Because our sample consisted of modest numbers of lesbian women (n=79), cisgender gay men (n=69), bisexual men (n=11), and bisexual women (n=154), we conducted preliminary MANOVA analyses comparing bisexual and gay men and bisexual and lesbian women, respectively, on study variables of interest with the goal of increasing statistical power. Because the overall F-statistics for bisexual and gay men [F(9, 70)=1.82, p = .08] and bisexual and lesbian women [F(9,216)=1.47, p=.16], respectively, did not significantly differ, data for bisexual and gay men and bisexual and lesbian women were combined, bringing the sample size of sexual minority men to 80 and the sample size of sexual minority women to 233.
Additionally, we examined differences in gender and sexual orientation on demographic variables. Individuals identifying as men were slightly but significantly younger (Mage = 20.7, SDage = .20) than individuals identifying as women (Mage = 19.8, SDage = .12, p > .001). There were also racial/ethnic differences between individuals identifying as women (28.9% indicated identification as a racial/ethnic minority) than men (37.9% indicated identification as a racial/ethnic minority). Then, LGB participants had slightly higher BMIs (MBMI = 25.6, SDBMI = .35) than straight counterparts (MBMI = 24.0, SDBMI = .22). Also, individuals identifying as LGB were significantly older (Mage = 21.7, SDage =.20) than their straight counterparts (Mage = 18.7, SDage = .13). The LGB participants were also more educated: straight participants were more likely to be currently enrolled in college (with 97% enrolled) in comparison to LGB participants (with 70% enrolled). Further, LGB participants were more likely to have either completed a bachelor’s degree or graduate/professional degree (21.4%) in comparison to straight participants (with only 1.8% reporting having these degrees). These modest sample asymmetries are likely secondary to recruitment strategies. Individuals identifying as heterosexual were predominantly drawn from introductory psychology classes, whereas sexual minority recruitment was also supplemented through additional outlets (e.g., Facebook, MTURK, etc.).
MANCOVA for differences in weight stigma and sociocultural influences
Factorial MANCOVA (Gender x Sexual Orientation) was conducted with BMI and race (White or other) as covariates and with outcome variables being the four UMB-FAT scales and the five SATAQ-4 scales. The values of Pillai’s Trace (aka multivariate partial eta-squared) were highest for gender (.154) and BMI (.144) and smallest for race (.024) and the interaction (.056) with sexual orientation having an intermediate value (.088).
Univariate tests showed that the MANCOVA was significant for every outcome variable, with proportions of variance (R2, aka h2) being .173 for media pressure, .121 for family pressure, .112 for attraction weight stigma, .088 for athletic internalization, .053 for thin internalization, and less than .05 for the remaining outcomes.
For weight bias, significant main effects for sexual orientation were observed for negative judgments [F(1,994)=8.14, p<.01], attraction [F(1,994)=30.64, p<.001], and equal rights [F(1,994)= 7.39, p<.01]. Heterosexual individuals demonstrated more weight bias toward persons who are obese than did sexual minority individuals. Furthermore, significant main effects for gender were observed for all forms of weight bias [negative judgment [F(1,994)=7.06, p<.01], attraction [F(1,994)=32.68, p<.001], social distance [F(1, 994)=9.33, p<.01], equal rights [F(1,994)=34.17, p<.001]. Men demonstrated significantly more bias towards overweight or obese persons than did women. No significant interaction effects were observed between gender and sexual orientation. See Table 1 for MANCOVA cell and marginal adjusted means and standard errors.
For sociocultural attitudes, significant main effects for gender were observed for all measures except for societal pressure from peers: (Family, [F(1,994)=33.11, p<.001]; Media, [F(1,994)=70.94, p<.001]; Internalization Thin, [F(1,994)=19.95, p<.001]; Internalization Athletic/Muscular, [F(1,994)=14.63, p<.010]). Women reported receiving significantly more pressure from family, the media, and showed more internalization of thin ideals. However, men demonstrated more internalization of the muscular ideals than their women counterparts. There was also a main effect for sexual orientation for societal pressure from the media [F(1,994)=16.82, p<.001] and internalization of muscular ideals [F(1,994)=37.90, p<.001]. Sexual minority individuals reported significantly more pressure from the media about their weight, but internalized standards of muscular ideals less compared to their heterosexual counterparts. There was a significant interaction between sexual orientation and gender for media pressure toward appearance [F(1,994)=23.71, p<.001] and muscular standards [F(1,994)=16.15, p<.001]. These interactions were probed with tests of the simple main effects of sexual orientation. Sexual minority status had a significant simple effect on athletic ideals among men, F(1.286) = 42.63, p < .001. partial h2 = .13). Internalization of athletic ideals was greater in straight men (adj M = 17.51) than in gay men (adj M = 13.29). Straight women had a slightly greater mean (adj M = 14.44) than did gay women (adj M = 13.63), but this difference fell short of statistical significance, F(1, 713) = 3.61, p = .058, partial h2 = .005. Media pressure regarding appearance was significantly greater among gay men (adj M = 13.23) than among straight men (adj M = 9.78), F(1, 286) = 27.96, p < .001, partial h2 = .09, but among women the effect of sexual orientation was nil, F(1, 712) = 0.45, p = .51, partial h2 = .001.
After adjusting for overlap with other predictors, BMI was positively associated with pressure from family (b = .31), peers (b = .21), media (b = .12), and internalization of the thin ideal (b = .10), but negatively associated with attraction bias (b = -.17), internalization of muscularity (b = -.12), distance bias (b = -.09), and judgment bias (b = -.08). After adjusting for overlap with other predictors, minority participants were less affected by the thin ideal (ES = .24), attraction bias (ES = .19), internalization of muscularity (ES = .18), and media pressure (ES = .14). The effect sizes (ES) here are the differences between adjusted means divided by the overall standard deviation.
Partial correlation and regression examining weight stigma and sociocultural influences
Partial correlations were utilized to examine the relationship between weight stigma and sociocultural influences controlling for the influence of racial/ethnic identity and BMI. Greater weight bias (negative judgments, social distance, equal rights, and attraction) was associated with internalization of the muscular/athletic ideal. Greater negative judgments about people who are overweight and perceiving people who are overweight or obese as unattractive was significantly associated with greater internalization of the thin ideal. The feeling that people with obesity are unattractive was significantly associated with individuals receiving significant family pressure toward appearance. Greater negative judgments toward people with obesity, a greater desire to socially distance from people with obesity, as well as the belief that people who are overweight and obese should have equal rights as people who are normal weight were significantly associated with greater pressure from peers to be thin. Interestingly, greater media pressure was significantly associated with less weight bias (negative judgements, distance, equal rights, and attraction).
Linear regression examining weight stigma, gender, sexual orientation, including two- and three-way interactions were examined. Overall, there was no evidence that gender identity or sexual orientation (gender x weight bias; sexual orientation x weight bias; gender x sexual orientation x weight bias) impacted the strength of the association between weight bias and sociocultural variables. Regression tables available upon request.