ArchSexBehav(2010)39:1207–1208 DOI10.1007/s10508-010-9634-z LETTER TO THE EDITOR The Importance of Measuring Internalized Homophobia/ Homonegativity MichaelW.Ross • B.R.SimonRosser • DerekSmolenski Publishedonline:13May2010 (cid:2)SpringerScience+BusinessMedia,LLC2010 NewcombandMustanski(2009)aretobecongratulatedfor theabilitytointerpretthemagnitudeofanyobservedassoci- attempting a meta-analysis of internalized homophobia/ ation,leavingonlythedirection. homonegativity(IH)andforrecognizingthepotentialimpor- Second,theninedifferentIHmeasuresusedaresimplynot tanceofthisconstructforHIV-relatedresearchinmenwhohave comparable,absentastudyinwhichthedifferentmeasuresare sex with men (MSM). However, we disagree with their key given to the same population and the degree of association conclusionthat‘‘researcherswouldbewell-servedtoconsider measured.Theoutcomemeasuresalsovaryacrossstudies,and abandoningtheinvestigationofIHasapredictorofriskysexual thereareinsufficientdatatodeterminehowtheoutcomeswere behaviorinordertofocusresourcesonmorepromisinglinesof actuallymeasuredandwhattheirdistributionswere.Thefact research.’’Therearesevenconcerns. that some IH scales factor into several subscales (Ross & First, their approach to meta-analysis used correlation Rosser,1996),somebutnotallofwhichmayberelatedtorisk coefficients;however,correlationcoefficientsinmeta-analy- behaviors,furthercomplicatesacomparisonofIHmeasures sesarenotandcannotbeeffectsizes.Suchstandardizedeffect (asNewcombandMustanskinote). sizes‘‘cannotreliablymeetthegoalsofmeta-analysis’’(Cum- Third,onestudyusedcompulsivesexualbehavior(CSB)as mings,2004,p.597)andGreenland,Maclure,Schlesselman, theoutcomemeasure.Thatisinappropriatesinceitsuperim- Poole, and Morgenstern (1991) note that the arguments posesadeterministicrelationshipbetweenCSBandriskysex, ‘‘against traditionally standardized coefficients apply with nottheappropriateprobabilisticone.Unfortunately,thestrong evengreaterforceagainsttheuseofcorrelationcoefficientsas correlationthereprobablycontributedtotheobservedeffect. measures of effect’’(p. 392). They are inappropriate to use Fourth,therecommendationnottostudyIHfurtherisbased because(1)theyareboundedbythedistributionalpropertiesof upontheirobservingnodirectrelationshipbetweenIHandHIV each individual study; (2) following from that, a one SD risk. However, Newcomb and Mustanski readily acknowl- changeinonesampleis,bydefinition,notequivalenttooneSD edgeandreferencestudiesshowingclearindirectrelationships inanotherstudy;(3)evenifacorrelationishigh,theeffectsize that,inturn,maybehelpfultothedesignofinnovativeinter- couldbequitemodest.Incontrast,alowerornocorrelation ventions. The existence of several moot correlations could couldindicateacurvilinearrelationshiporvariationarounda indicate conflicting pathways that haven’t been explored. strongerfixedeffect;and(4)analysisofcorrelationsnegates ShouldwedeterminesomecausalassociationofIHwithother proximal determinants of riskbehavior, then we will havea goodideaofhowtoappropriatelyinterveneonthosefactors. M.W.Ross(&) Ignoring IH in further risk behavior studies would prevent CenterforHealthPromotionandPreventionResearch, developmentofculturallyspecificandrelevantinterventions UniversityofTexasSchoolofPublicHealth,7000Fannin, forMSM. Suite2622,Houston,TX77030,USA e-mail:[email protected] Fifth,therathernarrowchoiceofoutcomesassumesthatIH hasadirecteffectonlyonsexualbehavior.Somedataalready B.R.S.Rosser(cid:2)D.Smolenski indicatethatIHoratleastonecomponentofitisassociated HIV/STIInterventionandPreventionStudiesProgram, withdrugandalcoholabuseinMSM,andweknowthatdrug DivisionofEpidemiologyandCommunityHealth,University ofMinnesotaSchoolofPublicHealth,Minneapolis,MN,USA andalcoholabusearerelatedtoriskysexualbehavior.Itwould 123 1208 ArchSexBehav(2010)39:1207–1208 makeclinicalaswellasresearchsensetoincludeotherrisk indirecthealthimpactsinnon-Westerncontextsandinnon- outcomes,suchasdrugandalcoholabuse,and measuresof WesternMSMsubcultures. mood and affect as effects of IH which may themselves In summary, while Newcomb and Mustanski have pro- translateintosexualriskbehavior.Further,theimpactofIHon duced a most thought-provoking article summarizing some drugandalcoholabuseandothernegativementalhealthout- studiesondirecteffectsofIHonHIVriskbehaviors,we comes(Rosser,Bockting,Ross,Miner,&Coleman,2008)are encourage researchers to explore IH further, particularly inthemselvesclinicallyimportantforMSM,whetherornot throughindirecteffectsandincontextswherehomonegativity theytranslateintoriskysexualbehavior,andweconcurwith appearsamajorbarriertoHIVpreventionforMSM.Wedonot NewcombandMustanskionthis. believe,astheysuggest,thatthe‘‘currentutilityofthiscon- Sixth,wealreadyhavedatathatsuggestthatIHoperates structforunderstandingsexualrisktakingofMSMiscalled throughinterveningvariables.Wedemonstrated(Ross,Ros- intoquestion’’forusebyHIV/STI,sexualityandmentalhealth ser,Neumaier,&thePositiveConnectionsTeam,2008)that researchers.OutsideoftheWest,homophobiaandinternalized IH is significantly associated with HIV/STI risk through at homophobiadostillmatter. least two intervening pathways (but has no direct effect). Serodiscordant unprotected analintercourse is a function of menbeinglessdisclosingoftheirHIVserostatus,itselfsig- References nificantlyassociatedwithIH;andlowercondomusethrough Cummings,P.(2004).Meta-analysisbasedonstandardizedeffectsis lowercondomself-efficacyisitselfalsosignificantlyassoci- unreliable.ArchivesofPediatricandAdolescentMedicine,156, ated with IH. There are almost certainly other important 595–597. indirectpathwaysthroughwhichIHoperatesonriskbehaviors Greenland,S.,Maclure,M.,Schlesselman,J.J.,Poole,C.,&Morgenstern, waitingtobeelucidated. H.(1991).Standardizedregressioncoefficients:Afurthercritiqueand reviewofsomealternatives.Epidemiology,2,387–392. Finally, the most troubling aspect of their recommenda- Newcomb,M.E.,&Mustanski,B.(2009).Moderatorsoftherelationship tiontoabandonIHandHIVriskresearchisthatthestudies between internalized homophobia and risky sexual behavior in reviewedwereoverwhelminglyNorthAmerican.Studiesnot menwhohavesexwithmen:Ameta-analysis.ArchivesofSexual inEnglishwereexcluded.Worldwide,homonegativityandIH Behavior.doi:10.1007/s10508-009-9573-8. Ross,M.W.,&Rosser,B.R.S.(1996).Measurementandcorrelatesof remainhugethreatstothewell-beingofMSM.Aswewrite, internalized homophobia: A factor analytic study. Journal of thereisawaveofhomonegativitysweepingAfrica.Uganda ClinicalPsychology,52,15–21. hasintroducedlegislationwhichprovidesforthedeathpenalty Ross,M.W.,Rosser,B.R.S.,Neumaier,E.R.,&thePositiveConnections formalehomosexualacts.InSenegal,itisreportedthatthe Team. (2008). The relationship of internalized homonegativity to unsafesexualbehaviorinHIVseropositivemenwhohavesexwith bodiesofgaymenhavebeendugupanddesecrated.Outside men.AIDSEducationandPrevention,20,547–557. NorthAmericaandselectedWestEuropeanandAustralasian Ross,M.W.,Smolenski,D.J.,Kajubi,P.,Mandel,J.S.,McFarland,W., countries,homonegativityappearsanoverwhelmingandever- & Raymond, H. F.(2010). Measurement of internalized homo- present physical and psychological threat to MSM. To rec- negativity in gay and bisexual men in Uganda: Cross-cultural propertiesoftheInternalizedHomonegativityScale.Psychology, ommend not studying IH and its still poorly understood HealthandMedicine,15,159–165. associationswithriskbehavior,simplybecausestudiesinthe Rosser,B.R.S.,Bockting,W.O.,Ross,M.W.,Miner,M.H.,&Coleman, U.S. failed to observe direct relationships between IH and E. (2008). The relationship between homosexuality, internalized sexual risk behavior, makes no sense. Our study on IH in homo-negativity,andmentalhealthinmenwhohavesexwithmen. JournalofHomosexuality,55,185–203. Ugandan MSM (Ross et al., 2010) will, we hope, stimulate other researchers to explore the construct and its direct and 123 ArchSexBehav(2010)39:1209–1211 DOI10.1007/s10508-010-9655-7 LETTER TO THE EDITOR The Importance of Measuring Internalized Homophobia/ Homonegativity: Reply to Ross, Rosser, and Smolenski (2010a) MichaelE.Newcomb • BrianMustanski Publishedonline:16July2010 (cid:2)SpringerScience+BusinessMedia,LLC2010 Ross, Rosser, and Smolenski (2010a) provide a commentary continuous,theproduct-momentcorrelationcoefficientis regardingourrecentlypublishedmeta-analysisoftherelation- thestraightforwardlyappropriateeffectsizestatistic.Indeed, ship between internalized homophobia (IH) and risky sexual virtuallyallsuchresearchfindingswillbereportedin behaviorinmenwhohavesexwithmen(MSM)(Newcomb& termsofthecorrelationcoefficientinoriginalstudies, Mustanski,2009).Rossetal.raisedanumberofmethodological soitwillrarelybesensibletoconsideranyothereffect andconceptualconcernsaboutthemeta-analysis,andtherefore sizestatistictorepresentthem’’(p.63). disagree withourconclusion thatresearchers wouldbewell- LipseyandWilsonfurtherdiscusstheneedtostandardizeES servedtofocustheirresourcesonotherpredictorsofriskysexual r usingFisher’srtoztransformationinordertoovercomeaprob- behaviorthataremorepromisingintermsoftheirpredictive lematicstandarderrorformulationwhichisusedinweighting ability.Wewouldliketoaddressanumberoftheseconcernsin effectsizesbasedonsamplesize.Byusingstandardizedand ordertofurthersupportourconclusionthattheliteraturedoesnot weightedES statistics,theanalystisabletomeaningfullycom- supportacurrentsignificantbivariateassociationbetweenIH r bine results across studies and make conclusions aboutcom- andriskysexualbehavior.Wefurtherbelievethatfewviable binedoveralleffectsize. modelshavebeenproposedtoexaminethepotentialindirect Rossetal.alsonotedthattheIHmeasuresusedinthemeta- effectsofIHonrisk. analysiswerenotcomparable.Whileitistruethattherehasbeen First,Rossetal.raiseconcernsabouttheuseofcorrelation somedisagreementintheliteratureabouthowbesttomeasureIH, coefficients (ES) in meta-analysis and state that‘‘correlation r oneofthepurposesofmeta-analysisistocombineresultsfrom coefficientsinmeta-analysisarenotandcannotbeeffectsizes.’’ multiplestudiesonthesamerelationshipinordertoaccountfor Infact,correlationcoefficientsarecommonlyusedaseffectsize methodological and measurement differences between these inmeta-analysisandareadequatemeasuresofeffectsizefor studies.CombiningstudiesusingmultiplemeasuresofIHthat meta-analysisgiventhatextrastepsaretakeninordertoover- mayassessslightlydifferentcomponentsoftheconstructallow come certain problematic properties of this statistic. In their theanalysttomakemoregeneralconclusionsaboutIHthanwould widely-usedbookonmeta-analysis,LipseyandWilson(2001) bepossiblebyexaminingeachstudyindividually.Assuch,weare stated: betterabletomakegeneralconclusionsaboutthecurrentstateof …whentheresearchfindingstobemeta-analyzedinvolve the literature on the association between IH and risky sexual bivariaterelationshipsinwhichboththevariablesare behaviorbecausewedidnotrelyonasinglemeasureofIH.Infact, itiscommonpracticeinmeta-analysistousedifferentmeasuresof thesameconstruct.Forexample,awidely-citedmeta-analysisof M.E.Newcomb(&) theeffectsofschool-basedinterventionprogramsonaggressive DepartmentofPsychology(MC285),UniversityofIllinois behavior included multiple different measures of aggressive atChicago,1007W.HarrisonStreet,Chicago,IL60607,USA behavior,such asphysicalaggression, externalizing problems, e-mail:[email protected] fighting,bullying,actingout,disruptiveness,and/orconductand B.Mustanski disciplineproblems(Wilson,Lipsey,&Derzon,2003). DepartmentofPsychiatry,UniversityofIllinoisatChicago, Chicago,IL,USA 123 1210 ArchSexBehav(2010)39:1209–1211 Whenmultiplemeasuresareused,itisalsopossibletotestfor betweenIHandriskysexualbehavior.Rossetal.state,‘‘Shouldwe differencesineffectsbetweenmeasures.Toconsiderthis,we determine some causal association of IH with other proximal includedthemostcommonlyusedmeasureofIH,theNungesser determinantsofriskbehavior,thenwewillhaveagoodideaof HomosexualAttitudesInstrument(NHAI),asamoderatorof howtoappropriatelyinterveneonthosefactors.’’Here,weseem therelationshipinafollow-upanalysis.Thisvariablewasnota toagreethatsuchproximalfactorsshouldbecomethefocusof significant moderator, indicating that the NHAI was not dif- scarcepublichealthresourcesovermoredistalfactorsasIHwould ferentiallyassociatedwithriskysexualbehaviorascomparedto thenbe.WherewedodisagreewithRossetal.isintheirstatement othermeasures. that‘‘IgnoringIHinfurtherriskbehaviorstudieswouldpre- Ross et al. also expressed concern about varying measures vent development of culturally specific and relevant interven- of risky sexual behavior, specifically the inclusion of a study tionsforMSM.’’First,webelievethereisadangerindeveloping assessingcompulsivesexualbehaviorastheoutcomevariable. interventioncontentbasedontheassumptionthattheinternali- Themajorityofthestudiesincludedinthemeta-analysisassessed zationofanti-gayandheterosexiststigmaiscentraltotheunder- unprotectedanalsexastheriskvariable,butseveralstudies standingofMSMriskbehaviors,whentheevidencedoesnotbear includedmeasuresofunprotectedvaginalandoralsexandnum- thisout.Thismayleadtoamisdirectionofscarceintervention berofsexualpartnersascomponentsofsexualrisk.Again,meta- resources.Forexample,itispossiblethattheexternalmanifes- analysisshouldhelptoaccountformeasurementdifferences tationsofthisstigma,suchaslegislationthatpreventstwomen acrossmultiplemeasuresofthesameorsimilarconstructs.While frombeingmarried,mayrepresentcriticalstructuralfactorsthat itistruethatcompulsivesexualbehaviorislessfrequentlyusedas increaseriskforHIVamongMSM.However,thispolicymaynot asexualriskvariable,removingthisstudyfromthemeta-analysis needtobepsychologicallyinternalizedbytheindividualinorder didnotalterthepatternorsignificanceofanyofthefindings, forittoaffectrisktaking.Inouropinion,thisdeservesfurther includingthemoderatingeffectofyearofdatacollectiononthe investigation. Second, we do not believe that a cultural factor relationshipbetweenIHandriskysexualbehavior. needstobecorrelatedwithariskoutcomeinordertobeconsidered Rosset al. suggested that, evenif there is no direct rela- for inclusion in making an intervention culturally specific and tionship between IH and risky sexual behavior, the indirect relevant. For example, in the CDC’s ADAPT framework for relationshipsreportedinsomestudiesrepudiateourconclu- adaptinganexistingHIVinterventiontoatargetgroup(McKleroy sionsaboutthevalueoffurtherresearchonthisassociation. etal.,2006),changesweredeemednecessaryinordertomake Inourarticle,wecitedstudiesthathavefoundevidencefor theinterventionculturallyappropriateandtoaddressuniquerisk indirecteffectsofIHonriskysexualbehaviorinMSM.How- factorsandbehavioraldeterminants.IfIHisacorepartofthe ever, relatively few studies have provided evidence for and culture,thenitmayverywellneedtobeaddressedinanHIV replicated these indirect effects, and none have provided a preventionprogramevenifitisnotariskdeterminant.Individuals strong rationale for why an indirect effect may exist in the maydisagreeabouttheextenttowhichIHiscurrentlyacorepart absenceofadirectbivariaterelationship.Such‘‘inconsistent ofgayculture. mediation’’occurswhenthe indirectand directeffectarein Rossetal.alsonotedthatourmeta-analysisonlyexamined oppositedirectionsandthereforecanceleachotheroutinthe onepotentialoutcomeassociatedwithIH,andthatanindirect estimation of the direct relationship (i.e., a ‘‘suppression’’ relationshipbetweenIHandriskysexualbehaviormayexistvia effect)(MacKinnon,Krull,&Lockwood,2000).Inthesemore theassociationbetweenIHandotheroutcomes,suchasmental rarelyreportedcasesofsuppression,itisparticularlyimportant health and substance use. As is noted in our discussion in the for researchers to explicate a reasonable model for under- meta-analysis of the relationship between IH and risky sexual standingthereasonandimplicationsforsuppression.Wefound behavior,wealsoconductedameta-analysisoftherelationship nosuchexplanationsorevenpatternsofcorrelationsconsis- between IH and internalizing mental health problems, which tent with suppression in the published papers we reviewed foundamoderateeffectsizeforthisrelationship(Newcomb& reportingindirecteffects.InreadingthearticlecitedbyRoss Mustanski,2010).WeagreewithRossetal.thattheeffectsofIH et al. as evidence for indirect effects, we found no report, onmentalhealthareimportanttoconsiderinbothresearchand description, or explanation of an indirect effect between IH clinicalwork.Wealsoconsideredconductingmeta-analysesof andsexualrisktaking(Rossetal.,2010b).Webelievemuch therelationshipsbetweenIHandalcoholanddruguse,butthere strongerempiricalevidenceandtheoreticaljustificationforthe werenotenoughstudiesontheserelationshipstowarrantmeta- existenceofanimportantindirectrelationshipbetweenIHand analysis.Whileweagreethattheremaybearelationshipbetween risky sexual behavior is needed before public health deci- IHandsubstanceuse,wecautionreadersfromdrawingthecon- sions should be made based on the existence of an indirect clusionthatthisisevidenceforanindirectlinkbetweenIHand relationship. riskysexualbehavior.Thedirectassociationbetweensubstance Furthermore,ifanindirectrelationshipwasfirmlyestablished, useandsexualriskhasnotbeenconsistentintheliterature,par- thismakesthecaseforfocusingresearchandinterventions ticularlyforalcoholuse(Cooper,2006),andconflictingfindings onthosemoreproximalprocessesthatmediatetherelationship havebeenreported.Similarly,theassociationbetweensexualrisk 123 ArchSexBehav(2010)39:1209–1211 1211 andmentalhealthoraffectivestateshasbeeninconsistent(Cre- promising in terms of empirical or theoretical support given paz&Marks,2001;Mustanski,2007). scarcepublicresourcesavailabletostudytheseprocesses. Finally,weagreewithRossetal.thatIHmaybedifferently experiencedinnon-Westernculturesandthatitplausiblymay predictriskysexualbehaviorinothercultures.Weonlyincluded References studies in our meta-analysis from Western cultures that were Cooper,M.L.(2006).Doesdrinkingpromoteriskysexualbehavior? conductedinEnglishasbothlanguageandculturemayaffectthe CurrentDirectionsinPsychologicalScience,15,19–23. experienceofIHandthusrenderitimpossibletomakemean- Crepaz,N.,&Marks,G.(2001).Arenegativeaffectivestatesassociated ingfulcomparisonsacrossstudies.Ourconclusionthat‘‘research- withHIVsexualriskbehaviors?Ameta-analyticreview.Health erswouldbewell-servedtoconsiderabandoningtheinvestigation Psychology,20,291–299. Lipsey,M.W.,&Wilson,D.B.(2001).Practicalmeta-analysis.Thou- of IH as a predictor of risky sexual behavior’’applies to the sandOaks,CA:SagePublications. countriesorculturesinwhichthestudiesincludedinthemeta- MacKinnon,D.P.,Krull,J.L.,&Lockwood,C.M.(2000).Equivalence analysiswereconducted.However,weencourageresearchersto ofthemediation,confoundingandsuppressioneffect.Prevention continuetobemindfulofthedifferencebetweenIH,whichisan Science,1,173–181. McKleroy,V.S.,Galbraith,J.S.,Cummings,B.,Jones,P.,Harshbarger, internalpsychologicalprocess,andexternalanti-gaybehaviors C.,Collins,C.,etal.(2006).Adaptingevidence-basedbehavioral andpolicies.Victimizationresultingfromhomophobiadoesnot interventions for new settings and target populations. AIDS always lead to IH, and determining which individuals under EducationandPrevention,18,59–73. whichconditionsinternalizeheterosexismwouldbehelpfulfor Mustanski,B.(2007).TheinfluenceofstateandtraitaffectonHIVrisk behaviors:AdailydiarystudyofMSM.HealthPsychology,26, cliniciansworkingwiththiscommunity.Wedoverymuchagree 618–626. withRossetal.abouttheimportanceofstudyingthehealthand Newcomb,M.E.,&Mustanski,B.(2009).Moderatorsoftherelationship development of LGBT individuals and MSM in non-Western betweeninternalizedhomophobiaandriskysexualbehaviorinmen contexts.ItisnotourintenttoarguethatIHisnotworthstudying who have sex with men: A meta-analysis. Archives of Sexual Behavior,doi:10.1007/s10508-009-9573-8. outsidetheWestinresearchonriskysexualbehaviororthatIH Newcomb,M.E.,&Mustanski,B.(2010).Therelationshipbetween doesn’tmatterasaconstructthatisrelevanttomentalhealthand internalized homophobia and internalizing mental health prob- socialjustice.However,itisourintenttoencourageresearchersto lems:Ameta-analyticreview(underreview). bemindfulofthescienceandrationaleinfluencingtheirdecisions Ross,M.W.,Rosser,B.R.S.,&Smolenski,D.(2010a).Theimportance ofmeasuringinternalizedhomophobia/homonegativity[Letterto todoso. the Editor]. Archives of Sexual Behavior, doi:10.1007/s10508- In conclusion, we stand by the conclusions of our meta- 010-9634-z. analysisaboutanoveralllackofdirectassociationbetweenIH Ross,M.W.,Smolenski,D.J.,Kajubi,P.,Mandel,J.S.,McFarland,W., andriskysexualbehaviorinMSMinstudiespublishedinEng- &Raymond,F.H.(2010b).Measurementofinternalizedhomo- negativity in gay and bisexual men in Uganda: Cross-cultural lish withWesternsamples. Wealso believe thereisnotwell propertiesoftheInternalizedHomonegativityScale.Psychology, developedevidenceorexplanationsforindirecteffectsinthe Health,andMedicine,15,159–165. absenceofdirecteffects.WhiletheexperienceofIHmayhave Wilson,S.J.,Lipsey,M.W.,&Derzon,J.H.(2003).Theeffectsof importantimplicationsforotherhealth-relatedoutcomes(e.g., school-based intervention programs on aggressive behavior: A meta-analysis.JournalofConsultingandClinicalPsychology,71, mentalhealth),wemaintainthatisitcriticalforfutureresearchto 136–149. focusoninvestigatingpredictorsofsexualriskthataremore 123 ArchSexBehav(2010)39:1213–1215 DOI10.1007/s10508-010-9658-4 LETTER TO THE EDITOR Depressive Symptoms Among Same-Sex Oriented Young Men: Importance of Reference Group RitchC.Savin-Williams • KennethM.Cohen • KaraJoyner • GerulfRieger Publishedonline:21July2010 (cid:2)SpringerScience+BusinessMedia,LLC2010 Arecentmeta-analysisindicatedstrongsupportforthecom- Udry & Chantala, 2005). Indeed, the effect size for male monlyobservedfindingthat,incontrasttoheterosexualmen, sexualorientationdifferenceappearstobeamongthelargest same-sexorientedmenareatgreaterriskformentalhealthprob- sex-dimorphicdiscrepanciesinthedevelopmentalliterature lems,especiallyaffectivedisorders(Kingetal.,2008).Themost (Bailey&Zucker,1995).Inarecentreview,Bailey(2009) commonexplanationsforthesefindingsaretheincreasedprev- notedthatsame-sexorientedmenare,onaverage,moresex alenceofsocietalprejudice,stigma,andviolencethatallegedly atypicalintheirself-concepts,motorbehavior,bodymove- generate‘‘minority-stress’’effects (King et al., 2008; Meyer, ments,occupationalcareers,andrecreationalintereststhan 2003;Saewyc,2007).Hatzenbuehler(2009,p.707)notedthat heterosexualmen.AccordingtoarecentFinnishtwinstudy, stigma-relatedstress‘‘getsundertheskin’’byinducingeleva- substantialgeneticeffectsaccountforvariationinbothsex- tions in emotional dysregulation, which, in turn, confers risk atypical behavior and adult sexual orientation, suggesting forincreasedpsychopathologyamongnonheterosexualpopu- that at least part of this co-variation is due to biological lations.Althoughthisexplanationisreasonable,fewother influences(Alankoetal.,2010). hypotheses have been proposed for the apparent elevation of Given therobustfindingthatinsomerespectssame-sex depressivesymptomsamongsame-sexattractedindividuals. orientedmenaremoresimilartoheterosexualwomenthan Thealternativeconsideredhereisthatmentalhealthdiscrep- heterosexualmen,wespeculatethatthebasicprincipleofa anciesamongmalesexualorientationgroupsareillusorybecause correctlychosenreferencegroupiscriticalinascertainingthe theyare derived frominappropriate(i.e., same biologicalsex) existence of male sexual orientation differences in mental groupcomparisons.Weadoptinasex-inversionperspective— health. To our knowledge, sex-atypical endorsement of that cross-sex comparisons are more suitable for situations in sexuallydimorphicmentalhealthpatternshasnotbeensys- whichnonheterosexualmenareassessedonsexuallydimorphic tematically addressed. We hypothesize that same-sex ori- variables. ented men more closely resemble mental health patterns An extensive literature documenting cross-sex behavior observedamongheterosexualwomenthanmen.Thiswould amongnumerousdomainshasconsistentlyreportedthat,asa particularlybethecaseformentalhealthdisordersthatare group,nonheterosexualmenscoresignificantlymoresimilar especiallysexdimorphic,suchasdepression.Indeed,ameta- totheoppositesexthanheterosexualmenonvariousmea- analysisofsexdifferencesindepressionrevealedthat,begin- sures of gender nonconformity (Bailey & Zucker, 1995; ninginearlyadolescenceandcontinuingthroughadulthood,the Cohen,2002;Lippa,2005;Rieger&Savin-Williams,2010; prevalence of depression is more than twice as high among womenasmen(Twenge&Nolen-Hoeksema,2002).Thissex differencewasreplicatedinKingetal.’s(2008)meta-analysis R.C.Savin-Williams(&)(cid:2)K.M.Cohen(cid:2)G.Rieger fordepression,anxiety,andsuicidality. HumanDevelopment,CornellUniversity,MVRHall, Ithaca,NY14853,USA To address this possibility, we analyzed data drawn from e-mail:[email protected] Wave3ofAddHealth,acomprehensive,representativeschool- basedstudyofU.S.youth(Harrisetal.,2003;Udry&Bearman, K.Joyner 1998). For Wave 3 (2001–2002), the project re-interviewed DepartmentofSociology,BowlingGreenStateUniversity, BowlingGreen,OH,USA 15,197oftheoriginalWave1in-homeparticipants,nowaged 123 1214 ArchSexBehav(2010)39:1213–1215 18–28years.TheresponseratesforWaves1and3were79% depressed.’’Itemswereratedona4-pointscale,rangingfrom and77%,respectively.Allincludedparticipantshadtoreport 0(never/rarely)to3(mostofthetime/allofthetime). the same sex at both waves to qualify for the sample. After Analyses incorporated survey design procedures to take excluding participants with missing data on main variables intoaccountAddHealth’sweighting(usingthegrandsample (sexualorientationanddepression)andnolongitudinalgrand longitudinal weight for participantsinWaves1 and3)and sample weight, the final sample was 13,992, of which 6,615 cluster design. Means and their 95% confidence intervals were heterosexual and nonheterosexual men and 6,351 were (CIs)werecalculatedtoexaminehowdepressivesymptoms heterosexual women (the remainder were nonheterosexual differedbysexandsexualorientation.Consideredasagroup women,notincludedinthisstudy). andthenasthreesexualorientationsubgroups(mostlyhet- Interviews were conducted by professional staff who erosexual,bisexual,andhomosexual),nonheterosexualmen enteredsubjects’answersonalaptopcomputer.Forsensitive were compared to heterosexual men. These comparisons questions, such as sexual orientation, interviewers handed werethenrepeatedwithheterosexualwomenasthereference participantsthelaptoponwhichtheycompletedacomputer- group. assistedself-interview;thisenabledtheyoungadultstoenter As expected, when compared to heterosexual men, non- responses to questions that appeared on screen and were heterosexual men and heterosexual womenreportedsignifi- heard on tape with earphones. Participants were asked to cantlyhighervalues for depressivesymptoms (Table1).By choose the description that best fits how they think about nonheterosexual subgroup, mostly heterosexual and homo- themselves. Options were 100% heterosexual (straight); sexualmenendorsedhighervaluesofdepressivesymptoms mostly heterosexual (straight), but somewhat attracted to thanheterosexualmen.Despitereportingthehighestdepres- peopleofyourownsex;bisexual—thatis,attractedtomen sive value, bisexual men did not differ significantly from andwomenequally;mostlyhomosexual(gay),butsomewhat heterosexualmalesbecauseoftheirsmallsamplesize.Non- attracted to people of the opposite sex; 100% homosexual heterosexual men as a group and all nonheterosexual sub- (gay);ornotsexuallyattractedtoeithermalesorfemales.The groupsdidnotdiffersignificantlyfromheterosexualwomenin small number of participants who failed to report attrac- depressivesymptoms. tionsweredropped(n=49).Duetothesmallsamplesizeof Giventhewidelydocumentedsexatypicalityofsame-sex mostly homosexuals, they were combined with 100% attracted men across multiple domains, it may not be sur- homosexuals.Depressivesymptomswereassessedwitha10- prising that the elevation of affective symptoms among item derivative CES-D scale (Meadows, Brown, & Elder, nonheterosexualmencloselyresembledthatofheterosexual 2006).Scalealphawas0.81.Includedwerequestionsabout women. Although this finding has not been previously thefrequencyofeventsduringthepastweeksuchas‘‘Youfelt reported, it can be detected by re-analyzing King et al.’s Table1 Sexualorientationand Males(n=6615) depressivesymptomsbymale andfemalereferencegroup Percentage/M SD 95%CI t Sexualorientation(males) 100%heterosexual(n=6238) 0.944 (.936–.952) – Allnonheterosexuals(n=377) 0.056 (.048–.064) – Mostlyheterosexual(n=204) 0.032 (.026–.038) – Bisexual(n=41) 0.006 (.003–.008) – Homosexual(n=132) 0.018 (.013–.023) – Depressivesymptomsa 100%heterosexualmales 4.61 0.07 (4.47–4.77) – Allnonheterosexualmales 6.05 0.31 (5.42–6.68) 4.65*** Mostlyheterosexualmales 6.11 0.48 (5.16–7.08) 3.22** Bisexualmales 6.32 0.97 (4.34–8.31) 1.81 Homosexualmales 5.84 0.45 (4.94–6.75) 2.69** **p\.01;***p\.001 100%heterosexualfemales(n=6351) 5.90 0.09 (5.71–6.10) – (two-tailedsignificancetest Allnonheterosexualmales 6.05 0.31 (5.42–6.68) 0.45 ofdifferenceindepressive symptoms 100%heterosexualmales 4.61 0.07 (4.47–4.77) -12.15*** a Absoluterange,0–30. Mostlyheterosexualmales 6.11 0.48 (5.16–7.08) 0.44 Symptomsbetweenthisgroup Bisexualmales 6.32 0.97 (4.34–8.31) 0.44 and100%heterosexualmales Homosexualmales 5.84 0.45 (4.94–6.75) -0.15 orfemales 123 ArchSexBehav(2010)39:1213–1215 1215 (2008)meta-analysis:thepercentofdepressionamongnon- atypicalbehaviorandadultsexualorientation.ArchivesofSexual heterosexual men (13.7%) was considerably closer to het- Behavior,39,81–92. Bailey,J.M.(2009).Whatissexualorientationanddowomenhaveone? erosexual women (10%) than to heterosexual men (5.9%). InD.A.Hope(Ed.),Nebraskasymposiumonmotivation:Vol.54: Thus, heterosexual women, rather than heterosexual men, Contemporaryperspectivesonlesbian,gay,andbisexualidentities may constitute a more appropriate reference group when (pp.43–63).NewYork:Springer. exploringfemale-typicalmentalhealthconcerns. Bailey,J.M.,&Zucker,K.J.(1995).Childhoodsex-typedbehaviorand sexualorientation:Aconceptualanalysisandquantitativereview. Whereas a minority-stress model would explain these DevelopmentalPsychology,31,43–55. findingsbyhighlightingasharedpatternofdevaluation,pre- Cohen, K. M. (2002). Relationships among childhood sex-atypical judice, and discrimination among nonheterosexual men and behavior, spatial ability, handedness, and sexual orientation in heterosexualwomen,webelievethatasex-inversionper- men.ArchivesofSexualBehavior,31,129–143. Harris,K.M.,Florey,F.,Tabor,J.,Bearman,P.S.,Jones,J.,&Udry,J. spectiveisamoreparsimoniousinterpretation.Theunderly- R.(2003).Thenationallongitudinalstudyofadolescenthealth: ing mechanisms, as well as reasons cross-sex shifts do not Research design.ChapelHill, NC:Carolina Population Center. extendtoallsex-dimorphicbehaviors(e.g.,desirefornovel Availableat:http://www.cpc.unc.edu/projects/addhealth/design. sexpartners),remainuncertain.WilsonandRahman(2005) Hatzenbuehler,M.L.(2009).Howdoessexualminoritystigma‘‘Get undertheskin’’?Apsychologicalmediationframework.Psycho- suggested that neuropsychological differences between the logicalBulletin,135,707–730. sexes—whichinvolvethehypothalamus,limbicsystem,and King,M.,Semlyen,J.,Tai,S.S.,Killaspy,H.,Osborn,D.,Popelyuk,D., neurotransmitters, as well as higher neurological centers— etal.(2008).Asystematicreviewofmentaldisorder,suicide,and may explain similarities between homosexual men and het- deliberate self harm in lesbian, gay and bisexual people. BMC Psychiatry,8.doi:10.1186/1471-244X-8-70. erosexualwomen.Weadditionallysuspectelevatedfemale- Lippa,R.A.(2005).Sexualorientationandpersonality.AnnualReview typical sex-dimorphic traits that lead to increased endorse- ofSexResearch,16,119–153. mentofaffectivesymptoms,suchasinterpersonalsensitivity, Meadows, S. O., Brown, J. S., & Elder, G. H. (2006). Depressive emotion-focused ruminative coping modality, and reporting symptoms,stress,andsupport:Gendered trajectories fromado- lescencetoyoungadulthood.JournalofYouthandAdolescence, style. 35,89–99. Oncere-contextualized,thedepressivesymptomsrepor- Meyer, I. H. (2003). Prejudice, social stress, and mental health in ted by nonheterosexual men are not remarkable or even lesbian, gay, and bisexual populations: Conceptual issues and unexpected.Byquestioningtheappropriatereferencegroup researchevidence.PsychologicalBulletin,129,674–697. Rieger,G.,&Savin-Williams,R.C.(2010).Sexualorientation,gender forcontrastinglevelsofmentalhealth,ourfindingscontrib- nonconformity,andwell-being.Manuscriptsubmittedforpubli- utetoagrowingcallfordepathologizingindividualswhoare cation. notheterosexual(Savin-Williams,2005;Savin-Williams& Saewyc, E. M. (2007). Contested conclusions: Claims that can (and Joyner,2010). cannot)bemadefromthecurrentresearchongay,lesbian,and bisexualteensuicideattempts.JournalofLGBTHealthResearch, 3,79–87. Acknowledgments ThisresearchuseddatafromAddHealth,apro- Savin-Williams,R.C.(2005).Thenewgayteenager.Cambridge,MA: gram project directed by Kathleen Mullan Harris and designed by HarvardUniversityPress. J.RichardUdry,PeterS.Bearman,andKathleenMullanHarrisatthe Savin-Williams,R.C.,&Joyner,K.(2010).Diversityamongsexual UniversityofNorthCarolinaatChapelHill,andfundedbygrantP01- orientationgroupsindepressivesymptomlevels:Examiningthe HD31921fromtheEuniceKennedyShriverNationalInstituteofChild mediatingroleofriskybehavior.Unpublishedmanuscript,Cornell Health and Human Development, with cooperative fundingfrom 23 University,Ithaca,NY. otherfederalagenciesandfoundations.Specialacknowledgmentisdue Twenge, J. M., & Nolen-Hoeksema, S. (2002). Age, gender, race, RonaldR.RindfussandBarbaraEntwisleforassistanceintheoriginal socioeconomic status, and birth-cohort differences on the Chil- design.InformationonhowtoobtaintheAddHealthdatafilesisavail- dren’sDepressionInventory:Ameta-analysis.JournalofAbnor- ableontheAddHealthwebsite(http://www.cpc.unc.edu/addhealth). malPsychology,111,578–588. No direct support was received from grant P01-HD31921 for this Udry,J.R.,&Bearman,P.S.(1998).Thenationallongitudinalstudyof analysis.Useofthisacknowledgmentrequiresnofurtherpermission adolescenthealth.RetrievedApril30,2003,fromhttp://www.cpc. fromthepersonsnamed. unc.edu/projects/addhealth/. Udry,J.R.,&Chantala,K.(2005).Riskfactorsdifferaccordingtosame- sexandopposite-sexinterest. JournalofBiosocial Science,37, References 481–497. Wilson,G.D.,&Rahman,Q.(2005).Borngay:Thepsychobiologyof sexorientation.London:PeterOwen. Alanko,K.,Santtila,P.,Witting,K.,Varjonen,M.,Jern,P.,Johansson, A., et al. (2010). Common genetic effects on childhood gender 123 ArchSexBehav(2010)39:1217–1219 DOI10.1007/s10508-010-9670-8 LETTER TO THE EDITOR The Right Comparisons in Testing the Minority Stress Hypothesis: Comment on Savin-Williams, Cohen, Joyner, and Rieger (2010) IlanH.Meyer Publishedonline:31August2010 (cid:2)SpringerScience+BusinessMedia,LLC2010 Savin-Williams,Cohen,Joyner,andRieger(2010)suggest moresuitableforsituationsinwhichnonheterosexualmenare that purported ‘‘mental health discrepancies among male assessedonsexuallydimorphicvariables.’’Thisclaimisbased sexual orientation groups are illusory.’’They suggest that onvastresearchthatshowedthatgay andbisexualmen are their findings regarding gay men’s depressive symptoms more like heterosexual women than heterosexual men on rebutminoritystressexplanationofsexualorientationmental variousmeasures(e.g.,Bailey,2009;Bailey&Zucker,1995). healthdisparities.Butasitpertainstominoritystresstheory, Forexample,Savin-Williamsetal.citeBailey(2009),whohas theirrationaleisflawed. showedthat‘‘same-sexorientedmenare,onaverage,moresex Minoritystressisafrequentlyusedframeworkforunder- atypical in their self-concepts, motor behavior, body move- standing observed health disparities between sexual minor- ments, occupational careers, and recreational interests than ity and heterosexual populations (Herek & Garnets, 2007). heterosexualmen.’’Followingthisrationale,Savin-Williams Minority stress suggests that (1) lesbians, gay men, and etal.goontoshowthatgayandbisexualmen’slevelofdepres- bisexuals(LGB)compriseadisadvantagedsocialgroupthatis sivesymptomsisnotdifferentfromthatofheterosexualwomen. subject to stigma and prejudice; (2) stigma and prejudice Thus, they conclude‘‘once recontextualized, the depressive relatedtosexualorientationpredisposeLGBtoexcessstress; symptomsreportedbynonheterosexualmenarenotremark- and(3)inturn,thisexcessstressmayleadtoadversehealth ableorevenunexpected.’’ outcomesand,thus,observedhealthdisparities(Meyer,2003). Evenifonewouldconcedethepremiseoftheargument— Studiesusingtheminoritystressperspectivehaveconsistently that‘‘same-sexorientedmenaremoresimilartoheterosexual documentedmentalhealthdisparitiesbetweenLGBandhet- womenthanheterosexualmen’’—thereisnomerittotheclaim erosexual populations. This is evidenced, for example, in thatthesefindingsunderminetheminoritystressperspective. studies that used meta-analysis to summarize this literature The main reasonfor this is that Savin-Williamset al. make (Kingetal.,2008;Meyer,2003). theirargumentbasedononesexualminoritysubgroup—gay Savin-Williamsetal.arguedthattheobserveddisparityis andbisexualmen,excludinglesbiansandbisexualwomen— faultybecauseresearchershaveusedaninappropriaterefer- andoneoutcome—depressivesymptoms,excludinganxiety ence group in studying mental health outcomes in gay and andsubstanceusedisorders.Thischerry-pickedcomparisonis bisexualmen.Typically,researcherscomparegayandbisex- not sufficient to test minority stress theory. Minority stress ualmenwithheterosexualmenwhentheydocumentmental rests on sociological theory that links social structure with healthdisparities;Savin-Williamsetal.suggestthatgayand healthoutcomes(throughtheimpactofstress).Therefore,it bisexualmenshouldbecomparedwithheterosexualwomen, makes predictions about differentially situated groups (dis- notmen.Thereasonforthisisthat‘‘cross-sexcomparisonsare advantaged versus advantaged groups) and predicts similar patternsacrossvariousmentaldisorders(Schwartz&Meyer, 2010).Savin-Williamsetal.erronbothcounts. I.H.Meyer(&) First, minority stress suggests that sexual minorities are MailmanSchoolofPublicHealth,ColumbiaUniversity, sociallydisadvantagedinoursocietyduetohomophobiaand AllanRosenfieldBuilding,722West168thStreet, heterosexismasagroup—thatis,acrossallsubgroups,such Room533,NewYork,NY10032,USA e-mail:[email protected] asthosedefinedbygender,race/ethnicity,etc.Forthestudy 123 1218 ArchSexBehav(2010)39:1217–1219 ofminoritystress,therefore,thegroupscomparedoughttobe heterosexualwomeninpatternsofsubstanceuseddisorders.In allsexualminorities—menandwomen—versusallhetero- fact,thisisnotthecase:gayandbisexualmenhaveamuch sexuals.Similarly,ifgenderinequalitywasstudied,theref- higherprevalenceofsubstanceuseproblems,includingsub- erence group for women would be men; if race/ethnic stance use disorders,than heterosexual women (King et al., inequality was studied, the reference group for race/ethnic 2008;Meyer,2003).Asaresult,whenallmentaldisordersare minorities would be whites. This is because in each com- considered,gayandbisexualmenhavehigherlevelsofdis- parisonweareinterestedintheaverageeffectonthedisad- orderthanheterosexualwomen.Thisfindingisconsistentwith vantagedversusadvantaged,thatis,acrossdiversesubgroups the minority stress hypothesis but not with Savin-Williams within (Schwartz & Meyer, 2010). Savin-Williams et al.’s etal.’sreferencegrouphypothesis. hypothesis is refuted when one examines LGB as a group In summary, examining the evidence on sexual orienta- versusheterosexualsasagroup.Consistentwiththeminority tion,stress,anddisorder,wefindthatsexualminoritieshave stress hypothesis, in such comparisons, LGB populations greaterexposuretostress(Meyer,Schwartz,&Frost,2008) have higher rates of mental disorders than heterosexuals andtheyhavetheexpectedresultanthigherratesofdisorder (Kingetal.,2008;Meyer,2003). whencomparedwithheterosexuals(Kingetal.,2008;Meyer, Second,minoritystress(andsocialstresstheorymoregen- 2003).WhenweexamineSavin-Williamsetal.’sreference erally)isasociologicaltheorythatpredictsthatdisadvantaged grouphypothesiscarefullyinlightofthisevidence,wemust social status affects the aggregate of mental disorders, rather refuteit. than any particular disorder—we are interested in whether a Finally,Savin-Williamsetal.contendthattheirarguments disadvantagedgroupmemberhasanydisorderthatiscausedby ‘‘contribute to a growing call for depathologizing individ- minoritystress.Thisisbecausesocialdisadvantageandresul- ualswho are notheterosexual.’’I rejectthe implication that tantstressarethoughttobegenericpathogens.Minoritystress minoritystresstheorypathologizesLGBindividuals.Minor- does not predict a specific impact on, say, depression versus itystresstheorypositionsthesourceofstress,andtherefore anxiety and substance use disorders. Although this is not an mentalhealthproblems,asstemmingfromprevailingsocie- infallible rule, researchers should provide a good reason to tal-levelsexualstigma,prejudice,anddiscriminationandnot excludeonedisorderoranother.Areasontoexcludeadisorder a reflection of individual traits. Although some politically- would be, for example, that stress has no part in causing the motivated persons may use evidence that LGB individuals disorder.Insuchahypotheticalsituation,ifstressdoesnotplaya have a higher prevalence of disorder than heterosexuals to causalrole,theresearchercouldnotreasonablyhypothesizethat pathologize, stigmatize, and discriminate against LGB per- minoritystresswouldleadtoexcessdisorderinthedisadvan- sons,suchargumentsaremisguidedastheydefylogic.During tagedgroup. the debates that led to the removal of homosexuality as a ThisprobleminSavin-Williamsetal.’sproposedreference mentaldisorderfromDSM-IIin1973,Marmor(1980)noted grouphypothesisisreminiscentofa1970sdebateinthesoci- howillogicalitistoassociatefindingsaboutprevalenceof ologyofmentalhealthaboutgenderrolestressanddisorder. pathologyinthegroupwithpathologizingthegroupitself: GoveandTudor(1976)claimedthatahigherlevelofdepression …thebasicissue…isnotwhethersomeormanyhomo- inwomencomparedwithmenisevidenceofwomen’sgender sexualscanbefoundtobeneuroticallydisturbed.Ina role stress. Critics argued, as we argue here, that selecting society like ours where homosexuals are uniformly depression alone of all mental disorders is wrong because it treatedwithdisparagementorcontempt—tosaynoth- providesopportunisticsupportforafavoritetheory.Whenall ing about outright hostility—it would be surprising mentaldisorderswerecompared,itbecameevidentthatwomen indeedifsubstantialnumbersofthemdidnotsufferfrom andmenhavesimilarlevelsofdisorders,refutingtheproposed animpairedself-imageandsomedegreeofunhappi- gender role hypothesis (Dohrenwend & Dohrenwend, 1976). ness with their stigmatized status….It is manifestly Analternativehypothesiswasofferedforthedifferingpatterns unwarrantedandinaccurate,however,toattributesuch ofdisorders,suggestingthatmenandwomen’sstressresponse neuroticism,whenitexists,tointrinsicaspectsofhomo- differed:Womeninternalizestress,resultinginhigherlevelsof sexualityitself.(p.400) mooddisorders,andmenexternalizestress,resultinginhigher levelsofsubstanceusedisordersandantisocialbehaviors(Ro- Minoritystresstheorypointstopathogenicsocialcondi- senfield,1999). tionsthatstigmatizeLGBpeopleandtreatthemasinferiorto If we relied on these gendered patterns of difference heterosexuals.Evenattheriskthatresearchfindingscanbe between internalizing and externalizing disorders that are misusedbysome,studiesonthepsychiatricepidemiologyof characteristicofmenandwomeninthegeneralpopulation,we LGB individuals are important to help guide funding by would expect, based on Savin-Williams et al.’s proposed governmentalandotheragenciesandtodirectresearchand hypothesis, that gay and bisexual men would be similar to preventionefforts. 123