abstract. This Essay reports the results of a survey experiment that we conducted on over eight hundred heterosexual respondents to compare associational attitudes toward gay men who engage in different types of sexual practices. Specifically, we randomly assigned respondents to hear one of three descriptions of a gay character, which differed only with regard to the character’s penetrative preference: top (preferring to penetrate one’s partner), bottom (preferring to be penetrated by one’s partner), and versatile (having an equal preference). Overall, we find that heterosexuals displayed heightened and statistically significant associational aversion toward versatile characters and, to a lesser degree, toward bottom characters, relative to respondents’ willingness to associate with top characters. We elaborate why heterosexuals seem to display systematically less associational aversion toward those men whose penetrative preference is most consistent with gender stereotypes. Based on those results, we revisit the notion, adopted by many courts, that Price Waterhouse sex-stereotyping doctrine cannot apply to sexuality claims because it would turn sexual orientation into a protected class after Congress has opted not to do so. Our results suggest that gender-motivated homophobia is not uniformly targeted toward all gay men or uniformly present among all who discriminate on the basis of sexual orientation. We also further consider why respondents were most averse to versatility, drawing a potential distinction between “trait opposition” and “trait intermediacy” gender violations. Finally, we discuss the implications of our findings for the broader LGBT movement in law and society.

authors. Ian Ayres is William K. Townsend Professor, Yale Law School. Richard Luedeman is a J.D. Candidate, Yale Law School, Class of 2014. For helpful comments, thanks to Tara Ayres, Matthew Nelson, Travis Pantin, Vicki Schultz, and seminar participants at the American Law and Economics annual conference. For excellent research assistance, thanks to Su Da.

Introduction

Many people are familiar with the scene in which an elder relative finds out that someone in the family is in a gay or lesbian relationship and timidly asks, “So which one of you is the man and which is the woman?” Some may also be familiar with the considerably less benign scene in which someone attempts to question a man’s masculinity by suggesting that he enjoys being anally penetrated, or to question a woman’s femininity by suggesting that she enjoys penetrating others. Employees have been harassed and discriminated against for “tak[ing] it up the ass.”1 These sorts of references, whether oblique and benign or direct and malicious, point to a certain cultural fascination with the dynamics of sexual penetration. Moreover, this fascination echoes the well-documented cultural importance that sexual penetrative dynamics had in ancient societies2 and continue to have in the LGBT community today.3 Indeed, many LGBT people carry strongly held preferences about their roles in sexual penetration, preferences that can become a substantial part of their identities within their communities.

There is, however, essentially no empirical understanding of whether the broader, modern-day American heterosexual population tends to judge people according to their “penetrative preferences.”4 Do heterosexuals harbor differential animus toward people with different penetrative preferences? Would heterosexuals be more averse to associating with an LGBT person with a certain penetrative preference? And what relevance could those attitudes and prejudices have to law and society? This Essay makes a preliminary attempt to answer those questions. To do so, we conducted a modest experiment in which we introduced respondents to fictional characters and described the characters’ penetrative preferences, with an eye to detecting whether different penetrative preferences might lead to different reactions from our heterosexual respondents.

In doing so, we also introduced respondents to the labels commonly used among gay men: one who prefers to be the penetrating partner is a “top,” one who prefers to be the receptive partner is a “bottom,” and one who readily engages in both is “versatile.”5 For this preliminary study, we focused primarily on reactions to gay male characters in order to test as many hypotheses as possible within a very short survey.6 Although our conclusions are stated in general terms and may well apply to attitudes towards lesbian and bisexual women, we have not sought to draw any particular conclusions about that topic with this first study, recognizing that there may be real differences in how female sexuality is perceived.7

What we found was that people did respond differently depending on which penetrative preference we assigned to the character. Respondents seemed to expect, and even to prefer in some cases, that male characters, whether gay or straight, be penetrators or “tops.” Our results provide a new way to interpret the reach of a key case in antidiscrimination law, Price Waterhouse v. Hopkins,8 which introduced the legal doctrine concerning gender stereotyping9 as a form of workplace discrimination. Our findings support the idea that some but not all heterosexual aversion to homosexuals and other gender-nonconforming groups may derive from gender-motivated prejudice.

Section I.A sets the backdrop for understanding why Price Waterhouse doctrine could be informed and modified by taking into account public attitudes towards penetrative preferences. We describe what we call the “Price Waterhouse dilemma”: On the one hand, Price Waterhouse says that gender stereotyping must not motivate employment decisions. Since bi/homosexuality defies predominant gender stereotypes, one might therefore expect Price Waterhouse to forbid employers from discriminating against bi/homosexual employees on the basis of sexual orientation. However, Congress has repeatedly failed to include sexual orientation as an explicitly protected category under Title VII. In deference to Congress, even the most progressive of courts have therefore only granted relief to bi/homosexual plaintiffs who focus on their nonsexual10 gender-nonconformity—such as their manner of speech or dress—rather than on their bi/homosexuality itself.

We argue that courts need not bracket sexuality altogether in order to show deference to Congress. Section I.B advocates distinguishing between sexual orientation and other related but separate dimensions of sexuality that could be perceived as gender-nonconforming,11 such as a particular penetrative preference. This distinction is important because although Congress has rejected the categorical protection of sexual orientation under Title VII, it has never addressed the question whether penetrative preference, as an independent aspect of sexuality, is protected. Section I.C then additionally argues that, even in the case of sexual orientation, a court can apply Price Waterhouse in a way that protects bi/homosexual plaintiffs, but stops short of categorically including bi/homosexuality within the protections of Price Waterhouse. Echoing Vicki Schultz’s discussion of same-sex sexual harassment,12 we advocate a greater focus on employers’ actual subjective motivations under Price Waterhouse, which would leave open the possibility that employers could express kinds of prejudice against homosexuality that are not gender-motivated (e.g., religiously based prejudice).A critical question on summary judgment then becomes how plausible it is that a given instance of alleged prejudice against a bi/homosexual plaintiff was gender-motivated. We discuss how our results regarding penetrative preferences bear on that question.

Part II describes in detail the methods and results of our experiment. Part III then concludes by reconnecting the results to Price Waterhouse doctrine. It also draws in further related questions about other gender-nonconforming groups, public opinion, and the future of the LGBT movement.

I. the price waterhouse dilemma and penetrative preferences

A. Current Approaches to the Dilemma

Justice Brennan’s plurality opinion in Price Waterhouse v. Hopkins13 turned the American legal system’s attention to the concept of gender stereotyping. He sweepingly declared, “[W]e are beyond the day when an employer could evaluate employees by assuming or insisting that they matched the stereotype associated with their group . . . .”14 The opinion held that “[i]n the specific context of sex stereotyping, an employer who acts on the basis of a belief that a woman cannot be aggressive, or that she must not be, has acted on the basis of gender” for the purposes of Title VII protections.15 Ann Hopkins’s aggressive personality was viewed negatively because she was a woman, whereas an aggressive man would have been viewed more positively. Price Waterhouse therefore seemed to hold that employers cannot punish employees for possessing a certain trait if they would not also punish a member of the other sex for possessing that same trait—in other words, employers cannot punish gender-nonconformity.

Conceptually, the reasoning behind Price Waterhouse would seem to permit an enormous range of discrimination claims.16 Gender stereotyping could potentially include any and all assumptions about the sexual attractions or behaviors that befit people of a given sex. Lesbian women, for example, are discriminated against for violating the stereotype that women must prefer to have sex with men. Moreover, gay men and lesbian women frequently are characterized or identified by a particular appearance or set of behaviors that does not accord with an observer’s gendered expectations.

Still, courts have remained hesitant to apply sex-stereotyping analysis to discrimination cases brought by gay men and lesbian women, insisting that “a gender stereotyping claim should not be used to ‘bootstrap protection for sexual orientation into Title VII.’”17 This may be because some courts have simply failed to acknowledge the logical connection between the cultural label “gay” and many gendered stereotypes.18 But even where courts explicitly acknowledge the connection between gender stereotyping and sexual-orientation discrimination, they refuse to extend Price Waterhouse that far.19

A highly salient rationale for this refusal seems to be a sense of judicial restraint: although Price Waterhouse’s interpretation of Title VII’s broad text could cover bi/homosexual plaintiffs, Congress has had numerous opportunities to expand Title VII to explicitly cover sexual orientation discrimination but has not done so. In Simonton v. Runyon, for example, the Second Circuit wrote, “[W]e are informed by Congress’s rejection, on numerous occasions, of bills that would have extended Title VII’s protection to people based on their sexual preferences.”20 This rationale invokes what some have called the “rejected proposal” rule of statutory interpretation: courts should disfavor interpretations that were considered, but rejected, by Congress as explicit amendments to the statute.21 Because Congress has for decades failed to pass the Employment Non-Discrimination Act (ENDA), which would give explicit employment protections to LGBT people,22 courts tend to believe that interpreting Title VII to cover sexual orientation would contravene congressional intent.

More generally, some courts seem to think that interpreting Title VII to cover sexual orientation would be so categorical and socially impactful as to step over the line between interpretation and amendment. Perhaps the most direct elaboration of this reasoning comes from Vickers v. Fairfield Medical Center, in which the Sixth Circuit wrote,

Ultimately, recognition of Vickers’ claim would have the effect of de facto amending Title VII to encompass sexual orientation as a prohibited basis for discrimination. In all likelihood, any discrimination based on sexual orientation would be actionable under a sex stereotyping theory if this claim is allowed to stand, as all homosexuals, by definition, fail to conform to traditional gender norms in their sexual practices.23

Recognizing this dilemma, courts have tasked themselves with the difficult duty of drawing a line based on the particular facts of each case between non-actionable sexual-orientation discrimination and actionable gender stereotyping.

One approach is to treat gender stereotyping and sexual-orientation discrimination as mutually exclusive. The court tries to determine what was actually motivating the employer’s actions: the plaintiff’s sexual orientation or the plaintiff’s other gender-nonconformity. As many scholars have pointed out, however, in practice this exercise gives courts license to dismiss perfectly valid claims of gender stereotyping merely because the plaintiff was bi/homosexual.24 Plaintiffs may try to avoid this fate by concealing their sexuality, emphasizing their nonsexual gender-nonconformity, or both.25 But once some courts decide that a plaintiff’s sexual orientation was in play, they often altogether refuse to allow the plaintiff to frame his or her complaints in terms of gender stereotyping. Although such plaintiffs may possess many gender-nonconforming traits, these courts generally conclude that the plaintiffs are being discriminated against primarily because employers have inferred, or been made aware of, the plaintiffs’ bi/homosexuality.26

Under this approach, the idea that a court can interpret Title VII liberally, à la Price Waterhouse,to cover gender-nonconforming plaintiffs, but cannot amend it to cover bi/homosexual plaintiffs, requires an analytical leap: The court has to conclude, despite the plaintiff’s possessing many gender-nonconforming traits distinct from his or her sexuality, that the discriminator was primarily concerned with the plaintiff’s sexuality. The court could reach such a conclusion by inferring that the nonsexual traits were merely the means by which the employer made inferences about the plaintiff’s sexuality, or that the nonsexual traits were eclipsed in salience by the plaintiff’s sexuality. Thus, in spite of the existence of nonsexual gender-nonconformity, the court leaps directly to the question of whether sexual orientation is a protected category, and concludes that it is not.

A second approach gives more serious consideration to so-called mixed-motive analysis. The court entertains the possibility that, although the plaintiff may have been known or assumed to be bi/homosexual, the employer still could have been responding, at least in substantial part, to the plaintiff’s nonsexual gender-nonconformity. In addition to its sex-stereotyping holding, Price Waterhouse also set out the first standard for mixed-motive cases, in which a defendant is shown to have relied on both permissible reasons (for example, job performance) and impermissible reasons (for example, gender stereotypes) in reaching a decision. Once a plaintiff “shows that gender played a motivating part in an employment decision,” the employer has an affirmative defense27: it can prove that “even if it had not taken gender into account,” its employment decisions would have been the same.28

The Civil Rights Act of 1991 then statutorily adopted mixed-motiveanalysis and clarified its scope. It allows plaintiffs to use any and all evidence, including indirect or circumstantial evidence, to show that gender stereotypes were in play in an employer’s decision-making process.29 Once the plaintiff shows that gender was a “motivating factor,” the employer has already committed an unlawful employment practice.30 If the employer successfully establishes the “same decision” affirmative defense, then the plaintiff cannot recover damages or an injunction mandating hiring, reinstatement, or promotion.31 The plaintiff can still, however, recover declarative relief, attorney’s fees, and other forms of injunctive relief.32

In light of mixed-motive analysis, some courts have reasoned that the mere existence of a defendant’s prejudice against bi/homosexuality should not rule out the possibility of concomitant gender stereotyping. In Prowel v. Wise Business Forms, Inc., the Third Circuit explained:

To be sure, the District Court correctly noted that the record is replete with evidence of harassment motivated by Prowel’s sexual orientation. Thus, it is possible that the harassment Prowel alleges was because of his sexual orientation, not his effeminacy. Nevertheless, this does not vitiate the possibility that Prowel was also harassed for his failure to conform to gender stereotypes.33

Of particular interest to the court was the fact that Brian Prowel had detailed many examples of “effeminate” behaviors, such as crossing his legs, filing his nails, speaking with a high-pitched voice, and having an interest in design, all in addition to the fact that his coworkers knew him to be gay.34 It was therefore plausible that, even after subtracting out his employer’s knowledge (or assumptions) about his sexual orientation, Prowel still would have been discriminated against.

In Centola v. Potter, seven years earlier, Judge Nancy Gertner of the United States District Court for the District of Massachusetts had already articulated a much more emphatic version of the same basic idea.35 Judge Gertner’s opinion stressed that evidence of prejudice against bi/homosexuality should in no way serve to obscure or nullify evidence of concomitant gender stereotyping. Acknowledging that the distinction between gender stereotyping and prejudice against bi/homosexuality can often be blurry, Judge Gertner reasoned that bi/homosexual plaintiffs should nonetheless be able to present facts that combine both sources of prejudice.36 As long as the defendants’ discrimination was based on gender stereotypes, she concluded, the fact that those prejudices coexisted with other lawful (that is, pure anti-bi/homosexual) prejudice was irrelevant.37

The idea here is that as long as a plaintiff can show that her failure to conform to gender stereotypes was a “motivating factor,” she can invoke Title VII. Under this theory, the court should simply bracket the matter of the plaintiff’s bi/homosexuality, considering it neither to support nor to undermine the plaintiff’s case. The defendant’s additional prejudice against bi/homosexuality may then be relevant to the employer’s partial affirmative defense: “Yes, I had gender stereotypes on the mind, but I would have made the same decision anyway because I knew the plaintiff was gay and did not like that.”

B. The Dilemma and Penetrative Preferences: Distinguishing Dimensions of Sexuality

In this Section, we argue that the Price Waterhouse doctrine ought to be expanded to recognize the potential for gender stereotyping of sexualtraits that are nonetheless distinct from sexual orientation. The upshot of opinions like Prowel and Centolais that drawing the plaintiff’s sexual traits into the calculation does not necessarily bring a Price Waterhouse claim to a dead end. But even under Prowel and Centola, the only escape from the dead end entails focusing only on those of the plaintiff’s traits that are gendered, but have nothing to do with the plaintiff’s sexuality. That is, a plaintiff has to argue that his or her employer was motivated by some gendered trait (e.g., manner of dress, pitch of voice) that was completely separate from the plaintiff’s sex life.

The Prowel-Centola approach may indeed direct Title VII’s protections to the most marginalized plaintiffs. Social psychology studies have found, unsurprisingly, that respondents generally have more positive views of gender-conforming (“straight-acting”) homosexuals than of gender-nonconforming homosexuals.38 The problem is that many gay men and lesbian women cannot put forward a discrimination claim based on their nonsexual gender-nonconformity because their gender violations, aside from the fact of their bi/homosexuality, may be comparatively subtle—that is, they are relatively “straight-acting.” But these men and women still face prejudice. The evidence is mixed, but at least among certain populations of respondents, aversive attitudes towards bi/homosexual people persist independent of any “extra” gender violations.39 Unfortunately, courts tend to leave these plaintiffs categorically without recourse under federal law. And although many states have adopted laws against sexual-orientation discrimination, many others have not.40

We want to suggest that the current understanding of the Price Waterhouse dilemma overstates the deference that is owed to Congress’s rejection of ENDA, even assuming that the “rejected proposal” rule is persuasive. The current understanding conflates the entire realm of sexuality with the sex of one’s partners. Both early and recent versions of ENDA have defined “sexual orientation” as “homosexuality, heterosexuality, or bisexuality,” and nothing more.41 Dictionaries vary slightly in their definitions, but Merriam-Webster, for example, defines “homosexuality” as “erotic activity with another of the same sex.”42 What Congress seems to have rejected, therefore, is the categorical protection of a class of individuals defined by whom they have sex with or have sexual desire for. Nowhere has Congress rejected an interpretation of Title VII that concerns how individuals have sex.

Courts should therefore have less hesitation about taking account of other dimensions of sexuality when they are (a) distinct from the sex of one’s desired or actual sexual partners, but still (b) gendered in some way and therefore within the reach of Price Waterhouse, and (c) plausibly the source of discrimination. We believe that penetrative preferences may be just such a dimension based on the kind of pejorative language seen in discrimination cases up to now (for example, about employees who “take it up the ass”).43 And, as we will explain, our empirical study supports that conclusion. Moreover, because of mixed-motive analysis, courts could take account of gender stereotyping of penetrative preferences, even though in the vast majority of cases employers would probably also be motivated by general prejudice against bi/homosexuality.

It is appropriate for courts to distinguish between discrimination based on sexual orientation and discrimination based on penetrative preference because there are real differences that make penetrative preference something more than just a proxy for bi/homosexuality. First, any particular penetrative preference is a category that would include members of all sexual orientations, including heterosexual. Although penetrative preferences that might be perceived as gender-nonconforming are probably more prevalent among bi/homosexual populations, nothing in principle prevents an employer from taking issue with the gender-nonconformity of heterosexual employees’ sex lives. Especially in the age of modern sexual equipment, nothing prevents heterosexuals from reversing conventional penetrative roles, and heterosexual couples are increasingly doing so.44

Second, drawing penetrative preference under the Price Waterhouse sex-stereotyping umbrella would not categorically extend protection to all bi/homosexual employees. There is diversity among gay men’s penetrative preferences,45 as well as their perceived penetrative preferences.46 Plaintiffs would have to show that gendered stereotypes about their particular penetrative preference (real or perceived) actually “motivated” their employers, and this will often be difficult to do. To begin with, in a world where many heterosexuals perhaps never need to contemplate penetrative roles at all, some employers may altogether lack ideation about penetrative preferences. Additionally, a given employer may not be prejudiced against the particular penetrative preference that it believes the employee in question holds. For example, an employer could be prejudiced against men who are the more sexually “female” partner in a way that would only reach bi/homosexual employees perceived to be bottoms.

With this understanding, the deference owed to congressional inaction is quite modest in scope: a court should not stop an employer from merely taking issue with the sex of a plaintiff’s romantic partners (or the targets of the plaintiff’s desires) because Congress has had opportunities to endorse that possibility but has chosen not to do so. Under Price Waterhouse, however, a court still may stop an employer from using gender norms about how people should have sex (what it means to “act like a (wo)man” in one’s sex life) as a basis for punishing its employees. This approach to the Price Waterhouse dilemma would allow courts to extend limited protection to plaintiffs who lack prominent nonsexual gender-nonconformity but are nonetheless targets of discrimination.

C. The Dilemma and Penetrative Preferences: Closer Examination of the Source of Prejudice Against Bi/Homosexuality

Our study of penetrative preferences will also shed light on a second, and broader, possibility: expanding the scope of Price Waterhouse to cover bi/homosexuality itself—that is, the sex of one’s actual or desired sexual partners—in some cases, while still stopping short of a categorical amendment to Title VII. In addition to what we advocated in Section I.B, we support the following idea: just as a court can separate prejudice against bi/homosexuality (i.e., discriminating against gay men as such) from prejudice against nonsexual gendered traits (e.g., discriminating against men with high-pitched voices), a court might also separate gender-motivated prejudice against bi/homosexuality (e.g., discriminating against gay men because “real” men are only attracted to women) from gender-indifferent prejudice against bi/homosexuality (e.g., discriminating against gay men on the presumption that they are promiscuous or psychologically troubled). Under this framing of Price Waterhouse, if a defendant’s prejudice against bi/homosexuality is gender-motivated, it should be impermissible, but if it is gender-indifferent, it should be permissible. This is indeed the same approach that Vicki Schultz has advocated for determining whether same-sex sexual harassment is or is not based on gender violations.47

The Price Waterhouse dilemma arises in part because of how prejudice against bi/homosexuality is conceptualized. One understanding of whether or not prejudice is gendered is strictly formal: if the prejudice could be made to disappear by switching the target’s sex, then it is gendered. In general, homosexuality fits that mold (though bisexuality is more complicated). Therefore, if a court adopts that formal understanding, any form of prejudice against homosexuality appears gendered. Framing the issue in this way understandably heightens the anxiety caused by the Price Waterhouse dilemma. It seems to present an all-or-nothing choice: either categorically limit the reach of Price Waterhouse or else categorically defy Congress.Ironically, scholarship advocating recognition of the conceptual connection between sexual-orientation discrimination and gender stereotyping48 may actually contribute to this anxiety.

But we must not forget that liability under Price Waterhouse, as with any disparate treatment theory of liability under Title VII, relies upon inferences about the discriminator’s actual, subjective state of mind.49As a theoretical matter, homosexuality is characterized by its defiance of the stereotype that people should have only opposite-sex partners. It does not follow from that, however, that gender stereotypes necessarily attained salience in the mind of any particular person accused of discrimination enough to have “played a motivating part in an employment decision.”50 Prejudice against a group that is defined by its defiance of a stereotype should be legally distinguished from prejudice against a group because of its defiance of that stereotype.

The law of “bona fide occupational qualifications” (BFOQs) under Title VII51 involves a similar distinction. We permit discrimination against women, as a group, when the employer’s reasons for discriminating are “reasonably necessary to the normal operation of that particular business or enterprise.”52Thus, even where discrimination against the targeted group is generally impermissible, the action is nonetheless legal where there is a permissible reason for discrimination. Because homosexuality is a category defined by sex, all forms of prejudice against homosexuality will rely on sex to define their target. However, not all forms of prejudice against homosexuality will arise because the discriminator wishes to punish gender-nonconformity. And because Price Waterhouse’s protections extend only to gender-nonconformity, other reasons to discriminate against homosexuality, though perhaps morally repugnant, are, like BFOQs, legal.

Employers who punish bi/homosexuality are not necessarily punishing gender-nonconformity. For example, they might instead be expressing their devotion to religious teachings.53 The AIDS panic contributes additional very plausible possibilities: fear of perceived uncleanliness or condemnation of perceived promiscuity.54 With other sources of prejudice against bi/homosexuality on the table, a court need not simply assume in all cases that the hostile sentiment has arisen from stereotypes about the sex of the people that men and women should be attracted to. Instead, a court may engage in the same sorts of credibility determinations that disparate-treatment law has always required: judging whether defendants are being honest when they claim that, although race or sex could have motivated their decisions, it did not.

This reformulation of the issues would address the Price Waterhouse dilemma because judges could simultaneously vindicate the values of Price Waterhouse and avoid making a categorical, quasi-legislative pronouncement on the legal status of sexual-orientation discrimination. Sexual-orientation discrimination would only receive partial protection, conditional on showing that an employer’s prejudice against bi/homosexuality was “motivated” at least in part by gender stereotypes. The employer could then either dispute that evidence altogether or raise the partial affirmative defense that it would have made the same decision due to other lawful factors, including gender-indifferent sources of prejudice against bi/homosexuality.

One rejoinder is that purportedly gender-indifferent sources of prejudice against bi/homosexuality only arise as an indirect way of expressing disapproval of gender violations. Indeed, there are deep questions about precisely what mental state should be labeled as “motivation,” to which we can provide no easy answer.55 Worse still, in some cases, purportedly gender-indifferent reasons to discriminate may be nothing more than a pretext like any other—a knowing attempt to conceal motives and deceive the court. Employers keen on continuing to discriminate against bi/homosexual employees could potentially evade liability by ensuring that any workplace expressions of prejudice against bi/homosexuality make reference principally to gender-indifferent sources of prejudice.

But these difficulties are common to any area of antidiscrimination law, which must operate in a world in which thought processes are shrouded, often deliberately. Courts can decide for themselves precisely what qualifies as “motivation,” and plaintiffs still have an opportunity to convince courts that the proffered gender-indifferent basis for the prejudice was pretextual.56 Additionally, incentivizing employers to eliminate gender-motivated pejorative language from the workplace would seem to serve the broad gender-equality goals invoked by Price Waterhouse,57 even when it leaves a particular bi/homosexual plaintiff without legal recourse. Indeed, that fact highlights how this proposal is different in kind from a proposal to simply add sexual-orientation discrimination to Title VII by judicial decree.

The dismissal and summary judgment phases of litigation are where the distinction between gender-motivated and gender-indifferent prejudice against bi/homosexuality could have its greatest practical impact. A motion to dismiss for failure to state a claim would require a court to decide whether it is facially plausible that an employer’s prejudice against an employee’s homosexuality was gender-motivated.58 Later in litigation, a summary judgment motion would require a court to decide whether there is a “genuine issue of material fact” as to whether gender stereotypes motivated the employer’s prejudice against the plaintiff’s bi/homosexuality.59 In many cases, the specific facts and circumstances brought out in the pleadings and during discovery may answer these questions decisively. But where the facts provide little guidance, the court might want to know how plausible gender-motivated prejudice against bi/homosexuality is in the abstract, relative to gender-indifferent prejudice against bi/homosexuality.

Here, social science could influence the outcomes. Social science could inform judges’ own personal intuitions about the likelihood that prejudice against bi/homosexuality—that is, prejudice against the fact of bi/homosexuality itself—is gender-motivated. Moreover, if plaintiffs could introduce social science evidence in the summary judgment process, they could perhaps shift those intuitions.60 If that happened, a greater proportion of cases could survive summary judgment and progress towards trial. Knowing that, and often wishing to avoid the expense and exposure of a trial, employers would be inclined to avoid discriminating against bi/homosexual employees for any reason.

Social science studies offer a method of exploring and demonstrating the abstract plausibility of gender-motivated prejudice against bi/homosexuality. Current empirical evidence is suggestive of gender-motivated prejudice against homosexuality, but not conclusive on the matter. We already know that a certain baseline level of prejudice exists towards otherwise gender-conforming homosexuals.61 And there is evidence that people are perceived as less gender- conforming when they are labeled as homosexual, compared to when they are not labeled as homosexual.62

What we do not know is whether the latter (gendered judgments about homosexuality in isolation) causes the former (prejudice towards homosexuality in isolation). That is, if we ignore other markers of gender-nonconformity (e.g., high-pitched voice) and look just at the mere fact of a person’s homosexuality, is there still prejudice that is motivated by the idea that homosexuality is gender-nonconforming? Alternatively, if we described a fictional man as gay or bisexual, but changed nothing else about him, would other people like him less than they would have otherwise because of what the additional information suggests about his gender-conformity? Ordinarily, we could test this question by lowering the fictional man’s gender-conformity and then measuring the change in people’s attitudes. If people liked him less (or more) as a result, we could infer that there was a causal relationship between his gender-conformity and how well he was liked.

For our purposes, the ideal case would be to manipulate the masculinity/femininity of bi/homosexuality itself. It is hard to imagine what this would look like, however, other than presenting different fictional characters, for example, at various points along the Kinsey Scale.63 The difficulty with such a setup is that the characters might be perceived as belonging to entirely different communities, rather than merely differing in the masculinity/femininity of their sexual attractions. For example, rather than drawing on gender stereotypes, we might end up drawing increasingly on stereotypes about and prejudice towards bisexuals, particularly the perception of bisexuals as promiscuous and untrustworthy or confused.64

One alternative is simply to manipulate nonsexual traits: a person’s voice, mannerisms, personality, etc. As we have seen, it is already fairly clear that manipulating these nonsexual markers of gender-conformity does indeed change people’s opinions of gay people.65 Then, knowing that people’s prejudice responds to nonsexual markers of gender-conformity, one might infer that prejudice also must respond to sexual markers of gender-conformity. But the rival theory would be that sexuality is different, that the sexual and the nonsexual might be conceptually distinct in people’s minds. People who readily create and police a mental gender spectrum for nonsexual behaviors might fail to do so for sexual behaviors because, for example, sexuality exists in such a private sphere.

Our study of attitudes towards penetrative preferences, which are of course sexual descriptors, has allowed us to draw a closer analogy to attitudes towards bi/homosexuality. This, in turn, has allowed us to make some progress on our fundamental question of the extent to which prejudice against bi/homosexuality responds to changes in perceived gender. By making comparisons only among attitudes towards each ofthe three penetrative preferences, we have been able to control for the baseline amount of prejudice against bi/homosexuality that might be attributable to other, gender-indifferent factors. To the extent that each penetrative preference is viewed as more or less gender-conforming, attitudinal differences brought about by each preference would then suggest something about the general population’s tendency to judge sexuality according to its gender-conformity. In other words, if we can make people dislike gay characters more by adding a dimension of sexuality that is perceived as gender-nonconforming, we need not reach far to infer that these people are also basing their dislike of bi/homosexuality, at least in part, on its gender-nonconformity.

This preliminary study of attitudes towards penetrative preferences was not designed in a way that can definitively demonstrate that prejudice towards sexuality is gender-motivated. One recent small-scale study found that respondents were able to guess gay men’s penetrative preferences from facial clues, and that the effect was mediated through the respondents’ judgments of the faces’ masculinity.66 But we have not probed the thought processes of respondents in order to uncover the actual subjective basis for their manifested prejudice. Indeed, future research into public attitudes towards gender and sexuality should endeavor to do so. What we can say, however, is that our results were suggestive of the idea that people do evaluate an individual’s sexuality based on how gender-conforming it seems, beyond looking at the biological sex of that individual’s partner. In addition to aspects of our results that support such a conclusion, which we will discuss, the cultural history of penetrative preferences themselves is suggestive of that idea.

Unlike sexual orientation categories, penetrative preferences do not seem to have acquired crystallized public narratives that could introduce strong confounding associations like those related to bisexuality. Despite the seeming cultural fascination with penetration, there is no evidence that heterosexuals understand people with different penetrative preferences as comprising separate subcommunities with distinct identities. Although gay men adopt these labels as fairly concrete identities67 and regularly communicate them to potential romantic matches,68 there is little scholarship on this topic, and none that we could find dealing with how these preferences are viewed by outsiders to the LGBT community.69 Within the popular media, we are unaware of significant mainstream discussion of penetrative preference identities outside of the LGBT community. Moreover, heterosexuals presently have little reason or opportunity to discuss penetrative preferences as identities distinct from sexual orientation, because heterosexuality carries with it a strong presumption that the male will be the penetrator.

Gender associations with penetrative preferences, on the other hand, could be formed on the spot and thereby impact people’s impressions. Penetrative preferences are readily gendered. At various points in history, a man’s perceived masculinity has been tied to whether he penetrated or was penetrated.70 Certainly within modern gay communities, bottom is perceived as the most “feminine” penetrative preference and top as the most “masculine.”71 Moreover, on a conceptual level, tops fit expectations of the masculine gender better than bottoms. To the extent that the archetypal male is heterosexual, he is also exclusively—or at least primarily—a penetrator, if only because of the cultural importance placed on the complementarity of male and female anatomy.

We do not wish to make assumptions about the appropriate gender classification of any particular penetrative preference and thereby to reify the very stereotypes that we are seeking to uncover. Rather, our project is to theorize the ways in which gender norms could possibly be influencing society’s views towards sexuality. One possibility is that gender norms might lead respondents to favor gay men who are penetrators, in accordance with stereotypes about the roles that men must play. There is, however, also reason to believe that bottoms might be favored because they would fulfill a stereotype that people who are attracted to men—whether women or men themselves—enjoy being penetrated. Finally, we could imagine a scenario in which versatile men would be most disfavored, as they occupy an intermediate gender status, much like bisexual and transgender people.72 Any of these results, however, would be consistent with a world in which gender-conformity affects the degree of aversion to particular sexual practices.

We have also been able to explore whether there are variables that can explain why prejudice against certain penetrative preferences is more prevalent in some circumstances than in others. If such variation exists, it would accord with a world in which some cases of prejudice against bi/homosexuality are gender-motivated to a greater degree than others. We have looked at two possible sources of variation: demographics and setting. On the basis of these data, we will consider whether certain subgroups of respondents are more or less likely than others, on average, to display gender-motivated prejudice. And we will consider whether the larger factual setting in which the penetrative preferences are presented changes the average person’s tendency to be prejudiced against a certain preference.

II. a preliminary study of attitudes towards penetrative preferences

A. Methods

Our study explored public attitudes through the use of online survey forms with an experimental design. Respondents were recruited through the eLab at the Yale School of Management, an online referral service that advertises the potential for participants to win small prizes for participation in various external surveys. Participants in our study (n = 949) were directed to an online survey hosted by Qualtrics, a data collection service, and told they would have a one-in-twenty-five chance of winning a twenty-five dollar Amazon.com gift card.

Once they entered the survey, respondents were asked to respond to a series of questions concerning a character named Tom, as well as a series of questions concerning characters named Laurie and Ron.73 Subjects were presented at random with one of four descriptions of Tom: the vignette either said he was a top, said he was a bottom, or said he was versatile—and described what those labels meant—or else left his penetrative preference unspecified. Comparisons made among the four versions of the Tom vignette are therefore made among four randomly assigned pools of respondents. Similarly, subjects were presented at random with one of four different versions of Laurie and Ron: Ron as a penetrator with “masculine” interests, Ron as penetrator with “feminine” interests, Laurie as penetrator with “feminine” interests, and Laurie as penetrator with “masculine” interests.

We should clarify that, in describing these characters or situations as more “feminine” or “masculine,” we are referring only to how they maybe seen by respondents. In relying on stereotypically female or male descriptors, we sought only to report on the potential ways in which these descriptors could impact respondents’ attitudes by calling to mind potential cultural associations. We do not endorse these associations ourselves or believe that any particular set of interests should be viewed as masculine or feminine.

In all versions of the Tom vignette, Tom was described in the following way:

Tom is a doctor living and working in a moderately sized city on the East Coast. He was raised in the South and still loves to barbecue when he gets the chance. He keeps his apartment as clean as possible, especially when he invites his friends over for cocktails. Tom is gay and single. Sometimes he’ll invite someone he’s dating home, and if things go well, they’ll end up having anal sex.

The purpose in describing Tom this way was to mix what could have been seen by respondents as “masculine” or rugged traits (being a doctor, being from the South, enjoying barbecuing) with what they might see as “feminine” or refined traits (living in a large East Coast city, being tidy, enjoying cocktails), so that respondents would be conflicted about how to characterize Tom’s gender. Thus, when we mentioned his penetrative preference, we might be able to tip the scales one way or the other.

The three versions in which Tom’s penetrative preference was mentioned differed only with regard to one of the following final sentences: (1) “Tom considers himself a ‘top’ in the bedroom, which means he really likes to penetrate the other guy anally, but he doesn’t enjoy being penetrated by the other guy.” (2) “Tom considers himself ‘versatile’ in the bedroom, which means he equally enjoys penetrating the other guy anally and being penetrated by the other guy.” (3) “Tom considers himself a ‘bottom’ in the bedroom, which means he really likes being penetrated anally by the other guy, but he doesn’t enjoy penetrating the other guy.”

For each of the four versions of the Tom vignette, we asked respondents the following questions, in random order: (1) “If Tom invited you to go to a barbecue festival with him, would you go?” (2) “If Tom invited you over for cocktails with some of your mutual friends, would you go?” (3) “Do you think Tom would ordinarily ‘pass’ as a straight man in a social situation?” There were six possible responses: Unlikely, At Least Somewhat Unlikely, Only Slightly Unlikely, Only Slightly Likely, At Least Somewhat Likely, and Likely.

The first two questions were selected as proxies for prejudice, in place of a more direct question about how much respondents liked Tom. We did this in order to mitigate the effect of putting respondents on the spot about potentially controversial judgments, and thereby to avoid social-desirability bias.74 With the third question, we sought to measure the respondents’ tendency to engage in stereotyping of gay men, even if those stereotypes did not translate into prejudice in the first two questions.

The purpose in asking about both the barbecue festival and the cocktail party was to see whether respondents’ aversion to a particular penetrative preference might be muted or magnified by the setting in which they were asked to accompany Tom. One possibility was that respondents might view being invited to someone’s house for cocktails as more intimate and “feminine” than being asked to a public event with traditionally rugged, “masculine” associations, which might make certain penetrative preferences more or less threatening to particular respondents.75

For the Laurie and Ron vignette, the description was as follows:

<Laurie/Ron> is an investment banker at a large firm in New York City. <Her husband/His wife> <Ron/Laurie> is a nurse in the suburb where they live. <Laurie/Ron> is a huge Yankees fan but seldom has time to go see a game. It’s a good thing <she/he> married <Ron/Laurie>, because <Laurie/Ron> has no idea how to keep up a household. They have a pretty conventional marriage, but sometimes <Laurie/Ron> likes to use sex toys to stimulate and penetrate <Ron/Laurie> anally, which <Ron/Laurie> also enjoys. They haven’t decided yet whether or not to have kids, but they love each other very much.

The four versions differed with respect to the marked fields. In two versions, the interests-related fields had Laurie as the investment banker who enjoys sports and is married to a nurse who keeps up the household; the other two had Ron in that role. Then, for each of those two versions, the sex-related fields had either Laurie or Ron as the penetrator and the other person as the penetrated.

For each of the four versions of the Laurie and Ron vignette, we asked respondents the following questions, in random order: (1) “Are Ron and Laurie a couple you can see yourself being friends with?” (2) “How likely do you think it is that Ron has occasional homosexual desires?” (3) “How likely do you think it is that Laurie has occasional homosexual desires?” There were six possible responses: Unlikely, At Least Somewhat Unlikely, Only Slightly Unlikely, Only Slightly Likely, At Least Somewhat Likely, and Likely.

After they completed the questions concerning these vignettes, respondents continued to a demographic survey of sex, age, sexual orientation, gender identity, race, Hispanic ethnicity, time spent in the United States, and highest level of educational attainment. Finally, respondents were asked to estimate what proportion of gay men are tops, bottoms, or versatiles.

B. Results

Our sample of eLab respondents, described in Table 1, was fairly diverse but not entirely representative of the broader population. After excluding the 121 respondents who identified as LGBT, we were left with 828 respondents. Of those, 523 were female and 305 were male. The average respondent’s age was 37.3 years, with a standard deviation of 15.3. Among respondents, 6% identified as Hispanic or Latino and 24% identified as having a non-white racial background. Ninety percent of respondents said they had spent the majority of their lives in the United States. For educational attainment, 10% had never attended college, 31% had attended some college, 29% were college graduates, and 31% had attended at least some graduate school. Particularly in light of the high level of educational attainment, we suspect that our results might be biased in favor of greater tolerance for gender violations.

Using a two-sided Mann-Whitney test for statistical significance, we made pair-wise comparisons of answers among the different versions of each vignette.76 We compared unspecified-Tom to top-Tom, bottom-Tom, and versatile-Tom across all three questions. We also compared top-, bottom-, and versatile-Tom to one another. Then we compared all versions of the Laurie and Ron vignette to one another across all three questions. Full p-values for these tests are listed in Tables 2-7. One, two, or three asterisks following the p-values denote where we found, respectively, weak (α = 0.10), ordinary (α = 0.05), or strong (α = 0.01) statistical significance.

Regarding respondents’ willingness to go to a barbecue festival with Tom, presented in Table 2, we saw only one strongly significant effect: people were significantly less likely to go to a barbecue festival with Tom when Tom was described as versatile, as compared to when Tom’s penetrative preference was unspecified (p = 0.003). But there was also a weakly significant tendency to disfavor bottom-Tom in comparison to unspecified-Tom (p = 0.077), as well as a weakly significant tendency to favor top-Tom over versatile-Tom (p = 0.066). Comparing the means for each response (computed for illustrative, but not inferential, purposes),77 Table 2 also shows an overall pattern of feeling most comfortable with unspecified-Tom, followed by top-Tom, then bottom-Tom, then versatile-Tom.

Results for the cocktail party question, reported in Table 3, also display the same ordinal ranking of means, with respondents most likely to attend with unspecified-Tom, followed by top-Tom, then bottom-Tom, then versatile-Tom. However, the gap between unspecified-Tom and all the other Toms widened. Additionally, respondents were significantly more likely to go to the cocktail party with unspecified-Tom than with versatile-Tom (p = 0.001) or bottom-Tom (p = 0.008). Most notably, we even saw a statistically significant difference between versatile-Tom and top-Tom (p = 0.046).

Finally, Table 4 reports on the ability of the various Toms to pass as straight. Again, we saw the same ordinal ranking of means, with the unspecified-Tom being most likely to pass, followed in descending likelihood by top-Tom, bottom-Tom, and versatile-Tom. Top-Tom, bottom-Tom, and versatile-Tom were all significantly less likely to be perceived as “passable” than unspecified-Tom, but the differentials between the three specified groups were smaller and not statistically significant. Although mentioning penetrative preference amplified how “gay” Tom was perceived to be, no particular role in anal sex was considered “gayer” than another.

Next, we considered whether subgroups of our study population differed in their prejudices towards specific versions of Tom. Based on earlier studies that have shown men and women to react differently to homosexuality and gender-nonconformity,78 we decided to control for respondents’ biological sex. Second, based on surveys showing generational shifts in attitudes towards homosexuality, we also sorted our study population into two age clusters, roughly at its median: those at least thirty-three years old and those thirty-two years old or younger.79 We then performed Mann-Whitney significance tests, like those performed on the entire study population, for each of the three subgroups. Results are presented in Tables 2a-2d (for the barbecue setting) and Tables 3a-3d (for the cocktail setting).

For the barbecue setting, we saw no significant results in any of the subgroups. Three of the subgroups—younger men (Table 2a), younger women (Table 2b), and older men (Table 2c)—did, however, display the same ordinal ranking as before: top-Tom, followed by bottom-Tom, followed by versatile-Tom. Older women (Table 2d) did not, instead favoring bottom-Tom ahead of top-Tom, though not at a statistically significant level. Although Tables 2a-2c show a fairly large gap between top-Tom and versatile-Tom, replicating what we saw in the overall study population, the difference was not statistically significant—perhaps because of the much smaller sample size.

For the cocktail setting, we did see significant results. Younger men (Table 3a) and younger women (Table 3b) again favored top-Tom over versatile-Tom. Like those two groups, older men (Table 3c) also had the highest mean ranking for top-Tom, followed by bottom-Tom, and then versatile-Tom. Because of the distribution of the responses, however, the Mann-Whitney test registered statistical significance for the top-bottom difference but not the top-versatile difference.80 Older women (Table 3d), by contrast, showed no statistically significant tendency to favor one penetrative preference more than the others and did not follow the same ordering of the means seen in all three of the other groups.

Overall, the Tom vignette points to prejudice against certain penetrative preferences. In several instances, both within the full study population and within certain subgroups, respondents favored top-Tom over versatile-Tom. Then there is the disfavoring of bottom-Tom by older men. Both of these results are bolstered by the fact that both versatile-Tom and bottom-Tom, but not top-Tom, elicited a significantly less favorablereaction than the baseline set by unspecified-Tom. Additionally, there is the overall trend in computed means, placing top-Tom first, versatile-Tom last, and bottom-Tom in the middle, which is of no statistical significance in isolation but which was very persistent across nearly all the comparisons we made.

The results also seem to suggest that prejudice against certain penetrative preferences is variable. It is true that changing the setting did not change the result within the entire study population: a tendency to favor top-Tom over versatile-Tom. But that tendency was stronger in the cocktail party setting. Moreover, after we separated the population into subgroups, we found several significant biases among the subgroups on the cocktail party question, whereas we found no biases with the barbecue question. Additionally, we found real differences among the subgroups in the cocktail party setting. In particular, we found that older women did not display the tendency (found among younger men and women and older men) to favor one penetrative preference over another.

We should emphasize that this finding does not mean that older women are less likely to discriminate against homosexuality, nor does it mean that younger women are equally likely to discriminate against homosexuality as younger men. Indeed, Tables 2a-2d and 3a-3d show a higher average rating among younger women than among younger men across all versions of Tom. The relative ratings of each version of Tom, however, suggest that young women nonetheless tended to have gender on the mind when evaluating sexual practices. We therefore might reasonably expect that, among young women who do have an overall negative view of homosexuality, some of that negativity arises from gender norms. The results among older women, meanwhile, show that certain contextual factors, such as the age of the discriminator, can make that less likely.

Regarding Laurie and Ron, we did see some tendency in the data to favor as friends the versions of the couple that had Ron as the penetrator (Table 5). That tendency was not, however, statistically significant. That is not especially surprising, given that the question (“Are Ron and Laurie a couple you can see yourself being friends with?”) was more direct than the questions we asked about Tom’s likability, and therefore more likely to bring about social-desirability bias. Also, the fact that respondents were presented with a question about Laurie and Ron’s joint likability as a couple differentiates this scenario further from the Tom vignette. Accordingly, the Laurie and Ron results are not substantially in conflict with the Tom results.

Particularly remarkable were the questions about Laurie and Ron’s latent homosexuality (Tables 6-7). Respondents were overwhelmingly more likely to think that both Ron (Table 6) and Laurie (Table 7) had “occasional homosexual desires” when Laurie was the penetrator. Every single pair-wise comparison between Ron-as-penetrator and Laurie-as-penetrator had a strongly significant p-value (p < 0.01) for both questions.

First of all, these results rule out the possibility that penetrative preferences are only culturally relevant in the typical homosexual context. Although the effect on actual prejudice was not statistically significant when measured by the first question, the latter questions show that penetrative preferences had an effect on our respondents’ perceptions even in the context of a heterosexual relationship. Laurie and Ron were described as a loving couple contemplating whether to have children, but respondents were persuaded that both Laurie and Ron were more likely to have homosexual desires merely because of the roles they played in bed. By comparison, respondents overall had almost no response to the differences in Ron and Laurie’s professions and interests.

These results do not directly demonstrate that penetrative preferences were gendered in our respondents’ minds, but they are suggestive of that conclusion. On a formal level—bracketing for a moment the question of respondents’ actual beliefs—these results do show that penetrative preferences are gendered to some extent. They are formally gendered because there is an interaction taking place between penetrative preferences and biological sex. Mentioning anal sex and sex toys might have had some baseline effect, but there was an additional effect that occurred only when the woman was the penetrator (or the man was the penetrated). This disparity strongly suggests that there was an expectation that the man would penetrate and the woman would be penetrated.

Saying that some cultural expectation was disrupted does not, however, tell us the precise ways in which the respondents’ views of Laurie and Ron changed when that expectation was disrupted. If we conducted this study again with the benefit of hindsight, we would ask direct questions about gender-conformity. At least in theory, respondents could have thought that a penetrating woman and penetrated man were just “gayer” than their counterparts with reversed roles, without actually believing that one possibility was more gender-nonconforming per se. But homosexuality and gender are culturally associated: people tend to think that gender-nonconforming people are gay81 and that gay people are gender-nonconforming.82 It therefore seems likely that, in associating Laurie-as-penetrator with homosexuality, respondents also associated it with some degree of gender-nonconformity.

Our final survey questions asked respondents to estimate the percentages of tops, bottoms, and versatiles among gay men, presented in Table 8. The average estimates for tops, bottoms, and versatiles were 31.5%, 27.9%, and 40.6%, respectively. Thus, the average heterosexual person believes that more than two-thirds of gay men are either bottoms or versatiles. And if they view both of those as gender-nonconforming preferences, then it further supports the notion that an employer might assume, even without knowing an employee’s penetrative preference, that a gay employee is “feminine” in the bedroom and therefore discriminate against him.

While the true proportion of these groups is only imprecisely known, the handful of published estimates range from 19-20% for tops, 26-35% for bottoms, and 47-54% for versatiles, among the gay male population.83 On average, therefore, our heterosexual respondents did correctly intuit that versatiles are most common, but tended to underestimate their prevalence in the population. More importantly, heterosexuals failed to perceive that there are likely to be more bottoms than tops in the population. Analyzing the proportion of respondent estimates falling outside the most credible prevalence ranges, which we have adopted as a reference point, we found that while only 9% of respondents underestimated the number of tops, a whopping 79% overestimated. For bottoms, 51% underestimated and 21% overestimated. And for versatiles, 58% underestimated and 18% overestimated.

Next, we ran an ordinary least squares regression to see whether there was a priming effect created by the vignettes themselves. For this regression, we included variables for the other non-Tom-related vignettes not discussed in this paper. Given the randomization of the vignettes, did seeing more or fewer tops/bottoms/versatiles in the vignettes make these types more salient in ways that impacted respondents’ estimates of the prevalence percentages? Table 9 provides some modest evidence of a priming effect by reporting that an increasing number of bottom vignettes caused respondents to lower their estimates for the percentage of tops.

III. discussion

A. Revisiting the Price Waterhouse Dilemma

Our preliminary study of heterosexuals’ attitudes towards penetrative preferences has allowed us to suggest two modifications to the premises underlying the current understanding of the dilemma between Price Waterhouse’s principlesand judicial deference to Congress: (1) There are real forms of gender-motivated prejudice against a person’s sexuality that are distinct from prejudice against having actual or desired partners of the same sex. (2) With regard to prejudice against the sex of someone’s partners, it is quite plausible that such prejudice is gender-motivated, but that is not a foregone conclusion. Both of these possibilities would allow courts to extend antidiscrimination protections—at least through the summary judgment phase—to a broader class of plaintiffs who are discriminated against because of their sexuality, without categorically covering homosexuality. Although our study was not designed to answer every question raised by those proposed modifications, the results do shed light on many of those questions.

What is perhaps most striking about the results is the mere fact that penetrative preference made any significant differences in attitudes whatsoever. Penetrative preference was not some irrelevant detail that added noise to the data. For both the Tom and the Laurie-and-Ron vignettes, people responded differently when the characters played a different role in anal sex. Although it would not be particularly surprising to find that mentioning anal sex colored respondents’ reactions, it is entirely different to see reactions that vary according to the particular role in anal sex being described.

Penetrative preference therefore stands as an example of a dimension of sexuality, aside from the sex of one’s partners, that might plausibly have a real effect on prejudice.84 Indeed, it might have an impact even when employers do not know gay employees’ penetrative preferences, because people may hold opinions about the general prevalence of each preference and tend to assume that gay men are sexually “feminine” unless there are indications to the contrary. Respondents’ disparate responses could not be explained by certain gender-indifferent prejudices (e.g., “My religious tenet is that all men who engage in anal sex are sinful.”), but we did not directly prove that respondents contemplated penetrative preferences in gendered terms. Still, we believe that the results are consistent with and suggestive of that conclusion. There is indeed an opportunity for further research into the actual thought processes underlying prejudice against various dimensions of sexuality.

To the extent that penetrative preferences are subjectively gendered, the tendency to favor tops and disfavor versatiles also speaks to the broader likelihood that prejudice towards bi/homosexuality itself is gender-motivated. This insight joins existing social psychology research in demonstrating a connection between prejudice and the manipulation of potentially gender-salient traits, but it does so by manipulating a distinctly sexual trait. We also saw, however, that this likely relationship between sexual gender-conformity and prejudice was not constant or inevitable. In several instances, contextual variables made prejudice against a particular penetrative preference less likely. A different setting (barbecue festival versus cocktail party) seemed to attenuate signs of prejudice. Additionally, older men had a different prejudice than younger men, and older women did not show any prejudice at all with regard to penetrative preferences.

Together, these results suggest that courts can and should interpret Price Waterhouse in a way that covers a broader range of bi/homosexual plaintiffs, including those who lack prominent outward gender-nonconformity. Courts should be open to claims about discrimination on the basis of dimensions of sexuality that are distinct from sexual orientation, such as penetrative preferences. And even with regard to sexual orientation, courts could expand Price Waterhouse to cover only those situations where prejudice against bi/homosexuality was gender-motivated. By making individualized determinations and interrogating the actual source of employers’ motivations, as courts do elsewhere in their antidiscrimination dockets, they could still avoid providing blanket coverage to all claims of sexual-orientation discrimination.

B. Further Legal Implications Arising from Versatility

That a versatile person, and not a top or bottom, was most often disfavored adds another layer to our analysis. We often think of “gender stereotyping” as an expectation that biological males will possess other male traits and biological females will possess other female traits, with little or no crossover. The basic way of violating that expectation is what we term “trait opposition”—that is, when a trait on one side of the masculine-feminine gender divide is adopted by a person thought to belong on the other side (e.g., a boy who paints his room pink).

At a higher level of abstraction, however, gender stereotypes represent an organizing societal principle around which people’s expectations of others’ behavior are built. Violating someone’s expectations tends to elicit anger or anxiety because the violator’s actions compromise the certainty of other people’s gendered behavior and identity.85 With that understanding, one’s traits need not be entirely oppositional to one’s sex in order to run afoul of gender stereotypes. Instead, a particular trait may meld elements of both genders into a single identity and defy categorization as masculine or feminine. This is what Kenji Yoshino has called an “intermediate” identity and what Ruth Colker has called a “hybrid” identity.86

Theoretically, any intermediate identity could be thought of as merely eclectic and broken down into a combination of oppositional traits and non-oppositional traits: a transgender person is not biologically intermediate; rather, some individual biological traits (e.g., genitalia) are female while others (e.g., chromosomes) are male. The question, however, is how the discriminator conceives of the trait—at what level does he or she analyze the trait? To the extent that people conceive of certain bundles of traits as collectively intermediate rather than merely eclectic, the concept of intermediacy might actually pose a more fundamental challenge than opposition, in that it challenges the very coherence of the masculine-feminine divide.87

Again, because our study did not examine subjective motivation, there is the possibility that the additional prejudice towards versatile-Tom was due entirely to some gender-indifferent source—for example, if respondents immediately attached the same promiscuity association to versatility as others have attached to bisexuality.88 However, penetrative versatility could well be disliked due to its intermediate position on the gender spectrum. It certainly has an intermediate character: we described a versatile person as someone who “equally enjoys penetrating the other guy anally and being penetrated by the other guy.” The most persuasive evidence for the existence of intermediacy prejudice comes from the shape of the results. If respondents had only been concerned with trait opposition, versatility might still have been disfavored, but not more than both top and bottom preferences. In oppositional terms, being versatile is conceptually less violative of gender norms than being a bottom because the opposition is incomplete; at most, being a bottom and being versatile would be equally disfavored. That versatility stood out in our results suggests that something other than, or in addition to, oppositional prejudice was involved.

Of course, up to now, courts interpreting Title VIIhave applied an opposition-based logic, perhaps with the notable exception of Shroer v. Billington89—that is, it is unlawful for an employer to punish an employee for possessing traits conventionally held by the opposite sex. Even transgender plaintiffs have had to work within that binary framework, being viewed, for example, as biological men with female traits, including female anatomy in some cases.90 It is uncertain, therefore, whether the current predominant understanding of gender stereotyping, with its oppositional rhetoric, would reach this potentially intermediacy-based prejudice against versatility. It may be that versatility could be protected as a partial form of gender opposition, but as with discrimination against transgender people, that view might miss the larger point that those thought to be gender-intermediates may face unique and higher levels of prejudice.

C. Public Opinion Implications

The results of our study also carry practical political implications for the contemporary LGBT rights movement, to the extent that the movement is constrained by the heterosexual majority’s knowledge, biases, and opinions. First, could greater knowledge and discussion of penetrative preferences shift public opinion towards acceptance of bi/homosexuality? And second, do the biases we detected in our study suggest anything about public opinion towards bisexual, transgender, and other subcommunities within the larger movement?

Combined with the results of the attitudinal questions, our respondents’ estimates of the percentage of each preference paint an interesting picture. As evidenced by the large variance in estimates, our respondents seemed to have fairly vague conceptions of what gay men actually do. But within these fluctuations, we found that respondents tended to underestimate the prevalence of versatiles and bottoms, sexual identities that they comparatively disfavored, and overestimated the prevalence of tops, the sexual identity that they comparatively favored. Heterosexuals’ ignorance therefore seems to allow them to carry a more gender-normative depiction of the gay community than is true, and may thereby inflate their overall opinion of the community. We also found a modest framing effect in that we could reduce respondents’ overestimation of tops by presenting them with more vignettes featuring bottoms. Perhaps, therefore, greater visibility of and openness about gay men’s sexual practices might actually lead some heterosexual observers to hold less favorable attitudes towards the community.

That the respondents found penetrative versatility to be particularly troubling also suggests that other intermediate groups such as bisexual and transgender people face deep attitudinal hurdles due to the nature of their gender violations. It suggests that even if the movement could erase decades of built-up narratives about what it means to be bisexual or transgender, these groups still might not be on equal footing with their homosexual allies. Perhaps even more troubling, these results further suggest that homosexuals may in fact possess very real incentives to distance themselves from the more gender-threatening elements of the LGBT movement.91 Indeed, some commentators have accused the LGBT establishment of doing precisely that.92 And to the extent that the LGBT movement is including bisexuality and transgender identity in its messaging, would it actually see stronger gains if it excluded them altogether?

Rather than pander to the current biases built into public perception, however, the LGBT rights movement could perhaps reshape those perceptions. One approach would be to attack the problem head-on and wage a public relations campaign directly aimed at raising the general public’s comfort level with intermediacy. An important tactic would be the ability to normalize intermediacy, to show that it is not something entirely alien to the average person’s existence. Allies of the movement could acknowledge the myriad ways in which they engage in trait opposition, but reframe this as a matter of an overall intermediate identity (e.g., “I’m not entirely a man or entirely a woman in how I dress, behave, or think,” rather than “I’m a man who happens to have a few feminine traits”).

Conclusion

Whether one is making a determination about the plausibility of gender-motivated employment discrimination or drawing up strategic plans for advancing the social status of gender-nonconforming groups, it is essential to understand how and why society favors some groups over others. In this Essay, we have argued that sexual behavior is indeed socially gendered beyond its connection to sexual orientation and sufficiently so to bring out actual prejudice, but that this is not always the case and will depend on the specific context. And by analogy, these same gender dynamics likely impact the ways in which many members of the broader LGBT community are perceived in certain situations.

Courts therefore should not categorically exempt employers’ attitudes towards sexual behavior in an effort to limit Price Waterhouse’s scope. To do so would be to disregard the possibility of blatant gender-motivated discrimination. Instead, courts should draw the line in accordance with the evidence of defendants’ actual motivations. Until Congress decides to amend Title VII, employers are free under federal law to discriminate against bi/homosexuality, but not in a way that relies upon the view that certain sexual practices make someone less of a man or less of a woman.

Our findings may also have relevance to participants in debates over LGBT rights. Many heterosexual people may be operating with a more gender-normative view of gay sexual practices than is accurate. Members of the LGBT rights movement should be aware of the possible differential effects of trait opposition and intermediacy on public opinion, and should consider whether and how that disparity ought to be addressed.

There are still many questions to be answered about what goes on in the minds of those who discriminate based on sex, gender, and sexuality. It is difficult to imagine how courts can properly conceive of what gender-motivated discrimination means without a systematic model of the ways in which gender categories shape perception, supported by a body of empirical findings. This Essay has aimed to make a small contribution to that model, in the hope that one day the values expressed in Price Waterhouse will be fully realized, both inside and outside the legal system.

Appendix

Table 1.

summary of respondent demographics

Table 2.

summary of responses to question “if tom invited you to go to a barbecue festival with him, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 2a.

among male respondents under age 33, summary of responses to question “if tom invited you to go to a barbecue festival with him, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 2b.

among female respondents under age 33, summary of responses to question “if tom invited you to go to a barbecue festival with him, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 2c.

among male respondents age 33 or older, summary of responses to question “if tom invited you to go to a barbecue festival with him, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 2d.

among female respondents age 33 or older, summary of responses to question “if tom invited you to go to a barbecue festival with him, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 3.

summary of responses to question “if tom invited you over for cocktails with some of your mutual friends, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 3a.

among male respondents under age 33, summary of responses to question “if tom invited you over for cocktails with some of your mutual friends, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 3b.

among female respondents under age 33, summary of responses to question “if tom invited you over for cocktails with some of your mutual friends, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 3c.

among male respondents age 33 or older, summary of responses to question “if tom invited you over for cocktails with some of your mutual friends, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 3d.

among female respondents age 33 or older, summary of responses to question “if tom invited you over for cocktails with some of your mutual friends, would you go?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 4.

summary of responses to question “do you think tom would ordinarily ‘pass’ as a straight man in a social situation?” and mann-whitney p-values testing for statistically significant differences in responses between versions of tom

Table 5.

summary of responses to question “are ron and laurie a couple you can see yourself being friends with?” and mann-whitney p-values testing for statistically significant differences in responses between versions of laurie and ron

Table 6.

summary of responses to question “how likely do you think it is that ron has occasional homosexual desires?” and mann-whitney p-values testing for statistically significant differences in responses between versions of laurie and ron

Table 7.

summary of responses to question “how likely do you think it is that laurie has occasional homosexual desires?” and mann-whitney p-values testing for statistically significant differences in responses between versions of laurie and ron

Table 8.

summary of responses to question “if you had to guess, what percentage of gay men do you think are ‘tops’ (only like to penetrate other guys), ‘bottoms’ (only like being penetrated by other guys), and ‘versatiles’ (equally enjoy both)?” and percentage of responses outside of “guesses of reality” range

Table 9.

ordinary least squares regression of respondents’ guesses about prevalence of each penetrative preference, as a function of how many characters of each preference they were exposed to throughout the experiment