groves 9453087.0005.001 in
« Prev section Next section »

Chapter 1: The War on Gender Equality is Working: The Bem Sex Role Inventory Across 25 Years

Judith L. Fischer
Texas Tech University
Cheryl A. Juergens
Gallatin City-County Health Department (MT)

Acknowledgements: This project was supported in part by Grant No. R01 HD 18864 “Network Supports and Coping in Adult Transitions” from NICHD to Judith Fischer & Donna Sollie. The authors declare that they have no conflicts of interest. We thank many people past and present who worked on this project.

The enculturated-lens theory of individual gender formation...situate[s] the individual from birth to death in a social and historical context containing the lenses of androcentrism and gender polarization (Bem, 1993, p. 138).

Decades ago the women’s movement provided encouragement that institutional barriers to gender equality were falling (Bem, 1993). But Bem’s 1993 writings also described the entrenched U.S. social world as one that was structured by white heterosexual men to maintain and defend male privilege. In such a patriarchy, men could not cross gender boundaries without heavy stigmatization; thus, for secure gender identity, “real” men had to suppress tendencies toward femininity within themselves. Although the majority of US men were not themselves powerful in society, Bem argued that they were able to avoid feeling emasculated by seeing themselves as more powerful than women and gay men. In support of this theorizing, Cheryan, Cameron Schwartz, Katagiri, and Monin (2015) reported that, compared to controls, men who felt threatened responded by increasing their self-reported masculinity and decreasing their femininity. As well, an androcentric societal structure meant that women worked rigorously to live up to the ideal of the “real” woman as someone who was sexually attractive to men (Bem, 1993). Men who threaten women in the context of relationships also regulate and contest women’s autonomy (DeShong, 2015). Furthermore, unchanging workplace constraints that reflect the policies of gendered institutions overrule individual desires (Pedulla & Thébaud, 2015).

When under perceived threat, it is unsurprising that powerful societal institutions and psychological processes push back (Wood & Eagly, 2002). The gender equality gains of the 1960s and 1970s challenged the security of the gender identities of men and women (Bem, 1993), and were met with backlash (Bem, 1993; Faludi, 1991). By 2010, the gender revolution was described by a scholar as “uneven and stalled” (England, 2010, p. 149) in that women were not moving up in their jobs and careers and women’s traditional jobs continued to be devalued and unattractive to men. To be a feminist became equated with being anti-men (Robnett, Anderson, & Hunter, 2012). Fueled in part by patriarchal beliefs, the New Right (Ruth, 1983), composed of people and organizations from the Radical Religious Right, Moral Majority, and New Conservatism, became a counter movement to gender equality. Concomitant with this patriarchal thrust were intensified restrictions on women’s reproductive choices (Finlay, 2006). The backlash of the late 1980s and early 1990s was reflected in the political and cultural successes of the New Right. Fitting into this context, the events of September 11, 2001 (9/11) provided a pretext for a pervasive, intense mainstream backlash (Anderson, 2015; Faludi, 2007; Lorber, 2005). As Faludi described in detail in her 2007 book The Terror Dream, the U.S. responded to a threat to the homeland by transforming it into a threat to the domestic home, an excuse to erase the gains of gender equality and restore real men to their purported“rightful” positions. Faludi described this 9/11 response as rooted in androcentric myths and delusions that conformed to a long tradition of such U.S. responses to real threats. Gilligan (2011) identified tensions surrounding femininity: in a democracy, to be caring is to be human, but in a patriarchy to be caring is dismissed as feminine.

This study was conducted to identify historical shifts in gender role identity. It is bookended by earlier times that saw increasing promotion of gender equality and later times that saw increasing pressures to reverse these gains through the revival of long-standing cultural identifications with patriarchy, androcentrism, and gender polarization. At its heart, if the war on gender equality is working it should reveal itself at a personal level. That is, self-identifications of masculinity and femininity would differ in ways reflecting greater polarization in recent times compared to earlier times. There are a variety of ways to index gender equality, such as work force participation, wages, and shared home tasks. We chose to consider gendered identifications because of the important personal and relational consequences to individuals in a world ordered by gender (Starr & Zurbriggen, 2016) and by binary conceptualizations of gender.

There are several explanations that could explain the changes/differences in masculinity and femininity in different time periods. First, is an historical times explanation. In this study, the term historical times refers to the year or time of measurement; it is shorthand for the cultural and societal themes evident at that time. According to Bem (1993), people in different historical times will differ based on the historical events and accompanying cultural expectations and norms of the time. She asserted that “the assignment of women and men to different and unequal positions in the social structure” accounted for conventional gendering of women and men and not childhood socialization (Bem, 1993, p. 135). Between 1982 and 2007 there were many important social changes such as increased: national average age of first marriage (U.S. Census Bureau, 2015), parents’age of first birth (Kirmeyer & Hamilton, 2011), participation of women in the workforce (Wood & Eagly, 2002), and time span between high school and the assumption of adult responsibilities (Arnett, 2000). As well, the events of 9/11 were seared in public consciousness (Faludi, 2007). At the time of measurement in the spring of 2007, these events were less than six years in the past and, as a result, the country was still at war. Thus, time of measurement is the focus of this research.

A second explanation for gender role changes is to consider cohort changes/differences. Birth cohorts are people defined by their year of birth or a span of years such as decade of birth. Birth cohorts are sometimes identified with names associated with particular generations even though exact years may vary from study to study (Parry & Urwin, 2011; Sessa, Kabacoff, Deal, & Brown, 2007). Grouping people into generations reflects the belief that each cohort“share[s] a different set of values and attitudes, as a result of shared events and experiences” (Parry & Urwin, 2011, p. 80), that they share cohort-defining events (that could be modified by parental practices, Stewart & Healy, 1989) or formative experiences when coming of age (Twenge, 1997). When college students are studied, time of coming of age and time of measurement can refer to the same thing. Thus, explanations based on formative experiences (Twenge, 1997) cannot be disentangled from historical time of measurement. However, when adults are studied longitudinally there is a period of separation between formative experiences and later ones.

Comparing two cohorts at one point in time and finding differences does not necessarily support a cohort explanation because members of these cohorts also differ in age. So, in addition to historical times and cohort/formative experiences as explanations of change/difference, there is a thirdexplanation, that refers to age or maturation effects (Parry & Urwin, 2011). Younger selves differ from older selves based on their physical and psychosocial stages of development and accumulations of life experiences. A single design cannot disentangle explanations based on age, historical times, or cohort/formative years.

Study Design Issues

The disentanglement of time of measurement, cohort, and age requires a multifaceted design. Schaie (1965) formulated procedures to study changes and differences and his explication of most efficient designs informs this study. As illustrated in Table 1, we chose three of Schaie’s designs: longitudinal, cross-sectional, and time lag. There are two historical time periods of measurement, 1982 and 2007, and two cohorts, late baby boomers (BB) and Generation Y (GY). Two ages are represented: college senior emerging adults (CS) and midlife adults (ML). The longitudinal component (row) is of a late baby boomer 1958-1962 birth cohort studied in 1982 and 2007 when they were approximately 22 and 47 years old (CS and ML). The cross-sectional (same historical time of measurement) component (column) includes the BB cohort and the GY cohort studied in 2007 when baby boomers were approximately 47 (ML) and the Generation Y group members were approximately 22 years old (birth year 1983 – 1987). The time lag component (diagonal) includes the BB cohort in 1982 compared with the GY cohort in 2007 with both groups approximately 22 years old (CS) at the differing times of measurement.

In the longitudinal design, the time of measurement differs but so do the ages of the participants. Thus, any within-cohort changes could be because of time or age. In the cross-sectional design, two or more groups of people from different birth cohorts participate at the same time, with time of measurement a constant but with varying birth cohorts and ages. Differences between cohorts could be because of cohort membership or age. Finally, when two groups of same-age people, but from different cohorts and different times of measurement (time lag) are assessed, the between-cohort differences could be due to cohort membership or time of measurement. The comparison of patterns of results from these three designs (longitudinal, cross-sectional, time lag) allows inferences about likely loci of changes/differences (Schaie, 1965). The pattern that would reflect the effects of time of measurement would include differences between BBCS and BBML and between BBCS and GYCS in predicted directions. This brief introduction to research design and to historical developments over the past several decades sets the stage for this study.

Table 1. Design of Study: Longitudinal, Cross Sectional, and Time Lag
Table 1. Design of Study: Longitudinal, Cross Sectional, and Time Lag

Development of Gender Role Identity in Social Contexts

Gender involves a number of dimensions such as the activities people like to do, their attitudes, and their personal-social attributes (DiDonato & Berenbaum, 2011). Bem’s (1981a) gender schema approach (as expressed in the opening quote from 1993) is very different from earlier conceptualizations of masculinity and femininity as a bipolar construct representing “relatively enduring traits which are more or less rooted in anatomy, physiology, and early experience” (Constantinople, 1973, p. 390). In the late 1960s, Bakan (1966) described two dimensions he termed agency (associated with men) and communion (associated with women). Furthermore, masculinity and femininity could be measured separately (Constantinople, 1973). The BSRI established that women and men could endorse positive characteristics of femininity and masculinity using items that were also freed from the requirement that they distinguish females from males (Bem, 1974). The idea of androgyny, that one person could endorse characteristics approved of for both men and women, provided a conceptual and measurement basis for revised thinking about what was appropriate for all individuals and what constituted a flexible gender role identity (Bem, 1977). Compared to 1982, by the year 2007 there were a number of forces pushing for greater gender polarization (Faludi, 2007). Greater polarization is recognized in a number of ways: men and women changing from their younger selves; men and women of the same age varying when measured at different times. The polarization we are interested in involves (a) men changing/different on femininity through declines/lower scores over time; (b) women changing/different on masculinity through declines/lower scores over time; and (c) declines/lower scores on androgynous expressions of self-identity.

Changes and Differences on BSRI Masculinity and Femininity Scale Scores

This section considers studies that used time lag, cross-sectional, and longitudinal designs from the 1980s to the mid-2000s. These studies included the BSRI masculinity and femininity scales with continuous scores (Bem, 1974). The purpose of this selective review is to weigh evidence for or against the proposition that self-endorsements of gender roles reflect the time of measurement and to gauge support for alternative explanations such as formative experiences or age-related changes. Generally excluded are the articles covered in the Twenge meta-analysis study (1997) described below. The participants in the reviewed research were in the U.S. and almost all were European American. Although our study is based in the U.S., these same struggles, with equality of men and women and of developing a culture free from androcentric views and patriarchal structures, are persistent and occur in many cultures.

Time Lag Designs

Twenge’s (1997) meta-analysis is a good starting point for time lag designs and for understanding differences across time in femininity and masculinity among college students. Twenge’s time lag design covered twenty-one years of BSRI administration to college students beginning in 1973 where “effects of the women’s movement were only beginning to be felt” (p. 305) to 1994 where many social changes toward equality had occurred. The earliest birth cohorts were 1949-1955 (early Baby Boomers) whereas the latest were 1970-1976 (early Generation X). In correlations of BSRI scores with year of administration, Twenge reported increased BSRI masculinity scores among women and men across time. There was also a suggestion that men’s femininity scores increased over time. Furthermore, men and women in the most recent years covered did not differ on masculinity, reflecting convergence, but there remained significant differences between men and women on femininity.

Twenge (1997) interpreted her results as reflecting differences in cohort experiences in formative years associated with different historical times. However, these findings are also congruent with an historical times interpretation. More recently, Donnelly and Twenge (2016) conducted meta-analyses on studies from 1993 to 2012 on masculinity, femininity, and androgyny. From 1993 to 2012, only women’s femininity scores decreased. The Donnelly and Twenge meta-analysis primarily embraced the years of backlash (Faludi, 1991) when not many changes over time would be expected. The decline in femininity among women is inconsistent with what might be expected.

Cross-Sectional Designs

In cross-sectional studies there are same time comparisons of people of different ages who also belong to different generations and represent varying experiences with historical events. For older birth cohorts, the women’s movement may have occurred too late to have an impact on the emerging adult socialization of gender roles (Strough, Leszczynski, Neely, Flinn, & Margrett, 2007). Among other cohorts the salience of the women’s movement would have been occurring at the same time as the development of identity and for still others (presumably younger) the women’s movement would have been the status quo or been under attack.

Age differences reflect not just cohort differences but also any differences that accrue due to the experiences of the maturing individual. The maturity principle predicts that people become more communal and more agentic as they take on adult roles (McAdams & Olson, 2010). Traditionally, agency is a match for work roles and communion a match for family/caregiving roles. When men perform work roles and women perform family roles, then agency would increase in men and communion would increase in women as they progress in these roles. But to the extent that men and women are in similar roles they would demonstrate similar agency and/or communion. But, if even “similar” roles are polarized (Suh, Moskowitz, Fournier, & Zuroff, 2004) men and women performing the same roles would still find men becoming more agentic and women more communion oriented. Hypothesizing from a different basis that presumes lessening of parental and procreation needs with increased age, Guttman’s parental imperative approach (1975) predicted a cross-over in gender roles. Older men would score higher on femininity than younger men and older women would score higher on masculinity than younger women.

Lemaster, Delaney, and Strough (2015) research compared three age groups in 2013 and found no differences in masculinity, femininity, or androgyny. With no age group or cohort differences, these results could provide support for an historical times explanation. But, without comparisons to earlier times, they shed no light on changes in gender polarization. Two cross-sectional studies of women (Erdwins, Tyer, & Mellinger, 1983; Mellinger & Erdwins, 1985) representing different ages/cohorts provided some support for the self-endorsement of gender roles consistent with the expectations for their cohort. Because time of measurement is constant, these cross-sectional studies shed light on age/cohort differences but not on time of measurement effects.

Using data collected two decades after the Erdwins et al. (1983) study, Strough et al. (2007) compared both men and women on BSRI masculinity and femininity in six age groups At their time of measurement in the mid-2000s, women’s masculinity scores varied by age: younger millennials and boomers were higher than the older preboomers, explained as a function of historical times when coming of age (Strough et al., 2007). These findings support the role of the gender equality movement of the 1960s and 1970s in framing gender role endorsement. However, it is also possible that there are age-related aspects at play such as role involvements in work and family at play (Erdwins et al., 1983; Mellinger & Erdwins, 1985). These cross-sectional studies do not allow conclusions about time of measurement.

Longitudinal Designs

A third approach to understanding gender roles that allows examination of time of measurement and age-related phenomenon is the longitudinal design where the same cohort(s) are studied over time. In this type of design, changes in scores on masculinity and femininity scales may be associated with (a) aging, (b) current life experiences (career, family, etc.), and/or (c) time of measurement. Yanico (1985) followed first-year college women from 1976 to 1980 and found no across-time differences on masculinity or femininity scores — as might be expected in this pre-backlash time. Across a 10-year time span encompassing the 1980s (a largely pre-backlash time period), Hyde, Krajnki, and Skuldt-Niederberger (1991) also found no across time differences in gender role categories. There are too few longitudinal studies to draw conclusions about effects across changing historical times.

In sum, the commentaries reviewed about the effects of the earlier movement toward gender equality and the effects of the more recent backlash presented compelling arguments. The research supported the idea that historical times represented by cultural messages at time of measurement were influences on self-endorsement of femininity and masculinity. Furthermore, cultural messages when coming of age, as well as age-related social roles, were also described as important to women’s and men’s self-reports of masculinity and femininity. Although there was strong support for the influences of historical times in earlier decades (Twenge, 1997) these results could be interpreted as time of measurement or time of coming of age (the explanation preferred by Twenge, 1997). And the studies that could shed light on effects of historical times were subject to alternative explanations. Taken together, there was tentative support for the idea that in post 9/11 times there would be gender conformity and polarization. But as noted in the review, a study with only a single design results in an inability to disentangle effects of time of measurement from birth cohort from age of participants. There are other limitations in the literature: (a) reliance on women in studies that included assessment of social roles of work and family; (b) use of nondiverse samples; (c) lack of assessment of the invariance of the measures across birth cohorts, time, and age; (d) heavy use of cross-sectional designs that were reviewed together in an overall time lag meta-analysis on college students from the mid-1970s to the mid-1990s (Twenge, 1997) and beyond (Donnelly & Twenge, 2016); (e) a scarcity of studies with measurements in more recent times; and (f) few studies of gender roles among the members of the millennial generation, many of whom came of age post 9/11.

Current Study and Hypotheses

Bem’s theory of gender schema (1981a, 1993) highlighted the importance of culture to the development and expression of gender role identity. Our primary interest is in demonstrating the impact of the war on equality during the recent backlash years on gender polarization. When the historical times encourage heightened awareness of and endorsement of traditional gender roles, there should be greater gender polarization compared to times that support gender equality. This study incorporates three designs, longitudinal, time lag, and cross-sectional, that allows inferences to be made about the impact of changes in society between two times of measurement (1982 and 2007). As well, the overall design allows checking the validity of the conclusions about time of measurement effects by consideration of alternative explanations based on either time period when growing up or age-related social roles. There are a number of advances in this study over previous research. For example, previous research did not use gender schema theory and many were atheoretical. We used the BSRI original, long form (Bem, 1974) as this has been used in most of the research. There is evidence that a short form has superior measurement qualities (Campbell, Gillaspy, & Thompson, 1997) and we come back to this in the discussion on measurement invariance. Our times of measurement span 25 years and reflect gender equality support (1982) and the post-9/11 backlash (2007). We included a sample from Generation Y (millennials) and later Baby Boomers. We included men as well as women to overcome the limitation of a number of studies that included only women.

Gender polarization is considered in two hypotheses about scores on masculinity, femininity, and androgyny: Hypothesis 1a, longitudinally, across time, compared to an earlier time period that encouraged gender equality, the more recent time period that discouraged gender equality would demonstrate decreased masculinity scores among baby boomer women and decreased femininity scores among baby boomer men; Hypothesis 1b, there would be decreases among baby boomers across time in androgyny. Hypothesis 2a predicts across time lags, college senior men would score lower on femininity and college senior women would score lower on masculinity in more recent times than in earlier times; and, Hypothesis 2b, androgyny would be lower in more recent times than in earlier times. Androgyny reflects a balanced and higher endorsement of masculine and feminine roles (Heilbrun & Pitman, 1979).

Because our samples are from two different cohorts assessed at two different times, there are a number of demographic characteristics that may distinguish these groups beyond the associations with work and family social roles noted in the literature review. We compare groups on demographics, such as ethnicity (see Harris, 1994, who questioned the validity of the BSRI for non-European American groups), religion, religiosity (Thompson & Remmes, 2002, who predicated their study on examining the notion that “masculinity thwarts people from embracing spirituality, whereas femininity promotes religious experience”, p. 521), and financial status (Mulholland, 1996, who associated entrepreneurialism with masculinity), and use these items as control variables in preliminary analyses.

Method

Participants and Procedures

The participants came from the same large Texas university in the southwestern U.S. Twenty-five years following the initial sampling of college seniors in 1982, the members of the 1982 cohort were invited to return in 2007. For analysis purposes, college seniors were restricted in age to 20-24 years, thereby including those who were more homogenous in age and who underwent college experiences at similar ages. Procedures in each recruitment period were approved by the institutional review board for the protection of human subjects; recruitment techniques depended upon year. All recruitment and data collection took place in spring and summer.

Participants recruited in 1982 understood that they were part of a longitudinal study, with five waves of measurement in 1982, 1983, 1984, 1990, and 2007. Only the 1982 and 2007 waves were used in this analysis (no comparable new group of college seniors was recruited until 2007). In 1982, random stratified sampling was used to include equivalent numbers of men and women. After letter and phone contact, participants met with project staff, filled out the questionnaire with the BSRI, were interviewed, and were paid $5.00.

In 2007, these participants were followed-up via letters and mailed surveys that included the long form BSRI (or for some, a short survey with no BSRI) with postage paid return envelopes. They were offered opportunities to enroll in drawings for cash rewards of $100 via a separate postage paid postcard. In all, 43% of the original participants in the larger study returned surveys. From these 180 returnees, participants were eliminated from this particular study if they were not part of the 1982 college senior group, if they had returned only the short survey (that omitted the BSRI), and if they were older or younger than the 18-24 age cut-offs used in this analysis. After these eliminations, there were 33 men and 29 women (n = 62) representing a late baby boomer cohort termed Group BB and designated as BBCS in 1982 and BBML in 2007.

In 2007, college undergraduates, including college seniors, were solicited via university emailed announcements, the university electronic announcement board, and in-person announcements in classes to complete a paper and pencil survey which included the BSRI among other measures. This approach meant almost every student had an equal opportunity to volunteer for the study. Rewards included random drawings for prizes donated by local businesses as well as drawings for monetary awards of $100. Recruitment of college seniors was monitored and data collection continued until similar numbers of men and women to those in the earlier longitudinal study were recruited. From a larger sample of 2007responding college seniors, participants were selected randomly within the constraints of having a sample similar in number and gender distribution to that of the baby boomer group. In 2007 there were 35 men and 30 women (n = 68) college seniors representing Generation Y termed Group GYCS. Groups BBCS and GYCS did not differ significantly in gender composition. In general, participants were European American and affiliated with Catholic/Protestant/Christian religions. College senior participants were less partnered than midlife adults and were less involved in full-time employment.

Measures

Dependent measures. The dependent variables were gender role measures from the 60-item Bem Sex Role Inventory (Bem, 1974). Sample masculinity items are: Defends own beliefs, Independent, Assertive. Sample femininity items were: Cheerful, Affectionate, Sympathetic. Items were answered on the original 7-point scale ranging from 1 = never or almost never true to 7 = always or almost always true in response to the prompt “Describe yourself”. The mean scale score was calculated and could range from 1 to 7. Measures of Cronbach’s alpha internal consistency reliability were acceptable, ranging from .66 to .92. Androgyny was calculated by the Heilbrun and Pitman (1979) procedure: (Masculinity+Femininity) – (|Masculinity-Femininity|).

To assess possible attrition effects, 2007 returnees from the baby boomer group were compared to nonreturnees on 1982 demographic characteristics and masculinity and femininity scale scores. There were no statistically significant differences between those who continued with the study and those who did not. When possible, analyses were conducted with and without covariates of ethnicity, religiosity, financial status, and relationships status. These preliminary analyses found nonsignificant covariates and patterns of significance and nonsignificance did not change across analyses. Therefore, the results reported are based on analyses without covariates.

Results

Testing Hypotheses

Table 2 includes means and standard deviations of masculinity and femininity by cohort group, year of measurement, and gender. As well, Cronbach’s alpha measures of internal consistency are also listed in Table 2. Due to societal backlash and stalled support for gender equality, the overall hypothesis is that the femininity, masculinity, and androgyny scores of those in 2007 would reflect greater polarization compared to those in 1982. In order to support this time of measurement hypothesis, the findings must fit a pattern that there are longitudinal differences (Hypothesis 1a; Hypothesis 1b) as well as time lag differences (Hypothesis 2a; Hypothesis 2b). Support for each hypothesis was examined in the context of multivariate analyses of variance and follow-ups. Figures 1 – 4 graph the results with longitudinal results in Figure 1, time-lag results in Figure 2, and cross-sectional results in Figure 3. Androgyny scores are plotted in Figure 4 for longitudinal, time-lag, and cross-sectional designs.

Hypothesis 1. Hypothesis 1a called for longitudinal differences across time among baby boomers. Time (1982 and 2007) and gender role scores (masculinity and femininity) were repeated measures and gender was a between-groups factor in the multivariate analyses (MANOVA). A main effect of time was expected and was found (Wilks λ = .76, p < .001, η2 = .24). The main effect of time was unmodified by interactions with BSRI, gender, or both. Overall, BSRI scores declined from 4.98 to 4.74.

Based on our specific hypothesis of declines in men’s femininity and women’s masculinity scores, follow-up tests were conducted. Within-gender analyses with paired t-tests of particular BSRI scores were calculated (see Figure 1). Across time, women’s masculinity scores declined (t(28) = 2.12, p < .05) from a mean of 4.93 to 4.70. Although women’s masculinity scores and femininity scores declined in similar magnitudes, the decline in femininity scores among women was not significant (4.97, 4.74). Among men, femininity scores declined (t(32) = 3.97, p < .001) from a mean of 4.72 to a mean of 4.35, but men’s masculinity scores did not decline significantly (5.31, 5.16). These patterns support the hypothesis of greater polarization over time periods from greater to lesser support for gender equality.

The Hypothesis 1b prediction, of a decline in androgyny scores over time, was supported (Wilks λ = .77, p < .001, η2 = .23). Androgyny scores were higher in 1982 than in 2007 (9.07 > 8.48), true for both men and women. In sum, the longitudinal design results supported greater gender polarization in more recent times than in more distant times.

Hypothesis 2. Hypothesis 2a called for time lag comparisons that would support greater gender polarization among college seniors in 2007 than in 1982 (GYCS, BBCS). The BSRI scores were treated as repeated measures, time of measurement and gender were treated as between group factors. There was a significant three-way interaction (Wilks λ = .94, p < .01, η2 = .06). The follow-up findings (see Figure 2) to this three-way interaction demonstrated, as predicted, men’s scores on femininity were significantly higher in 1982 than in 2007 (4. 72 > 4.32) (F(1,66) = 8.14, p < .01, η2 = .11), and women’s scores on masculinity were higher in 1982 than in 2007 (4.93 > 4.56) (F(1,57) = 4.26, p < .05, η2 = .07). As expected, there were no significant differences on women’s scores on femininity or men’s scores on masculinity by time of measurement.

With respect to the within person polarization prediction of Hypothesis 2b, there was marginal support (F(1,123) = 3.79, p = .05, η2 = .03). As predicted, androgyny scores were higher in 1982 than in 2007 (9.07 > 8.72), true for both men and women. Taken together, the time-lag findings support the hypothesis that there would be greater gender polarization in 2007 than in 1982.

Alternatives. As hypothesized, the longitudinal and time lag results provided support for gender polarization across a 25-year time span. But significant cross-sectional differences could compromise this interpretation. Therefore, we examined scores in the cross-sectional design in 2007 where time of measurement was constant with a MANOVA where cohort (BBML and GYCS) and gender were between group factors and BSRI measures were repeated. There were no cohort effects either as a main effect or in interaction with BSRI or gender. There was a weak three-way interaction of cohort by BSRI by gender (Wilks λ = .97, p = .045, η2 = .03). Follow-ups within gender indicated there were no significant effects when comparing groups by BSRI scores (for means see Table 2 and Figure 3). Finally, androgyny scores did not differ by cohort (F(1,60) = 1.28, ns, η2 = .01), gender (F < 1.00), or cohort by gender (F < 1.00). That is, the overall score on androgyny of midlife adults in 2007 (M = 8.48) was similar to the overall androgyny score of college seniors in 2007 (M = 8.72), true for men and women. Taken together, cross-sectional comparisons between midlife adults and college seniors in 2007 pose no challenges to the conclusions of the longitudinal and time-lagged results that there was greater gender polarization in 2007 than in 1982.

Table 2. Means, Standard Deviations, and Cronbach’s Alpha of 20-item BSRI Femininity, Masculinity, Androgyny Scale Scores By Time of Measurement, Group, and Gender
Table 2. Means, Standard Deviations, and Cronbach’s Alpha of 20-item BSRI Femininity, Masculinity, Androgyny Scale Scores By Time of Measurement, Group, and Gender

Discussion

The past fifty decades demonstrate fluctuations in support for gender equality. The 1960s and early 1970s saw the rise of a movement for gender equality (Bem, 1993; England, 2010). By the late 1980s and early 1990s this movement was countered by the New Right backlash (England, 2010; Faludi, 1991). And into this emergent backlash, the events of September 11, 2001, were fashioned into a potent push for traditional gender roles, for the reemergence of the gender polarization patterns characteristic of androcentric and patriarchal cultures (Faludi, 2007; Lorber, 2005). Bem’s (1981a, 1993) gender schema theory of gender role development argued that the adoption of conventional gender roles was more a product of the assigned roles in the social structure than of childhood socialization. In the language of this study, 1982 and 2007 represented two time periods characterized by different views of the roles of women and men. In 1982, gender equality was more culturally accepted, but in 2007, gender polarization was more strongly promoted. This study was designed to test the hypothesis that time of measurement would be a factor in the self-endorsed gender role scores of women and men, that the war on gender equality would be evident in such personal arenas as self-descriptions on masculine and feminine characteristics.

Because past research was characterized by designs that could not disentangle time of measurement from experiences growing up or from experiences as an adult, we incorporated three designs in this study (Schaie, 1965). The first design was longitudinal, comparing the same cohort from 1982 to 2007. Initially, these late baby boomers were all college seniors but, by 2007, they were midlife adults. The second design was time lag, comparing a cohort of college seniors in 1982 (the late baby boomers) with a cohort of college seniors in 2007 (Generation Y), groups who participated in different historical times and who came of age in markedly different times. The third design was cross-sectional with two groups measured in 2007. In this third design, one group was the GY cohort of college seniors and the other was of the baby boomer midlife adults. Each group measured in 2007 came of age in different times. They were also of differing ages with varying engagements in social roles of career and partnering.

Through triangulation of the resulting patterns across these designs, support for a time of measurement hypothesis was apparent. Such support is seen if, in both the longitudinal and time lag designs, there were expected differences between those assessed in 2007 from those assessed in 1982. Greater polarization was found (lower scores on femininity among men, lower scores on masculinity among women, and lower scores on androgyny among men and women). Furthermore, alternative explanations of age/maturation/imperatives were ruled out as there were no contradictory findings in the cross-sectional design analyses. The results extended the Twenge (1997) findings of increases in gender roles when her time frame began at the stirrings of the women’s movement to a period of greater support for gender equality (1974 to 1994). Although Twenge (1997) favored a coming of age explanation, our findings reinterpret her results as reflecting a time of measurement effect. Gender polarization had decreased over the time frame she analyzed compared to the time frame of this study which saw an increase in polarization. With respect to gender role scores, we conclude that changes in historical times affect people regardless of when they came of age.

The effect of time of measurement would appear to rest on women demonstrating differences or changes in masculinity and men demonstrating differences or changes in femininity, changes in the characteristics generally considered more evident in the other gender, at least when considered through a binary lens. Neither men nor women in this study showed increases or decreases on characteristics generally considered more evident in the same gender. However, in the more recent meta-analysis, Donnelly and Twenge (2016) reported that women’s femininity scores decreased from 1993 to 2012, a nineteen-year span that was generally within the backlash years. Because of the time span covered it is difficult to interpret these findings. By contrast, Twenge’s (1997) earlier analysis and ours spanned changing historical times. Generally, it appeared that over the past several decades, the effect of the war on gender equality is to suppress expression of roles more closely associated with the other gender than with the same gender. In times of gender equality backlash and when under threat, men suppress the feminine (Cheryan et al., 2015). In times of support for gender equality, the way people become more gender equal is for men to acknowledge more feminine qualities and women to acknowledge more masculine qualities.

As we look at the cross-sectional findings that compared college seniors and midlife adults in 2007, there were no cohort differences on masculinity, femininity, or androgyny despite the considerable differences in age and in social roles of career involvement and partnering. Thus, both cohorts “got” the gender polarization message of the times. These findings are consistent with the study by Lemaster el al. (2015), which compared three age cohorts in 2013 and found no differences on masculinity, femininity, or androgyny. The cross-sectional and longitudinal findings could have provided support for a maturity hypothesis (McAdams & Olson, 2010) had older men demonstrated higher scores on masculinity and older women demonstrated higher scores on femininity than younger men and women. There was no support for such an age-related hypothesis. Nor was there support for the parental imperative theorizing of Guttman (1975) that there would be a cross-over effect with men becoming more feminine and women becoming more masculine with age.

Writing in 1987, Helson and Moane speculated that midlife adults might find themselves more similar to current college seniors than to themselves when they had been college seniors. Indeed, results of this study provide support for this time of measurement prediction. As well, our findings offer the opportunity to more fully parse meanings behind the historical times perspective. Birth cohorts have been identified by generational groupings that provide a short hand to describe those who experience similar events and cultural and societal expectations. These generations develop their identities, gender roles and other perceptions of self as they come of age. However, the study’s lack of differences between midlife adults in 2007 compared to emerging adults in 2007 failed to confirm the importance of experiences when growing up. In support of Bem’s (1993) theorizing, current cultural contexts, regardless of generations, seemed to be strong enough to override variations in experiences growing up. For example, the midlife adults and college seniors in 2007 all shared the larger post 9/11 cultural context which Faludi (1991), Finlay (2006), and Lorber (2005) described. Indeed, history is marked by events that are turning points for many in the society, not just a particular cohort. If the commentaries we cited about the effects of recent history are correct, the attacks of 9/11, coming in the midst of a rising cultural backlash against the advances of women, may have been one of those periods of time that affected people regardless of generation. Thus, it becomes important to consider that what appear to be coming of age experiences or age-related roles instead may be more reflective of the contexts in which people are situated (Bem, 1993).

Of interest in these findings is the lack of effects (in analyses of covariance) of social role differences in such areas as marriage/partnering and career as an explanation for differences in gender roles. In the literature, studies that looked at social role involvements and gender roles were primarily cross-sectional and involved women only (e.g., Erdwins et al., 1983; Mellinger & Erdwins, 1985). By contrast, our study included men and women. A major difference between the samples used in these studies and ours was that most of the previous literature included birth cohorts that ended with late baby boomers. Our study included late boomers and millennials.

In a different time, with men and women equally involved in work and careers, with supportive social policies that allow and encourage such equality (Bem, 1993), gender polarization should disappear. However, Bem (1993) encouraged broader thinking on the subject. In order to dismantle gender polarization, she argued that androcentrism must be taken apart as well. Americans should “reconstruct their social institutions to be so inclusive of both male and female experience that neither gender is automatically advantaged or disadvantaged by the social structure” (Bem, 1993, p. 186). In such a world, concepts of masculinity and femininity and androgyny, among others, would be banished from the “cultural consciousness” (p. 192-193). For example, as Ferree (2010) noted, workplaces, such as law firms, may demand overwork that currently finds more women than men opting out. As well, some careers, such as those in information technology, allow men to retain typical gender schemas but women in the field adopt atypical gender schemas (Lemons & Parzinger, 2007). The movements of the 1960s for peace, civil rights, women’s rights, gay liberation (as these movements were called) were all challenges to patriarchy and to an androcentric and gender polarized world. At this writing, post-2016 Presidential election, the backlash dream of a patriarchal, androcentric, gender polarized atmosphere is reflected in the 45th president, the cabinet, and advisory appointments. As well, there is a notable resurgence of women and men who reject such arrangements. The day after the 2017 U.S. inauguration, women’s marches around the world brought over a million protesters together to assert the importance of human rights (Stein, Hendrix, & Hauslohner, 2017).

Limitations and Strengths

Although the study enjoyed a number of strengths, such as, different cohorts, different time periods, a prospective longitudinal design, and use of the 20-item BSRI scales to allow comparison with other research, there were some limitations. Effect sizes were generally small to moderate. We used binary categories: men and women; femininity and masculinity. As well, the study lacked diversity in ethnicity, social class, religious affiliation, and geography. Almost all participants were European American, all went to college, and recruitment was from only one university in one state in the U.S. These characteristics of our study limit generalizability to other experiences and regions of the country (Konrad & Harris, 2002) or to other countries (e.g., Colley, Mulhern, Maltby, & Wood, 2009). Gender intersects with race, class, ethnicity, and culture (Ferree, 2010; Stewart, Winter, Henderson-King, Henderson-King, 2015) but the lack of variability in our study precluded investigation of intersectionality. Goldberg, Kelly, Matthews, Kang, and Sumaroka’s (2012) research on college students of diverse ethnicities in 1988 and 2010 examined views on work-care benefits and costs. Such research is a reminder of the importance of including diverse samples in research on attitudes about gender roles, and by extension, on self-perceptions of gender roles. In this study, a preliminary analysis on the more diverse 2007 sample of college seniors indicated that ethnicity was not a significant predictor of gender role self-endorsements. However, the numbers of participants in our study precluded identifying possible differences among non-European American participants.

Although groups in this study differed as expected on social roles and age, we did not study people across their own life transitions which could help to tease out if and how these variables are associated with masculinity and femininity (McHugh & Frieze, 1997). The study was not experimental in that participants were not randomly assigned to groups in which societal or personal messages were manipulated. Nonetheless, the experimental study results of the Cheryan et al. (2015) experiment, in which men’s experiences were manipulated, supported the idea that men’s self-endorsement of femininity was responsive to threat.

Because our design allowed disentangling coming of age from time of measurement, we were able to comment on timing of socialization effects. Socialization effects during childhood and coming of age were not supported. But socialization effects may have been present in different contexts, with different areas of development beyond the exclusive focus in this study on gender roles. As noted by others (e.g., Donnelly & Twenge, 2016), the items on the BSRI could have different meanings in different decades, that is, they might not have demonstrated invariance over time. With 20 items per scale, our sample size was too small to test for invariance. However, the 10-item per scale BSRI (trimmed from the 20-item per scale measure) (Bem, 1981b) was testable for invariance. When we analyzed the 10-item version for metric and scalar invariance, these were acceptable in longitudinal, time-lagged, and cross-sectional analyses (Fischer & Juergens, 2015). As well, parallel analyses of the 10-item scales produced results that were similar to those reported here for the 20-item version. Thus, we conclude that the comparisons in this study were based on similar meanings of items across the 25 years.

In the future, use of such scales will be valuable in documenting continuity and discontinuity in the self-endorsements of gender roles as historical times change. Research will be able to track for whom and under what circumstances there are changes in gender polarization. When the future embraces the dismantling of androcentrism and gender polarization, then such measurements will no longer be needed.

Compliance with Ethical Standards

The authors declare that they have no conflicts of interest. All research was approved by the university’s institutional review board. Informed consent procedures were followed.

References

  • Anderson, K. J. (2015). Modern misogyny: Anti-feminism in a post-feminist era. New York, NY: Oxford University Press.
  • Arnett, J. J. (2000). Emerging adulthood: A theory of development from the late teens through the twenties. American Psychologist, 55, 469-480. doi: 10.1037/0003-066X.55.5.469
  • Bakan, D. (1966). The duality of human existence: An essay on psychology and religion. Oxford, England: Rand McNally.
  • Bem, S. L. (1974). The measurement of psychological androgyny. Journal of Counseling and Clinical Psychology, 42(2), 155-162. doi:10.1037/h0036215
  • Bem, S. L. (1977). On the utility of alternative procedures for assessing psychological androgyny. Journal of Consulting and Clinical Psychology, 45(2), 196-205. doi:10.1037/0022-006X.45.2.196
  • Bem, S. L. (1981a). Gender schema theory: A cognitive account of sex typing. Psychological Review, 88(4), 354-364. doi:10.1037/0033-295X.88.4.354
  • Bem, S. L. (1981b). The Bem Sex-Role Inventory: A professional manual. Palo Alto, CA: Consulting Psychologists Press.
  • Bem, S. L. (1993). The lenses of gender: Transforming the debate on sexual inequality. New Haven, CT: Yale University Press.
  • Campbell, T., Gillaspy, J. J., & Thompson, B. (1997). The factor structure of the Bem Sex-Role Inventory (BSRI): Confirmatory analysis of long and short forms. Educational and Psychological Measurement, 57(1), 118-124. doi:10.1177/0013164497057001008
  • Cheryan, S., Schwartz Cameron, J., Katagiri, Z., & Monin, B. (2015). Manning up: Threatened men compensate by disavowing feminine preferences and embracing masculine attributes. Social Psychology, 46(4), 218-227. doi:10.1027/1864-9335/a000239
  • Colley, A., Mulhern, G., Maltby, J., & Wood, A. M. (2009). The short form BSRI: Instrumentality, expressiveness and gender associations among a United Kingdom sample. Personality and Individual Differences, 46(3), 384-387. doi:10.1016/j.paid.2008.11.005
  • Constantinople, A. (1973). Masculinity-femininity: An exception to a famous dictum?. Psychological Bulletin, 80(5), 389-407. doi:10.1037/h0035334
  • DeShong, H. F. (2015). Policing femininity, affirming masculinity: Relationship violence, control and spatial limitation. Journal of Gender Studies, 24(1), 85-103. doi:10.1080/09589236.2013.833087
  • DiDonato, M. D., & Berenbaum, S. A. (2011). The benefits and drawbacks of gender typing: How different dimensions are related to psychological adjustment. Archives of Sexual Behavior, 40(2), April, 457-463. doi:10.1007/s10508-010-9620-5
  • Donnelly, K., & Twenge, J. M. (2016). Masculine and feminine traits on the Bem Sex-Role Inventory, 1993-2012: A cross-temporal meta-analysis. Sex Roles, online. doi: 10.1007/s11199-016-0625-y
  • England, P. (2010). The gender revolution: Uneven and stalled. Gender & Society, 24(2), 149-166. doi:10.1177/0891243210361475
  • Erdwins, C. J., Tyer, Z. E., & Mellinger, J. C.(1983). A comparison of sex role and related personality traits in young, middle-aged, and older women. International Journal of Aging and Human Development, 17(2), 141-152. doi:10.2190/6YQQ-EDQ3-KFAR-03Q3
  • Faludi, S. (1991). Backlash: the undeclared war against American women. New York, NY: Crown.
  • Faludi, S. (2007). The terror dream: myth and misogyny in an insecure America. New York, NY: Henry Holt & Co.
  • Ferree, M. M. (2010). Filling the glass: Gender perspectives on families. Journal of Marriage and Family, 72(3), 420-439. doi:10.1111/j.1741-3737.2010.00711.x
  • Finlay, B. (2006). George W. Bush and the war on women: Turning back the clock on progress. London: Zed Books.
  • Fischer, J. L., & Juergens, C. A. (2015). Bem Sex Role Inventory: Differences between men and women across 25 years. National Council on Family Relations Annual Meeting, November, Vancouver, British Columbia.
  • Gilligan, C. (2011). Joining the resistance. Oxford, England: Polity Press.
  • Goldberg, W. A., Kelly, E., Matthews, N. L., Kang, H., Li, W., & Sumaroka, M. (2012). The more things change, the more they stay the same: Gender, culture, and college students’ views about work and family. Journal of Social Issues, 68(4), 814-837. doi:10.1111/j.1540-4560.2012.01777.x
  • Guttman, D. L. (1975). Parenthood: Key to the comparative psychology of life? In N. Datan & L. Ginsberg (Eds.) Life-span developmental psychology: Normative life crises (pp. 98- 119). San Diego, CA: Academic Press.
  • Harris, A. C. (1994). Ethnicity as a determinant of sex role identity: A replication study of item selection for the Bem Sex Role Inventory. Sex Roles, 31(3-4), 241-273. doi:10.1007/BF01547717
  • Heilbrun, A. B., & Pitman, D. (1979). Testing some basic assumptions about psychological androgyny. Journal of Genetic Psychology, 135, 175-188. doi: 10.1080//00221325.1979.10534069
  • Helson, R., & Moane, G. (1987). Personality change in women from college to midlife. Journal of Personality and Social Psychology, 53(1), July, 176-186. doi:10.1037/0022-3514.53.1.176
  • Hyde, J. S., Krajnik, M., & Skuldt-Niederberger, K. (1991). Androgyny across the life span: A replication and longitudinal follow-up. Developmental Psychology, 27, 516-519. doi:10.1037/0012-1649.27.3.516
  • Kirmeyer, S. E., & Hamilton, B. E. (2011). Childbearing differences among three generations of U.S. women. NCHS data brief, no 68. Hyattsville, MD: National Center for Health Statistics.
  • Konrad, A. M., & Harris, C. (2002). Desirability of the Bem Sex-Role Inventory items for women and men: A comparison between African Americans and European Americans. Sex Roles, 47(5/6), 250-271. doi.10.1023/A:1021386727269
  • Lemaster, P., Delaney, R., & Strough, J. (2015). Crossover, degendering, or...? A multidimensional approach to life-span gender development. Sex Roles, doi:10.1007/s11199-015-0563-0
  • Lemons, M. A., & Parzinger, M. (2007). Gender schemas: A cognitive explanation of discrimination of women in technology. Journal of Business and Psychology, 22(1), 91-98. doi:10.1007/s10869-007-9050-0
  • Lorber, J. (2005). Breaking the bonds: Degendering and feminist change. New York, NY: W. W. Norton.
  • McAdams, D. P., & Olson, B. D. (2010). Personality development: Continuity and change over the life course. Annual Review Of Psychology, 61, 517-542. doi:10.1146/annurev.psych.093008.100507
  • McHugh, M. C., & Frieze, I. H. (1997). The measurement of gender-role attitudes: A review and commentary. Psychology of Women Quarterly, 21, 1-16. doi:10.1111/j.1471-6402.1997.tb00097.x
  • Mellinger, J. C., & Erdwins, C. J. (1985). Personality correlates of age and life roles in adult women. Psychology Of Women Quarterly, 9(4), 503-514. doi:10.1111/j.1471-6402.1985.tb00899.x
  • Mulholland, K. (1996). Entrepreneurialism, masculinities and the self-made man. In D. L. Collinson, J. Hearn, D. L. Collinson, J. Hearn (Eds.) Men as managers, managers as men: Critical perspectives on men, masculinities and managements (pp. 123-149). Thousand Oaks, CA, US: Sage Publications, Inc.
  • Parry, E., & Urwin, P. (2011). Generational differences in work values: A review of theory and evidence. International Journal of Management Reviews, 13(1), 79-96. doi:10.1111/j.1468-2370.2010.00285.x
  • Pedulla, D. S., & Thébaud, S. (2015). Can we finish the revolution? Gender, work-family ideals, and institutional constraint. American Sociological Review, 80(1), 116-139. doi:10.1177/0003122414564008
  • Robnett, R. D., Anderson, K. J., & Hunter, L. E. (2012). Predicting feminist identity: Associations between gender-traditional attitudes, feminist stereotyping, and ethnicity. Sex Roles, 67(3-4), 143-157. doi:10.1007/s11199-012-0170-2
  • Ruth, S. (1983). A feminist analysis of the New Right. Women’s Studies International Forum, 6(4), 345-351. doi:10.1016/0277-5395(83)90026-2
  • Schaie, K. W. (1965). A general model for the study of developmental problems. Psychological Bulletin, 64(2), 92-107. doi:10.1037/h0022371
  • Sessa, V. I., Kabacoff, R. I., Deal, J., & Brown, H. (2007). Generational differences in leader values and leadership behaviors. The Psychologist-Manager Journal, 10(1), 47-74. doi:10.1080/10887150709336612
  • Starr, C. R., & Zurbriggen, E. L. (2016). Sandra Bem’s gender schema theory after 34 years: A review of its reach and impact. Sex Roles, doi:10.1007/s11199-016-0591-4
  • Stein, P., Hendrix, S., & Hauslohner, A. (2017, January 22). Women’s marches: More than one million protesters vow to resist President Trump. The Washington Post. Retrieved from http://www.washingtonpost.com
  • Stewart, A. J., & Healy, J. M. (1989). Linking individual development and social changes. American Psychologist, 44(1), 30-42. doi:10.1037/0003-066X.44.1.30
  • Stewart, A. J., Winter, D. G., Henderson-King, D., & Henderson-King, E. (2015). How politics become personal: Sociohistorical events and their meanings in people’s lives. Journal of Social Issues, 71(2), 294-308. doi:10.1111/josi.12111
  • Strough, J., Leszczynski, J. P., Neely, T. L., Flinn, J. A., & Margrett, J. (2007). From adolescence to later adulthood: Femininity, masculinity, and androgyny in six age groups. Sex Roles, 57(5-6), 385-396. doi:10.1007/s11199-007-9282-5
  • Suh, E. J., Moskowitz, D. S., Fournier, M. A., & Zuroff, D. C. (2004). Gender and relationships: Influences on agentic and communal behaviors. Personal Relationships, 11(1), 41-59. doi:10.1111/j.1475-6811.2004.00070.x
  • Thompson, E. J., & Remmes, K. R. (2002). Does masculinity thwart being religious? An examination of older men’s religiousness. Journal for the Scientific Study of Religion, 41(3), 521-532. doi:10.1111/1468-5906.00135
  • Twenge, J. M. (1997). Changes in masculine and feminine traits over time: A meta-analysis. Sex Roles, 36(5-6), 305-325. doi:10.1007/BF02766650
  • U.S. Census Bureau (2015). Current population survey, March and annual social and economic supplements, 2014 and earlier. Downloaded from: http://www.census.gov/hhes/families/files/ms2.csv.
  • Wood, W., & Eagly, A. H. (2002). A cross-cultural analysis of the behavior of women and men: Implications for the origins of sex differences. Psychological Bulletin, 128(5), 699-727. doi:10.1037/0033-2909.128.5.699
  • Yanico, B. J. (1985). BSRI scores: Stability over four years for college women. Psychology of Women Quarterly, 9(2), 277-283. doi:10.1111/j.1471-6402.1985.tb00878.x
« Prev section Next section »