The concept of Subtle Sexism was introduced in the 1990ies when overt sexism was found to be on the decline. Subtle sexism was defined rather vaguely as harmful treatment of women that goes unnoticed because it is conceived as normal behavior. The Modern Sexism Scale was devised to measure this kind of sexism but the validation did not include such behavior. Neither has the scale’s ability to measure sexism defined as subtle or covert sexist beliefs been documented. Further, the scale has paradoxical behavior in that it will show no sexism when discrimination is universal and universal sexism when there is no discrimination.
A concept of subtle sexism over and above the lack of belief in discrimination has not been articulated, and the Modern Sexism Scale is not a scientific tool to measure subtle or covert sexism. It assesses beliefs about discrimination against women. Its relation to sexism is the simple one that if a respondent does not believe discrimination is a problem the respondent is classified as sexist. There is nothing subtle about this.
In the past, an effective approach to subtle sexism has been to describe it concretely and get it recognized as sexism. That would be a strategy to revive. As to the Modern Sexism Scale, it may be helpful for assessing beliefs about discrimination where discrimination is known to be present.
Benokraitis & Feagin  introduced the division of sexism into of blatant, subtle and covert forms. This typology has been widely used, although the meaning has shifted over time. Despite reservations that subtle sexism by nature of being subtle could be difficult to document, Benokraitis & Feagin described 9 types: condescending chivalry, supportive discouragement, friendly harassment (including flattery), subjective objectification, radiant devaluation, liberated sexism, benevolent exploitation, considerate domination, and collegial exclusion, all of which were richly illustrated with examples.
Today, it is striking how little of what Benokraitis & Feagin included under the rubric of subtle discrimination would still be regarded as subtle. A female faculty member reported in an interview that a female colleague had been in a heated discussion with a male colleague. That latter “finally exploded, “What that young lady needs is a good spanking” . If this and other examples were subtle in the 1980ies, then Benokraitis & Feagin certainly accomplished a mission of bringing them out in the open and transferring them to the category of overt sexism.
In 1997, Swim and Cohen introduced a questionnaire to measure Subtle Sexism. They defined Subtle Sexism as “openly unequal and harmful treatment of women that goes unnoticed because it is conceived to be customary or normal behavior” . The main justification for the introduction of the new measure was the decline over the preceding years in sexist beliefs as measured by then prevalent tools such as the Attitudes Toward Women Scale. Overt sexism was on the decline but other kinds of sexism were felt to have taken over. Parallels were drawn to racism where a trend had been observed from overt racism to more subtle or covert forms. The theory was that the same transition had happened with sexism.
The phenomenon was illustrated with a single example which found that although men and women rated female and male leaders equally, their non-verbal behavior toward the female leaders was more negative than toward the male leaders . No other examples of behavior were given. Without further theorizing, the authors claimed that Subtle Sexism could be measured with the new questionnaire, the Modern Sexism Scale. The new version of Subtle Sexism was thus from the outset rather nondescript, and it was basically defined by reference to another novelty, the Modern Sexism Scale.
Whereas the definition of one unknown by another introduced an element of circularity, it would have been possible to identify the sexist behavior of interest (subtle sexist behavior) and show that the scale measured this behavior, and then define Subtle Sexism as what the scale was measuring. In the following I will argue that nothing of the sort was done.
For twenty years, the Modern Sexism Scale has been used by scientists as well as practitioners who have accepted the premise that it measures Subtle Sexism. It is typically introduced by statements like “The [Modern Sexism Scale] …. measures covert and subtle, as opposed to overt and blatant, forms of sexism” , and “we administered the Modern Sexism Scale …, a well-validated instrument” . But does anyone really know what Subtle Sexism is? I will argue below that the concept is as vague as ever.
The Modern Sexism Scale
A main feature of the Modern Sexism Scale is that it is supposed to measure Subtle Sexism whereas it seems to measure something else. Examination of a questionnaire should start with a thorough read-through of the actual questions – an evaluation of the face validity. This would seem to be a commonplace, but as judged by the gap between what people claim of a questionnaire and what is often clear from reading the questions, this is hardly a superfluous remark. So here are the 8 questions of the Modern Sexism Scale:
1. Women often miss out on good jobs due to sexual discrimination.•
2. It is rare to see women treated in a sexist manner on television.
3. Society has reached the point where women and men have equal opportunities for achievement.
4. It is easy to understand the anger of women’s groups in America.•
5. Over the past few years, the government and news media have been showing more concern about the treatment of women than is warranted by women’s actual experiences.
6. Discrimination against women is no longer a problem in the United States.
7. On average, people in our society treat husbands and wives equally.
8. It is easy to understand why women’s groups are still concerned about societal limitations of women’s opportunities.• (cited form )
Each item is scored on a Likert-type scale. Items with asterixis are reverse coded.
The first 5 are grouped as denial of the existence of discrimination against women, the next 2 as feeling resentment about complaints about discrimination, and the last as feeling resentment about special favors for women . Since all questions weigh the same in the scale, the questions about belief in the existence of discrimination drive the mean score.
This should lead to the preliminary conclusion that the latent variable that is being measured is the respondent’s belief that women are being discriminated in the United States.
How did Swim et al. get from this face value of the questionnaire to concluding that it measures Subtle Sexism? In essence, they did not.
Conventionally, the validation of a scale is the demonstration that the scale measures what it purports to measure. Here, one would expect to see demonstrated that the Modern Sexism Scale actually measures subtle sexism as defined by the authors, namely “openly unequal and harmful treatment of women that goes unnoticed because it is conceived to be customary or normal behavior”.
The original validation of the Modern Sexism Scale  included 3 elements
- Demonstration that men scored higher on the scale than women
- Demonstration that the scale correlated with humanitarian-egalitarian values
- Demonstration that the scale correlated with overestimation of the percentage of females in male-dominated occupations
A later publication  added:
- Demonstration that the scale was different from the Attitude Toward Women Scale
- Demonstration that the scale was correlated with the Attitude Toward Women Scale
- Demonstration that the scale correlated with affective reactions to different categories of women and men, for example feminists and chauvinists
- Demonstration that the scale correlated with perceptions of what constitutes sexual harassment
That’s all there was. There was absolutely nothing about sexist behavior, let alone sexist behavior that goes unnoticed.
Here’s a thought experiment: I claim that I’ve invented a test for Parkinson’s disease. This disease is more common in men than in women. My validation of the test will now have as a main element whether it is positive in more men than women. Also, I will check whether it performs differently from a test for diabetes, and whether it correlates with this latter test. I’ll find a couple of other laboratory tests that do not measure quite the same as the new test. If these conditions are met, I will conclude that the test diagnoses Parkinson’s disease, without taking any interest in whether the subjects I have tested actually have Parkinson’s disease or not. Would you be happy to be diagnosed with Parkinsonism by this test? You shouldn’t. But this is essentially the “validation” that was done of the Modern Sexism Scale at its introduction. A non-validation.
There is no reason to belabor this point. The simple fact of the matter is that the validation made at the introduction of the scale was irrelevant. It did not address “harmful treatment of women that goes unnoticed because it is conceived to be customary or normal behavior” which the scale was supposed to measure.
Later studies of relevance to validation
For the Modern Sexism Scale to be validated it would have to be shown that the scores be highly correlated with the behavior that it is claimed to measure, i.e. “unequal and harmful treatment of women that goes unnoticed because it is conceived to be customary or normal behavior”. That would require an study setup radically different form the usual validation of one questionnaire by another.
There is a study which has used a relevant setup – and it was not in the theory’s favor. Greenawalt  studied 155 young female athletes’ preferences for coaches, and their sexism as measured on the Modern Sexism Scale. She also interviewed 10 of the athletes who preferred male couches. This would seem to be a good setup for a validation of the Modern Sexism Scale, since there are not only in-depth interviews to cross-check the questionnaire responses but also actual behavior, the choice of coach.
Greenawalt found that the athletes’ scores on the Modern Sexism Scale were in general low, i.e. they were generally in agreement that society discriminates against women (and thus according to the theory they did not harbor modern sexist views). Still, 80% of them preferred male couches.
In the interviews, the 10 women laid out their reasons for preferring male coaches in a way that Greenawalt interpreted as clearly sexist. In other words, although the athletes agreed that women miss out on jobs because of discrimination, they did not believe that their own preference for male coaches over female coaches represented discrimination.
If we accept that this constitutes sexist behavior, the read-out the Modern Sexism Scale of no sexism stands out as a blatant failure of the instrument. The questionnaire is supposed to address exactly the setting where overt sexist statements are rare but sexist behavior still prevalent. Here the sexist behavior is universal by design, but the scale shows no sexism.
Three other studies are at least marginally relevant to the validation question. Two correlated the Modern Sexism Scale to respondents’ evaluations of résumés [5,8]. In the study by Moss-Racusin et al., an application for a position as laboratory manager was sent to a sample of 127 professors. All professors received the same application, the only difference being that the application had been randomly assigned a male or a female name. Participants were asked to rate the applicant’s competence, hireability and potential salary and mentoring that would be offered. The professors’ “subtle bias” against women was measured with the Modern Sexism Scale. The main result were that male applicant were rated higher than female applicants (by both male and female professors). The Modern Sexism Score was associated with less support for the female applicants. The latter finding might indicate that the professors who deny discrimination are more prone to discriminate – which is indeed thought provoking.
The study setup by Ingalls  has clear similarities to Moss-Racusin et al. but the findings are opposite. Participants were presented with 3 résumés that should be ranked. One group evaluated the résumés with names on, the other groups made blind evaluations. The main finding was that the résumés with female names were ranked higher than the male named résumés in the non-blind setting, whereas there was no difference when the evaluations were made blinded. The evaluations did not co-vary with scores on the Modern Sexism Scale. The participants in Ingalls’ study were predominantly female college students.
 includes a Study 2, where the authors correlated the Modern Sexism Scale with a behavior, the use of sexist language. Respondents were instructed to describe 3 dilemmas in their own words, and the number of descriptions where they used “he” and “she” was the dependent variable “sexist language”. The main result was that respondents who scored in the lower third on the Modern Sexism Scale used such language in an average of 0.68 out of the 3 situations, whereas respondents in the upper third averaged 1.21. The difference was statistically significant.
The question we are interested in is, however, not statistical significance, but whether the study demonstrated that the Modern Sexism Scale measures subtle sexist behavior, here whether it measures sexist language as defined in the paper. The answer is clearly no. The reason is that the correlation between the Modern Sexism score and the Sexist Language score was only 0.15 . This may be statistically significant, but it falls way below what would be needed for an assessment to be a measurement of something.
Paradoxical behavior of the Modern Sexism Scale
The lack of validation is however not the only problem with the Modern Sexism Scale. Let us pretend for a moment that we live in a society with absolute equality between men and women. Discrimination is a thing of the past. Do women miss out on good jobs due to sexual discrimination in this society? No, we’ll have to strongly disagree (5 points, reverse coded item). Is it rare to see women treated in a sexist manner on television? Yes, indeed it is, so we strongly agree (5 points). Society has reached the point where women and men have equal opportunities for achievement? Strongly agree (5 points). By now we have earned a mean score of 5 on the Modern Sexism Scale, i.e. maximally sexist, and we aren’t going to fare better with the rest of the questions.
This cursory inspection of the questions tells us one important thing about the questionnaire, namely that if no discrimination is present in a population, the scale will give a strong read-out of sexism. It is like having a thermometer that shows our child has fever and must stay home from school, when the child is completely healthy. We would immediately replace this thermometer with one that actually measured the temperature.
Let us nonetheless make another thought experiment: Now imagine a society in which discrimination is universal and everybody can see it. Do women miss out on good jobs because of discrimination? Yes, strongly agree (1 point, reverse coded item). Is it rare to see women treated in a sexist manner on television? No, strongly disagree, we see it all the time (1 point). Has society reached the point where women and men have equal opportunities for achievement? Not at all (1 point). We have now earned a mean of 1 point on the Modern Sexism Scale, which is minimum, i.e. the Modern Sexism Scale has measured no sexism even though discrimination is universal. So now our thermometer shows no fever when our child is sick, and we send it to school.
How did everything get upside down here? The Modern Sexism Scale measures the respondent’s belief that women are being discriminated. That is straightforward. The problem lies in the interpretation that lies inherent in the scale namely that if respondents do not believe there is discrimination, they are sexist. That only works under the condition that 1. there is discrimination and 2. people do not want to admit it. In these circumstances it may make sense to interpret lack of belief in discrimination as a kind of sexism (although it is still far from proven that the sexism has the form of harmful behavior that goes unnoticed).
But that places a burden of proof on the researcher who wants to use the scale. Before using the scale and interpreting it in terms of sexism the researcher must make good that these two conditions are met – which begs the point of using the scale. When it has already been established by other means that there is discrimination that people do not admit, what is then the point of administering the questionnaire?
Other hints to what Subtle Sexism might be
The proponents of the scale may object that I’m taking things too literally. “Subtle sexism” is a theoretical construct which is not to be understood as any specific behavior. It is akin to subtle racism and we must try to become aware of what it is in our daily life although it is difficult to pin down, exactly because it is subtle. But although subtle, it exerts great influence on what we do, and it is harmful to women.
Apart from being a rather nebulous theory, it would not help the Modern Sexism Scale. Rather on the contrary, it would amount to admitting that it is not possible to match the theoretical construct of Subtle Sexism to empirical validation data, and then there is no scale. Furthermore, this line of argumentation would indicate that the validation was a sham.
A related objection could be raised on the grounds that the theory is really about covert sexist beliefs, rather than behavior. For example Swim & Hyers talk about the scale as a measure “of subtle or covert sexist beliefs” . This does not solve the problem either. Now it would have to be shown that the questions about belief in discrimination in reality measure secrets which the respondents do not want to disclose. Iit would be no small feat if such an instrument had indeed been developed, but all we get to know about these beliefs is that they are covert and sexist. They are so clandestine that even those who developed an instrument to measure them, are unable to describe them. But without bringing these covert beliefs out in the open and documenting that the scale is able to measure them there is no validation, and the concept of “subtle or covert sexist beliefs” remains obscure. It would be akin to having a thermometer which quite obviously measures the child’s temperature, and claiming that the temperature discloses something else, a hitherto unknown disease of grave consequence. That would nor fly in medicine. It shouldn’t in social science either.
The Modern Sexism Scale as a measure of Attitudes towards discrimination
No data have been presented to verify that the Modern Sexism Scale measures Subtle Sexism whether defined as behavior or covert beliefs. In addition, the scale exhibits paradoxical behavior under circumstances that cannot easily be identified.
As argued above, the candidate for a measured construct is “belief in discrimination”. In view of later findings that the scale scores correlated with:
Evaluation of what is sexist language and use of such language [9,11]
Emotional reactions to feminists, chauvinists, and other person categories 
Evaluation of what constitutes sexual harassment 
“attitudes towards discrimination” or “attitudes to feminism” might fit even better. One of the attractions of these labels would be that they do not signal any particular response being more acceptable than another. “Sexist” is a derogatory term, and labelling a person as sexist is obviously a moral judgement. Respondents who volunteer to fill in a questionnaire are thus routinely labelled sexist by the researcher. The research ethics of this is questionable.
The issues pertaining to equal opportunities for men and women are complex. Material inequalities in income and power-positions are obvious, but the causes of inequalities are intricate and have been debated for decades. A person with insufficient intellectual power to grasp the complexities of these issues (whom of us can claim to fall outside this category?) might get a high score, but would that justify the researcher passing derogatory moral judgements on the respondent? Likewise, a person who has never experienced or heard of female family members, friends, or acquaintances who were negatively discriminated against when applying for a job, might be inclined to disagree that women often miss out on job opportunities because of discrimination. This respondent may be ill informed about the state of affairs outside her own circles, but being ill informed is not a moral deficiency.
Hidden in plain sight
In conclusion, the Modern Sexism Scale has never been demonstrated to measure anything but attitudes to discrimination, and the concept of Subtle Sexism remains obscure. It is one of those things that are so obvious that they can be difficult to see. All it takes is, however, to ask the question: where is the data that shows that this questionnaire, which asks about beliefs in discrimination, in reality measures “harmful treatment of women that goes unnoticed because it is conceived to be customary or normal behavior”, alternatively clandestine sexist beliefs? The answer is that there is no data. Later research has not strengthened the case, rather to the contrary. It is easy to see that the scale has paradoxical behavior when discrimination is either absent or universal.
Unfortunately, the Modern Sexism Scale continues to be used, and study results continue to be referenced as if they showed something about sexism. The study by Moss-Racusin et al. which used the Modern Sexism Scale has become a mainstay in the literature about bias against women in academia. The study has been referenced 1605 times (Google Scholar as of April 2019), more than the paper which introduced the scale . At the same time, the meaning of Subtle Sexism has shifted further away from Benokraitis & Feagin’s original formulation, and is increasingly coalescing with the idea of unconscious bias against women, a development made possible by the nondescript nature of both terms.
Rather than continuing on this ill-founded path, I suggest that one take example of Benokraitis & Feagin. Their effective approach to handling subtle sexism was to describe it concretely, bring it out in the open and get it recognized as sexism. Let us see what it is. As to the Modern Sexism Scale, it may be helpful for measuring beliefs about discrimination where discrimination is known to be present but not for measuring sexism.
1. Benokraitis NV, Feagin JR. Modern sexism. Englewood Clifs: Prentice-Hall; 1986.
2. Swim JK, Cohen LL. Overt, Covert, And Subtle Sexism: A Comparison Between the Attitudes Toward Women and Modern Sexism Scales. Psychology of Women Quarterly. 1997 Mar 1;21(1):103–18.
3. Butler D, Greis FL. Nonverbal affect responses to male and female leaders: Implications for leadership evaluations. Journal of personality and social psychology. 1990;58(1):48–59.
4. Yoon E, Adams K, Hogge I, Bruner JP, Surya S, Bryant FB. Development and Validation of the Patriarchal Beliefs Scale. Journal of Counseling Psychology. 2015;62(2):264–79.
5. Moss-Racusin CA, Dovidio JF, Brescoll VL, Graham MJ, Handelsman J. Science faculty’s subtle gender biases favor male students. Proceedings of the National Academy of Sciences. 2012 Oct 9;109(41):16474–9.
6. Swim JK, Aikin KJ, Hall WS, Hunter BA. Sexism and racism: Old-fashioned and modern prejudices. Journal of personality and social psychology. 1995;68(2):199–214.
7. Greenawalt NJ. Modern Sexism and Preference for a Coach Among Select National Collegiate Athletic Association Division I Female Athletes: A Quantitative and Qualitative Analysis. Theses and Dissertations (All). 539. http://knowledge.library.iup.edu/etd/539; 2012.
8. Ingalls SA. The Effects of a Blind Selection Process on Gender Discrimination in Applicant Selection. Electronic Theses, Projects, and Dissertations. 715. http://scholarworks.lib.csusb.edu/etd/715; 2018.
9. Swim JK, Mallett R, Stangor C. Understanding subtle sexism: Detection and use of sexist language. Sex roles. 2004;51(3–4):117–28.
10. Swim JK, Hyers LL. Sexism. In: Nelson T, editor. Handbook of prejudice, stereotyping, and discrimination. New York: Psychology Press; 2009.
11. Parks JB. Attitudes toward women mediate the gender effect on attitudes toward sexist language. Psychology of Women Quarterly. 2004;28(3):233–9.