Childhood Experiences and Mental Health of Sexual Minority Adults: Examining Three Models

Abstract

Three models of recalled childhood gender nonconformity (GNC) and maltreatment are proposed to explain disparities in current psychological distress and lifetime suicidality among sexual minority individuals, using a United States probability sample of cisgender lesbian/gay (n=701), bisexual (n=606), and other (e.g., queer, n=182) adults. Indirect effects indicated that lesbian/gay individuals were more likely than bisexual individuals to experience maltreatment and suicidal ideation as childhood GNC increased. Other indirect effects found that bisexual individuals reported more psychological distress and greater likelihood of lifetime suicidal ideation and attempts than lesbian/gay individuals as maltreatment increased. The direct effects of sexual orientation were stronger than the indirect effects via maltreatment or GNC, with bisexual individuals reporting more maltreatment, distress, and suicidality than lesbian/gay individuals. Significant findings for individuals with other identities were similar to those of bisexual individuals. Adjusted findings were comparable for women and men. The findings indicated that sexual minority individuals reported experiences consistent with sexual minority stress during early developmental periods, before being aware of their sexual orientation. It is necessary to understand the early lived experiences of sexual minority individuals, differences between lesbian/gay and bisexual individuals in those experiences, and their implications for adaptation.

Lesbian/gay, bisexual, and other sexual minority individuals are at elevated risk for poor health relative to heterosexual peers (National Academies of Sciences, Engineering, and Medicine, 2020), with the health disparity attributed to experiencing minority stress for being a sexual minority (Meyer, 2003). Both lesbian/gay and bisexual individuals are each more likely to have poor mental and physical health relative to heterosexual peers, as United States (US) representative samples of the population have found (Cochran et al., 2016; Fredriksen-Goldsen et al., 2017; Kerridge et al., 2017; Ward et al., 2015). In addition, poor health is more common among bisexual than lesbian/gay individuals, as documented by mental and physical health studies (Kerridge et al., 2017; Pharr, 2021) and by meta-analytic studies on mental health (Ross et al., 2018; Salway et al., 2019). Minority stress has been adapted to explain the health disparities of bisexual individuals, with monosexism referring to experiencing prejudice and discrimination from heterosexual and lesbian/gay individuals for being attracted to both rather than one biological sex (McLean, 2018). However, a pathway to poor mental and physical health also must exist for lesbian/gay individuals to account for their poorer health relative to heterosexual individuals. The principle of equifinality asserts more than one possible pathway to the same adaptive outcome (Cicchetti & Rogosch, 1996). In addition, the different pathways for lesbian/gay and bisexual individuals suggests three possible health outcomes. First, the pathway for lesbian/gay individuals may not be as strong as the one for bisexual individuals. Second, the health effects of the two pathways may be relatively equal, ensuring nonsignificant findings between lesbian/gay and bisexual individuals. Third, the poorer health of lesbian/gay individuals relative to bisexual individuals may be limited to a few health markers. The findings by Kerridge and colleagues (2017) and Pharr (2021) support the first two possibilities. However, they cannot address the last possibility because not all markers of poor health were examined.

The minority stress explanations for poor health rest on the premise that individuals know their sexual minority status (Meyer, 2003). However, a broad definition of minority, gay-related, or sexual minority stress refers to experiencing society’s stigmatization for being or being perceived to be a sexual minority (Rosario et al., 2002). Therefore, sexual minority stress may be experienced in childhood, before any self-awareness of a minority status. This may happen if the perpetrator consciously or unconsciously infers or suspects a child may be a nascent sexual minority.¹

Childhood sexual minority stress differs from Meyer’s (2003) minority stress for developmental reasons. For example, internalized heterosexism, identity concealment, and other markers of proximal stress do not exist because awareness of one’s sexual orientation is lacking. For another example, the mediational processes added to minority stress theory (Hatzenbuehler, 2009) are irrelevant, given their cognitive, coping, and social functioning demands exceed the capabilities of children. Consequently, the extant literature on minority stress, its health correlates, and potential mediators and moderators is not pertinent to childhood minority stress, except in the most general sense that stress has negative consequences for health.

We underscore that a great deal of childhood sexual minority stress may have been experienced by the time an individual becomes aware of possibly being a sexual minority. This means that allostatic load or the “wear and tear” of stress on the individual has been increasing over time, taxing physiological systems and compromising mental and physical health (Juster et al., 2010; McEwen, 1998; McEwen & Seeman, 1999).

The effects of childhood sexual minority stress may be noteworthy for two other reasons. First, this stress has been experienced during a developmental time when cognitive, coping, emotion-regulation capabilities, and other individual resources are limited. Second, childhood is a sensitive developmental period because early experiences may have cascading effects over time and across domains, scaffolding the individual’s continuing development or generating vulnerabilities that are carried forward into adulthood (Masten & Cicchetti, 2010). Consequently, early experiences may account for subsequent health disparities between lesbian/gay and bisexual individuals.

Given both lesbian/gay and bisexual individuals report poorer mental and physical health than heterosexual peers (e.g., Fredriksen-Goldsen et al., 2017; Kerridge et al., 2017), different childhood pathways to health are likely for each group, with a pathway accounting for poorer health among lesbian/gay individuals and another for poorer health among bisexual individuals, even for the same health outcome. In this article, we propose three models concerning the roles of childhood gender nonconformity (GNC) and childhood maltreatment in the health of lesbian/gay and bisexual adults.

Model 1a: Sexual Orientation → GNC → Maltreatment → Poor Health

Model 1a specifies that sexual orientation is implicated in childhood GNC. It adds that GNC leads to experiences of childhood maltreatment, and maltreatment results in poor mental and physical health. The model is premised on societal beliefs that link GNC with homosexuality. In societies in which homosexuality is stigmatized, the link between GNC and homosexuality may prove stressful for children who may be sexual minority in the future. We attribute the link to the connection between masculinity and heterosexuality (Diefendorf & Bridges, 2020). Gender and sexuality are intertwined in ways that support and reinforce gendered roles and heterosexuality, while stigmatizing deviations from these sociocultural objectives (Szymanski & Henrichs-Beck, 2014). The socialization is hypothesized to occur via modelling and reinforcement (Bussey & Bandura, 1999). Its effects are apparent early in life (Egan & Perry, 2001).

Sexual orientation and childhood GNC.

Recalled childhood GNC is more common among sexual minority than heterosexual peers, as literature reviews have documented (Bailey & Zucker, 1995; Zucker, 2008). In an observational study based on home videos of children, adult raters judged children who grew up to be sexual minority as more GNC² in childhood than heterosexual controls (Rieger et al., 2008). A longitudinal study found that mothers rated children who would be sexual minority in adolescence as more GNC than heterosexual children at ages 2.5, 3.5, and 4.75 years (Xu et al., 2019). As may be inferred, GNC extends over time from childhood through subsequent development (e.g., Golombok et al., 2012).

GNC varies with sexual orientation, and studies have predominantly found lesbian/gay individuals to be more GNC than other sexual minority individuals. A study found that lesbian/gay individuals were most GNC in behaviors, followed by bisexual individuals, then mostly heterosexual individuals, with completely heterosexual individuals being the least GNC (Roberts et al., 2012). Although another study supported these findings (Baams, 2018), a representative sample of the US population found that GNC in young adulthood was higher among lesbian/gay individuals than bisexual, mostly heterosexual, and heterosexual individuals, with no significant differences among the last three groups (Li et al., 2016).

Sexual orientation and childhood maltreatment.

Childhood emotional, physical, and sexual abuse is more common among sexual minority than heterosexual individuals, as documented by a meta-analysis (Friedman et al., 2011).

Childhood GNC and maltreatment.

Despite improving attitudes toward and greater acceptance of sexual minority individuals (Pew Research Center, 2020), homonegative prejudice and discrimination persist in the US (Herek & McLemore, 2013), Europe (Štulhofer & Rimac, 2009), and elsewhere (Pew Research Center, 2020). In 2019, 21% of the US population agreed that homosexuality was unacceptable (Pew Research Center 2020). Homonegative attitudes may affect how parents respond to a child whom they infer or suspect may be sexual minority by means of GNC. Studies have found that parents are less accepting of and respond negatively to offspring who are GNC (D’Augelli et al., 2006; van Beusekom et al., 2015).

Negative parental reactions may occur before a child is aware of being a sexual minority. The parent may maltreat the child to eliminate the GNC behaviors and “reorientate” the inferred sexual orientation. The parent may ignore the child, becoming increasingly less accessible, which may render the child vulnerable to abuse by others. The negative parental behaviors may range from relatively benign to severe and cover the breadth of abuse. Research has found that GNC is related to experiencing emotional and physical abuse (Baams, 2018; Roberts et al., 2012) and, in the general or predominantly heterosexual population, to being sexually abused (Assink et al., 2019).

We hypothesize that lesbian/gay individuals are more likely than bisexual individuals to experience childhood maltreatment because lesbian/gay are more GNC, and those who are more GNC are at elevated risk for maltreatment.

Relations of childhood GNC and maltreatment to health.

We hypothesize that negative reactions provoked by GNC and inherent to maltreatment are related to poor health. Indeed, GNC is related to psychological distress among sexual minority youth (D’Augelli et al., 2006; Roberts et al., 2012; Toomey et al., 2010). Furthermore, in a representative sample of the US population, GNC during young adulthood was related to depression symptomatology, independent of sexual orientation (Li et al., 2016). It also was related to suicidality, regardless of sexual orientation (Fitzpatrick et al., 2005). In some studies, the relation between GNC and psychological distress has been found only among men (Lippa, 2008; Skidmore et al., 2006), consistent with evidence that gender atypicality is less tolerated among men than women, regardless of sexual orientation (Lippa, 2008).

Studies of adverse childhood experiences (ACEs), a composite of maltreatment and family dysfunction, find that ACEs are related to negative changes in physiological systems (for a narrative review of the literature, see Danese & McEwen, 2012), including genetic damage via shortening of telomeres (Shalev et al., 2013). In addition, ACEs have been related to increased likelihood of recent suicidal ideation and attempts among lesbian, gay, and bisexual youth (Clements-Nolle et al., 2018). More important for us, studies suggest that ACEs may mediate the relation between sexual orientation and health. In a representative sample of the US population, the disparity between sexual minority and heterosexual adults in depression (OR = 2.42) decreased after adjusting for ACEs (OR = 1.88) (Austin et al., 2016). Another US representative sample found that ACEs reduced the sexual orientation disparity in psychological distress between sexual minority and heterosexual adults (Blosnich & Andersen, 2015). Still another US representative sample found that adjusting for child physical abuse reduced the disparities between lesbian/gay and heterosexual individuals and between bisexual and heterosexual individuals in depression symptomatology and a composite of suicidal ideation and attempts (McLaughlin et al., 2012).

Model 1b: Sexual Orientation → GNC → Poor Health

The proposed Model 1a just discussed implicates childhood maltreatment as a response to a child’s GNC. However, not all parents maltreat their GNC children, even if they still attempt to steer them toward more gender typical behaviors. Despite this possibility, the GNC child may experience negative reactions by non-household members, as the non-familial environment asserts itself. The literature finds that children who are more gender conforming report higher self-worth and more peer social competence, and their peers report liking and accepting them more (Egan & Perry, 2001). Moreover, children who are more GNC experience more (peer-reported) internalizing symptomatology when pressured to be more gender typical (Yunger et al., 2004). The resulting variant Model 1b restricts mediation to GNC, hypothesizing that lesbian/gay individuals have poorer health than bisexual individuals because they were more GNC in childhood.

Implications of Models 1a and 1b.

These models potentially provide a pathway to poorer health for lesbian/gay relative to bisexual individuals. Childhood GNC is essential to both models, but it is not as extreme or common for bisexual individuals as it is for lesbian/gay individuals (Roberts et al., 2012). A different pathway to poor health must exist for bisexual individuals.

Model 2: Sexual Orientation → Maltreatment → Poor Health

We propose that bisexual individuals experience poorer mental and physical health than lesbian/gay individuals because they experience more childhood maltreatment. This occurs regardless of childhood GNC, which is why GNC does not appear in the model depicted in the heading. In other words, even after adjusting for GNC, bisexual individuals will report poorer health than lesbian/gay individuals, given their maltreatment experience.

Maltreatment is more common among bisexual than lesbian/gay individuals, in keeping with Model 2. Pooled representative samples of the US population (Merrick et al., 2018) indicate that bisexual individuals are more likely than lesbian/gay peers to report child emotional abuse (59% vs. 47%), physical abuse (35% vs. 31%), and sexual abuse (35% vs. 23%). A meta-analysis found that bisexual individuals were 25% more likely than lesbian/gay peers to ever attempt suicide, although no significant difference was found in lifetime ideation (Salway et al., 2019). These and comparable mental health findings by others (Kerridge et al., 2017; Ross et al., 2018) suggest maltreatment may operate independent of GNC to explain differences in health between bisexual and lesbian/gay individuals.

Sex Differences

Sex differences in the models’ variables have been studied. Female and male sexual minority individuals do not differ significantly on childhood GNC, but parental reactions to such behaviors are more negative for male than for female sexual minority children (D’Augelli et al., 2008). Although sexual abuse is more prevalent among girls than boys in the general population (Assink et al., 2019), sexual abuse is more common among male than female sexual minority individuals relative to heterosexual peers (Friedman et al., 2011). This finding raises another concern: Do the relations among the models’ variables vary by sex? We address these concerns after examining the models.

Summary of Models and Hypotheses

We propose different mental and physical health pathways for bisexual and lesbian/gay individuals, with each pathway capturing different childhood experiences. Our three models do not mean that bisexual individuals, on average, may not have poorer health than lesbian/gay individuals. The models address two different issues: First, why is each group susceptible to poor health as a result of childhood experiences? Second, what mechanisms potentially explain their poor health risks relative to each other? The latter implies that the pathway to poor health differs between the two groups, with one pathway resulting in poor health for bisexual individuals and another pathway for poor health for lesbian/gay individuals. Accordingly, bisexual and lesbian/gay individuals may differ on health relative to each other.

Models 1a and 1b propose that GNC and maltreatment (1a) or just GNC (1b) explain how lesbian/gay individuals may have poor health. Parents and other individuals respond negatively to GNC children (D’Augelli et al., 2006; van Beusekom et al., 2015), and lesbian/gay individuals are more GNC in childhood than are bisexual individuals (Roberts et al., 2012). Consequently, we precede tests of Models 1a and 1b with examination of the nested model that concerns the relations among sexual orientation, GNC, and maltreatment.

The third model, Model 2, proposes that the poorer health that has been found among bisexual relative to lesbian/gay individuals (e.g., Kerridge et al., 2017) is attributed to maltreatment, independent of GNC. Figure 1 depicts Models 1a, 1b, and 2.

Figure 1.

Hypothesized Models of the Mediating Roles of Childhood Gender Nonconformity and Childhood Maltreatment in the Relations Between Sexual Orientation and Health. (Note. LG = lesbian/gay and BI = bisexual; GNC = gender nonconformity. For Model 1b, maltreatment, which concerns household maltreatment, is irrelevant. For Model 2, GNC is irrelevant. The direction of the hypothesized relation as positive (+) or negative (−) is depicted.)

Although the health literature reviewed above for the models pertains to mental health, specifically psychological distress and suicidality, the theoretical models apply more broadly to both mental and physical health, given early childhood experiences. Maltreatment increases allostatic load (Danese & McEwen, 2012), which has negative implications for mental and physical health (Juster et al., 2010; McEwen, 1998; McEwen & Seeman, 1999).

Finally, we examine sex differences. We do not expect to find significant sex differences in our models once adjustments are made for potential confounding variables.

Methods

Participants

A US probability sample was drawn from all 50 states and the District of Columbia to participate in Generations, a study of cisgender lesbian/gay and bisexual adults. Eligible individuals were cisgender; identified as lesbian, gay, or bisexual; belonged to 1 of 3 generations; identified as Hispanic, non-Hispanic White or Black, or multiethnic/multiracial; reported at least a sixth-grade education; and spoke English sufficiently well to complete the telephone screener in English. Of 1,563 individuals, 45 were excluded because 27 identified as transgender and 18 were under age 18 years or provided an age in the screener discrepant with their reported year of birth in the survey. The transgender individuals were moved into a parallel study focusing on transgender individuals. The analytic sample contained 1,518 unweighted cases, which approximately represented 1,514 weighted cases.

The Gallup Organization recruited the sample from March 2016 to March 2018, using its dual-frame sampling procedure, which included random dialing of landline and cellular telephones. Individuals provided oral consent to be screened for eligibility over the telephone. Eligible participants completed the survey online or as hard copy, and received a $25 gift certificate. The study was approved by the IRB of all involved institutions. Details on the participants and recruitment procedures are available elsewhere (Meyer et al., 2020).

Measures

Sexual orientation.

Respondents were asked, “Which of the following best describes your current sexual orientation:” straight/heterosexual, lesbian, gay, bisexual, queer, same-gender loving, or other. Those who indicated other (n=71) could provide a response. Those responses were recoded when possible (e.g., “dyke” into lesbian). The remaining responses were pansexual, anti-label, asexual, and other. We excluded 30 respondents who identified as asexual (n=19) or heterosexual (n=11). The heterosexual individuals indicated they were lesbian, gay, or bisexual on the screener, but heterosexual on the survey. The final sexual-minority categories were lesbian/gay, bisexual, and other. The other category consisted of individuals who identified as queer, same-gender loving, pansexual, anti-label, or other.

Childhood GNC.

Four items were selected from a measure of recalled GNC (Zucker et al., 2006). The items assessed gendered behavior as a child with respect to favorite toys and games, imitated or admired television or movie characters, role enacted in fantasy play, and feeling of self. The response scales ranged from masculine (1=always masculine or always boys or men) to feminine (5=always feminine or always girls or women), with the midpoint representing equally masculine and feminine. In addition, systematic missing was represented by responses of not applicable or neither for behaviors in which the individual did not engage (e.g., toys were not gendered). Responses were recoded to 0–4 and scored by biological sex. These 4 items were validated by others (e.g., Roberts et al., 2012). We submitted them to principal component factor analysis. A single factor emerged with an eigenvalue of 2.82 that explained 70.6% of the variance. All items loaded .80 or higher (α=.86). The mean was computed and multiplied by 4 to produce a sum of the items. Higher scores indicated higher childhood GNC.

Childhood maltreatment.

Events occurring before age 18 years were assessed with items from the Behavioral Risk Factor Surveillance System’s ACEs module (Centers for Disease Control and Prevention, 2010). Six items assessed emotional/psychological, physical, and sexual abuse (e.g., “how often did your parents or adults in your home ever slap, hit, beat, kick or physically hurt you in any way”). We dichotomized the 3-point response scale of never, once, or more than once to never (0) and ever (1) before summing the scores to assess the number of maltreatment experiences from 0 to 6 (α=.75).

Psychological distress.

We used the Kessler-6 scale (Kessler et al., 2002) to assess how often, in the past 30 days, respondents experienced 6 symptoms (e.g., felt nervous, hopeless), using a response scale ranging from all of the time (1) to none of the time (5). Responses were reversed coded when necessary and rescaled from 0 to 4 before being summed (α=.89). Scores ranged from 0 to 24, with higher scores indicating more distress. Missing data (n=27) on any item were imputed using predictive mean matching (Little, 1988).

Suicidality.

We adapted items (Ursano et al., 2014) to assess lifetime suicidality: “Did you ever in your life have thoughts of killing yourself?” and subsequently asked, “Did you ever make a suicide attempt (i.e., purposefully hurt yourself with at least some intention to die)?” Responses were no (0) and yes (1).

Biological sex and gender identity.

Respondents were queried about their “sex assigned at birth, on your original birth certificate” as female or male. They were asked to describe their current gender identity as woman, man, transgender woman/male-to-female (MTF), transgender man/female-to-male (FTM), nonbinary/genderqueer. Only cisgender individuals were included in the study, as indicated above.

Ethnicity and race.

Eligibility restrictions assured sufficient numbers of Hispanic, non-Hispanic Black, and non-Hispanic White individuals in each generation; we henceforth refer to the last 2 as “White” or “Black” for ease of presentation. Non-Hispanic Asians were excluded a priori, given their small representation in the US population, which, when coupled with the relatively smaller population of lesbian, gay, and bisexual individuals, ensured that few Asian lesbian, gay, and bisexual potential participants would be identified. Eligible respondents were re-assessed in the survey. Nineteen individuals were excluded (e.g., only identified as Asian in the survey, but not on the screener).

Generations.

Individuals were assigned to one of 3 age cohorts defined by major events in lesbian, gay, and bisexual history in the US: The Stonewall Inn riots of 1969, the formation of the AIDS organization, ACT UP, in 1987, and the Massachusetts’ Supreme Court ruling legalizing same-sex marriage in 2003. Respondents who were aged 7–13 years during one of these events were included in the corresponding younger (aged 18–25 years), middle (aged 34–41), and older (aged 52–61) cohort. We used generations rather than age in analyses because the study was designed to recruit members of these generations.

Income.

Participants indicated their gross annual household income based on wages and all other sources. Twelve income categories were created. The midpoint for each category was calculated, adjusted for household size, and scaled for 3-person households (Pew Research Center, 2015). Based on the 2016 US median household income of $57,617 for a household of 3 (United States Census Bureau, 2017), the adjusted incomes were categorized as lower (< $38,244), middle ($38,244--$115,234), and upper (>$115,234).

Data Analysis

Sampling weights were used in all statistical analyses to generalize to a national lesbian, gay, and bisexual population. Analyses were conducted in R version 3.6.1 using the survey package for estimation with sampling weights (Lumley, 2004), except for the mediation models which used a different R package (see below).

Descriptive statistics were generated. Sexual minority group differences were assessed using F-test and effect size η² for ANOVA models for continuous variables, and χ² test of independence and effect size Cramér’s V for categorical variables. Pearson correlations were computed. Total effects of sexual minority groups on mental health were estimated with linear regression for psychological distress and logistic regression for suicidal ideation and attempts. Predictors were sexual minority groups, childhood GNC, childhood maltreatment, and covariates (sex, ethnicity/race, generations, and income groups). Unless indicated otherwise, the lesbian/gay individuals composed the reference group in all analyses, including the mediational analyses.

Mediation models.

Possible mechanisms of sexual orientation on mental health via GNC and maltreatment were examined using the general approach developed by Imai, Keele, Tingley, and Yamamoto (2010) and programmed in the R package mediation. This framework approaches mediation by estimating 2 statistical models. In the first model, the mediator(s) (GNC or maltreatment) is regressed on the primary predictor of interest (sexual minority groups) and the covariates. In the second model, the (mental health) outcomes are each regressed on the mediator(s), primary predictor, and covariates. Based on this procedure, the average causal mediation effect (ACME or indirect effect) and the average direct effect (ADE) are obtained for each sexual minority group using 1,000 Monte Carlo simulations while accommodating differences in the distributions of mediators and outcomes when estimating ACME and ADE (Imai, Keele & Tingley, 2010). This approach circumvents limitations of the product of coefficients approach when dealing with mediators and outcomes with different distributions (Imai, Keele & Tingley, 2010).

For the regression and mediation models concerning distress, we standardized GNC, maltreatment, and distress. We also tested moderating sex effects. Power analyses indicated adequate power to detect any effects other than meaningless ones.

Results

The weighted sample of approximately 1,514 individuals consisted of lesbian/gay (n=701), bisexual (n=606), and other (n=182) individuals. Nineteen unweighted cases lacked sexual orientation data. They represent approximately 25 weighted individuals, which sum to 1,514 weighted cases when added to the numbers of lesbian/gay, bisexual, and other individuals.

We found a relatively large number of individuals who self-identified as other than lesbian, gay, or bisexual. They were classified as “other” on sexual orientation. Although we had no hypotheses about these individuals, we included them in the analyses for exploratory purposes. They and the bisexual individuals were compared to the lesbian/gay individuals in the adjusted analyses.

Unadjusted Relations

Descriptive statistics appear in Table 1. Significant mean differences indicated that bisexual individuals were less GNC in childhood than lesbian/gay or other individuals, but they experienced more childhood maltreatment than lesbian/gay individuals. Bisexual individuals also reported more recent psychological distress and lifetime suicidality (i.e., suicidal ideation and attempts) than lesbian/gay individuals. Other individuals did not differ significantly from bisexual individuals on maltreatment or any mental health outcome, but they reported more psychological distress and suicidality than lesbian/gay individuals.

Table 1.

Descriptive Statistics for Weighted Sample of Sexual Minority Individuals.

	Total (n = 1,513.9)	Lesbian/gay (n = 700.7)	Bisexual (n = 605.9)	Other (n = 182.2)	Effect Size
	M (SD)	M (SD)	M (SD)	M (SD)	η2
Childhood gender nonconformity (GNC)	10.96 (3.44)	11.41 (3.80)a	10.14 (2.90)b	11.58 (3.12)a,c	0.03***
Childhood maltreatment	2.10 (1.79)	1.89 (1.77)a	2.37 (1.82)b	2.07 (1.73)a,b	0.01***
Psychological distress (past 30 days)	8.82 (5.62)	7.03 (5.16)a	10.37 (5.54)b,c	10.01 (5.62)c	0.06***
Age	30.86 (13.30)	35.39 (14.86)a	26.85 (10.33)b	27.71 (10.64)b	0.08***
	% (n)	% (n)	% (n)	% (n)	Cramér’s V
Suicidal ideation, ever	74.17 (1112)	66.16 (457.50)a	79.27 (477.10)b	84.59 (152.48)b,c	0.17***
Suicide attempt, ever	27.54 (411.6)	21.47 (147.3)a	32.59 (196.1)b	34.86 (63.16)b,c	0.13***
Sex:
Female	59.99 (908.3)	39.02 (273.5)a	76.75 (465.0)b	81.40 (148.34)b	0.40***
Male	40.01 (605.6)	60.98 (427.3)	23.25 (140.9)	18.60 (33.89)
Generation:
Older	17.39 (263.3)	29.78 (208.7)a	6.39 (38.71)b	8.36 (15.24)b	0.22***
Middle	20.86 (315.9)	22.19 (155.5)a	20.48 (124.10)a	19.08 (34.77)a
Younger	61.74 (934.70)	48.03 (336.6)a	73.13 (443.07)b	72.56 (132.22)b
Ethnicity/race:
Non-Hispanic White	60.14 (900.9)	58.34 (405.09)a	63.72 (380.86)a	51.54 (93.70)a	0.08*
Non-Hispanic Black	13.63 (204.2)	15.37 (106.75)a,b	10.87 (64.98)b	17.11 (31.11)a
Hispanic	10.94 (163.9)	13.04 (90.54)a	8.50 (50.80)a	12.38 (22.50)a
Multiethnic/multiracial	15.29 (229.0)	13.25 (91.99)a	16.90 (101.03)b	18.97 (34.49)a,b
Income:
Upper	25.27 (382.6)	35.22 (246.8)a	17.36 (105.2)b	15.67 (28.55)b	0.16***
Middle	35.41 (536.1)	33.57 (235.3)a	34.38 (208.3)a,b	43.30 (78.90)b
Lower	39.32 (595.3)	31.20 (218.7)a	48.26 (292.4)b	41.03 (74.77)b,c

Note. Of the total sample, approximately 25 cases were missing data on sexual orientation. The other category contains individuals who identified as queer, same-gender loving, pansexual, anti-label, or other. M = mean and SD = standard deviation. Effects sizes are eta square (η²) for ANOVA and Cramér’s V for chi square. Significant F or χ² is indicated after the effect size. Significant sexual orientation pairwise differences (p < .05) are noted by differing superscripts.

^*p < .05.

^**p < .01.

^***p < .001.

Several sociodemographic differences are noted in Table 1. For example, significantly fewer women than men were lesbian/gay, but more women than men were bisexual or other. These and other sociodemographic differences indicated the need to adjust for the demographic variables in the multivariate analyses,

The bivariate relations (Table 2) indicated that individuals who were more GNC in childhood reported more maltreatment and were more likely to have ever ideated or attempted suicide. Those with more maltreatment experienced more recent psychological distress and lifetime suicidality.

Table 2.

Pearson Correlations.

	1	2	3	4	5	6	7
1. Bisexual v. lesbian/gay and other	1.00	---	−0.19***	0.12***	0.24***	0.10***	0.09**
2. Other v. bisexual and lesbian/gay		1.00	0.07**	−0.01	0.08*	0.09***	0.06
3. Childhood gender nonconformity			1.00	0.10**	0.04	0.07*	0.07*
4. Childhood maltreatment				1.00	0.22***	0.15***	0.26***
5. Psychological distress (past 30 days)					1.00	0.35***	0.29***
6. Suicidal ideation, ever						1.00	0.34***
7. Suicidal attempt, ever							1.00
Demographic characteristics
Male v. female	−0.29***	−0.17***	−0.23***	−0.11***	−0.18***	−0.07*	−0.11***
Older v. middle and younger generation	−0.24***	−0.09***	−0.07**	−0.01	−0.28***	−0.14***	−0.07**
Middle v. older and younger generation	−0.01	−0.02	0.01	.14***	−0.10***	−0.08*	−0.04
Black v. Hispanic, multiethnic/multiracial, and White	−0.07*	0.04	0.03	0.07	0.01	−0.06	0.03
Hispanic v. Black, multiethnic/multiracial, and White	−0.07*	0.02	0.05	0.09**	−0.01	−0.05	−0.01
Multiethnic/multiracial v. Black, Hispanic, and White	0.03	0.04	0.01	0.05	0.09**	0.06*	0.05
Upper v. middle and lower income	−0.16***	−0.08***	−0.07*	−0.18***	−0.24***	−0.14***	−0.11***
Middle v. upper and lower income	−0.01	0.06	−0.04	−0.05	−0.02	0.06	−0.03

Note. For every categorical variable, the first group mentioned = 1 and the group(s) mentioned after “v” = 0. The contrasts of categorical variables with more than 2 categories represent the coding necessary to generate dummy coded variables in the adjusted, multivariate analyses.

^*p < 05.

^**p < .01.

^***p < .001.

Adjusted Relations

Model variables.

Table 3 provides the adjusted relations for mental health. Bisexual and other individuals reported more recent psychological distress and lifetime suicidal ideation and attempts than lesbian/gay individuals. With each additional point increase in childhood GNC, individuals were 18% more likely to have a history of suicidal ideation. Those who experienced more maltreatment reported poorer mental health.

Table 3.

Linear Regression of Psychological Distress and Logistic Regression of Suicidality.

	Psychological Distress		Suicide Ideation, Ever		Suicide Attempt, Ever
	β	SE	OR	(95% CI)	OR	(95% CI)
Bisexual	0.32***	0.08	1.60*	(1.08, 2.35)	1.40†	(0.94, 2.08)
Other	0.30**	0.10	2.74***	(1.61, 4.67)	1.68*	(1.01, 2.79)
Childhood GNC	0.01	0.03	1.18*	(1.00, 1.38)	1.03	(0.99, 1.09)
Childhood maltreatment	0.20***	0.03	1.53***	(1.27, 1.84)	1.37***	(1.25, 1.50)
Male	−0.05	0.07	1.21	(0.87, 1.68)	0.90	(0.64, 1.27)
Black	−0.13	0.10	0.52**	(0.32, 0.82)	0.88	(0.53, 1.47)
Hispanic	−0.13	0.10	0.50**	(0.30, 0.81)	0.71	(0.40, 1.23)
Multiethnic/multiracial	0.05	0.09	0.96	(0.60, 1.54)	1.06	(0.71, 1.60)
Older generation	−0.62***	0.07	0.50***	(0.35, 0.73)	0.76	(0.51, 1.14)
Middle generation	−0.41***	0.08	0.51***	(0.35, 0.73)	0.60**	(0.41, 0.88)
Upper income	−0.35***	0.09	0.73	(0.49, 1.07)	0.76	(0.48, 1.19)
Middle income	−0.23***	0.08	1.12	(0.76, 1.65)	0.85	(0.59, 1.22)

Note. Reference groups are lesbian/gay for sexual orientation, female for sex, White for ethnicity/race, younger cohort for generation, and lower income for income group. Not shown are tests of three-way interactions predicting each mental health outcome and involving sexual orientation (SO) by childhood gender nonconformity (GNC) by sex and SO by maltreatment by sex. These nonsignificant interactions were adjusted by all possible two-way interactions among SO, childhood GNC, maltreatment, and sex, as well as their main effects. The two-way interactions also were examined after adjusting for the main effects. Only 1 of the 15 two-way interactions, SO by GNC, was significant. For all interactions, adjustments were made for sex, ethnicity/race, generations, and income groups.

^†p < .10

^*p< .05

^**p < .01

^***p < .001.

GNC as mediator of the relations between sexual orientation and maltreatment.

Figure 2 shows that lesbian/gay individuals reported higher childhood GNC on average than bisexual and other individuals. Higher GNC was related to more maltreatment. The indirect effect (Table 4) indicated that higher GNC statistically explained why lesbian/gay individuals experienced more maltreatment than bisexual or other individuals (part of Model 1a). The direct effect found that bisexual individuals still experienced more maltreatment than lesbian/gay individuals after adjusting for GNC.³

Figure 2.

Model of Sexual Orientation, Childhood Gender Nonconformity, and Childhood Maltreatment. (Note. BI = bisexual, OT = other, and LG = lesbian/gay, the reference group; GNC = gender nonconformity. Standardized estimates (β) reported, as the data were standardized. The total effects of sexual orientation on maltreatment appear in parentheses and direct effects outside parentheses. All analyses were adjusted for covariates. *p < .05.)

Table 4.

Mediation Results

Outcome	ACME via childhood GNC	ACME via Maltreatment	Proportion Mediated (Pm) by childhood GNC	Proportion Mediated (Pm) by Maltreatment	ADE of Sexual Orientation
	β (95% CI)	β (95% CI)	Pm (95% CI)	Pm (95% CI)	β (95% CI)
Maltreatment
Bisexual v. LG	−.07 (−.11, −.03)***	---	−.30 (−.74, −.11)***	---	.30 (.18, .42)***
Other v. LG	−.03 (−.06, −.01)***	---	−.23 (−6.20, 5.83)	---	.07 (−.11, .22)
Psychological distress
Bisexual v. LG	.00 (−.01, .02)	.05 (.02, .08)***	.01 (−.04, .05)	.14 (.06, .22)***	.32 (.21, .44)***
Other v. LG	−.00 (−.02, .01)	.01 (−.02, .05)	−.01 (−.09, .04)	.05 (−.09, .18)	.30 (.15, .45)***
	B (95% CI)	B (95% CI)	Pm (95% CI)	Pm (95% CI)	B (95% CI)
Suicidal ideation, ever
Bisexual v. LG	−0.02 (−0.04, 0.00)*	0.01 (0.00, 0.03)***	−0.32 (−1.82, −0.04)*	0.15 (0.05, 0.37)***	0.08 (0.03, 0.14)***
Other v. LG	−0.01 (−0.02. 0.00)*	0.00 (−0.01, 0.01)	−0.05 (−0.15, −0.01)*	0.01 (−0.07, 0.09)	0.15 (0.09, 0.20)***
Suicide attempt, ever
Bisexual v. LG	−0.02 (−0.03, 0.00)†	0.02 (0.01, 0.04)**	−0.27 (0.09, 0.66)**	0.27 (0.09, 0.66)***	0.06 (0.01, 0.12)**
Other v. LG	−0.01 (−0.02, 0.00)†	0.00 (−0.02, 0.03)	−0.07 (−0.42, 0.02)†	0.03 (−0.33, 0.29)	0.10 (0.03, 0.19)*

Note. ACME = average causal mediation effect; GNC=gender nonconformity; ADE = average direct effect; CI=confidence interval; LG = lesbian/gay, the reference group. The proportion mediated (Pm) is based on 1,000 Monte Carlo simulations. Standardized estimates (β) for ACME and ADE are presented for maltreatment and psychological distress as outcomes, as all variables involved in these analyses were standardized prior to the mediational analyses. Unstandardized estimates (B) are provided for suicidal outcomes, as variables were not standardized prior to conducting those analyses. All analyses were adjusted for sex, ethnicity/race, generations, and income groups. Negative proportion of mediated effect indicates that the direct and indirect effects have opposite directional signs. The proportion is an absolute value.

^†p < .10.

^*p < 05.

^**p < .01.

^***p < .001.

GNC and maltreatment as mediators of the relations between sexual orientation and psychological distress.

GNC was not significantly related to distress in the adjusted model. Only maltreatment was related to distress (Figure 3, panel A), with more maltreatment linked to more distress. In keeping with Model 2, the indirect effect indicated that more maltreatment statistically explained why bisexual individuals reported more distress than lesbian/gay individuals (Table 4). Independent of GNC and maltreatment, bisexual and other individuals reported more distress than lesbian/gay individuals. In addition, the indirect paths of sexual orientation on distress via both mediators (rather than just through GNC or maltreatment) were nonsignificant for bisexual or other vs lesbian/gay individuals.

Figure 3.

Models of the Mediating Roles of Childhood Gender Nonconformity and Childhood Maltreatment in the Relations Between Sexual Orientation and Health. (Note. BI = bisexual, OT = other, and LG = lesbian/gay, the reference group; GNC = gender nonconformity. Estimates for panel A are standardized (β), as all variables were standardized. For suicidality, panels B and C, estimates are unstandardized (B), as variables were not standardized. The total effects of sexual orientation on distress and suicidality appear in parentheses and direct effects outside parentheses. All analyses were adjusted for covariates. †p < .10. *p < .05.)

GNC and maltreatment as mediators of the relations between sexual orientation and lifetime suicidality.

Higher GNC (OR=1.1, 95% CI=1.0, 1.3) and more maltreatment (OR=1.3, 95% CI=1.2, 1.4) were related to suicidal ideation (Figure 3, panel B), indicating that different pathways to ideation were found for lesbian/gay and bisexual individual. Per Model 1b, the indirect effects found that higher GNC statistically explained why lesbian/gay individuals ideated (Table 4). Per Model 2, more maltreatment statistically explained why bisexual individuals ideated. However, the indirect effects of sexual orientation on suicidal ideation via both GNC and maltreatment were not significant for bisexual or other vs. lesbian/gay individuals.

As illustrated in panel C of Figure 3, more maltreatment was associated with a higher likelihood of suicide attempts (OR=1.4, 95% CI=1.3, 1.5). In keeping with Model 1b, higher GNC explained why lesbian/gay individuals were likely to attempt suicide, although the finding was marginally significant. Per Model 2, the significant indirect effect showed that bisexual individuals were likely to attempt suicide, given more maltreatment (Table 4). Just as for suicidal ideation, the direct effects found that bisexual and other individuals were more likely to attempt suicide than lesbian/gay individuals. Furthermore, the indirect effects of sexual orientation on attempts via both GNC and maltreatment were nonsignificant for bisexual or other vs lesbian/gay individuals.

Sex differences.

Unadjusted analyses found that women reported significantly poorer mental health than men (Table 2), but adjusted analyses found no significant sex differences in mental health (Table 3). The moderating role of sex in the adjusted associations of sexual minority groups with the mediators and health outcomes was nonsignificant for the three-way interactions of sex by sexual minority groups by GNC or maltreatment with each mental health outcome. Only 1 of 15 two-way interactions of sex moderating the relations between sexual minority groups and GNC, maltreatment, or any outcome was significant.

Discussion

Consistent findings document sexual orientation disparities in health (e.g., Institute of Medicine, 2011; National Academies of Sciences, Engineering, and Medicine, 2020). This evidence has generated a robust body of research that examines correlates of health among sexual minority individuals, concentrating on sexual minority stress. Most studies focus on current or relatively recent stress, largely overlooking childhood, the period when many individuals are unaware of being a sexual minority. We examined the relations between childhood experiences of GNC and maltreatment to mental health years later, examining 3 models with data from the first US probability sample of sexual minority individuals. We proposed and found different pathways to poor mental health for lesbian/gay and bisexual individuals. This does not mean that bisexual individuals, on average, may not experience poorer health than lesbian/gay individuals. It does mean that one is able to understand why members of one group experience poorer health relative to members of the other group and why both groups report poorer health than heterosexual peers (e.g., Kerridge et al., 2017).

Our findings indicate that lesbian/gay and bisexual individuals have distinct early experiences that have important implications for subsequent health and adaptation. We encourage sensitivity to these possibilities and urge investigators to design studies that adequately capture distinctions between lesbian/gay and bisexual individuals and, thereby, as we did, extend the literature beyond the ubiquitous comparison of sexual minority to heterosexual peers.

Preliminary Findings

Unadjusted relations were consistent with the literature and our theoretical models. Lesbian/gay individuals reported higher childhood GNC than bisexual individuals, consistent with prior findings (Roberts et al., 2012). Childhood GNC was associated with more childhood maltreatment for lesbian/gay individuals, as hypothesized (Model 1a). In addition, bisexual individuals reported more maltreatment than lesbian/gay individuals, as proposed (Model 2). The maltreatment findings are consistent with those of pooled representative samples of the US population, in which lesbian/gay and bisexual individuals experienced more adverse childhood experiences (ACEs) than heterosexual peers, with the highest number of ACEs reported by bisexual individuals (Merrick et al., 2018). Disaggregation of ACEs found that bisexual individuals were more likely than lesbian/gay individuals to experience maltreatment (Merrick et al., 2018).

The Three Models

Tests of the theoretical models underscored the role of childhood GNC or maltreatment in sexual-minority adult psychological distress and lifetime suicidality, results that are consistent with studies of sexual minority and heterosexual populations (Austin et al., 2016). In keeping with Model 1b, we found support for the GNC model for lesbian/gay individuals, in which more childhood GNC statistically explained why lesbian/gay individuals ever experienced suicidal ideation. The same pattern of findings was marginally significant for lifetime suicide attempt. However, GNC was irrelevant for explaining recent (i.e., past 30 days) psychological distress of lesbian/gay individuals. The distress finding may be a consequence of the connectedness and support enjoyed by lesbian/gay individuals in the sexual minority community (Frost & Meyer, 2012). Our significant results particularly applied to bisexual individuals, whose reports of recent distress and lifetime suicidality were statistically explained by experiencing more maltreatment, as hypothesized by Model 2. Taken together these findings suggest different developmental sequelae for mental health vulnerabilities among sexual minority people, suggesting that sexual minority stress in childhood may operate differently for lesbian/gay and bisexual individuals.

Our study affirms the hypothesized connection between GNC and maltreatment for lesbian/gay individuals, which may explain why GNC has been linked to psychological distress and suicidality regardless of sexual orientation (Fitzpatrick et al., 2005; Li et al., 2016; van Beusekom et al., 2020). Results highlight the fundamental role of GNC in sexual minority stress for sexual minority people, in which violations of gender norms generate negative responses by others (Gordon & Meyer, 2007; Roberts et al., 2012; Thoma et al., 2021). Sexual minority individuals trace such experiences to childhood. Because lesbian/gay individuals are more GNC in childhood than bisexual individuals (Roberts et al., 2012), they are at risk for maltreatment and subsequent poor health.

Yet, elevated childhood maltreatment seems to be a driving mechanism for mental health vulnerability for bisexual individuals. For two related reasons, caution is needed before concluding that maltreatment is an external reaction to a perceived nascent bisexuality based on our analyses. First, maltreatment was not connected for bisexual individuals by means of childhood GNC. Second, other observable traits not examined here may link a nascent bisexual status to maltreatment. A potential candidate is temperament because it is biologically based (Rothbart et al., 2000) and apparent early in life (Gonda et al., 2009; Greenwood et al., 2012; Kawamura et al., 2010). It is also foundational for personality (McAdams, 2019). A study found that more common among bisexual than lesbian/gay or heterosexual individuals is the volatile temperament, which is characterized by being “disperse, unquiet, hasty, and disorganized” (Guerim et al., 2015). We suggest that such behaviors in children may frustrate parents, which may lead some parents to maltreat their children. Clearly, the “problem” is not the child, but rather the adult’s inability to cope appropriately with the child. This is unfortunate on its face, but also because such temperament may account for the greater openness to experience that has been observed among bisexual relative to lesbian/gay or heterosexual individuals, as a meta-analysis found (Allen & Robson, 2020). Openness to experience, a Big Five personality trait, is associated with creativity, intelligence, aesthetic preferences, and liberal political attitudes, as a narrative review detailed (Schwaba, 2019). These findings and our speculations regarding their potential ability to explain why maltreatment is connected to health among bisexual adults highlight the need for additional research and deeper theoretical understanding of the developmental processes and mechanisms that may distinguish the experiences of bisexual individuals from those of lesbian/gay individuals (Allen & Robson, 2020; Guerim et al., 2015; Swan & Habibi, 2018).

Sexual Minority and Sex Differences

Comparison of lesbian/gay, bisexual, and other individuals in a national probability sample is relatively unique. Many prior studies compare sexual minority to heterosexual individuals and do not test differences among sexual minority groups, with only a minority comparing lesbian/gay with bisexual individuals (e.g., Andersen & Blosnich, 2013; Balsam et al., 2005; Kerridge et al., 2017; Kuyper & Fokkema, 2011).

Patterns for other individuals (approximately 12% of the sample), when significant, were most similar to results for bisexual individuals, underscoring the distinctive pathways for mental health for monosexual and plurisexual individuals, at least when considering childhood GNC and maltreatment. Future research on sexual minority individuals who endorse nontraditional sexual identities (e.g., Goldberg et al., 2020) will illuminate ways that these groups differ (or not) as a subgroup of sexual minority people. So far, the evidence indicates some similarity between the new sexual identities and the traditional identities of lesbian, gay, and bisexual. The pansexual identity includes cisgender and non-cisgender individuals, and it is more common among younger than older adults (Greaves et al., 2019). Among cisgender persons, pansexual individuals are similar to bisexual individuals on sexual attractions and behaviors (Morandini et al., 2017; Timmins et al., 2021). Of cisgender individuals, those who identify as queer are somewhat more similar to lesbian/gay individuals on sexual attractions (Morandini et al., 2017; Timmins et al., 2021).

We found no evidence of sex differences in adjusted mental health. More important, sex did not moderate the adjusted associations of sexual minority groups with GNC, maltreatment, and mental health, indicating that the relations among these variables were similar for women and men. Consequently, the mental health implications of childhood processes addressed in this report were comparable for women and men.

Limitations

In addition to the limitations mentioned above, 4 others are noted. First, although childhood GNC has been linked to sexual orientation (e.g., Rieger et al., 2008), more theoretical and empirical work is needed on childhood maltreatment. Despite the robust literature on maltreatment, measurement is not specific to sexual minority status. More important, childhood maltreatment requires a link to sexual minority status to account for subsequent health disparities by nascent sexual orientation. We examined childhood GNC, but other mechanisms may exist, such as temperament. Such a link is particularly needed for bisexual individuals for whom childhood maltreatment was not explained. Second, individuals retrospectively reported their childhood GNC and maltreatment experiences. Such reports for childhood GNC have been significantly related to video evidence of childhood GNC (Rieger et al., 2008). However, a meta-analysis found low agreement between retrospective and prospective reports of childhood maltreatment (Baldwin et al., 2019). Further, the summary maltreatment measure used here does not distinguish types or severity of maltreatment, for which there may be distinct mental health consequences (McCabe et al., 2020). Despite the potential limitations of our measure of childhood maltreatment, the ability of such reports to predict subsequent poor health bolsters confidence in its validity (Danese & Widom, 2020; Newbury et al., 2018; Reuben et al., 2016). Third, the design is cross-sectional. Although all data were collected at the same time, the predictors were based on reports of childhood experiences. Until we have a study that begins tracking children and parents into the children’s adulthood, we will continue to have recollected reports of childhood experiences. Fourth, despite its advantages as a national probability sample, our sample is limited to individuals from 3 age cohorts and 3 ethnic/racial groups in the US. However, the sample is comparable to US representative samples in that, for example, more women identified as bisexual than lesbian (Copen et al., 2016; Kerridge et al., 2017).

Conclusions

Findings from our study suggest that sexual minority individuals are likely to experience sexual minority stress in childhood, a developmental period during which sexual orientation is nascent. Such experiences in childhood are related to subsequent poor health. In addition, the findings illuminate different pathways for lesbian/gay and bisexual individuals between nascent sexual orientation and their subsequent health, underscoring differences in their early experiences. The results point to the relevance of life course-informed understandings of sexual minority stress, which may contribute to a deeper understanding of the mechanisms explaining the health of sexual minority individuals and of health disparities between these individuals and their heterosexual peers. The implications may extend beyond health, given the cascading effects of early experiences across multiple domains of adaptation.