students who actively take part in bullying, are victimized by it, or who to withdraw from it. The present study focused on defending behavior. Evidence from other studies suggests that girls may be more likely than boys to defend victims. Social identity theory was tested in this study for its ability to explain this gender difference in defending behavior. Further delineation of the Problem The United States Department of Education defines bullying as behavior that is 1) aggressive in intent, 2) repeated, 3) unprovoked, and 4) based on a power imbalance between victim and bully (Furlong, Morrison, & Greif, 2003). These four components are commonly included in definitions of bullying found in research literature (Greene, 2003), and the ideas of power imbalance and repetition are widely prevalent in bullying definitions (Smith & Brain, 2000). The repetition and power imbalance components make bullying a form of aggression that is distinct from one -time assaults, fights between equals, or random acting out behaviors. According to existing bullying literature, bullying appears to be normative for child groups around the world, though it is generally considered socially unacceptable (Smith & Brain, 2000). Research suggests that bullying occurs in most schools (e.g., Salmivalli & Voeten, 2004). Children as young as 3 to 5 years are involved in bullying (Crick, Casas, & Ku, 1999). According to prevalence literature from the United States and around the world, bullying appears to be the most common type of aggression during elementary and middle school (Greene, 2003).
30 A review by Marini, Spear, and Bombay (1999) confirm s findings that bullying confers negative consequences both upon children who bully and those who are bullied. Their review attests that children who bully are at risk for legal, mental health, and drug related problems. Meanwhile, children who are bulli ed experience physical symptoms of illness, life long isolation, drug abuse, suicide, and decreased school performance including drop -out apparently in relation to victimization (Marini et al., 1999). Awareness of the associated consequences of bullying has prompted researchers to examine the once ignored phenomenon. The scientific study of bullying began in Norway in the 1970s, with other Scandinavian countries contributing heavily to the topic early on (Greene, 2003). Smith and Brain (2000) identify the publication of the book Aggression in the Schools: Bullies and Whipping Boys (Olweus, 1978) as representing a milestone emerging from Scandinavia during that era. The study of bullying has since spread throughout North America, the Pacific region, the United Kingdom, Ireland, other European countries, and developing countries (Salmivalli & Voeten, 2004; Smith & Brain, 2000). Furlong, Morrison, and Grief (2003) have concluded, however, that research and political interest in bullying was slow to arise in the United States. The primary focus of bullying related research has changed over time from examining only bullies and their victims, without consideration of context, to regarding bullying as a group phenomenon with various roles played by most childre n. As far back as 1973, Dan Olweus studied group mechanisms related to bullying (Olweus, 1978). However, such social psychological study of bullying was not then the trend. In the 1980s and 1990s, there was a substantial research focus on children who bullied or who were bullied, with the aim of finding characteristics that distinguished them from normal children (Greene, 2003; Smith & Brain, 2000). Although peers are present in 88% of bullying situations (Hawkins, Pepler, & Craig,
31 2001), researchers had typically approached bullying as a problem of only bullies and victims (Greene, 2003; Smith & Brain, 2000). The narrow focus on perpetrators and targets of bullying was evident not only in research studies but also in intervention efforts (Furlong, Morrison, & Greif, 2003). Almost all counseling interventions in the past have focused on intervening with children who committed or were targeted by acts of bullying (Greene, 2003). Furlong and colleagues (2003) conclude that state governments sometimes e nact bullying legislation in reaction to school violence as depicted in the media, rather than referring to current bullying literature. The resultant laws may lean toward punishing bullies instead of counseling them or attending to the broader sociologic al conditions that perpetuate bullying (Furlong et al., 2003). Fortunately the trend of conceptualizing bullying as a problem of individual perpetrators and victims is changing. Researchers are now paying more attention to how bullying behavior is influ enced by peers, schools, families, and larger institutions (Furlong et al., 2003). A growing consensus of research seems to suggest that interventions on an individual level have had limited effect (Greene, 2003). Addressing bullying as a group activity is becoming more commonplace in the literature. Much of the recent research on bullying focuses on group processes (Rodkin & Hodges, 2003; Salmivalli, 1999). Currently, numerous examples of group-based interventions can be found in program evaluation lit erature (Merrell, Gueldner, Ross, & Isava, 2008; Smith, Schneider, Smith, & Ananiadou, 2004). An important result of the increased attention to group processes in bullying is the identification (Olweus, 2001) and empirical verification (Salmivalli et al., 1996) of specific roles that children play in the process of bullying. Unique among these roles is one in which children defend victims of bullying. Unfortunately, interest in children who defend has been slow to
32 grow. Even though a few research studies examine the defender role, they generally do not focus on defending but rather discuss it as part of the context of bullying. The limited research on children who defend is reflected in their apparent omission from school anti -bullying programs. This a uthor found no literature describing interventions that incorporate d children who we re known to defend, even though researchers have recommended using children who defend as key figures in anti -bullying interventions. Salmivalli (1999) recommend ed the use of children who defend as models, mentors, and counselors in school anti bullying programs. However, later interventions developed or evaluated by that researcher (Salmivalli, 2001; Salmivalli, Kaukiainen, & Voeten, 2005) did not employ children who defe nd. In one of those programs, Salmivalli (2001) used peer counselors, but selected them using criteria other than their defender behavior. It appears that it is difficult to incorporate the defender role into practice, perhaps because of its inconspicuou sness in research literature. The State of Research on Defending Though research studies rarely focus specifically on defending, findings related to defending are available. Some research literature identifies various qualities of individuals who defend and investigates contextual variables that may relate to defending. What follows is a report of the data available regarding individual characteristics and contextual influences of children who defend, and a detailed description of the relationship betwee n gender and defending. The information presented is exhaustive of the literature referring to the bullying related behavior termed defending and the children termed defenders, though the information is small in scope and volume compared to that avail able about children who bully or are bullied. Th e following review suggest s that children who adopt a defending role exist in measurable amounts, and that they appear to differ in terms of individual characteristics and
33 social environment from children w ho assume other bullyingrelated roles. It also reasons that attention to contextual variables influencing defending behavior should be extended beyond peer influences to adult influences. Furthermore, it conclude s that the internalization of peer and a dult derived gender norms wa s an obvious target for a study designed to overcome the shortfalls of current defender -related research. Individual Tendencies and the Defender Role Children who defend appear to differ from children who exhibit other bullying -related behaviors in regard to certain individual characteristics. These individual characteristics include high trait agreeableness, low aggression, competent moral awareness, views against bullying, strong empathic ability, heightened inhibitory con trol, enhanced comprehension of others motives, and proficient emotion regulation. Defenders, when personal attributes are examined, also display high self -esteem and a highly generalized positive self -concept. As might be predicted, defenders seem to be more empathic than children who commit or abet the act of bullying. Gini, Albiero, Benelli, and Alto (2007), studying 318 Italian children aged 12 14, found empathic responsiveness to correlate positively with defending and negatively with bullying. This finding held for boys but not for girls in the sample. Maeda (2003) found similar results for both genders. Studying 293 10to 13 -year -old children in the United States, she found that defenders were significantly more empathic than were children identified as bullies or as their followers. Perhaps not surprising is that defender scale scores in a Finnish sample of children 9 12 years old were positively correlated with a scale of anti bullying attitudes (Salmivalli & Voeten, 2004). In addition children who defend appear less aggressive than children who bully or who support bullying. Maeda (2003) measured children on two types of aggression, reactive and proactive. Reactive aggression is an emotion -driven retaliation against someone perceive d to
34 have done harm. Proactive aggression is aggression initiated to achieve status or material goods. In Maedas (2003) study, children known to defend scored lower on proactive aggression than those known to bully, assist bullying, or reinforce it. Ch ildren recognized as defenders in the study also scored lower on reactive aggression than children who bully. In addition, for boys only, children identified as defenders measured lower on reactive aggression than children who assist or reinforce bullying (Maeda, 2003). Maeda (2003) also found that children who defend are good at understanding the cognitive and emotional motives of other peoples behavior. On a scale to measure this cognitive and emotional perspective-taking, children categorized as def enders scored significantly higher than children who bully directly (i.e., children who directly assault or insult victims, as opposed to those who engage in social sabotage and veiled teasing). Ginis (2006) study of 204 students in Italy aged 8 to 11 p rovided partial support for Maedas (2003) findings on perspective -taking. To measure perspective taking ability, researchers presented various scenarios to participants and asked them to infer the cognitive, emotional, and moral motives of the behaviors of the characters in the scenarios. Correct inferences were tallied to create a perspective taking score. Children identified as defenders scored higher on cognitive perspective taking than did those identified as victims. Scores on the defender scale c orrelated positively with cognitive, emotional, and moral perspective -taking (Gini, 2006). Perhaps contrary to Maeda (2003) was the finding that the only other role score to correlate with perspective -taking was the bully scale score, which correlated pos itively with cognitive and emotional, though not moral, perspective taking (Gini, 2006). Even though there are inconsistencies between the findings of these studies, their combined results suggest a strong interpersonal awareness in children who defend, a nd perhaps in children who commit acts of
35 social bullyingan awareness that is less common in children playing other roles. It seems important that children who defend appear to be better than children who bully on moral perspective taking, despite their similarity on two other perspective taking dimensions (Gini, 2006). Students who defend appear morally different from children who bully in another way. The Gini (2006) study also found that children who defend seem to have more concern for morals than children who engage in or promote bullying. The investigator used a moral disengagement scale to measure childrens tendency to minimize and justify violence and aggression. Children identified as defenders were less likely than bullies, assistants, an d reinforcers to disengage morally; and defender scale scores correlated negatively with moral disengagement scale scores (Gini, 2006). However, the reported results of this study did not indicate that children in the defender role differed from children in other nonaggressive roles (i.e., victim, outsider) on the variable of moral disengagement. Self restraint, or inhibitory control, is another variable on which children who defend seem to differ from children who bully. Monks, Smith, and Swettenham (20 03) gave 104 4to 6 -year olds in London schools an exercise intended to measure their ability to overcome previously learned question answer responses in order to answer new, counterintuitive test items correctly and quickly (e.g., looking at a flash card picture of a sun and, as instructed, quickly saying night). Children categorized as defenders scored higher on this task than children identified as bullies. The authors describe the task as requiring inhibitory control (Monks et al., 2003). Perhap s in the same category as inhibitory control, another form of self -control seems evident in these children. According to a scale designed to measure emotion regulation skills, defenders
36 seem better able to regulate their emotions than children who bully, help the bully, or cheer the bully (Maeda, 2003). Children who defend appear to have healthier self -esteem than children in other roles. Salmivalli, Kaukiainen, Kaistaniemi, and Lagerspetz (1999) measured 316 14to 15 year -old Finnish schoolchildren on peer -evaluated self -esteem, self -evaluated self -esteem, and peer evaluated defensive egotism (i.e., protective inflation of ones sense of self superiority). Then the authors used those three self -esteem dimensions to create five self -esteem profiles: Defensive self -esteem, self -belittlers, genuine self -esteem, humble pride, and low self esteem. Defensive self -esteem describes children who scored high on defensive egotism and above average on the other two dimensions. Self -belittlers were children w ho scored above average on peer evaluated self -esteem, very low on self -evaluated self -esteem and not high on defensive egotism. Genuine self -esteem denotes children who scored high on peer and self -evaluated self -esteem, but not on defensive egotism. Humble pride signifies children who scored high in self evaluated self -esteem, very low on peer -evaluated self -esteem and not high on defensive egotism. Finally, low self -esteem indicates children who scored low in all three dimensions. Results show ed that, for boys, but not for girls, a high defending score was most positively associated with the category of genuine high self -esteem. This is notable because, for boys at least, so -called genuine self -esteem differentiates children who defend not onl y from children in aggressive roles, but also from children who fall into other role behavior groups including children who remain outside of bullying scenarios (Salmivalli et al., 1999). Another attribute by which children who defend seem to stand out fro m other non -bullying children is self -concept. Self -concept differs from self -esteem in that self -esteem is an overall personal self -judgment whereas self -concept describes ones views about selected personal
37 qualities. Salmivalli (1998) measured 316 Fi nnish schoolchildren, 14 and 15 years old, on six self -concept areas: Academic, behavioral, emotional, social, family related and physical. Defender scores correlated with all the self -concept area scores except for physical self -concept. This lack of co rrelation between defender scores and physical self -concept scores distinguished defender scores from bully, assistant, and reinforcer scores (which correlated positively with physical self -concept scores), and from outsider and victim scores (which correl ated negatively with physical self -concept scores). Moreover, of all the bullyingrelated role -behavior scores, only defending scores correlated positively with academic and emotional self -concept scores. Therefore, three self -concept dimensions (i.e., physical, academic, and emotional) distinguished children who defend from children in all other identified roles, not just those in the aggressive roles (Salmivalli, 1998). An important and stable individual characteristic that has been studied in relatio n to defending behavior is Big Five agreeableness. Tani, Greenman, Schneider, and Fregoso (2003) administered the Big Five Personality Inventory and the defender scale to 232 8to 10 year -old students of two Italian schools. Children identified as def enders scored higher than children in all other roles on the dimension of agreeableness, which includes altruism as a component (Tani et al., 2003). Another individual variable that relates to defending is biological gender, which will be discussed in det ail later in the chapter. Social -Contextual Variables and the Defender Role The social world of children who defend differs from the environment of other children in several ways. Defenders appear to be more well liked than other children, they appear nev er to be socially isolated, they associate with other children who defend, their classrooms exhibit anti bullying norms, and their defending behavior seems to rely more on social ties than does behavior characterizing other roles. The following review wil l show that these discoveries raise
38 questions about unexplored social aspects of childrens lives (i.e., adult influences, gender norms) that may influence a tendency to defend. Children who defend appear to be more popular than nondefending children, inc luding children not identified in any role. One study conducted in Spain (Monks, Ruiz, & Val, 2002) assessed the sociometric status of 92 children who were 4 to 6 years old. The children were asked to nominate a few peers whom they liked most and liked least. Teachers were asked to name children whom they thought classmates liked most and liked least. To assign bullying related roles to the children based on peer -nomination, Monks et al. (2002) adapted Salmivalli and colleagues (1996) bullyingroles questionnaire into cartoon depictions of bullying related role behavior. They had children look at each cartoon and asked children to nominate some peers who engaged in the depicted behaviors, and then to report whether or not they themselves engage d in the depicted behaviors. Then the cartoon version of the questionnaire was adapted into Spanish text in order to allow teachers to nominate children whom they thought fit the textual descriptions of bullyingrelated role behaviors. The final adapted and translated scales allowed for the selection of children into five categories: Aggressors, victims, defenders, supporters (i.e., a category comprised of assistants and reinforcers) and bystanders (i.e., children who were not selected into any of the oth er four role categories). Results were assessed for sociometry (i.e., popularity) based both on peer -, teacher and self -nomination into the various roles. When reviewing the findings of the Monks and colleagues (2002) study, it is important to note that the method of assessing popularity and role behavior used in the study may have affected the studys results. That is, self and teacher -reports returned no significant results. Only peer nominations of popularity and defender role behavior resulted in s ignificant findings.
39 Children selected by peers as defenders received more like most peer nominations than did those identified as bystanders. Peer nominated defenders also received fewer like least peer nominations than aggressors and supporters. T he peer nominated defender (and supporter) scores also correlated positively with the number of like most peer nominations (Monks et al., 2002). Monks and colleagues (2002) results imply that children who defend are more accepted and less rejected tha n children in any other role, including children who did not meet any threshold criterion for identification in a role. However, it is important also to discuss the apparent effect of nomination strategy (i.e., peer, self, and teacher) on the results of t he study. Peer reported popularity and peer reported defending returned significant findings, but self and teacher reports on those variables did not. This method -specific result may suggest that the traditional peer rating paradigm used to identify chil dren who defend presents a confound. Specifically, there may be a relationship between a childs popularity and his or her salience in the minds of children asked to rate peers on defender behavior. The peer report method of selection may be conflated wi th the nominated childs popularity. Children might simply name popular children as defenders and unpopular children as victims and aggressors. Salmivalli and colleagues (1998) suspect this confound, stating that some of the consistency in the peer eval uations of childrens social behavior might reflect the stability of their social reputation among peers, not their behavior per se (p. 214). On the other hand, the inconsistent results in Monks et al. (2002) may arise from the cartoon adaptation of the d efender scale for the pre -literate participants or the subsequent textual reconstruction of the scale for teachers, rather than from problems with the peer report identification method. In addition, it may be that peers know more than teachers do about th eir
40 classmates. Fortunately, other research studies exist to support the notion that defenders are popular children, although the results from these studies do not necessarily abate concerns about peer -report methodologies. Goossens, Olthof, and Dekker (2006) measured the sociometric status of over 240 school children in the Netherlands and followed up on the same children two years later. Their ages ranged from 8 to12 years in the first year of sampling, and 10 to14 years in the second. The researchers classified children into five sociometric categories: Popular, rejected, neglected, controversial and average. They then used various methods of grouping children into bullyingrelated behavior roles. Each method incorporated a modification of Salmivalli and colleagues (1996) original scale or of its administration. Examples of selection methods compared in Goossens et al. (2006) included using various cutoff scores, having children nominate only a few peers they thought appropriate for each item of the scale, and having children rate all of their peers for each item of the scale, as in the original scale procedure. Regardless of the method used, children identified as defenders were the only children who fell into the category of popular both years, and they were the only children who did not fall into any of the other sociometric categories either year (Goossens et al., 2006). Sociometric status was also assessed in the original Salmivalli and colleagues (1996) study, which is credited for publishing the first version of the bullying role scales. The participants were 573 12to 13 year -old Finnish schoolchildren. Children identified as defenders received more like most peer nominations than children in all other roles, and they received few like least nominations. These researchers also categorized children as popular, rejected, neglected, controversial, and average. Popular children scored higher on the defender scale than did children in other sociometric categories.
41 In addition to being adm ired, children who defend appear least likely to be friendless. Salmivalli, Huttunen, and Lagerspetz (1997) had children aged 11 to 12 sort themselves and their classmates into friendship groups. The children did this by drawing maps of peer groups, in cluding themselves in any group to which they believed they belonged. To depict groups, children would draw classmates in clusters and circle the clusters. Children could depict someone as not existing in any peer group by drawing them outside of all cluster circles. Analysis combining the friendship maps indicated that children who defend were the only children who were always found to have a network of friends, though their friendship networks appeared to be smaller than those of children who bully, as sist, or reinforce (Salmivalli et al., 1997). In addition to enjoying an ample quantity of admirers and friends, children who defend appear to befriend children of a predictable type. Salmivalli and colleagues (1997) found that children who defend tend to associate mainly with each other and that they do not appear to associate with bullies or bullies friends. What is more, a study by Salmivalli, Lappalainen, and Lagerspetz (1998) suggests that defending behavior may relate, more than do other role behaviors, to the composition of a childs friendship group. They sampled children twice over two years, first in sixth grade (ages 12 13) and again in eighth grade (ages 1415). They found that a childs current friendships with children who defend related p ositively to his or her current frequency of defending behavior, but that a childs prior defending behavior did not predict current defending. In contrast, prior bullying and victim behavior predicted current bullying and victim behavior (Salmivalli et a l., 1998). Of note was that predicting defender behavior using current friends roles alone was more reliable than was predicting bully, victim, and other roles using a combination of friends roles and prior behavior (Salmivalli et al., 1998). This
42 appa rently strong link between social sphere and defender behavior seemingly stronger than the social -sphere tie to other role behaviors may suggest a greater reliance of the defender role upon social context, compared to other bullying related roles. In addition to each other, children who defend usually have friends who are identified as victims and outsiders (Salmivalli, Huttunen, & Lagerspetz, 1997). Ironically, despite their popularity among peers and camaraderie with like others, children who defend do not appear to receive peer support for their behavior in a form comparable to that which children who bully receive from assistants and reinforcers. Approximately 16 studies (e.g., Camodeca & Goossens, 2005; Salmivalli et al., 1996) have identified r oles in which children specifically support acts of bullying either by physically assisting or otherwise reinforcing (i.e., cheering, inciting, watching) children who bully, but researchers have not found such supporters for children who defend. This appe ars unfortunate considering the conclusion above that defending behavior may rely heavily on peer support. Another contextual variable that predicts a child being in the defender role is a classrooms set of bullyingrelated norms, which are defined as s tudents expectations about the social consequences of pro or anti bullying behaviorsin their classroom (Salmivalli and Voeten, 2004, p. 248). These expectations may be set verbally or non-verbally, by teachers or students. Norms that are anti -bullyin g appear to contribute to defender behavior (Salmivalli & Voeten, 2004). Interestingly, girls seem to rely more on classroom anti bullying norms for their defender role stability than do boys, even though girls appear more stable overall in the defender r ole than do boys (Salmivalli, Lappalainen, & Lagerspetz, 1998). It is fortunate that a relationship between gender and norms has been explored in relation to defending behavior. However, the norms studied so far have not been gender -specific. In
43 additi on, the apparent lack of defender -supportive roles, and the presence of bully -supportive roles, may either result from an actual lack of peer support for defending or from a lack of defender -focused research. A clear lack of research interest in defenders appears evident, giving practitioners, educators, and other concerned adults little empirical information about how to promote and support defending. The preceding review shows some advancement in the study of defending behavior, including the examinati on of both individual and social -contextual variables. The study of social context helps to begin describing the conditions under which defending occurs; but social variables that have been studied are limited to peer ecology, often excluding adult influe nces. In addition, what has been reported about the social environment of children who defend raises new questions about the effect of gender socialization on defending. A detailed account of findings related to gender and defending will further suggest that there is a need to examine social gender norms in relation to defending. Gender and Defending This section reviews evidence that girls are more likely to defend than boys and that some variables relate to girls and boys defending differently. Bec ause most defender related research only examines defending as part of the spectacle of bullying, reports of gender differences in defending do not always include findings specific to the defender role. Researchers often explicitly state that significant gender differences are found for the defender role, even when not including a statistic to the role. However, some research reports include statistics specific to gender differences in defending. Presented here are findings of gender differences in defen ding. When available, relevant statistics are included to show the sizes of gender effects. Ages and grades are included to show the relevance of these gender results to the population of interest in the present study (i.e., children in middle school).
44 T ypically, studies that have used versions of the defender scale have found girls more likely to be identified in the role of defender. Five studies state that more girls than boys occupy the defender role, without providing statistics specific to that role. Camodeca and Goossens (2005) found that, in seventh and eighth grade (mean age = 11), girls were categorized in the roles of defender, outsider, and victim significantly more often than boys. These authors did not report separate statistics for defend ers, choosing to report the statistic for the three roles grouped together. Goossens, Olthof, and Dekker (2006) compared two methods for determining threshold scores for categorizing children into specific roles. One categorization method included assig ning children to roles based on relative threshold scores for each classroom, ensuring that each classroom had children representing every role. The other categorization method involved setting three absolute score thresholds (i.e., 10%, 15%, 20%) represe nting how many peers believed a participant should be categorized into one of the bullying related roles. The researchers used these relative and absolute thresholds to assign children from the same sample to roles at ages 8 12 and again at ages 1014. At both data -collection points and for all threshold scores, girls were significantly more often categorized than boys as defenders. In this study, Chi squares were offered for the overall effect for all roles, not for the defending role specifically. In a similar study using only relative threshold scores, Salmivalli and colleagues (1998) noted that girls (in eighth grade, ages 14 15) were significantly more likely to be defenders and outsiders than were boys, giving no specific estimation for the defendin g role. Salmivalli et al. (1996) also reported an overall Chi -square indicating significant gender differences for the roles in general, not specifically for the defender role. However, they specifically reported the proportion of girls and boys (in sixth grade, ages 12 13) in the defender
45 role, showing that 30.1% of girls and only 4.5% of boys were categorized as defenders. In another study, Salmivalli and colleagues (1997) reported that 28.4% of the girls versus 4.6% of boys (sixth grade; ages 1112) were classified as defenders, but did not test these differences against a selected probability of chance effects. Four studies to date have reported findings that describe significant differences specifically between boys and girls defending. Salmivall i and Voeten (2004) found that, in grades four and five, girls were more likely to defend than were boys. In grade four (ages 9 -10), the regression coefficient representing the higher prevalence of girls than boys defending was .53, SE = .13. In grade fi ve (ages 10 11) the regression coefficient was .72, SE = .16. In grade six (ages 1112) the regression coefficient was 1.14, SE = .22 (Salmivalli & Voeten, 2004). They further reported that gender was the strongest predictor of role behaviors, compared t o several other variables (i.e., individual anti -bullying attitude, group anti -bullying norms, and group normative indifference to bullying). Salmivalli et al. (1999) reported that, in eighth grade (ages 14 15), girls had significantly higher defender scores than boys, F (1,271) = 31.76, p < .001. Menesini and colleagues (2003) found in a sample of Italian middles school students, aged 11 14, that girls were reported to defend significantly more often that boys, F (1,279 = 34.4). Salmivalli, Kaukiainen, a nd Voeten (2005) found boys had significantly lower defending scores than girls according to both self and peer report, in grades four (ages 9 10) and five (ages 1011). With the range of possible defender scores being 1 to 1, self reports in grade four (ages 9 10) yielded regression coefficients of -.33, SE = .06, and in grade five (ages 10 11) the regression coefficient was -.21, SE = .08, indicating the drop in self reported defender score associated with being male. For peer reports in grade four (a ges 9 10), regression coefficients
46 were -.76, SE = .11, and in grade five (ages 10 11) -.75, SE = .11, indicating the drop in peer reported defender score associated with being male. Another finding that held true for both self and peer report was that g irls were significantly more stable in the defender role than were boys (Salmivalli et al., 1998). In this study, boys categorization in the defender role in a particular year (sixth grade, ages 12-13) did not correlate with their presence in that role t wo years later (eighth grade, ages 14 15), while girls did, as evidenced by correlations found with peer report, r = .47, p < .001, and self report, r = .48, p < .001. Some studies did not find that girls defend more often than boys. Sutton and Smith ( 1999) showed mixed gender results for defender behavior. Students in other studies typically rate all of their classmates on bullying related role behaviors. However, participants in this study only scored same -gender classmates on role behaviors. Boys scored higher than girls (ages 7 10) on the defender scale, t (191) = 2.05, p < .05, but a table indicates that a higher percentage of girls than boys were actually categorized into the defender role. Of the remaining studies using the defender scale, Maed a (2003) tested for and did not find significant relationships between gender and defender role categorization in fifth and sixth -graders (aged 1013); and Salmivalli (1998) and Tani et al. (2003) did not consider gender effects in their study. Unlike o ther studies, Rigby and Johnson (2006) did not use reported defending behavior as a dependent variable, but asked children whether they would object to bullying in two imaginary scenarios. Their participants were students in the Australian school system i n primary school (years 6 and 7) and secondary school (years 8 and 9). In primary school, girls claims that they would interfere were stronger than those of boys in both verbal, 2(4) = 11.26, p < .05, and
47 physical bullying scenarios, 2(4) = 13.85, p < .01. There were no significant gender differences in claims of past intervention in bullying situations. An audio -visual method for observing bullying events, designed by Pepler a nd Craig (1995) to be unobtrusive and naturalistic, was employed to observe defending in first to sixth grade children in Toronto in a 2 -year study (Hawkins, Pepler, & Craig, 2001). Results showed more boys than girls were observed defending. This findi ng contradicts the survey research cited above in which girls appear to defend more than boys, and it appears that naturalistic observation is a more valid measure of defending behavior than child reports. However, Hawkins and colleagues (2001) also note d that a majority (i.e., 61%) of the children present at observed bullying situations were boys, and that children were significantly more likely to intervene when a bully or victim shared their gender. When the number of boys and girls present in observe d bullying situations was taken into account, no gender differences were evident in defending behavior. Nor was a significant relationship found between gender and type of defending (i.e., aggressive and nonaggressive). Although this finding does not con tradict other studies finding girls to defend more than boys, it fails to support them. It is, however, possible that the Hawkins et al.s (2001) selection of children to be observed created a sampling bias. The authors only applied surveillance to stud ents who were nominated by peers as bullies, victims, or bully -victims children around whom bullying was thought most likely to occur (Hawkins et al., 2001). This procedure may have caused more of the sampling to be conducted among bullying circles than i f children were selected at random for observation. It is possible that bullying circles differ from other social circles in defending frequency, type (i.e., physical, verbal), or timing (i.e., before, during, after bullying). In addition,
48 children were aware that they were being observed, which may have caused differences in actual defending behavior. The finding that girls defending is more dependent than boys defending on classroom anti -bullying norms (Salmivalli et al., 1998) begins the process of d etermining the connection between social norms, gender, and defending. Furthermore, gender differences in the relationship between normative pressure to help victims and childrens stated willingness to defend have been found. Rigby and Johnson (2006) ex amined imagined or anticipated defending behavior in relation to gender and the expectations of others. Their results showed that a childs reported willingness to intervene in an imagined bullying scenario varied according to his or her level in school ( i.e., primary vs. secondary), gender, and expectations of selected others. Boys reports of willingness to defend were more related to friends expectations to defend than to their parents expectations, which were more related to girls expressed willing ness to defend (Rigby & Johnson, 2006). These findings suggest that the expectations of others affect girls and boys defending differently. Furthermore, girls perceived more pressure from parents and friends to help victims (Rigby & Johnson, 2006). M ore research is required to determine whether the expectations of others are internalized as gender identity in a way that influences reports of actual (i.e., not imagined) defending behavior. The present study investigate d how gender, social pressure, an d gender identity combine to predict defender behavior in children. A review of the application of both individual and social context variables demonstrates that individual differences paint only a partial picture of defending behavior, and that knowledg e of social context influences will help create a more detailed representation of defending behavior. Gender is an individual, arguably fixed variable that has been shown to strongly
49 predict defending behavior. However, gender is also a socially construc ted variable, which may change as social environments change. It is this changeable aspect of gender that may be important in improving childrens gender related behavior. Because caregiving, educating, and counseling are social activities intended to fa cilitate positive growth or change, the social components of the relationship of gender to defending behavior are essential for empowering adults who care about improving safety at schools. Social Identity Theory and Defending The overarching proposition o f s ocial i dentity t heory (SIT; Tajfel & Turner, 1986) is that humans have a drive to achieve a positive self view based on membership in a highly regarded group. Many types of groups can be considered important to a persons identity, including religious, political, economic, racial, and other groups. Among those types of groups, one often stands out as salient for a particular person in helping define his or her identity. For example, a person may be more inclined to compare him or herself to others ba sed on racial group identity rather than religious group identity. Of interest in the present study wa s that SIT ha d been used to postulate the effects of ones gender group identity. Gender identity is usually defined as an unarticulated, global sense of ones maleness or femaleness that is acquired early in life and is considered to be relatively impermeable to change (Lorenzi -Cioldi & Doise, 1990; p77). In the present study, gender identity wa s considered to be partly determined by principles of social identity theory. The present study examine d the relationship between gender identity and defending behavior. SIT ha d been used in the past to explain attitudes of dominance and acts of bullying. If the theory applies broadly to processes related to bullying, it should have predict ed results in this study related to defending behavior. Its application to gender identity help ed determine which variables might have mediate d or moderate d the apparent relationship between gender and defending.
50 Various methods can be used to enhance group-based self -esteem when an in group identity is not providing satisfactory esteem to its members. One can change groups or members can act to improve a groups image. SIT postulates a firm but flexible group members hip. The theory asserts that individuals not perceiving positive self -identity from an in groups identity will sometimes transfer groups. The possibility for individuals to change groups is called social mobility (Tajfel & Turner, 1986). However, there are times when it may be difficult to change groups. For example, some may feel that changing ones gender grouping is difficult. In such a situation, a person can strengthen his or her own sense of personal identity by helping enhance his or her group s status and then working to ensure continued membership in that group. Acting in ways that help enhance group status when individual social mobility is not possible is termed social change (Abrams & Hogg, 1990; Tajfel & Turner, 1986). Social change oft en involves comparing ones ingroup to a superior out -group on a dimension different from that on which the out -group excels. This strategy is called social creativity When a group is outmatched on a particular attribute, social creativity allows tha t group to highlight an attribute on another dimension to boost in-group esteem. Take, for example, the assumption that males succeed more at domination than do females (Sidanius & Prato, 1999), which is corroborated by the ample findings that boys are mo re likely than girls to bully. Social creativity would allow girls to choose another dimension by which to show their value, one on which they may be considered superior to males. So, if relationship and empathy are considered positive female attributes (e.g., Hall & Halberstadt, 1980), then social creativity may involve girls emphasizing the relationally empathetic acts of befriending and supporting victims when girls are presumably outmatched on domination. In a sense, defending allows girls to gain gr ound by taking the higher ground.
51 Sometimes social change involves social competition in which an in -group directly challenges the dominant status of another group in a way that could improve the status of the in group vis -vis the salient out-group. Th e aim of social competition is either to win out over the out group or to deconstruct the social assumptions that permit the dominant status of the out group (Abrams & Hogg, 1990). Schmitt, Branscombe, and Kappen (2003) found that women are more likely to endorse social -dominance oriented beliefs when considering inequality that favors women. This finding accords with SITs concept of social competition, in which group members might favor direct rivalry with a relevant out -group to enhance group status. Perhaps then, defending can also be seen as a method of direct social competition. When a person defends, he or she may be challenging the bully seeking to undermine his or her authority. If this is a cross -gender challenge, in which a girl is intervenin g with a male bully, it can be seen as bolstering the female gender groups position over the male group through direct rivalry. However, even if it is not a cross gender challenge, female defending behavior may still be seen as social competition as def ined by SIT. To be bullying, aggression must be committed by the powerful upon the powerless. In challenging the enforcement of power differences, defenders are challenging the very idea of hierarchy the social concept that keeps one gender below another Instead of competing to beat another group, girls who defend may be competing to equalize groups, thereby gaining increased relative status for their gender. The salience of a particular in -group/out group dimension is also important to personal positiv e self -concept. People belong to many in-groups and they stress some in-group memberships over others. Specific to gender identity, Lorenzi Cioldi and Doise (1990) discuss how gender is a more salient identifying group for some people than for others. Sex typed individuals are males and females whose self -conceptions vary mainly along a stereotypical and
52 bipolar masculinity -femininity dimension. They make use of gender schema to a greater extent than individuals classified as androgynous or undifferent iated (Lorenzi Cioldi & Doise, 1990; p. 77). Therefore, some people may not work to enhance the status of their gender group so much as of another in -group (e.g., religious, academic, regional). Therefore, in the present study the increase of defendin g behavior according to these SIT strategies (i.e., social mobility, social creativity, social competition) was expected to be higher in girls who strongly identif ied with their female identity. Similarly, boys tendency to avoid defending behavior was ex pected to be less extreme if their most salient group identity wa s not gender based. If males are thought to be dominating (Sidanius & Prato, 1999), a lower level of male identity would have predict ed a lessened propensity to commit acts of domination (i. e., bullying) and to avoid counter behaviors (i.e., defending). Meanwhile, if females are considered sensitive, those females who we re highly gender typed would have been more likely to commit acts of sensitivity (i.e., defending) than females who d id not hold their gender identity as primary. The literature contains evidence of the effect of gender socialization variables upon views about dominance. Gender differences in peoples orientation toward social dominance appear to be moderated by ones level of gender identity (Wilson & Liu, 2003), such that higher gender identity in males and lower gender identity in females is associated with greater favoring of social dominance. Schmitt, Branscombe, and Kappen (2003) find that, while men are significantly m ore oriented toward social dominance than are women, sexism scores mediate this relationship to the point of insignificance. Findings regarding peoples orientation toward social dominance can be interpreted as relevant to bullying, because the repetitively enforced power imbalance in the definition of bullying implies that it is an act of domination. Gender
53 socialization variables in the present study should, therefore, have also modif ied the relationship between gender and defending since defending wa s considered behavior that opposes domination. It is not entirely an individuals decision whether he or she changes in-groups, emphasizes a particular in -group attribute, competes directly with a salient out -group, or experiences a particular in -group mem bership as most salient. Social identity and the status of ones in group are based upon many peoples perceptions: that of the individual, those of other in -group members, and even those of out -group members (Tajfel & Turner, 1986). Groups include and e xclude persons based on their behavior. For example, an in -group will likely retain an in -group member who bullies if the in-group has pro -bullying norms, and it will retain an in group member who intervenes in bullying if the in-group has pro -fairness no rms (Ojala & Nesdale, 2004; cf. Sentse, Scholte, Salmivalli, & Voeten, 2007). Groups also reward or punish behavior depending on whether it is thought to enhance or detract from group distinctiveness and favor. For example, disregarding pro bullying and pro -fairness norms, in -groups will retain a member who bullies out group members who appear too similar to the in-group, threatening group distinctiveness (Ojala & Nesdale, 2004). Therefore, physical enforcement of in -group boundaries is thought to occur. This threat of violence or exile indicates that ones behavior is strongly influenced by other people, whether they are in -group members or not. Consequently, using SIT to predict defending behavior requires being aware of the normative pressures that come from inside and outside of a participants in group. Normative pressure (i.e., from mother, father, teachers, and friends) to help victims wa s the variable selected to measure such influences from inside and
54 outside the in group. The present study i nclude d a comparison of the gender in-group and out group normative pressure (i.e., from parents) for their relationship to defending behavior. Various SIT strategies of self and group -promotion frame d predictions in the present study regarding gender dif ferences in defending behavior. If strong support for SIT wa s to be construed from the findings of this study, normative pressure to help victims and gender identity should have modif ied the importance of biological gender in predicting defending behavior Support for the Need for the Study Children in the defender role do not appear to stay in the defender role as time passes or as friends change. They often assume other roles, sometimes becoming victims of bullying (Brown, Birch, & Kancherla, 2005; Salm ivalli, Lagerspetz, Bjrkqvist, sterman, & Kaukiainen, 1996; Sutton & Smith, 1999), other times becoming outsiders (Goossens, Olthof, & Dekker, 2006; Sutton & Smith, 1999) and sometimes becoming assistants and reinforcers of bullies (Sutton & Smith, 1999) Why might many children abandon the defending role in favor of roles thought to be less socially desirable? Why might many children remain in the bully role, though it is widely advertised as a problem? A contention of this dissertation wa s that the a ct of gathering and disseminating information about a behavior may increase the likelihood of that behavior occurring and that relative abundance of bullying research and intervention compared to that for defending should, therefore, be diminished. Wheth er a behavior is productive or destructive, increasing awareness about it may promote it. A meta analysis conducted by Paik and Comstock (1994) of 217 studies dating from 1957 to 1990 confirms the common claim that watching television violence increases a ggressive behavior in the watcher. More related to the effect of research and intervention is Ringolds (2002) discussion of the boomerang effect in which public campaigns intended to reduce unhealthy behaviors actually increase the probability of the behavior occurring.
55 Several mechanisms are offered to explain the boomerang effect. However, the one most in keeping with the tenets of social identity theory involves normalizing a negative behavior. Public health campaigns can prime awareness in the t arget audience of the prevalence of a negative behavior. Audience members might perceive this prevalence as confirmation that the negative behavior is normative within their group, and the need to conform to group expectations increases the occurrence of that behavior in those who have heard the warnings (Ringold, 2002). Evidence exists that a more auspicious use of the boomerang effect can be attained by normalizing positive health behaviors to increase their likelihood (e.g., yoga; Rimal, Lapinski, Coo k, & Real, 2005). Similarly, research, publicity, and interventions highlighting either bullying or defending may promote the target behavior. For example, activities designed to target and reduce bullying may unintentionally increase its occurrence. Fo rtunately, it might be that research, publicity, and intervention focusing on defending will promote that behavior. The present study wa s intended to add to the small body of work that highlights defending. Summary of Major Points Bullying appears to be a fairly common and harmful phenomenon. Research on bullying has evolved from narrowly focusing on the bully and victim to considering the social environment that perpetuates it and from intervening on an individual level to intervening on a group level. This greater emphasis on social context has led to the identification of various bullying related roles, including a role in which children defend victims. However, little is known about children who defend, and recommendations (Salmivalli, 1999) to rec ruit these children as peer leaders in interventions appear to have been largely ignored. Nevertheless, research evidence about children who defend is available, which shows them to be empathic, moral, confident, and well liked children. Importantly, ch ildren who defend are more likely to be girls than boys. Learning more about gender differences in defending behavior
56 may have implications for practitioners and educators. Social identity theory was considered an apt framework for predicting the influen ce of social context on gender differences in defending behavior. Defending behavior wa s expected to relate to socially constructed gender identity and correlational findings we re expected to help confirm this relationship. The next chapter describe s the procedures and instruments that were used to investigate these predictions.
57 CHAPTER 3 METHODOLOGY Defending is an attempt to intervene in the process of bullying or to ameliorate its effects on a victim. Research on defending in schools suggests that girls are more likely than boys to defend victims. The purpose of the present study was to investigate whether ones gender identity and normative pressure to help victims modify the relationship between gender and levels of defending behavior. This wa s considered a test of the suitability of social identity theory (SIT; Tajfel & Turner, 1986) for explaining gender differences in defending behavior. The present study use d a sample of middle school students in a town in the rural Southeastern United Stat es. Children were administered a measure of defending, as well as measures of gender identity, normative social pressure to help victims, and biological gender. In other research, age appears to moderate the relationship between normative pressure to hel p victims and a childs stated willingness to defend. Therefore, age was planned for use as a control variable if a correlation were suggested between it and defending Analyses were conducted to test whether the relationship between gender and defending wa s mediated or moderated by the other predictor variables. Participants Various age groups have been studied in regard to their defender behavior. Studies requiring no reading ability adaptation of the defender scale have included participants with ages ranging from 9 (Salmivalli & Voeten, 2004) to 15 years old (Salmivalli, 1998; Salmivalli, Kaukiainen, Kaistaniemi, & Lagerspetz, 1999; Salmivalli, Lappalainen, & Lagerspetz, 1998). Other scales used in the present study appear be most appropriate for use on a middle school sample. Therefore, students in sixth, seventh, and eighth grade were sought for participation.
58 Availability sampling was used. All students supplying informed consent from their parents were included in the study, if present on the day of data collection To determine the minimum sample size, correlations between defender scores and other variables, as reported in prior studies, were examined. The multiple regression analysis involving the greatest number of variables involves the final questions in which a model is constructed of all variables and interactions discussed in other analyses. This analysis includes k = 10 predictor terms. The data available for a predictive variable in a multiple regression analysis, which was the negative relationship between a childs defending scores and his or her friends bullying scores (Salmivalli et al., 1998). On a power analysis table (Algina & Olej nik, 2003), selecting the tailed tests indicated the need for a sample size of 104 participants, after adjusting for k = 10 possible predictor terms. Assuming a 20% response rate of parental informed consent for their children's participation, the researcher sought to solicit participation from 525 students to achieve a sample size of 105. An accuracy table (Algina & Olejnik, 2003) indicated that a correlation from this sample size, n = 105, would estimate a population correlation within .15 to .20. In order to strive for a demographically representative sample of middle school students, children from four middle sc hools were sampled. The first middle school wa s part of a K 12 university affiliated research school, and the county school board in the same community manages the other three middle schools. The universityaffiliated middle school population is thought be demographically similar to the local population, but not representative in regards to behavior issues. This is because, though student admissions aim to build a demographically
59 representative student body, children are also screened to exclude those wh o evidence severe behavior problems. Limitations in generalizability of the results are discussed in chapter 5 within context of how comparable the three schools are to each other and how comparable the final sample is to middle school students in the reg ion. Sampling Procedures Upon receiving approval from the university institutional review board, from local school administrators, and from teachers, the researcher sought informed consent from students and parents until an adequate number of students were eligible to participate in the study. Please see Appendix A for a draft of the informed consent document. The survey was administered to children in paper and -pencil format. Data collection took several weeks Instructions and scales appear ed in the su rvey packet in the following order: Participant Instructions for Survey, Right to Refuse, and Informed Consent (Appendix A); Defender Scale (Appendix B); Scale of Normative Pressure to Help Victims (Appendix C); Childrens Personal Attributes Questionnaire Feminine -Masculine Scale (Appendix D); Demographic Form (Appendix E); and Contact Instructions for Participants (Appendix F). The demographic form ask ed for date of birth, biological gender, and race/ethnic/cultural identification. The demographic form appear ed after the other scales to avoid priming gender at the beginning of the survey. Birth dates were solicited as two items, asking the year first, to avoid the potential for respondents to make the error of writing the current year instead of their birth year. When participants were finished, the researcher or a proxy collect ed each survey directly from each student, leaving each student with contact instructions for the principal researcher, the research advisor, and a local crisis center. If partic ipants indicate d confusion regarding any part of the survey, the researcher made note of such feedback anonymously.
60 Design The study use d a cross -sectional, correlational, multivariate survey design. Three predictor variables (i.e., gender, gender identi ty, and normative pressure to help victims) and one control variable (i.e., age) were assessed for their relationships to the dependent variable (i.e., self reported defending). Gender was converted to a point -biserial format to permit correlational analy ses. Bivariate analyses were conducted for all variables, as were multivariate analyses using every combination of the predictor variables possible. All relevant two -way interactions among the predictor and control variables were tested. T hese interacti ons we re hypothesized in relation to SIT and to aid discussion of methodological limitations and suggestions for future research. Causal interpretations were not made from results of the present study but causal possibilitie s are used to suggest future causal research. In addition to correlational analyses described, primary factor analysis was conducted to test the defender scale for unidimensionality when administered as a self -report measure. Measures Reliability and validity information were present ed here to show why the instruments selected for the study were optimal for measuring the constructs of interest. The constructs of interest were defending behavior, gender identity, normative pressure to defend, and biological gender. Defending behavior was assessed using the self report defender scale (Salmivalli, 1998; Salmivalli, Kaukiainen, Kaistaniemi, & Lagerspetz, 1999; Salmivalli, Lappalainen, & Lagerspetz, 1998). Gender identity was assessed using the feminine -masculine scale of the Childrens Personal Attributes Questionnaire (CPAQ -FM; Hall & Halberstadt, 1980). Normative pressure to defend was assessed by adapting a series of questions developed by Rigby and Johnson (2006). Biological gender was asked on a demographic questionnaire.
61 The Self -Report Defender Scale Self reported defending was the dependent variable of interest in the present study. Refer to Appendix B to view the selected self report defender scale. The defender scale is one of six scales on the Participant Role Questionnaire (e.g., Salmivalli et al., 1996; Salmivalli et al., 1998). The other scales on that questionnaire measure bullying, assisting, reinforcing, remaining outside, and being victimized. The other role scales were not used in the present study Only the defen der scale was considered appropriate for the present study Analyses involve d correlating defender scores with predictor variables. Various revisions of the defender scale exist, using both peer and self report to estimate defending behavior. Evidence i s here presented supporting the self report use of the defender scale to estimate defending behavior as a continuous variable. As described below, self reports and peer reports on defender scales have been found to correlate positively; and because of thi s relationship, data from previous peer -report administrations provides some support for the use of the self report defender scale. Also described below is also supportive evidence derived directly from self report use of the defender scale. Finally, the author present s reasons that the selected version of the self report defender scale (Salmivalli, 1998; Salmivalli et al., 1998; Salmivalli et al., 1999) is most appropriate version for the present study Defending is prevalent enough in child populations to measure without prohibitive sample sizes. Studies using the original defender scale scoring method (Salmivalli et al., 1996) estimate the proportion of defenders in various populations to range between 15.6% and 27.5% (Goossens, Olthof, & Dekker, 2006; Maeda, 2003; Salmivalli, 1998; Salmivalli et al., 1996; Salmivalli et al., 1998; Sutton & Smith, 1999). Studies exploring alternative scoring methods return a wider prevalence range for defenders: 5% (Goossens et al., 2006) to 46.1% (Sutton & Smith, 1999).
62 Criterion -related validity for the defender scale has been shown in relation to variety of individual characteristics: genuine self -esteem, aggression, moral sensibility, anti bullying attitudes, empathy, self -control, social cognition, emotion regula tion, trait agreeableness, and various self -concept dimensions. In addition, criterion-related validity has been established in connection with various social context variables: popularity, having defenders as friends, and classroom anti -bullying norms. Furthermore, factor analysis of peer report scales (Camod eca & Goossens, 2005; Goossens et al. 2006; Salmivalli, Huttunen, & Lagerspetz, 1997; Salmivalli et al., 1998; Sutton & Smith, 1999; Tani, Greenman, Schneider, Fregoso, 2003) confirm that the defend er scale measures a single behavioral role dimension that is distinct from other bullying related behavior roles (i.e., bully, assistant, reinforcer, and outsider). Self report defender scale scores correlate significantly and positively with peer report defender scores, r = .46, p = .01, and negatively with peer report bully, r = -.14, p = .01, reinforcer, r = -.21, p = .01, and assistant, r = .19, p = .01, scales (Salmivalli et al., 1996). The scale is appropriate for use on middle school students. N orming of the first version of the defender scale employed a sample of fourththrough sixth -graders, and the scale was designed for use on children aged 10 through 14 (Violence Institute of New Jersey, 2007). Other versions of the scale, including the se lected version, have been administered to children in middle school grades, international equivalent grades, or on children of middle school age (Gini, 2006; Gini et al., 2007; Goossens et al., 2006; Maeda, 2003; Salmivalli, 1998; Salmivalli et al., 1996; Salmivalli et al., 1997; Salmivalli et al., 1998; Salmivalli et al., 1999; Salmivalli, Kaukiainen, & Voeten, 2005; Salmivalli & Voeten, 2004). The version of the defender scale chosen for the present study wa s a six item scale used in three studies (Salmivalli, 1998; Salmivalli et al., 1998; Salmivalli et al., 1999). Self -report data
63 we re available from two of those studies (Salmivalli et al., 1998; Salmivalli et al., 1999). All three studies were conducted with eighth -grade participants. This version o f the defender scale ha d been used more often than other versions. In addition, it appear ed to be the only self report defender scale used in a study that includes a factor analysis (Salmivalli et al., 1998), confirming that the defender scale measure d a unified construct separate from the constructs measured by the other bullying scales in the same questionnaire. However, it wa s possible that only peer report data was used in the factor analysis, because peer report was the authors primary interest in t hat study. Factor analysis of item responses to the selected scale were conducted for the present study Internal reliability for the selected scale was reported (Salmivalli et al., 1998) as a range containing coefficient alphas of all of the bullying r reliabilities may have been for peer report use of the scale. Internal reliability based on self report use of the selected scale was analyzed in the present study Test retest reliability, ha d been researched fo r the selected self report defender scale. Test retest reliability over two years for the selected self report defender scale was significant, r = .37, p < .001, and identical to that reported for the corresponding six -item peer report defender scale (Sal mivalli et al., 1998). Self report defender scores on the selected scale correlate significantly and in expected ways with several variables. For girls, though not for boys, self report defender scores correlated positively with peer evaluated self -esteem r = .30, p < .001 (Salmivalli et al., 1999). In addition, expected gender differences (i.e., similar to peer report) occur in the two-year stability of self reported defender behavior, in that girls have significantly stable defender scores, r = .48, p < .001, while boys do not (Salmivalli et al., 1998).
64 Although peer report and self report use of the defender scale may estimate defending behavior in different ways, some evidence supports self report use of the defender scale. Using self report, results of a study by Salmivalli and colleagues (1998) have shown that children who retain the same classmates over two years have stable defender scores, r = .55, p < .01, while students acquiring different classmates do not; but the finding is opposite using pe er report. The self report -based defender scale findings are more consistent than the peer report -based defender scale findings are with the same studys self and peer report stability findings for other roles. Evidence of a possible limitation of the self -report defender scale exists in findings suggesting that children may be inflating their own defender scores (Salmivalli et al., 1996; Salmivalli et al., 1998; Sutton & Smith, 1999). Self assessment, therefore, may be subject to a social desirabilit y bias. Evidence contradicting a social desirability bias exists. For example, peer -identified defenders appear to underestimate their own defending behavior (Sutton & Smith, 1999). In addition, defender role stability does not seem to be different for self and peer assessment (Salmivalli et al., 1998). Furthermore, scores on an arguably undesirable behavior, reinforcing, are higher for self-report than peer report (Salmivalli et al., 1996). Moreover, peer reported bullies more often identify themsel ves as bullies, reinforcers, and assistants than as any other role; whereas peer reported defenders more often identify themselves as defenders or outsiders (Salmivalli et al., 1996). Finally, peer and self reported defending have been shown to correlate significantly with each other (Salmivalli et al., 1996). Therefore, it does not appear that social desirability bias explains all of the differences apparent between self and peer reported defending. A direct test of social desirability bias wa s not wi thin the scope of the present study. Hawkins and colleagues (2001) found that children are significantly more likely to intervene in a bullying situation when a bully or victim shares their gender. To explore this
65 potential relationship for use as a pos sible control variable and for interpretation in an SIT context, half of the participants w ere asked to complete the defender scale in reference to situations in which they witnessed a boy bullying another boy, and the other half complete d the scale in ref erence to situations during which they witnessed a girl bullying another girl. This situation -specific prompt was accomplished by rotating two forms of the survey among participants regardless of participant gender. Thus participants were assigned to r ecalling actual situations in which they either shared the gender of the bully and victim, or did not. Gender Identity To measure participants level of gender socialization (i.e., gender identity), the researcher administer ed the long form of the Children s Personal Attributes Questionnaire Feminine Masculine Scale (CPAQ FM; Hall & Halberstadt, 1980). Please see Appendix D for a copy of the CPAQ -FM. The CPAQ -FM is one of three scales contained on Hall and Halberstadts (1980) questionnaire. The other tw o scales in the questionnaire, Feminine (CPAQ -F) and Masculine (CPAQ -M), we re not considered appropriate to the questions of the present study Those scales measure masculinity and femininity as gender attributes that can coexist in a person of any sex; w hile the present study investigate d a social drive to see gender groups as having mutually exclusive attributes, as described in gender applications of SIT. Regarding the appropriateness of the CPAQ FM for measuring this tendency, Hall and Halberstadt (19 80) state It is not surprising that the Feminine -Masculine scale should show more marked gender differences than the two unipolar scales, which by definition consist of attributes that are socially desirable for both males and females (p. 279), whereas t he one dimensional CPAQ -FM scale contains attributes considered differentially desirable for males and females. Similarly, the research questions of the present study regard ed whether defender behavior wa s differentially desirable on a social level for gi rls and boys.
66 Higher scores on the CPAQ -FM denote higher masculinity and lower femininity, and lower scores denote the opposite (Hall & Halberstadt, 1980). In a study of children aged 8 11, the internal consistency of the CPAQ & Halberstadt, 1980). Internal consistency of the scale was also examined in the present study The CPAQ -FM scale has significant and good test retest reliability, r = .46, over one year (i.e., grades 34 in the first year, grades 4 6 in the next year; H all & Halberstadt, 1980). Evidence supports criterion -related validity for the scale. CPAQ -FM scores correlate significantly, r = .86, with scores on the adult version of the scale, when both are completed by a single adult group (Hall & Halberstadt, 19 80). CPAQ -FM scores also correlate significantly with mothers descriptions of their children on the CPAQ -FM, r = .33, in thirdthrough sixth graders. Furthermore, the CPAQ -FM scale significantly correlates positively with the masculine scale, r = .51 ( Hall & Halberstadt, 1980). In addition, the CPAQ -FM discriminates gender in expected directions, such that higher (i.e., more masculine) scores occur in boys, with reported effect size of .90 SD in a sample of children in grades three through six. This discriminative effect emerged also for older children using shorter versions of the scale (Hall & Halberstadt, 1980). As expected by the scales developers, higher masculinity scores on the CPAQ -FM significantly predicted higher scores on some self -concep t dimensions for both boys and girls. In the same article, studies using the short version of the CPAQ FM which correlates positively with the long form revealed expected relationships between girls scores on the scale and scores on two other scales: observer rated assertiveness, r = .36, and observer -rated dependency, r = -.40. In addition, discriminant validity was supported by the confirmation of a hypothesized weak correlation between the scale scores and intelligence scores (Hall & Halberstadt, 19 80).
67 Normative Pressure to Help Victims Rigby and Johnson (2006) used a four item scale (Rigby, 2005) to determine a childs perception that his or her mother, father, friends, and teacher would expect him or her to support a victim or support a bully in a given bullying situation. Lower scores on this scale represent a perceived expectation to support a bully and higher scores represent a perceived expectation to support a victim. Each of the four items asks a child to indicate the expectations of a single source (e.g., mother) of normative pressure. Participants are asked to omit a question regarding parental expectations if that parent is not living. The scale appears to function as four single item subscales in the studies where it has been used (Ri gby, 2005; Rigby & Johnson, 2006), and no internal consistency data is available for the scale as a whole. An expanded version of the scale w as created for the present study I nternal consistency for the entire scale w as examined in the present study as were internal consistenc ies of subscales derived from expanding each of the single items. Please see Appendix C for a copy of the scale of normative pressure to help victims, adapted for the purposes of this study. The original scale, or procedure, was administered to children in their primary school years (i.e., sixth and seventh, mean age 11.5 years) and secondary school years (i.e., eighth and ninth, mean age 13.5) in Australian schools (Rigby & Johnson, 2006). The reported mean ages correspond to e xpected ages for sixth seventh eighth -, and ninth -graders in US schools; therefore the findings we re transferable to the sample of the present study Findings related to the scale appear to accord with SIT when it is presumed that a female identity in cludes a social norm of compassion while a male identity contains a norm of power. The level of normative pressure in the Rigby and Johnson (2006) study differed by a childs gender and the identified source of normative pressure. In primary school, sign ificantly more girls than boys thought their fathers and friends would expect them to support a victim. In
68 secondary school, significantly more girls than boys thought their friends would expect them to support a victim (Rigby & Johnson, 2006). Claims by children that they would support a victim in a bullying scenario varied by gender and normative pressure. For boys in primary, r = .47, and secondary school, r = .38, but not for girls, normative pressure from friends correlated significantly with willin gness support a victim. For girls in primary school, but not for boys, normative pressure from mothers, r = .21, and fathers, r = .36, correlated significantly with commitment to support a victim. This effect held for girls in secondary school for normat ive pressure from both mother, r = .25, and father, r = .35, and a relationship was found in secondary school between boys expected pressure from their fathers and their willingness to support a victim, r = .25 (Rigby & Johnson, 2006). It also appeared i n the study that, with exception, boys willingness to help victims was more related to friends expectations, while girls willingness to help victims was more related to parental expectations. Because the relationship between normative pressure and cla ims of willingness to oppose bullying differed by gender in Rigby and Johnsons (2006) study, gender differences we re expected in the relationship between normative pressure and actual reported defending. Correlational analyses include d tests to determine whether an interaction exists between normative pressure (i.e., for each source) and gender when predicting defender scores. An extensive search conducted by this author suggests that the scale of normative pressure to help victims wa s apparently the mo st thoroughly validated procedure for assessing the specialized construct necessary for the present study : that of social pressure to defend, as exerted by adults and other children. For the purposes of this study, adaptations were included to permit the measurement of internal reliability of the measure of normative pressure from each independent source (i.e., mother, father, friends, and teachers). The original studies showed
69 photographs of bullying scenarios thus assigning students to different situat ions, and asked the children to respond to the normative pressure items. In the present study each of the four items were administered four times as applied to four different bullying scenarios derived from the textual prompt consistently used with the d efender scale, and photographs were used. The reason for replacing photographs with prompts derived from the defender scale wa s to ensure that all scales related to bullying we re based on the same definition of the phenomenon. However, the definition de scribes four broad categories of bullying, providing the opportunity to prompt children to respond to normative pressure scale items in four different bullying situations: direct physical, direct verbal/social, indirect physical, and indirect social/verbal. This adaptation likely add ed to the reliability of measuring normative pressure to defend, but also allow ed for the measurement of internal consistency for each source (i.e., mother, father, friends, and teachers). The adapted scale consist ed of 16 tot al items instead of 4, and 4 items for each pressure source instead of 1. Furthermore, because the terms friends and teacher may be too broad to describe those persons most socially significant to participants, the terms best friends and favorite te achers have been substituted. Both terms were stated in the plural to ensure that the scale prompt ed children to assess group pressures as opposed pressure from a selected individual, th us keeping closer to constructs described in SIT. Moreover, the pre dominant age -group of best friends was asked in the demographics questionnaire to ensure that child -on-child pressure was assessed, and the predominant gender of best friends and favorite teachers was ascertained in order to determine if gender in -group pr essure and gender out -group pressure differ ed in their relationship to defending behavior. These intended improvements to the assessment of
70 normative pressure to help victims were assessed for reliability, and for validity via a factor analysis and conten t analysis Age Age appears related to scores of defending and to normative pressure to help victims. Thus the age of a participant was anticipated as a controlling variable (see demographics form, Appendix E). In Rigby and Johnsons (2006) study, s eemin g age effects appeared in boys regarding normative pressure to help victims, such that primary school boys willingness to object in a bullying situation did not correlate to pressure from either parent, but secondary school boys willingness to object did correlate with pressure from their fathers. Older girls in the study were also significantly more likely than the younger girls to perceive that their mothers, fathers, and friends expected them to support a victim (Rigby & Johnson, 2006). These age effects in past research suggest ed that it could have been important to control for the effects of age on the predictor variables relationships to defending behavior in the present study Though significant correlations between each source of pressure and defending were reported by gender and age group in Rigby and Johnson (2006), it was not reported whether the correlations differed significantly by gender and age group. The present study tested for interactions between age and gender for each normati ve pressure source. Various increases and decreases, related to age or grade level, appear to occur in the number of children identified as defenders in studies. Salmivalli and Voeten (2004) found that children in the fifth and sixth grades were less li kely to be defenders than children in the fourth grade. Conversely, Salmivalli and colleagues (1998) found that students were more likely to be defenders in the eighth grade than they were in the sixth grade. Therefore, age and grade level were planned a s control variables in the event that defending correlated with age or differed by grade level.
71 Analyses Various correlational analyses were employed in the study. For all analyses described, assumptions of homoscedasticity, linearity, conditional normal ity, and independence were tested, and corrections made where appropriate and feasible. Zero -order correlational analyses were conducted between defending and all predictor and control variables (i.e., participant gender, gender identity, normative pressure to help victims, shared gender between participant and bully or victim, normative pressure source gender for parents, and age). In addition, zero order relationships were tested between participant gender and various proposed mediating variable s in ord er to diagram paths of mediation from gender through each of the other predictor variables. Multiple regression analyses were also conducted. The first multiple regression analysis include d participant gender as the predictor variable and gender identit y as the mediating variable, with defending as the dependent variable. All paths (i.e., predictor -dependent, predictor -mediating, mediating dependent) are described as correlations, including correlations between predictor and dependent variables before a nd after adding the mediating variable. The effect of mediation was analyzed using a method described by MacKinnon, Lockwood, Hoffman, West, and Sheets (2002) including a test for significance; and a SAS program (Taborga, Cheong, & MacKinnon, 2000) was e mployed to carry out the analyses. This method was used for all mediation analyses. Possible c ausal inferences are discussed tentatively In addition, four multiple regression analyses include d participant gender as the independent variable and normati ve pressure to help victims from each pressure source (i.e., mother, father, friends, teachers) as mediating variables, with defending as the dependent variable, controlling for gender identity. Another multiple regression analysis include d participant gender as the predictor variable and shared gender between the participant and the bully or victim as the mediating variable, with
72 defending as the dependent variable. Shared gender was numerically coded for point biserial calculations so that the two poss ibilities (i.e., participant wa s same gender as bully or victim or participant wa s not the same gender as bully or victim) were regarded as levels for correlational analysis. Still another multiple regression analysis was conducted to examine a potential interaction between normative pressure to help victims and the gender match between the person providing the pressure and the participant (i.e., a variable representing whether the participant and the person providing normative pressure we re in the same g ender in -group). In -group status was numerically coded for point -biserial calculations so that the two possibilities (i.e., gender in group or gender out -group) were regarded as levels for correlational analysis. Evidence of a an interaction between the two variables were to be construed as support ing the prediction, based on SIT s concept of in -group favoritism that perceived normative pressure from in -group (i.e., same gender) persons and out -group (i.e., opposite gender ) persons would have different relationships to ones behavior. A similar analysis explore d whether the age (i.e., child, mixed, adult) of a par ticipant s best friends would moderate the relationship between normative pressure and defending. Finally, all possible subsets regression a nalysis was conducted to determine if gender wa s diminished in or excluded from the best predictive combination of the predictor variables and hypothesized interaction term s A significant change in the relationship between biological gender and defending were to be considered an important step toward asserting that social influences may influence defending in concert with or more directly than biological gender. Methodological Limitations Certain limitations were anticipated in the present study. Several were expected from the use of surveys to collect behavioral data. The self report defender scale score may possibly
73 represent not only actual defending behavior, but also a participants likelihood of responding in a socially desirable way. In addition, the scale of normative pressure to help victims measures only a child's perception of the expectations of his or her significant relations. These perceptions may be different from expectations reported by those significant relations, if such data were to be collected. Causal interpretations of results obtained in this study were made in all possible causal directions and implications for practice based on the results were phrased and read with tentativeness, describing alternative explanations for sign ificant findings. Research that manipulates gender identity salience and a childs perception of normative pressure to help victims is needed for causal conclusions. It wa s considered that this correlational study could serve as a springboard for such ca usal research. It wa s expected that random selection of participants would not be conducted, due to the desirability of achieving an adequate sample size with a possibly low return rate of informed consent by parents. As such, results were probably suscep tible to a bias reflecting differences between families who would complete and return the required informed consent forms and those who would not. Such differences would possibly affect responses to survey items. Possibilities such as this were to be con sider ed when interpreting results of the present study, particularly if an extremely low return rate require d the researcher to settle for a convenience sample.
74 CHAPTER 4 RESULTS This chapter presents results of surveys administered to students in four middle schools in a rural southeastern university county. The dependent variable addressed in the surveys regarded a childs tendency to defend victims of bullying, while the candidate predictor and moderating variables were gender, gender identity, norma tive pressure from parents, friends, and teachers to help victims of bullying, and demographics. Demographics are reported first, followed by reliability and validity information about the measurement instruments, inspection of potential control variables and finally the results of analyses pertinent to the studys research questions. Sample Demographics Participants were 274 students from three public middle schools and one university affiliated research middle school in a rural county in the Southeaster n United States Examination of the difference in betas ( DFB ETAS ), or the difference in the parameter estimate caused by the removal of an observation, indicated no undue influence of outliers, and therefore no participants were excluded from the study. Participants were excluded from a specific analysis if missing data was applicable to that analysis. One hundred sixty-eight participants (61%) were female and 101 (37%) male. Five participants (2%) did not indicate their gender. Eighty -nine students (32%) were in 6th grade, 64 (23%) in 7th, and 121 (44%) in 8th. The mean age of participants was 12.97 years ( SD = .97, Median age = 13.07). Students ages ranged from 10.55 to 14.95 years. Twenty-four participants did not indicate an age. Students were pe rmitted to endorse as many racial -cultural identities as they thought appropriate from a list including African American, White, Asian, Pacific Islander, Hispanic/Latino(a), Native American/American Indian, Alaskan Native, and Other. Because of
75 multiple s elections from several children, counts total greater than the total sample size (N=274). A total of 75 students (27%) identified as African American, 175 (64%) as white, 19 (7%) as Asian, 5 (2%) as Pacific Islander, 45 (16%) as Hispanic/Latino(a), 20 (7% ) as Native American/American Indian, 3 (1%) as Alaskan Native, and 31 (11%) as Other. The numbers of children choosing a single identifying category were 51 (19%) African American, 130 (47%) White, 8 (3%) Asian, 1 (.4%) Pacific Islander, 20 (7%) Hispanic /Latino(a), 1 (.4%) Alaskan Native, and 2 (1%) Other. No participant identified solely as Native American/American Indian. Fifty -five (20%) participants selected multiple racial cultural identities. Six students (2%) did not indicate a racial -cultural i dentity. The sample makeup seems to resemble the population of the county sampled with the important exception of gender. The U.S. Census Bureau (2008a) estimates the proportion of females and males aged 10 to 14 in the county sampled to be 48% and 52% respectively. Analysis of U.S. Census Bureau (2008a) data provides the following percentages for racial/categories (i.e., either alone or in combination with others) for children in the sampled county between ages 10 and 14 as of July 2007 indicate to be 33% Black or African American, 64% White, 5% Asian, .08% Native Hawaiian or Pacific Islander, 7% Hispanic, and 1% American Indian or Alaskan Native. Those selecting themselves to be in a single category were 31% Black/African American Alone, 62% White Al one, and 3% Asian Alone, excluding groups comprising less than 1% of the population. For all ages, 6.2% of the population was some other [single] race (U.S. Census Bureau, 2008b). Reliability and Validity of Instruments For each instrument, the assessme nt of reliability and validity were required. One reason for these analyses wa s the absence of certain analyses in past uses of some of the instruments. Another reason wa s that the use of an instrument for the particular aims of any study is likely to
76 be unique in some way. Knowledge of instrument psychometrics for a particular use allows for informed discussion of results. Therefore, internal reliability and factor structure was assessed for each variable in this study that was measured by a scale. Internal consistency was calculated for the Self -Report Defender Scale, Normative Pressure to Help Victims, and the Childrens Personal Attributes Questionnaire Feminine Masculine Scale (CPAQ FM). Factor analysis was also conducted to explore whether instr uments represented a single dimension where appropriate (i.e., Defender Scale, CPAQ -FM) and several dimensions where appropriate (i.e., Normative Pressure). Factor analyses have not been conducted in prior studies either on the Self -Report Defender Scale or on the measure of Normative Pressure used in this study. Defender Scale Cronbachs coefficient alpha for internal consistency was computed for the self report defender scale using SAS 9.1 statistical analysis software, the software used for all analys es in this study. Internal consistency for standardized Defender Scale scores was sufficient, = .692. In addition, removing any item from the scale appeared to reduce internal consistency or not increase it. Principal factor analysis for the scale suggested that it is appropriate to interpret the scale as representing a single factor. This sup port for the scales unidimensionality is tentative due to the sample size required for a reliable factor analysis. Childrens Personal Attributes Questionnaire Feminine -Masculine Scale Internal consistency for the CPAQ -FM was low, = .419. Principal fac tor analysis was also run on the Childrens Personal Attributes Questionnaire Feminine -Masculine scale (CPAQ FM) to investigate whether the scale appears to represent a single construct when used on the current sample. This analysis was conducted both for the complete 13 item scale and for the more internally consistent 9 -item scale. Using the proportion criterion, three factors were
77 retained for the complete scale. The retention of multiple factors did not support the unidimensionality of the CPAQ -FM. A nalyses of the effect of deleting each scale item indicated that deleting four variables (i.e., I believe very strongly in God, I am interested in what goes on outside my home and in the world, I like math and science a lot, and When something very bad happens, I get very upset and forget what is the best thing to do.) would result in a scale with a higher internal consistency, = .559. As in the full scale, principal factor analysis did not support the unidimensionality of the CPAQ -FM with items deleted, as two factors were retained using the proportion criterion. The short CPAQ -FM as described by Hall and Halberstadt (1980) r esembles the present studys modified shorter form tested above, but with additional items removed: It is very important to me to have my parents or other grownups take care of me so nothing bad can happen to me, It is very important to me that people t hink that I am good, I like to take charge of things, and I am very good at getting my way with my friends. For the present study, the internal consistency, = .452, for Hall and Halberstadts (1980) short form was higher than that of the long form, but lower than that of this studys modified form. Principal factor analysis indicated the retention of a single factor, supporting the unidimensionality of Hall and Halbertstadts (1980) short CPAQ -FM. This unidimensionality of construct was not considered necessary for the purposes of the present study, as the scale was intended to measure adherence to stereotypes associated differentially with femininity and m asculinity, even if those stereotypes encompass more than one construct. Sufficient i nternal consistency is a prerequisite to discussions of validity. An internally inconsistent scale cannot be said to measure a particular construct. Therefore, the decis ion was
78 made to employ the modified form of the CPAQ -FM because it had the highest internal consistency. In selecting the form of the CPAQ -FM to be used in analyses, issues of content and criterion validity were considered, as well as issues of internal consistency. It is difficult to argue for the removal or inclusion of any items of a gender identity scale based on content validity alone. For example, an item such as I like to take charge of things, may sound stereotypically masculine in some areas and eras. However, in a 21st century university -centered county, in which examples of female leadership are plentiful, such an item may lose its connection to gender. This example may be arguable, though, because many people may experience biases in fav or of male leadership, even in the community described. The content effects of removing items from a scale published in 1980, therefore, were not clear at least in terms of how the items related to gender identity directly. However, Chapter 5 contains a detailed description of factors represented by items in all three scales arguing the content equivalence of this studys modified scale to Hall and Halberstadts (1980) two forms Criterion validity of items was explored by correlating forms of the scale with gender, and by correlating individual items with gender. Analyses showed that correlation s between gender and the three forms of the scale do not differ by more than .04 In addition, a scale consisting only of items that significantly predict gend er had the lowest internal consistency of all the scale forms and did not predict gender better than the other scales. Criterion validity, as measured by those analyses was not helpful in determining which version of the scale was most appropriate for us e in the present study. The best evaluation of the effect on construct validity of removing items from the long CPAQ -FM may come from examining the short form created by the developers. As noted above, the items removed to create the modified CPAQ -FM were half of those removed by Hall
79 and Halberstadt (1980) to create the short form of the scale, only more items were removed to create the short scale. The modified scale, therefore, returns four items that were removed from the long CPAQ FM by its authors w hen they created the short CPAQ FM. This similarity to both scales supports the notion that the content of the modified scale fully covers the construct domain conceptualized by the developers. A comparison of factors represented by items in the three scales will be presented in Chapter 5. Normative Pressure to Help Victims An adapted procedure was used to measure a childs normative pressure to help victims of bullying. This modified scale was used to determine a childs perception of the expectations of his parents, friends, and teachers that he or she help victims or help bullies. The original procedure consisted of four one item subscales. The adaptation of the scale provided a sixteen -item scale with four subscales in which the expectations of fo ur people or groups of people (i.e., mother, father, best friends, and favorite teachers) were reported for four bullying situations (i.e., hitting and shoving, name -calling and teasing, ostracizing, and stealing. This adaptation of the original procedure necessitated preliminary reliability and validity analysis of the instrument as used in the study. Principal factor analysis retained five factors using the proportion criterion. Using varimax orthogonal rotation for easier interpretation of factor loadi ngs and searching for factor loadings of .5 or higher allowed for the clear selection of all items into a factor. Factor 1 included items indicating normative pressure from the mother and father in two direct bullying situations (i.e., hitting/shoving and name -calling/teasing). Factor 2 included all items related to normative pressure from favorite teachers. Factor 3 included all items related to normative pressure from best friends. Factor 4 included items indicating normative pressure from both parent s as related to ostracizing. Factor 5 included items indicating normative pressure from both parents as
80 related to stealing. These factors largely resembled the subscales as used in Rigby (2005), Rigby and Johnson (2006), and as intended for use in this study. However, instead of indicating unique expectations from mother, father, friends, and teachers, perceived parental expectations seemed to cluster together. Parental expectations clustered together based on the type of bullying described rather than on parental gender as the researcher expected. Even so, i n the design for the current study, perceived parental expectations were assessed as planned, separately for mother and father in order to address the studys research questions. Internal consiste ncy was examined for the total NPHV score, for each NPHV factor, and NPHV from each parent. Internal consistencies were = .903 for total NPHV, = .872 for NPHV from parents regarding direct bullying, = .855 for NPHV from teachers, = .829 for NPHV from friends, = .871 for NPHV from parents regarding ostracizing, and = .899 for NPHV from parents regarding stealing. NPHV from mother and father were = .8 34 and .8 28 respectively. The factors must be interpreted with caution due to the sample size required for reliable fa ctor analysis. Descriptive Statistics Descriptive data for the dependent and independent varia bles appear in Table 4 1. Possible score ranges were 0 to 2 for defending, 1 to 4 for the CPAQ FM and 4 to 20 for normative pressure Higher scores denote more frequent defending, more masculine gender identity, and a higher normative pressure to defend victims. Table 4 2 shows the frequency of students determined to be in the same age or gender in-group as their best friends, and the same gender in -group as their favorite teachers, as well as the frequency of students assigned to the two bullying situa tions (i.e., same gender as bully and victim, different gender as bully and victim) Table 4 3 displays the percents of students reporting that significant persons in their lives expect them to support victims of bullying.
81 This stud y carried the intent of determining whether pressure from childaged friends explained defending behavior. Some students indicated friendship groups that included only children, some only adults, and some a mix of children and adults. Only three participants indicated having f riendship groups consisting entirely of adults. Removing these three participants did not appear to make a difference in analyses related to friends normative pressure Therefore, it was considered safe to remove these observations from subsequent analy ses related to normative pressure from friends thus ensur ing the inclusion only of participants whose friendship groups had children in them Relationships of Defending Scores to Controls and Predictors Assumptions of homoscedasticity, linearity, and cond itional normality were tested by examining studentized residual plots, studentized residual numerical data, and the nature of the dependent variable. Independence was examined by testing for differences in defending scores among the four different school s and two different school types (i.e., public vs. university affiliated). In addition, tolerance and variance inflation rates were observed to detect multicollinearity of predictors in a model. Unless otherwise stated, assumptions were met for each of t he analyses discussed. No significant differences were found in defending scores either among schools or school types. Defending Score Relationships to Control Variables Control variables of interest to the study were student age, student grade level, and whether or not a bullying situation depicted a bully and victim that was the same gender (i.e., bullying situation in -grouping). Though not an anticipated control variable, a variable of interest for further discussion of the studys generalizability was a students racial cultural identity. As with all dichotomous independent variables in the study, point biserial formatting of the variable
82 bullying situation in -grouping allowed for inclusion of such dichotomous variables in regression analyses. Zer o -order linear regression was conducted to explore the relationship between the continuous or dichotomous control variables and the independent variable: defending. Analysis of variance was conducted to determine if grade -level differences and differences among racial cultural identities, both individual and combined, occurred in defending scores. No relationship was found between defending and age Nor were defending scores found to relate to the gender sameness of the participant to the bully and vic tim in scenarios participants were asked to recall No differences in defending scores emerged among grade levels. Therefore, the anticipated inclusion of age, grade level, and bullying situation in-grouping as control variables in analyses was not carri ed out. Individual racial -cultural identities were analyzed both a s dichotomous identities (e.g., African -American vs. not African -American) and in comparison to other racial -cultural identities (e.g., African -American vs. White). Membership in any raci al -cultural identity did not predict defending scores, nor were differences in defending scores found among racial -cultural identities. The implications for the generalizability of the study with regard to racial -cultural identities will be explored in Ch apter 5. Defending Score Relationships to Predictors, Mediators, and Moderators The first of this studys research question asks, Does gender, gender identity, or normative pressure to help victims (i.e., from mother, father, friends, or teacher) predict defending behavior? Zero -order regression analyses were conducted to explore the relationships between defending and gender, normative pressure, and gender identity. For point biserial formatting, female participant gender was coded 0 and male gender wa s coded 1. A significant negative relationship was found between defending and being male, r (266) = -.13, p <
83 .05. A significant negative relationship was found between defending scores and masculine gender identity using the modified CPAQ -FM, r (250) = .13, p < .05. Significant positive relationships were found between defending and normative pressure from mother, r (266) = .29, p < .0001, from father, r (260) = .32, p < .0001, from best friend(s), r (260) = .44, p < .0001, and from favorite teacher(s), r (265) = .24, p < .0001. Please see Table 4 4 for a summary of these relationships. These analyses support an affirmative response to research question one such that defending was more likely in females than in males, in children with more feminine gender identity, and in children who believed that their mothers, fathers, best friends, and favorite teachers expected them to support victims of bullying. Mediation Analyses Evidence of mediation occurs when three conditions are met: first, evidence of a signi ficant relationship between the predictor and presumed mediator must exist; second, a significant relationship between the presumed mediator and the dependent variable must be evident when the predictor is controlled for ; third, a significant zero order r elationship between the predictor and dependent variable must become non-significant when controlling for the mediator (Baron & Kenny, 1986). M ediation analyses were carried out with defending as the dependent variable, gender as the predictor, and, each in turn, gender identity and normative pressure as mediators. The exclusion of observations due to missing data in responses from the CPAQ -FM normative pressure from mother, and normative pressure from best friends scales resulted in the disappearance o f the significant relationship between gender and defending. In order to restore this significant relationship to view and so permit the testing of proposed mediators missing responses from items in those scales were replaced by the average response for each item, and observations were not excluded.
84 Th is studys second research question asks, Does gender identity mediate the relationship between gender and defending behavior? Figure 4 1 shows that, though the other two conditions are met, gender identity does not significantly predict defending when controlling for gender Therefore, support for research question two cannot be inferred. The third research question asks, Does normative pressure to help victims (i.e., from mother, father, best fri ends, or favorite teacher s ) mediate the relationship between gender and defending behavior? As shown in Figure 4 2 only normative pressure from best friends meets all three of Baron and Kennys (1986) criteria for a mediating variable Limitations emer ge because causal inferences are not applicable. However, the finding constitutes tentative support for the prediction that normative pressure from friends mediates the relationship between gender and defending behavior Moderation Analyses Several modera tion analyses were planned in order to explore research questions and to provide inferential results that help compare results of this study to data reported in past studies. Results related directly to the present studys research questions are presented first, followed by those which are used for purposes of comparison to other research. The fourth research question asks, Does gender in -group pressure differ from gender out group pressure in its relationship to defending behavior? This fourth research question requires the testing of gender in-grouping as a moderator between normative pressure and defending behavior. Baron and Kenny (1986) explain that moderation is tested by regressing a dependent variable (e.g., defending) on a predictor (e.g., norm ative pressure), a moderator (e.g., gender ingrouping), and an interaction between predictor and moderator (e.g., normative pressure X gender in-grouping). In order to infer a moderating influence, the interaction term must significantly predict the depe ndent variable and the inference is clearer if the moderator
85 correlates with neither the predictor nor the dependent variable (Baron & Kenny, 1986). Gender in -grouping was conceived for this study as a continuous variable, ranging from low gender ingrouping to high gender in -grouping. That is, a students valued friendship group might consist entirely of persons of his or her gender, or partially, or not at all. This concept only matter ed where mixed gender in -grouping wa s possible (i.e., friends, teachers). Regressing defending scores on normative pressure, gender in-grouping, and an interaction of the two revealed n o moderating effect of gender in-grouping for normative pressure from mother, father, best friends, or favorite teachers in predicting def ender scores. The fifth research question asks, Does the presence of adults in a friendship group moderate the relationship between normative pressure and defending? No evidence for a moderating effect of best friends age emerged in the relationship be tween defending and normative pressure from mother, father, best friends, or favorite teachers A moderating effect was found in planned interaction analys e s that w ere carried out for comparison to previous findings and not necessarily for application to social identity theory. The relationship s between defending and normative pressure from mother, father, best friends, and favorite teachers w ere examined to see if any of them varied as a function of participant gender. A significant interaction was indicated, though not meeting Baron and Kennys (1986) criteria for added clarity for friend pressure and teacher pressure. Friend pressure predicted defending more strongly in boys r (96) = .62, p < .0001 than in girls r (159) = .30 p < .0 001. Teacher pre ssure did not predict defending for girls, but did so for boys r (96) = .35, p < .0 01. Figure 4 3 depicts interactions between gender and normative pressure. The same moderation analyses were conducted with participant age as the moderator normative pr essure from all measured sources as predictor variables, and defending as the
86 dependent variable A significant moderating effect emerged again not especially markedly clear according to Baron and Kenny (1986), for pressure from friends such that the rel ationship between defending and friend pressure was higher for children above the samples mean age, r (129) = .47, p < .0001 than for children at or below the mean age, r (109) = .35, p < .001. Please see Figure 4 4 for a graph of the interaction between a ge and normative pressure from friends. Combining Variables in Predicting Defending The sixth research question asks, When all independent variables are combined to predict defending behavior, is the predictive value of participant gender changed? Becau se this analysis contained many variables, many observations had missing data. T he number of observations removed for missing data was large enough to make the planned analysis invalid for exploring this research question, because the relationship between gender and defending was not significant prior to adding in the remaining variables. Replacing a large amount of missing data also appeared an unsound choice. The planned regression analysis of all possible subsets s uggested that the combination of 1 6 of the predictors yield ed the largest adjusted squared multiple correlation coefficient available for the studys variables. G ender was included among the 16 retained predictors. However, multicollinearity was evident in the high variance inflation facto rs of several variables: gender, father pressure, teacher pressure, and almost all interactions. With the exception of gender, variables were removed one at a time in decreasing order of variance inflation rate until a model with no evident multicollinear ity emerged. Because all possible subsets regression selects variables at a more liberal significance level than .05, v ariables left in each model were tested for significant contribution to the explanation of the variance in defending scores, using a pro bability cutoff of .05 with a Bonferroni adjustment for family-wise comparisons. The final model contained one significant predictor the term representing an interaction between age and
87 normative pressure from friends Table 4 5 shows results from each successive model. This process did not select a better model for predicting defending scores than the single variable representing normative pressure from best friends Summary This chapter presented the findings of single and multiple regression analyses to answer six research questions concerning the relationship between childrens defending behavior and their gender, gender identity, and normative pressure to defend from significant persons. In one analysis the central relationship, that between gende r and defending, was not significant before adding mediators, so that conclusions regarding the mediating power of the added predictors could not be made For the remaining questions, tentative findings were reported Chapter 5 presents a discussion of t hese results, including limitations and suggestions for future research. In addition, implications for counseling theory and practice are discussed.
88 Table 4 1. Descriptive statistics for dependent and independent variables. Variable Na Mean Std. d ev iation Range Low High Defending (dependent) Girls Boys 272 166 101 1.09 1.13 1.02 .42 .40 .45 0 .00 0 .00 0 .00 2 .00 2 .00 2 .00 Normative pressure from mother Girls Boys 26 9 165 100 16.87 16.99 16.67 2.4 5 2.4 5 2.4 7 9 .00 9 .00 12 .00 20 .00 20.00 20 .00 Normative pressure from father Girls Boys 26 3 1 16 98 16.5 8 16.77 16.28 2.53 2.53 2.54 9 .00 9 .00 12 .00 20 .00 20.00 20 .00 Normative pressure from friends Girls Boys 265 163 98 15.03 15.51 14.26 2. 79 2.73 2.68 8 .00 9 .00 8 .00 20 .00 20.00 20 .00 Normative pressure from teachers Girls Boys 26 8 167 97 17. 8 5 17.91 17.71 2.5 4 2.5 8 2.51 4 .00 4 .00 8 .00 20 .00 20.00 20 .00 CPAQ FM long form Girls Boys 252 149 99 2.51 2.41 2.65 .35 .33 .34 1.62 1.62 1.77 3.46 3.23 3.46 CPAQ FM modified (this study) Girls Boys 252 149 99 2.52 2.40 2.70 .47 .44 .45 1.33 1.33 1.67 3.56 3.56 3.56 CPAQ FM Short Form Girls Boys 252 1 49 99 2.68 2.52 2.93 .55 .50 .53 1.40 1. 40 1.60 4.00 3.80 4.00 a Differences in sample size indicate exclusion of observations with missing variables. Sample sizes broken down by gender do not sum to equal sample sizes not broken down by gender.
89 Table 4 2. Frequencies for best friends age, best friends gender (in-group vs. out -group), favorite teachers gender (in group vs. out group), and assignment of bullying gender scenario (in -group vs. out group, rotated forms given to the sample). Variable Frequency Percent Best Friends age Mostly children or teens (not mostly adults) An even mix of children and adults Mostly adults (18 or over) (Missing data) 180 82 3 (9) 66 30 1 (3) Best Friends Gender Same gender as participant (in -group) Mixed Different gender as participant (out -group) (Missing data) 107 151 9 (7) 39 55 3 (3) Favorite Teachers Gender Same gender as participant (in -group) Mixed Different gender as participant (out -group) (Missing data) 73 156 37 (8) 27 57 14 (3) Assignment of bullying gender scenario (rotated) Same gender as bully and victim (in-group) Different gender as bully and victim (out -group) ( Missing data) 134 135 (5) 49 49 (2)
90 Table 4 3. Percents of stude nts reporting parents, best friends, and favorite teachers as expecting them to help victims in various bullying situations, by participant gender. Mother Father Friends Teachers Girls Boys 92 91 90 88 85 66 95 92 Table 4 4. Pearson product -moment co rrelations between predictors and defending scores (two tailed tests). Variable r df Biological gender (0 = female, 1 = male) .13 266 Gender i dentity (CPAQ FM modified form ) Girls Boys .13 .05 ns .17 ns 250 147 98 Norma tive pressure from mother Girls Boys .29** .22 .38*** 266 162 99 Normative pressure from father Girls Boys .32** .31*** .32 ** 260 158 97 Normative pressure from best friends Girls Boys .44** .30**** .63**** 260 159 96 Normative pressure from favorite teachers Girls Boys .24** .15 ns .36 *** .265 164 96 p < .05, ** p < .01, *** p < .001, **** p < .0001 ns = not significant
91 Table 4 5 Semi -partial c orrelations to defending behavi or for variables in b est initial model by R-square selection followed by subsequent models created by successively disqualif ying predictors for high multicollinearity (VIF) Predictor T erms Initial M odel Model 2 Model 3 Model 4 Model 5 No high VIF Gender 16 .15 16 18* 18 01 Age a Bullying /victim in group 07 .08 09 09 08 07 Bully/victim gender a Mother pressure a Father pressure 12 .17 11 11 15 15 Friend pressure a Teacher pressure 15 06 08 10 10 16 CPAQ 08 .08 09 09 09 09 Mother gender in group a Father gender in group a Friend gender in group a Teacher gender in group a African American 13 .14 13 13 13 11 White a Asian 11 10 08 08 08 07 Pacific Islander a Hispanic a Native American 09 09 08 07 07 06 Alaskan Native a Other race/culture a Mother pressure X gender I n group a Father pressure X gender I n group a Friend pressu re X gender I n group a Teacher pressure X gender in G roup 07 05 05 05 05 07 Gender X mother pressure 14 15 05 06 a Gender X father pressure 16 15 a Gender X friend pressure 19 18 15 15 19 a Gender X teache r pressure 08 04 02 a Age X mother pressure a Age X father pressure 14 a Age X friend pressure 13 11 14 13 13 28 Age X teacher pressure 15 08 10 10 10 16 Model adjusted R 2 .2 5 23 .2 1 .21 .21 18 All models had a sig nificant adjusted R 2 *Significant correlation after Bonferroni adjustment when controlling for other variables in model. aMissing correlations indicate variable was removed from model, and all subsequent models.
92 Figure 4 1. Mediation diagram showing zero-order correlations (and semi -partial correlations) with gender as the predictor, gender identity as the mediator, and defending as the dependent variable. N=268. p < .05 **** p < .0 001 Figure 4 2. Mediation diagrams showing zero -order correlations (and semi -partial correlations) for normative pressure from A) Mother, N=266. B) Father, N=258. C) Friends, with substituted data, N=262. D ) Teachers, N=263. p < .05 ** p < .01, ***p < .001, *** p < .0 001 Gender .12* ( .12) .18* .03 A Defending Normative Pressure (Mother) (.18**) Gender .14* ( .13*) .23**** .04 D Defending Normative Pressure (Favorite Teachers) (.23***) Gender .13* ( .10) .32**** .10 B Defending Normative Pressure (Father) (.31****) Gender .13* ( .05) .40**** .20** C Defending Normative Pressure (Best Friends) (.38****) Gender .13* ( .09) .13* .33**** Defending Gender Identity CPAQ -FM (Modified) ( .09)
93 00.5 1 1.5 2 48 12 16 20 Friend Pressure Defending Score Girls Boys 0 0.5 1 1.5 2 4 8 12 16 20 Teacher Pressure Defending Score Girls Boys Figure 4 3. Interaction s between participant gender and normative pressure from A) best f riends and B) favorite teachers predicting defending scores. Abscissa scales begin at four, and not zero, in order to depict the range of possible scores from identified scales. A B
94 0 0.5 1 1.5 2 4 8 12 16 20 Friend Pressure Defending Score 12 y/o 13 y/o 14 y/o Figure 4 4. Interaction in predicting defending scores between age and normative pressure from friends Ab scissa scale begins at four, and not zero, in order to depict the range of possible scores from the identified scale.
95 CHAPTER 5 DISCUSSION As summarized in Chapters 1 and 2, t he relationship between gender and a childs tendency to defend victims of bull ying has been explored in thirteen studies that are known to this author. Nine of those studies find that girls appear more likely than boys to defend. However, no theory has been explored in relation to this apparent link between gender and defending This study tested whether facets of s ocial i dentity t heory (SIT; Tajfel & Turner, 1986) explained the relationship between gender and defending, by examining gender identity and normative pressure to help victims as potential mediating variables A theore tical framework for understanding the relationship between gender and defending will help counselors, teachers, parents, and policymakers to encourage helpful behaviors in both genders. In addition, this researchers attempt to explore the applicability o f SIT to gender differences in defending offers a new theoretical direction for future research. Overview of Study and Discussion of Findings This study surveyed 274 middle school children 168 of whom were girls 101 of whom were boys and 5 of whose gend er was unknown. The proportion of girls to boys in the sample was different than that reported for the studied population by the U.S. Census Bureau (2008a) which estimate d the ratio to be nearly 1:1. The proportion of participants in selected racial cul tural identities was similar to the proportion in the population of the county being sampled. Though participants did not directly represent the population sought in terms of gender, the sample s racial/cultural demographics d id not pose a barrier to the generalizabilit y of the results of the study, although other barriers may exist. Students completed a survey packet including scale s of self reported defending, self reported gender identity, and a childs perception of expectations (i.e., normative press ure) from his mother, father, best friends, and favorite teachers
96 to help victims of bullying versus supporting perpetrators, when witnessing four styles of bullying. Included in the instrument packets were forms that described two varying situations: one in which a boy is bullying another boy, and another in which a girl is bullying a nother girl. Forms were passed out in an alternating manner regardless of the sex of the participant. This pattern of instrumentation distribution result ed in half of stude nts read ing of a bullying situation in whi ch they were gender in grouped with the bully and victim, and the other half being gender out -grouped. In addition, a demographics form asked the age grouping of each students best friends, and the gender com position of each students best friends and favorite teachers. This allowed for the investigation of age and gender in-grouping as moderators in the relationship between normative pressure and defending behavior. Zero -order and multiple regression analys es were carried out to explore all relationships pertinent to research questions. Hypothesized control variables were not used because they did not correlate with the independent variable, defending. In addition, defending did not differ among the four different schools sampled or between school types (i.e., public vs. research), and therefore evidence of violating the assumption of independence was not obtained from those analyses. Furthermore, the dependent variable was continu ous and distributions di d not appear skewed or kurtotic, allowing for more confidence that the assumption of independence had been met. Studentized residual plots appeared to be shapeless, maintaining support for assumptions of linearity, homoscedasticity, and conditional normal ity. In predicting defending behavior, in teractions between normative pressure and gender, and between normative pressure and age were carried out for descriptive purposes and for comparison to results of Rigby and Johnson (2006). Although authors of thi s prior study did not test for interactions, their reports showed higher predictive power of parental pressure among
97 girls than among boys, higher predictive power of friend pressure among boys, and no apparent differences for teacher pressure. The presen t study finds higher predictive power for both friend and teacher pressure among boys than for girls, and no differences for parental pressure In addition, interaction analyses in the present study found that the relationship between defending and friend pressure increased with age, a moderating effect not explored in Rigby and Johnsons (2006) report. Although replication of these results are necessary, it appears that social norms may be influential for boys and older students, and that targeting the s ocial environments of those two groups may improve interventions intended to encourage defending behavior Gender Results indicated a somewhat weak significant negative relationship between defending and gender indicating that girls were more likely to sel f-report defending behaviors than were boys. This relationship was tenuous enough that the removal of observations with missing data for other variables resulted in the apparent loss of a significant relationship between gender and defending. This occurr ed in t hree cases: when removing observations because of missing responses to the CPAQ -FM (i.e., gender identity) scale, and doing the same for the scale s of normative pressure from mothers and normative pressure from best friends. It did not occur when r emoving observations for missing responses to scales of normative pressur e from father or favorite teachers. This represents a conditional corroboration of results from previous studies reporting genders link to defending behavior The seeming change i n the relationship between gender and defending when removing some observations, but not when removing others, may imply a systematic difference that was not explored in this study, especially considering that the relationship was restored when missing res ponses were replaced with the means for those responses and deleted observations were retained S c hafer and Graham (2002) view the treatment of missing data as an opportunity to
98 make inferences about the missingness of data as relates to a particular po pulation rather than to retain observations otherwise deleted. They do, however, acknowledge that mean substitution may accurately estimate missing responses. Importantly, in this study, substituting missing observations made it possible to conduct pl anned analyses to address their corresponding research questions without the apparent alteration of correlation coefficients caused by removing observations The evident fragility of the relationship between gender and defending is not consistent with Salm ivalli and Voetens (2004) finding that gender was the most powerful predictor of defending in comparison to various bullying -related attitudes and norms. However, about 30% of the studies reviewed by this researcher fail to find a relationship between ge nder and defending. The present study is more consistent with the remaining studies that do find a relationship Future research might explore cultural reasons for the differences among studies in the reported strength of relationships between gender and defending, as these studies have been conducted in various countries. In addition, methodological differences vary from study to study, including how defending is measured. Gender Identity The long form of the CPAQ -FM (Hall & Halberstadt, 1980) gender identity scale was originally intended for use in the present study. However, the long form had extremely low internal consistency as indicated in Chapter 4. Removing several items to create this studys modified form increased internal consistency appre ciably, without altering the overall content dimensions covered by Hall and Halberstadts (1980) long and short form s Evidence of the similarity in content between this studys modified form and the original long and short forms comes from two previous studies (Absi -Semaan, Crombie, & Freeman, 2008; Thomson & Zand, 2005), which included factor analys e s o f the long and short forms. All
99 but one of the items removed from the long scale in the present study to make the modified scale were also removed by Ab si Semaan and colleagues (2008) for failing to meet clear factor loading criteria. Moreover, t here were no item content dimensions (i.e., factors) included in Hall and Halberstadts (1980) forms of the scale that were not also represented in the present s tudys modified form. All three forms had items representing low emotionality leadership independence, and interpersonal factors described by Absi -Semaan and colleagues ( 2008) and Thomson and Zand (2005). Furthermore, the items removed from the long form to make the present studys modified form were only a portion of the items removed by the scales authors to make their short version of the scale. All results were obtained using the present studys modified form of the scale. Gender accounted for 11% of the variance of gender identity scores. Males scored as more masculine, and females as more feminine, on the scale. Being female or more feminine predicted more frequent defending. G ender and gender identity, when modeled separately, each ac counted for about 2% of the variance in defending scores. Neither variable predicted defending when both were included in the model. Normative Pressure to Help Victims The title of this variable connotes that high scores on the instrument denote higher normative pressure (i.e., high expectation) to help victims, but fails to communicate that low scores indicate high normative pressure to help bullies, and middling scores convey pressure to do nothing. More pertinent to interpreting correlational results i s that the title of the measure also does not reveal that items in the scales asked students what others would expect them to do not necessarily what others would pressure them to do or what behavior others would exemplify as a norm
100 Normative pressure ex plained more variance in defending scores than did any other variable in the study with all normative pressure sources significantly explaining variance in defending behavior w hen controlling for gender Specifically, m others pressure explain ed 3 %, fat hers pressure explain ed 10%, best friends pressure explained 14%, and favorite teachers pressure explained 5% of the variance in defending scores. Controlling for normative pressure resulted in reducing the predictive value of gender for defending, exc ept in the case of teachers. However, only best friends normative pressure was predicted by gender. Gender accounted for about 4% of the variance in scores on normative pressure from friends. Girls were more likely than boys to report that friends expe cted them to help victims of bullying. Thus, friends normative pressure was the only variable to show evidence of being a mediator. A n attempt to combine all predictor terms in order to predict defending scores resulted in the retention only of a predi ctor term representing an interaction between age and normative pressure from friends, though it did not account for a higher amount of variance in defending scores than did normative pressure from friends alone. Therefore, not only was normative pressure from friends the only variable found to mediate the relationship between gender and defending but it was also the most robust and powerful predictor of defending scores in the study Furthermore, n ormative pressure from best friends was not moderated by the age or gender makeup of friends providing normative pressure, and neither was pressure from parents or teachers. If the above findings hold in future studies it may imply that friends of any gender and age, even adult friends have perhaps more of an influence on middle school students defending behavior than parents or teachers, a childs gender, or a childs gender identity. However, c onsidering another direction of causality reveals that children who defend may be more likely to
101 see others expec tations as pro -defending a prospect not explored in this study It is possible that gender influences defending and that defending then influences friends expectations Even so, the relationship between defending and friend pressure is much stronger th an that between defending and gender. Implications for Theory Although the studys findings lend veracity to the notion that social influences are conspicuously related to childrens tendency to help the victims of bullying, t he studys findings only parti ally support using a gender identity application of social identity theory (SIT; Tajfel & Turner, 1986) to explain gender differences in this helpful behavior. Importantly, gender identity did not predict defending behavior when controlling for gender. O n the other hand gender identity predicted defending behavior as well as gender did and gender did not predict defending behavior when controlling for gender identity. T he studys findings may lend support to SIT s relevan ce in understanding gender diffe rences in defending behavior, if not confirming gender identity as a mediator between sex and defending. G ender sameness of the participant to the mother, father, best friends, or favorite teachers did not moderate the relationship between normative pressu re from those people and defending behavior. This expected moderation was to explore whether pressure from the same gender persons was differentially influential than pressure from different gender persons. Based on SITs principal of social competition, members of an in group should act in concert to better themselves in comparison with a salient out -group, paying more attention to in -group norms than to pressure from out -group persons. However, this reading of SIT ignores the possibility that out group people might pressure a person toward the same behaviors as his or her in -group member s would, in order that the out -group can promote its own distinctiveness. Social competition may not necessarily require that in -group pressure differ in influence from out -group
102 pressure. Conversely, the notion of in-group favoritism does suggest a higher influence from in group members than that from out -group members upon a participant, and so a lack of support for gender applications of social identity is concluded from this finding, even though future clarifications are needed. What is clearer about the studys findings is that normative pressure from any source appeared to be a more powerful predictor of defending than gender, and that friend pressure appeared to m ediate genders influence on defending. Males and females appeare d differentially expected to defend and they also report ed different frequencies of defending behavior. This finding supports the gender identity aspect of SIT in explaining v ariations in defending behavior. Taken together, the studys findings imply that gender identity and normative pressure predict defending behavior, and that friends expectations may mediate the relationship between gender and defending in middle school students These findings parallel those from adult populations in which gender identity moderates gender differences in social dominance orientation (Wilson & Liu, 2003) and that sexism scores mediate the relationship between gender and social dominance orientation (Schm itt et al., 2003) Thus, a gender reading of SIT in understanding defending behavior appears partially supported by the results of this study This conclusion of partial support for gender applications of SIT assumes that normative pressure and gender i dentity caused defending behavior and not the reverse. Furthermore, one would have to assume that self -esteem is gained by responding in conformance to social gender norms in order to infer support for SIT from the results of this study. In addition, the study did not test factors that cause variances in gender identity, except for gender itself. Further possibilities are discussed in the sections concerning the studys limitations and implications for future research, which follow implications for pract ice.
103 Implications for Practice Defending, as defined and measured in this study, not only involves opposition to bullying, but also support to victims; and so p ro -defending interventions in schools may provide more relief than anti -bullying interventions. This study promotes new ways of thinking about promoting defending. Based on th e present stud ys findings regarding applying a gender identity theory to defending behavior, counselors and educators may consider looking at peer pressure and gender stereot yping as areas of intervention. In this study, parents and favorite teachers did not appear to have differential expectations for girls and boys to defend, but favored peers did. According to this studys findings, n ormative pressure from parents, teach ers, and friends may be very influential upon a students defending behavior, regardless of the influen tial persons gender. The outlook for normative pressure to defend appears more favorable in this study than in Rigby and Johnson s (2006) study. In the present study, 66% to 8 5 % of participants reported that their friends expected them to support victims of bullying, whereas the range was 35% to 74% in Rigby & Johnson (2006). Percents of reported expectations from parents, friends, and teachers are sho wn by gender on Table 4 3 for the present study The present studys findings may expand and refine Ahmed and Braithwaite s (2004) conclusion that cooperation between families and schools may best help relieve school bullying problems. D etermin ing which f amily members and school professionals should be solicited to participate in interventions and where the target area for intervention should be may be prerequisites to planning interventions It may be that organized efforts by pro -defending, influentia l parents and teachers should focus on intervening with the closest friends of children assessed as potential defenders of victims Similar to findings about bullying in Salmivalli and colleagues (1997) study of social networks defending may be a frien dship-group supported activity, and those researchers
104 recommendations regarding restructuring peer groups may be seconded by findings of the present study. If promoting defending behavior in schools is an important goal, then it might serve well for scho ol personnel to play a more deliberate role in child friendship group formation. Past interventions have often used peer mentors who befriend troubled students or mediate disputes. These peer -mentoring programs have shown varying levels of success in de creasing bullying behavior. This researcher believes that peer -mentors may be more effective if selected for their higher tendency to defend, a practice that does not appear to have been tried. Furthermore, classrooms, break times, extracurricular clubs and other opportunities to introduce children to each other can be organized to inten tionally disconnect children who bully from those who support their bullying, negate the solitude of victims, and increase interaction between pro -defending and neutral students. Furthermore, friendships in middle schools often begin in earlier grades, and perhaps friendship group formation should begin at that level, and continue throughout the later grades. This is not merely a use of peer mediator s to help students i n disputes and personal dilemmas, but a broader intervention, in which children are intentionally although not rigidly organized in groups that dilute the most negative peer influences. However, gender, gender identity, and gender -based pressure we re the predictors focused on in the present study. Being female and having a more feminine gender identity predicted increase d levels of defending compared to being male and having a more masculine gender identity respectively ; and best friends expect ed girls to defend more and boys to defend less. Considering additionally the finding that pressure from best friends and favorite teachers we re much more related to defending behavior in boys than in girls, it seems possible that interventions intended to bring boys level of defending up to the level of girls may create
105 meaningful improvements in programs designed to increase defending behavior in schools Given these findings parents and teachers might wish to examine and break down gender stereotypes that l ead boys to defend less than girls, and which may lead friends to expect less defending out of boys. More proactively male stereotypes can be reframed to suggest defending as emerging from male characteristics such that boys will come to expect themselves to defend, as will their friends. This may be difficult. For example, it is not easy to conceive of how aggression and dominance stereotypes about males can be reframed to encourage nonaggressive, non -dominating support for victims. In order to do s o, a deeper qualitative analysis of gender stereotypes than could be conducted in this study will be necessary. These implications are made with the intent that practitioners and their allies understand how little is known about defending, and with a recom mendation that counselors, educators, and policymakers begin to become familiar with the research on defending. Furthermore, school pro -defending interventions can also be explorations, using a scientist-practitioner approach to uncover more information a bout defending than existing research provides Limitations of the Study and Implications for Research A discussion of the studys limitations will precede a discussion of i mplications for research, as some research implications flow naturally from the lim itations. This section reviews limitations related to the type of study, availability of data and instrument psychometrics Limitations The current study was a correlational study in which predictors were not manipulated by the researcher. Therefore, di rectional assumptions of causality in mediation diagrams can be misleading. Though gender cannot be said to be caused by any of the predictors of interest, the mediator and independent variable in any of the analyses can be meaningfully reversed in direct ion Furthermore, this study relies on self report for all variables, gender, gender identity,
106 others expectations, and demographics of participants and their significant others. Of primary concern is the inherent confound in asking students to guess at the expectations of their significant others. A persons gender or level of defending behavior could easily influence how a person thinks others expect him or her to behave. Furthermore, the study is cross -sectional, involving a single observation of t he sample as opposed to more than one sampling over a space of time. A cross -sectional design prevents the assessment of stability of the behaviors or measurements, as well as the relationships and treatment effects, relevant to the study. For example, a nalyses of the present study cannot help determine the stability of defending scores, or that of the relationship between defending scores and scores of normative pressure from best friends. In addition, reporting on either ones own defending behavior or others expectations may be subject to a social ly desir able respon se because children may want to make themselves look good in the eyes of researchers or be afraid to be caught reporting negatively on their mothers, fathers, friends, and teachers. Systema tic inaccuracies in reporting the age and gender of ones best friends are also possible and the implications of these upon interpreting the studys results are unknown. The sample size obtained to conduct this study was considered small for purposes of fa ctor analysis of instruments and for analyzing differences among racial/cultural identities, schools and grade levels. In addition, boys were underrepresented in the sample, compared to the population studied. Furthermore, schools, grade levels and sch ool types did not contribute the same number of observations. Sample sizes from the four different schools ranged from 52 to 87 participants. The number of participants from different grade levels ranged from 64 in the seventh grade to 121 in the eighth grade. The number of students from the two types of school s
107 w as 52 from the universityaffiliated research middle school and 222 from the public middle schools. These schools were selected based on the researchers ability to contact representatives of e ach school, which was partly based on the researcher's relationships with persons in the community. Therefore, certain types of school s may have been recruited or volunteered to participate, even though variety w as sought. Unintentional patterns of sel ection may also have occurred at the classroom and individual levels. C lassroom groups were selected partially based on school personnel willingness and availability and the relationships between the researchers school contacts and the teachers solicited for classroom participation. S tudent participation was based somewhat on their willingness to volunteer, parents completion of consent forms, and the students return of completed parental consent forms At least two students complained of having left their completed parental consent forms at home. Students who participated in the study may have been meaningfully different from students who did not. Similarly, students with missing responses to certain items on the questionnaire may be meaningfully di fferent from both students who left other items blank and students who responded to all items completely. For example, 24 students left out the necessary data to compute their a ges, and between one and six responses were missing for all items on the gende r identity measure. Not only do these missing responses raise concerns about the characteristics of students leaving items blank, but also about the clarity of items Internal reliability was not consistent across scales. Cronbachs alphas ranged from .56 for the gender identity scale to .69 for the defending scale to between .83 and .86 for normative pressure subscales. Differences among scales internal reliability may underestimate their relationships to each other, or bias analyses of the contribut ion of multiple predictors. The low
108 internal reliability of the gender identity scale, in particular, makes discussions about its validity difficult. An important assumption of this study is that defending behavior is effective, prosocial, and nonviolent. As such, implications for practice were framed with a pro -defending point of reference. However, defending might sometimes be ineffective, antisocial, or violent. A final limitation of the study comes from the use of the terms defending and defender, which may imply to some the establishment of affinity with victimized students and enmity against those who bully, or at least the protection of the victimized child but not of the child who bullies. The assumption that caregiving by peers is require d for victims but not for children who bully is not substantiated by prior research that describes undesirable consequences of bullying for children in both positions. Implications for Research Opportunities to answer questions that are raised by the prese nt study and its limitations may stimulate further research. For example, causal studies can be designed to manipulate levels of normative pressure from mother, father, best friends, and favorite teachers to help victims. Textual p rompts can place partic ipants in different groups including pressure to support bullying, pressure to do nothing, and pressure to help victims. For example, students can be given instructions that include text reading either Recent research indicates that most students want t heir friends to support victims of bullying, or Recent research indicates that most students want their friends to support bullies. Then group differences could be evaluated for a causal connection to defending score s if random sampling and assignment are conducted. Thus, the directional assumptions of the present studys mediation analyses could be supported or contradicted. Adequate debriefing would be recommended for minors participating in such a study as some will be asked to believe they are expected to support aggressive behaviors
109 Furthermore, causal effects can be assessed better through the use of multiple measurements over time. Therefore, adaptations of the present study that include multiple observations of the same sample, or a subset thereof, can assist in describing the relationship between normative pressure and defending behavior. In addition, repeated measurements will assist in determining the stability of relationships and variables pertinent to the study. For example, the stab ility of defending behavior has been examined in other cultures (e.g., Salmivalli et al., 1998) but not in the United States Studies exploring the susceptibility of the self report defender scale to social desirability bias can help develop a self -report instrument with even better demonstrated reliability and validity to measure actual defending behavior Further efforts to define defending might accompany the refinement of this measure. The more clearly and consistently researchers begin to assess def ending, the more confidence research consumers can be that findings from various studies are related to an agreed upon construct. Use of s uch an instrument may extend to practice implications, aiding in the assessment of schools and individual students, a nd possibly guiding individual counseling and school -wide intervention efforts. The same type of validation effort can be conducted regarding the measurement of normative pressure to help victims, which also seem s vulnerable to socially desirable responding. In addition, a re wording of scale items is needed to clearly reflect the construct alluded to in the title. T he word expect is interpretable as both want or anticipate, and wording should be chosen that reflects others social influence on an individual s behavior, not others predictions of a persons behavior based on what they have seen him or her do in the past (see Appendix D) In point of fact, the normative pressure measurement procedure used in the present study was adapted from a four item procedure created by Rigby (2005) who did not allege it to have the
110 properties of an intentionally developed scale. In a time when researchers recommend the exploration of socioecological contexts of bullying and defending, a valid way to measure s ocial influences on those behaviors could contribute to both research and practice. Replication of the present study is recommended with specific improvements. Prerequisite to replication is further development of defending, gender identity, and normati ve pressure scales with better demonstrated validity, representing widely agreed upon definitions of the phenomenon while maintaining regional applicability. Furthermore, researchers may wish to follow Schafer and Grahams (2002) suggestion to incorporate theoretically -grounded predictions about what missing values might occur in their studies, and what those missing values might imply about their populations of interest Assumptions about defending behaviors effectiveness may be tested by examining the influence of changes in defending scores upon measures of relevant variables such as school safety and academic achievement. Qualitative research is recommended to determine the perceived effects of defending from the point of view of students, teachers, and counselors. Changing the terms defender and defending to a term that does not connote taking sides is recommended, particularly for qualitative studies, but also for the purpose of representing a broader set of behaviors that involve helping vict ims without necessarily opposing children who bully. Obtaining l arge sample sizes also will aid in the assessment of effects of nested populations: individuals within classrooms within schools within systems within the population of interest. Random selection of participants and random assignment to conditions of gender identity, normative pressure, and bullying situations (i.e., gender of bully and victim, type of bullying behavior) are recommended for future studies
111 Summary This chapter presented an o verview of the study, a discussion of the findings, implications for theory and practice, limitations of the study, and finally implications for research. It was concluded by this researcher that gender identity and gender stereotypes, as envisaged by som e readings of social identity theory, could have an effect on childrens helping behaviors toward victims of bullying. Results also suggested that valued friends of any age or sex may play a crucial role in the gender differences apparent in children s de fending behavior, and that parents and favored teachers might be able to play a conscious role in shaping the way friends shape friends behavior. Research directions for the future should focus on cultivating agreement among researchers on the definition and effects of defending in schools, on experimentally testing the causal assumptions of the present study and on creating instruments related to defending, social pressure, and gender identity that can be used by practitioners and educators to improve s chool atmosphere and safety.
112 APPENDIX A PARENTAL CONSENT Department of Counselor Education PO Box 117046 University of Florida Gainesville, FL 32611-7046 Parental Consent Dear Parent/Guardian: My name is James R. Porter. I am a doctoral student in th e Department of Counselor Education at the University of Florida, conducting research on Childrens Responses to School Bullying, under the supervision of Dr. Sondra Smith-Adcock. The purpose of this study is to help determine if social environments help e ncourage or discourage children from helping victims of bullying. The results of the study may help parents, teachers, counselors, and school administrators better understand the kind of social environments that are related to childrens helpful behaviors. These results may not directly help your child today, but may benefit future students and families. With your permission, I would like to ask your child to volunteer for this research. Children will remain anonymous in this study, as will all people in their social environments. Students will be asked anonymously about their responses to bullying in different situations and about their social groups. The questionnaires will be distributed, explained, and collected by me or a trained proxy (teacher, c ounselor, or other qualified volunteer) at your childs school during (added here will be the period agreed upon by researcher and school personnel) The 30 -minute procedure will take place once during the month of September, October, or November. Childr en will not place their names on any questionnaire materials. Questionnaire packets will be number -coded in case forms are separated. Once a questionnaire is collected, it cannot be traced back to the student or to a childs parents, friends, teachers, o r classmates. Results will be reported in the form of group averages and other group data. No individual information will be reported. Participation or non -participation in this study will not affect the childrens grades or placement in any programs. Children will not be required to miss class work or will be permitted to make up any missed work. You and your child have the right to withdraw consent for your child's participation at any time without consequence. Children do not have to answer any que stions they do not wish to answer. There are no known risks or immediate benefits to the participants. Participants will be compensated twice by being permitted to choose from an assortment of thank -you gifts agreed upon by researchers, teachers, school s taff, and administrators. Your child will be able to choose from this assortment first upon handing me your signed consent form, and again upon completing the questionnaire packet. Group results of this study should be available in December upon your req uest. If you have any questions about this research protocol, please contact me or my faculty supervisor, Dr. Smith -Adcock, at 3920731. Questions or concerns about your child's rights as a
113 research participant may be directed to the IRB02 office, Univers ity of Florida, Box 112250, Gainesville, FL 32611, (352) 3920433. James R. Porter, M.Ed., Ed.S., Principal Investigator I have read the procedure described above. I voluntarily give my consent for my child, _________________________________, to pa rticipate in James R. Porter's study of childrens responses to school bullying. I have received a copy of this description. ____________________________ ___________ Parent / Guardian Date ____________________________ ___________ 2nd Parent / Wi tness Date
APPENDIX B STUDENT ASSENT Department of Counselor Education PO Box 117046 University of Florida Gainesville, FL 32611 7046 Student Assent Dear Student, Your parents have agreed to let you participate in this research study if you also volu ntarily agree. This is not a test or a quiz. There are no wrong answers. This is not a part of school and your grades will not be affected whether you agree to participate or not. This is a questionnaire, a series of questions for you to answer anonymously, that will help me to understand peoples behavior during bullying situations. I am very interested in reading your responses. If you agree to participate, please answer questions with complete honesty. PLEASE DO NOT PLACE Y OUR NAME ANYWHERE ON THE QUESTIONNIARE. THAT WAY, YOU CAN BE ABSOLUTEL Y ANONYMOUS AND YOUR ANSWERS CANNOT BE TR ACED TO YOU. Your privacy is very important to me. If later you feel it would help to discuss any feelings that come from participating in this research study, please follow the instructions that I will leave the last page with every participant when I collect the questionnaires. Participation in this study is entirely voluntary. Everyone who completes the questionnaire will receive their choice of (List assortment of compensation items here) just as you received for returning your parents signed permission to us. If at anytime during the questionnaire you feel you no longer wish to participate, you may stop at any time. However, I think that you will find the qu estions interesting and that you will be eager to have your answers counted with everyone elses. If you have trouble understanding any of the items, please raise your hand and I will help clarify the item as best as I can. When you are finished with th e questionnaire, please check over each item on the questionnaire to make sure there are no items left blank unintentionally. After this, please close your packet and turn it facedown so that I will know when everyone is finished. Out of respect for the privacy of your classmates, please do not look at anyone elses desk until I collect all packets. The entire questionnaire takes about 20-minutes or less. If you agree to participate in the study, please sign below and I will give you your questionnaire p acket. Please keep one copy of this letter. James R. Porter, M.Ed., Ed.S., Principal Investigator I __________________________________voluntarily agree to participate in James R. Porter's study of Childrens Responses to School Bullying. I have received a copy of this description. ____________________________ ___________ Signature Date
115 APPENDIX C THE SELF REPORT DEFENDER SCAL E M ALE CONDITION ( ADAPTED FROM SALMIVALLI, 1998; SA LMIVALLI ET AL., 199 8; SALMIVALLI ET AL. 1999) Bullying is one child be ing exposed repeatedly to harassment and attacks from one or several other children; harassment and attacks may be, for example, shoving or hitting the other one, calling him/her names or making jokes about him/her, leaving him/her outside the group, taking his/her things, or any other behavior meant to hurt the other one It is not bullying when two students with equal strength or equal power have a fight, or when someone is occasionally teased, but it is bullying when the feelings of the same student are intentionally and repeatedly hurt. Please keep the above definition in mind when answering questions about your behavior when bullying takes place. THINK BACK TO TIMES WHEN YOU HAVE SEEN ONE BOY BULLYING ANOTHER BOY. Evaluate your own typical behavior in that situation by circling 0 for never, 1 for sometimes, or 2 for often for the following items. Never Sometimes Always 1 Comforts the victim in the bullying situation. 0 1 2 2 Tells others to stop bullying. 0 1 2 3 Says to the others that bullying is stupid. 0 1 2 4 Tries to make the others stop bullying. 0 1 2 5 Comfo rts the victim afterward. 0 1 2 6 Encourages the victim to tell the teacher about the bullying. 0 1 2
116 APPENDIX D THE SELF REPORT DEFENDER SCAL E FEMALE CONDITION ( ADAPTED FROM SALMIVALLI, 1998; SA LMIVALLI ET AL., 1998; SALMIVALLI ET AL., 1999) Bullying is one child being exposed repeatedly to harassment and attacks from one or several other children; harassment and attacks may be, for example, shoving or hitting the other one, calling him/her names or making jokes about him/her, leaving him/her outside the group, taking his/her things, or any other behavior meant to hurt the other one It is not bullying when two students with equal strength or equal power have a fight, or when someone is occasionally teased, but i t is bullying when the feelings of the same student are intentionally and repeatedly hurt. Please keep the above definition in mind when answering questions about your behavior when bullying takes place. THINK BACK TO TIMES WHEN YOU HAVE SEEN ONE GIRL BU LLYING ANOTHER GIRL. Evaluate your own typical behavior in that situation by circling 0 for never, 1 for sometimes, or 2 for often for the following items. Never Sometimes Always 1 Comforts the victim in the bullying situation. 0 1 2 2 Tells others to stop bullying. 0 1 2 3 Says to the others that bullying is stupid. 0 1 2 4 Tries to make the others stop bullying. 0 1 2 5 Comforts the victim afterward. 0 1 2 6 Encourages the victim to tell the teacher about the bullying. 0 1 2
117 APPENDIX E NORMATIVE PRESSURE T O HELP VICTIMS (ADAP TED FROM RIGBY & JOHNSON, 2006) Tell what various people would EXPECT YOU to do when each type of bullying takes place by circling the number that indicates how strongly they would expect you to support the bully or the victim. (Please omit any question about a parent that parent is no longer alive.) SITUATION 1 of 4: Imagine that you are witnessing a bullying situation, and that there are others witnessing it also. You see ONE CHILD SHOVING OR HITTING THE OTHER ONE Strongly Support Do S upport Strongly support the the nothing the support the BULLY bully victim VICTIM 1. My mother would expect me to : 1 2 3 4 5 2. My father would expect me to : 1 2 3 4 5 3. My best friend(s) would expect me to : 1 2 3 4 5 4. My favorite teacher(s) would expect me to : 1 2 3 4 5 SITUATION 2 of 4: Imagine that you are witnessin g a bullying situation, and that there are others witnessing it also. You see ONE CHILD CALLING THE OTHER CHILD NAMES OR MAKING JOKES ABOUT HIM/HER Strongly Support Do Support Strongly support the the nothing the support the BULLY bully victim VICTIM 1. My mother would expect me to : 1 2 3 4 5 2. My father would expect me to : 1 2 3 4 5 3. My best friend(s) would expect me to : 1 2 3 4 5 4. My favorite teacher(s) would expect me to : 1 2 3 4 5
118 Tell what various people would EXPECT YOU to do when each type of bullying takes place by circling the number that indicates how strongly they would expect you to support the bully or the victim. (Please omit any question about a parent that parent is no longer alive.) SITUATION 3 of 4: Imagine that you are witnessing a bullying situation, and that there are others witnessing it also. You see ONE CHILD KEEPING ANOTHER CHILD FROM BEING PART OF A GROUP Strongly Support Do Support Strongly support the the nothing the support the BULLY bully victim VICTIM 1. My mother would expect me to : 1 2 3 4 5 2. My father would expect me to : 1 2 3 4 5 3. My best friend(s) would expect me to : 1 2 3 4 5 4. My f avorite teacher(s) would expect me to : 1 2 3 4 5 SITUATION 4 of 4: Imagine that you are witnessing a bullying situation, and that there are others witnessing it also. You see ONE CHILD TAKING ANOTHER CHILDS THINGS Strongly Support Do Support Strongly support the the nothing the support the BULLY bully victim VICTIM 1. My mother would expect me to : 1 2 3 4 5 2. My father would expect me to : 1 2 3 4 5 3. My best friend(s) would expect me to : 1 2 3 4 5 4. My favorite teacher(s) would expect me to : 1 2 3 4 5
119 APPENDIX F CHILDRENS PERSONAL ATTRIBUTES QUESTIONN AIRE FEMININE MASCULINE SCALE (HAL L & HALBERSTADT, 198 0) Please indicate how true of you each of the following statements is. Circle one number for each item. (Please note that lower numbers mean Very true of me and higher numbers mean Not at all true of me.) Not at A little Mostly Very all true true true true of me of me of me of me me 1 I believe very strongly in God 2 It is hard to hu rt my feelingsa b.... 3 I am oft en very pushy with other peoplea b. 4 It is very important to me to have my parents or other grown ups take care of me so nothing bad can happen to me. a 5 I am very interested in what goes on outside my home and in t he world. 6 I like math and science a lot..... 7 I am a quiet person *a b. 8 It is very important to me that people think I am good*a 9 I like to take charge of thingsa 10. I cry when things upset me *a b 11. I am not good at fixing things or working with tools *a b 12. When something very bad happens, I get very upset and forget what is the best thing to do 13. I am very good at getting my way with my friends. a 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 *Item reverse coded. a Items included in the present studys mod ified form of the scale. bItems included in Hall and Halberstadts (1980) short form of the scale.
120 APPENDIX G DEMOGRAPHICS Please tell us a little about yourself. Answer the following questions honestly by circling the response that is most true or by writing in the answer. 1. Sex: Male Female 2 What year were you born? 3 On what month and day is your birthday? Month _________________________Day______ 4 I am in grade: 6 7 8 5 I identify most with the following group (circle as many as apply to you) : African American White Asian Pacific Islander Hispanic/Latino(a) Native American/American Indian Alaskan Native Other 6 My best friend(s) are (circle one ): Mostly children or teens (not mostly adults) An even mix of children and adults Mostly adults (18 or over) 7 My best friend(s) are (circle one ): Mostly Male Mixed Mostly Female 8 My favorite teacher(s) are (circle one ): Mostly Male Mixed Mostly Female
121 LIST OF REFERENCES Abrams, D., & Hogg, M.,A. (1990). Social identity theory : Constructive and critical adv ances London: Harvester Wheatsheaf. Absi -Semaan, N ., Crombie, G., & Freeman, C. (1993). Masculinity and femininity in middle childhood: Developmental and factor analyses. Sex Roles, 23(3 -4), 187206. Ahmed, E., & Braithwaite, V. (2004). Bullying and vict imization: cause for concern for both families and schools. Social Psychology of Education, 7, 3554. Algina, J., & Olejnik, S. (2003). Sample size tables for correlation analysis with applications in partial correlation and multiple regression analysis. M ultivariate Behavioral Research, 38(3), 309323. Anderson, M., Kaufman, J., Simon, T. R., Barrios, L., Paulozzi, L., Ryan, G., et al. (2001). School associated violent deaths in the united states, 1994 1999. JAMA: Journal of the American Medical Association, 286(21), 2695. Baron, R. M., & Kenny, D. A. (1986). The moderator -mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6), 11731182. Brown, S. L., Birch, D. A., & Kancherla, V. (2005). Bullying perspectives: Experiences, attitudes, and recommendations of 9 to 13-year olds attending health education centers in the united states. Journal of School Health, 75(10), 384392. Camodeca, M., & Goossens, F. A. (2005). Children's opinions on effective strategies to cope with bullying: The importance of bullying role and perspective. Educational Research, 47(1), 93 105. Carney, J. V. (2000). Bullied to death: Perceptions of peer abuse and suicida l behaviour during adolescence. School Psychology International, 21(2), 213. Crick, N. R., Casas, J. F., & Ku, H. (1999). Relational and physical forms of peer victimization in preschool. Developmental Psychology, 35(2), 376385. Dempsey, A. G., & Storch E. A. (2008). Relational victimization: The association between recalled adolescent social experiences and emotional adjustment in early adulthood. Psychology in the Schools, 45(4), 310 322. Ford, T. E., Stevenson, P. R., Wienir, P. L., & Wait, R. F. (2 002). The role of internalization of gender norms in regulating self -evaluations in response to anticipated delinquency. Social Psychology Quarterly, 65(2), 202212. Furlong, M. J., Morrison, G. M., & Greif, J. L. (2003). Reaching an American consensus: Reactions to the special issue on school bullying. School Psychology Review, 32(3), 456470.
122 Gillies Rezo, S., & Bosacki, S. (2003). Invisible bruises: Kindergartners' perceptions of bullying. International Journal of Children's Spirituality, 8(2), 163. G ini, G. (2006). Social cognition and moral cognition in bullying: What's wrong? Aggressive Behavior, 32(6), 528539. Gini, G., Albiero, P., Benelli, B., & Alto, G. (2007). Does empathy predict adolescents' bullying and defending behavior? Aggressive Beha vior, 33(5), 467476. Gini, G., & Pozzoli, T. (2006). The role of masculinity in childrens bullying. Sex Roles, 54 (7), 585588. Goossens, F. A., Olthof, T., & Dekker, P. H. (2006). New participant role scales: Comparison between various criteria for ass igning roles and indications for their validity. Aggressive Behavior, 32(4), 343357. Greene, M. B. (2003). Counseling and climate change as treatment modalities for bullying in school. International Journal for the Advancement of Counselling, 25(4), 293 302. Hall, J. A., & Halberstadt, A. G. (1980). Masculinity and femininity in children: Development of the children's personal attributes questionnaire. Developmental Psychology, 16(4), 270280. Harris Interactive, & GLSEN. (2005). From teasing to torment : School climate in America, A survey of students and teachers New York: GLSEN. Hawkins, D. L., Pepler, D. J., & Craig, W. M. (2001). Naturalistic observations of peer interventions in bullying. Social Development, 10(4), 512527. Kim, Y. S., Leventhal, B. L., Koh, Y., Hubbard, A., & Boyce, W. T. (2006). School bullying and youth violence: Causes or consequences of psychopathologic behavior? Archives of General Psychiatry, 63 (9), 10351041. Laflamme, L., Engstrm, K., Mller, J., Alldahl, M., & Hallqvis t, J. (2002). Bullying in the school environment: An injury risk factor? Acta Psychiatrica Scandinavica, 106, 20 25. Leary, M. R., Kowalski, R. M., Smith, L., & Phillips, S. (2003). Teasing, rejection, and violence: Case studies of the school shootings. A ggressive Behavior, 29(3), 202214. Lorenzi Cioldi, F., & Doise, W. (1990). Levels of analysis and social identity. In D. Abrams, & M. Hogg A. (Eds.), Social identity: Constructive and critical advances (pp. 71 88). London: Harvester Wheatsheaf.
123 Maeda, R (2004). Empathy, emotion regulation, and perspective taking as predictors of children's participation in bullying. (Doctoral Dissertation, ProQuest Information & Learning). Dissertation Abstracts International Section A: Humanities and Social Sciences, 6 4 (11), 39573957. (UMI 3111101) Retrieved from http://search.ebscohost.com.lp.hscl.ufl.edu/login.aspx?direct=true&db=psyh&AN=200499009 050&site=ehost live Marini, Z., Spear, S., & Bombay, K. (1999). Peer victimization in middle childhood: Characteristics, causes, consequences, of school bullying. Brock Education, 9, 32 47. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychological Methods, 7(1), 83 104. McLaughlin, C., Arnold, R., & Boyd, E. (2005). Bystanders in schools: What do they do and what do they think? Factors influencing the behaviour of English students as bystanders. Pastoral Care in Education, 23(2), 17 22. Menesini, E., Codecasa, E., Benelli, B., & Cowie, H. (2003). Enhancing children's responsibility to take action against bullying: Evaluation of a befriending intervention in Italian middle schools. Aggressive Behavior, 29(1), 1 14. Merrell, K. W., Gueldner, B. A., Ross, S. W., & Isava, D. M. (2008). How effective are school bullying intervention programs? A meta analysis of intervention research. School Psychology Quarterly, 23(1), 26 42. Monks, C., Ruiz, R., & Val, E. (2002). Unjustified aggression in preschool. Aggressive Behavior, 28(6), 458476. Monks, C. P., Smith, P. K., & Swettenham, J. (2003). Aggressors, victims, and defenders in presch ool: Peer, self -, and teacher reports. MerrillPalmer Quarterly, 49 (4), 453469. Nansel, T. R., Overpeck, M., Pilla, R. S., Ruan, W. J., Simons Morton, B., & Scheidt, P. (2001). Bullying behaviors among US youth. JAMA: Journal of the American Medical Asso ciation, 285(16), 2094. Ojala, K., & Nesdale, D. (2004). Bullying and social identity: The effects of group norms and distinctiveness threat on attitudes towards bullying. British Journal of Developmental Psychology, 22(1), 19 35. Olweus, D. (1978). Aggr ession in the schools : Bullies and whipping boys Washington; New York: Hemisphere Pub. Corp; distributed solely by Halsted Press. Olweus, D. (2001). Peer harassment: A critical analysis and some important issues. In J. Juvonen, & S. Graham (Eds.), Peer harassment in school: The plight of the vulnerable and victimized (pp. 3 20). New York: The Guilford Press.
124 Paik, H., & Comstock, G. (1994). The effects of television violence on antisocial behavior: A meta analysis. Communication Research, 21(4), 516546. Pepler, D. J., & Craig, W. M. (1995). A peek behind the fence: Naturalistic observations of aggressive children with remote audiovisual recording. Developmental Psychology, 31(4), 548. Rigby, K. (2005). Why do some children bully at school? The contrib utions of negative attitudes towards victims and the perceived expectations of friends, parents and teachers. School Psychology International, 26(2), 147 161. Rigby, K., & Johnson, B. (2006). Expressed readiness of Australian schoolchildren to act as byst anders in support of children who are being bullied. Educational Psychology, 26(3), 425440. Rigby, K., & Slee, P. (1999). Suicidal ideation among adolescent school children, involvement in bully victim problems, and perceived social support. Suicide & Li fe -Threatening Behavior, 29(2), 119. Rimal, R. N., Lapinski, M. K., Cook, R. J., & Real, K. (2005). Moving toward a theory of normative influences: How perceived benefits and similarity moderate the impact of descriptive norms on behaviors. Journal of Health Communication, 10(5), 433450. Ringold, D. J. (2002). Boomerang effects in response to public health interventions: Some unintended consequences in the alcoholic beverage market. Journal of Consumer Policy, 25(1), 27 63. Rodkin, P. C., & Hodges, E. V E. (2003). Bullies and victims in the peer ecology: Four questions for psychological and school professionals. School Psychology Review, 32(3), 384400. Salmivalli, C. (1998). Intelligent, attractive, well behaving, unhappy: The structure of adolescents self -concept and its relations to their social behavior. Journal of Research on Adolescence (Lawrence Erlbaum), 8 (3), 333354. Salmivalli, C. (1999). Participant role approach to school bullying: Implications for interventions. Journal of Adolescence, 22 (4), 453. Salmivalli, C. (2001). Peer -led intervention campaign against school bullying: Who considered it useful, who benefited? Educational Research, 43(3), 263278. Salmivalli, C., Huttunen, A., & Lagerspetz, K. M. J. (1997). Peer networks and bullyi ng in schools. Scandinavian Journal of Psychology, 38(4), 305312. Salmivalli, C., Kaukiainen, A., Kaistaniemi, L., & Lagerspetz, K. M. J. (1999). Self -evaluated self -esteem, peer -evaluated self -esteem, and defensive egotism as predictors of adolescents' participation in bullying situations. Personality and Social Psychology Bulletin, 25 (10), 12681278.
125 Salmivalli, C., Kaukiainen, A., & Voeten, M. (2005). Anti -bullying intervention: Implementation and outcome. British Journal of Educational Psychology, 75(3), 465487. Salmivalli, C., Lagerspetz, K., Bjrkqvist, K., sterman, K., & Kaukiainen, A. (1996). Bullying as a group process: Participant roles and their relations to social status within the group. Aggressive Behavior, 22(1), 1 15. Salmivalli, C., Lappalainen, M., & Lagerspetz, K. M. J. (1998). Stability and change of behavior in connection with bullying in schools. Aggressive Behavior, 24(3), 205218. Salmivalli, C., & Voeten, M. (2004). Connections between attitudes, group norms, and behaviour in bullying situations. International Journal of Behavioral Development, 28(3), 246258. Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7(2) 147177. Schmitt, M. T., Branscombe, N. R., & Kappen D. M. (2003). Attitudes toward groupbased inequality: Social dominance or social identity? British Journal of Social Psychology, 42(2), 161. Schmitt, N. (1996). Uses and abuses of coefficient alpha. Psychological Assessment, 8(4), 350353. Sentse, M. Scholte, R., Salmivalli, C., & Voeten, M. (2007). Person Group dissimilarity in involvement in bullying and its relation with social status. Journal of Abnormal Child Psychology, 35(6), 10091019. Sidanius, J., & Pratto, F. (1999). Social dominance : An intergroup theory of social hierarchy and oppression. Cambridge, UK ; New York: Cambridge University Press. Smith, J. D., Schneider, B. H., Smith, P. K., & Ananiadou, K. (2004). The effectiveness of whole -school antibullying programs: A synthesis of eval uation research. School Psychology Review, 33 (4), 547560. Smith, P. K., & Brain, P. (2000). Bullying in schools: Lessons from two decades of research. Aggressive Behavior, 26(1), 1 9. Sutton, J., & Smith, P. K. (1999). Bullying as a group process: An ad aptation of the participant role approach. Aggressive Behavior, 25(2), 97 111. Taborga, M. P., Cheong, J. & MacKinnon, D. P. (2000). Technical assistance report: Mediation analysis. Retrieved June 10, 2008, from http://www.public.asu.edu/~davidpm/ripl/Mediation_Analysis.PDF Tajfel, H., & Turner, J. C. (1986). The social identity of intergroup behavior. In S. Worchel, & W. G. Austin (Eds.), Psychology of intergroup relations (2n d ed., pp. 7 24). Chicago: Nelson -Hall.
126 Tani, F., Greenman, P. S., Schneider, B. H., & Fregoso, M. (2003). Bullying and the big five. School Psychology International, 24(2), 131. Tersahjo, T., & Salmivalli, C. (2003). "She is not actually bullied." the discourse of harassment in student groups. Aggressive Behavior, 29(2), 134 154. Thomson, N., & Zand, D. (2005). Analysis of the Childrens personal attributes questionnaire (short form) with a sample of African American adolescents. Sex Roles, 52 (3), 237244. Turner, J. C., & Reynolds, K. J. (2003). Why social dominance theory has been falsified. British Journal of Social Psychology, 42, 199206. Twemlow, S. W., Fonagy, P., Sacco, F. C., & Brethour Jr., J. R. (2006). Teachers who bully students: A hidden trauma. International Journal of Social Psychiatry, 52(3), 187198. U.S. Census Bureau. (2008a August 6). County estimates by age, sex, race, and Hispanic Origin [Data file]. Retrieved December 18, 2008, from http://www.census.gov/popest/counties/asrh/files/cc -est2007alldata 12.csv U.S. Census Bureau. (2008b). American FactFinder [Data access service]. Retrieved January 22, 2009 from http://factfinder.census.gov/home/saff/main.html?_lang=en Violence Institute of New Jersey. (2007). Searchable inventory of instruments assessing violent behavior and related constructs in children and ad olescents. Retrieved June 10, 2008, from http://vinst.umdnj.edu/VAID/TestReport.asp?Code=PRQ Vossekuil, B., Fein, R. A., Reddy, M., Borum, R. & Modzeleski, W. (2002). The final report and findings of the safe school initiative: Implications for the prevention of school attacks in the united states. Retrieved June 10, 2008, from http://opi.mt.gov/pdf/safeschools/Prevent AttacksReprt.pdf Wilson, M. S., & Liu, J. H. (2003). Social dominance orientation and gender: The moderating role of gender identity. British Journal of Social Psychology, 42(2), 187.
127 BIOGRAPHICAL SKETCH James R. Porter received his Doctor of Philosop hy from the University of Florida in the s pring of 2009. He obtained both a Master of Education and a Specialist in Education degree in mental health counseling from the University of Florida. He earned Bachelor of the Arts degree s in t heater and p sychol ogy from the University of South Florida. He took a break from his time as a student to work as an actor in the Tampa Bay area and to serve a term of enlistment in the United States Marine Corp. James thrives on face to -face counseling, believing that t herapy is a collaborative effort in which both client and counselor have much to offer James also enjoys statistical analysis, and hopes to expand his career in dual directions that involve direct counseling and rigorous research related to peace buildin g and conflict resolution. James lives in Gainesville, Florida with his wife and daughter.