Or

Dening and Measuring Cyberbullying Within the Larger Context of BullyingVMJaa Ceb Mc Ned Gr

ArtKe

A

PuMetoRewhirredebeyingbuAcraveCoincresDirmitioho

adpsy

*tive

Journal of Adolescent Health 51 (2012) 5358

105doiictimizationichele L. Ybarra, M.P.H., Ph.D.a,*, Danah Boyd, Ph.D.b,c, Josephine D. Korchmaros, Ph.D.a, andy (Koby) Oppenheim, M.A.d

nter for Innovative Public Health Research, Irvine, Californiaicrosoft Research, Boston, Massachusettsw York University, Department of Media, Culture and Communication, New York City, New Yorkaduate Center, City University of New York, New York City, New York

icle history: Received April 20, 2011; Accepted December 22, 2011ywords: Bullying; Cyberbullying; Measurement

B S T R A C T

rpose: To inform the scientic debate about bullying, including cyberbullying, measurement.thods: Two split-form surveys were conducted online among 617-year-olds (n 1,200 each)inform recommendations for cyberbullying measurement.sults:Measures that use theword bully result in prevalence rates similar to eachother, irrespective ofether a denition is included, whereas measures not using the word bully are similar to each other,spective of whether a denition is included. A behavioral list of bullying experiences without either anition or the word bully results in higher prevalence rates and likely measures experiences that areond thedenition of bullying. Follow-upquestions querying differential power, repetition, andbully-over time were used to examine misclassication. The measure using a denition but not the wordllyappearedtohavethehighestrateoffalsepositivesand,therefore,thehighestrateofmisclassication.oss two studies, an average of 25% reported being bullied at leastmonthly in person comparedwith anrage of 10%bullied online, 7% via telephone (cell or landline), and 8%via textmessaging.nclusions:Measures of bullying among English-speaking individuals in the United States shouldlude the word bully when possible. The denition may be a useful tool for researchers, butults suggest that it does not necessarily yield amore rigorousmeasure of bullying victimization.ectly measuring aspects of bullying (i.e., differential power, repetition, over time) reducessclassication. To prevent double counting across domains, we suggest the following distinc-ns: mode (e.g., online, in-person), type (e.g., verbal, relational), and environment (e.g., school,me). We conceptualize cyberbullying as bullying communicated through the online mode.

2012 Society for Adolescent Health and Medicine. All rights reserved.

IMPLICATIONS ANDCONTRIBUTIONS

Denitions and measures ofcyberbullying vary widely,contributing to an inconsis-tency in ndings across stud-ies. Data from two split-formstudies suggest that using theword bully in the surveymeasure, and including a fol-low-up question about differ-ential power, reduces self-misclassication of youth.Implementing these two as-pects of measurement in bul-lying surveys will better alignndings across future studies.

Cyberbullying and harassment victimization are signicant social and behavior problems, and substance use [14]. Depend-

See Editorial p. 3iginal articleolescent health issues, as they are associated with concurrentchosocial problems including depressive symptomatology,

ingpre

thecomprolyi

Address correspondence to:Michele Ybarra, M.P.H., Ph.D., Center for Innova-Public Health Research, 555 El Camino Real #A347, San Clemente, CA 92672.E-mail address:Michele@InnovativePublicHealth.org (M. Ybarra).

4-139X/$ - see front matter 2012 Society for Adolescent Health and Medicine. All r:10.1016/j.jadohealth.2011.12.031www.jahonline.orgon the denition, measure, and methodology used, recentvalence rates range between 9% [5] and 72% [6].Denitions of cyberbullying vary widely [7], contributing toinconsistency in ndings across studies. A lack of consensusplicates cross-study comparisons and, thus, limits researchgress. Some researchers treat cyberbullying as a type of bul-ng, equivalent to physical and relational bullying [8]; others

ights reserved.

treasentoonnicceltin

oping[10[2,[1,badifbethepremebeingstrAlsbepepo

thiphthawhingof

Ga

tiometwwo[23yovarsurbutertimageriemoimweale

inwh19cybthrrel

occNocomtiothitofronuwh

Me

20thealscauheprotio

Par

on6plesurstuin

butherecologroNopastada

thenuanwh

TabDem

D

FA

R

H

M.L. Ybarra et al. / Journal of Adolescent Health 51 (2012) 535854at it as an environment, equivalent to school [9]. If it is treateda type of bullying (e.g., cyber vs. relational bullying) or as anvironment (e.g., online vs. at school), measures are vulnerabledouble counting (e.g., relational bullying online; being bulliedlinewhile physically located at school). If treated as a commu-ation mode (i.e., in person, text messaging, voice [landline orl phone], or online), however, cyberbullying becomes a dis-ct and meaningful category.Even when dening cyberbullying the sameway, researcherserationalize it differently across studies. Some measure it us-a simple question (e.g., have you been cyberbullied?),11]. Others use a denition (i.e., we say bullying is. . .)3,9], a list of behavioral experiences [4,5,1216], or both6,8,1720]. Drawbacks exist for each approach. A denition-sed measure may challenge respondents whose experiencesfer from the denition. It also assumes that the denition willread and understood. Using theword bully necessitates thatparticipant adopt the label of being bullied [21]. It alsosumes that the respondent and researcher share the sameaning of bully. Because of rapid changes in technology,havioral lists, although providing concrete examples of bully-, are vulnerable to constant iteration unless lists are con-ained to experiences that are universal across environments.o, unless coupled with a denition or follow-up questions,havioral lists likely measure general aggression instead of re-titive bullying between two or more people of differentialwer.To avoid the word bully, some use synonyms (e.g., meanngs) [6,20]. In a 14-country comparison of 67 words andrases used to describe bullying, Smith and colleagues reportt the terms bullying and picking on cluster together,ereas the words harassment, intimidation, and torment- relate to each other in a different cluster [22]. Thus, the usesynonyms may not always connote bullying.

p in the Literature

To better understand how variations in denition and opera-nalization affect prevalence rates and to identify the bestthod for measuring cyberbullying, we report the results ofo related studies. Study 1 examines the relative impact of therd bullying and an adapted version of Olweus denition] on prevalence rates. It questions whether the likelihood ofuth admission of being bullied, and adoption of that label,ies by the appearance of theword bully or a denition in thevey question. Study 2 examines how well reported rates ofllying align with Olweus [23] three main denitional charac-istics of bullying: differential power, repetitiveness, and overe. We identify which measure results in the highest percent-of accurate self-classication (i.e., those who say their expe-

nce occurred over time, repeatedly, and by someone withre power than them). Given the use of other terms to approx-ate the word bullying (e.g., mean things and harassment),also examine the impact of one word harassment on prev-nce rates.Olweus [23] denition of bullying has foundwide acceptancethe research literature. Based on the number of researcherso are using an adapted version of this denition [2,3,8,9,17], this seems true of cyberbullying as well. We conceptualizeerbullying under the larger rubric of bullying. We proposeee mutually exclusive components: (1) type (e.g., physical,ational), (2) mode of communication through which bullyingurs (e.g., in person, online), and (3) environment (e.g., school).t all components need to be included, but they should not bebined even if space or budget is limited to avoid uninten-

nal double counting of victimization rates. To our knowledge,s is the rst study to propose measurement to be constrainedthese three distinct domains. The recommendation arisesm extensive consultations among the authors about the co-ndrum attributable to counting victimization across spaces inich youth engage.

thods

Two separate mini-surveyswere conducted online: January10 (study 1) andMay2010 (study 2) (A third survey examiningimpact of question order on prevalence rates of bullying waso conducted. Differences were not statistically signicant. Be-se of space limitations in the journal, results are excludedre, but they are available upon request from the authors). Thetocol was reviewed and approved by Chesapeake Institu-nal Review Board. C &R Research administered the surveys.

ticipants

Respondentswere randomly selected from a 30,000-memberline panel. For each survey, 1,200 youth between the ages of17 years were recruited. Younger youth (69 years) com-ted the survey with their parent; older youth completed thevey alone. A waiver of parental consent was obtained for thedies. Youth assent was required to participate95% assentedstudy 1 and 97% in study 2.The survey research rm routinely conducts monthly omni-ses (i.e., surveys). Participants who took part in an omnibus inpast 3 months were excluded from the current months

ruitment pool. The sample was purposefully balanced on bi-gical sex (50% female) and age-groups (300 youth in eachup: 68 years; 911 years; 1214 years; and 1517 years).other eligibility criteria were applied. As shown in Table 1,

rticipants were an average of 12 years (mean: 11.9 years,ndard deviation: 3.5 years). About 70% were white (weightedta).The response rate (i.e., the number of people who clicked onsurvey invitation link in the e-mail divided by the total

mber of survey invitation e-mails sent) for study 1 was 32%d for study 2 was 39%. Survey completion rates among thoseo started the survey were about 93% for each study. Rates are

le 1ographic characteristics of study participants

emographic characteristics Study 1 (n 1,140) Study 2 (n 1,167)

emale 49.6% (581) 48.6% (582)ge (years)68 22.1% (285) 22.7% (291)911 22.3% (275) 23.3% (290)1214 26.1% (289) 24.9% (290)1517 29.5% (291) 29.1% (296)ace/ethnicityWhite 81.2% (926) 79.8% (931)Black/African American 3.9% (45) 5.6% (65)Some other race 5.0% (57) 5.1% (60)Hispanic 6.6% (75) 6.8% (79)Decline to answer 3.2% (37) 2.7% (32)ousehold income $50,000 29.2% (333) 27.3% (319)

wi[24

difodsevciaon

Me

yeaneantimtioarepo

dif

yodoteakicrum

abab

A H Abt

Thmethesay

wirephaoth

hoingschpe

Infolatyo

twanpriitewe

expsomyoingroyotru

occmu(cenetur

ing

diftiowo(prdewamof

buqustrthapohape

ron

Da

yogeoticspostuestgogobuwaitiv

Re

itefacgiryeaiteoftho

M.L. Ybarra et al. / Journal of Adolescent Health 51 (2012) 5358 55thin the expected range of well-conducted online surveys,25].To examine relative differences in prevalence rates based onferentmeasures eldedwithin one sample, a split-formmeth-ology was useda random subsample was assigned to one oferal possible forms of the measure. Internal validity is cru-l for valid split-form studies; external validity less so. Thus, anline panel is as acceptable as other sampling procedures.

asures

All questions used a 5-point scale and referred to the pastr. Youth were categorized into one of three groups: (1)ver, (2) less frequently thanmonthly (i.e., once or a few times),d (3) monthly or more often (i.e., a few times a month, a fewes a week, every day, or almost every day). This categoriza-n grouping was determined prima facie to reect those whobullied repetitively and over time (at least monthly) as op-

sed to those aggressed on less frequently.In study 1, youth were randomly assigned to one of fourferent forms of the survey question.The denition + word bully form read as follows: We say aung person is being bullied when someone repeatedly says ores mean or nasty things to them. Examples include beingsed repeatedly or having nasty or cruel things said; being hit,ked, or pushed around; being excluded or left out; or havingors spread.We are not talking about times when two young people ofout the same strength ght or tease each other. We are askingout things that:

re repeated (happen more than once)appen over time (more than just 1 day)re between people of different power or strengththis mighte physically stronger, socially more popular, or some otherype of strength.

ese things can happen anywhere, like at school, online, via textssaging, at home, or other places young people hang out. Inlast 12months, howoften have others bullied you by doing oring the following things to you?

The denition-only formwas the same as aforementioned, butth a modied rst sentence: Sometimes people your ageeatedly say or do mean or nasty things to each other. It alsod a modied question: In the last 12 months, how often haveers done or said the following things to you?The bully-only form read as follows: In the last 12 months,w often have others bullied you by doing or saying the follow-things to you? These things can happen anywhere, like atool, online, via textmessaging, at home, or other places youngople hang out.The nal form presented neither the denition nor the word:the last 12 months, how often have others done or said thelowing things to you? These things can happen anywhere, likeschool, online, via text messaging, at home, or other placesung people hang out.Note that the denition provided did not differentiate be-eenmode, environment, and type because thiswas unimport-t for participants to consider. Instead, participants weremed to think about bullying experiences broadly, and then them response options forced the differentiation (i.e., modesre queried separately from types of bullying).All youth were then presented the same behavioral list oferiences: (1) hit, kicked, pushed, or shoved you around; (2)eone made threatening or aggressive comments to you; (3)

uwere calledmean names; (4) youweremade fun of or teaseda nasty way; (5) you were not let in or you were left out of aup because someone was mad at you or was trying to hurtu; (6) someone spread rumors about you, whether they weree or not; and (7) some other way.Youth who said that at least one of these experiences hadurred to them in the past year were asked the mode of com-nication through which it occurred: in person, by phone callll or land line), by text message, or online (e.g., e-mail, socialtwork site, or instantmessenger). Response options were cap-ed with the same 5-point frequency scale.Budget limitation prevented the inclusion of the third bully-context, environment, in study 1.In study 2, youth were randomly assigned to one of fourferent forms: (1) denition the word bully, (2) the deni-n-only, (3) the word bully-only, and (4) denition therds bully and harassment. Based on ndings from study 1esented later in the text), a form that included neither thenition nor the word bully was not included. The denitions modied slightly from study 1 to improve readabilityore than once and more than just 1 day were used insteadare repeated and happen over time.Across all four forms, youth who indicated they had beenllied through at least one mode were asked three follow-upestions: (1) Was it by someone who had more power orength than you? This could be because the person was biggern you, had more friends, was more popular, or had morewer than you in another way, (2) Was it repeated, so that itppened again and again?, and (3) Did it happen over a longriod? We mean more than a week or so?Because of budget limitations, victimization type and envi-ment were not queried in study 2.

ta analysis

Data were weighted to match the U.S. online population ofuth on biological sex, age, race/ethnicity, household income,graphy, and county size. The design-based Pearson 2 statis-was used to determine difference between categorical re-nses, while taking into account survey weighting [26]. Indy 2, we calculated the positive predictive value (PPV) toimate each measures rate of misclassication. PPV requires ald standard bywhich ameasure is compared.Wedened theld standard as being consistent with Olweus components ofllyingdifferential power, with repetition, and over time. PPVs calculated as the True Positives/(True Positives False Pos-es).

sults

Conrmatory factor analysis suggested that the behavioralms measuring the different types of bullying loaded on onetor, bullied (Table 2). Loadings were similar for boys andls and younger (e.g., 68-year-olds) and older (e.g., 1517-r-olds) youth (data available upon request). The behavioralms were combined to reect youth who reported at least onethe seven types of bullying victimization had occurred versusse who reported none of the seven types of bullying experi-

ensub

The

Stuwede

TabFacfor

T

H

T

MNSRS

TabCom

Q

B

C

NACom e (e.g.,usin onb h o den

a

b

c

d

e

f

M.L. Ybarra et al. / Journal of Adolescent Health 51 (2012) 535856ces had occurred. This variable, behavioral list, was used insequent analyses.

effect of measurement text and order on prevalence rates

dy 1. As shown in Table 3, reported rates of victimizationre highest in the split-form measure that used neither thenition nor theword bully and lowest for the two forms that

le 2tor loadings for iterative principal components Conrmatory Factor Analysisthe latent variable bully

ype of experience Study 1

Denition bully

Denition-only

Bully-only

Neither

n 281 n 287 n 286 n 286

it, kicked, pushed,shoved

.65 .72 .63 .68

hreatening/aggressivecomments

.73 .80 .79 .86

ean names .80 .82 .79 .75asty teasing .81 .87 .79 .72ocial exclusion .72 .74 .66 .80umors .72 .78 .67 .73omething else .59 .64 .58 .75

le 3parison of 12-month prevalence rates of bullying based on measure text and o

uestion type Study 1

Denition bully

Denition-only

Bully-only

Neither dnor bull

n 281 n 287 n 286 n 286

ehavior list (any)Never 26% 22% 26% 19%Monthly 40% 39% 38% 38%Monthly 34% 39% 35% 42%

ommunication modeAny modeNever 42%c 32%d 41%e 27%Monthly 35% 40% 42% 43%Monthly 23% 27% 16% 30%

In personNever 46%c 37%d 46%e 32%Monthly 33% 38% 38% 41%Monthly 20% 25% 16% 27%

TelephoneNever 88%c 84% 91%e 76%Monthly 9% 9% 6% 14%Monthly 3% 7% 3% 9%

Text messageNever 88%abc 77%d 88%e 79%Monthly 5% 14% 10% 11%Monthly 6% 9% 2% 10%

OnlineNever 84%c 76%d 86%e 74%Monthly 9% 16% 10% 14%Monthly 7% 7% 4% 12%

not applicable.parisons of prevalence rates within each study (e.g., study 1) and question typg 2 tests. No statistically signicant differenceswere noted between: 1)Denitiarassment; or 3) Denition-only and neither. The following notations are used tDenition bully and denition-only.Denition bully and bully-only.Denition bully and neither.Denition-only and bully-only.Bully-only and neither.Denition-only and denition bully harassment.d theword bully in the introduction to the survey question.ferences between different forms are noted in the Table. Formple, rates for the denition-only form were statisticallyistinguishable from the form that used neither denition norword bully, Reliable differences were noted for communi-ion mode between the bully-only form and the form thatd neither denition nor the word bully, as well as betweendenition bully form and the form that used neither the

nition nor bully for bullying via phone. The denition lly form also differed reliably from the denition-only formbullying via text messaging.

dy 2. Twelve-month prevalence rates for bullying using thenition bully and bully-only formswere similar to thoseorted in study 1. Rates observed for the form that used thenition bully harassment were similar to those ob-ved for the denition bully form, suggesting that thewordrassment does not denote additional context beyond bully.

ferential misclassication

In study 2, youth who endorsed any type of bullying experi-ce were asked follow-up questions to determine whether theeriences included (1) differential power, (2) repetition, andoccurrence over time. To calculate the PPV, endorsement of

Study 2

ion Denition bully

Denition-only

Bully-only

Denition bully harassment

n 291 n 290 n 290 n 296

NA NA NA NA

43%a 20%df 42% 41%34% 25% 32% 34%22% 55% 25% 25%

46%a 21%df 45% 45%34% 31% 32% 33%20% 48% 23% 22%

91%a 69%df 90% 87%6% 12% 4% 9%4% 19% 6% 4%

87%a 67%df 85% 86%8% 10% 9% 10%6% 23% 6% 5%

85%a 67%df 80% 79%8% 8% 12% 13%7% 25% 8% 9%

behavioral list, online) tested for statistically signicant difference (p .05)ully andDenitionbullyharassment; 2)Bully-only andDenitionbullyote statistically signicant differences noted:useDifexaindthecatusethedebufor

Studerepdeserha

Dif

enexp(3)

rder

enity

allOlwrepthrwhfalsim

beivicfollaboendquon

Relstumoleavia

Dis

intexpeacmeirreeithperseaouin tuserieprede

forqu

abbeadfer

difbuquwhcridenInstimmorequisexpabitheratdotrarat

TabPos

Gstd

only

hly M

T

1) 1

D

) 1)

Per nitioncrit y subtPPV lse po

a criter

b er.

M.L. Ybarra et al. / Journal of Adolescent Health 51 (2012) 5358 57three was dened as the gold standard, consistent witheus denition of bullying. As shown in Table 4, 9%11%orted being bullied monthly or more often and also met theee criteria. Between 0% and 3% youth met all three criteriaile also reporting it occurred less often thanmonthly. Rates ofse positives, and therefore the positive predictive value, wereilar across all four question forms.We also examined a briefer follow-up series. Youthwho reportedng bullied at least monthly implicitly meet the criteria for beingtimized both over time and repeatedly. Then, the only additionalow-up question needed to determine the gold standard is oneut differential power. When the gold standard is redened asorsing the question about differential power and reporting a fre-ency of monthly ormore often, PPV ranges from47% (denition-ly) to 65% (bully-only; see Table 4).

ative rates of bullying by communicationmode. Across the twodies, an average of 25% reported being bullied at leastnthly in person. An average of 10% reported being bullied atst monthly online, 7% via telephone (cell or landline), and 8%text messaging.

cussion

Across two split-form online surveys of youth 617 years old, theroduction text appears to affect endorsementof bully victimizationeriences.Measures that use theword bully affect rates similar toh other, irrespective of whether a denition is included, whereasasures that do not use the word bully are similar to each other,spective of whether a denition is included. This suggests thater the participants are not reading the denition or that it is notsonally meaningful. The denition may be a useful tool for re-rchers, but these results suggest that it doesnot yield amore rigor-smeasureofbullyingvictimization. If readingburdenwerean issuehe current survey, it is likelymore so an issue in other surveys thatlonger denitions. Furthermore, a behavioral list of bullying expe-

nceswithouteitheradenitionor thewordbully results inhighervalence rates and likelymeasures experiences that are beyond thenition of bullying.Victimization rates are strikingly similar acrossmeasurementms when the three Olweus [23] criteria-based follow-upestions are applied. All forms, therefore, seem to be equally

le 4itive predictive value of four different measures of bullying

oldandardenition

Denition bully Denition-only Bully-

Monthly Monthly PPV Monthly Monthly PPV Mont

hree criteriameta

Yes 3% (8) 9% (22) 18% 0% (1) 11% (39) 17% 1% (4)No 31% (96) 13% (39) 25% (87) 44% (110) 31% (94ifferential

powerb

Yes 12% (39) 14% (36) 59% 10% (30) 22% (70) 47% 14% (43No 22% (65) 8% (25) 16% (58) 33% (79) 18% (55

centages are based upon the entire sample. For example, 3% of all youth in theDeeria. Percentages for youth who did not report being bullied can be calculated b positive predictive value, calculated as the true positives/(true positives faThe gold standard is the measure (eitherMonthly or Monthly) all threeFor example, denition bully PPV for all three criteria (2289639).The gold standard is frequency of monthly or more often differential powFor example, denition bully PPV for all three criteria (36)/(36 25).le to identify true positives (i.e., those who say they haveen bullied and endorse all three follow-up questions). Thus,ding these three follow-up questions seems to neutralize dif-ences in the question forms.Thebenetof using all three follow-upquestions is the ability toferentiate between the few participants who report repetitivellying that occurs over a relatively short period (i.e., less fre-ently thanmonthly) andalsomeet the three criteria versus thoseo report bullying over a short period but do not meet the threeteria. In the current studies, this reects 0%3% of the respon-ts. The drawback of using three questions is participant burden.teadofusingseparatequestions toqueryrepetitivenessandovere, one could classify participants who report being bulliednthlyormore frequently as, bydenition,meeting thecriteriaofpetitiveness and over time. In this case, only one follow-upestion toquerydifferentialpower isneeded.Again, thedrawbackmissing the respondents who have intense but shorter-termeriences.Whenthisgold standard is applied, thePPV, that is, thelity toaccurately identifyvictimizationcases, suggests thatusingdenition without the word bully may result in the higheste of false positives (i.e., thosewho say theyhave beenbullied butnot endorse the differential power follow-up question). In con-st, the word bully may elicit responses that lead to the lowestes of misclassication. Researchers should consider including atst this one follow-up question in the same way that follow-upsfunctional impairment are now included for many Diagnosticd StatisticalManual ofMental Disorders (DSM IV)measures [27].We used the same behavioral lists across all communicationdes.When lists are used,we advocate for universal lists acrossmunication mode and environment to allow for compari-s of youth experience online and off-line. This alsomakes theasure transcendent of technology. Adolescent uses of technol-y will continue to change at a rapid pace, but this should notect our denition of bullying or the associated behavioral lists.

plications for cyberbullying

Similar to pan-European data [20], twice as many youth in ourdies report bullying in person compared with each of the othermunication modes. Despite the rapid uptake of technologies,

aditional face-to-face communication still is the dominantdeofbullying.Conceptualizingmodeasacomponentofbullying

Denition bully harassment

All (combined)

onthly PPV Monthly Monthly PPV Monthly Monthly PPV

0% (25) 17% 2% (6) 11% (32) 22% 2% (19) 10% (118) 18%5% (47) 32% (88) 14% (50) 30% (365) 21% (246)

7% (47) 65% 13% (37) 16% (50) 61% 12% (149) 17% (203) 56%8% (25) 21% (57) 9% (32) 19% (235) 14% (161)

bully study reported being bullied less thanmonthly and alsomet all threeracting the four shown groups of youth from 100%.sitives).ia met (i.e., [1] differential power, [2] repetition, and [3] over time).leaforan

mocomsonmeogaff

Im

stucomtrmo

permits us to count rates of online experiences separately fromthomaofplebleothtudwanonopoho12knrepapp

Limthe[20weyoalppode8-y[22merepanresThdepri

Co

geswoingeithandbubeaddconideingfollpro(e.gmesho

Ac

HDDean

Institute of Child Health and Human Development. The authorsthaCrotiretic

Re

[1

[2

[3

[4

[5

[6

[7

[8

[9

[10

[11

[12

[13

[14

[15

[16

[17

[18

[19

[20

[21

[22

[23

[24

[25

[26

[27

M.L. Ybarra et al. / Journal of Adolescent Health 51 (2012) 535858se occurring in person, allowing for direct comparisons. Someybe concerned that components of theOlweus-baseddenitionbullying may not translate well to the online context. For exam-, the concept of repetition onlinemay be different, as it is possi-to have a picture posted or rumorwritten once, yet sharedwithers over and over again. Although different in potential magni-e, this seems similar to a rumor scrawled once on a bathroomll for many people to see repetitively. Traditionally, we wouldt say that thismeets thedenitionof repetitionoff-line, and it ist clear why it should online. Similarly, anonymity has beeninted to as something unique to the online world. This assumes,wever, that all bullying off-line is done by known people. In fact,% of youth reporting being bullied at school say they do notow who their bully is [9]. Although lower than the 46% whoortnotknowingwhotheironlinebully is, the issueofanonymitylies beyond online spaces.

itations. Findings are specic to English-speaking youth inUnited States. Not all languages have the word bullying,22]. Different ndings would likely emerge if a similar studyre conducted among non-English-speaking individuals. Be-nd an examination of internal consistency (i.e., Cronbachsha and conrmatory factor analysis) of the behavioral items,ssible variation by agewas not examined. It is possible that thenition and word bully have different inuences on anear-old individual compared with an 18-year-old individual]. Also, the calculation of PPV requires a gold standard toasure against; a stronger standard would have been observerort. Finally, somemay wonder why other more sophisticatedalyses were not used. For example, the application of itemponse theory is currently being debated in bullying research.e discussion centers on whether being bullied reects an un-rlying trait; if it is not, then item response theory is inappro-ate. This question is beyond the scope of this article.

ntribution

Findings from twomini-surveys of youth sampled nationally sug-tthatmeasuresforEnglish-speakingindividualsshouldincludetherdbullywhenpossible.Thedenitionseemslesscritical inaffect-prevalence rates. A behavioral list of bullying experienceswithouteradenitionorthewordbullyresultsinhigherprevalencerateslikely measures experiences that are beyond the denition ofllying.Thewordharassment,althoughresearchersconsider it tomeaningfully different from bullying, does not seem to connoteitionalcontext foryouthbeyondbully. Its inclusiontohelpyouthceptualize bullying, or to give themadifferent termwithwhich tontify, does not affect bullying rates. Furthermore, directlymeasur-aspects of differential power and repetition over time throughow-upquestions reduces themisclassicationofyouth. Finally,wepose three mutually exclusive components of bullying: (1) type.,physical), (2)communicationmode(e.g.,online),and(3)environ-nt (e.g., home). Not all components must be included, but theyuld not bemerged to prevent double counting.

knowledgments

The project described was supported by award number R01-057191 from the National Institute of Child Health and Humanvelopment. The content is solely the responsibility of the authorsddoes not necessarily represent the ofcial viewsof theNationalnk Dr. Kimberly Mitchell, Ms. Amanda Lenhart, and Dr. Donnass for their input in the study design;Ms. Tonya Prescott for herless help with formatting and proofreading; and the study par-ipants for their participation in this study.

ferences

] Raskauskas J, Stoltz AD. Involvement in traditional and electronic bullyingamong adolescents. Dev Psychol 2007;43:56475.

] Smith PK,Mahdavi J, CarvalhoM, et al. Cyberbullying: Its nature and impactin secondary school pupils. J Child Psychol Psychiatry 2008;49:37685.

] Slonje R, Smith PK. Cyberbullying: Another main type of bullying? ScandJ Psychol 2008;49:14754.

] Ybarra ML, Diener-West M, Leaf PJ. Examining the overlap in internetharassment and school bullying: Implications for school intervention. JAdolesc Health 2007;41:S4252.

] Wolak J, Mitchell KJ, Finkelhor D. Online victimization of youth: 5 years later.Alexandria, VA:National Center forMissing andExploited Children, 2006.

] Juvonen J, Gross EF. Extending the school grounds?Bullying experiencesin cyberspace. J Sch Health 2008;78:496505.

] Tokunaga RS. Following you home from school: A critical review and syn-thesis of research on cyberbullying victimization. Comput Hum Behav2010;26:27787.

] Wang J, Iannotti RJ, Nansel TR. School bullying among adolescents in theUnited States: Physical, verbal, relational, and cyber. J Adolesc Health 2009;45:36875.

] Ybarra M, Mitchell K, Espelage D. Comparisons of bully and unwantedsexual experiences online and ofine among a national sample of youth. In:Ozdemir O, Ovali F, eds. Pediatrics/Book 2 (in press).

] Li Q. Cyberbullying in schools: A research of gender differences. Sch PsycholInt 2006;27:15770.

] Livingstone S, BoberM. UK kids go online: Final report of key project ndings.London, UK: LSE Research Online, 2005. Available at: http://eprints.lse.ac.uk/399/01/UKCGO_Final_report.pdf. Accessed February 12, 2011.

] Aricak T, Siyahhan S, Uzunhasanoglu A, et al. Cyberbullying among Turkishadolescents. Cyberpsychol Behav 2008;11:25361.

] Hinduja S, Patchin JW. Ofine consequences of online victimization: Schoolviolence and delinquency. J Sch Violence 2007;6:89112.

] Patchin JW, Hinduja S. Bullies move beyond the schoolyard: A preliminarylook at cyberbullying. Youth Violence Juv Justice 2006;4:14869.

] Finkelhor D, Mitchell KJ, Wolak J. Online victimization: A report on thenations youth. Alexandria, VA: National Center for Missing and ExploitedChildren, 2000.

] Topu C, Erdur-Baker O, Capa-Aydin Y. Examination of cyberbullying expe-riences among Turkish students from different school types. CyberpsycholBehav 2008;11:6438.

] Katzer C, Fetchenhauer D, Cyberbullying BF. Who are the victims? A com-parison of victimization in internet chat rooms and victimization in school.Media Psychol 2009;21:2536.

] Kowalski RM, Limber SP. Electronic bullying amongmiddle school students.J Adolesc Health 2007;41:S2230.

] Dehue F, Bolman C, Vllink T. Cyberbullying: Youngsters experiences andparental perception. Cyberpsychol Behav 2008;11:21723.

] Livingstone S, Haddon L, Grzig A, et al. Risks and safety on the internet: Theperspective of European children. In: Full Findings. London, England: LSE,EU Kids Online, 2011. Available at: http://eprints.lse.ac.uk/33731/. Ac-cessed February 12, 2011.

] Marwick A, Boyd D. The drama! Teen conict, gossip, and bullying innetworked publics. Paper Presented at: A Decade in Internet Time: Sympo-sium on the Dynamics of the Internet and Society; Sept 22 2011; Oxford,England. Available at: http://papers.ssrn.com/sol3/papers.cfm?abstract_id1926349. Accessed October 12, 2011.

] Smith PK, Cowie H, Olafsson RF, et al. Denitions of bullying: A comparisonof terms used, and age and gender differences in a fourteen country inter-national comparison. Child Dev 2002;73:111933.

] Olweus D. The Revised Olweus Bully/VictimQuestionnaire. Bergen, Norway: Re-searchCenter forHealth Promotion [HIMIL], University of Bergen, 1996.

] Cook C, Heath F, Thompson RL. Ameta-analysis of response rates inweb- orinternet-based surveys. Educ Psychol Meas 2000;60:82136.

] Kaplowitz MD, Hadlock TD, Levine R. A comparison of web andmail surveyresponse rates. Public Opin Q 2004;68:94101.

] StataCorp. Stata statistical software. Release 11. College Station, TX: Stata-Corp LP, 2009.

] American Psychiatric Association. Diagnostic and statistical manual ofmental disorders, 4th edition. Washington, DC: American Psychiatric Asso-ciation, 2000.

Defining and Measuring Cyberbullying Within the Larger Context of Bullying VictimizationGap in the LiteratureMethodsParticipantsMeasuresData analysis

ResultsThe effect of measurement text and order on prevalence ratesStudy 1Study 2

Differential misclassificationRelative rates of bullying by communication mode

DiscussionImplications for cyberbullyingLimitations

ContributionAcknowledgmentsReferences