Journal of Abnormal Psychology Copyright 2005 by the American Psychological Association

2005, Vol. 114, No. 3, 471– 476 0021-843X/05/$12.00 DOI: 10.1037/0021-843X.114.3.471

Mental Health of Lesbian, Gay, Bisexual, and Heterosexual Siblings:

Effects of Gender, Sexual Orientation, and Family

Kimberly F. Balsam and Theodore P. Beauchaine Ruth M. Mickey and Esther D. Rothblum
University of Washington University of Vermont

Self-identified lesbian, gay male, and bisexual (LGB) individuals were recruited via convenience
sampling, and they in turn recruited their siblings (79% heterosexual, 19% LGB). The resulting sample
of 533 heterosexual, 558 lesbian or gay male, and 163 bisexual participants was compared on mental
health variables and their use of mental health services. Multilevel modeling analyses revealed that sexual
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

orientation predicted suicidal ideation, suicide attempts, self-injurious behavior, use of psychotherapy,
This document is copyrighted by the American Psychological Association or one of its allied publishers.

and use of psychiatric medications over and above the effects of family adjustment. Sexual orientation
was unrelated to current psychological distress, psychiatric hospitalizations, and self-esteem. This is the
1st study to model family effects on the mental health of LGB participants and their siblings.

Keywords: lesbian, gay, bisexual, mental health, sexual orientation

Recent research on lesbian, gay, and bisexual (LGB) mental disorders (e.g., Jorm, Korten, Rodgers, Jacomb, & Christensen,
health has used representative population-based samples for the 2001). Thus, combining groups may obscure differences between
first time, as a result of the inclusion of items about sexual bisexual women and men’s mental health and that of lesbians and
orientation in some national surveys. These studies have found that gay men.
people engaging in same-gender sexual behavior and/or identify- One way to locate large samples of people who self-identify as
ing as LGB are at higher risk for mental health disorders, including LGB is to recruit participants via LGB community events and
depression, anxiety, substance abuse, and suicide ideation and organizations. Such methods of data collection (e.g., Morris,
attempts. LGBs also utilize mental health services at higher rates Waldo, & Rothblum, 2001) result in large sample sizes of LGB
(cf. Cochran, 2001, for a review). Population-based samples are a people but typically lack a control group because heterosexual
major improvement in LGB research, but there are some limita- samples cannot be located through the same sources. Moreover,
tions to data from national surveys. First, the number of people LGB convenience samples have limitations in generalizability. For
who identify as nonheterosexual is extremely small. For example, example, self-identified LGB individuals may be more open about
of the 2,917 respondents who answered a single item about sexual their sexual orientation, and greater “outness” has been shown to
orientation in a recent national survey (Cochran, Sullivan, & Mays, relate to positive mental health (Morris et al., 2001). Yet there are
2003), only 41 identified their sexual orientation as lesbian or gay some advantages to studying people who self-identify as LGB via
and 32 as bisexual. Second, researchers often combine data from community sources. They are eager participants in studies for
gay and bisexual men and from lesbian and bisexual women in which the research team is known within the community, which
order to increase power. However, preliminary research has shown may help explain why national surveys obtain so few people who
that bisexuals may be at especially high risk for mental health self-identify as LGB.
In sum, national representative studies yield small numbers of
LGB respondents who are often combined into a single group to
Kimberly F. Balsam and Theodore P. Beauchaine, Department of Psy- increase statistical power. Although convenience samples yield
chology, University of Washington; Ruth M. Mickey, Department of large numbers of LGB respondents, they are nonrepresentative and
Mathematics and Statistics, University of Vermont; Esther D. Rothblum, lack a heterosexual control group. In the present study, we sought
Department of Psychology, University of Vermont.
to recruit both LGB respondents and their siblings via convenience
This study was funded by grants from the Lesbian Health Fund of the
Gay and Lesbian Medical Association, the Scrivner Award of the Ameri-
sampling. Unlike members of other oppressed groups (e.g., Afri-
can Psychological Foundation, and a University Research and Scholarship can Americans, immigrants), most LGBs have siblings who are
Grant from the University of Vermont, all awarded to Esther D. Rothblum; members of the dominant group (heterosexuals). Moreover, sib-
and grants from the Society for the Scientific Study of Sexuality, the lings are typically comparable on race, ethnicity, age, and parental
Society for the Psychological Study of Social Issues, the Scrivner Award socioeconomic status. Although the sibling methodology has been
of the American Psychological Foundation, and the Child and Adolescent used among lesbians and their sisters (Rothblum & Factor, 2001),
Psychology Foundation, all awarded to Kimberly F. Balsam. it has not been used in samples including both men and women.
We thank Laura Gordon, Ali Oun, Colleen Nilsen, Mary Mellows,
Thus, the role that sexual orientation plays in the mental health of
Kirsten Glennon, and Julianne Hellmuth for their assistance in collecting
siblings within families has not been evaluated adequately.
these data. We also thank Susan Cochran for her helpful comments on a
draft of this article. The goal of the current study was to recruit self-identified LGB
Correspondence concerning this article should be addressed to Kimberly individuals and their siblings (heterosexual or LGB) and then
F. Balsam, Department of Psychology, University of Washington, Box compare LGBs with heterosexuals within this sample on mental
351525, Seattle, WA 98195. E-mail: [email protected] health variables and the use of mental health services, while

471
 472 BRIEF REPORTS

controlling for sibling variance. We hypothesized that LGB indi- was no significant age difference between index participants and their
viduals would report more psychological distress, suicidal ide- siblings.
ation, suicide attempts, and use of mental health services than Total participants (index participants plus siblings) included 805
would their heterosexual counterparts. Given Jorm et al.’s (2001) (64.2%) women and 449 (35.8%) men. On the basis of self-reported sexual
orientation (heterosexual, bisexual, lesbian, or gay), 533 (42.5%) identified
study, we also hypothesized that bisexual women and men would
as heterosexual, 163 (13.0%) as bisexual, and 558 (44.5%) as lesbian or
report more psychological distress, suicidal ideation, and suicide
gay. Among women, 348 (43.2%) identified as heterosexual, 125 (15.5%)
attempts than lesbians and gay men. Researchers have used sib- as bisexual, and 332 (41.2%) as lesbian. Among men, 185 (41.2%) iden-
lings to investigate familial aggregation of sexual orientation (e.g., tified as heterosexual, 38 (8.5%) as bisexual, and 226 (50.3%) as gay.
Bailey, Dunne, & Martin, 2000) but rarely to investigate sexual Although most sibling participants identified as heterosexual, 19.2% iden-
orientation differences in mental health and other psychological tified as LGB. The sample was overwhelmingly European American
variables. (91.7%). Participants of color included 1.1% African American, 0.5%
Asian American, 2.5% Latino, 0.6% Native American, 2.6% biracial, and
0.8% who identified as other.
Method The questionnaire mailed to participants was entitled “Sisters and Broth-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ers Project” and did not indicate that the study focused on sexual orienta-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Large paid advertisements were placed in prominent national and state tion. Instructions stated: “This survey is being distributed in order to learn
LGB periodicals and in periodicals for LGB people of color. In addition, how the lives of adult siblings are similar or different. There is little
announcements were sent to LGB organizations listed in the resource book information about sisters and brothers and how their lives change in
Gayellow Pages (2001). The announcement was placed on LGB Web sites, adulthood.”
sent to LGB electronic mailing lists, and distributed by LGB friends and Demographic information included sex, race/ethnicity, age, and educa-
colleagues. The text of ads and announcements stated: “University LGB tional level. Sexual orientation was assessed by asking participants to
research team is looking for volunteers to complete a survey about how the self-report whether they identified as heterosexual, bisexual, lesbian, or
lives of adult sisters and brothers are similar or different. To participate, gay.
please contact . . . and indicate the number of siblings. You do not need to Psychological distress was measured by using the Brief Symptom In-
be out to your siblings to participate in this study.” Thus, the ads did not ventory (BSI; Derogatis & Spencer, 1982), a 53-item self-report scale
indicate that the study was about mental health. When interested partici- designed to measure a wide range of symptoms associated with psycho-
pants from these LGB resources contacted us, they were asked how many pathology. The BSI contains an overall score, the Global Severity Index
siblings might participate. We then mailed questionnaires and postage-paid (GSI), consisting of nine subscales (Somatization, Obsessive–Compulsive,
return envelopes to the original respondents (index participants) and their Interpersonal Sensitivity, Depression, Anxiety, Hostility, Phobic Anxiety,
siblings or, if they wished, mailed all questionnaires to the original respon- Paranoid Ideation, and Psychoticism). Each item on the BSI represents one
dents for them to mail to their siblings. In some cases, LGB participants symptom. Participants rate how much discomfort they have felt as a result
had siblings who were themselves LGB. To cast a wide net and not exclude of that symptom over the past week on a 5-point Likert scale. Responses
any siblings, we sent questionnaires to LGB siblings as well (such siblings range from 0 (not at all) to 4 (extremely). The GSI, the most sensitive
were included in the pool of LGB participants). We did not specify that measure of psychopathology, had an alpha of .96 in the present data set.
siblings had to be full biological siblings in order to participate in the study.
A total of 2,354 questionnaires were requested by index participants
Suicidal Ideation, Suicide Attempts, and Self-Injurious
(those who contacted us) for themselves and their siblings; 1,274 were
returned (54.1%). Of the 796 questionnaires requested by index partici- Behavior
pants for themselves, 620 (77.9%) were returned. Index participants re- Questions assessing suicidal ideation and suicide attempts, both before
quested questionnaires for 1,558 siblings, and 645 were returned (41.3%). and since the age of 18, were included. Additionally, participants were
However, the response rate of siblings may be an underestimation because asked if they ever engaged in intentional self-injurious behaviors without
we do not know how many siblings actually got these questionnaires. Many the intention of committing suicide.
index participants had more than one sibling, and we received completed
questionnaires from some but not all of them. Thus, even though the
response rate was different, the actual number of index participants and Therapy, Psychiatric Hospitalization, and Use of
siblings was roughly equal. Overall, questionnaires were sent out to 790 Psychiatric Medications
families of siblings. Of these, 421 families (53.3%) had the index partic-
ipant and at least one other sibling return questionnaires. Twenty question- Participants were asked (a) whether they had ever been in counseling or
naires were excluded from further analyses because participants did not therapy, (b) whether they were hospitalized for mental health problems
indicate gender or sexual orientation or indicated that they were before age 18, and (c) whether they were hospitalized for mental health
transgender. problems since age 18. They were also asked if they had ever been
Index participants were asked to list the gender, age, and relationship to prescribed psychiatric medications.
them (full biological, half biological, step, foster, adopted, or other) of all
of their siblings. Index participants who returned questionnaires reported Rosenberg Self-Esteem Scale
having a total of 1,576 siblings (839 sisters and 737 brothers). Of these,
1,300 (82.5%) were full biological siblings, 173 (11.0%) were half siblings, Self-esteem was assessed by using the Rosenberg Self-Esteem Scale
and 103 (6.5%) were step-siblings, foster siblings, adopted siblings, or (Rosenberg, 1965), which includes 10 items that assess self-acceptance and
other. The mean number of siblings (including the index participant) per self-worth. Items phrased negatively were reverse-scored for a total score.
family was 3.65. On average, index men had 1.24 brothers and 1.51 sisters; This measure has been used frequently in research and had an alpha of .89
index women had 1.26 brothers and 1.36 sisters. The difference in gender in the present data set.
of siblings between index men and index women was not significant.
The number of siblings who returned questionnaires included 145 broth- Satisfaction With Life Scale
ers and 256 sisters of index women and 78 brothers and 128 sisters of index
men. An average of 2.4 participants per family (including the index Life satisfaction was assessed by using the Satisfaction With Life Scale
participant) returned questionnaires (range ⫽ 0 –12; median ⫽ 2). There (Diener, Emmons, Larsen, & Griffin, 1985), a 5-item measure designed to
 BRIEF REPORTS 473

assess global aspects of life satisfaction. The Satisfaction With Life Scale assessed by constructing orthogonal contrast codes comparing (a)
correlates with other well-established measures of subjective well-being heterosexuals (2) versus both gay men and lesbians (⫺1) and
and had an alpha of .91 in the present data set. bisexuals (⫺1), and (b) gay men and lesbians (⫺1) versus bisex-
uals (1). In the latter contrast, heterosexuals were coded as 0.
Results These nested contrasts enabled us to evaluate the independent
We first examined whether LGBs who were index participants effects of belonging to a sexual minority group and of subgroup
differed from LGBs recruited from siblings. A series of t tests differences within sexual minority groups. The orthogonal contrast
comparing the two groups found no significant difference on any vectors were entered as Level 1 fixed effects. The significance of
dependent measure. these contrasts was indicated by differences in slopes at Level 2.
Descriptive statistics for all measures are presented by sex and To account for possible differential sex effects across sexual
sexual orientation in Table 1. Because participants were nested orientation groups, we also included two Sex ⫻ Sexual Orientation
within families, all data were analyzed by using multilevel mod- interaction terms included at Level 1, one for heterosexuals versus
eling, conducted in HLM Version 5.05 (Raudenbush, Bryk, LGBs and the other for gay men and lesbians versus bisexuals. In
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Cheong, & Congdon, 2000). For each outcome measure, two-level each case, the sex and sexual orientation vectors outlined above
This document is copyrighted by the American Psychological Association or one of its allied publishers.

random intercepts models were constructed. Sexual orientation were multiplied by one another to create the interaction term. All
effects were modeled at Level 1. Because age, sex, and education Level 1 predictors were uncentered.
could account for variance in the outcome measures, we used a At Level 2, we were interested in examining family effects on
forward-stepping procedure to determine which of these variables, the Level 1 intercepts, which represented within-family function-
if any, also belonged in Level 1 models. This involved testing age, ing on each dependent variable. In addition, mean familial adjust-
sex, and education effects separately, before adding sexual orien- ment (averaged GSI scores within families) was included as a
tation predictors. Age, sex, and education were retained in the full Level 2 covariate to both assess and control for the influence of
multilevel models only if they provided significant prediction family mental health on the outcome measures. The only Level 1
when tested in isolation ( p ⬍ .01). Sexual orientation effects were parameter that was allowed to vary at Level 2 was the intercept

Table 1
Sample Descriptive Statistics by Sexual Orientation and Sex

Male Female

Heterosexual Bisexual Gay Heterosexual Bisexual Lesbian

Variable M SD M SD M SD M SD M SD M SD

Continuous variables

Age 35.8 10.9 35.2 10.3 39.3 11.7 36.8 11.4 31.6 10.3 36.8 11.1
Educationa 4.04 1.4 3.82 1.3 4.53 1.4 4.00 1.3 4.52 1.3 4.69 1.2
Number of siblings 3.27 2.1 2.74 1.8 2.77 1.7 3.00 1.7 2.37 1.7 2.75 1.8
Global Severity Indexb 0.45 0.44 0.70 0.63 0.53 0.44 0.52 0.48 0.64 0.56 0.55 0.45
Somatization 0.26 0.39 0.52 0.62 0.37 0.51 0.43 0.52 0.49 0.57 0.45 0.50
Obsessive–Compulsive 0.68 0.63 1.05 0.82 0.80 0.59 0.87 0.71 0.98 0.76 0.88 0.63
Interpersonal Sensitivity 0.52 0.67 0.76 0.89 0.74 0.73 0.71 0.74 0.85 0.84 0.78 0.80
Depression 0.50 0.64 0.80 0.77 0.68 0.69 0.55 0.67 0.81 0.86 0.64 0.73
Anxiety 0.39 0.53 0.73 0.76 0.53 0.55 0.52 0.60 0.73 0.70 0.59 0.60
Hostility 0.56 0.59 0.66 0.66 0.44 0.49 0.55 0.59 0.60 0.72 0.48 0.51
Phobic Anxiety 0.16 0.32 0.35 0.57 0.25 0.44 0.21 0.46 0.24 0.46 0.27 0.48
Paranoid Ideation 0.62 0.73 0.76 0.84 0.57 0.68 0.51 0.66 0.49 0.72 0.48 0.58
Psychoticism 0.37 0.55 0.62 0.76 0.42 0.51 0.36 0.53 0.47 0.55 0.39 0.50
Rosenberg Self-Esteem 33.5 4.7 31.6 5.4 32.9 4.9 32.2 5.1 31.1 5.5 32.4 5.3
Life satisfaction 21.8 3.1 20.4 3.3 21.4 3.2 22.0 3.1 21.2 3.3 21.9 3.2

Dichotomous variables

Self-injurious behavior (%) 12.6 34.3 15.5 13.2 40.2 24.5

Suicide ideation ⬍ 18 years (%) 14.8 42.9 40.5 20.5 41.8 33.8
Suicide attempt ⬍ 18 years (%) 3.3 17.1 10.5 6.2 17.4 7.0
Suicide ideation ⱖ 18 years (%) 18.0 31.4 41.1 19.7 39.3 38.4
Suicide attempt ⱖ 18 years (%) 3.3 11.4 10.5 4.4 10.7 7.9
History of therapy (%) 43.8 69.4 72.4 54.8 85.2 82.9
Psychiatric hosp. ⬍ 18 years (%) 6.9 2.9 1.8 0.0 3.3 1.8
Psychiatric hosp. ⱖ 18 years (%) 2.2 5.7 6.8 3.8 6.6 5.5
Psychiatric med. history (%) 17.5 36.1 35.5 28.1 44.3 39.9

Note. hosp. ⫽ hospitalization; med. ⫽ medication.

a
1 ⫽ some or no high school; 2 ⫽ high school degree; 3 ⫽ some college; 4 ⫽ college degree; 5 ⫽ some graduate or professional school; 6 ⫽ graduate
or professional degree. b The Global Severity Index and the nine subscales below it are from the Brief Symptom Inventory.
 474 BRIEF REPORTS

term. This approach was taken because the random slope param- Discussion
eters were not reliable in some of the more complex models. Thus,
all Level 1 slopes were fixed, creating random intercepts models We hypothesized that LGB individuals would report greater
(see Raudenbush & Bryk, 2002). In cases of binary outcome psychological distress, suicidal ideation, suicide attempts, and use
measures (e.g., presence vs. absence of suicide attempts), nonlin- of mental health services than would their heterosexual counter-
ear Bernoulli models were specified. For all analyses, there were parts, even when controlling for sibling variance. This hypothesis
1,235 participants nested within 639 families. Participants who was only partially supported. Sexual orientation significantly pre-
were missing data on any variable (n ⫽ 19) were excluded from dicted suicidal ideation, suicide attempts, self-injurious behavior,
analyses.1 and use of mental health services, but not psychological distress as
Results from the HLM analyses appear in Table 2. Because of indexed by the BSI subscales.
the large number of comparisons and the large sample size, only In general, our hypothesis that bisexuals would show greater
effects with p values less than or equal to .01 are interpreted. psychological distress than lesbians and gay men was not sup-
Because Level 1 analyses of the BSI subscales uncovered no ported in the current study, although bisexuals did report engaging
sexual orientation effects or Sex ⫻ Sexual Orientation interactions, in more self-injurious behavior. Their rates of suicidal ideation and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

suicide attempts, however, were not higher than the other groups.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

these findings are not reported. Among the remaining variables,

several sexual orientation effects emerged. These are summarized Thus, it may be that researchers can feasibly combine samples of
in turn below. sexual minority participants in comparisons with heterosexuals.
On the other hand, this result must be interpreted with caution,
given the relatively small number of bisexuals in the current
Men Compared With Women sample. Furthermore, the recruitment method of this study may
have yielded a relatively psychologically healthy sample of bisex-
As Table 2 indicates, only two sex effects were found. Women ual men and women who are self-identified and active within
of all sexual orientation groups were more likely to report that they bisexual organizations. This may not be reflective of bisexual men
had been in therapy than their male counterparts, and men were and women in the general population, who experience relatively
more likely to have experienced a psychiatric hospitalization be- greater invisibility and lack the within-group community support
fore age 18. This latter effect was qualified by a Sex ⫻ Sexual that lesbians and gay men have begun to enjoy in recent years.
Orientation interaction, described below. Whereas previous studies have compared LGB adults with
unrelated heterosexuals and found small but significant differences
Heterosexuals Compared With Sexual Minorities in mental health, the current study is the first to model family
effects of a sample of lesbian, gay male, bisexual, and heterosexual
As also indicated in Table 2, sexual minority status was predic- siblings. Results indicate that families vary in mental health, such
tive of both suicidal ideation and attempts, before and after age 18. that siblings, even those who differ in sexual orientation, are
Significant sexual orientation effects were also found for self- comparable in general psychological distress, self-esteem, and life
injurious behavior, histories of psychotherapy, and psychiatric satisfaction. Nevertheless, LGBs still use mental health services
medications. In each case, sexual minorities scored higher than did more and are at higher risk for suicidal ideation, suicide attempts,
heterosexuals. and self-injurious behavior than are heterosexual siblings. One
potential explanation is that although familial factors impact men-
tal health for all siblings in a family, LGB siblings must also
Lesbians and Gay Men Compared With Bisexuals contend with “minority stress” associated with their sexual orien-
tation (Meyer, 2003). Future research with LGBs and their siblings
The only contrast comparing lesbians and gay men with bisex-
might examine how other within-family processes, such as victim-
uals that was significant was for self-injurious behavior. Bisexuals
ization and social support, might impact mental health differently
were more likely to report histories of self-injury than were gays or
according to sexual orientation. Additionally, the possibility that
lesbians.
cultural factors might impact self-reports of mental health prob-
lems should be explored. Savin-Williams (2001) suggested that
Sex ⫻ Sexual Orientation Interaction Effects LGB youth may overreport suicide attempts in response to a
cultural script that says that being LGB is associated with distress
Only one significant effect was found by using the Sex ⫻ and being suicidal; this may have played a role in the retrospective
Heterosexual versus LGB interaction term or the Sex ⫻ Lesbians reports of the LGB participants in the current study. On the other
and Gay Men versus Bisexual interaction term. Here, the latter hand, such cultural factors may influence actual suicidal behaviors
interaction term indicated that heterosexual males were more as well.
likely than any other group to report a psychiatric hospitalization LGB participants were more likely to be consumers of mental
before age 18. health services than were heterosexuals. Although few studies of
LGB mental health include measures of utilization, those that do
Sibling Variance
1
There were 9 participants for whom family identification code was
As indicated by the significant Level 2 effects of family adjust- missing or unclear. We included these participants and treated them as
ment, averaged sibling functioning as assessed by GSI scores individual respondents with a family size of one. When we reran all
significantly predicted all of the psychiatric variables reported in analyses excluding these 9 participants, the overall pattern of results was
Table 2. identical.
 This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Table 2
Multilevel Modeling Analyses of Sexual Orientation and Family Adjustment Effects on Mental Health

Level 1 fixed effects Level 2

family effects
Sex ⫻
Heterosexual Gay/lesbian Heterosexual Family
Age Education Sex vs. LGB vs. bisexual vs. LGB adjustmenta

Variable Coeff. SE Coeff. SE Coeff. SE Coeff. SE Coeff. SE Coeff. SE Coeff. SE

Self-injurious behavior ⫺.046 .007** — — .030 .084** .513 .178* ⫺.040 .104 .866 .196**
Suicide ideation ⬍ 18 years ⫺.021 .006** — — ⫺.468 .090** ⫺.044 .156 .209 .107 .960 .193**
Suicide attempt ⬍ 18 years — — — ⫺.449 .140* .225 .219 .265 .164 1.19 .223**
Suicide ideation ⱖ 18 years .018 .006* — — ⫺.304 .084** ⫺.156 .153 ⫺.018 .099 1.24 .204**
Suicide attempt ⱖ 18 years .031 .009** — — ⫺.415 .158* ⫺.025 .228 .157 .195 1.63 .259**
History of therapy .015 .005* .149 .052* .654 .167** ⫺.037 .076** ⫺.041 .189 ⫺.108 .098 1.01 .232**
BRIEF REPORTS

Psychiatric hosp. ⬍ 18 years — ⫺.625 .155** ⫺1.57 .348** .294 .165 .064 .419 ⫺1.86 .198** 1.07 .318**
Psychiatric hosp. ⱖ 18 years — — — ⫺.381 .168 ⫺.031 .252 .258 .204 1.38 .257**
Psychiatric med. history .024 .006** — — ⫺.364 .081** .303 .153 .156 .095 1.07 .185**
Rosenberg Self-Esteem .042 .012** — — .357 .154 ⫺.590 .320 ⫺.383 .184 ⫺5.82 .423**
Life satisfaction — — — .067 .106 .056 .230 .049 .125 ⫺2.92 .251**

Note. Dashes indicate variables that were omitted from the final multilevel modeling equation because they were nonsignificant when tested during forward stepping. The Sex ⫻ Lesbians and Gay Men
versus Bisexual interaction term was not included in the table because no significant effects were found for this variable. LGB ⫽ lesbian, gay male, and bisexual; Coeff. ⫽ coefficient; hosp. ⫽
hospitalization; med. ⫽ medication.
a
Averaged within-family Global Severity Index score.
* p ⱕ .01. ** p ⱕ .001.
475
 476 BRIEF REPORTS

consistently find that LGB individuals are high utilizers of psy- ences should include diagnostic interviews to determine whether
chotherapy. In the current study, this difference appeared with these sibling patterns are also true when examining mental health
psychiatric medications as well. High use of mental health services disorders, both lifetime and current. A final limitation is the use of
has usually been interpreted as a sign of psychological distress relatively general questions about suicidal ideation and attempts.
(e.g., Cochran, 2001), but Morgan (1992) found that norms within Following Savin-Williams’s (2001) example, future research on
the LGB community make mental health services more acceptable. sexual orientation should ask more specific, behaviorally anchored
In the present study, use of mental health services among hetero- suicidality questions to more accurately assess risk among LGB
sexuals was also high (though significantly lower than that of adults.
LGBs). This may indicate a “contagion effect,” in which LGBs
describe their use of such services in positive terms to their References
siblings, with the result that these heterosexuals, too, make use of
mental health services. For example, Morgan found that 28% of Bailey, J. M., Dunne, M. P., & Martin, N. G. (2000). Genetic and envi-
her sample of heterosexual women (who were unrelated to the ronmental influences on sexual orientation and its correlates in an
lesbians in her study) had ever been in therapy, whereas Rothblum Australian twin sample. Journal of Personality and Social Psychology,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

78, 524 –536.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

and Factor’s (2001) study of sisters found that over half of het-
Cochran, S. D. (2001). Emerging issues in research on lesbians and gay
erosexual sisters of lesbians had been in therapy.
men’s mental health: Does sexual orientation really matter? American
The current method of recruiting siblings from LGB index Psychologist, 56, 931–947.
participants raises several issues regarding sampling. On the pos- Cochran, S. D., Sullivan, J. G., & Mays, V. M. (2003). Prevalence of
itive side, it is an easy way of obtaining heterosexuals for com- mental disorders, psychological distress, and mental health services use
parison purposes, which has been a challenge in prior LGB re- among lesbian, gay, and bisexual adults in the United States. Journal of
search. Even though the questionnaire return rate of LGB index Consulting and Clinical Psychology, 71, 53– 61.
participants was higher than that of siblings, many participants had Derogatis, L. R., & Spencer, P. M. (1982). Brief symptom inventory (BSI):
more than one sibling who was heterosexual, yielding roughly Administration, scoring, & procedures manual—I. Baltimore: Author.
equal numbers of heterosexuals and LGBs. Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The
Satisfaction With Life Scale. Journal of Personality Assessment, 49,
There were more lesbian than gay male respondents in the study,
71–75.
and sisters were more likely to respond than brothers, reflecting the
Gayellow pages. (2001). New York: Renaissance House.
fact that women are more likely to volunteer for research than men. Greene, B. (1994). Lesbian women of color: Triple jeopardy. In L. Comas-
A major limitation is the lack of ethnic and racial diversity. The Diaz & B. Greene (Eds.), Women of color: Integrating ethnic and gender
goal of this study was to include a diverse sample, and strong identities in psychotherapy (pp. 389 – 427). New York: Guilford Press.
efforts were made to recruit potential participants of color. Nev- Jorm, A. F., Korten, A. E., Rodgers, B., Jacomb, P. A., & Christensen, H.
ertheless, only 8.3% of the sample were members of an ethnic (2001). Sexual orientation and mental health: Results from a community
minority group, making this study comparable to the majority of survey of young and middle-aged adults. British Journal of Psychiatry,
studies on LGB populations. The requirement of the participation 180, 423– 427.
of a sibling may have served as a barrier to diversity. Although Meyer, I. M. (2003). Prejudice, social stress, and mental health in lesbian,
gay, and bisexual populations: Conceptual issues and research evidence.
participants were assured that they could give a survey to a
Psychological Bulletin, 129, 674 – 697.
heterosexual sibling without outing themselves, people of color,
Morgan, K. S. (1992). Caucasian lesbians’ use of psychotherapy. Psychol-
who may be more skeptical of research in general and less out to ogy of Women Quarterly, 16, 127–130.
their families, may have been even less likely to participate (cf. Morris, J. F., Waldo, C. R., & Rothblum, E. D. (2001). A model of
Greene, 1994). predictors and outcomes of outness among lesbian and bisexual women.
Another sampling issue to consider is whether the yoked sibling American Journal of Orthopsychiatry, 71, 61–71.
design (i.e., index participants answering an ad and then recruiting Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models:
their siblings) confounded the study results in any way. One Applications and data analysis methods (2nd ed.). Thousand Oaks, CA:
possibility is that index participants, who actively volunteered, Sage.
may have been psychologically healthier than their siblings (who Raudenbush, S. W., Bryk, A. S., Cheong, Y. F., & Congdon, R. (2000).
HLM5: Hierarchical linear and nonlinear modeling. Lincolnwood, IL:
were more passive volunteers), thereby biasing the sexual orien-
Scientific Software International.
tation differences toward the null. Conversely, it could be argued
Rosenberg, M. (1965). Society and the adolescent self-image. Princeton,
that the yoked sibling method biases sexual orientation differences NJ: Princeton University Press.
away from the null, if index participants are more psychologically Rothblum, E. D., & Factor, R. (2001). Lesbians and their sisters as a
minded than their siblings and therefore more likely to report control group: Demographic and mental health factors. Psychological
mental health symptoms. Although the impact of sampling method Science, 12, 63– 69.
should be further examined in future research, it is important to Savin-Williams, R. C. (2001). Suicide attempts among sexual-minority
note that initial comparisons of the LGB index participants with youths: Population and measurement issues. Journal of Consulting and
LGB siblings showed no differences on any dependent measure in Clinical Psychology, 69, 983–991.
the study, indicating that the yoked design may not have con-
founded results. Received February 3, 2003
Mental health measures were self-report checklists of symp- Revision received October 19, 2004
toms. Future studies of within-family sexual orientation differ- Accepted October 25, 2004 䡲