bivariate analysis of gender and homonegativity, we found that there was weak evidence of a bivariate relationship (men and women did not differ significantly in their homonegative attitudes). However, in Model 2 (the full model), we find that gender does provide a significant effect while controlling for other variables. This evidence now supports the second hypothesis that males are more homonegative. Further analysis was conducted to unveil what variables contributed to this suppressor effect. Models 3 through 8 are those being used to discuss the relationship between gender and homonegativity. Model 3 demonstrates that there is no evidence of a relationship between gender and homonegativity (even while controlling for the year the data were collected from). However Models 4, 5 and 6 all show that the relationship becomes significant when we control for age, education, and religious fundamentalism. As noted in bivariate analysis, women in the sample tended to be older, more religiously fundamental, and have less educated, which all contribute to higher levels of homonegativity. When controlling for any one of these factors, the previously masked relationship becomes prevalent.

      To add to the complexity of this relationship, the gender role scale was reintroduced into the model. In bivariate analysis we found that women in the sample tended to have more egalitarian attitudes about gender roles. When the gender role scale was reintroduced in the presence of gender, education, age, or any combination of the three, we found that gender was again no longer significant. The lack of relationship remained intact until religious fundamentalism was controlled for8. In essence, the gender role scale better explains gender differences in homonegativity until the level of religious


8 For simplicity, not all model combinations were shown