RESEARCH ARTICLE

Demographic Differences in the Prevalence, Co-Occurrence, and Correlates of Adolescent Bullying at School

KELLIE E. CARLYLE, PhD, MPHa KENNETH J. STEINMAN, PhD, MPHb

ABSTRACT
BACKGROUND: Despite a large literature on bullying, few studies simultaneously examine different dimensions of the phenomenon or consider how they vary by demographic characteristics. As a result, research ndings in this area have been inconsistent. This article focuses on 2 dimensions of bullying behaviorsaggression and victimizationand examines demographic variation in their prevalence, co-occurrence, and association with other health outcomes. METHODS: School-based surveys were administered to a census of 6th-12th graders in 16 school districts across a large metropolitan area in the United States (n = 79,492). A 2-factor scale assessed repeated experiences with bullying aggression and victimization. RESULTS: Both dimensions of bullying tended to be more common among younger, male, African American and Native American students. There were, however, several exceptions as well as considerable variation in the magnitude of demographic differences. Most youth involved with bullying were either perpetrators or victims, but not both. For example, only 7.4% of all youths were classied as bully/victims. Substance use was more strongly associated with aggression, whereas depressive affect was more strongly associated with victimization. CONCLUSIONS: Researchers should distinguish different dimensions of bullying and consider how they vary by demographic characteristics. In particular, repeated aggression and victimization largely involve different students and may require distinct approaches to prevention. Keywords: research; bullying; violence.
Citation: Carlyle KE, Steinman KJ. Demographic differences in the prevalence, co-occurrence, and correlates of adolescent bullying at school. J Sch Health. 2007; 77: 623-629.

Assistant Professor, (kcarlyle@asu.edu), Hugh Downs School of Human Communication, Arizona State University, PO Box 871205, Tempe, AZ 85287. Assistant Professor, (ksteinman@sph.osu.edu), School of Public Health, The Ohio State University, 438 Cunz Hall, 1841 Millikin Rd, Columbus, OH 43210.

Address correspondence to: Kellie Carlyle, (kcarlyle@asu.edu), Hugh Downs School of Human Communication, Arizona State University, PO Box 871205, Tempe, AZ 85287.

Journal of School Health

November 2007, Vol. 77, No. 9

2007, American School Health Association

623

ew studies in the United States have had samples large and diverse enough to systematically examine demographic differences in different dimensions of adolescent bullying. The term bullying includes a range of behaviors that are repeated over time1 (eg, hitting, excluding from group activities, and spreading rumors). Adolescents can experience bullying as a perpetrator and/or victim. This study examines the prevalence, co-occurrence, and correlates of different dimensions of bullying and explores how they may vary by age, gender, and ethnicity. By identifying inconsistent ndings in the research literature and testing them on a large, diverse data set, this article will help guide future efforts to understand the universal and particular aspects of adolescent bullying. DEMOGRAPHIC DIFFERENCES Age Research has fairly consistently indicated that bullying decreases over time for middle and high school students.2-5 This downward trend in bullying is supported by similar national trends in physical ghting.6 Still, some studies have found that bullying and victimization increase with age.7 One explanation is that, while bullying does tend to decrease in general from early to late adolescence, the prevalence rates temporarily peak in the middle school. Thus, increases may be found based on the grade levels chosen for comparison. Consistent with previous research, we hypothesize that the prevalence and co-occurrence of both bullying and victimization will decrease overall with age, regardless of sex, or ethnicity. Gender Previous research suggests that bullying is more common among males than females.8-10 However, numerous studies have found no gender differences, and some suggest that results may be inuenced by gender role stereotypes and how aggression itself is measured (see Underwood et al11 for a review). Nonetheless, the general trends in male and female bullying behaviors are reasonably well supported. As such, we expect that males will bully and be victimized more than females. Ethnicity Whereas physical ghting appears to be more common among African American (39.7%) and Hispanic (36.1%) than white (30.5%) high school students,6 ethnic group differences in bullying are less well established. Notably, few studies have had samples large enough to include comparisons of more than 2 ethnic groups. Some research has suggested
624
d

that the bullying-ethnicity relationship may be more dependent upon specic racial dynamics1dynamics that may be school or community specic and not necessarily apply to ethnic groups as an aggregate. Given the sparse number of studies on the topic, we assume a null hypothesis: there are no ethnic group differences. Correlates Adolescents victimized by bullying can experience sociopsychological harm12,13 including higher levels of depression.2 Bullying is also associated with other problem behaviors in general,7 such as substance use.14 It is not clear, however, how these associations vary across different dimensions of bullying behaviors. The co-occurrence of aggressive behavior and substance use, for example, may both reect a broad personal orientation toward antisocial behavior15 or could be 1 way that adolescents cope with victimization and peer rejection.16,17 Guided by the limited literature in this area, we hypothesize that perpetration will correlate with substance use and that victimization will correlate with depression. Accordingly, this article aims to ll a gap in the literature by examining a large, ethnically diverse data set with consistent measures and a systematic statistical approach. Doing so may help reconcile inconsistencies in previous research and guide future research efforts. METHOD Data were collected using the Primary Prevention Awareness Attitude and Use Survey18 (PPAAUS). The PPAAUS was developed by the Safe and Drug Free Schools Consortium of Franklin County, Ohio, to assess adolescent risk behaviors and their determinants to guide public health policy in the county. The PPAAUS has been administered to all 6th-12th graders in Franklin County (metropolitan Columbus), Ohio, every 3 years since 1988. Data for the present study are based on surveys administered during fall 2003. Trained teachers and school staff administered the PPAAUS during the second class period in 16 public school districts (55 high schools and 91 middle schools), 6 private schools, and 36 Catholic schools. Passive parental consent was used, and students were given the option of not participating in the survey. Students completed the surveys anonymously, with the only identiers being the respondents school building and grade. A readability analysis of the instrument items used in this study indicated a Flesch-Kincaid grade level of 6.6. Flesch-Kincaid readability tests are widely used and have a high correlation with other readability tests.19 The Institutional
d

Journal of School Health

November 2007, Vol. 77, No. 9

2007, American School Health Association

Review Board of The Ohio State University approved the analyses for the present study.

Table 1. Standardized Factor Loadings from Principal Components Analysis 13 Survey Items*
Factor Loadings

Measures To study bullying perpetration as well as victimization, we employed 13 survey items that asked about frequency of behaviors during the past year. For each of the constituent items, the original response options included never, once, 2-3 times, 4 or more times. Principal components analysis of a random 20% subset of the sample yielded 2 factors with eigenvalues less than 1 that explained 47.6% of the variance in the constituent items. One factor included 7 items relating to perpetration (eg, During the past year at school, how often have you pushed others around to make them afraid?) and a second factor included 6 items related to victimization (eg, During the past year at school, how often has someone verbally attacked you?). These results were replicated on a second random 20% subset with nearly identical results. Both scales included measures of direct (eg, . . . how often have you threatened to beat someone up?) as well as indirect (eg, . . . how often have you told lies or spread false rumors about someone?) types of aggression. Analyses offered little support for distinguishing subscales of indirect and direct aggression. As such, the perpetration and victimization scales each included measures of both. The reliability of the scales was acceptable with Cronbach as of .82 and .74 for the perpetration and victimization scales, respectively. Table 1 presents the factor loadings for the constituent items. Because these scales yielded continuous measures of aggression and victimization, they were not appropriate for measuring bullying as a repeated pattern of behavior.1 To distinguish the phenomenon from lower levels of aggression/victimization, we classied as bullying any time a youth responded 4 or more times to at least one of the constituent items. Similar measurement approaches have been used previously.7,10,20 To examine the validity of this measurement strategy, we constructed alternative measures of bullying perpetration and victimization by dening as bullies and victims any youths whose scores fell in the upper quintile of the continuous scales. In Results, we discuss how results of analyses using the alternative measures differ from those of the primary measures. To assess whether perpetration and victimization are differentially related to established correlates of bullying, we constructed measures of depressive affect and substance use. Depressive affect was assessed by 2 variables (a = .61) reporting the frequency of feeling happy and depressed (range = 1 almost never to 3 most of the time). The
Journal of School Health
d

1 During the past year at school . . . How often have you told lies or spread false rumors about someone? How often have you pushed others around to get something you want? How often have you pushed others around to make them afraid? How often have you threatened to beat up someone? How often have you hit someone with your fists or beat up someone? How often have you taken money or things by force from people? How often have you left someone out of a group or activity to hurt them? How often has someone taken money or things directly from you using force, a weapon, or threats? How often have other students spread lies or false rumors about you? How often has someone physically attacked you? How often has someone verbally attacked you? How often has someone left you out of a group or activity to hurt you? How often have you feared for your physical safety?
*Factor loadings less than 0.10 are left blank.

2 0.152

0.51 0.776 0.821 0.772 0.72

0.665

0.587 0.57

0.677 0.101 0.647 0.695 0.729

0.683

substance-use variable was constructed by principal components analysis of 3 items measuring the frequency of cigarette, alcohol, and marijuana use (range = 1 never used to 6 use about every day). The factor model explained 73% of the variance in the constituent items, with loadings ranging from .84 to .88. Because the distribution of factor scores was highly asymmetrical, we applied a log transformation to reduce its departure from a normal distribution (mean = 0.15, SD = 0.33, skewness = 0.92, and kurtosis = 0.4521). Procedure We performed a series of cross-tabular analyses to examine how bullying varied by grade, gender, and ethnicity. A series of logistic regression models tested the association of both dimensions of bullying with substance use and depressive affect. In each model, we entered each of the demographic characteristics (ie, grade, gender, and ethnicity) as well as substance use and depressive affect. We then tested all 2-way interactions by comparing model v2 statistics
d

November 2007, Vol. 77, No. 9

2007, American School Health Association

625

in logistic regression models with the interaction term (and the main effects) versus models with only the main effects. For example, we regressed a yes/ no measure of bullying perpetration on grade, gender, ethnicity, substance use, and depression in step 1 and an interaction term (eg, grade by substance use) in step 2. Whenever signicant interaction terms were identied, we then performed stratied analyses (eg, separate models for males and females) to specify the exact nature of the differences. Given our large data set and the greater likelihood of type I errors with repeated statistical tests on the same variables, p values of less than .001 were viewed as statistically signicant.21 RESULTS This section presents the demographic characteristics of the sample, followed by the prevalence and co-occurrence of bullying perpetration and victimization by grade, gender, and ethnicity. Demographics The 2003 PPAAUS included data from 79,492 students representing 81.4% of the enrolled student population of participating schools and 96.7% of those completing the questionnaire.18 The most common reasons for failing to participate in the study included being chronically absent, home schooled, or otherwise not enrolled in school. As such, the data are only representative of students who regularly attended schools in Franklin County in 2003. The sample was evenly split between males (49.3%) and females (50.7%) and included white (63.0%), African American (20.6%), Hispanic (2.2%), Asian (3.1%), and Native American (0.7%) youth. In addition, 10.4% of respondents described themselves as multiracial, other, or refused to respond. Sixth graders comprised 16.8% of the sample, 7th/8th graders 32.9%, 9th/10th 28.5%, and 11th/12th 21.7%. Prevalence Overall, 20.1% of youth in the study reported having been bullied in the past year (Table 2). Victimization was somewhat more common among 6th-8th graders and males, although gender differences were only signicant among whites and Asians. Ethnic differences were modest, except for the much higher rates among Native American youths. No other interactions among demographic variables were detected. Findings for bullying perpetration presented a much more complex picture. Overall, 18.8% of the youth reported bullying perpetration during the past year (Table 2). Perpetration was most common
626
d

Table 2. Prevalence of Bullying Aggression and Victimization: Differences by Demographic Group*

n Overall Grade 6 7 8 9 10 11 12 Gender Male Female Ethnicity White African American Hispanic Asian Native American 78,068 13,077 12,926 12,789 11,730 10,541 9099 8008 37,676 39,142 49,535 15,863 1689 2464 570 Perpetration 18.80% 16.3 20.0 22.3 19.8 19.3 16.7 15.6 23.3 14.3 15.5 27.7 17.4 11.6 30.9 Victimization 20.10% 22.8 23.1 22.5 18.3 18.6 16.7 15.5 22.3 17.9 19.5 19.6 16.8 16.5 27.5

*ns may not sum to overall total because of missing data. Differs from overall column prevalence by p , .001.

among seventh through ninth graders and among males. Both African American and Native American youth reported much higher rates of perpetration, whereas rates among Asian youth were lower. Higher order analyses (not shown) found that gender differences in perpetration varied across different ethnic groups, Breslow-Day v2(4) = 28.22, p , .001. Among whites, Hispanics, and Asians, for example, perpetration was about twice as common among males compared with females. For other groups, however, gender differences were less pronounced. Among African Americans, 22.8% of females reported perpetration compared with 32.5% of males. Gender differences also varied by grade, Breslow-Day v2(6) = 31.05, p , .001, such that they were greatest among older youths. Comparing 8th and 12th graders, for instance, perpetration among females declined from 17.6% to 10.5%, whereas males only declined from 26.7% to 20.9%. Finally, ethnic group differences in perpetration diminished markedly between 6th and 12th grades. Through 9th grade, African American youths were about twice as likely as others youths to perpetrate bullying, yet by 12th grade, there was no signicant difference. Native American youth, however, departed from this trend, as their rates of perpetration remained high among youth in older grades. Even in 12th grade, 33.9% of youths in this group reported bullying others. Co-Occurrence We examined the co-occurrence of bullying perpetration and victimization and tested whether their
d

Journal of School Health

November 2007, Vol. 77, No. 9

2007, American School Health Association

co-occurrence varied across demographic groups. Over two thirds (68.5%) of youths reported neither bullying nor being bullied by others. Being a victim only (12.7%) was slightly more common than being a perpetrator only (11.4%), with an additional 7.4% of all youths being classied as bully/victims. Cooccurrence of perpetration and victimization varied by grade, being most common in 8th grade (9.0%) and least common in 11th (6.3%) and 12th (6.2%) grades. Co-occurrence was also more common among Native American (12.8%) and African American (9.1%) youth and least common among Asian (5.3%) youth. These differences, however, were largely associated with the overall prevalence of perpetration and victimization in each group. When we limited analyses to youths who had some exposure to bullying (ie, perpetration and/or victimization), less than one quarter (23.6%), were classied as both bully and victim. Analyses detected no signicant differences by grade, v2(6) = 16.16, p = .013, or ethnicity, v2(4) = 5.89, p = .21, although males (25.8%) were more likely than females (20.6%) to be classied as bully/ victims, v2(1) = 90.00, p , .001. Correlates of Bullying Behaviors A series of logistic regression models estimated how substance use and depressive affect were associated with bullying perpetration and victimization, controlling for age group, ethnicity, and gender. Preliminary analysis identied signicant interaction effects that indicated the need to stratify analyses across certain demographic variables. Specically, the association of bullying victimization with depressive affect and substance use varied by gender. Both depressive affect and substance use were positively associated with victimization, although the effects were somewhat stronger among females than among males (Table 3). For both genders, depressive affect had larger adjusted odds ratio (AOR) than did substance use. For bullying perpetration, the association with substance use varied by grade. Table 4 presents results

Table 4. AORs* (with 95% Confidence Intervals) for Depressive Affect and Substance Use Testing Association with Adolescent Bullying Perpetration: Models at 6th, 8th and 12th Grades
Grade 6 Depressive affect 1.84 Substance use 24.71 Model fit/ calibration, n Hosmer-Lemeshow, 31.94 v2 (p) Area under the receiver operating characteristic curve 8 12

(1.66-2.04) 1.62 (1.48-1.77) 1.60 (1.41-1.81) (18.73-32.61) 9.68 (8.23-11.39) 6.1 (4.99-7.54) 10,672 10,782 6956 (,.01) 0.75 12.41 (.09) 0.75 4.43 (.82) 0.72

*Adjusted for gender and ethnicity.

for logistic regression models at 6th, 8th, and 12th grades. In each model, the AORs for depressive affect was much smaller than those for substance use. Notably, the magnitude of the AORs for substance use diminished with increasing grade levels. The coefcients for the sixth-grade model should be interpreted with caution given the questionable model t. Results Using Alternative Bullying Measures To examine the validity of our measures of bullying, we replicated all analyses using alternative measures. For both perpetration and victimization, 90% of youths were similarly classied across the primary and alternative measures. Inconsistent classication of bullying perpetrators was more common among males (12.0%), African American, (12.3%) and Native American (15.6%) youth and was less common among sixth graders (8.0%). Inconsistent classication of bullying victims was not associated with grade, ethnicity, or gender. Overall, results of analyses using the alternative measure (available from the authors) were very similar to those reported above. DISCUSSION Overall, 28.2% of students reported involvement with some type of bullying behavior, which is consistent with other prevalence reports.22 Our ndings build on previous research by distinguishing different dimensions of bullying and systematically examining differences by age, gender, and ethnicity. Although the large, diverse nature of the sample permitted analyzing such distinctions, it is important to interpret the ndings with caution as the sample is taken from 1 metropolitan area, which may limit the generalizability. This section highlights a few key ndings, discusses how they relate to the previous literature, and proposes directions for future research.
d

Table 3. AORs* (with 95% Confidence Intervals) for Depressive Affect and Substance Use Testing Association with Adolescent Bullying Victimization: Models for Males and Females
Males Depressive affect Substance use Model fit/calibration, n Hosmer-Lemeshow, v2 (p) Area under the receiver operating characteristic curve
*Adjusted for grade and ethnicity.

Females 2.78 (2.64-2.93) 1.87 (1.67-2.04) 33,744 21.11 (.01) 0.69

2.42 (2.31-2.55) 1.57 (1.44-1.72) 32,242 12.90 (.11) 0.65

Journal of School Health

November 2007, Vol. 77, No. 9

2007, American School Health Association

627

Males reported higher levels of both types of bullying behaviors. Similarly, males were more likely than females to be classied as bully/victims. However, while both depressive affect and substance use were positively associated with victimization, the effects were somewhat stronger among females than among males. One possible explanation for this difference is that females may engage in more psychological forms of aggression, whereas males use more physical; the consequences of which may differ. Although the nature of the measures used in this study do not allow for this explanation to be examined here, studies have found that females both use and experience more indirect aggression.11,12 We detected several persistent ethnic group differences across bullying behaviors. One compelling explanation for this focus on the minority status of an ethnic group in particular schools. However, both Asian and Native American youth represented small proportions of the student body in each school, yet they reported very different rates of bullying behavior. Similarly, white and African American students often constituted either the majority or a large minority of students within different schools in the study. Yet, the students from these groups also reported different levels of bullying behavior. One possible explanation relates to previous ndings that ethnic minorities perceive that minorities were more likely than majority students to experience bullying.23 Consistent with a social constructionist view, Harris24 found that aggression is affected by cultural factors. Previous studies often note that many bullies are also victims.7 Several of our ndings suggest that it may be more appropriate to consider these dimensions of bullying as separate phenomena. Overall, only 7.4% of all youths were classied as bully/victims, which is similar to others who have found that between 3.9% and 8.2% are bully/victims.7,25 Moreover, our ndings suggest that bullying perpetration and victimization represent 2 distinct, if modestly related phenomena, and should be studied as such. Our analyses of correlates of bullying behaviors provide further support for the value of distinguishing perpetration from victimization. Substance use was more strongly associated with aggression than with victimization (Table 3). This is consistent with previous ndings14,17 and problem-behavior theory,15 which suggests that problem behaviors serve a functional purpose that aids in achieving a specic goal such as coping with rejection. Importantly, these are preliminary ndings and future research should investigate whether substance use is leading to bullying behaviors or bullying behaviors are leading to increased substance use. Consistent with previous literature,12,13 depressive affect was somewhat more strongly associated with victimization than
628
d

with aggression. However, as with substance use, future research should examine the nature of this relationship further. Implications for Prevention The demographic differences illustrated in this article have important implications for prevention, particularly in the areas of program targeting and message tailoring. Targeting is a means of dividing a population based on a variety of dimensions,26 including sex, age, and ethnicity. For example, our results suggest that younger adolescents tend to be victimized at higher rates. As such, interventions aimed at coping skills would best be targeted at these younger students. Also, Native American students experienced both forms of bullying behaviors at a higher rate than all other ethnicities, which suggests that this would be a meaningful subgroup to target with a culturally relevant intervention. Similarly, tailoring refers to messages personalized at the individual level27 and may be based on bully or victim status, substance use, or other characteristics specic to the adolescent. For example, a students involvement in bullying behaviors can be immediately assessed in an online survey, the results of which determine the subsequent prevention messages the adolescent receives. This type of computerbased tailoring has been used successfully in other areas of adolescent health28 and provides a potentially rich avenue for future research with its ability to provide a cost-effective way to target bullies, victims, and the small subset of bully/victims. Future research should also qualitatively examine the nature of the ethnic differences (eg, varying cultural norms, social contexts, coping skills) to make more specic recommendations for ethnic targeting. Similarly, future prevention programs should use interactive technologies to tailor specic risk reduction messages. To date, no programs have been designed specically for particular ethnic groups nor has the impact of existing programs adapted for various ethnic or cultural factors been systematically evaluated.7 Future work by both researchers and practitioners should address this gap. Such programs should be modeled after the effective bullying prevention programs implemented in the United States and other countries (see Olweus29 for an overview). Strengths and Limitations By analyzing both perpetration and victimization using a large sample that allowed for multiple comparisons and more in-depth analysis of interactions than has previously been examined, this article was able to add further renement to the existing literature on bully/victims. Moreover, the sample was diverse enough to compare 5 different ethnic groups.
d

Journal of School Health

November 2007, Vol. 77, No. 9

2007, American School Health Association

Our measures of bullying differ from those used by other studies in that we did not nd support for separating direct and indirect measures of aggression and victimization and we used the threshold of 4 or more times to qualify as bullying behavior. In many ways, this is a limitation of the literature in general given the lack of an agreed-upon scale for measuring bullying behaviors.30 Another limitation is that despite the large sample size, our study is based on a single urban county that may not be representative of other places in the United States. To the extent that different cultural and social climates inuence the prevalence of bullying as well as its determinants,29 our ndings may lack generalizability. Other studies, for instance, report that some but not all types of adolescent risk behaviors may vary by region.31,32 Also, the cross-sectional design of this study precluded our ability to test for the causal effects of bullying behavior on substance use and depressive affect.

CONCLUSIONS This article lls a gap in the literature through its examination of the various associations between perpetration and victimization and whether these associations vary by age, gender, and ethnicity. Overall, the ndings support the importance of distinguishing bullying behaviors and considering how they vary by demographic characteristics. Future research should examine the social and contextual factors surrounding these demographic differences and use these ndings to further rene prevention programming.

REFERENCES
1. Espelage DL, Swearer SM. Research on school bullying and victimization: what have we learned and where do we go from here? School Psych Rev. 2003;32:365-383. 2. Seals D, Young J. Bullying and victimization: prevalence and relationship to gender, grade level, ethnicity, self-esteem, and depression. Adolescence. 2003;38:735-747. 3. Rigby K, Slee PT. Bullying among Australian school children: reported behavior and attitudes toward victims. J Soc Psychol. 1994;131:615-627. 4. Landau S, Bjorkqvist K, Lagerspetz KM, et al. The effect of religiosity and ethnic origin on direct and indirect aggression among males and females: some Israeli ndings. Aggress Behav. 2002;28:281-299. 5. Eisenberg ME, Neumark-Sztainer D, Perry CL. Peer harassment, school connectedness, and academic achievement. J Sch Health. 2003;73:311-316. 6. Centers for Disease Control and Prevention. Surveillance summaries. MMWR Morb Mortal Wkly Rep. 2004;53(SS-2):651-655. 7. Haynie DL, Nansel T, Eitel P, et al. Bullies, victims, and bully/ victims: distinct groups of at-risk youth. J Early Adolesc. 2001;21:29-49. 8. Olweus D. Bully/victim problems in school: facts and intervention. Eur J Psychol Educ. 1995;12:495-510.

9. Schafer M, Werner NE, Crick NR. A comparison of two approaches to the study of negative peer treatment: general victimization and bully/victim problems among German schoolchildren. Br J Dev Psychol. 2002;20:281-306. 10. Mouttapa M, Valente T, Gallaher P, et al. Social network predictors of bullying and victimization. Adolescence. 2004;39: 315-335. 11. Underwood MK, Galen BR, Paquette JA. Top ten challenges for understanding gender and aggression in children: why cant we all just g et along? Soc Dev. 2001;10:248-266. 12. Crick NR, Casas JF, Nelson DA. Toward a more comprehensive understanding of peer maltreatment: studies of relational victimization. Curr Dir Psychol Sci. 2002;11:98-102. 13. Crick NR, Grotpeter JK. Relational aggression, gender, and social-psychological adjustment. Child Dev. 1995;66:710-722. 14. Berthold KA, Hoover JH. Correlates of bullying and victimization among intermediate students in the Midwestern USA. Sch Psychol Int. 2000;21:65-78. 15. Jessor R. Problem-behavior theory, psychosocial development, and adolescent problem drinking. Br J Addict. 1987;82:331-342. 16. Cooper ML, Frone MR, Russell M, et al. Drinking to regulate positive and negative emotions: a motivational model of alcohol use. J Pers Soc Psychol. 1995;69:990-1005. 17. Rusby JC, Forrester KK, Biglan A, et al. Relationships between peer harassment and adolescent problem behaviors. J Early Adolesc. 2005;25:453-477. 18. Safe and Drug Free Schools Consortium. 2003 primary prevention attitude awareness and use survey website. 2004. Available at: http://www.edcouncil.org/programs/drugfree/ ppaaus/web/table_of_contents.htm. Accessed September 1, 2005. 19. Paasche-Orlow M, Taylor H, Brancati F. Readability standards for informed consent forms as compared with actual readability. N Engl J Med. 2003;348:721-726. 20. Eslea M, Menesini E, Morita Y, et al. Friendship and loneliness among bullies and victims: data from seven countries. Aggress Behav. 2003;30:71-83. 21. Neter J, Wasserman W, Kutner MH. Applied Linear Statistical Models. 3rd ed. Boston, Mass: Irwin; 1990. 22. US Department of Health and Human Services. U.S. Teens in Our World. Rockville, Md: U.S. Department of Health and Human Services; 2003. 23. Siann G, Callaghan M, Glissov P, et al. Who gets bullied? The effect of school, gender and ethnic group. Educ Res. 1994;36: 123-134. 24. Harris MB. Ethnicity, gender, and evaluations of aggression. Aggress Behav. 1995;21:343-357. 25. Unnever JD. Bullies, aggressive victims, and victims: are they distinct groups? Aggress Behav. 2005;31:153-171. 26. Atkin CK. Theory and principles of media health campaigns. In: Rice RE, Atkin CK, eds. Public Communication Campaigns. 3rd ed. Thousand Oaks, Calif: Sage, 2001:49-68. 27. Bulger DW, Smith AB. Message tailoring: an essential component for disease management. Dis Manag Health Outcomes. 1999; 5:127-134. 28. Roberto AJ, Zimmerman RS, Carlyle KE, Abner EL. The effects of a computer-based HIV, STDs, and pregnancy prevention program on rural adolescents. J Health Comm. 2007;12:53-76. 29. Olweus D. A prole of bullying. Educ Leadersh. 2003;60:12-17. 30. Crothers LM, Levinson EM. Assessment of bullying: a review of methods and instruments. J Couns Dev. 2004;82:496-503. 31. Beauvais F. Comparison of drug use rates for reservation Indian, non-reservation Indian and Anglo youth. Am Indian Alsk Native Ment Health Res. 1992;5:32-37. 32. Johnston LD, OMalley PM, Bachman JG, Schulenberg JE. Monitoring the Future National Survey Results on Drug Use, 1975-2004. Volume I: Secondary School Students (NIH Publication No. 05-5727. Bethesda, Md: National Institute on Drug Abuse; 2005.

Journal of School Health

November 2007, Vol. 77, No. 9

2007, American School Health Association

629