19
masculine ad condition received significantly higher ratings on masculinity than users portrayed
in the feminine ad condition (F(1, 333)= 132.19, p < .01, M
masculine
= 5.58, M
feminine
= 4.24,),
whereas users portrayed in the feminine ad condition received significantly higher ratings on
femininity than users portrayed in the masculine ad condition (F(1, 333)= 79.05, p < .01,
M
masculine
= 3.90, M
feminine
= 4.80). Therefore, the results of the manipulation checks were
satisfactory.
Self-ratings on Masculinity and Femininity
Subjects rated how ideal it would be if they had the 20 characteristics in the masculinity
subscale and the 20 characteristics in the femininity subscale of Bem’s (1974) Sex Role
Inventory on a 7-point Likert basis. Using median split criteria suggested by Spence and
Helmreich (1978), subjects were categorized into four groups. Subjects who scored high on
both masculinity and femininity were categorized as androgynous (N = 104, 48% of them were
male). Subjects who scored high on masculinity and low on femininity were categorized as
masculine (N = 64, 56% of them were male). Subjects who scored low on masculinity and high
on femininity were categorized as feminine (N = 66, 33% of them were male). Finally, subjects
who scored low on both masculinity and femininity were categorized as undifferentiated (N =
100, 62% of them were male). Cronbach’s reliability alphas for masculinity and femininity
were satisfactory at .90 and .84 respectively.
Ad-self-congruency on Masculinity and Femininity
Two ad-self-congruency scores were calculated, one for masculinity and the other for
femininity. Ad-self-congruency on masculinity was first calculated by subtracting the mean
ratings of portrayed users on masculinity from the mean self-ratings on masculinity. The sum
was then squared given that the study was concerned only with the degree of discrepancy or
congruency, not the direction. Therefore, a larger number indicated higher levels of
Convention