The Problem with Square Pegs: Sexual Orientation Concordance as a Predictor of Depressive Symptoms

Abstract

The author uses a nationally representative sample of cisgender young adults to examine the relationship between sexual orientation concordance and the prevalence of depressive symptoms. In these analyses, the author differentiates between those with an exclusive identity (100 percent gay or 100 percent straight) and those with a nonexclusive identity (“mostly gay,” “mostly straight,” or bisexual). Among those with an exclusive identity, the author differentiates between those with behavior and attraction that is in line with (concordant) or goes against (discordant) a claimed gay or straight identity. Those with a concordant sexual orientation report significantly lower depressive symptoms scores than do those with either a discordant sexual orientation or a nonexclusive identity. When accounting for orientation, concordance is significantly associated with depressive symptoms for straight- but not gay-identified young adults. These findings generally hold for women, but not for men when change in identity is controlled for.

Keywords

depressive symptoms depression mental health sexual orientation

Like many fields, research on sexual orientation is characterized by multiple and inconsistent measures. For example, a man who identifies as 100 percent gay (“identity”), is attracted to both men and women (“attraction”), and has had sex only with women (“behavior”) has an exclusively gay identity, nonexclusive attraction, and exclusively straight behavior. Such “discrepancies” within an individual’s sexual experiences pose a problem for researchers who want to link sexual orientation with mental health outcomes. In the past, researchers have largely treated these discrepancies as noise and responded by dichotomizing sexual orientation or treating dimensions of sexual orientation (identity, behavior, and attraction) as equivalent. However, more recently the field has taken to making use of these data, whose complicated nature is a more accurate reflection of sexual orientation itself.

My study adds to this new line of research by examining how specific patterns of sexual orientation identity, behavior, and attraction are related to depressive symptoms. Specifically, I examine sexual orientation concordance, defined as having sexual attraction and a history of behavior that matches one’s current sexual orientation identity. For example, a woman who identifies as exclusively gay, and has only same-sex attraction and behavior is classified as gay concordant. In contrast, a woman who identifies as exclusively gay and is attracted to, or has had sex with, men is gay discordant. The opposite holds for those who identify as exclusively straight. Those with a nonexclusive identity (“mostly gay,” “mostly straight,” or bisexual) are not categorized in terms of concordance but are themselves a group of interest. This is a break from past research, which either treated this group as inherently concordant (Gattis, Sacco, and Cunningham-Williams 2012; Zhao 2010) or as equivalent to an exclusive identity (Talley et al. 2015). My conceptualization of concordance speaks to broader issues in the literature about how we categorize individuals by sexual orientation.

The few existing studies examining the relationship between discordance of sexual orientation and mental health find that discordance is associated with worse mental health (Gattis et al. 2012; Zhao 2010). However, these studies are few and limited. My study extends their work in several ways. First, I am the first to test the relationship between discordant sexual orientation and mental health for both gay- and straight-identified young adults. Second, I use the National Longitudinal Study of Adolescent and Adult Health (Add Health), a nationally representative sample of cisgender young adults.¹ This is important because convenience samples of LGBT communities, more commonly used in the field, may oversample respondents at greater risk for mental health problems (Savin-Williams 1994). Third, I argue for a conceptualization of discordance based on the presence of behavior or attraction that is actively at odds with a claimed sexual orientation identity. Last, I include a measure of discordance that is based on all past behavior as well as one based on only past-year behavior. This is important because it allows me to partly control for the effects of change in identity.

Background

Sexual Orientation and Mental Health

The demedicalization of homosexuality in 1973 led researchers away from the assumption, and later debate, of homosexuality as a disorder to a new research question: are nonheterosexuals² at greater risk for mental health disorders than heterosexuals (Meyer 2003)? Theory and meta-analyses of past research suggest that the answer is yes. Nonheterosexuals are at higher risk for depression, anxiety, and hazardous drinking, among other outcomes (Cochran 2001; King et al. 2008; Meyer 2003). Even the most cautious of the meta-analyses concluded that where differences do exist, nonheterosexuals are worse off in terms of their mental health outcomes (Meyer 2003). Beyond this broad claim, researchers have also established that nonheterosexuals are at greater risk for specific disorders. For example, one analysis found that nonheterosexuals are at more than twice the risk of having experienced depression in the past 12 months than are heterosexuals (King et al. 2008).

The literature in the area is far from purely descriptive. Instead, much attention has been given to explaining why this disparity in mental health exists. One now classic explanation is “minority stress,” which argues that stigma, discrimination, and prejudice cause nonheterosexuals stress, which then contributes to negative mental health (Meyer 1995, 2003). Although not all research explicitly supports this model, much of the literature is dedicated to the study of sexual stigma (Herek 2004) and its internalization (Herek and Garnets 2007). Other research focuses on nonheterosexuals’ experiences of discrimination on the basis of their sexual orientation identity (Savin-Williams 1994), as well as discrimination at the individual and structural levels (Cochran 2001). Although the precise theories, terms, and cited mechanisms may vary, the consensus is that nonheterosexuals are at greater risk for mental health problems because of the societal assumption of heterosexuality and hostility toward nonheterosexuality. Despite the breadth of high-quality work on the topic, however, there are reasons to believe that our understanding of the process is limited.

One limitation to research on sexual orientation is the difficulty of measuring sexual orientation, which is conceived of as multidimensional (attraction, behavior, and identity) and fluid over time (Laumann et al. 1994; Ott et al. 2011). The multidimensionality of sexual orientation means that estimates of the prevalence of nonheterosexuals depend on which dimension is used and how it is operationalized. For example, one study found that nonheterosexuals make up between 1 percent and 15 percent of the population depending on what dimension is used and what cutoffs define categories (Savin-Williams and Ream 2007). Furthermore, the relationship between the three dimensions themselves is imprecise. For example, some studies show no clear pattern among dimensions and only a small overlap (about 15 percent) among all three (Igartua et al. 2009; Laumann et al. 1994). This means that few nonheterosexuals are internally consistent on all three dimensions (e.g., having nonheterosexual identity, attraction, and behavior). Last, measuring sexual orientation is complicated by gender, with differences found between men and women³ in the prevalence of nonheterosexuality (Zhao 2010). For example, women are more likely to report a nonexclusive sexual orientation than are men, who tend to cluster around the extremes (i.e., 100 percent straight or 100 percent gay) (Igartua et al. 2009).

A second limitation is the use of nonrepresentative samples, the consequence of studying a relatively small, hidden group (Cochran 2001; King etal. 2008). Until quite recently, most research on nonheterosexuals used convenience samples drawn from LGBT communities and organizations (e.g., Herek 2004; Talley et al. 2015), making statistical generalization about the relative risk for specific disorders for nonheterosexuals difficult (Cochran 2001; Russell and Joyner 2001; Savin-Williams 1994). Specifically, identity-based sampling may lead to undersampling those for whom a nonheterosexual identity is not as salient. Those who identify as straight but have same-sex attraction or behavior—in other words, those whose identities do not match their attraction or behavior—would similarly be excluded. Consequently, the observed relationship between nonheterosexuality and mental health outcomes in these studies may represent only a distinct subset of nonheterosexuals (Savin-Williams 1994). These limitations become especially problematic when researchers dichotomize sexual orientation into heterosexual and nonheterosexual. In seeking to make broad claims about all nonheterosexuals, they implicitly, and falsely, treat them as a homogeneous group (Bostwick and Hequembourg 2013). When this happens, it becomes possible for one subgroup with high risk for mental health disorders to drive the association for the entire nonheterosexual category.

These limitations carry into research linking sexual orientation to mental health. For example, three recent meta-analyses reviewed and combined previous work to determine whether nonheterosexuals are at greater risk for mental disorders (Cochran 2001; King et al. 2008; Meyer 2003). These meta-analyses combined studies in which sexual orientation was measured by attraction, identity, behavior, or some combination thereof. Given how little overlap these measures have, it is unlikely that nonheterosexuals were reliability categorized across the all the studies included. Out of necessity, the meta-analyses treated sexual orientation as binary, comparing all individuals classified as nonheterosexual by the given study’s measure with their heterosexual counterparts.⁴ These analyses, although well formulated to answer their research question, were limited by the simplicity of the question they sought to answer. The availability of new data and the outcropping of more specific sexual orientation identities in recent years has led researchers to tackle a more complicated question: For whom does sexual orientation matter for mental health? And why?

The Importance of Exclusivity

One line of research aimed at answering this question calls for attention to the exclusivity, rather than simply the direction, of sexual orientation (Bostwick and Hequembourg 2013; Lindley, Walsemann, and Carter 2012). Exclusivity refers to whether a dimension of sexuality is oriented toward only one sex. Within each dimension of sexual orientation (identity, attraction, and behavior) individuals can have an exclusive or nonexclusive⁵ orientation. For example, a man who identifies as 100 percent straight, is attracted to both men and women, and has sex only with women has an exclusively straight identity, nonexclusive attraction, and exclusively straight behavior. Recent research has shown that distinct and surprising patterns emerge when the exclusivity and the gender of the respondent is considered. For example, one study found that while lesbian-identified women had a relatively high probability of mood disorders, women with exclusively gay behavior had the lowest risk for mood disorders, even lower than women with exclusively straight behavior (the same pattern holds for attraction) (Bostwick et al. 2010). Another study of young adults found no difference in depressive symptoms between exclusively gay or exclusively straight young adults. However, women who had nonexclusive identity, attraction, or behavior had a higher prevalence of depressive symptoms than did those with exclusively straight orientation (Lindley et al. 2012). These studies suggest that for those wishing to study depressive symptoms among young adults, there is an important distinction between exclusive and nonexclusive identities and behaviors.

Those with nonexclusive identities may experience more, or at least different, stigma than do those who identify as exclusively gay because of the societal belief that sexual orientation is oriented toward only one sex, or monosexism (Bostwick and Hequembourg 2013). This belief leads people, within both the general public and the LGBT community, to question the existence and validity of bisexuality itself (Balsam and Mohr 2007; Bostwick and Hequembourg 2013; Hequembourg and Brallier 2009; Ripley et al. 2011; Ross, Dobinson, and Eady 2010). When they do, the identity is invalidated and is no longer useful as a symbol that facilitates interaction with others. Furthermore, monosexism on the part of the LGBT community deprives bisexuals of an important benefit of accepting a nonheterosexual identity: access to a supportive community of “experientially similar others” (Lindley et al. 2012; Thoits 2011). Thus, bisexuals are subject to the negative consequences of a stigmatized identity (discrimination, legal inequality, etc.) and but also not have access to the benefits (e.g., a supportive community or a symbol for explaining action).

Along with bisexuals, there is growing evidence that those who identify as “mostly straight” or “mostly gay” may have distinct sexual orientations (Savin-Williams and Vrangalova 2013; Vrangalova and Savin-Williams 2012). Although limited by small group sizes (especially for men, who are more likely to claim exclusive identities), the available research suggests that the mostly straight have rates of mental and physical health disorders that are higher than heterosexuals and slightly lower than bisexuals (Vrangalova and Savin-Williams 2014). There is also some evidence that, at least for young women, those who identify as mostly straight and mostly gay have very similar depressive symptoms scores, which fall higher than the exclusively straight-identified and exclusively gay-identified but lower than bisexuals (Lindley et al. 2012). This is important to research on sexual orientation and mental health because it suggests that exclusivity, rather than orientation, of sexuality may be driving mental health outcomes. It also suggests against combining those who identify as “mostly” straight or gay with their exclusive counterparts (Lindley et al. 2012).

Although past research (Bostwick et al. 2010; Lindley et al. 2012) has made important contributions using nationally representative data sets to link measures of sexual orientation and mental health, these studies were limited in their failure to compare nonheterosexual groups with one another. This is symptomatic of the conception of sexual orientation as an ordinal continuum with heterosexuality as the base category, the “normal” from which researchers measure how far people stray. Beyond its theoretical or normative implications, this obscures potential differences among nonheterosexual groups (Cragun and Sumerau 2015). I extend prior research by comparing those with nonexclusive identities (“mostly gay,” “mostly straight,” or bisexual) to both the exclusively gay- and straight-identified. This allows me to test to what extent exclusivity rather than orientation drives differences in mental health in young adults. In doing so I differentiate between the effects of exclusivity, orientation, and concordance (discussed below) on the mental health of young adults. I hypothesize that respondents with nonexclusive identities will have higher depressive symptoms scores than those with either gay- or straight-exclusive identities, irrespective of concordance.

The Importance of Concordance

A second line of research has reflected a growing interest in concordance: the degree to which sexual orientation identity, behavior, and attraction match. Although operationalization varies, generally sexual orientation is thought to be concordant if attraction and behavior match a claimed identity (e.g., a woman identifies as gay and has only same-sex attraction and behavior) and discordant if they do not (e.g., a woman identifies as gay but has female-male attraction and/or behavior). Following Laumann et al.’s (1994) finding of a lack of overlap among the three dimensions of sexual orientation, scholars have started to study concordance (Igartua et al. 2009), as well as call for the study of the relationship between concordance and mental health (Lindley et al. 2012). Although studies answering this call are still few (Gattis etal. 2012; Igartua et al. 2009; Savin-Williams and Ream 2007; Zhao 2010), they have shown concordance to be a potentially promising mechanism linking sexual orientation and mental health.

Cognitive dissonance theory would suggest that a dissonant relationship between one’s sexual orientation identity and one’s history of sexual behavior or attraction would cause stress and might contribute to mental health disorders such as depression (Festinger 1957; Talley et al. 2015). People strive for internal consistency; when theirbehaviors do not match their images of themselves—their identities—they feel higher levels of stress. The more important the identity in question, the more stressful dissonant behavior can be. The importance placed by society on sexual orientation identities means that sexual orientation discordance will be especially stressful (Festinger 1957; Gamson 1995). Given this, I hypothesize that having a discordant, as opposed to a concordant, sexual orientation identity will be associated with more depressive symptoms. The few studiesthat have examined the relationship between concordance and mental health support this hypothesis.

Two recent studies have examined the concordance between sexual orientation identity and behavior or attraction in relation to mental healthoutcomes. Zhao (2010) compared straight-concordant, straight-discordant, and nonheterosexual-identified youths. She found that adolescents with nonheterosexual identities were at greater risk for suicidality than straight-concordant adolescents but no significant difference between straight-discordant and straight-concordant youth. For women, there may have been a protective effect of exclusive and concordant same-sex behavior and attraction. In a second study, Gattisetal. (2012) compared straight-concordant, straight-discordant, and nonheterosexual-identified adults. They found that straight-discordant women were at greater risk for a depressive episode than both straight-concordant women and women who identified as nonheterosexual. This research illustrates that concordance is associated with worse mental health but that this relationship may vary by gender.

Another related body of research examines sexual orientation concordance and health behaviors, especially drinking (Drabble, Midanik, and Trocki 2005; Poston and Baumle 2010; Talley etal. 2015). Although not directly a measure of mental health, drinking and other health risk behaviors may be used in an attempt to ameliorate the effects of cognitive dissonance (Talley et al. 2015). For example, Talley et al. (2015) argued for cognitive dissonance as a mechanism in their study of sexual minority women. They found that greater sexual orientation discordance was associated with increased risk for hazardous drinking, presumably because sexual minority women engage in hazardous drinking to reduce dissonance. Other studies have found a similar relationship among the straight-identified (Drabble et al. 2005; Poston and Baumle 2010). The relationship between discordance and hazardous health behaviors provides further support for the hypothesis that discordant sexual orientation is negatively related to mental health.

Conceptualizing Sexual Orientation Concordance

Although scholars agree that concordance measures whether attraction and/or behavior matches identity, differences in conceptualization and operationalization exist. For example, some studies ignore the possibility of discordance among the gay-identified (Gattis et al. 2012; Zhao 2010), instead grouping together all respondents without a straight-exclusive identity. Gattis etal. (2012) labeled as “gay concordant” any respondent who identified as gay, lesbian, or bisexual, regardless of history of sexual behavior. The authors reasoned that “gay/lesbian and bisexual men and women live in a society where heterosexuality is normative; therefore, it is not surprising for them to have engaged in at least one opposite-sex sexual encounter, particularly if it occurred before coming out” (p. 1187). The fact that heterosexual behavior is normative, however, does not necessarily make it less stressful for those who identify as gay or lesbian to engage in. This is a hypothesis to be tested rather than an assumption to be made. I address this limitation by differentiating between gay-concordant and gay-discordant identities. In doing so, I test the hypothesis that discordance will be associated with more depressive symptoms but that this relationship will be weaker for those with an exclusively gay identity than those with an exclusively straight identity.

Talley et al. (2015) took another track and treated concordance as ordinal. They calculated concordance by subtracting a five-category measure of behavior or attraction (1 = only women, 2 = mostly women, 3 = equally men and women, 4 = mostly men, 5 = only men) from a five-category measure of identity. Although this measure incorporates more information than my three-category version, it falsely assumes that a lack of behavior implied by an identity is equally discordant to behavior excluded by an identity. That is, by their measure a woman who identifies as only straight but has had sex with mostly men, but some women, would have the same level of discord as a woman who identifies as mostly straight but has had sex equally with men and women. Theoretically, respondents with a nonexclusive identity could feel stress because they have not validated that identity with matching behavior. However, I argue that dissonance associated with the lack of behavior implied by an identity is different from, and even lesser than, the presence of behavior that goes against that identity. This distinction is particularly important for sexual orientation because of the precariousness of the heterosexual identity, which can be threatened by any same-sex behavior or attraction (Bostwick and Hequembourg 2013), and the stigma associated with non-exclusive identities.

Researchers also vary by whether they look at discordant attraction and behavior simultaneously (Zhao 2010) or as separate types of concordance (Gattis et al. 2012; Talley et al. 2015). The three dimensions of sexual orientation may be distinct, but they do not occur in a vacuum, nor are they independent. At any one point in time, each individual falls somewhere on theoretical continua of sexual orientation identity, behavior, and attraction. Location on one dimension is related to the other two dimensions, and the meaning and consequences of an individual’s orientation on one dimension depend on the other two. In other words, the relationship between any one dimension and mental health outcomes will be moderated by the other two dimensions. When considering concordance, the relationship between identity and mental health is moderated concurrently by identity and attraction. Although considering attraction and behavior separately allows a more nuanced understanding of whether discordant attraction or behavior has a larger effect, it falsely excludes either attraction or behavior from the model. For this reason, I include attraction and behavior in the same measure of concordance.

Another area in which measures of discordance vary is in the time span of sexual behavior included (Gattis et al. 2012; Talley et al. 2015; Zhao 2010). For example, some studies include all past behavior (Zhao 2010), while others use behavior in the past five years (Talley et al. 2015). Both long- and short-term measures of sexual behavior contribute to our understanding of concordance and mental health. A concordance measure comparing current identity to a long-term measure of behavior, such as all past behavior, could capture change in identity rather than current discord (Talley et al. 2015). For a woman who identified as gay at 17, past behavior with the same sex was not (at the time) discordant, even if she currently identifies as straight. This is potentially problematic because we know that change in sexual orientation identity, in particular change toward more same-sex-oriented identities, is associated with increased risk for depressive symptoms (Everett 2015). However, a long-term measure offers larger group sizes and potentially captures discord a short-term measure would miss. It is unlikely that all or even most of the discordant respondents, using the measure of all past behavior, are classified as such as the result of change in identity. Second, it is still possible that acting on a former identity, and therefore having a discordant history of behavior, causes some stress. Given two young adults whose identities changed, there may be greater stress associated with having acted on the former identity.

Short-term measures limit the possibility of identity change acting as a spurious variable and have two other advantages. The first is that recent behavior should be more salient to current identity and therefore should have a stronger relationship with mental health. Second, the more time has passed, the more time respondents have had to use methods to lessen dissonance. For example, respondents could explain past discordant behavior as that of a younger self engaged in sexual experimentation, thus reducing the threat the behavior poses to current identity. This is a less likely explanation for behavior in the past year, especially because our respondents (at 24–32 years old) have largely exited the life stages of adolescence and emerging adulthood, when sexual experimentation is more common and accepted (Arnett 2000). To take advantage of the benefits of both long- and short-term measures of behavior, I calculate concordance using both all past behavior and behavior in the past year.

In the following analysis, I consider the relationship between sexual orientation concordance and depression, something only two prior studies have examined. I contribute to the literature inseveral ways. Unlike many studies on sexual orientation and mental health, I use nationally representative data, thus avoiding oversampling respondents at greater risk for depressive symptoms (Savin-Williams 1994). Furthermore, I differ from past studies by examining concordance among both straight- and gay-identified young adults and comparing nonheterosexual groups with one another. I also include two measures of concordance: one using all past behavior and one using behavior from the past year. In my analysis, I first examine the relationship between a three-category concordance variable and depression. I then examine concordance among the gay- and straight-identified, to see if any differences between the two identities are evident. I also run my analyses for men and women separately. This allows me to compare men and women, whom past research has suggested may be differently affected by discordance (Gattis et al. 2012; Zhao 2010).

Data and Methods

Data

I use restricted-use data from wave IV of Add Health, along with demographic variables from wave I and sexual orientation identity at wave III. Add Health follows a nationally representative sample of cisgender adolescents who were in grades 7 to 12 during the 1994–1995 school year. The third wave was collected in 2001–2002, while the fourth wave was collected in 2007–2008, when respondents were aged 24 to 32. I include in the analysis only those respondents who had valid weights and values for the variables used in each analysis. Respondents with no past sexual behavior (n = 386) were excluded from analysis because they could not be characterized as either concordant or discordant. There was no significant difference in sexual orientation attraction or identity between those who had and had not engaged in sexual behavior. Most respondents with no sexual behavior both identified as exclusively straight and reported only female-male attraction. I also exclude respondents on the basis of sexual orientation identity, dropping those who were missing data or identified as not sexually attracted to men or women. My final analytic sample for the models using all past behavior to calculate concordance consists of 13,859 young adults. The models using only past-year behavior to calculate concordance excluded 1,490 respondents who had not reported having sexual behavior in the past year, for a sample size of 12,369.

Sexual Orientation and Concordance

Sexual behavior is defined in the survey as “genital stimulation” and respondents are coded as ever having had behavior with partners of the same sex only, female-male only, or with both. Respondents were asked if they were romantically attracted to women and men and were coded as having same-sex attraction, female-male attraction, or both. Last, respondents were asked if they think of themselves as

(1) 100% heterosexual (straight) (2) mostly heterosexual (straight), but somewhat attracted to people of the same sex (3) bisexual that is, attracted to men and women equally⁶ (4) mostly homosexual (gay) but somewhat attracted to people of the opposite sex, or (5) 100% homosexual (gay).

Respondents were categorized as exclusively gay or straight if they identified as 100 percent gay or straight. Respondents who identified as “mostly gay,” “mostly straight,” or bisexual were categorized as having a nonexclusive identity.

I conceptualize of discordance as having attraction or behavior that is actively excluded by one’s current sexual orientation identity. This means that those with a nonexclusive identity (“mostly gay,” “mostly straight,” or bisexual) are not categorized in terms of concordance because their identity does not exclude romantic attraction to, or behavior with, men nor women. Thus, my concordance variable has three nominal categories: discordant, concordant, and nonexclusive. If a woman identifies as exclusively gay, is attracted only to women, and only reports having sex with women, she is classified as gay concordant. In contrast, if she identifies as exclusively gay but is attracted to men and/or reports having had sex with men, she is classified as gay discordant. The opposite is true for those who identify as exclusively straight. I do not count a lack of behavior or attraction implied by identity as discord. If a woman identifies as bisexual but has only had sex with men, the lack of same-sex behavior is not considered discordant. I calculate concordance using all past behavior as well as only behavior from the past year.

Figure 1 displays the scheme used to categorize respondents in terms of concordance and identity.Group sizes for both past-year-only and all-past-behavior measures are included. Concordance is a three-category variable (concordant, discordant, and nonexclusive identity), but Figure 1 shows concordance categories split up by gay or straight identity for descriptive purposes. The largest group is concordant (all past behavior, n = 11,527; past-year behavior, n = 10,723) made up largely of those who identify as straight (all past behavior, n = 11,423; past-year behavior, n = 10,568). Nonexclusive is second largest (all pastbehavior, n = 1,724 past-year behavior, n= 1,499) and discordant is the smallest (all past behavior, n = 613; past-year behavior, n= 147). Looking at Figure 1, we see that of those who identify as gay, about half report discordant attraction or history of behavior. This drops to 8 percent when only past-year behavior is considered. The percentage discordant is much smaller for those who identify as straight (about 5 percent all past behavior, 1 percent past-year behavior). Even though discordance is relatively rare for those who identify as straight, gay identity is so uncommon (n = 196) that any concordance group containing both gay and straight will be largely dominated by the scores of the straight-identified respondents. For this reason, I show analyses with the three-category concordance variable and with concordance split by gay and straight identity, for comparison purposes. I do so only for concordance calculated using all past behavior, because the group sizes become too small to do so when using only past-year behavior (i.e., gay concordant, n = 13). Although statistically significant differences between the gay identity groups will be harder to obtain because of their smaller sizes, these groups do offer the chance to check if the pattern holds for both gay- and straight-identified individuals.

Figure 1.

Sexual Orientation Concordance Categorization by Identity, Attraction, and Behavior.

Depressive Symptoms

I measure depressive symptoms using a nine-item version of the Center for Epidemiological Studies–Depression scale (Radloff 1977). Respondents were asked how often in the past seven days they felt sad, felt too tired to do things, enjoyed life, and so on, with responses coded from 0 = “never or rarely” to 4 = “most of the time or all of the time.” Per convention, I reverse-coded the positively worded items and added the nine items to create a scale ranging from 0 to 36 (Cronbach’s α = .79 in wave I and .81 in wave IV). Because the scale has a skewed distribution, I ran all models using both untransformed data and a log transformation of the depression scale. Both yielded similar results, so for ease of interpretation, I present models without log transformation in this article. The wave IV measure of depressive symptoms is my outcome variable, and I include a measure for depressive symptoms at wave I as a control.

Controls

Following Lindley et al. (2012), I control for gender, age, race, and highest level of education. To control for gender, I use an indicator variable for female. Age is a continuous variable measured in whole years. Race is a four-category variable and includes non-Hispanic white, non-Hispanic black, Hispanic, and other. Highest level of education is a five-category variable with the categories of less than high school, high school degree, some college, bachelor’s degree, and greater than bachelor’s degree. I include controls for political orientation and religious attendance because nonheterosexual-identified Americans are generally more liberal and less religious than the general population (Pew Research Center 2013). I want to see whether this pattern holds among concordance categories and also to control for any effects that religious and political tendencies have on the relationship between sexual orientation and mental health. Political orientation is a four-category variable with the values of conservative, middle of the road, liberal, and refused/do not know. Religious attendance is a binary measure of whether respondents had attended a religious service in the past year. Male, non-Hispanic white, less than high school, conservative, and no attendance serve as the omitted comparison groups.

To lessen the possibility that identity change acts a spurious variable, I also include an indicator of whether a respondent’s identity at wave IV was the same as at wave III. When combined with the past-year-only behavior measure of concordance, this means that only if a respondent had changed his or her identity from the wave III value, acted on this new identity during the past 12 months, and then changed his or her identity back to the wave III value would that individual be classified as discordant because of a change in identity. I also control for depressive symptoms at wave I to ensure that what I am capturing is the effect of discord rather than any underlying relationship between depression and sexuality.

Results

Table 1 shows descriptive statistics for the total sample, as well as for the concordance (all past behavior) categories, subdivided by gay and straight identity. A little over half (53 percent) of my sample are female. Whites make up a little over half of my sample, followed by 20 percent blacks and 16 percent Hispanics. The average respondent is 28 years old and had a depressive symptoms score of 5.9 at wave I. Almost half (45 percent) of my sample had some college, while 23 percent had a high school degree or less, and almost a third (32 percent) had a bachelor’s degree or higher. About a quarter of my sample identified as conservative (24 percent) or liberal (27 percent), while 44 percent identified as middle of the road. A majority (70 percent) reported attending religious services, and only 11 percent reported a different sexual orientation identity in wave III as in wave IV.

Table 1.

Demographic Characteristics by Sexual Orientation Identity and Concordance (All Past Behavior).

	Straight Concordant	Straight Discordant	Gay Concordant	Gay Discordant	Nonexclusive Identity	Total
Gender
Male	.51	.36	.79	.45	.18	.47
Female	.49	.64	.21	.55	.82	.53
Race
White	.54	.48	.45	.45	.59	.54
Black	.20	.22	.16	.27	.15	.20
Hispanic	.16	.21	.26	.15	.15	.16
Other	.10	.09	.12	.12	.11	.10
Age (years)	28.57	28.62	28.68	28.62	28.22	28.53
Depressive symptoms at wave III	5.69	6.91	6.06	7.21v	6.97	5.91
Education
Less than high school	.07	.12	.03	.10	.07	.07
High school diploma	.16	.21	.08	.19	.12	.16
Some college	.44	.44	.48	.44	.48	.45
Bachelor’s degree	.20	.15	.24	.18	.19	.20
More than bachelor’s degree	.12	.07	.16	.09	.13	.12
Political orientation
Conservative	.25	.25	.10	.18	.13	.24
Middle of the road	.45	.40	.22	.26	.40	.44
Liberal	.24	.27	.65	.54	.42	.27
Refused/don’t know	.06	.08	.03	.03	.05	.06
Religious attendance
Doesn’t attend	.28	.31	.38	.46	.40	.30
Attends	.72	.69	.62	.54	.60	.70
Identity change
Changed	.03	.16	.35	.68	.54	.11
n	11,423	516	99	97	1,724	13,890

Looking at discordance (all past behavior) andidentity, both straight-concordant and gay-discordant groups are about half female, the nonexclusive identity group is about 80 percent female, the straight-discordant group is more than 60 percent female, and gay concordant is only about 20 percent female (Table 1). Nonexclusive identity has the highest percentage white, at almost 60 percent, while gay concordant and discordant have the lowest at 45 percent. Gay concordant is made up of more than a quarter Hispanics, while gay discordant is more than a quarter black. In terms of education, the discordant categories have the least educated respondents, while gay concordant has the most educated, followed by nonexclusive identity. Gay concordant has the most liberal respondents, while straight concordant and straight discordant are the most conservative. Gay discordant has the largest proportion of respondents who do not attend religious services, while straight concordant has the smallest. Both discordant groups were more likely to report a different identity in wave IV as in wave III, compared with their concordant counterparts (16 percent vs. 3 percent for straight, 68 percent vs. 35 percent for gay).

I hypothesized that (1) those with a nonexclusive identity would have the highest depressive symptoms scores, (2) those with discordant identities would have higher scores than concordant identities, and (3) concordance would matter more for those with a straight than a gay identity. Table 2 gives the mean depression scores for concordance (all past behavior), concordance (past-year behavior), and concordance (all past behavior) divided by identity subgroups. My hypotheses are partly supported by the mean scores. The nonexclusive identity category has the highest scores when all past behavior is used. However, when only past-year behavior is used, discordant have the highest mean scores. Similarly, although in most cases the discordant group has higher mean scores than its concordant counterpart, this pattern does not hold for gay-identified men. My third hypothesis, concordance will matter more for those with a straight than a gay identity, is hard to judge by simply looking at mean scores. However, it does seem as if it may be true for men (5.9 and 4.6 straight discordant and concordant vs. 5.0 and 5.1 gay discordant and concordant).

Table 2.

Mean Depression Scores by Concordance, Identity, and Gender.

	All Respondents			Women only			Men only
	M	SD	n	M	SD	n	M	SD	n
Concordance (all past behavior)
Concordant	4.9	3.9	11,522	5.2	4.1	5,605	4.6	3.7	5,917
Discordant	6.3	4.7	613	6.7	5.0	385	5.7	4.2	228
Nonexclusive identity	6.8	4.7	1,724	7.0	4.8	1,412	6.0	4.0	312
Concordance (past-year behavior)
Concordant	4.9	3.9	10,723	5.3	4.1	5,184	4.6	3.7	5,539
Discordant	7.0	4.9	147	7.7	5.3	88	6.0	4.2	59
Nonexclusive identity	6.8	4.7	1,499	7.0	4.8	1,252	5.9	3.9	247
Concordance and identity (all past behavior)
Straight concordant	4.9	3.9	11,423	5.2	4.1	5,585	4.6	3.7	5,838
Straight discordant	6.4	4.8	516	6.7	5.0	332	5.9	4.3	184
Gay concordant	5.0	3.6	99	4.3	4.2	20	5.1	3.4	79
Gay discordant	6.0	4.6	97	6.8	5.0	53	5.0	4.0	44
Nonexclusive identity	6.8	4.7	1,724	7.0	4.8	1,412	6.0	4.0	312

When we look at the concordant groups subdivided by gay and straight identity⁷ (see “Concordance and identity [all past behavior]”), we see again that discordance is associated with higher depression scores, for both those who identify as straight and those who identify as gay. Interestingly, the means and standard deviations are more similar within concordance categories than within identity. For example, the gay-concordant group has a mean and standard deviation (5.0 and 3.6) that are closer to straight concordant (4.9 and 3.9) than to gay discordant (6.0 and 4.6). This supports the use of a three-category concordance classification as opposed to one subdivided by gay or straight identity.

The pattern in mean depression scores among concordance categories by gay and straight identity holds for women but not for men (see Table 2). Looking at the mean depressive scores for women, we see that they follow the pattern of the full model in which nonexclusive has the highest mean, followed by discordant and then concordant categories, regardless of gay or straight identity. However, men present an exception to this pattern in that gay concordant has a slightly higher mean depression score than does the gay-discordant category.

To further examine the relationship between concordance and depression, I regressed a continuous depressive symptoms score on my concordance variables. Table 3 shows four models, each run separately using three configurations of my concordance variable. Model 1 contains the full sample but none of the control variables. Model 2 has all respondents and all control variables. Models 3 and 4 include all control variables but are estimated for only female and only male respondents. Each model is run first using concordance (all past behavior), then concordance (all past behavior) subdivided by identity, and last concordance (past-year behavior). Table 3 presents the results using (straight) concordant as the omitted category. However, I run all analyses using each concordance group as the omitted category in turn, which I discuss below.

Table 3.

Linear Regression of Depression Scores on Concordance and on Concordance by Identity.

	Model 1	Model 2	Model 3	Model 4
	(Total Sample, No Controls)	(Total Sample, All Controls)	(Women Only, All Controls)	(Men Only, All Controls)
Concordance (all past behavior)
Concordant	—	—	—	—
Discordant	1.47***	1.26***	0.91*	0.32
Nonexclusive identity	1.78***	0.78**	1.36***	0.66
n	13,859	11,416	6,227	5,189
Concordance (past-year behavior)
Concordant	—	—	—	—
Discordant	2.29***	1.85**	2.74**	0.60
Nonexclusive identity	1.81***	1.15***	1.30***	0.69
n	12,369	10,182	5,479	4,703
Concordance and identity (all past behavior)
Straight concordant	—	—	—	—
Straight discordant	1.59***	0.85**	1.03**	0.51
Gay concordant	0.42	0.32	−0.73	0.36
Gay discordant	0.77	−0.16	0.06	−0.49
Nonexclusive identity	1.78***	1.16***	1.33***	0.64
n	13,859	11,416	6,227	5,189

p < .05, **p < .01, and ***p < .001 (two-tailed test).

Using the three-category concordance (all past behavior and past-year behavior) variables, I find that respondents with discordant and nonexclusive identities have significantly higher depression scores than respondents with concordant identities. When I run the regression using nonexclusive identity as the omitted category, I find no statistically significant difference between the discordant and nonexclusive identity categories in any of the models. I hypothesized that (1) those with nonexclusive identities would have higher scores than both those with discordant and concordant identities and (2) that those with discordant identities would report more depressive symptoms than those with concordant identities. The results described above partly support my hypotheses, in that those with nonexclusive and discordant identities have higher scores than those with concordant identities. However, those with nonexclusive identities do not have higher scores than those with discordant identities.

Looking at concordance and identity in models 1 and 2, the most consistent finding is that straight-concordant respondents have significantly lower depressive symptom scores than do straight-discordant or nonexclusive respondents (p < .001). In model 1, nonexclusive respondents have higher scores than do gay-concordant respondents (p < .05), but this loses significance when controls are added (model 2). In model 2, with all controls, nonexclusive respondents have significantly higher scores than do gay-discordant respondents (p < .05), a difference that is only marginally significant (p < .10) in model 1. In neither model is there a significant difference between the depressive symptoms scores of gay-concordant and gay-discordant respondents. This supports my third hypothesis, because discord is statistically significantly associated with higher depressive symptoms for straight- but not gay-identified respondents.

The findings for the total sample are largely mirrored for women. Analyses using both past-year and all past behavior show that women with concordant identities report fewer depressive symptoms than do women with nonexclusive or discordant identities (p < .05). There is no statistically significant difference between discordant and nonexclusive groups. When identity is considered, the only significant finding is that straight-concordant women report fewer symptoms than straight-discordant and nonexclusive women (p< .05). Thus, women follow the pattern of the total population.

However, these findings do not hold for men, who show no statistically significant differences between concordance groups (p < .05). The only group differences among men to reach even marginal significance are (1) between nonexclusive and concordant when only concordance is considered and (2) between nonexclusive and straight concordant when identity is added in. The results described above are from a model (model 4) containing a control for different sexual orientation identities reported in waves III and IV. When the model is run without this control, the results for men mirror those of women and the larger population. That is, if change in identity is not controlled for, the straight-concordant group has significantly fewer depressive symptoms than the nonexclusive (p < 0.001) and straight-discordant (p < .05) groups. The concordant group has fewer depressive symptoms than the nonexclusive group (p < .001) for both past-year and all past behavior and fewer than the discordant group for all past (but not past-year) behavior. This shift when change in identity is controlled for suggests that change in sexual orientation explains the relationship between concordance and depression for men, in a way it does not seem to for women (for whom analyses including a control for identity change are substantively the same as those without).

Discussion

My study builds on recent scholarship that focuses on unpacking the complexities of sexual orientation in relation to mental health. I hypothesized that young adults with a nonexclusive identity are at greater risk for depressive symptoms than those with an exclusive identity. My findings partly supported this hypothesis. Nonexclusive young adults reported higher depressive symptom scores than those with straight-concordant identities, and in some models this held for gay-concordant and gay-discordant identities as well. However, there is no evidence that nonexclusive respondents have higher scores than straight-discordant respondents. I also hypothesized that those with discordant identities would have higher depressive symptoms scores than those with concordant identities. Results from a model using a simple concordance variable support this hypothesis. However, dividing respondents into subgroups by gay or straight identity shows that it holds only for the straight-identified. This is in line with my third hypothesis, that concordance would matter more for the mental health of straight- than gay-identified respondents.

Gender further complicates the picture. Although women largely mirror the results of the general population, men do not (at least not when a change in sexual orientation identity is controlled for). In the full model containing all controls, there are no significant group differences for men. However, if the control for a change in reported sexual orientation identity is removed, men’s models show similar results to women and the total sample. This suggests that change in sexual orientation identity, rather than discord, is associated with depressive symptoms for men.

Concordance may act as a mediating variable between something else and mental health. For example, concordance could be mediating the effect of internalized stigma or a more homophobic environment. Respondents who have internalized the stigma of nonheterosexuality or who live in an area with greater homophobia may be less likely to claim a nonheterosexual identity, despite a history of same-sex behavior. If this were the case, then the patterns we see in Table 3 may be driven partly by a relationship between internalized stigma or homophobic environment and depression.

It is possible that the difference between gay concordant and gay discordant is not statistically significant because of the normality of heterosexual behavior or attraction. As Gattis et al. (2012) pointed out, it is not surprising that many gay-identified young adults have a history of female-male behavior given that heterosexuality is accepted and expected. I found that discordance is more common among gay-identified young adults than among the straight-identified. Because of this, it may be more accepted and viewed asnormal among gay- identified young adults. Alternatively, it is possible that gay-discordant respondents have more readily available and accepted tools for lessening dissonance. For example, they may be more likely to believe that sexual orientation is fluid, in which case past behavior or attractions would not affect their current identity. Or they may reconstruct their past, defining their past behavior as pre–coming out: the actions of a past self (Mead 1929). Straight-identified young adults may seek to lessen dissonance by claiming experimentation. However, because of the widely held belief that any same-sex behavior or attraction makes one nonheterosexual, straight-identified persons face the chance that others will not believe them (Bostwick and Hequembourg 2013). People may be more likely to question the heterosexuality of someone with past same-sex behavior than the homosexuality of someone with past female-male behavior.

Limitations

My study is limited by small group sizes, which make it difficult to find statistically significant differences among nonheterosexual groups. My data are also limiting in that they do not capture sexual orientation in the level of detail or nuance that current theory would require. For example, although my data can somewhat capture change in sexual orientation identity, this is only through a five-category measure taken at two time points six years apart. Thus, I cannot speak to the fluidity of sexual orientation, nor to those whose identity falls outside the five identity categories provided. Given that these categories are limited, assume a gender binary, and allow a respondent to choose only one sexual orientation, I am unable to say how much my results reflect those whose identities and experiences fall outside of this categorization. Although this is a serious limitation, it is one that is difficult to avoid when using a large survey aimed at a broad population. These types of questions are better answered using qualitative studies or surveys targeted at LGBT populations.

It is also important to note that this study is generalizable only to a specific cohort of cisgender young adults. This is not a weakness, as I believe this cohort has had a distinct experience growing up in a time of rapid change in attitudes toward homosexuality. However, future research should examine the effects of concordance and exclusivity on depression in different age groups. There is also some evidence to suggest that women are more likely to respond to stress with depressive symptoms than men (Piccinelli and Wilkinson 2000). Therefore, it is possible that by using depressive symptoms as my only outcome, I am underestimating the negative mental health effects associated with discordance for men. Future research should address this by testing the relationship between concordance and a range of mental health outcomes.

Conclusion

This study makes an important contribution to the literature by examining the effects of concordance on depressive symptoms in a nationally representative sample of cisgender young adults. The finding that nonexclusive groups have a higher risk for depression points to the need for future research and policy focusing on this group. The differences found between men and women suggest future researchers examine the mental health effects of sexual orientation identity change in men. This study also provides strong evidence against grouping all nonheterosexuals together when examining the relationship between sexual orientation and mental health outcomes. This is particularly important given the finding of no significant difference between gay- and straight-identified young adults within concordance categories. It is time we move away from the heterosexual/nonheterosexual binary. Instead, we should adjust the way we conceptualize and operationalize sexual orientation to allow us to understand the more nuanced relationship between sexual orientation and mental health. With improvements in methods and available data on nonheterosexuals, it is time we stop trying to force square pegs into round holes.

Footnotes

Acknowledgements

I would like to thank Clem Brooks, Arthur Alderson Maria Gerth, and especially Peggy Thoits and Peter Lista for their comments and suggestions on various drafts of this article.

Notes

References

Arnett

Jeffrey Jensen

. 2000. “Emerging Adulthood: A Theory of Development from the Late Teens through the Twenties.” American Psychologist 55(5):469–80.

Balsam

Kimberly F.

Mohr

Jonathan J.

2007. “Adaptation to Sexual Orientation Stigma: A Comparison of Bisexual and Lesbian/Gay Adults.” Journal of Counseling Psychology 54(3):306–19.

Bockting

Walter O.

Miner

Michael H.

Swinburne Romine

Rebecca E.

Hamilton

Autumn

Coleman

Eli

. 2013. “Stigma, Mental Health, and Resilience in an Online Sample of the US Transgender Population.” American Journal of Public Health 103(5):943–51.

Bostwick

Wendy B.

Boyd

Carol J.

Hughes

Tonda L.

McCabe

Sean Esteban

. 2010. “Dimensions of Sexual Orientation and the Prevalence of Mood and Anxiety Disorders in the United States.” American Journal of Public Health 100(3):468–75.

Bostwick

Wendy

Hequembourg

Amy L.

2013. “Minding the Noise: Conducting Health Research among Bisexual Populations and Beyond.” Journal of Homosexuality 60(4):655–61.

Cochran

S. D.

2001. “Emerging Issues in Research on Lesbians’ and Gay Men’s Mental Health: Does Sexual Orientation Really Matter?” American Psychologist 56(11):931–47.

Cragun

Ryan T.

Sumerau

J.Edward

. 2015. “The Last Bastion of Sexual and Gender Prejudice? Sexualities, Race, Gender, Religiosity, and Spirituality in the Examination of Prejudice toward Sexual and Gender Minorities.” Journal of Sex Research 52(7):821–34.

Drabble

Laurie

Midanik

Lorraine T.

Trocki

Karen

. 2005. “Reports of Alcohol Consumption and Alcohol-related Problems among Homosexual, Bisexual and Heterosexual Respondents: Results from the 2000 National Alcohol Survey.” Journal of Studies on Alcohol 66(1):111–20.

Everett

Bethany

. 2015. “Sexual Orientation Identity Change and Depressive Symptoms: A Longitudinal Analysis.” Journal of Health and Social Behavior 56(561):37–58.

10.

Festinger

Leon

. 1957. A Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press.

11.

Gamson

Joshua

. 1995. “Must Identity Movements Self-Destruct? A Queer Dilemma.” Social Problems 42(3):390–407.

12.

Gattis

Maurice N.

Sacco

Paul

Cunningham-Williams

Renee M.

2012. “Substance Use and Mental Health Disorders among Heterosexual Identified Men and Women Who Have Same-sex Partners or Same-sex Attraction: Results from the National Epidemiological Survey on Alcohol and Related Conditions.” Archives of Sexual Behavior 41(5):1185–97.

13.

Hequembourg

Amy L.

Brallier

Sara A.

2009. “An Exploration of Sexual Minority Stress across the Lines of Gender and Sexual Identity.” Journal of Homosexuality 56(3):273–98.

14.

Herek

Gregory M.

2004. “Beyond ‘Homophobia’: Thinking about Sexual Prejudice and Stigma in the Twenty-first Century.” 1(2):6–24.

15.

Herek

Gregory M.

Garnets

Linda D.

2007. “Sexual Orientation and Mental Health.” Annual Review of Clinical Psychology 3:353–75.

16.

Igartua

Karine

Thombs

Brett D.

Burgos

Giovani

Montoro

Richard

. 2009. “Concordance and Discrepancy in Sexual Identity, Attraction, and Behavior among Adolescents.” Journal of Adolescent Health 45(6):602–8.

17.

King

Michael

Semlyen

Joanna

Tai

Sharon See

Killaspy

Helen

Osborn

David

Popelyuk

Dmitri

Nazareth

Irwin

. 2008. “A Systematic Review of Mental Disorder, Suicide, and Deliberate Self Harm in Lesbian, Gay and Bisexual People.” BMC Psychiatry 8:70.

18.

Laumann

Edward O.

Gagnon

John H.

Michael

Robert T.

Michaels

Stuart

. 1994. The Social Organization of Sexuality. Chicago: University of Chicago Press.

19.

Lindley

Lisa L.

Walsemann

Katrina M.

Carter

Jarvis W.

2012. “The Association of Sexual Orientation Measures with Young Adult’s Health-related Outcomes.” American Journal of Public Health 102(6):1177–85.

20.

Mead

George Herbert

. 1929. “The Nature of the Past.” Pp. 235–42 in Essays in Honor of John Dewey, edited by Cross

New York: Henry Holt.

21.

Meyer

Ilan H.

1995. “Minority Stress and Mental Health in Gay Men.” Journal of Health and Social Behavior 36(1):38–56.

22.

Meyer

Ilan H.

2003. “Prejudice, Social Stress, and Mental Health in Lesbian, Gay, and Bisexual Populations: Conceptual Issues and Research Evidence.” Psychological Bulletin 129(5):674–97.

23.

Ott

Miles Q.

Corliss

Heather L.

Wypij

David

Rosario

Margaret

Austin

S. Bryn

. 2011. “Stability and Change in Self-reported Sexual Orientation Identity in Young People: Application of Mobility Metrics.” Archives of Sexual Behavior 40(3):519–32.

24.

Pew Research Center. 2013. A Survey of LGBT Americans: Attitudes, Experiences and Values in Changing Times. Washington, DC: Pew Research Center.

25.

Piccinelli

Marco

Wilkinson

Greg

. 2000. “Gender Differences in Depression: Critical Review.” British Journal of Psychiatry 177:486–92.

26.

Poston

Dudley

Baumle

Amanda

. 2010. “Patterns of Asexuality in the United States.” Demographic Research 23:509–30.

27.

Radloff

Lenore Sawyer

. 1977. “A Self-report Depression Scale for Research in the General Population.” Applied Psychological Measurement 1(3):385–401.

28.

Richards

Christina

Bouman

Walter Pierre

Seal

Leighton

Barker

Meg John

Nieder

Timo O.

T’Sjoen

Guy

. 2016. “Non-binary or Genderqueer Genders.” International Review of Psychiatry 28(1):95–102.

29.

Ripley

Matthew

Anderson

Eric

McCormack

Mark

Adams

Adrian

Pitts

Robin

. 2011. “The Decreasing Significance of Stigma in the Lives of Bisexual Men: Keynote Address, Bisexual Research Convention, London.” Journal of Bisexuality 11(2–3):195–206.

30.

Ross

Lori E.

Dobinson

Cheryl

Eady

Allison

. 2010. “Perceived Determinants of Mental Health for Bisexual People: A Qualitative Examination.” American Journal of Public Health 100(3):496–502.

31.

Russell

S. T.

Joyner

2001. “Adolescent Sexual Orientation and Suicide Risk: Evidence from a National Study.” American Journal of Public Health 91(8):1276–81.

32.

Savin-Williams

R. C.

1994. “Verbal and Physical Abuse as Stressors in the Lives of Lesbian, Gay Male, and Bisexual Youths: Associations with School Problems, Running Away, Substance Abuse, Prostitution, and Suicide.” Journal of Consulting and Clinical Psychology 62(2):261–69.

33.

Savin-Williams

Ritch C.

Ream

Geoffrey L.

2007. “Prevalence and Stability of Sexual Orientation Components during Adolescence and Young Adulthood.” Archives of Sexual Behavior 36(3):385–94.

34.

Savin-Williams

Ritch C.

Vrangalova

Zhana

. 2013. “Mostly Heterosexual as a Distinct Sexual Orientation Group: A Systematic Review of the Empirical Evidence.” Developmental Review 33(1):58–88.

35.

Talley

Amelia E.

Aranda

Frances

Hughes

Tonda L.

Everett

Bethany

Johnson

Timothy P.

2015. “Longitudinal Associations among Discordant Sexual Orientation Dimensions and Hazardous Drinking in a Cohort of Sexual Minority Women.” Journal of Health and Social Behavior 56(2):225–45.

36.

Thoits

Peggy A.

2011. “Mechanisms Linking Social Ties and Support to Physical and Mental Health.” Journal of Health and Social Behavior 52(2):145–61.

37.

Vrangalova

Zhana

Savin-Williams

Ritch C.

2012. “Mostly Heterosexual and Mostly Gay/Lesbian: Evidence for New Sexual Orientation Identities.” Archives of Sexual Behavior 41:85–101.

38.

Vrangalova

Zhana

Savin-Williams

Ritch C.

2014. “Psychological and Physical Health of Mostly Heterosexuals: A Systematic Review.” Journal of Sex Research 51(4):410–55.

39.

Zhao

Yue.

2010. “The Association of Sexual Identity, Attraction, and Behavior with Suicidal Ideation and Attempts among Adolescents.” Unpublished manuscript, McGill University, Montreal, QC, Canada.