Introduction

Over the past decades, definitions of sexual orientation have evolved and expanded and the ways in which sexual minorities identify have changed dramatically. In recent years, there has been a steady increase in the number of women who identify as bisexual and who report same-sex sexual behavior or attraction (Chandra et al., 2011; Copen et al., 2016; Richters et al., 2014; Twenge et al., 2016). For example, between 2001 and 2013, the proportion of women identifying as bisexual almost doubled in both Australia (Australian Study of Health and Relationships: Richters et al., 2014) and the US (National Survey of Family Growth [NSFG]: Chandra et al., 2011; Copen et al., 2016). During the same period, there were substantial increases in the proportion of women reporting sexual experiences with both sexes in the two countries (Richters et al., 2014; Twenge et al., 2016), as well as an increase in the number of women reporting attraction to both sexes in Australia (Richters et al., 2014). Similar trends have been documented in other countries, such as the UK (Mercer et al., 2013) and the Netherlands (Kuyper & Vanwesenbeeck, 2009).

To obtain more nuanced assessments of sexual identity, some social and health surveys have begun offering a greater range of response options within questions on sexual orientation—for example, providing the intermediate categories of “mainly heterosexual” and “mainly lesbian/gay” in addition to the more common response options of “heterosexual,” “bisexual,” and “lesbian/gay.”Footnote 1 Recent studies using these more refined classifications have shown that women who identify as mainly heterosexual have higher exposure to violence (Hughes et al., 2010a, b), higher rates of alcohol and other drug use (Hughes et al., 2015), and poorer health and well-being (Vrangalova & Savin-Williams, 2014). These findings highlight the importance of understanding which women identify as mainly heterosexual in population surveys, with implications for the development of targeted prevention and intervention strategies.

Most repeated surveys of large, cross sections of populations (e.g., the General Social Survey, NSFG) only offer respondents the choice of the three traditional sexual identity labels. Thus, we know little about changes in the prevalence of the mainly heterosexual category among women over time. We also have limited information about sociodemographic differences between women who choose the “mainly heterosexual” identity label and those who choose the adjacent labels of “exclusively heterosexual” or “bisexual.” Drawing on the facilitative environments model (Katz-Wise & Hyde, 2017), we theorize that women’s propensity to identify as “mainly heterosexual” compared to “exclusively heterosexual” is on the rise and is correlated with sociodemographic traits that approximate women’s position within the social structure. We subsequently provide novel empirical evidence related to these assumptions by modeling unique survey data from three national cohorts of Australian women.

Who Identifies as Mainly Heterosexual?

Empirical evidence on the adoption of a mainly heterosexual identity among women remains patchy and heavily parochial to the U.S. When given the option, more U.S. women select a “mostly heterosexual” label than all the other minority sexual identities combined (Vrangalova & Savin-Williams, 2012). For example, in the National Longitudinal Study of Adolescent Health (Add Health), 16% of women identified as mostly heterosexual—four times the combined number identifying as bisexual, mostly lesbian, and exclusively lesbian (Savin-Williams et al., 2012). Evidence from longitudinal studies suggests that the proportion of women identifying as mostly heterosexual increases steadily across adolescence and young adulthood (e.g., Calzo et al., 2017; Savin-Williams et al., 2012), but little is known about the prevalence of this identity label at later life-course stages. Descriptive statistics from the National Study of Health and Life Experiences of Women showed that the proportions of women identifying as mostly heterosexual decreased with age, with 52.1% of mostly heterosexual women in the 21–34-year-old age range and only 4.2% in the age range of 65 + years (Hughes et al., 2015). However, these data were cross-sectional and therefore may reflect cohort rather than age differences. Further, the data are now 20 years old.

Information about the sociodemographic characteristics of women who select a mainly heterosexual identity label is also scarce and comes largely from simple descriptive analyses. For example, Mollborn and Everett (2015) compared the sociodemographic characteristics of mostly and exclusively heterosexual women in Add Health using bivariate techniques. Compared to exclusively heterosexual women, mostly heterosexual women were significantly less likely to be Black, have completed college, have grown up in a religiously conservative neighborhood, or have attended a rural school, and significantly more likely to be white, have highly educated parents, have been victimized, and have grown up in a neighborhood with a higher proportion of college-educated people. Accompanying evidence indicates that, on average, the relative strength of same-sex (compared to different-sex) sexual attraction and relative frequency of same-sex (compared to different-sex) sexual behavior reported by mostly heterosexual women are higher than those reported by exclusively heterosexual women, but lower than those of bisexual and lesbian women (Thompson & Morgan, 2008; Vrangalova & Savin-Williams, 2012).

The present study leverages unique Australian data to add further insights into the emergent body of scholarly knowledge about women who choose mainly heterosexual as a label to describe their sexual identity. For the sake of convenience, we refer to these women as “mainly heterosexual women” or “women who identify as mainly heterosexual.” First, we contribute by generating evidence on the mainly heterosexual category outside the U.S. While a handful of non-U.S. studies have provided preliminary, descriptive results (Calzo et al., 2018; Kuyper & Bos, 2016; Martin-Storey et al., 2019; Peter et al., 2017), ours is the first study to be conducted in Australia. Second, the present study is the first to investigate changes in both the prevalence and the sociodemographic correlates of mainly heterosexual identification across multiple cohorts of women. This provides an important and unique window into social change in regard to sexual minority identification. Third, we generate new knowledge on the relative importance of sexual attraction and behavior in distinguishing mainly heterosexual women from those selecting adjacent sexual identity labels.

Theoretical Framework and Expectations

Models of sexual orientation/identity development increasingly acknowledge the importance of contextual influences. The facilitative environments model (Katz-Wise & Hyde, 2017) is an empirically driven conceptual framework based on a qualitative study of sexual orientation development and change. According to this model, how individuals experience, understand, and label their sexual identity depend on influences that cut across three concentric levels—the individual, interpersonal, and societal levels. Environmental influences exist at both the interpersonal and societal levels and shape the availability and visibility of different sexual orientation categories/labels. Based on the facilitative environments model, we expected that some sociodemographic characteristics may be associated with social contexts that enable a full expression of women’s sexual diversity, whereas others may be associated with social contexts that constrain it.

First, we expected age cohort to be associated with women’s propensities to identify as “mainly” compared to “exclusively” heterosexual. Changes to the sociopolitical context over the past two decades mean that younger cohorts of women have come of age in environments that are arguably more facilitative of identifying as anything other than “exclusively heterosexual.” For example, in countries such as Australia, the U.S., and the UK, public attitudes are increasingly supportive of sexual-minority people and the rights of same-sex couples (Mercer et al., 2013; Perales & Campbell, 2018; Twenge et al., 2016). Over the same period, the legal rights of sexual minorities have been expanded, culminating in the passing of same-sex marriage laws in, for example, the US (2015) and Australia (2017). Thus, women from younger cohorts may feel more comfortable acknowledging some degree of same-sex sexuality when labeling their sexual orientation than women from older cohorts.

The likelihood of women identifying outside the exclusively heterosexual norm varies also by life-course stage. The contexts of young adulthood are considered particularly conducive to the exploration of same-sex sexuality and the development of a non-heterosexual identity (Rosario, 2019). In contrast, heteronormative pressures and/or a desire to align with normative femininity may increase as women plan for and transition into motherhood (Budnick, 2016; Diamond, 2008; Ross et al., 2017; Silva, 2019). The reverse pathway is also plausible, with motherhood and marriage being also a consequence of—not just a precursor to—exclusive heterosexuality. Regardless of which of these pathways apply, each leads us to expect that the likelihood of exclusively heterosexual identification compared to mainly heterosexual identification would be higher among women who are older, are married, and have children.

Earlier research findings indicate that cultural background is intertwined with sexual identity. For instance, U.S. research shows that women from some immigrant groups are less likely to adopt a non-heterosexual identity than US-born women (England et al., 2016; Silva & Evans, 2020). Hence, we expected that the likelihood of selecting a mainly heterosexual identity label compared to an exclusively heterosexual identity label would be lower among women born in a non-English-speaking country than among Australian-born women.

Concerning socioeconomic status, previous studies have reported that sexual minority women are, on average, socioeconomically worse off than heterosexual women, with bisexual women typically reporting the poorest economic status (Badgett, 2018; Charlton et al., 2018; Mollborn & Everett, 2015). Thus, we expected that the socioeconomic position of mainly heterosexual women in the sample would be better than that of lesbian and bisexual women, but worse than that of exclusively heterosexual women. Geographic context was also hypothesized to be an important factor. In Australia, community attitudes toward sexual minorities are significantly less positive in regional and remote areas than in major cities (Perales & Campbell, 2018). This may have implications for the way in which women in different geographic locations label their sexual orientation. It may also prompt sexual minority women to migrate from higher stigma to lower stigma areas—a phenomenon observed among gay and bisexual men in the US (Keene et al., 2017; Lewis, 2014). This led us to expect that women living outside Australia’s major cities would exhibit a greater propensity than women living in a major city to identify as exclusively heterosexual.

Further, we expected that women’s sexual attractions and sexual behavior would align to some extent with their responses to the sexual identity question. As such, we expected that mainly heterosexual women would be more likely to report being “attracted more often to males (but sometimes to females)” and having had “sexual experiences more often with males (but sometimes with females)” than exclusively heterosexual women. In contrast, we expected that mainly heterosexual women would be less likely to report being “attracted more often to males (but sometimes to females)” and having had “sexual experiences more often with males (but sometimes with females)” than bisexual women.

Finally, we considered the possibility that the associations between the sociodemographic factors discussed above and women’s “mainly heterosexual” identity differ in strength and direction across age cohorts. On the one hand, associations between women’s micro-level social environments and their propensity to identify as mainly heterosexual may be less pronounced in younger age cohorts. As previously discussed, women from older age cohorts grew up in societal contexts characterized by more pervasive stigmatization of non-heterosexuality. These women’s awareness of different sexual orientation categories and confidence to identify outside the heterosexual norm may be strongly tied to their exposure to certain facilitative environments (e.g., university). In contrast, women from younger age cohorts came of age in substantially more supportive societies. For them, information about and support of minority sexual orientation labels may have been much greater, rendering their exposure to micro-level facilitative environments less important for their adoption of a non-heterosexual identity (including a mainly heterosexual identity).

On the other hand, the progressive liberalization of sociopolitical contexts may have been experienced unevenly across different groups of women. For example, women belonging to certain religious or ethno-migrant groups may have comparatively lower access to knowledge about same-sex sexuality and experience lower support to adopt a non-heterosexual identity. From this perspective, certain sociodemographic factors may be more strongly associated with a propensity to self-label as mainly heterosexual in younger compared to older cohorts. While we did not make specific predictions about which of the two perspectives would prevail, our analyses shed empirical light on these matters.

Method

Participants

We use data from three national age cohorts of women taking part in the Australian Longitudinal Study on Women’s Health (ALSWH). The first two cohorts, born in 1946–1951 and 1973–1978, respectively, were recruited to the study in 1996. Stratified random samples of women within each cohort were selected from the Australian Medicare database and invited to participate (for details, see Brown et al., 1998; Dobson et al., 2015; Women’s Health Australia, 2017). Medicare Australia is the national health-insurance scheme covering all citizens and permanent residents. Women living outside major cities were oversampled at a ratio of 2:1. When calculating descriptive statistics for these cohorts, sample weights were used to adjust for this initial oversampling.

In Wave 1, there were 13,714 women in Cohort 1 (women born 1946–1951) and 14,247 women in Cohort 2 (women born 1973–1978). Comparisons between the study samples and national census data on a range of sociodemographic variables indicated that women in the study were broadly representative of Australian women of the same ages. Data have been collected from these two cohorts of women approximately every three years via self-completed questionnaires, first by mail and, since 2012, with an option to complete online. Response rates have ranged from 81 to 92% for Cohort 1 and from 57 to 69% for Cohort 2. Since the study commenced, Cohort 1 was asked about their sexual identity on one occasion (2001), when they were 50–55 years of age. Cohort 2 was asked on three occasions (2000, 2003, and 2012), when they were 22–27 years, 25–30 years, and 34–39 years of age, respectively. In addition, we use data from a third cohort of women born in 1989–1995 and recruited to the ALSWH in 2013 (Loxton et al., 2018).

Recruitment of Cohort 3 relied on multiple methods, including traditional and online (social) media outlets and peer referral. Women who were born during the target years and were eligible for Medicare were invited to enroll in the study.Footnote 2 Cohort 3 included 17,012 women at baseline. Since 2013, data have been collected from these women annually via an online survey, with response rates ranging from 55 to 70%. In this cohort, sexual identity information has been collected on four occasions (2013, 2014, 2015, and 2017), with women being 18–23 years of age in the first of these waves, and 22–28 years of age in the last.

Measures

Outcome Variable: Sexual Identity

Sexual identity was consistently measured with the following item: “Which of these most closely describes your sexual orientation?” The seven possible responses were: [1] “I am exclusively heterosexual,” [2] “I am mainly heterosexual,” [3] “I am bisexual,” [4] “I am mainly homosexual (lesbian),” [5] “I am exclusively homosexual (lesbian),” [6] “I don’t know,” and [7] “I don’t want to answer.” This is similar to questions used in US studies such as the Growing Up Today Study and Add Health. For our analyses, women who did not know or did not want to answer the question were coded as missing, and the remaining responses were coded as a set of dummy variables denoting each sexual identity category. An additional option (“Other”) was added to the sexual identity item in the most recent wave of data collection for Cohort 3 (year 2017). This option was selected by 1.6% of women (n = 132). There are several reasons why a woman might select the “Other” option (e.g., she might identify as asexual, queer or pansexual), but it is impossible to know for certain which applies. Therefore, to avoid misclassification and facilitate comparability across waves and cohorts, these cases were coded as missing and excluded from analysis. In addition, between 1 and 3% of women did not respond to the sexual identity item in each wave and these women were also coded as missing.

Explanatory Variables: Sociodemographic Characteristics

We examined the associations between women’s self-reported sexual identity and a range of theoretically informed sociodemographic variables. We included variables that were consistently measured across the three cohorts and available in the survey waves in which sexual identity was measured. Table 1 shows descriptive statistics for all variables for the full sample and for each cohort separately. Tables 5, 6, and 7 in the Appendix show descriptive statistics for each cohort by sexual identity group.

Table 1 Descriptive statistics
Full size table

Women’s cultural background was approximated by country of birth (Australia, another English-speaking country, a non-English-speaking country), whereas women’s life-course stage was captured by measures of age (in whole years), marital status (married, in a de facto relationship, divorced/separated/widowed, never married) and parental status (yes/no). We measured socioeconomic status using educational attainment (less than high school, high school, certificate/diploma, university degree), employment/student status (working only, studying only, both working and studying, neither working nor studying) and financial stress (“Over the last 12 months, how stressed have you felt about the following areas of your life? Money,” with possible responses ranging from [1] “not at all stressed” to [5] “extremely stressed”).Footnote 3 The role of geographic context was captured through an area of residence variable (major city, inner regional, outer regional, remote/very remote, living overseas). In models using multiple waves of data, we also controlled for survey year.

Explanatory Variables: Sexual Attraction and Sexual Behavior

In Wave 5 of Cohort 3 (2017), women were asked for the first time about their sexual attractions and behavior. Sexual attractions were measured with the following item: “Which of the six statements best describes you? I am sexually attracted…” with the following possible response options: [1] “Only to females”; [2] “More often to females”; [3] “Equally to both”; [4] “More often to males”; [5] “Only to males”; [6] “Never to anyone”; and [7] “Don’t want to answer.” Due to small cell sizes, we combined the “Only to females” and “More often to females” groups. Responses were then coded as a set of six dummy variables denoting sexual attraction.

Sexual behavior was measured with the following item: “Which statement best describes you? I have had sexual experiences…” with the response options [1] “Only with females”; [2] “More often with females”; [3] “Equally with both”; [4] “More often with males”; [5] “Only with males”; [6] “No experience”; and [7] “Don’t want to answer.” Again, the “Only with females” and “More often with females” groups were combined due to small cell sizes.Footnote 4 A set of six dummy variables denoting sexual behavior was then created from the responses.

Analytic Approach

To examine the sociodemographic factors associated with selecting each of the sexual identity labels, we used cohort-specific multinomial logistic regression models (Wooldridge, 2010). Because the ALSWH is a panel study, we have multiple rows of data (hereon referred to as “observations”) for each individual participating in the study. The models for Cohorts 2 and 3 include a random effect to account for this nesting of multiple observations within the same women (Rabe-Hesketh & Skrondal, 2008).Footnote 5 To examine associations between sexual attractions and behaviors and sexual identity, we fitted standard multinomial logistic regression models—as these models rely only on a single wave of data. Across all models, “mainly heterosexual” was used as the baseline category of the outcome variable from which we compared results from the other sexual identity categories.

For ease of interpretation, the estimated model coefficients were expressed as relative risk ratios (RRRs). RRRs give the ratio of the probability of falling into a given category of the outcome variable (e.g., bisexual) over the probability of falling into the baseline category of the outcome variable (i.e., mainly heterosexual) associated with a one-unit increase in an explanatory variable. RRRs greater than one indicate that a given explanatory variable is associated with a higher likelihood that the outcome is realized, compared to the baseline outcome. Conversely, RRRs smaller than one indicate that a given explanatory variable is associated with a lower likelihood that the outcome is realized, compared to the baseline outcome. All analyses were conducted using Stata 16 software.

Results

Sociodemographic Factors

In a first set of analyses, we explored changes in the prevalence of the mainly heterosexual identity label across the different cohorts, as well as associations between this sexual identity label and sociodemographic variables for each cohort. These descriptive analyses pooled all available information from the women across waves of the study and were thus based on a total of 76,930 observations from 36,665 women with non-missing data on the sexual identity and sociodemographic variables. Specifically, the analyses included 9149 observations from 9149 women in Cohort 1, 24,722 observations from 11,014 women in Cohort 2, and 43,059 observations from 16,502 women in Cohort 3.

Table 1 and Fig. 1 show quite dramatic differences in the prevalence of the different sexual identity groups across the three cohorts. The proportion of women identifying as exclusively heterosexual differed markedly across the three cohorts, from 97.5% of observations in Cohort 1, to 91.5% in Cohort 2, and to 63.3% in Cohort 3. Meanwhile, the prevalence of the mainly heterosexual identity label increased across cohorts (from 1.2% of observations in Cohort 1, to 6.3% in Cohort 2, and to 26% in Cohort 3). Similarly, the prevalence of bisexual identity increased from just 0.1% of observations in Cohort 1, to 0.9% in Cohort 2, and to 8.6% in Cohort 3. Mainly lesbian identity was rarely selected in any of the cohorts. Yet even this identity showed an upward cross-cohort trend, increasing from 0.2% of all observations in Cohort 1, to 0.4% in Cohort 2, and to 1.1% in Cohort 3. Exclusively lesbian identity was the only category to maintain a relatively similar prevalence across the three cohorts, comprising 1.1% of all observations in Cohort 1, and 0.9% in each of the other two cohorts. Together, these descriptive results suggest that exclusively heterosexual identification is in decline among successive cohorts of Australian women. Critically, such decline appears to be largely driven by a concomitant increase in mainly heterosexual identification. Figure 1 depicts these trends and the prevalence of various sexual identities by cohort and year. Overall, this shows that cross-cohort changes in the prevalence of mainly heterosexual identification far exceed within-cohort changes.

Fig. 1
figure 1

Prevalence of sexual identity categories across cohorts and over time. Notes: ALSWH data; Cohort 1 (women born 1946–1951): year 2001; Cohort 2 (women born 1973–1978): years 2000, 2003, 2012; Cohort 3 (women born 1989–1995): years 2013, 2014, 2015, 2017

Full size image

Table 2 shows results from the multinomial logistic regression models, with “mainly heterosexual” as the baseline outcome. While all five sexual identity groups were included as outcomes in the models, for simplicity we only display coefficients for the groups directly adjacent to the “mainly heterosexual” category (i.e., “exclusively heterosexual” and “bisexual”). Coefficients for the “mainly lesbian” and “exclusively lesbian” groups are given in the Appendix Table 8.

Table 2 Relative risk ratios from multinomial logistic regression models of sexual identity
Full size table

A number of sociodemographic factors showed consistent associations with women’s propensity to identify as exclusively heterosexual compared to mainly heterosexual across the three cohorts. The likelihood of identifying as exclusively heterosexual compared to mainly heterosexual was higher among women who were studying rather than working (RRRs = 0.75–0.88) and who reported greater levels of financial stress (RRRs = 0.83–0.92), but lower among women living in a regional area compared to a major city (RRRs = 1.17–1.35). While the pattern of results was consistent across the three cohorts, the parameters for Cohort 1 were not always statistically significant at conventional levels (p < 0.10) due to small cell sizes in that cohort. There was also a trend toward a higher likelihood of identifying as exclusively heterosexual compared to mainly heterosexual among women with children across cohorts (RRRs = 1.06–1.27), but this was only statistically significant for Cohort 3.

There were few sociodemographic variables consistently associated with the propensity to identify as bisexual compared to mainly heterosexual across the cohorts. Across the three cohorts, there was a higher likelihood of bisexual identity compared to mainly heterosexual identity among women who were neither working nor studying (RRRs = 1.33–2.17) and immigrants from English-speaking countries (RRRs = 1.12–2.84), and a lower likelihood of bisexual compared to mainly heterosexual identity among women who moved overseas compared to those living in a major Australian city (RRRs = 0.64–0.65). However, these three trends only reached statistical significance in Cohort 3.

Perhaps more interesting, for some of the sociodemographic variables examined associations with sexual identity diverged across the cohorts. For example, compared to women who had not completed high school (Year 12), those with a university degree had a lower likelihood of identifying as exclusively heterosexual compared to mainly heterosexual in Cohort 1 (RRR = 0.37, p < .01), but a higher likelihood in Cohort 3 (RRR = 1.21, p < .01). Compared to Australian-born women, women born in a non-English-speaking country had a lower likelihood of identifying as bisexual compared to mainly heterosexual in Cohort 1 (RRR =  < 0.01, p < .01) and Cohort 3 (RRR = 0.72, p < .10), yet there was a non-significant trend in the opposite direction in Cohort 2.

Intersections between Identity, Attraction, and Behavior

Results of a second set of analyses provide unique insights into the associations between women’s sexual attraction and behavior and mainly heterosexual identification. These analyses were based on a subsample of our full sample⁠—women from Wave 5 (2017) of Cohort 3 (n = 7,741). This was the only occasion⁠ in the ALSWH in which women were asked about their sexual attractions and sexual behavior (in addition to their sexual identity). At the time these data were collected, the women were 22–28 years of age.

Table 3 shows sexual attraction and behavior frequencies by sexual identity for the subset of ALSWH respondents who were asked about all three sexual orientation dimensions (Cohort 3, Wave 5). Clear trends in patterns of sexual attraction emerged across the three identity groups. The vast majority of exclusively heterosexual women reported being attracted only (89%) or mostly (10%) to males. Among mainly heterosexual women, these proportions were reversed: Just under 10% reported being attracted only to males, while 86% reported being attracted mostly to males. An additional 3% said they were attracted equally to both sexes. The greatest within-group variability in attractions was observed for bisexual women, with 46% being attracted more often to males, 40% being attracted equally to both sexes, and 14% being attracted more often to females.

Table 3 Descriptive statistics: sexual attraction and behavior
Full size table

In regard to sexual behavior, all three groups reported having had more sexual experiences with men than women. The majority of exclusively heterosexual women (82%) reported having had sexual experiences with men only, compared to 46% of mainly heterosexual women and 17% of bisexual women. Among exclusively heterosexual women, 13% reported some same-sex sexual experience. Same-sex sexual experience was more common among mainly heterosexual women, with 50% reporting that their sexual experiences were mostly with men (but sometimes with women). An additional 1% reported sexual experiences equally with both sexes. Bisexual women were the most likely of the three groups to report same-sex experience: 69% reported having had sexual experiences mostly with men (but sometimes with women), 9% equally with both, and 4% mostly or only with other women. Exclusively heterosexual women were the most likely to report having had no sexual experience (5%), followed by mainly heterosexual (3%) and bisexual (1%) women.

Table 4 shows results of three multinomial logistic regression models, with sexual identity as the outcome (baseline outcome category = “mainly heterosexual”) and sexual attraction and behavior as the explanatory variables (first separately and then simultaneously). While all five sexual identity groups were included as outcomes in the models, for simplicity we only display RRRs for the groups directly adjacent to the “mainly heterosexual” category (i.e., “exclusively heterosexual” and “bisexual”).

Table 4 Relative risk ratios from multinomial logistic models of sexual identity regressed on sexual attraction and behavior
Full size table

The results indicated that sexual attraction and sexual behavior were significantly and uniquely associated with sexual identity in women, in a manner consistent with expectations. When attraction and behavior were considered simultaneously (Model 3), women were significantly more likely to identify as exclusively heterosexual than mainly heterosexual if they were attracted to males only, compared to being attracted mostly to males (RRR = 68.37, p < .01). In contrast, the likelihood that women identify as bisexual compared to mainly heterosexual was significantly higher when women were attracted only/mostly to females (RRR = 31.31, p < .01) or equally to both sexes (RRR = 25.05, p < .01), compared to being attracted mostly to males.

Controlling for sexual attraction, sexual behavior was also significantly associated with sexual identity, although the model coefficients were not as large as those for attraction. For example, women were significantly more likely to identify as exclusively heterosexual compared to mainly heterosexual if they reported sexual experiences with males only (RRR = 3.32, p < .01) or no sexual experience (RRR = 3.32, p < .01) rather than sexual experiences mostly with males. Conversely, women who reported having had sexual experiences equally with both sexes were significantly more likely than those who reported having had more experience with males to identify as bisexual, compared to mainly heterosexual (RRR = 2.02, p < .05).

A comparison of the pseudo-R2 statistics across the three models provided evidence that sexual attraction was more strongly associated with sexual identity than was sexual behavior. When sexual attraction was the only of the two constructs included in the model (Model 1), approximately 54% of the variation in sexual identity was accounted for, compared to a much lower 22% when sexual behavior was the only of the two constructs in the model (Model 2). When sexual behavior was added to the model with sexual attraction (Model 3), the amount of variation explained increased only slightly, to 57%.

Discussion

Our results add to a small but growing body of literature about women who identify as “mainly heterosexual.” The need for greater understanding of this group of women is motivated by the fact that multiple studies have found that they are particularly vulnerable to poor health, victimization, and substance use (Hughes et al., 2010a, b, 2015; Vrangalova & Savin-Williams, 2014).

First, we confirm with Australian data what have been found in studies conducted in the U.S. (Calzo et al., 2017; Savin-Williams et al., 2012); the percentage of women who identify as “mainly heterosexual” has increased among younger age cohorts, seemingly at the expense of fewer women identifying as “exclusively heterosexual.” In fact, the increases observed in our Australian data appear to be more pronounced than those reported previously in U.S. studies. In Australia, the percentage of women who identified as mainly heterosexual increased from ∼1% in the oldest cohort (born 1946–1951) to ∼26% in the most recent one (born 1989–1995). As a comparison point, ~ 11% of young women in the U.S. identified as mostly heterosexual in 2001–2002 (Wave III Add Health: Savin-Williams et al., 2012) compared to ~ 15% in 2005–2007 (Growing Up Today Study: Calzo et al., 2017). Due to differences in the data and samples across studies, these comparisons should nevertheless be interpreted with caution.

The large number of young women who choose this label in the youngest cohort is striking and warrants greater understanding of who these women are and what their experiences are like. Consistent with the facilitative environments model (Katz-Wise & Hyde, 2017), we posit that one important factor is the increasing acceptance of sexual minority people in Australia. Although same-sex marriage was not legal in Australia until late 2017, support for sexual minorities had increased exponentially in previous years (Perales & Campbell, 2018). These changes may have made women more comfortable describing themselves as something other than “exclusively heterosexual.”

Second, this study was the first to use multiple age cohorts to examine the demographic characteristics of women who select a mainly heterosexual identity, relative to those who select other identity categories. Based on the facilitative environments framework, we hypothesized that some women may occupy social locations that enable or constrain full expression of their sexual orientation. Consistent with expectations, we found that exclusively heterosexual women were more likely to be married, less likely to be subjected to financial stress, and more likely to live in a non-metropolitan area than mainly heterosexual women. However, as we have noted elsewhere in the paper, it is plausible that these associations run in either or both directions (i.e., social locations may precede sexual locations, or vice versa). For example, exclusively heterosexual identification may be a precursor to higher levels of heterosexual marriage. Alternatively, heterosexual marriage may encourage women to identify as exclusively heterosexual regardless of their past or current sexual attractions, due to cultural prescriptions of monogamy and as a strategy to reduce cognitive dissonance between behavior and identity.

In contrast to our predictions, we did not observe a uniform positive association between women’s non-English-speaking background and their likelihood of identifying as exclusively heterosexual compared to mainly heterosexual. Contrary to our expectations, women from the oldest cohort who were born in a non-English-speaking country were significantly less likely than Australian-born women to identify as exclusively compared to mainly heterosexual. This pattern of results may reflect the drastic changes in the countries of origin of Australian migrants that have taken place over recent decades. Among recent migrant cohorts, there is an over-representation of women from Asian countries in which public support for same-sex relations is low (e.g., China or India). In contrast, the older migrant cohorts include comparatively larger shares of women from European countries where support for same-sex relations is higher (e.g., Germany or the Netherlands). Further, it is possible that women from older migrant cohorts—who have spent a longer amount of time in Australia—have embraced Australian values concerning sexual diversity to a larger degree than women from more recent migrant cohorts. The curvilinear findings concerning bisexuality are more difficult to explain, with women from non-English-speaking backgrounds in the older and younger cohorts more likely to identify as “mainly heterosexual” than “bisexual,” but those in the middle cohort being as likely to identify as “mainly heterosexual.”

The unique properties of the ALSWH data also allowed us to identify important cross-cohort differences in the sociodemographic correlates of mainly heterosexual identification. One such difference concerned the role of age. Specifically, we found no significant association between age and mainly heterosexual identification among the older cohorts. Within the youngest cohort, however, an additional year of age was associated with a lower likelihood of identifying as exclusively rather than mainly heterosexual, and with a greater likelihood of identifying as bisexual than mainly heterosexual. These findings indicate that, among women in more recent cohorts, the odds of reporting a minority sexual identity increase with age. Although it is not entirely clear what factors are driving this age gradient, the facilitative environments framework points to some likely candidates. Judging from the ages at which women in the youngest cohort were surveyed (18–28 years), these may include leaving the parental home for school or work and other life experiences that challenge assumptions of exclusive heterosexuality (Rich, 1980; Tolman, 2006).

We also observed cohort differences in the association between women’s education and identifying as mainly heterosexual. Among the older cohorts, a university degree was associated with being less likely to identify as exclusively than mainly heterosexual. However, in the younger cohorts, having university-level educational qualifications was associated with being more likely to identify as exclusively than mainly heterosexual. For older cohorts, a college environment may have been the one “liberal” space women had access to that allowed them to question or explore their sexuality. Younger cohorts of women may find support for sexual minority identities more broadly in their own communities and therefore feel safer to identify with a non-exclusively heterosexual identity. These processes, however, do not fully explain why women with degree-level qualifications in the youngest cohort were significantly more likely (rather than equally likely) to identify as exclusively heterosexual compared to mainly heterosexual. This remains an open question. The pattern of results may nevertheless be affected by some women in the youngest cohort having started, but not yet completed, their university studies (and hence not selecting the “university degree” option in the survey). It is also possible that it takes some time for women to adopt more fluid views concerning their own sexuality after completing a degree.

A further contribution of this study was to consider how sexual attractions and behaviors were associated with women’s propensity to identify as mainly heterosexual. Consistent with previous research (Vrangalova & Savin-Williams, 2012), our findings indicated that the one-dimensional sexual attraction and behavior continuums in the ALSWH surveys align with the sexual identity continuum. This was especially the case for sexual attraction, which explained the majority of the variance in sexual identity in multivariable models. The fact that women’s sexual orientation labels were more strongly associated with their attractions than their behavior is not surprising. First, behavior is subject to constraints of opportunity in a way that attraction is not: women are free to feel attracted to whoever they like, but to act upon their attractions they must first find a suitable partner. Second, same-sex behavior arguably constitutes a more stigmatizing departure from heteronormativity than private same-sex attractions. Particularly, some both-sex attracted women may be reluctant to engage in same-sex behavior due to fear of social repercussions. Thus, they may describe their sexual orientation as “mainly heterosexual” or “bisexual,” yet they may only ever engage sexually with men. In addition, sexual behavior can be motivated by a variety of reasons aside from desire. For example, research on the phenomenon of “straight girls kissing” at college parties has found that women report a range of motivations for these public displays of behavior, including a desire for male attention, experimentation, fun, and social pressure (Rupp & Taylor, 2010; Yost & McCarthy, 2012).

Despite the strong associations with sexual attractions, our results suggest that other (unobserved) factors may be important in women’s propensity to identify as mainly heterosexual. Our model that included the full set of sociodemographic characteristics, sexual attraction, and sexual behavior as explanatory variables accounted for 57% of the variance in women’s sexual identity. Hence, 43% of the variance in women’s sexual identity remains unexplained by these factors, suggesting that other characteristics may be important in influencing the choice of a sexual identity label. While we were unable to examine these with the available data, the facilitative environments framework (Katz-Wise & Hyde, 2017) coupled with recent empirical studies indicates that psychosocial factors such as family and gender-roles attitudes, political conservatism, socialization experiences during childhood, religiosity, ethnic group, and internalized homophobia or biphobia can also shape how women label their sexual orientation (Mollborn & Everett, 2015; Silva, 2018, 2019). Further, while we examined the additive effects of sexual attractions, sexual behavior, and sociodemographic factors on women’s sexual orientation labels, future research could also consider interactive effects. It is possible, for example, that women attracted to both women and men are inclined to label their sexual orientation in different ways depending on their other social locations (e.g., their ethnicity, religion or disability status)—a proposition that is consistent with intersectionality theory (Bowleg, 2012; Silva & Evans, 2020).

Strengths and Limitations

This study featured a number of strengths: It leveraged robust data from three national cohorts of Australian women to provide unique evidence on historical changes in the prevalence and correlates of “mainly heterosexual” identification among women; it generated findings from an additional country, thereby expanding the empirical body of knowledge beyond the US; and it generated novel insights into the interconnections between women’s sexual identity, attractions, and behavior. Notwithstanding these strengths, several data-driven study limitations must be acknowledged.

First, although the ALSWH is unique in its ability to answer our research questions, there are significant longitudinal inconsistencies within and across its different cohorts. For instance, the three cohorts differed substantially in their overall sample sizes, the size of the sexual minority subsamples, the availability of information on sexual orientation, and—importantly—the ages at which women were asked about their sexuality. The oldest cohort, specifically, was only asked about their sexual orientation in a single wave of data collection, in which a small minority of women selected the mainly heterosexual (n = 101, 1.2%) or bisexual (n = 13, 0.1%) categories. Further, women in this cohort were much older than women in the other cohorts. Collectively, these inconsistencies mean that differences between cohorts should be interpreted with caution. Nevertheless, our results underscore that cross-cohort changes in the prevalence of the mainly heterosexual category far exceed within-cohort changes over time—suggesting that genuine cohort differences, rather than differences in the ages at which different cohorts were surveyed, are responsible for the observed trends.

Second, our analyses identifying associations between the three major dimensions of sexual orientation (identity, attractions, and behavior) were based on a single wave of data from the youngest cohort, as the requisite information was only available in that subsample. Finally, information on gender identity was completely absent from the ALSWH data. Thus, we were unable to disentangle the experiences of cisgender women from those of gender minority women, including those who identify as transgender, non-binary, or agender.

Third, given the nature of our data, the findings reported here were restricted to women, and the extent to which they can be generalized to men remains to be established. On the one hand, it is possible that the observed increase in “mainly heterosexual” identification and the correlates of selecting this sexual identity category are similar among men. Indeed, non-sexual social factors have been associated with the sexual identities adopted by men (see e.g., Silva, 2018; Silva & Whaley, 2018). Nevertheless, it is unlikely that the magnitude of social change pertaining to mainly heterosexual identification has been as strong among men as women. Not only are women’s sexualities potentially more “plastic” (Baumeister, 2000), “fluid” (Diamond, 2008; Diamond et al., 2017), and “non-specific” (Chivers, 2010) than men’s, but also women have “less to lose” concerning their social status when they deviate from exclusive heterosexuality (England et al., 2016). Indeed, while a progressively larger share of women identify as “bisexual” in the U.S., an analogous trend among men has not taken place (England et al., 2016). It follows that future empirical studies should focus on exploring the issues covered in the present study among cohorts of men.

Conclusions

This study has generated novel and robust evidence contributing to a body of scholarly knowledge documenting the fluidity of sexual orientation identities over time. Specifically, our findings from a relatively progressive country—Australia—demonstrate that the intermediate sexual identity response option “mainly heterosexual” is gaining significant traction among women. Consistent with the facilitative environments framework, the adoption of this identity label over its neighboring labels (i.e., “exclusively heterosexual” and “bisexual”) is structured along important sociodemographic axes—including education, socioeconomic status, geographic context, life-course stage, and other domains of sexual orientation. Yet these associations are also in flux, with evidence that they have changed across successive cohorts of Australian women. Understanding the factors that contribute to individuals’ labeling of their own sexual orientation and the consequences of those labels on their health and life experiences requires paying close attention to these and other processes of social change.