Correspondence should be addressed to Gerald Jones, Box 18425, Los Angeles CA 90018-0425
Sexism toward women scores of university students at the beginning of an introductory Women's and Men's Studies (Gender Studies) course were compared with sexism levels of the same students at the end of the course, and with sexism levels of controls. Differences in sexism levels associated with Bem's sex role categories were investigated. Sexism levels among students in every category, except undifferentiated, were significantly lower at the end of the course, while controls showed no pre/post differences. As expected, males were higher than females both at pretest and at posttest and sexism in both sex groups decreased significantly by equivalent amounts; but androgynous males in the introductory course were lower in sexism than sex-typed males, and androgynous females were lower than sex-typed females. Undifferentiated subjects decreased in sexism least during the course, and cross-typed and androgynous males decreased most. The relationship of sexism to several additional variables not included in formal hypotheses also is discussed. These variables were age, religion, ethnicity, SES and grade earned in the Gender Studies course.
Sexism is the oppression of women, brought about by discriminatory actions and attitudes based on an assumption that females are subordinate (Jaggar & Struhl, 1978). Specifically, the present study follows the detailed definition used by Benson and Vincent (1980) in their development of the instrument used here to measure sexism. For them, sexism includes 1) Attitudes that women are genetically inferior to men, 2) Support for the premise that men should have greater rights and power, 3) Support for anti-female sex discrimination, 4) Hostility for women who engage in traditionally masculine behavior or fail to fulfil traditional female roles, 5) Lack of support and empathy for women's liberation movements and the issues involved in such movements, (6) Utilization of derogatory labels and restrictive stereotypes in describing women, and 7) Evaluation of women on the basis of physical attractiveness and willingness to treat women as sexual objects.
The locus of this study was an undergraduate introductory course which combines Women's and Men's studies in an interdisciplinary overview. The course, offered at a major, urban, private university, is team-taught by one female and one male professor, and supported by a female and a male Teaching Assistant. At the time of the present study, there were two such teams offering separate (but parallel) courses, serving a total of approximately 375 students per semester.
The study is exploratory and collected demographic data including age, year in school, ethnic background, religion and SES, but was also designed to test six hypotheses. The principal expectation of the present study (Hypothesis 1) is that students completing the introductory Women's and Men's Studies course will have significantly lower levels of sexism at the end of the semester when compared with their own sexism results from the beginning of the semester, as well as when compared to a control group of similar students.
The investigators expected to find sex differences in sexism levels, similar to those found in the development and testing of the Sexist Attitudes Toward Women Scale (Benson & Vincent, 1980). Hypothesis 2, therefore, states that males will be higher in sexism than females both before and after the course. Hypothesis 3 states that separately measured sexism levels of both women and men will decrease by equivalent amounts.
Since the introduction of an instrument to measure androgyny and other sex roles (Bem, 1974) over 100 studies have shown that gender role is more significant than biological sex in predicting and evaluating behavior (Bem, 1984). This notion led to Hypothesis 4, that sex-typed individuals will be higher in sexism than androgynous or cross-typed individuals. (According to Bem, 1984, results associated with undifferentiated individuals are inconsistent and less predictable.) Again, however, the expectation that the course would have an effect on all groups of students led the investigators to Hypothesis 5, that all gender- role categories of individuals will show significant decreases in sexism at the end of the course.
Since sexism is ubiquitous in American society, affecting and being promoted by both men and women (Tavris & Wade, 1984), it seems likely to assume that the reduction of sexism (through a Gender Studies course) would be equally likely whether the faculty member interacting with and evaluating the students were a female or a male. (In the present study the lectures were shared equally by a female and a male professor, but the students were grouped for discussion and exam evaluation with either a female or a male teaching assistant for the duration of the course.) Hypothesis 6 states that students with a female Teaching Assistant (TA) would not differ from those with a male TA in their reduction of sexism at the end of the course.
A majority of the subjects were Caucasian (62%), the balance being Asian/Pacific (17%), Hispanic (10%), Black (10%) and Other (1%). Most students were lower division undergraduates; median age of the sample was 19 years. The control group of 61 females and 59 males was obtained from three undergraduate courses, each of which, like the Gender Studies course, fulfilled basic requirements common to most university students and could be expected to reflect a cross-section of students: introductory Anthropology, English composition, and introductory Political Science.
Sexist Attitudes Toward Women Scale (SATWS). This is a 40- item Likert-scaled instrument developed by Benson and Vincent (1980) in which the subject is asked to rate statements regarding the role and position of women in society on a scale from 1 (strongly disagree) to 7 (strongly agree). Each subject receives a single score on a scale from 40 to 280, the latter score representing the highest level of sexism.
Scoring: Items occasionally left blank by subjects could potentially affect the total score, since blank items are scored as zero (0); therefore, an average item score was computed based only on those items marked, and then the item score was multiplied by 40, yielding a prorated sexism score based on actual responses, regardless of the number of blanks left by the subject. (This correction was not part of the scoring procedure in the SATWS pilot tests as reported by Benson and Vincent, 1980.)
The Bem Sex Role Inventory (BSRI). This widely-used instrument, developed by Bem (1974, 1981), uses a checklist of 60 adjectives, and yields two scores, one for femininity and one for masculinity, for each individual. Subjects were classified in the present study using a median split technique, as suggested by Bem (1977), resulting in assignment into one of six categories: androgynous, female sex-typed, male sex-typed, female cross-sex- typed, male cross-sex-typed and undifferentiated. The present study divided the androgynous and undifferentiated categories by sex, resulting in a total of eight categories --- four for females and four for males.
One potential problem exists with the design of the study, though the investigators have made every attempt to minimize its effect. It is possible that the pretest itself, which deals with attitudes towards women, might have sensitized the subjects to feminist issues, thereby influencing the outcome of the posttest. The authors feel this type of sensitizing effect is unlikely, for the following reasons: 1) the instructions were designed to give the impression that many departments, other than the Gender Studies program, were involved in the study; 2) the participants took only 20 minutes during a regular class session filling out the questionnaires; and 3) the questionnaire (SATWS) focuses on attitudes; by contrast, every attempt is made during each semester to remind students that the course is based on scholarship and evidence, and that personal reactions and attitudes are not what is being taught.
For data analysis, ANOVAs have been used when possible to minimize the error terms. In Hypotheses 2, 4 and 6, where analysis of sexism levels at pretest or at posttest, not changes in sexism levels over time, are of interest, separate ANOVAs are performed (sexism [dep. var.] X sex X sex of TA for Hypotheses 2 and 6; sexism [dep. var.] X gender role for Hypothesis 4). In Hypotheses 3 and 6, where change in sexism over time in dichotomous groups is of interest, a single ANOVA using a difference score variable (posttest sexism minus pretest sexism) is performed (sexism difference [dep. var.] X sex X sex of TA). In Hypotheses 3 and 5, where a comparison of two or more groups' posttest sexism with their own pretest sexism is of interest, individual t-tests for dependent samples are performed. Note that these t-tests do not result in an increase of the error term, because between-group comparisons are not of interest to this Hypothesis. The tests used in the "Additional Results" section follow similar guidelines.
At the end-of-semester posttest, femininity and masculinity levels for control and experimental subjects were equivalent, although sexism levels (with the effects of the pretest removed through an analysis of covariance) were significantly lower for students in the experimental group than for controls (F[1,441] = 19.66, p<.001). Also, t-tests for dependent samples showed that posttest sexism levels in the control group were equivalent to their own pretest levels, while posttest sexism levels in the experimental group (M = 111.6) were significantly lower than their pretest sexism levels (M = 119.2), t(354) = -7.61, p<.001. Hypothesis 1 was solidly confirmed.
The prediction of Hypothesis 2 that males would be higher in sexism than females was confirmed, as shown in Table 1.
Summaries of ANOVAs for Hypotheses 2 and 6:
Sexism (dep. var.) X sex X sex of TA
(Experimental Subjects Only)
Main signif. Main signif.
Effect d.f. F of F Effect d.f. F of F
Sex 1 89.01 <.001 Sex 1 40.04 <.001
Sex of TA 1 0.17 (.68 ns) Sex of TA 1 0.01 (.93 ns)
TOTAL d.f. 404 TOTAL d.f. 354
Means of Significant Main Effects
Sex: Female Male Female Male
Mean: 105.46 138.73 99.04 130.19
SD: 27.98 34.38 28.98 36.43
n: 244 161 212 143
Hypothesis 3 also was confirmed. In t-tests for dependent samples, the 212 females in the experimental group showed a significant decrease of 6.8 points (Ms = 105.8 at pretest and 99.0 at posttest), t(211) = -5.62, p<.001, and the 143 males in the experimental group showed a significant decrease of 8.9 points (Ms = 139.1 at pretest and 130.2 at posttest), t(142) = -5.14, p<.001. The magnitude of the difference (i.e., the difference between the 6.8 point decrease for females and the 8.9 point decrease for males) was shown to be equivalent in a Sexism Difference (dep. var.) X Sex X Sex of TA ANOVA in which there were no significant effects or interactions.
Results pertinent to Hypotheses 4 and 5 are presented in Table 2. Both at pretest and posttest, androgynous males were
Experimental Group Mean Sexism Levels (SATW Scale)
at Pretest and Posttest by Posttest Gender Role Category
Gender Role Mean Sexism
Classification Means Difference
(n) Pretest Posttest (decrease) t p
Sex-typed Male (53) 147.9 139.1 -8.8 -3.31 <.005
(39) 140.6 135.9 -4.7 -1.58 n.s.d.
Androgynous Male (28) 138.6 128.7 -9.9 -2.15 <.05
(41) 112.8(a) 109.0 -3.7 -1.79 n.s.d
Cross-typed Male (22) 119.5(a) 104.4 -15.1 -3.16 <.01
Sex-typed Female (69) 110.7 103.1 -7.6 -3.89 <.001
(39) 102.9 95.5 -7.4 -2.26 <.05
(63) 97.7 90.2 -7.5 -3.04 <.005
(a)-Except for these two values (which changed relative position at
posttest), all other means are arranged in rank order.
lower than sex-typed males in sexism, but not lower than sex- typed females (Sexism [dep. var.] X Gender Role ANOVA), thus failing to confirm Hypothesis 4. Same-sex-only rank-order comparisons showed that androgynous males were lower than sex- typed males, and androgynous females were lower than sex-typed females, though these differences were not significant at the .05 level (Scheffe test). Sex-typed males were highest in sexism in all analyses, and sex-typed females were consistently relatively low, with androgynous females lowest of all.
On the other hand, Hypothesis 5 was partially confirmed. The sexism level in each gender role category was lower at posttest, and t-tests for dependent samples reached significance in all but two categories --- undifferentiated males and females. (Pretest and posttest means, t values and significance levels are reported in Table 2). Cross-sex-typed males decreased more than other groups, while female and male undifferentiated groups decreased least.
Hypothesis 6 was confirmed. ANOVAs revealed no significant main effects either in sexism levels (see Table 1) or in amount of decrease in sexism (ANOVA reported above regarding Hypothesis 3) between those students who had a female TA and those who had a male TA.
Additional Results . Several variables were included in data collection and analyzed without formal hypotheses. These variables included age, religion, ethnic background (race), SES, and grade (evaluation) earned from the course.
The median age was 19 years, with 94% of subjects between 17 and 22. An ANOVA at pretest which included experimental and control subjects revealed an interaction between sex and age which approached significance (p=.058), in which older females are lower in sexism than younger females, while older males are higher in sexism than younger males. Posttest interactions between age and sex in the experimental subjects were not significant.
With regard to religion of the experimental subjects, ANOVAs revealed no significant differences between groups at pretest or at posttest. Within religious groups, however, t-tests showed that three of the five groups decreased significantly (p<.001) at the posttest. Table 3 displays the number of subjects, means and t-statistics, as well as a breakdown of the pre- and posttest means by sex.
Experimental Group Pretest and Posttest Means
by Religious Group
Means Females Males
(n) Pre Post Decrease t p Pre Post Pre Post
(27) 117.4 115.5 -1.9 -0.59 n.s.d. 104.3 103.6 131.5 128.2
(119) 119.6 113.6 -6.0 -3.51 <.001 106.4 101.9 143.9 135.2
(7) 133.7 111.6 -22.1 -2.01 (.091) 133.0 124.7 134.0 106.4
(102) 119.6 111.1 -8.4 -4.62 <.001 104.6 96.4 144.7 135.9
(76) 119.8 108.6 -11.2 -5.02 <.001 109.4 98.0 132.0 121.0
(14) 112.2 107.9 -4.3 -.77 n.s.d. 98.4 91.9 130.7 129.3
Note: Groups are arranged in rank order according to posttest means.
Three of the religious groups warrant further comment. The number of Moslems was too small to produce reliable t-test results, but among the seven subjects the decrease in sexism of over 22 points is remarkable. The 27 Jewish students constituted a sufficient, though small, group for statistical purposes, but their decrease of 1.9 points was not significant. Fundamentalist Christians were too few to generate reliable statistics (n = 14), but they did decrease by 4.3 points, and were the lowest in sexism of all groups both at pretest and at posttest.
In the breakdown of religious group means by sex shown in Table 3, it can be noted that all categories decreased by five points or more, except Jewish females and Fundamentalist Christian males. By far the largest decrease in these categories was among Moslem males (n = 5). These findings, and all the data related to religious groups point to the need for further research regarding the relationship of religious affiliation and sexism. Such research should be careful to specify the "other" category in more detail than did the present study, especially noting the distinction between "religious other" and "non- religious".
Subjects were classified into five ethnic groups, though "Native American" (Indian) was too small (n = 3) to warrant comparisons with other groups. ANOVAs with post-hoc Scheffe (.05) comparisons revealed that between the remaining four groups, Asians were significantly higher in sexism than the other three relatively equivalent groups -- Blacks, Caucasians, and Hispanics -- both in the pretest analysis, which included both experimentals and controls (F[3, 515] = 6.92, p<.001), and in the posttest analysis, which included experimental subjects only (F[3, 347] = 4.75, p<.005). All four ethnic groups showed significant decreases in sexism (significance of t-tests for dependent samples at p<.05 or better) when posttest means were compared to pretest.
With respect to SES, subjects were categorized into four groups according to their estimates of family income (parent or self): under $15,000 per year (n = 41), $15 to 30,000 (n = 53), $30 to 50,000 (n = 83), and over $50,000 (n = 167). There were no significant patterns or interactions due to SES level when sexism levels were analyzed at pretest (experimental and control subjects combined) or at posttest (experimental subjects only). However, like all other breakdowns of the experimental group, subjects in each of the four SES categories showed significant (p<.05 or better) pre-to-posttest decreases in sexism levels.
In an analysis of differences associated with the grade each experimental group subject received at the end of the course, an ANOVA showed that students who earned higher grades had significantly lower posttest sexism levels than students with lower grades, F(3, 215) = 8.2, p<.001. Posttest sexism means were 134.1 for D students (n = 16), 114.2 for C students (n = 76), 98.2 for B students (n = 110), and 96.7 for A students (n = 14). This trend was the same for females and for males.
It is not clear whether the change in sexist attitudes would be reflected in students' behavior as well. It is one thing to change an opinion, and quite another to have an effect on a person's actions. This question awaits further research, including the development of measures and methodology for assessing sexist behavior and its potential change.
What facilitated the reduction of sexism is not clear. The content of the course, with its focus on analyzing discrimination, sex differences and related issues, may have caused students to rethink their assumptions and adopt less sexist attitudes based on new information. It may be, however, that role modeling was occurring among those students who never before had come into extended contact with feminists.
The question of whether course content or role modeling accounts for lower sexism levels after a Gender Studies course is addressed in part by the tendency for female students with a female Teaching Assistant (TA) and male students with a male TA to decrease in sexism more than students with an other-sex TA. Further research should investigate such factors as discussion group dynamics, same-sex vs. mixed sex discussion groups, and students' expressions of identification with the discussion leader(s).
The present study found higher sexism levels among men than among women. It must be noted that our sexism results for men and women even at posttest (Ms of 135.5 and 96.97, respectively --- see Table 1) were higher than those of male and female college students in other published work. Benson and Vincent (1980) found sexism means in college students of 102.45 for males and 85.23 for females. These differences should be interpreted in the light of our use of an averaging technique, as described above, to correct for items left blank. Nonetheless, it is possible that our sample, drawn from one large, urban university, is not representative of college students in general.
The present study found, contrary to expectations, that all androgynous students were not less sexist than all sex-typed students: androgynous males were below sex-typed males in sexism, but above sex-typed females; androgynous females were below all other groups. Androgyny clearly is advantageous in terms of lower sexism levels, particularly when females are compared to females and males to males, but sexism apparently cannot be assumed to be linked simply with all stereotyped roles, given the relatively low sexism of sex-typed females.
Future research should explore the relationship between sexism and gender role, particularly in cross-typed and undiffer entiated subjects. In the present study, undifferentiated females and males showed the smallest decreases, neither of which reached statistical significance.
The striking contrast of Blacks, Hispanics and Caucasians (essentially equivalent posttest sexism means of 99.5, 103.9, and 110.7, respectively) with the significantly higher levels of Asians (posttest mean of 126.5) needs further investigation. (See results of ANOVA and post hoc comparison, reported above.) It is possible that the presence in our sample of a number of non-American Asians made a difference in the results, as other ethnic groups contained few if any non-Americans; the design of the present study did not provide for this analysis.
Finally, there is the intriguing finding of an inverse relationship between success in the course, in terms of letter grade, and posttest sexism. This is similar to previous research which noted a relationship between sexism and High School GPA (Bayer, 1975) and between sexism and overall College GPA (Etaugh, 1975). Are high-achieving students less sexist than average or lower-achieving students? Do students' sexism levels affect their ability to perform academically in Gender Studies courses? Or do instructors tend to grade more favorably those students perceived as less sexist? Research into this interaction could extend into many other areas involving controversial course content, including courses dealing with racism, nationalism, sexual orientation or religion.
Sexism is an undesirable remnant of our past. It is comforting to know that the means exist at least to reduce sexism in academic situations. One of the first results of this knowledge might well be to encourage Women's and Men's studies curricula at the High School level, and even earlier. The inclusion of Gender Studies viewpoints in all disciplines also is warranted, since such viewpoints contribute to the effort to eliminate sexism, even as they help to move academia toward more accurate representations of women's and men's co-participation in society.
Bem, S.L. (1974). The measurement of psychological androgyny. Journal of Consulting and Clinical Psychology, 42, 155-162.
Bem, S.L. (1977). On the utility of alternative procedures of assessing psychological androgyny. Journal of Consulting and Clinical Psychology, 45, 196-205.
Bem, S.L. (1981). The Bem Sex-Role Inventory Professional Manual. Palo Alto, CA: Consulting Psychologists Press.
Bem, S.L. (1984). Androgyny and gender schema theory: A conceptual and empirical integration. In T.B. Sondregger (Ed.), Nebraska Symposium on Motivation: Psychology and Gender, Lincoln: University of Nebraska Press.
Benson, P.L., and Vincent, S. (1980). Development and validation of the sexist attitudes toward women scale. Psychology of Women Quarterly, 5, 276-291.
Brod, H. (1987). The new men's studies: From feminist theory to gender scholarship. Hypatia: A Journal of Feminist Philosophy, 2, 179-196.
Deckard, B.S. (1975). The Women's Movement: Political, Socioeconomic, and Psychological Issues. New York: Harper and Row.
Etaugh, C. (1975). Biographical predictors of college students' attitudes toward women. Journal of College Student Personnel, 16, (4), 273-276.
Ferree, M.M., & Hess, B.B. (1985). Controversy and Coalition: The New Feminist Movement. Boston: Twayne Publishers.
Jaggar, A.M., & Struhl, P.R. (1978). Feminist Frameworks: Alternative Theoretical Accounts of the Relations Between Women and Men. New York: McGraw-Hill.
Tavris, C., & Wade, C. (1984). The Longest War: Sex Differences in Perspective (2nd Edition). New York: Harcourt, Brace, Jovanovich.