LAY, Circuit Judge, with whom Judge John R. Gibson joins in Part IIA.

This is an appeal brought by Marcia Hocevar (Hocevar) from the district court’s grant of summary judgment in favor of Purdue Frederick Company (Purdue), her former employer, in a sexual harassment and retaliation claim brought under Title VII of the Civil Rights Act of 1964, 42 U.S.C. §§ 2000e et seq. The district court found that the plaintiff was a member of a protected class (a female) but that she failed as a matter of law to demonstrate a genuine dispute of material fact on both of her claims. In addition to denying her claim of retaliation, the court found that Hocevar had not shown a hostile work environment because it concluded that the alleged harassment was neither pervasive nor severe. For the reasons stated below, the majority of the court (Judges Beam and Gibson) affirm the grant of summary judgment on the hostile work environment claim; a different majority of the court (Judges Lay and Gibson) reverse and remand for trial on Hoce-var’s retaliation claim.

I. BACKGROUND

The record shows that Marcia Hocevar began working at Purdue in August 1988 as a pharmaceutical sales representative. While working for Purdue in Minnesota between 1988 and 1992, Hocevar consistently out performed her then co-worker Timothy Amundsen (Amundsen) and was often ranked in the top sales percentile nationally. Hocevar was promoted three times in five years, the final promotion being to the position of sales training manager at corporate headquarters in Norwich, Connecticut. Hocevar’s bonuses reflect her good sales record, and her performance was rated at the highest possible level.

In June 1994, Hocevar transferred to Minnesota due to her impending marriage where she was placed under the supervision of Amundsen, the new district manager, and took over the sales territory previ*724ously assigned to him.1 Despite Hocevar’s history of top-notch performance evaluations, Amundsen rated Hocevar at the lowest possible level in October and November 1994. Amundsen gave Hocevar an additional adverse rating in February 1995.

In March 1995, Amundsen accused Hocevar of lying and making false sales reports. A company investigation concluded no wrongdoing on Hocevar’s part. Shortly thereafter, in July 1995, Amundsen again gave Hocevar the lowest possible performance rating despite the fact that she demonstrated a sales growth of seven percent. At some point, Purdue took away a portion of Hocevar’s sales territory — an area including the world renowned Mayo Clinic and LaCrosse, Wisconsin.2 This action was taken by Amundsen despite the fact that Hocevar exceeded Amundsen’s own prior sales record in the same territory and received bonuses for exceeding sales quota. These areas remained un-staffed for three months following removal from Hocevar’s territory.

On August 11, 1995, Amundsen recommended Hocevar for probation based on her past year’s performance. Following an automobile accident, Hocevar took disability leave from August 16, 1995, until September 15, 1995. Despite her absence, Hocevar again met her sales quota and earned a bonus. Hocevar took additional disability leave on October 21, 1995, and requested a part-time work schedule accommodation. Amundsen denied her request. As a result, Hocevar was unable to return to full-time work and remained on disability leave until her termination on June 7,1996.

Following Hocevar’s return to Minnesota in 1994, Amundsen engaged in hostile behavior in the workplace over a two-year period: he distributed sexually explicit material at business meetings; he made threats of violence towards female staff members; he constantly referred to women as “bitches,” “fucking bitches,” and “fat fucking bitches,”3 he told stories of animal violence (e.g., placing a loaded gun in the mouth of a dog that wandered into his yard); he told jokes at meetings that were derogatory towards women and contained profanity; he introduced a new employee as the “fucking new guy;” and claimed that new pharmaceutical products were so exciting a physician would be “creaming his jeans” to get them. Hocevar also testified .that Amundsen exhausted a portion of a staff meeting by playing an audiotape of the Jerky Boys which contained obscene, vulgar, and sexually explicit “prank” phone calls to businesses on topics such as genital warts.

Hocevar also testified that in April 1992, Purdue Regional Manager Paul Ka-sprzycki (Kasprzycki) had made sexual advances toward her at a bi-regional meeting in Denver, Colorado. She testified that she was afraid to report complaints to Kasprzycki (Amundsen’s supervisor) due to incidents of Kasprzycki making unwelcome sexual advances towards her, including pulling her toward him resulting in “full body contact” during what began as a consensual “fast” dance that led into a “slow” dance. She testified that Ka-*725sprzycki’s made “very clear his wish to have a sexual relationship” with her and made suggestive comments about being available for a sexual relationship. Hoce-var testified that Kasprzycki’s advances were even more explicit when no witnesses were around. According to Hocevar, this was not an isolated incident, as Kasprzycki had previously made “unwelcome and uninvited” sexual advances toward her following a Purdue national meeting in New Orleans in January 1992. Then, in front of nearly 150 people Kasprzycki made statements at a bi-regional meeting in April 1995 implying a female manager had a sexual device in her hand and, in a separate incident, that he would be engaging in a sexual liaison in his hotel room later that day with three female sales representatives that had just performed a singing skit. Additionally, she describes an incident at a national meeting in Texas in 1993 involving two other Purdue District Managers, Dan Mackavoy and Dick Silverman. Hocevar stated that the district managers talked throughout her presentation; after-wards, she approached them about their “rude” behavior, to which Mackavoy responded: “We were talking about what great legs you have.”

In yet another incident, also following a Purdue bi-regional meeting, Hocevar and six male and female co-workers were discussing Susan Faludi’s book Backlash: The Undeclared War Against American Women (discussing public reaction to successful working women). During this conversation, a male employee called Hocevar a “bitch” and the then new district manager, Kelly Bartlett, became “very angry” and “exploded” stating: “You women, since when are women always right and men are always wrong? If your women’s movement had its way, every woman would be working and our children would be being raised in communes.” The incident was so upsetting that Kathy Kiekhae-fer (Kiekhaefer) and a co-worker were crying and were “scared” and concerned at the prospect of working for a manager with such a feeling of hostility toward working women.

In October 1995, Hocevar complained to Dennis Merlo, a Purdue managerial employee, about Amundsen’s inappropriate behavior, foul language, and stories of animal violence. On December 20, 1995, Hocevar’s attorney notified Purdue of her intention to file a complaint against Purdue with the Minnesota Department of Human Rights alleging sexual harassment. The letter also voiced concerns about the “ongoing sexual harassment” of Hocevar and other women at Purdue. In January 1996, another female employee, Kiekhae-fer, filed a claim of sexual harassment with Purdue, which prompted Purdue to investigate the complaints.

Danielle Nelson (Nelson), Purdue’s Vice President of Equal Employment Opportunity Compliance and Human Resources Administration, conducted an investigation into the complaints of sexual harassment. Nelson found that Amundsen’s extensive use of profanity and off-color jokes violated company policy and was “unprofessional behavior.” Nelson concluded, however, that no sexual harassment occurred. Despite Nelson’s determination that no sexual harassment occurred, Purdue directed Amundsen — under threat of termination— to take a three month unpaid leave of absence during which he would receive counseling and management training. Thereafter, Nelson and James Lang (Lang), Purdue’s National Sales Manager, traveled to Amundsen’s district, informed the employees that Amundsen’s language was inappropriate and unacceptable, and trained employees on Purdue sexual harassment complaint procedures. After the Nelson/Lang visit, Hocevar’s co-worker Mary Beck-Johnson testified that workplace conduct “absolutely changed” — “personal” matters were no longer discussed and inappropriate language was no longer used at meetings.

In mid-April 1996, Amundsen returned from the unpaid leave of absence. On May 2, 1996, Hocevar filed a charge of sexual *726harassment with the Equal Employment Opportunity Commission (EEOC) and Purdue terminated her a little over a month later by letter dated June 7, 1996. Hocevar thereafter filed her claim alleging retaliatory discharge on July 10, 1996. Hocevar now appeals the district court’s grant of summary judgment in favor of Purdue.

II. DISCUSSION

A. Retaliation

The district court found that Hocevar made a prima facie case of retaliation following her claim of sexual harassment with the EEOC. It found that (1) Hocevar had engaged in a statutorily protected activity;4 (2) an adverse employment action occurred; and (3) the adverse employment action was causally linked to the protected activity.5 Nonetheless, the district court found that Purdue articulated a nondiscriminatory reason for termination, namely, the company’s need to restaff her vacant position.6 The district court furthermore found that Hocevar had not shown evidence that Purdue’s reason was pretextual.

Based on our de novo review of the record, we reverse the grant of summary judgment on Hocevar’s retaliation claim. The overall record establishes the plaintiff has demonstrated sufficient evidence, if believed, that the reason given for her discharge simply masked the true reason for the discharge — retaliation for filing an EEOC charge. See Furnco Const. Corp. v. Waters, 438 U.S. 567, 577, 98 S.Ct. 2943, 57 L.Ed.2d 957 (1978).7 The record demonstrates several factors from which a trier of fact might infer that retaliation was the true reason for her discharge:

(1) the close proximity in time between Hocevar’s discharge and Amundsen’s return to work after his three-month suspension without pay for sexual harassment as reported by Hocevar and others;

(2) the close proximity of Hocevar’s filing the EEOC claim of sexual harassment and her discharge;

(3) that long prior to the company’s reason for Hocevar’s discharge, the Mayo Clinic account was taken away from the plaintiff and that account was unserviced for over three months thus depriving Hocevar of substantial sales commission;

(4) that Hocevar was targeted by Amundsen and Kasprzycki for preprobation in July 1995;

(5) that defendant had earlier refused to accommodate plaintiffs work restriction *727following a car accident, when defendant’s own employment expert testified that such accommodation could have occurred;

(6) that Amundsen had required Hoce-var to call him every day with a special report about her sales calls; no other employee was required to do so;

(7) that Kathy Kiekhaefer testified that employees who complained about their manager “eventually were gone from the organization altogether.”

Hocevar’s allegations, if proven true, evidence a long history of unfavorable actions by Amundsen against her virtually from the moment she was placed under his supervision. During the period in which Amundsen gave her the lowest possible performance ratings and placed her on probation, Hoeevar consistently exceeded Purdue sales quotas and received bonuses. Hocevar’s sales performance exceeded quota notwithstanding the fact that Amundsen removed a lucrative portion of her sales territory and despite the offensive work environment and heightened scrutiny by Amundsen. After Hoeevar notified Purdue of Amundsen’s offensive behavior, Amundsen was forced to take an unpaid leave of absence. Less than a month after his return, Hoeevar filed a sexual harassment claim and shortly thereafter was terminated.

Giving Hoeevar the benefit of all favorable inferences on summary judgment, we hold there exists sufficient inference that the company’s sudden need to restaff the Mayo Clinic territory was indeed questionable and that Hocevar’s filing of her complaint with the EEOC was the motivating act that caused Amundsen to discharge her.

It is not for this court, nor for the district court, to weigh the evidence and decide whether Purdue’s proffered reason was true. As long as there exists conflicting evidence upon which reasonable men and women might differ, we find sufficient evidence of pretext to survive the motion of summary judgment. Under the circumstances, we find the district court erred in granting summary judgment on Hocevar’s retaliation claim. We, therefore, reverse the grant of summary judgment on Hoce-var’s claim of retaliation for the exercise of protected activity.

LAY, Circuit Judge,

dissenting.

I dissent from the grant of summary judgment on the hostile work environment claim.

Title VII makes it unlawful for an employer “to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual’s race, color, religion, sex, or national origin.” 42 U.S.C. § 2000e-2(a)(l). Title VII protects “more than ‘terms’ and ‘conditions’ in the narrow contractual sense.” Faragher v. City of Boca Raton, 524 U.S. 775, 786, 118 S.Ct. 2275, 141 L.Ed.2d 662 (1998) (quoting Oncale v. Sundowner Offshore Servs., Inc., 523 U.S. 75, 78, 118 S.Ct. 998, 140 L.Ed.2d 201 (1998)) (internal quotation omitted). The Act evinces Congress’ intention to define discrimination in the broadest possible terms, and neither enumerates specific discriminatory practices nor defines the breadth of actionable illegal activities. See Hall v. Gus Const. Co., 842 F.2d 1010, 1014 (8th Cir.1988). In interpreting the scope of activities prohibited under Title VII, the Supreme Court instructs that hostile work environment harassment occurs when “the workplace is permeated with ‘discriminatory intimidation, ridicule, and insult’ that is ‘sufficiently severe or pervasive to alter the conditions of the victim’s employment and create an abusive working environment.’” Harris v. Forklift Systems, Inc., 510 U.S. 17, 21, 114 S.Ct. 367, 126 L.Ed.2d 295 (1993) (citations omitted and emphasis added).

In a sexual harassment suit, in order to establish a claim of hostile work environment, a plaintiff must show (1) membership in a protected group; (2) the occur*728rence of unwelcome harassment; (3) a causal nexus between the harassment and membership in the protected group; and (4) that the harassment affected a term, condition or privilege of employment. See Carter v. Chrysler Corp., 173 F.3d 693, 700 (8th Cir.1999). If the harassment is perpetrated by a supervisor and the employee suffers a tangible employment action (e.g., demotion, undesirable reassignment, or discharge), the employer is vicariously liable for the supervisor’s sexual harassment of the employee. See Faragher, 524 U.S. at 807-08, 118 S.Ct. 2275; Burlington Industries, Inc. v. Ellerth, 524 U.S. 742, 765, 118 S.Ct. 2257, 141 L.Ed.2d 633 (1998). Important to my analysis is the Supreme Court’s recent statement that sexually harassing behavior perpetrated by a supervisor has a “greater power to alter the environment” than similar actions of mere co-workers. Faragher, 524 U.S. at 805, 118 S.Ct. 2275.

The inquiry at summary judgment is “whether the evidence presents a sufficient disagreement to require submission to a jury or whether it is so one-sided that one party must prevail as a matter of law.” Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 251-52, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986). As I previously indicated, in considering a motion for summary judgment, the district court should not weigh the evidence, make credibility determinations, or attempt to determine the truth of the matter, id. at 249, 106 S.Ct. 2505, but instead should give all reasonable inferences to the non-moving party. Id. at 255, 106 S.Ct. 2505. To survive summary judgment, Hocevar need only submit “ ‘sufficient evidence supporting a material factual dispute that would require resolution by a trier of fact.’ ” Austin v. Minnesota Mining & Mfg. Co., 193 F.3d 992, 994 (8th Cir.1999), (quoting Hase v. Missouri Div. of Employment Sec., 972 F.2d 893, 895 (8th Cir.1992)). Summary judgment is inappropriate where “reasonable minds could differ as to the import of the evidence.” Anderson, 477 U.S. at 250, 106 S.Ct. 2505. It is under this framework that we should review the district court’s grant of summary judgment to Purdue on Hocevar’s claims of hostile work environment.8

Hocevar asserts that during her tenure at Purdue, she endured a constant litany of vulgar and inappropriate behavior. As previously set forth, the behavior was perpetrated in large part by Amundsen, her direct supervisor, and, in small part, by two other Purdue managers.

As the district court found, there is no dispute that Hocevar, a female, is a member of a protected group. See Carter, 173 F.3d at 700 (female plaintiff member of protected group). Our inquiry, therefore, should turn to the second element of a hostile work environment claim: whether she was subject to unwelcome harassment. Conduct is “unwelcome” where it is “uninvited and offensive.” Bales v. Wal-Mart Stores, Inc., 143 F.3d 1103, 1108 (8th Cir.1998); see Moylan v. Maries County, 792 F.2d 746, 749 (8th Cir.1986) (conduct is unwelcome where employee neither solicited it nor invited it, and regarded it as undesirable or offensive).

That the conduct in question is unwelcome is “[t]he gravamen of any sexual harassment claim.” Meritor Sav. Bank, FSB v. Vinson, 477 U.S. 57, 68, 106 S.Ct. 2399, 91 L.Ed.2d 49 (1986). In determining whether conduct is “unwelcome,” we *729should consider whether the plaintiff indicated, by her conduct, that the alleged harassment was unwelcome. Quick v. Donaldson Co., 90 F.3d 1372, 1378 (8th Cir.1996) (citing Meritor, 477 U.S. at 68, 106 S.Ct. 2399). This is a fact question for the jury and turns largely on credibility determinations. See Meritor, 477 U.S. at 68, 106 S.Ct. 2399 (question of whether conduct is unwelcome presents difficult proof problems turning largely on credibility determinations committed to trier of fact); see also Quick, 90 F.3d at 1378.

The district court focused its inquiry on whether the incidents of harassment were “offensive” and concluded that they failed to constitute an offensive environment due to their infrequent use.9 Under review of the record, I find sufficient evidence and inference therefrom that there was an ongoing use of sexual vulgarity directed at the plaintiff as well as all women employees in general. Based on the evidence set forth below, the plaintiff has certainly met the threshold of proof of pervasiveness as a matter of law. The ultimate determination as to whether the harassment was pervasive must be made by the jury.10

The record in the present case shows Amundsen’s use of sexual vulgarity occurred throughout the workplace, on sales calls and during meetings. Hocevar’s female co-worker, Kiekhaefer, indicates Amundsen repeatedly referred to women as “bitches,” used the “F” word in virtually every other sentence, called clients “fuckers” and “assholes,” and routinely referred to female nurses and female physicians as “fucking bitches.” Similar testimony is in the record from another female employee of Amundsen’s, Mary Beck-Johnson, indicating Amundsen “routinely” used the terms “bitch” and “fuck” in meetings. Further, Hocevar testified that sexually explicit behavior occurred at meetings chaired by Amundsen, which Amundsen either condoned or failed to stop, and that sexually suggestive comments were made by two Purdue managers, one of whom subjected Hocevar to unwelcome physical contact during a consensual dance, which made Hocevar “extremely uncomfortable.” Under the record presented, these vulgar attacks cannot be simply regarded as “offhand” or isolated incidents.

The defendant, as does Judge Beam, relies on the fact that Hocevar herself had used the words “bitch” and “fuck” on occasion in the workplace. On this basis, it is argued that the words are not unwelcome by Hocevar. Hocevar, however, qualifies her use of these words by saying that they were not used in the same context that Amundsen had used them. There is a world of difference between the use of the *730infrequent swear word in the workplace, not actionable when not directed to a specific gender, and direct words demeaning to women in general. While Hocevar’s infrequent use of foul language may indeed, when presented to a jury, diminish her claim that the behavior of Amundsen and others was “unwelcome,” it in no way bars her claim as a matter of law. I am unaware of any case that precludes a plaintiff from arguing that the employer’s constant use of sexually charged language and off-color jokes is unwelcome merely because the plaintiff at times engaged in swearing. Such a reading is inconsistent with the mandate that courts consider the totality of the circumstances of a case. Faragher, 524 U.S. at 787, 118 S.Ct. 2275. Further, Judge Beam’s analysis utterly fails to address Amundsen’s threats of violence, his dissemination of sexually explicit material at meetings, his condonation of sexually graphic behavior at meetings, and the behavior of other Purdue managers, all apparently because Hocevar admitted to the infrequent use of foul language. The record further shows that Hocevar’s swearing was not directed as a demeaning word of harassment at any person or group of people. It is one thing that an employee use vulgarity in his or her general communication; it is quite another when the vulgarity is directed at á specific social group who reasonably could find it to be demeaning to their own self-being.

Even if one concedes that use of foul language by an employee can diminish a claim that the harassment was unwelcome and subjectively offensive, evidence of Hocevar’s reaction could still support a finding that Amundsen’s behavior was unwelcome. See Burns v. McGregor Electronic Industries, Inc., 989 F.2d 959, 964 (8th Cir.1993) (agreeing with district court that plaintiff having posed nude for a nationally distributed magazine does not lead inevitably to conclusion that workplace harassment was welcome); Bales, 143 F.3d at 1108-09 (plaintiffs reaction to harassment sufficient to support jury finding that behavior was unwelcome); see also Caviness v. Nucor-Yamato Steel Co., 105 F.3d 1216, 1223 (8th Cir.1997) (“[I]t stretches credulity to conceive that a reasonable jury might have thought [plaintiffs] welcomed from their co-workers the conduct detailed in the evidence at trial.”).

Hocevar testified that she suffered fear, depression, anxiety and self-doubt as a result of Amundsen’s behavior, including his degrading and demeaning criticism of her work performance. If she was not offended by this, as I think any reasonable person would be, it is difficult to explain that both she and her co-worker were under the continuing care of a psychologist and that Hocevar was treated with Prozac for anxiety and depression. Hocevar’s complaint to Purdue manager Dennis Mer-lo is also “reaction” evidence revealing that Hocevar viewed Amundsen’s sexually derogatory language unwelcome. Further, the record contains evidence that Hocevar, Kiekhaefer, and a third female co-worker were “scared” and upset to the point of tears following the incident surrounding the discussion of Susan Faludi’s book Backlash: The Undeclared War Against American Women. Giving her the benefit of all favorable inferences, this conduct could be found by a jury to support a finding that the harassment was unwelcome and that it was subjectively offensive.11 Precedent supports such a finding. See Harris, 510 U.S. at 22, 114 S.Ct. 367 (recognizing that Title VII bars discriminatory conduct that affects a reasonable person’s psychological well-being); Kopp v. Samaritan Health Sys., Inc., 13 F.3d 264, 269 (8th Cir.1993) (psychological harm *731is a relevant factor in hostile work environment analysis); cf. Jenson v. Eveleth Taconite Co., 130 F.3d 1287, 1304 (8th Cir.1997) (recognizing that a callous pattern and practice of degrading sexual harassment in the workplace can destroy self-esteem of women exposed to it).

The third element of a claim of hostile work environment requires evidence of a causal nexus between the harassment suffered and Hoeevar’s membership in a protected group. See Carter, 173 F.3d at 700. At the summary judgment stage, a plaintiff may prove harassment is “based on sex” by presenting evidence that members of one sex were the primary targets of harassment. Quick, 90 F.3d at 1378 (evidence that members of one sex were primary targets of harassment sufficient to show conduct was gender based for purposes of summary judgment) (quoting Kopp, 13 F.3d at 269-70). Whether harassing conduct is based on sex is determined by inquiring “whether ‘members of one sex are exposed to disadvantageous terms or conditions of employment to which members of the other sex are not exposed.’ ” Quick, 90 F.3d at 1379 (quoting Harris, 510 U.S. at 25, 114 S.Ct. 367) (Ginsburg, J., concurring).

This court has recently reaffirmed that gender-based insults, such as the term “bitch,” may give rise to an inference of discrimination based on sex, see Carter, 173 F.3d at 700, and has rejected the notion that an employee must be propositioned, touched offensively, or harassed by sexual innuendo in order to have been sexually harassed. See Quick, 90 F.3d at 1379 (citing Burns, 989 F.2d at 964). Additionally, we have held that intimidation and hostility toward women in general can result from conduct other than explicit sexual advances.12 See Hall, 842 F.2d at 1014. Moreover, it is well settled that verbal abuse, violence, or physical aggression may constitute sexual harassment, see Quick, 90 F.3d at 1379 (citing Burns, 989 F.2d at 964-65), and that such need not be explicitly sexual in nature. See Carter, 173 F.3d at 700-01. (“All instances of harassment need not be stamped with signs of overt discrimination to be relevant under Title VII if they are part of a course of conduct which is tied to evidence of discriminatory animus.”); see also, Williams v. General Motors Corp., 187 F.3d 553, 565-66 (6th Cir.1999) (gender-specific epithets such as “slut” and “fucking women” can support an inference that the comments were motivated by gender).

On the question of causal nexus, the district court essentially reasoned that because the offensive behavior occurred in front of both men and women and was not specifically directed at Hocevar, it failed to evidence the harasser’s thoughts toward a particular gender. Similarly, Judge Beam holds that because Amundsen indiscriminately used the crude adjective “fucking” when referring to both men and women, the term somehow loses its sexual connotation and cannot be used to show the language was causally linked to gender. He finds that “[tjhis is not a case where Amundsen used the term bitch as a synonym for female-specific characteristics of which he did not approve.” This reasoning is inconsistent with this court’s decision in Kopp where we found sufficient evidence of actionable harassment based on sex to survive summary judgment where women were more frequently exposed to harassment than men, despite that abuse was rarely couched in terms of sex or gender and was used in front of both men and women. Kopp, 13 F.3d at 269-70. Judge Beam’s opinion further fails to appreciate the inherently sexual *732nature of the profane term,13 and also fails to address Amundsen’s chronic characterization of women as “bitches,” “fucking bitches,” and “fat fucking bitches.”

Contrary to Judge Beam’s conclusion,14 Hoeevar presents the hypothetical case we considered in Kriss where a supervisor’s constant use of the word “bitch” was directed only at women. It is apparent that Amundsen used the term “bitch” throughout the workplace in a pejorative manner to describe women who were rude to him or behaved in some way that displeased him.15 This term also has a distinctively negative connotation when used to describe women.16

Giving Hoeevar the benefit of all reasonable inferences, there is little doubt that the extensive use of the gender-specific pejorative “bitch,” coupled with the sexually explicit and offensive term “fuck,” could support a finding that the harassment was based on sex. To hold otherwise is an unprecedented endorsement of the sexually insulting behavior presented. See Burns, 989 F.2d at 965 (vulgar and offensive epithets, including but not limited to “bitch,” are “ ‘widely recognized as not only improper but as intensely degrading, deriving their power to wound not only from their meaning but also from the disgust and violence they express phonetically.’ ”) (citations and internal quotations omitted).

The Supreme Court recently reaffirmed the “severe or pervasive” test articulated in Harris, defining a sexually objectionable environment to mean “one that a reasonable person would find hostile or abusive, and one that the victim in fact did perceive to be so.” Faragher, 524 U.S. at 787, 118 S.Ct. 2275; see Ellerth, 524 U.S. at 754, 118 S.Ct. 2257. Once there is evidence of improper conduct and subjective offense, the question of whether the conduct rose to a persuasive level of abuse is largely one for the jury. See Howard v. Burns Bros., Inc., 149 F.3d 835, 840 (8th Cir.1998); see also O'Shea v. Yellow Technology Services, Inc., 185 F.3d 1093, 1098 (10th Cir.1999) (“the severity and pervasiveness evaluation is particularly unsuited for summary judgment because it is ‘quintessentially a question of fact’ ”) (quoting Beardsley v. Webb, 30 F.3d 524, 530 (4th Cir.1994)).

*733In determining whether a work environment is sufficiently severe or pervasive to alter the terms or conditions of employment, we must look at the totality of the circumstances, including factors such as “ ‘frequency of the discriminatory conduct; its severity; whether it is physically threatening or humiliating, or a mere offensive utterance; and whether it unreasonably interferes with an employee’s work performance.’ ” Faragher, 524 U.S. at 787-88, 118 S.Ct. 2275 (quoting Harris, 510 U.S. at 23, 114 S.Ct. 367). Evidence of psychological harm to the plaintiff is also a relevant factor, Harris, 510 U.S. at 23, 114 S.Ct. 367, as is evidence of harassment of plaintiffs co-workers. See Howard, 149 F.3d at 838 (harassment of plaintiffs coworkers relevant to show pervasiveness of hostile environment).

Under this framework, we should consider the fourth element of a claim of hostile work environment harassment, namely, whether the harassment Hocevar suffered affected a term, condition or privilege of her employment. This court has held that in the context of Title VII “conditions of employment” may be altered by harassment if the employee is discouraged from remaining on the job, Smith v. St. Louis Univ., 109 F.3d 1261, 1264 (8th Cir.1997), or the harassment caused economic injury, affected the employee’s psychological well-being, detracted from job performance, or kept the employee from advancing in her career. Quick, 90 F.3d at 1378. Title VII is violated when a “workplace is permeated with ‘discriminatory intimidation, ridicule and insult’ that is ‘sufficiently severe or pervasive to alter the conditions of the victim’s employment and create an abusive working environment,’” where if the environment is objectively and subjectively perceived as hostile or abusive. Id. (quoting Harris, 510 U.S. at 21, 114 S.Ct. 367) (emphasis added).

On the question of whether Hocevar showed sufficient evidence that the harassment was “severe or pervasive” to alter a term or condition of employment, the district court found the alleged incidents were neither sufficiently pervasive nor directed at Hocevar. 17 In reaching this conclusion, the district court concluded that playing the Jerky Boys tape was not offensive conduct; that Amundsen’s stories of animal violence are not severe enough to constitute a hostile environment; that Amundsen’s referral to a female doctor as a “fucking bitch” and introduction of a new employee as a “fucking new guy” at most offended the person the comments were directed at, which was not Hocevar. On this point, Judge Beam recognizes that the use of foul language may have been pervasive. Because he concludes that the offensive language is not based on sex, however, he sets this evidence aside, then proceeds to consider whether the remaining facts of Hocevar’s case are sufficiently severe or pervasive. This approach errs because it fails to consider the totality of the circumstances, see Harris, 510 U.S. at 23, 114 S.Ct. 367 (all evidence concerning abusiveness of a plaintiffs working condition is relevant), and imposes a per se test requiring harassment be “directed at” plaintiff to be actionable. This approach has been rejected by an appellate court in Vinson v. Taylor, 753 F.2d 141, 146 (D.C.Cir.1985), which expressly held that “[e]ven a woman who was never herself the object of harassment might have a Title VII claim if she were forced to work in an atmosphere in which such harassment was pervasive.” Id. (emphasis added). The Vinson court reached this conclusion in light of EEOC Decision No. 71-909, 25 Ohio App.2d 141, *734267 N.E.2d 814, 3 Fair Empl.Prac.Cas. (BNA) at 269-70 (1971) in which the EEOC found reasonable cause to find a Title VII violation where a white employee was discharged for befriending African American co-workers.18

Title VII provides employees the “right to work in an environment free from discriminatory intimidation, ridicule, and insult.” Meritor, 477 U.S. at 65, 106 S.Ct. 2399 (emphasis added). The EEOC Guidelines defining sexual harassment do not limit sexual harassment to only those actions that are directed at the plaintiff. See 29 C.F.R. § 1604.11 (1999) (“Unwelcome sexual advances, requests for sexual favors, and other verbal or physical conduct of a sexual nature constitute sexual harassment when ... such conduct has the purpose or effect of unreasonably interfering with an individual’s work performance or creating an intimidating, hostile, or offensive working environment.”) (emphasis added).

I find no case that dictates, as Judge Beam suggests, that only behavior directed at the plaintiff, such as sexual advances, may support a claim of hostile work environment sexual harassment. Cf. Breeding, 164 F.3d at 1159 (considering supervisor’s fondling of genitals in view of various employees, including plaintiff, and inappropriate comments made 'in front of men and women); Howard, 149 F.3d at 838 (considering harassment of employees other than plaintiff relevant to show pervasiveness of hostile environment); Kopp, 13 F.3d at 270 (reversing grant of summary judgment where male physician used gender-specific foul language in front of numerous employees, both male and female, only one incident of which was directed at plaintiff); Jenson v. Eveleth Taconite Co., 824 F.Supp. 847 (D.Minn.1993) (class plaintiffs prevailed on hostile work environment claim where much of derogatory and insulting language used by men was directed at women in general). The fact that the bulk of the harassing behavior occurred in Hocevar’s workplace in her presence yet was directed at all women present does not, as a matter of law, preclude a finding of a hostile work environment. This is particularly true where the harassment occurred at the hands of Hocevar’s direct supervisor, in light of the Supreme Court’s recent statement that harassing behavior perpetrated by a supervisor has a “greater power to alter the environment” than similar behavior of mere co-workers. Faragher, 524 U.S. at 805, 118 S.Ct. 2275.

It is inconceivable at the summary judgment stage to suggest that the language used by Timothy Amundsen, Hocevar’s supervisor, coupled with the behavior of two Purdue managers, is insufficient as a matter of law to support a claim of hostile work environment harassment under Title VII. Justice Scalia’s recent observation is appropriate here, “[t]he real social impact of workplace behavior often depends on a constellation of surrounding circumstances, expectations, and relationships which are not fully captured by a simple recitation of the words used or the physical acts performed.” Oncale, 523 U.S. at 81-82, 118 S.Ct. 998.

I disagree with Judge Beam who perceives this to be a case of isolated or sporadic incidents of “mere offensive utterances” in the workplace. Assuming *735Hocevar’s allegations as trae at this early stage of summary judgment, the record reveals a clear pattern of pervasive offensive behavior tinged with gender animus. Hocevar provides evidence that she was physically afraid of Amundsen, particularly after an incident in which he punched a fellow employee, and that she found his demeaning behavior toward women so humiliating that she sought psychiatric treatment and medication. The allegations, if found true by a jury, are sufficient to permit a finding that the cumulative effect of Amundsen’s conduct, along with that of other Purdue managers, was sufficiently severe or pervasive to create a hostile work environment based on sexual harassment.

I therefore dissent from the affirmance granting summary judgment for hostile work environment in violation of Title VII.

BEAM, Circuit Judge, with whom Judge John R. Gibson joins in the result reached in Part IIA.

Marcia Hocevar appeals the district court’s19 grant of summary judgment in favor of Purdue Frederick Company (Purdue) and Timothy Amundsen, resulting in the dismissal of her Title VII claims of hostile work environment and retaliation.20 The district court found that Hocevar had not established a prima facie case of hostile work environment and that she had not demonstrated a retaliation claim.

I. BACKGROUND

I relate the relevant facts in the light most favorable to Hocevar.21 Hocevar worked as a sales representative for Purdue. During Hocevar’s employment, her supervisor, Timothy Amundsen, constantly used the words “bitch,” “fuck,” and “asshole,” and sometimes used combinations of these words. Aside from her claims of constant offensive language, Hocevar also asserts four specific instances of inappropriate conduct by Amundsen. First, Amundsen called a female client who treated him rudely a “fat fucking bitch.” Second, Amundsen called a new male employee a “fucking new guy” throughout a business meeting. Third, at a business meeting, Amundsen played a tape of the Jerky Boys, a set of crude, so-called comedians whose routine often includes offensive language. Fourth, Amundsen said that Purdue’s clients would “cream then-jeans” when they found out about a new product that Purdue had developed.

Hocevar cited four other incidents of sexual harassment involving other company officials. First, in the spring of 1992, while having drinks, several company employees engaged in a heated" argument about Susan Faludi’s book Backlash. During that argument, a company official expressed negative feelings about the feminist movement and another company offi*736cial called Hocevar a “bitch.”22 Second, in January of 1993, another company official made sexual advances toward her and pulled her close to have full-body contact during a dance at a company gathering. Third, at a company meeting in the spring of 1993, two other company officials talked during a presentation that Hocevar was giving. At the end of the presentation, Hocevar confronted them about their rude behavior and one of the men told her that they had been talking about “what great legs” Hocevar had. Fourth, in April of 1995, after a skit performed by three female employees at a company gathering, a company official23 suggested to the room of 150 people that he would be having a sexual liaison with the three women later that evening. During this same gathering, the company official also made a comment that suggested a female employee had a sexual device in her hand.

In August of 1995, Hocevar was injured in a car accident in which she received injuries that kept her off work for several weeks. On September 18, 1995, Hocevar returned to work. However, she was only able to work for a little more than a month before having to take another absence because of continuing pain from injuries sustained in the accident. On December 15, 1995, Purdue sent a letter to Hocevar in which the company expressed concern about her continued absence. Five days later, Hocevar’s attorney responded with a letter that outlined Hocevar’s complaints about Amundsen’s conduct. On May 2, 1996, Hocevar filed a complaint with the EEOC. About a month after the complaint was filed, Purdue terminated Hocevar.

II. DISCUSSION

A. Hostile Work Environment

To succeed on a claim of hostile work environment created by her supervisor, Hocevar has to prove the elements of such a case. These elements are: (1) that she is a member of a protected group; (2) that she was subject to unwelcome sexual harassment; (3) that the harassment was based on sex; and (4) that the harassment affected a term, condition, or privilege of employment. See Phillips v. Taco Bell Corp., 156 F.3d 884, 888 (8th Cir.1998). Purdue has an affirmative defense to liability or damages when no tangible employment action is taken if: (a) Purdue exercised reasonable care to prevent and correct promptly any sexually harassing behavior; and (b) Hocevar unreasonably failed to take advantage of any preventive or corrective opportunities provided by Purdue or to avoid harm otherwise. See Faragher v. City of Boca Raton, 524 U.S. 775, 807, 118 S.Ct. 2275, 141 L.Ed.2d 662 (1998). Hocevar is a member of a protected group. However, Hocevar has failed to establish that the alleged behavior was unwelcome, the discrimination was based on sex, or that the harassment affected a term, condition, or privilege of employment.24

Hocevar has not demonstrated that Amundsen’s use of offensive language was unwelcome. A plaintiff must indicate by her conduct that the alleged harassment was unwelcome. See Quick v. Donaldson Co., 90 F.3d 1372, 1378 (8th Cir.1996). A plaintiff cannot create a genuine issue of material fact with regard to unwelcome behavior when she engages in the conduct complained about. See Scusa v. Nestle U.S.A. Co., 181 F.3d 958, 966 (8th Cir.1999). Hocevar’s own testimony indicates that Amundsen’s use of offensive language was not unwelcome because she used the offensive language herself. Hocevar admitted that she also called the new coworker the “fucking new guy” at the busi*737ness meeting. She further admitted that she used the words “bitch” and “fuck” around both Amundsen and other Purdue employees. I find that these actions on the part of Hocevar vitiate her contention that the mere use of these words was unwelcome.

Hocevar also failed to establish that the discrimination was based on sex. Harassing conduct constitutes discrimination based on sex when members of one sex are exposed to disadvantageous terms or conditions of employment to which members of the other sex are not exposed. See Montandon v. Farmland Indus. Inc., 116 F.3d 355, 358 (8th Cir.1997). Hocevar failed to demonstrate that the language complained about was based on sex. Offensive language was used to describe both men and women. While Amundsen described a female client who had treated him rudely as a “fat fucking bitch,” he also referred to a new male employee as a “fucking new guy”. Offensive language was used in front of both men and women at company meetings and the Jerky Boys tapes were played in front of both men and women. The use of foul language in front of both men and women is not discrimination based on sex. See id. at 358; see also Scusa, 181 F.3d at 965.

Hocevar claims that Amundsen’s use of the term “bitch” itself shows a discriminatory attitude toward females. Gender-based insults may create an inference that discrimination was based on sex. See Carter v. Chrysler Corp., 173 F.3d 693, 700 (8th Cir.1999). However, mere use of the word “bitch,” without other evidence of sex discrimination, is not particularly probative of a general misogynist attitude. See Kriss v. Sprint Communications Co., 58 F.3d 1276, 1281 (8th Cir.1995).

In this ease, Hocevar has presented no additional evidence demonstrating that Amundsen’s use of the word “bitch” connotes a misogynist attitude. This is not a case where Amundsen used the term bitch as a synonym for female-specific characteristics of which he did not approve. See id. at 1281 (noting that evidence where supervisor used word “bitch” as synonym for “complain” would provide stronger evidence of sex harassment because that would demonstrate that supervisor associated complaining with females). Neither is this a case where Amundsen blamed Hocevar’s sexuality for his use of the word “bitch.” Carter, 173 F.3d at 701 (holding that use of sexual epithets is evidence of sexual harassment when co-employee claims he used sexual epithets because plaintiff dressed provocatively and put “her ass up in our faces”). Nor is this a case in which Amundsen engaged in a litany of obscene name calling against Hocevar. See Burns v. McGregor Elec. Indus. Inc., 989 F.2d 959, 964 (8th Cir.1993) (finding discrimination based on sex when male co-worker called female plaintiff a “bitch,” “asshole,” “slut,” and “cunt”). Because Hocevar has failed to present any additional evidence to bolster her contention that Amundsen’s pervasive use of the term “bitch” shows his misogynist attitude, I find that Hocevar cannot demonstrate the harassment was based on sex.

Finally, Hocevar cannot show that the harassment was sufficiently severe or pervasive so as to alter a term, condition, or privilege of employment. See Taco Bell, 156 F.3d at 888. “Conduct that is not severe or pervasive enough to create an objectively hostile or abusive environment-an environment that a reasonable person would find hostile or abusive-is beyond Title VII’s purview.” Oncale v. Sundowner Offshore Services, Inc., 523 U.S. 75, 81, 118 S.Ct. 998, 140 L.Ed.2d 201 (1998). Factors to consider when determining whether sexual harassment is sufficiently severe or pervasive include: “the frequency of the discriminatory conduct; its severity; whether it is physically threatening or humiliating, or a mere offensive utterance; and whether it unreasonably interferes with an employee’s work performance.” Harris v. Forklift Sys. Inc., 510 U.S. 17, 23, 114 S.Ct. 367, 126 L.Ed.2d 295 (1993). More than a few *738isolated instances are required. See Kimzey v. Wal-Mart Stores, Inc., 107 F.3d 568, 573 (8th Cir.1997). While the use of foul language may have been pervasive, I have already concluded that it was neither unwelcome nor based on sex.

This leaves Hocevar with four events that might constitute unwelcome behavior based on sex: (1) the Backlash incident; (2) the “great legs” incident; (3) the dancing incident; and (4) the skit incident. I will assume that all of these incidents could constitute unwelcome behavior based on sex. However, these incidents were clearly not pervasive because they occurred over at least a three-year period. In addition, a few inappropriate comments and an unwanted slow dance do not amount to particularly severe conduct that was threatening or humiliating.

I have little doubt that Amundsen’s behavior was boorish and unprofessional. But, Title VII is not a general civility code. See Faragher, 524 U.S. at 788, 118 S.Ct. 2275. The simple fact is that the cases on which Hocevar relies involved far, far more evidence than Hocevar has presented. See Rorie v. United Parcel Serv., 151 F.3d 757 (8th Cir.1998) (reversing summary judgment against plaintiff where supervisor patted female employee on back, brushed up against her, told her she “smelled good,” always “came-on” to her, and asked her about co-worker’s penis size); Howard v. Burns Bros. Inc., 149 F.3d 835 (8th Cir.1998) (affirming jury verdict where co-employee always used sexual innuendos, told plaintiff she had nice legs, brushed her buttocks, told jokes involving lewd gestures, and touched the buttocks of and talked “nasty” to other female employees); Hall v. Gus Const. Co., 842 F.2d 1010 (8th Cir.1988) (upholding judgment when plaintiffs’ male co-workers made repeated requests for sex and touched plaintiffs’ breasts and thighs). While I sympathize with Hocevar’s having to endure Amundsen’s conduct, her assertions fall far short of proof of a hostile work environment.

B. Retaliation

To establish a prima facie retaliation case, Hocevar must prove that: (1) she engaged in protected activity; (2) Purdue took adverse action against her; and (3) there is a causal connection between the two. See Scott v. County of Ramsey, 180 F.3d 913, 917 (8th Cir.1999). If Hocevar establishes a prima facie case, a presumption of retaliation arises, and the burden then falls on Purdue to advance a legitimate reason for the adverse employment action. See id. If Purdue advances a legitimate reason, the presumption drops out and Hocevar has the burden of demonstrating intentional retaliation. See id. If there is no direct proof of retaliation, as here, a claimant may, under some circumstances, advance indirect proof by evidence that the so-called legitimate reason is merely a pretext for unlawful retaliatory conduct. See id. I find that Hocevar established a prima facie case of retaliation. However, I also find that Purdue advanced a legitimate reason for Hocevar’s termination and that Hocevar presented no evidence of pretext.

Hocevar engaged in protected activity when she lodged a complaint with the EEOC on May 2, 1996. Purdue then took an adverse employment action against Hocevar by terminating her on June 7, 1996.25 Finally, she established an inference of a causal connection because her termination closely followed the filing of her EEOC complaint and also closely followed Amundsen’s return from a three-month suspension received, in part, be*739cause of Hocevar’s complaints to Purdue officials. See Smith v. Riceland Foods, Inc., 151 F.3d 813, 819-20 (8th Cir.1998) (causal connection established by circumstantial evidence, including close proximity of time between plaintiffs engagement in protected activity and the adverse employment action).

However, Purdue presented a legitimate reason for Hocevar’s termination. In its termination letter, Purdue expressly noted that Hocevar was dismissed because of the need to re-staff her territory due to her lengthy absence. Hocevar does not dispute her absence from work for more than seven months, and I have little doubt that Purdue has a legitimate need to have its sales territories covered. Thus, I find it was legitimate for Purdue to terminate Hocevar in order to re-staff her vacant sales territory.

Hocevar argues that the legitimate reason advanced by Purdue is a pretext for retaliation because: (1) Purdue failed to re-staff another important sales territory for three months and (2) a similarly situated employee who did not complain to the EEOC was not terminated by Purdue. As an initial matter, it does not appear that Hocevar made these arguments to the district court. See Womack v. City of Bellefontaine Neighbors, 193 F.3d 1028, 1032 (8th Cir.1999) (declining to address arguments first advanced on appeal). However, even assuming that these arguments were advanced below, she has not provided evidence of pretext.

The fact that Purdue left another territory unstaffed for three months does not amount to pretext. Perhaps if Purdue had terminated Hocevar after three months, this would be somewhat persuasive. However, Hocevar’s territory was left unstaffed for seven months-more than double the amount of time that Hocevar claims another territory was left unstaffed. Because of the significant difference in the amount of time that Hocevar’s territory remained un-staffed, Purdue’s failure to re-staff another territory for three months provides no evidence of pretext.

Hocevar’s contention about an allegedly similarly situated employee also fails. Hocevar presented evidence that Purdue did not terminate another sales representative who also expressed concern about Amundsen’s conduct but who did not file an EEOC complaint. Where the only evidence of pretext is disparate treatment when compared to another employee, a plaintiff must show that the other employee was similarly situated in all relevant aspects. See Scott, 180 F.3d at 917. In this case, the other sales representative and Hocevar were not similarly situated in all relevant aspects because the other sales representative was never absent from work. Thus, Purdue’s retention of the other sales representative provides no support for Hocevar’s allegations of pretext.

III. CONCLUSION

For the foregoing reasons, I would affirm the district court’s dismissal of the hostile work environment claim. I would also affirm the district court’s decision on the retaliation claim.

JOHN R. GIBSON, Circuit Judge,

concurring specially in affirmance of grant of summary judgment on hostile work environment claim.

I concur separately in the decision to affirm judgment against Hocevar on her hostile environment claim. I do not concur in Judge Beam’s opinion, because I believe it engages in fact finding, see supra at 725-26, and an unnecessary semantic dissection of the language in question. I recognize that our decision in Kriss v. Sprint Communications Co., 58 F.3d 1276, 1281 (8th Cir.1995), stated that “the word ‘bitch’ ... is not an indication of a general misogynist attitude,” but I do not read Kriss as establishing a test as to whether this word is inherently of a sexually harassing nature. It is beyond question that the repetitive use of the word in this case was demeaning to females, and the disci*740pline which Purdue Frederick imposed on Amundsen, who used the word, shows that the company recognized the utterance was improper.

Even unquestionably offensive words do not necessarily make a hostile work environment, without considering the context. See Oncale v. Sundowner Offshore Servs., Inc., 523 U.S. 75, 80, 118 S.Ct. 998, 140 L.Ed.2d 201 (1998) (use of words having sexual content or connotations not necessarily discrimination because of sex). I have examined the whole record and conclude that the facts taken in the light most favorable to Hocevar do not add up to a hostile work environment case.

At the outset, it is necessary to focus on the key facts. I believe that the only conduct on the record that might be substantial enough to alter a term, condition, or privilege of employment, see Harris v. Forklift Sys., Inc., 510 U.S. 17, 21-22, 114 S.Ct. 367, 126 L.Ed.2d 295 (1993), is Amundsen’s use of sexual vulgarities. The other allegations about Amundsen’s conduct, such as his involvement in a bar fight and his statements to employees at sales meetings that if sales did not improve, he was “going to kill you” or “kill your dog,” do not appear to be gender-related. See Oncale, 523 U.S. at 80, 118 S.Ct. 998 (to be actionable, harassment must be discrimination because of sex). Hocevar’s complaints about people other than Amundsen are based on incidents that were simply too few and far between to make a hostile environment.

Hocevar alleges that Amundsen chronically used foul language, specifically the words “fuck” and “bitch.” In a letter dated December 20, 1995, Hocevar’s counsel notified Purdue Frederick that Amundsen had engaged in ongoing sexual harassment against Hocevar. In January 1996 Purdue Frederick investigated the complaint. The investigation revealed that the members of the Viking District interviewed had “all participated in the use of profanity and, from time to time, off-color jokes. And the general consensus was that they carried it too far.” In particular, the investigator concluded that Hocevar herself had participated in the profanity and improper jokes. Hocevar admitted at her deposition in this case that she used the same offensive language around Amundsen and other sales representatives. A Purdue personnel representative traveled to the Viking District in February to inform the employees there that the firm expected a high level of professionalism and that the investigation revealed that “perhaps there had been some deterioration in that professional standard” with the use of profanity and jokes. Amundsen was put on a three-month leave and required to undergo counseling and training. This discipline resulted in an improvement in Amundsen’s language, as well as that of the other employees.

Hocevar complains most specifically of an incident in September 1994, at Dr. Ku-bics’s office, in which Amundsen became enraged at being treated disrespectfully by a female doctor. Amundsen repeatedly referred to the doctor as a “fat fucking bitch,” and talked about how he would like to “slam her one” and make her fear him.

In deciding when inappropriate conduct rises to the level of a hostile environment that changes a term or condition of employment, courts must look at all the circumstances. “These may include the frequency of the discriminatory conduct; its severity; whether it is physically threatening or humiliating, or a mere offensive utterance; and whether it unreasonably interferes with an employee’s work performance.” Harris, 510 U.S. at 23, 114 S.Ct. 367 (emphasis added). In this regard, it is important that the conduct complained of consists only of “offensive utterances” of a type Hocevar herself engaged in at work, and that to the extent the gender-based utterances were threatening or abusive, they were not aimed at Hocevar, but at the female doctor, who was not a Purdue employee.

*741“We have considered harassment of employees other than the plaintiff to be relevant to show pervasiveness of the hostile environment.” Howard v. Burns Bros., Inc., 149 F.3d 835, 838 (8th Cir.1998) (emphasis added). However, in the cases in our circuit where we have considered conduct directed at others in upholding sexual harassment claims, that conduct augmented evidence of harassment directed at the plaintiff, see, e.g., id. (evidence of harassment of others augmented evidence of physical contact of plaintiff and chronic innuendos); Hall v. Gus Const. Co., 842 F.2d 1010, 1015 (8th Cir.1988) (each plaintiff endured abuse). Abuse directed at a third party is part of the picture, but it is less significant than abuse directed at the plaintiff. See Gleason v. Mesirow Fin., Inc., 118 F.3d 1134, 1144-45 (7th Cir.1997) (“Second hand harassment” not as great an invasion as harassment directed at plaintiff); Black v. Zaring Homes, Inc., 104 F.3d 822, 826 (6th Cir.1997) (fact that most comments not directed at plaintiff contributes to conclusion of insufficiency of evidence). But see Leibovitz v. New York City Transit Auth., 4 F .Supp.2d 144, 150-53 (E.D.N.Y.1998) (upholding hostile environment verdict based entirely on harassment of others). Here, the conduct was directed at someone who was not a Purdue Frederick employee, and who was not even present to hear the hostile remarks. Therefore, even considering the evidence of Amundsen’s reaction to the female doctor, Hocevar did not start out with a strong case. The standards for establishing a hostile environment are set high so that Title VII “does not become a ‘general civility code.’ ” Faragher v. City of Boca Raton, 524 U.S. 775, 788, 118 S.Ct. 2275, 141 L.Ed.2d 662 (1998) (quoting Oncale, 523 U.S. at 80, 118 S.Ct. 998).

But Hoeevar’s claim truly struck the shoals when she admitted she used the sort of language she now complains of. In Scusa v. Nestle U.S.A. Co., 181 F.3d 958, 967 (8th Cir.1999), we held that a plaintiff could not show others’ workplace conduct was subjectively offensive when she did the same thing herself. “Appellant’s evidence of a hostile work environment falls flat in light of the fact that she engaged in the very type of conduct about which she now complains.... ” Id. at 967. Accord Gleason, 118 F.3d at 1146.

For these reasons, I concur in affirming the summary judgment entered against Hocevar on her hostile environment claim.

Justice Brennan

announced the judgment of the Court and delivered an opinion, in which Justice Marshall, Justice Blackmun, and Justice Stevens join.

Ann Hopkins was a senior manager in an office of Price Waterhouse when she was proposed for partnership in 1982. She was neither offered nor denied admission to the partnership; instead, her candidacy was held for reconsideration the following year. When the partners in her office later re*232fused to repropose her for partnership, she sued Price Waterhouse under Title VII of the Civil Rights Act of 1964, 78 Stat. 253, as amended, 42 U. S. C. §2000e et seq., charging that the firm had discriminated against her on the basis of sex in its decisions regarding partnership. Judge Gesell in the Federal District Court for the District of Columbia ruled in her favor on the question of liability, 618 F. Supp. 1109 (1985), and the Court of Appeals for the District of Columbia Circuit affirmed. 263 U. S. App. D. C. 321, 825 F. 2d 458 (1987). We granted certiorari to resolve a conflict among the Courts of Appeals concerning the respective burdens of proof of a defendant and plaintiff in a suit under Title VII when it has been shown that an employment decision resulted from a mixture of legitimate and illegitimate motives. 485 U. S. 933 (1988).

I

At Price Waterhouse, a nationwide professional accounting partnership, a senior manager becomes a candidate for partnership when the partners in her local office submit her name as a candidate. All of the other partners in the firm are then invited to submit written comments on each candidate— either on a “long” or a “short” form, depending on the partner’s degree of exposure to the candidate. Not every partner in the firm submits comments on every candidate. After reviewing the comments and interviewing the partners who submitted them, the firm’s Admissions Committee makes a recommendation to the Policy Board. This recommendation will be either that the firm accept the candidate for partnership, put her application on “hold,” or deny her the promotion outright. The Policy Board then decides whether to submit the candidate’s name to the entire partnership for a vote, to “hold” her candidacy, or to reject her. The recommendation of the Admissions Committee, and the decision of the Policy Board, are not controlled by fixed guidelines: a certain number of positive comments from partners will not guarantee a candidate’s admission to the partnership, nor will a specific *233quantity of negative comments necessarily defeat her application. Price Waterhouse places no limit on the number of persons whom it will admit to the partnership in any given year.

Ann Hopkins had worked at Price Waterhouse’s Office of Government Services in Washington, D. C., for five years when the partners in that office proposed her as a candidate for partnership. Of the 662 partners at the firm at that time, 7 were women. Of the 88 persons proposed for partnership that year, only 1 — Hopkins—was a woman. Forty-seven of these candidates were admitted to the partnership, 21 were rejected, and 20 — including Hopkins — were “held” for reconsideration the following year.1 Thirteen of the 32 partners who had submitted comments on Hopkins supported her bid for partnership. Three partners recommended that her candidacy be placed on hold, eight stated that they did not have an informed opinion about her, and eight recommended that she be denied partnership.

In a jointly prepared statement supporting her candidacy, the partners in Hopkins’ office showcased her successful 2-year effort to secure a $25 million contract with the Department of State, labeling it “an outstanding performance” and one that Hopkins carried out “virtually at the partner level.” Plaintiff’s Exh. 15. Despite Price Waterhouse’s attempt at trial to minimize her contribution to this project, Judge Ge-*234sell specifically found that Hopkins had “played a key role in Price Waterhouse’s successful effort to win a multi-million dollar contract with the Department of State.” 618 F. Supp., at 1112. Indeed, he went on, “[n]one of the other partnership candidates at Price Waterhouse that year had a comparable record in terms of successfully securing major contracts for the partnership.” Ibid.

The partners in Hopkins’ office praised her character as well as her accomplishments, describing her in their joint statement as “an outstanding professional” who had a “deft touch,” a “strong character, independence and integrity.” Plaintiff’s Exh. 15. Clients appear to have agreed with these assessments. At trial, one official from the State Department described her as “extremely competent, intelligent,” “strong and forthright, very productive, energetic and creative.” Tr. 150. Another high-ranking official praised Hopkins’ decisiveness, broadmindedness, and “intellectual clarity”; she was, in his words, “a stimulating conversationalist.” Id., at 156-157. Evaluations such as these led Judge Gesell to conclude that Hopkins “had no difficulty dealing with clients and her clients appear to have been very pleased with her work” and that she “was generally viewed as a highly competent project leader who worked long hours, pushed vigorously to meet deadlines and demanded much from the multidisciplinary staffs with which she worked.” 618 F. Supp., at 1112-1113.

On too many occasions, however, Hopkins’ aggressiveness apparently spilled over into abrasiveness. Staff members seem to have borne the brunt of Hopkins’ brusqueness. Long before her bid for partnership, partners evaluating her work had counseled her to improve her relations with staff members. Although later evaluations indicate an improvement, Hopkins’ perceived shortcomings in this important area eventually doomed her bid for partnership. Virtually all of the partners’ negative remarks about Hopkins — even those of partners supporting her — had to do with her “inter*235personal skills.” Both “[s]upporters and opponents of her candidacy,” stressed Judge Gesell, “indicated that she was sometimes overly aggressive, unduly harsh, difficult to work with and impatient with staff.” Id., at 1113.

There were clear signs, though, that some of the partners reacted negatively to Hopkins’ personality because she was a woman. One partner described her as “macho” (Defendant’s Exh. 30); another suggested that she “overcompensated for being a woman” (Defendant’s Exh. 31); a third advised her to take “a course at charm school” (Defendant’s Exh. 27). Several partners criticized her use of profanity; in response, one partner suggested that those partners objected to her swearing only “because it’s a lady using foul language.” Tr. 321. Another supporter explained that Hopkins “ha[d] matured from a tough-talking somewhat masculine hard-nosed mgr to an authoritative, formidable, but much more appealing lady ptr candidate.” Defendant’s Exh. 27. But it was the man who, as Judge Gesell found, bore responsibility for explaining to Hopkins the reasons for the Policy Board’s decision to place her candidacy on hold who delivered the coup de grace: in order to improve her chances for partnership, Thomas Beyer advised, Hopkins should “walk more femininely, talk more femininely, dress more femininely, wear make-up, have her hair styled, and wear jewelry.” 618 F. Supp., at 1117.

Dr. Susan Fiske, a social psychologist and Associate Professor of Psychology at Carnegie-Mellon University, testified at trial that the partnership selection process at Price Waterhouse was likely influenced by sex stereotyping. Her testimony focused not only on the overtly sex-based comments of partners but also on gender-neutral remarks, made by partners who knew Hopkins only slightly, that were intensely critical of her. One partner, for example, baldly stated that Hopkins was “universally disliked” by staff (Defendant’s Exh. 27), and another described her as “consistently annoying and irritating” (ibid.); yet these were people who had had very little contact with Hopkins. According to *236Fiske, Hopkins’ uniqueness (as the only woman in the pool of candidates) and the subjectivity of the evaluations made it likely that sharply critical remarks such as these were the product of sex stereotyping — although Fiske admitted that she could not say with certainty whether any particular comment was the result of stereotyping. Fiske based her opinion on a review of the submitted comments, explaining that it was commonly accepted practice for social psychologists to reach this kind of conclusion without having met any of the people involved in the decisionmaking process.

In previous years, other female candidates for partnership also had been evaluated in sex-based terms. As a general matter, Judge Gesell concluded, “[candidates were viewed favorably if partners believed they maintained their femin[injity while becoming effective professional managers”; in this environment, “[t]o be identified as a ‘women’s lib[b]er’ was regarded as [a] negative comment.” 618 F. Supp., at 1117. In fact, the judge found that in previous years “[o]ne partner repeatedly commented that he could not consider any woman seriously as a partnership candidate and believed that women were not even capable of functioning as senior managers —yet the firm took no action to discourage his comments and recorded his vote in the overall summary of the evaluations.” Ibid.

Judge Gesell found that Price Waterhouse legitimately emphasized interpersonal skills in its partnership decisions, and also found that the firm had not fabricated its complaints about Hopkins’ interpersonal skills as a pretext for discrimination. Moreover, he concluded, the firm did not give decisive emphasis to such traits only because Hopkins was a woman; although there were male candidates who lacked these skills but who were admitted to partnership, the judge found that these candidates possessed other, positive traits that Hopkins lacked.

The judge went on to decide, however, that some of the partners’ remarks about Hopkins stemmed from an imper*237missibly cabined view of the proper behavior of women, and that Price Waterhouse had done nothing to disavow reliance on such comments. He held that Price Waterhouse had unlawfully discriminated against Hopkins on the basis of sex by consciously giving credence and effect to partners’ comments that resulted from sex stereotyping. Noting that Price Waterhouse could avoid equitable relief by proving by clear and convincing evidence that it would have placed Hopkins’ candidacy on hold even absent this discrimination, the judge decided that the firm had not carried this heavy burden.

The Court of Appeals affirmed the District Court’s ultimate conclusion, but departed from its analysis in one particular: it held that even if a plaintiff proves that discrimination played a role in an employment decision, the defendant will not be found liable if it proves, by clear and convincing evidence, that it would have made the same decision in the absence of discrimination. 263 U. S. App. D. C., at 333-334, 825 F. 2d, at 470-471. Under this approach, an employer is not deemed to have violated Title VII if it proves that it would have made the same decision in the absence of an impermissible motive, whereas under the District Court’s approach, the employer’s proof in that respect only avoids equitable relief. We decide today that the Court of Appeals had the better approach, but that both courts erred in requiring the employer to make its proof by clear and convincing evidence.

II

The specification of the standard of causation under Title VII is a decision about the kind of conduct that violates that statute. According to Price Waterhouse, an employer violates Title VII only if it gives decisive consideration to an employee’s gender, race, national origin, or religion in making a decision that affects that employee. On Price Waterhouse’s theory, even if a plaintiff shows that her gender played a part in an employment decision, it is still her burden to show that the decision would have been different if the employer had *238not discriminated. In Hopkins’ view, on the other hand, an employer violates the statute whenever it allows one of these attributes to play any part in an employment decision. Once a plaintiff shows that this occurred, according to Hopkins, the employer’s proof that it would have made the same decision in the absence of discrimination can serve to limit equitable relief but not to avoid a finding of liability.2 We conclude that, as often happens, the truth lies somewhere in between.

*239A

In passing Title VII, Congress made the simple but momentous announcement that sex, race, religion, and national origin are not relevant to the selection, evaluation, or compensation of employees.3 Yet, the statute does not purport to limit the other qualities and characteristics that employers may take into account in making employment decisions. The converse, therefore, of “for cause” legislation,4 Title VII eliminates certain bases for distinguishing among employees while otherwise preserving employers’ freedom of choice. This balance between employee rights and employer prerogatives turns out to be decisive in the case before us.

Congress’ intent to forbid employers to take gender into account in making employment decisions appears on the face of the statute. In now-familiar language, the statute forbids *240an employer to “fail or refuse to hire or to discharge any individual, or otherwise to discriminate with respect to his compensation, terms, conditions, or privileges of employment,” or to “limit, segregate, or classify his employees or applicants for employment in any way which would deprive or tend to deprive any individual of employment opportunities or otherwise adversely affect his status as an employee, because of such individual’s . . . sex.” 42 U. S. C. §§2000e-2(a)(l), (2) (emphasis added).5 We take these words to mean that gender must be irrelevant to employment decisions. To construe the words “because of” as colloquial shorthand for “but-for causation,” as does Price Waterhouse, is to misunderstand them.6

But-for causation is a hypothetical construct. In determining whether a particular factor was a but-for cause of a given event, we begin by assuming that that factor was present at the time of the event, and then ask whether, even if that factor had been absent, the event nevertheless would have transpired in the same way. The present, active tense of the operative verbs of § 703(a)(1) (“to fail or refuse”), in contrast, turns our attention to the actual moment of the *241event in question, the adverse employment decision. The critical inquiry, the one commanded by the words of § 703(a)(1), is whether gender was a factor in the employment decision at the moment it was made. Moreover, since we know that the words “because of” do not mean “solely because of,”7 we also know that Title YII meant to condemn even those decisions based on a mixture of legitimate and illegitimate considerations. When, therefore; an employer considers both gender and legitimate factors at the time of making a decision, that decision was “because of” sex and the other, legitimate considerations — even if we may say later, in the context of litigation, that the decision would have been the same if gender had not been taken into account.

To attribute this meaning to the words “because of” does not, as the dissent asserts, post, at 282, divest them of causal significance. A simple example illustrates the point. Suppose two physical forces act upon and move an object, and suppose that either force acting alone would have moved the object. As the dissent would have it, neither physical force was a “cause” of the motion unless we can show that but for one or both of them, the object would not have moved; apparently both forces were simply “in the air” unless we can identify at least one of them as a but-for cause of the object’s movement. Ibid. Events that are causally overdetermined, in other words, may not have any “cause” at all. This cannot be so.

We need not leave our common sense at the doorstep when we interpret a statute. It is difficult for us to imagine that, in the simple words “because of,” Congress meant to obligate a plaintiff to identify the precise causal role played by legitimate and illegitimate motivations in the employment decision she challenges. We conclude, instead, that Congress meant *242to obligate her to prove that the employer relied upon sex-based considerations in coming to its decision.

Our interpretation of the words “because of” also is supported by the fact that Title VII does identify one circumstance in which an employer may take gender into account in making an employment decision, namely, when gender is a “bona fide occupational' qualification [(BFOQ)] reasonably necessary to the normal operation of th[e] particular business or enterprise.” 42 U. S. C. §2000e-2(e). The only plausible inference to draw from this provision is that, in all other circumstances, a person’s gender may not be considered in making decisions that affect her. Indeed, Title VII even forbids employers to make gender an indirect stumbling block to employment opportunities. An employer may not, we have held, condition employment opportunities on the satisfaction of facially neutral tests or qualifications that have a disproportionate, adverse impact on members of protected groups when those tests or qualifications are not required for performance of the job. See Watson v. Fort Worth Bank & Trust, 487 U. S. 977 (1988); Griggs v. Duke Power Co., 401 U. S. 424 (1971).

To say that an employer may not take gender into account is not, however, the end of the matter, for that describes only one aspect of Title VII. The other important aspect of the statute is its preservation of an employer’s remaining freedom of choice. We conclude that the preservation of this freedom means that an employer shall not be liable if it can prove that, even if it had not taken gender into account, it would have come to the same decision regarding a particular person. The statute’s maintenance of employer prerogatives is evident from the statute itself and from its history, both in Congress and in this Court.

To begin with, the existence of the BFOQ exception shows Congress’ unwillingness to require employers to change the very nature of their operations in response to the statute. And our emphasis on “business necessity” in disparate-*243impact cases, see Watson and Griggs, and on “legitimate, nondiscriminatory reason[s]” in disparate-treatment cases, see McDonnell Douglas Corp. v. Green, 411 U. S. 792, 802 (1973); Texas Dept. of Community Affairs v. Burdine, 450 U. S. 248 (1981), results from our awareness of Title VII’s balance between employee rights and employer prerogatives. In McDonnell Douglas, we described as follows Title VII’s goal to eradicate discrimination while preserving workplace efficiency: “The broad, overriding interest, shared by employer, employee, and consumer, is efficient and trustworthy workmanship assured through fair and racially neutral employment and personnel decisions. In the implementation of such decisions, it is abundantly clear that Title VII tolerates no racial discrimination, subtle or otherwise.” 411 U. S., at 801.

When an employer ignored the attributes enumerated in the statute, Congress hoped, it naturally would focus on the qualifications of the applicant or employee. The intent to drive employers to focus on qualifications rather than on race, religion, sex, or national origin is the theme of a good deal of the statute’s legislative history. An interpretive memorandum entered into the Congressional Record by Senators Case and Clark, comanagers of the bill in the Senate, is representative of this general theme8 According to their memorandum, Title VII “‘expressly protects the employer’s right to insist that any prospective applicant, Negro or white, must meet the applicable job qualifications. Indeed, the very purpose of title VII is to promote hiring on the basis of job qualifications, rather than on the basis of race or color.’”9 110 Cong. Rec. 7247 (1964), quoted in Griggs v. *244Duke Power Co., supra, at 434. The memorandum went on: “To discriminate is to make a distinction, to make a difference in treatment or favor, and those distinctions or differences in treatment or favor which are prohibited by section 704 are those which are based on any five of the forbidden criteria: race, color, religion, sex, and national origin. Any other criterion or qualification for employment is not affected by this title.” 110 Cong. Rec. 7213 (1964).

Many other legislators made statements to a similar effect; we see no need to set out each remark in full here. The' central point is this: while an employer may not take gender into account in making an employment decision (except in those very narrow circumstances in which gender is a BFOQ), it is free to decide against a woman for other reasons. We think these principles require that, once a plaintiff in a Title VII case shows that gender played a motivating part in an employment decision, the defendant may avoid a finding of liability 10 only by proving that it would have made the same *245decision even if it had not allowed gender to play such a role. This balance of burdens is the direct result of Title VII’s balance of rights.

Our holding casts no shadow on Burdine, in which we decided that, even after a plaintiff has made out a prima facie case of discrimination under Title VII, the burden of persuasion does not shift to the employer to show that its stated legitimate reason for the employment decision was the true reason. 450 U. S., at 256-258. We stress, first, that nei*246ther court below shifted the burden of persuasion to Price Waterhouse on this question, and in fact, the District Court found that Hopkins had not shown that the firm’s stated reason for its decision was pretextual. 618 F. Supp., at 1114-1115. Moreover, since we hold that the plaintiff retains the burden of persuasion on the issue whether gender played a part in the employment decision, the situation before us is not the one of “shifting burdens” that we addressed in Burdine. Instead, the employer’s burden is most appropriately deemed an affirmative defense: the plaintiff must persuade the factfinder on one point, and then the employer, if it wishes to prevail, must persuade it on another. See NLRB v. Transportation Management Corp., 462 U. S. 393, 400 (1983).11

Price Waterhouse’s claim that the employer does not bear any burden of proof (if it bears one at all) until the plaintiff has shown “substantial evidence that Price Waterhouse’s explanation for failing to promote Hopkins was not the ‘true .reason’ for its action” (Brief for Petitioner 20) merely restates its argument that the plaintiff in a mixed-motives case *247must squeeze her proof into Burdine’s framework. Where a decision was the product of a mixture of legitimate and illegitimate motives, however, it simply makes no sense to ask whether the legitimate reason was “the ‘true reason’ ” (Brief for Petitioner 20 (emphasis added)) for the decision — which is the question asked by Burdine. See Transportation Management, supra, at 400, n. 5.12 Oblivious to this last point, the dissent would insist that Burdine’s framework perform work that it was never intended to perform. It would require a plaintiff who challenges an adverse employment decision in which both legitimate and illegitimate considerations played a part to pretend that the decision, in fact, stemmed from a single source — for the premise of Burdine is that either a legitimate or an illegitimate set of considerations led to the challenged decision. To say that Burdine’s evidentiary scheme will not help us decide a case admittedly involving both kinds of considerations is not to cast aspersions on the utility of that scheme in the circumstances for which it was designed.

*248B

In deciding as we do today, we do not traverse new ground. We have in the past confronted Title VII cases in which an employer has used an illegitimate criterion to distinguish among employees, and have held that it is the employer’s burden to justify decisions resulting from that practice. When an employer has asserted that gender is a BFOQ within the meaning of § 703(e), for example, we have assumed that it is the employer who must show why it must use gender as a criterion in employment. See Dothard v. Rawlinson, 433 U. S. 321, 332-337 (1977). In a related context, although the Equal Pay Act expressly permits employers to pay different wages to women where disparate pay is the result of a “factor other than sex,” see 29 U. S. C. § 206(d)(1), we have decided that it is the employer, not the employee, who must prove that the actual disparity is not sex linked. See Corning Glass Works v. Brennan, 417 U. S. 188, 196 (1974). Finally, some courts have held that, under Title VII as amended by the Pregnancy Discrimination Act, it is the employer who has the burden of showing that its limitations on the work that it allows a pregnant woman to perform are necessary in light of her pregnancy. See, e. g., Hayes v. Shelby Memorial Hospital, 726 F. 2d 1543, 1548 (CA11 1984); Wright v. Olin Corp., 697 F. 2d 1172, 1187 (CA4 1982). As these examples demonstrate, our assumption always has been that if an employer allows gender to affect its decision-making process, then it must carry the burden of justifying its ultimate decision. We have not in the past required women whose gender has proved relevant to an employment decision to establish the negative proposition that they would not have been subject to that decision had they been men, and we do not do so today.

We have reached a similar conclusion in other contexts where the law announces that a certain characteristic is irrelevant to the allocation of burdens and benefits. In Mt. Healthy City Bd. of Ed. v. Doyle, 429 U. S. 274 (1977), the *249plaintiff claimed that he had been discharged as a public school teacher for exercising his free-speech rights under the First Amendment. Because we did not wish to “place an employee in a better position as a result of the exercise of constitutionally protected conduct than he would have occupied had he done nothing,” id., at 285, we concluded that such an employee “ought not to be able, by engaging in such conduct, to prevent his employer from assessing his performance record and reaching a decision not to rehire on the basis of that record.” Id., at 286. We therefore held that once the plaintiff had shown that his constitutionally protected speech was a “substantial” or “motivating factor” in the adverse treatment of him by his employer, the employer was obligated to prove “by a preponderance of the evidence that it would have reached the same decision as to [the plaintiff] even in the absence of the protected conduct.” Id., at 287. A court that finds for a plaintiff under this standard has effectively concluded that an illegitimate motive was a “but-for” cause of the employment decision. See Givhan v. Western Line Consolidated School Dist., 439 U. S. 410, 417 (1979). See also Arlington Heights v. Metropolitan Housing Development Corp., 429 U. S. 252, 270-271, n. 21 (1977) (applying Mt. Healthy standard where plaintiff alleged that unconstitutional motive had contributed to enactment of legislation); Hunter v. Underwood, 471 U. S. 222, 228 (1985) (same).

In Transportation Management, we upheld the NLRB’s interpretation of § 10(c) of the National Labor Relations Act, which forbids a court to order affirmative relief for discriminatory conduct against a union member “if such individual was suspended or discharged for cause.” 29 U. S. C. § 160(c). The Board had decided that this provision meant that once an employee had shown that his suspension or discharge was based in part on hostility to unions, it was up to the employer to prove by a preponderance of the evidence that it would have made the same decision in the absence of this impermissible motive. In such a situation, we empha*250sized, “[t]he employer is a wrongdoer; he has acted out of a motive that is declared illegitimate by the statute. It is fair that he bear the risk that the influence of legal and illegal motives cannot be separated, because he knowingly created the risk and because the risk was created not by innocent activity but by his own wrongdoing.” 462 U. S., at 403.

We have, in short, been here before. Each time, we have concluded that the plaintiff who shows that an impermissible motive played a motivating part in an adverse employment decision has thereby placed upon the defendant the burden to show that it would have made the same decision in the absence of the unlawful motive. Our decision today treads this well-worn path.

C

In saying that gender played a motivating part in an employment decision, we mean that, if we asked the employer at the moment of the decision what its reasons were and if we received a truthful response, one of those reasons would be that the applicant or employee was a woman.13 In 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.

Although the parties do not overtly dispute this last proposition, the placement by Price Waterhouse of “sex stereotyping” in quotation marks throughout its brief seems to us an insinuation either that such stereotyping was not present in this case or that it lacks legal relevance. We reject both pos*251sibilities. As to the existence of sex stereotyping in this case, we are not inclined to quarrel with the District Court’s conclusion that a number of the partners’ comments showed sex stereotyping at work. See infra, at 255-256. As for the legal relevance of sex stereotyping, we are beyond the day when an employer could evaluate employees by assuming or insisting that they matched the stereotype associated with their group, for “‘[i]n forbidding employers to discriminate against individuals because of their sex, Congress intended to strike at the entire spectrum of disparate treatment of men and women resulting from sex stereotypes.’” Los Angeles Dept. of Water and Power v. Manhart, 435 U. S. 702, 707, n. 13 (1978), quoting Sprogis v. United Air Lines, Inc., 444 F. 2d 1194, 1198 (CA7 1971). An employer who objects to aggressiveness in women but whose positions require this trait places women in an intolerable and impermissible catch 22: out of a job if they behave aggressively and out of a job if they do not. Title VII lifts women out of this bind.

Remarks at work that are based on sex stereotypes do not inevitably prove that gender played a part in a particular employment decision. The plaintiff must show that the employer actually relied on her gender in making its decision. In making this showing, stereotyped remarks can certainly be evidence that gender played a part. In any event, the stereotyping in this case did not simply consist of stray remarks. On the contrary, Hopkins proved that Price Water-house invited partners to submit comments; that some of the comments stemmed from sex stereotypes; that an important part of the Policy Board’s decision on Hopkins was an assessment of the submitted comments; and that Price Waterhouse in no way disclaimed reliance on the sex-linked evaluations. This is not, as Price Waterhouse suggests, “discrimination in the air”; rather, it is, as Hopkins puts it, “discrimination brought to ground and visited upon” an employee. Brief for Respondent 30. By focusing on Hopkins’ specific proof, however, we do not suggest a limitation on the possible ways *252of proving that stereotyping played a motivating role in an employment decision, and we refrain from deciding here which specific facts, “standing alone,” would or would not establish a plaintiff’s case, since such a decision is unnecessary in this case. But see post, at 277 (O’Connor, J., concurring in judgment).

As to the employer’s proof, in most cases, the employer should be able to present some objective evidence as to its probable decision in the absence of an impermissible motive.14 Moreover, proving “ ‘that the same decision would have been justified ... is not the same as proving that the same decision would have been made.’” Givhan, 439 U. S., at 416, quoting Ayers v. Western Line Consolidated School District, 555 F. 2d 1309, 1315 (CA5 1977). An employer may not, in other words, prevail in a mixed-motives case by offering a legitimate and sufficient reason for its decision if that reason did not motivate it at the time of the decision. Finally, an employer may not meet its burden in such a case by merely showing that at the time of the decision it was motivated only in part by a legitimate reason. The very premise of a mixed-motives case is that a legitimate reason was present, and indeed, in this case, Price Waterhouse already has made this showing by convincing Judge Gesell that Hopkins’ interpersonal problems were a legitimate concern. The employer instead must show that its legitimate reason, standing alone, would have induced it to make the same decision.

I — I I — I h-<

The courts below held that an employer who has allowed a discriminatory impulse to play a motivating part in an employment decision must prove by clear and convincing evidence that it would have made the same decision in the ab*253sence of discrimination. We are persuaded that the better rule is that the employer must make this showing by a preponderance of the evidence.

Conventional rules of civil litigation generally apply in Title VII cases, see, e. g., United States Postal Service Bd. of Governors v. Aikens, 460 U. S. 711, 716 (1983) (discrimination not to be “treat[ed]. . . differently from other ultimate questions of fact”), and one of these rules is that parties to civil litigation need only prove their case by a preponderance of the evidence. See, e. g., Herman & MacLean v. Huddleston, 459 U. S. 375, 390 (1983). Exceptions to this standard are uncommon, and in fact are ordinarily recognized only when the government seeks to take unusual coercive action— action more dramatic than entering an award of money damages or other conventional relief — against an individual. See Santosky v. Kramer, 455 U. S. 745, 756 (1982) (termination of parental rights); Addington v. Texas, 441 U. S. 418, 427 (1979) (involuntary commitment); Woodby v. INS, 385 U. S. 276 (1966) (deportation); Schneiderman v. United States, 320 U. S. 118, 122, 125 (1943) (denaturalization). Only rarely have we required clear and convincing proof where the action defended against seeks only conventional relief, see, e. g., Gertz v. Robert Welch, Inc., 418 U. S. 323, 342 (1974) (defamation), and we find it significant that in such cases it was the defendant rather than the plaintiff who sought the elevated standard of proof — suggesting that this standard ordinarily serves as a shield rather than, as Hopkins seeks to use it, as a sword.

It is true, as Hopkins emphasizes, that we have noted the “clear distinction between the measure of proof necessary to establish the fact that petitioner had sustained some damage and the measure of proof necessary to enable the jury to fix the amount.” Story Parchment Co. v. Paterson Parchment Paper Co., 282 U. S. 555, 562 (1931). Likewise, an Equal Employment Opportunity Commission (EEOC) regulation does require federal agencies proved to have violated *254Title VII to show by clear and convincing evidence that an individual employee is not entitled to relief. See 29 CFR § 1613.271(c)(2) (1988). And finally, it is true that we have emphasized the importance of make-whole relief for victims of discrimination. See Albemarle Paper Co. v. Moody, 422 U. S. 405 (1975). Yet each of these sources deals with the proper determination of relief rather than with the initial finding of liability. This is seen most easily in the EEOC’s regulation, which operates only after an agency or the EEOC has found that “an employee of the agency was discriminated against.” See 29 CFR § 1613.271(c) (1988). Because we have held that, by proving that it would have made the same decision in the absence of discrimination, the employer may avoid a finding of liability altogether and not simply avoid certain equitable relief, these authorities do not help Hopkins to show why we should elevate the standard of proof for an employer in this position.

Significantly, the cases from this Court that most resemble this one, Mt. Healthy and Transportation Management, did not require clear and convincing proof. Mt. Healthy, 429 U. S., at 287; Transportation Management, 462 U. S., at 400, 403. We are not inclined to say that the public policy against firing employees because they spoke out on issues of public concern or because they affiliated with a union is less important than the policy against discharging employees on the basis of their gender. Each of these policies is vitally important, and each is adequately served by requiring proof by a preponderance of the evidence.

Although Price Waterhouse does not concretely tell us how its proof was preponderant even if it was not clear and convincing, this general claim is implicit in its request for the less stringent standard. Since the lower courts required Price Waterhouse to make its proof by clear and convincing evidence, they did not determine whether Price Waterhouse had proved by a preponderance of the evidence that it would have placed Hopkins’ candidacy on hold even if it had not per*255mitted sex-linked evaluations to play a part in the decision-making process. Thus, we shall remand this case so that that determination can be made.

i>

The District Court found that sex stereotyping was permitted to play a part” in the evaluation of Hopkins as a candidate for partnership. 618 F. Supp., at 1120. Price Water-house disputes both that stereotyping occurred and that it played any part in the decision to place Hopkins’ candidacy on hold. In the firm’s view, in other words, the District Court’s factual conclusions are clearly erroneous. We do not agree.

In finding that some of the partners’ comments reflected sex stereotyping, the District Court relied in part on Dr. Fiske’s expert testimony. Without directly impugning Dr. Fiske’s credentials or qualifications, Price Waterhouse insinuates that a social psychologist is unable to identify sex stereotyping in evaluations without investigating whether those evaluations have a basis in reality. This argument comes too late. At trial, counsel for Price Waterhouse twice assured the court that he did not question Dr. Fiske’s expertise (App. 25) and failed to challenge the legitimacy of her discipline. Without contradiction from Price Waterhouse, Fiske testified that she discerned sex stereotyping in the partners’ evaluations of Hopkins, and she further explained that it was part of her business to identify stereotyping in written documents. Id., at 64. We are not inclined to accept petitioner’s belated and unsubstantiated characterization of Dr. Fiske’s testimony as “gossamer evidence” (Brief for Petitioner 20) based only on “intuitive hunches” (id., at 44) and of her detection of sex stereotyping as “intuitively divined” (id., at 43). Nor are we disposed to adopt the dissent’s dismissive attitude toward Dr. Fiske’s field of study and toward her own professional integrity, see post, at 293-294, n. 5.

*256Indeed, we are tempted to say that Dr. Fiske’s expert testimony was merely icing on Hopkins’ cake. It takes no special training to discern sex stereotyping in a description of an aggressive female employee as requiring “a course at charm school.” Nor, turning to Thomas Beyer’s memorable advice to Hopkins, does it require expertise in psychology to know that, if an employee’s flawed “interpersonal skills” can be corrected by a soft-hued suit or a new shade of lipstick, perhaps it is the employee’s sex and not her interpersonal skills that has drawn the criticism.15

Price Waterhouse also charges that Hopkins produced no evidence that sex stereotyping played a role in the decision to place her candidacy on hold. As we have stressed, however, Hopkins showed that the partnership solicited evaluations from all of the firm’s partners; that it generally relied very heavily on such evaluations in making its decision; that some of the partners’ comments were the product of stereotyping; and that the firm in no way disclaimed reliance on those particular comments, either in Hopkins’ case or in the past. Certainly a plausible — and, one might say, inevitable — conclusion to draw from this set of circumstances is that the Policy Board in making its decision did in fact take into account all of the partners’ comments, including the comments that were motivated by stereotypical notions about women’s proper deportment.16

*257Price Waterhouse concedes that the proof in Transportation Management adequately showed that the employer there had relied on an impermissible motivation in firing the plaintiff. Brief for Petitioner 45. But the only evidence in that case that a discriminatory motive contributed to the plaintiff’s discharge was that the employer harbored a grudge toward the plaintiff on account of his union activity; there was, contrary to Price Waterhouse’s suggestion, no direct evidence that that grudge had played a role in the decision, and, in fact, the employer had given other reasons in explaining the plaintiff’s discharge. See 462 U. S., at 396. If the partnership considers that proof sufficient, we do not know why it takes such vehement issue with Hopkins’ proof.

Nor is the finding that sex stereotyping played a part in the Policy Board’s decision undermined by the fact that many of the suspect comments were made by supporters rather than detractors of Hopkins. A negative comment, even when made in the context of a generally favorable review, nevertheless may influence the decisionmaker to think less highly of the candidate; the Policy Board, in fact, did not simply tally the “yesses” and “noes” regarding a candidate, but carefully reviewed the content of the submitted comments. The additional suggestion that the comments were made by “persons outside the decisionmaking chain” (Brief for Petitioner 48) — and therefore could not have harmed Hopkins — simply ignores the critical role that partners’ comments played in the Policy Board’s partnership decisions.

Price Waterhouse appears to think that we cannot affirm the factual findings of the trial court without deciding that, instead of being overbearing and aggressive and curt, Hopkins is, in fact, kind and considerate and patient. If this is indeed its impression, petitioner misunderstands the theory *258on which Hopkins prevailed. The District Judge acknowledged that Hopkins’ conduct justified complaints about her behavior as a senior manager. But he also concluded that the reactions of at least some of the partners were reactions to her as a woman manager. Where an evaluation is based on a subjective assessment of a person’s strengths and weaknesses, it is simply not true that each evaluator will focus on, or even mention, the sáme weaknesses. Thus, even if we knew that Hopkins had “personality problems,” this would not tell us that the partners who cast their evaluations of Hopkins in sex-based terms would have criticized her as sharply (or criticized her at all) if she had been a man. It is not our job to review the evidence and decide that the negative reactions to Hopkins were based on reality; our perception of Hopkins’ character is irrelevant. We sit not to determine whether Ms. Hopkins is nice, but to decide whether the partners reacted negatively to her personality because she is a woman.

V

We hold that when a plaintiff in a Title VII case proves that her gender played a motivating part in an employment decision, the defendant may avoid a finding of liability only by proving by a preponderance of the evidence that it would have made the same decision even if it had not taken the plaintiff’s gender into account. Because the courts below erred by deciding that the defendant must make this proof by clear and convincing evidence, we reverse the Court of Appeals’ judgment against Price Waterhouse on liability and remand the case to that court for further proceedings.

It is so ordered.

Justice White,

concurring in the judgment.

In my view, to determine the proper approach to causation in this case, we need look only to the Court’s opinion in Mt. Healthy City Bd. of Ed. v. Doyle, 429 U. S. 274 (1977). In Mt. Healthy, a public employee was not rehired, in part *259because of his exercise of First Amendment rights and in part because of permissible considerations. The Court rejected a rule of causation that focused “solely on whether protected conduct played a part, ‘substantial’ or otherwise, in a decision not to rehire,” on the grounds that such a rule could make the employee better off by exercising his constitutional rights than by doing nothing at all. Id., at 285. Instead, the Court outlined the following approach:

“Initially, in this case, the burden was properly placed upon respondent to show that his conduct was constitutionally protected, and that his conduct was a ‘substantial factor’ — or, to put it in other words, that it was a ‘motivating factor’ in the Board’s decision not to rehire him. Respondent having carried that burden, however, the District Court should have gone on to determine whether the Board had shown by a preponderance of the evidence that it would have reached the same decision as to respondent’s reemployment even in the absence of the protected conduct.” Id., at 287 (footnote omitted).

It is not necessary to get into semantic discussions on whether the Mt. Healthy approach is “but-for” causation in another guise or creates an affirmative defense on the part of the employer to see its clear application to the issues before us in this case. As in Mt. Healthy, the District Court found that the employer was motivated by both legitimate and illegitimate factors. And here, as in Mt. Healthy, and as the Court now holds, Hopkins was not required to prove that the illegitimate factor was the only, principal, or true reason for petitioner’s action. Rather, as Justice O’Connor states, her burden was to show that the unlawful motive was a substantial factor in the adverse employment action. The District Court, as its opinion was construed by the Court of Appeals, so found, 263 U. S. App. D. C. 321, 333, 334, 825 F. 2d 458, 470, 471 (1987), and I agree that the finding was supported by the record. The burden of persuasion then *260should have shifted to Price Waterhouse to prove “by a preponderance of the evidence that it would have reached the same decision ... in the absence of” the unlawful motive. Mt. Healthy, supra, at 287.

I agree with Justice Brennan that applying this approach to causation in Title VII cases is not a departure from, and does not require modification of, the Court’s holdings in Texas Dept. of Community Affairs v. Burdine, 450 U. S. 248 (1981), and McDonnell Douglas Corp. v. Green, 411 U. S. 792 (1973). The Court has made clear that “mixed-motives” cases, such as the present one, are different from pretext cases such as McDonnell Douglas and Burdine. In pretext cases, “the issue is whether either illegal or legal motives, but not both, were the ‘true’ motives behind the decision.” NLRB v. Transportation Management Corp., 462 U. S. 393, 400, n. 5 (1983). In mixed-motives cases, however, there is no one “true” motive behind the decision. Instead, the decision is a result of multiple factors, at least one of which is legitimate. It can hardly be said that our decision in this case is a departure from cases that are “inapposite.” Ibid. 1 also disagree with the dissent’s assertion that this approach to causation is inconsistent with our statement in Burdine that “[t]he ultimate burden of persuading the trier of fact that the defendant intentionally discriminated against the plaintiff remains at all times with the plaintiff.” 450 U. S., at 253. As we indicated in Transportation Management Corp., the showing required by Mt. Healthy does not improperly shift from the plaintiff the ultimate burden of persuasion on whether the defendant intentionally discriminated against him or her. See 462 U. S., at 400, n. 5.

Because the Court of Appeals required Price Waterhouse to prove by clear and convincing evidence that it would have reached the same employment decision in the absence of the improper motive, rather than merely requiring proof by a preponderance of the evidence as in Mt. Healthy, I concur in the judgment reversing this case in part and remanding. *261With respect to the employer’s burden, however, the plurality seems to require, at least in most cases, that the employer submit objective evidence that the same result would have occurred absent the unlawful motivation. Ante, at 252. In my view, however, there is no special requirement that the employer carry its burden by objective evidence. In a mixed-motives case, where the legitimate motive found would have been ample grounds for the action taken, and the employer credibly testifies that the action would have been taken for the legitimate reasons alone, this should be ample proof. This would even more plainly be the case where the employer denies any illegitimate motive in the first place but the court finds that illegitimate, as well as legitimate, factors motivated the adverse action.*

Justice O’Connor,

concurring in the judgment.

I agree with the plurality that, on the facts presented in this ease, the burden of persuasion should shift to the employer to demonstrate by a preponderance of the evidence that it would have reached the same decision concerning Ann Hopkins’ candidacy absent consideration of her gender. I further agree that this burden shift is properly part of the liability phase of the litigation. I thus concur in the judgment of the Court. My disagreement stems from the plurality’s conclusions concerning the substantive requirement of causation under the statute and its broad statements regarding the applicability of the allocation of the burden of proof applied in this case. The evidentiary rule the Court adopts today should be viewed as a supplement to the careful framework established by our unanimous decisions in McDonnell Douglas Corp. v. Green, 411 U. S. 792 (1973), and Texas Dept. of Community Affairs v. Burdine, 450 U. S. 248 (1981), for use in cases such as this one where the employer has created uncertainty as to causation by knowingly giving *262substantial weight to an impermissible criterion. I write separately to explain why I believe such a departure from the McDonnell Douglas standard is justified in the circumstances presented by this and like cases, and to express my views as to when and how the strong medicine of requiring the employer to bear the burden of persuasion on the issue of causation should be administered.

I

Title VII provides in pertinent part: “It shall be an unlawful employment practice for an employer ... to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual’s race, color, religion, sex, or national origin.” 42 U. S. C. §2000e-2(a) (emphasis added). The legislative history of Title VII bears out what its plain language suggests: a substantive violation of the statute only occurs when consideration of an illegitimate criterion is the “but-for” cause of an adverse employment action. The legislative history makes it clear that Congress was attempting to' eradicate discriminatory actions in the employment setting, not mere discriminatory thoughts. Critics of the bill that became Title VII labeled it a “thought control bill,” and argued that it created a “punishable crime that does not require an illegal external act as a basis for judgment.” 100 Cong. Rec. 7254 (1964) (remarks of Sen. Ervin). Senator Case, whose views the plurality finds so persuasive elsewhere, responded:

“The man must do or fail to do something in regard to employment. There must be some specific external act, more than a mental act. Only if he does the act because of the grounds stated in the bill would there be any legal consequences.” Ibid.

Thus, I disagree with the plurality’s dictum that the words “because of” do not mean “but-for” causation; manifestly they *263do. See Sheet Metal Workers v. EEOC, 478 U. S. 421, 499 (1986) (White, J., dissenting) (“[T]he general policy under Title VII is to limit relief for racial discrimination in employment practices to actual victims of the discrimination”). We should not, and need not, deviate from that policy today. The question for decision in this case is what allocation of the burden of persuasion on the issue of causation best conforms with the intent of Congress and the purposes behind Title VII.

The evidence of congressional intent as to which party should bear the burden of proof on the issue of causation is considerably less clear. No doubt, as a general matter, Congress assumed that the plaintiff in a Title VII action would bear the burden of proof on the elements critical to his or her case. As the dissent points out, post, at 287, n. 3, the interpretative memorandum submitted by sponsors of Title VII indicates that “the plaintiff, as in any civil case, would have the burden of proving that discrimination had occurred.” 110 Cong. Rec. 7214 (1964) (emphasis added). But in the area of tort liability, from whence the dissent’s “but-for” standard of causation is derived, see post, at 282, the law has long recognized that in certain “civil cases” leaving the burden of persuasion on the plaintiff to prove “but-for” causation would be both unfair and destructive of the deterrent purposes embodied in the concept of duty of care. Thus, in multiple causation cases, where a breach of duty has been established, the common law of torts has long shifted the burden of proof to multiple defendants to prove that their negligent actions were not the “but-for” cause of the plaintiff’s injury. See e. g., Summers v. Tice, 33 Cal. 2d 80, 84-87, 199 P. 2d 1, 3-4 (1948). The same rule has been applied where the effect of a defendant’s tortious conduct combines with a force of unknown or innocent origin to produce the harm to the plaintiff. See Kingston v. Chicago & N. W. R. Co., 191 Wis. 610, 616, 211 N. W. 913, 915 (1927) (“Granting that the union of that fire [caused by defendant’s *264negligence] with another of natural origin, or with another of much greater proportions, is available as a defense, the burden is on the defendant to show that. . . the fire set by him was not the proximate cause of the damage”). See also 2 J. Wigmore, Select Cases on the Law of Torts § 153, p. 865 (1912) (“When two or more persons by their acts are possibly the sole cause of a harm, or when two or more acts of the same person are possibly the sole cause, and the plaintiff has introduced evidence that one of the two persons, or one of the same person’s two acts, is culpable, then the defendant has the burden of proving that the other person, or his other act, was the sole cause of the harm”).

While requiring that the plaintiff in a tort suit or a Title VII action prove that the defendant’s “breach of duty” was the “but-for” cause of an injury does not generally hamper effective enforcement of the policies behind those causes of action,

“at other times the [but-for] test demands the impossible. It challenges the imagination of the trier to probe into a purely fanciful and unknowable state of affairs. He is invited-to make an estimate concerning facts that concededly never existed. The very uncertainty as to what might have happened opens the door wide for conjecture. But when conjecture is demanded it can be given a direction that is consistent with the policy considerations that underlie the controversy.” Malone, Ruminations on Cause-In-Fact, 9 Stan. L. Rev. 60, 67 (1956).

Like the common law of torts, the statutory employment “tort” created by Title VII has two basic purposes. The first is to deter conduct which has been identified as contrary to public policy and harmful to society as a whole. As we have noted in the past, the award of backpay to a Title VII plaintiff provides “the spur or catalyst which causes employers and unions to self-examine and to self-evaluate their employment practices and to endeavor to eliminate, so far as *265possible, the last vestiges” of discrimination in employment. Albemarle Paper Co. v. Moody, 422 U. S. 405, 417-418 (1975) (citation omitted). The second goal of Title VII is “to make persons whole for injuries suffered on account of unlawful employment discrimination.” Id., at 418.

Both these goals are reflected in the elements of a disparate treatment action. There is no doubt that Congress considered reliance on gender or race in making employment decisions an evil in itself. As Senator Clark put it, “[t]he bill simply eliminates consideration of color [or other forbidden criteria] from the decision to hire or promote.” 110 Cong. Rec. 7218 (1964). See also id., at 13088 (remarks of Sen. Humphrey) (“What the bill does ... is simply to make it an illegal practice to use race as a factor in denying employment”). Reliance on such factors is exactly what the threat of Title VII liability was meant to deter. While the main concern of the statute was with employment opportunity, Congress was certainly not blind to the stigmatic harm which comes from being evaluated by a process which treats one as an inferior by reason of one’s race or sex. This Court’s decisions under the Equal Protection Clause have long recognized that whatever the final outcome of a decisional process, the inclusion of race or sex as a consideration within it harms both society and the individual. See Richmond v. J. A. Croson Co., 488 U. S. 469 (1989). At the same time, Congress clearly conditioned legal liability on a determination that the consideration of an illegitimate factor caused a tangible employment injury of some kind.

Where an individual disparate treatment plaintiff has shown by a preponderance of the evidence that an illegitimate criterion was a substantial factor in an adverse employment decision, the deterrent purpose of the statute has clearly been triggered. More importantly, as an evidentiary matter, a reasonable factfinder could conclude that absent further explanation, the employer’s discriminatory motivation “caused” the employment decision. The employer has *266not yet been shown to be a violator, but neither is it entitled to the same presumption of good faith concerning its employment decisions which is accorded employers facing only circumstantial evidence of discrimination. Both the policies behind the statute, and the evidentiary principles developed in the analogous area of causation in the law of torts, suggest that at this point the employer may be required to convince the factfinder that, despite the smoke, there is no fire.

We have given recognition to these principles in our cases which have discussed the “remedial phase” of class action disparate treatment cases. Once the class has established that discrimination against a protected group was essentially the employer’s “standard practice,” there has been harm to the group and injunctive relief is appropriate. But as to the individual members of the class, the liability phase of the litigation is not complete. See Dillon v. Coles, 746 F. 2d 998, 1004 (CA3 1984) (“It is misleading to speak of the additional proof required by an individual class member for relief as being a part of the damage phase, that evidence is actually an element of the liability portion of the case”) (footnote omitted). Because the class has already demonstrated that, as a rule, illegitimate factors were considered in the employer’s decisions, the burden shifts to the employer “to demonstrate that the individual applicant was denied an employment opportunity for legitimate reasons.” Teamsters v. United States, 431 U. S. 324, 362 (1977). See also Franks v. Bowman Transportation Co., 424 U. S. 747, 772 (1976).

The individual members of a class action disparate treatment case stand in much the same position as Ann Hopkins here. There has been a strong showing that the employer has done exactly what Title VII forbids, but the connection between the employer’s illegitimate motivation and any injury to the individual plaintiff is unclear. At this point calling upon the employer to show that despite consideration of illegitimate factors the individual plaintiff would not have been hired or promoted in any event hardly seems “unfair” or *267contrary to the substantive command of the statute. In fact, an individual plaintiff who has shown that an illegitimate factor played a substantial role in the decision in his or her case has proved more than the class member in a Teamsters type action. The latter receives the benefit of a burden shift to the defendant based on the likelihood that an illegitimate criterion was a factor in the individual employment decision.

There is a tension between the Franks and Teamsters line of decisions and the individual disparate treatment cases cited by the dissent. See post, at 286-289. Logically, under the dissent’s view, each member of a disparate treatment class action would have to show “but-for” causation as to his or her individual employment decision, since it is not an element of the pattern or practice proof of the entire class and it is statutorily mandated that the plaintiff bear the burden of proof on this issue throughout the litigation. While the Court has properly drawn a distinction between the elements of a class action claim and an individual disparate treatment claim, see Cooper v. Federal Reserve Bank of Richmond, 467 U. S. 867, 873-878 (1984), and I do not suggest the wholesale transposition of rules from one setting to the other, our decisions in Teamsters and Franks do indicate a recognition that presumptions shifting the burden of persuasion based on evidentiary probabilities and the policies behind the statute are not alien to our Title VII jurisprudence.

Moreover, placing the burden on the defendant in this case to prove that the same decision would have been justified by legitimate reasons is consistent with our interpretation of the constitutional guarantee of equal protection. Like a disparate treatment plaintiff, one who asserts that governmental action violates the Equal Protection Clause must show that he or she is “the victim of intentional discrimination.” Burdine, 450 U. S., at 256. Compare post, at 286, 289 (Kennedy, J., dissenting), with Washington v. Davis, 426 U. S. 229, 240 (1976). In Alexander v. Louisiana, 405 U. S. 625 (1972), we dealt with a criminal defendant’s allegation that *268members of his race had been invidiously excluded from the grand jury which indicted him in violation of the Equal Protection Clause. In addition to the statistical evidence presented by petitioner in that case, we noted that the State’s “selection procedures themselves were not racially neutral.” Id., at 630. Once the consideration of race in the decisional process had been established, we held that “the burden of proof shifts to the State to rebut the presumption of unconstitutional action by showing that permissible racially neutral selection criteria and procedures have produced the monochromatic result.” Id., at 632.

We adhered to similar principles in Arlington Heights v. Metropolitan Housing Development Corp., 429 U. S. 252 (1977), a case which, like this one, presented the problems of motivation and causation in the context of a multimember decisionmaking body authorized to consider a wide range of factors in arriving at its decisions. In Arlington Heights a group of minority plaintiffs claimed that a municipal governing body’s refusal to rezone a plot of land to allow for the construction of low-income integrated housing was racially motivated. On the issue of causation, we indicated that the plaintiff was not required

“to prove that the challenged action rested solely on racially discriminatory purposes. Rarely can it be said that a legislature or administrative body operating under a broad mandate made a decision motivated solely by a single concern, or even that a particular purpose was the ‘dominant’ or ‘primary’ one. In fact, it is because legislators and administrators are properly concerned with balancing numerous competing considerations that courts refrain from reviewing the merits of their decisions, absent a showing of arbitrariness or irrationality. But racial discrimination is not just another competing consideration. When there is a proof that a discriminatory purpose has been a motivating factor in the deci*269sion, this judicial deference is no longer justified.” Id., at 265-266 (citation omitted).

If the strong presumption of regularity and rationality of legislative decisionmaking must give way in the face of evidence that race has played a significant part in a legislative decision, I simply cannot believe that Congress intended Title VII to accord more deference to a private employer in the face of evidence that its decisional process has been substantially infected by discrimination. Indeed, where a public employee brings a “disparate treatment” claim under 42 U. S. C. § 1983 and the Equal Protection Clause the employee is entitled to the favorable evidentiary framework of Arlington Heights. See, e. g., Hervey v. Little Rock, 787 F. 2d 1223, 1233-1234 (CA8 1986) (applying Arlington Heights to public employee’s claim of sex discrimination in promotion decision); Lee v. Russell County Bd. of Education, 684 F. 2d 769, 773-774 (CA11 1982) (applying Arlington Heights to public employees’ claims óf race discrimination in discharge case). Under the dissent’s reading of Title VII, Congress’ extension of the coverage of the statute to public employers in 1972 has placed these employees under a less favorable evidentiary regime. In my view, nothing in the language, history, or purpose of Title VII prohibits adoption of an evidentiary rule which places the burden of persuasion on the defendant to demonstrate that legitimate concerns would have justified an adverse employment action where the plaintiff has convinced the factfinder that a forbidden factor played a substantial role in the employment decision. Even the dissenting judge below “[had] no quarrel with [the] principle” that “a party with one permissible motive and one unlawful one may prevail only by affirmatively proving that it would have acted as it did even if the forbidden motive were absent.” 263 U. S. App. D. C. 321, 341, 825 F. 2d 458, 478 (1987) (Williams, J. dissenting).

*270II

The dissent’s summary of our individual disparate treatment cases to date is fair and accurate, and amply demonstrates that the rule we adopt today is at least a change in direction from some of our prior precedents. See post, at 286-289. We have indeed emphasized in the past that in an individual disparate treatment action the plaintiff bears the burden of persuasion throughout the litigation. Nor have we confined the word “pretext” to the narrow definition which the plurality attempts to pin on it today. See ante, at 244-247. McDonnell Douglas and Burdine clearly contemplated that a disparate treatment plaintiff could show that the employer’s proffered explanation for an event was not “the true reason” either because it never motivated the employer in its employment decisions or because it did not do so in a particular case. McDonnell Douglas and Burdine assumed that the plaintiff would bear the burden of persuasion as to both these attacks, and we clearly depart from that framework today. Such a departure requires justification, and its outlines should be carefully drawn.

First, McDonnell Douglas itself dealt with a situation where the plaintiff presented no direct evidence that the employer had relied on a forbidden factor under Title VII in making an employment decision. The prima facie case established there was not difficult to prove, and was based only on the statistical probability that when a number of potential causes for an employment decision are eliminated an inference arises that an illegitimate factor was in fact the motivation behind the decision. See Teamsters, 431 U. S., at 358, n. 44 (“[T]he McDonnell Douglas formula does not require direct proof of discrimination”). In the face of this inferential proof, the employer’s burden was deemed to be only one of production; the employer must articulate a legitimate reason for the adverse employment action. See Furnco Construction Corp. v. Waters, 438 U. S. 567, 577 (1978). The plaintiff must then be given an “opportunity to demonstrate *271by competent evidence that the presumptively valid reasons for his rejection were in fact a coverup for a racially discriminatory decision.” McDonnell Douglas, 411 U. S., at 805. Our decision in Texas Dept. of Community Affairs v. Burdine, 450 U. S. 248 (1981), also involved the “narrow question” whether, after a plaintiff had carried the “not onerous” burden of establishing the prima facie case under McDonnell Douglas, the burden of persuasion should be shifted to the employer to prove that a legitimate reason for the adverse employment action existed. 450 U. S., at 250. As the discussion of Teamsters and Arlington Heights indicates, I do not think that the employer is entitled to the same presumption of good faith where there is direct evidence that it has placed substantial reliance on factors whose consideration is forbidden by Title VII.

The only individual disparate treatment case cited by the dissent which involved the kind of direct evidence of discriminatory animus with which we are confronted here is United States Postal Service Bd. of Governors v. Aikens, 460 U. S. 711, 713-714, n. 2 (1983). The question presented to the Court in that case involved only a challenge to the elements of the prima facie case under McDonnell Douglas and Burdine, see Pet. for Cert. in United States Postal Service Bd. of Governors v. Aikens, O. T. 1981, No. 81-1044, and the question we confront today was neither briefed nor argued to the Court. As should be apparent, the entire purpose of the McDonnell Douglas prima facie case is to compensate for the fact that direct evidence of intentional discrimination is hard to come by. That the employer’s burden in rebutting such an inferential case of discrimination is only one of production does not mean that the scales should be weighted in the same manner where there is direct evidence of intentional discrimination. Indeed, in one Age Discrimination in Employment Act case, the Court seemed to indicate that “the McDonnell Douglas test is inapplicable where the plaintiff presents direct evidence of discrimination.” Trans World *272Airlines, Inc. v. Thurston, 469 U. S. 111, 121 (1985). See also East Texas Motor Freight System, Inc. v. Rodriguez, 431 U. S. 395, 403-404, n. 9 (1977).

Second, the facts of this case, and a growing number like it decided by the Courts of Appeals, convince me that the evidentiary standard I propose is necessary to make real the promise of McDonnell Douglas that “[i]n the implementation of [employment] decisions, it is abundantly clear that Title VII tolerates no . . . discrimination, subtle or otherwise.” 411 U. S., at 801. In this case, the District Court found that a number of the evaluations of Ann Hopkins submitted by partners in the firm overtly referred to her failure to conform to certain gender stereotypes as a factor militating against her election to the partnership. 618 F. Supp. 1109, 1116— 1117 (DC 1985). The District Court further found that these evaluations were given “great weight” by the decisionmakers at Price Waterhouse. Id., at 1118. In addition, the District Court found that the partner responsible for informing Hopkins of the factors which caused her candidacy to be placed on hold, indicated that her “professional” problems would be solved if she would “walk more femininely, talk more femininely, wear make-up, have her hair styled, and wear jewelry.” Id., at 1117 (footnote omitted). As the Court of Appeals characterized it, Ann Hopkins proved that Price Waterhouse “permitted] stereotypical attitudes towards women to play a significant, though unquantifiable, role in its decision not to invite her to become a partner.” 263 U. S. App. D. C., at 324, 825 F. 2d, at 461.

At this point Ann Hopkins had taken her proof as far as it could go. She had proved discriminatory input into the decisional process, and had proved that participants in the process considered her failure to conform to the stereotypes credited by a number of the decisionmakers had been a substantial factor in the decision. It is as if Ann Hopkins were sitting in the hall outside the room where partnership decisions were being made. As the partners filed in to consider *273her candidacy, she heard several of them make sexist remarks in discussing her suitability for partnership. As the decisionmakers exited the room, she was told by one of those privy to the decisionmaking process that her gender was a major reason for the rejection of her partnership bid. If, as we noted in Teamsters, “[presumptions shifting the burden of proof are often created to reflect judicial evaluations of probabilities and to conform with a party’s superior access to the proof,” 431 U. S., at 359, n. 45, one would be hard pressed to think of a situation where it would be more appropriate to require the defendant to show that its decision would have been justified by wholly legitimate concerns.

Moreover, there is mounting evidence in the decisions of the lower courts that respondent here is not alone in her inability to pinpoint discrimination as the precise cause of her injury, despite having shown that it played a significant role in the decisional process. Many of these courts, which deal with the evidentiary issues in Title VII cases on a regular basis, have concluded that placing the risk of nonpersuasion on the defendant in a situation where uncertainty as to causation has been created by its consideration of an illegitimate criterion makes.sense as a rule of evidence and furthers the substantive command of Title VII. See, e. g., Bell v. Birmingham Linen Service, 715 F. 2d 1552, 1556 (CA11 1983) (Tjoflat, J.) (“It would be illogical, indeed ironic, to hold a Title VII plaintiff presenting direct evidence of a defendant’s intent to discriminate to a more stringent burden of proof, or to allow a defendant to meet that direct proof by merely articulating, but not proving, legitimate, nondiscriminatory reasons for its action”). Particularly in the context of the professional world, where decisions are often made by collegial bodies on the basis of largely subjective criteria, requiring the plaintiff to prove that any one factor was the definitive cause of the decisionmakers’ action may be tantamount to declaring Title VII inapplicable to such decisions. See, e. g., Fields v. Clark University, 817 F. 2d 931, 935-937 *274(CA1 1987) (where plaintiff produced “strong evidence” that sexist attitudes infected faculty tenure decision, burden properly shifted to defendant to show that it would have reached the same decision absent discrimination); Thompkins v. Morris Brown College, 752 F. 2d 558, 563 (CA11 1985) (direct evidence of discriminatory animus in decision to discharge college professor shifted burden of persuasion to defendant).

Finally, I am convinced that a rule shifting the burden to the defendant where the plaintiff has shown that an illegitimate criterion was a “substantial factor” in the employment decision will not conflict with other congressional policies embodied in Title VII. Title VII expressly provides that an employer need not give preferential treatment to employees or applicants of any race, color, religion, sex, or national origin in order to maintain a work force in balance with the general population. See 42 U. S. C. §2000e-2(j). The interpretive memorandum, whose authoritative force is noted by the plurality, see ante, at 243, n. 8, specifically provides: “There is no requirement in title VII that an employer maintain a racial balance in his work force. On the contrary, any deliberate attempt to maintain a racial balance, whatever such a balance may be, would involve a violation of title VII because maintaining such a balance would require an employer to hire or refuse to hire on the basis of race.” 110 Cong. Rec. 7213 (1964).

Last Term, in Watson v. Fort Worth Bank & Trust, 487 U. S. 977 (1988), the Court unanimously concluded that the disparate impact analysis first enunciated in Griggs v. Duke Power Co., 401 U. S. 424 (1971), should be extended to subjective or discretionary selection processes. At the same time a plurality of the Court indicated concern that the focus on bare statistics in the disparate impact setting could force employers to adopt “inappropriate prophylactic measures” in violation of § 2000e-2(j). The plurality went on to emphasize that in a disparate impact case, the plaintiff may not simply *275point to a statistical disparity in the employer’s work force. Instead, the plaintiff must identify a particular employment practice and “must offer statistical evidence of a kind and degree sufficient to show that the practice in question has caused the exclusion of applicants for jobs or promotions because of their membership in a protected group.” 487 U. S., at 994. The plurality indicated that “the ultimate burden of proving that discrimination against a protected group has been caused by a specific employment practice remains with the plaintiff at all times.” Id., at 997.

I believe there are significant differences between shifting the burden of persuasion to the employer in a case resting purely on statistical proof as in the disparate impact setting and shifting the burden of persuasion in a case like this one, where an employee has demonstrated by direct evidence that an illegitimate factor played a substantial role in a particular employment decision. First, the explicit consideration of race, color, religion, sex, or national origin in making employment decisions “was the most obvious evil Congress had in mind when it enacted Title VII.” Teamsters, 431 U. S., at 335, n. 15. While the prima facie case under McDonnell Douglas and the statistical showing of imbalance involved in a disparate impact case may both be indicators of discrimination or its “functional equivalent,” they are not, in and of themselves, the evils Congress sought to eradicate from the employment setting. Second, shifting the burden of persuasion to the employer in a situation like this one creates no incentive to preferential treatment in violation of §2000e-(2)(j). To avoid bearing the burden of justifying its decision, the employer need not seek racial or sexual balance in its work force; rather, all it need do is avoid substantial reliance on forbidden criteria in making its employment decisions.

While the danger of forcing employers to engage in unwarranted preferential treatment is thus less dramatic in this setting than in the situation the Court faced in Watson, it is far from wholly illusory. Based on its misreading of *276the words “because of” in the statute, see ante, at 240-242, the plurality appears to conclude that if a decisional process is “tainted” by awareness of sex or race in any way, the employer has violated the statute, and Title VII thus commands that the burden shift to the employer to justify its decision. Ante, at 250-252. The plurality thus effectively reads the causation requirement out of the statute, and then replaces it with an “affirmative defense.” Ante, at 244-247.

In my view, in order to justify shifting the burden on the issue of causation to the defendant, a disparate treatment plaintiff must show by direct evidence that an illegitimate criterion was a substantial factor in the decision. As the Court of Appeals noted below: “While most circuits have not confronted the question squarely, the consensus among those that have is that once a Title VII plaintiff has demonstrated by direct evidence that discriminatory animus played a significant or substantial role in the employment decision, the burden shifts to the employer to show that the decision would have been the same absent discrimination.” 263 U. S. App. D. C., at 333-344, 825 F. 2d, at'470-471. Requiring that the plaintiff demonstrate that an illegitimate factor played a substantial role in the employment decision identifies those employment situations where the deterrent purpose of Title VII is most clearly implicated. As an evidentiary matter, where a plaintiff has made this type of strong showing of illicit motivation, the factfinder is entitled to presume that the employer’s discriminatory animus made a difference to the outcome, absent proof to the contrary from the employer. Where a disparate treatment plaintiff has made such a showing, the burden then rests with the employer to convince the trier of fact that it is more likely than not that the decision would have been the same absent consideration of the illegitimate factor. The employer need not isolate the sole cause for the decision; rather it must demonstrate that with the illegitimate factor removed from the calculus, sufficient business reasons would have induced it to take the same employment *277action. This evidentiary scheme essentially requires the employer to place the employee in the same position he or she would have occupied absent discrimination. Cf. Mt. Healthy City Bd. of Ed. v. Doyle, 429 U. S. 274, 286 (1977). If the employer fails to carry this burden, the factfinder is justified in concluding that the decision was made “because of” consideration of the illegitimate factor and the substantive standard for liability under the statute is satisfied.

Thus, stray remarks in the workplace, while perhaps probative of sexual harassment, see Meritor Savings Bank v. Vinson, 477 U. S. 57, 63-69 (1986), cannot justify requiring the employer to prove that its hiring or promotion decisions were based on legitimate criteria. Nor can statements by nondecisionmakers, or statements by decisionmakers unrelated to the decisional process itself, suffice to satisfy the plaintiff’s burden in this regard. In addition, in my view testimony such as Dr. Fiske’s in this case, standing alone, would not justify shifting the burden of persuasion to the employer. Race and gender always “play a role” in an employment decision in the benign sense that these are human characteristics of which decisionmakers are aware and about which they may comment in a perfectly neutral and nondiscriminatory fashion. For example, in the context of this case, a mere reference to “a lady candidate” might show that gender “played a role” in the decision, but by no means could support a rational factfinder’s inference that the decision was made “because of” sex. What is required is what Ann Hopkins showed here: direct evidence that decisionmakers placed substantial negative reliance on an illegitimate criterion in reaching their decision.

It should be obvious that the threshold standard I would adopt for shifting the burden of persuasion to the defendant differs substantially from that proposed by the plurality, the plurality’s suggestion to the contrary notwithstanding. See ante, at 250, n. 13. The plurality proceeds from the premise that the words “because of” in the statute do not embody any *278causal requirement at all. Under my approach, the plaintiff must produce evidence sufficient to show that an illegitimate criterion was a substantial factor in the particular employment decision such that a reasonable factfinder could draw an inference that the decision was made “because of” the plaintiff’s protected status. Only then would the burden of proof shift to the defendant to prove that the decision would have been justified by other, wholly legitimate considerations. See also ante, at 259-260 (White, J., concurring in judgment).

In sum, because of the concerns outlined above, and because I believe that the deterrent purpose of Title VII is disserved by a rule which places the burden of proof on plaintiffs on the issue of causation in all circumstances, I would retain but supplement the framework we established in McDonnell Douglas and subsequent cases. The structure of the presentation of evidence in an individual disparate treatment case should conform to the general outlines we established in McDonnell Douglas and Burdine. First, the plaintiff must establish the McDonnell Douglas prima facie case by showing membership in a protected group, qualification for the job, rejection for the position, and that after rejection the employer continued to seek applicants of complainant’s general qualifications. McDonnell Douglas, 411 U. S., at 802. The plaintiff should also present any direct evidence of discriminatory animus in the decisional process. The defendant should then present its case, including its evidence as to legitimate, nondiscriminatory reasons for the employment decision. As the dissent notes, under this framework, the employer “has every incentive to convince the trier of fact that the decision was lawful.” Post, at 292, citing Burdine, 450 U. S., at 258. Once all the evidence has been received, the court should determine whether the McDonnell Douglas or Price Waterhouse framework properly applies to the evidence before it. If the plaintiff has failed to satisfy the Price Waterhouse threshold, the case should be decided under the principles enunciated in McDonnell Douglas and Burdine, *279with the plaintiff bearing the burden of persuasion on the ultimate issue whether the employment action was taken because of discrimination. In my view, such a system is both fair and workable, and it calibrates the evidentiary requirements demanded of the parties to the goals behind the statute itself.

I agree with the dissent, see post, at 293, n. 4, that the evidentiary framework I propose should be available to all disparate treatment plaintiffs where an illegitimate consideration played a substantial role in an adverse employment decision. The Court’s allocation of the burden of proof in Johnson v. Transportation Agency, Santa Clara County, 480 U. S. 616, 626-627 (1987), rested squarely on “the analytical framework set forth in McDonnell Douglas,” id., at 626, which we alter today. It would be odd to say the least if the evidentiary rules applicable to Title VII actions were themselves dependent on the gender or the skin color of the litigants. But see ante, at 239, n. 3.

In this case, I agree with the plurality that petitioner should be called upon to show that the outcome would have been the same if respondent’s professional merit had been its only concern. On remand, the District Court should determine whether Price Waterhouse has shown by a preponderance of the evidence that if gender had not been part of the process, its employment decision concerning Ann Hopkins would nonetheless have been the same.

Justice Kennedy,

with whom The Chief Justice and Justice Scalia join,

dissenting.

Today the Court manipulates existing and complex rules for employment discrimination cases in a way certain to result in confusion. Continued adherence to the evidentiary scheme established in McDonnell Douglas Corp. v. Green, 411 U. S. 792 (1973), and Texas Dept. of Community Affairs v. Burdine, 450 U. S. 248 (1981), is a wiser course than creation of more disarray in an area of the law already difficult for the bench and bar, and so I must dissent.

*280Before turning to my reasons for disagreement with the Court’s disposition of the case, it is important to review the actual holding of today’s decision. I read the opinions as establishing that in a limited number of cases Title VII plaintiffs, by presenting direct and substantial evidence of discriminatory animus, may shift the burden of persuasion to the defendant to show that an adverse employment decision would have been supported by legitimate reasons. The shift in the burden of persuasion occurs only where a plaintiff proves by direct evidence that an unlawful motive was a substantial factor actually relied upon in making the decision. Ante, at 276-277 (opinion of O’Connor, J.); ante, at 259-260 (opinion of White, J.). As the opinions make plain, the evidentiary scheme created today is not for every case in which a plaintiff produces evidence of stray remarks in the workplace. Ante, at 251 (opinion of Brennan, J.); ante, at 277 (opinion of O’Connor, J.).

Where the plaintiff makes the requisite showing, the burden that shifts to the employer is to show that legitimate employment considerations would have justified the decision without reference to any impermissible motive. Ante, at 260-261 (opinion of White, J.); ante, at 278 (opinion of O’Connor, J.). The employer’s proof on the point is to be presented and reviewed just as with any other evidentiary question: the Court does not accept the plurality’s suggestion that an employer’s evidence need be “objective” or otherwise out of the ordinary. Ante, at 261 (opinion of White, J.).

In sum, the Court alters the evidentiary framework of McDonnell Douglas and Burdine for a closely defined set of cases. Although Justice O’Connor advances some thoughtful arguments for this change, I remain convinced that it is unnecessary and unwise. More troubling is the plurality’s rationale for today’s decision, which includes a number of unfortunate pronouncements on both causation and methods of proof in employment discrimination cases. To demonstrate the defects in the plurality’s reasoning, it is nec*281essary to discuss, first, the standard of causation in Title VII cases, and, second, the burden of proof.

I

The plurality describes this as a case about the standard of causation under Title VII, ante, at 237, but I respectfully suggest that the description is misleading. Much of the plurality’s rhetoric is spent denouncing a “but-for” standard of causation. The theory of Title VII liability the plurality adopts, however, essentially incorporates the but-for standard. The importance of today’s decision is not the standard of causation it employs, but its shift to the defendant of the burden of proof. The plurality’s causation analysis is misdirected, for it is clear that, whoever bears the burden of proof on the issue, Title VII liability requires a finding of but-for causation. See also ante, at 261, and n. (opinion of White, J.); ante, at 262-263 (opinion of O’Connor, J.).

The words of Title VII are not obscure. The part of the statute relevant to this case provides:

“It shall be an unlawful employment practice for an employer—

“(1) to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual’s race, color, religion, sex, or national origin.” 42 U. S. C. § 2000e-2(a)(1) (emphasis added).

By any normal understanding, the phrase “because of” conveys the idea that the motive in question made a difference to the outcome. We use the words this way in everyday speech. And assuming, as the plurality does, that we ought to consider the interpretive memorandum prepared by the statute’s drafters, we find that this is what the words meant to them as well. “To discriminate is to make a distinction, to make a difference in treatment or favor.” 110 Cong. Rec. 7213 (1964). Congress could not have chosen a clearer way *282to indicate that proof of liability under Title VII requires a showing that race, color, religion, sex, or national origin caused the decision at issue.

Our decisions confirm that Title VII is not concerned with the mere presence of impermissible motives; it is directed to employment decisions that result from those motives. The verbal formulae we have used in our precedents are synonymous with but-for causation. Thus we have said that providing different insurance coverage to male and female employees violates the statute by treating the employee “ ‘in a manner which but-for that person’s sex would be different.’” Newport News Shipbuilding & Dry Dock Co. v. EEOC, 462 U. S. 669, 683 (1983), quoting Los Angeles Dept. of Water and Power v. Manhart, 435 U. S. 702, 711 (1978). We have described the relevant question as whether the employment decision was “based on” a discriminatory criterion, Teamsters v. United States, 431 U. S. 324, 358 (1977), or whether the particular employment decision at issue was “made on the basis of” an impermissible factor, Cooper v. Federal Reserve Bank of Richmond, 467 U. S. 867, 875 (1984).

What we term “but-for” cause is the least rigorous standard that is consistent with the approach to causation our precedents describe. If a motive is not a but-for cause of an event, then by definition it did not make a difference to the outcome. The event would have occurred just the same without it. Common-law approaches to causation often require proof of but-for cause as a starting point toward proof of legal cause. The law may require more than but-for cause, for instance proximate cause, before imposing liability. Any standard less than but-for, however, simply represents a decision to impose liability without causation. As Dean Prosser puts it, “[a]n act or omission is not regarded as a cause of an event if the particular event would have occurred without it.” W. Keeton, D. Dobbs, R. Keeton, & D. Owen, Prosser and Keeton on Law of Torts 265 (5th ed. 1984).

*283One of the principal reasons the plurality decision may sow confusion is that it claims Title VII liability is unrelated to but-for causation, yet it adopts a but-for standard once it has placed the burden of proof as to causation upon the employer. This approach conflates the question whether causation must be shown with the question of how it is to be shown. Because the plurality’s theory of Title VII causation is ultimately consistent with a but-for standard, it might be said that my disagreement with the plurality’s comments on but-for cause is simply academic. See ante, at 259 (opinion of White, J.). But since those comments seem to influence the decision, I turn now to that part of the plurality’s analysis.

The plurality begins by noting the quite unremarkable fact that Title VII is written in the present tense. Ante, at 240-241. It is unlawful “to fail” or “to refuse” to provide employment benefits on the basis of sex, not “to have failed” or “to have refused” to have done so. The plurality claims that the present tense excludes a but-for inquiry as the relevant standard because but-for causation is necessarily concerned with a hypothetical inquiry into how a past event would have occurred absent the contested motivation. This observation, however, tells us nothing of particular relevance to Title VII or the cause of action it creates. I am unaware of any federal prohibitory statute that is written in the past tense. Every liability determination, including the novel one constructed by the plurality, necessarily is concerned with the examination of a past event.1 The plurality’s analysis of verb tense serves only to divert attention from the causation requirement that is made part of the statute by the “because *284of” phrase. That phrase, I respectfully submit, embodies a rather simple concept that the plurality labors to ignore.2

We are told next that but-for cause is not required, since the words “because of” do not mean “solely because of.” Ante, at 241. No one contends, however, that sex must be the sole cause of a decision before there is a Title VII violation. This is a separate question from whether consideration of sex must be a cause of the decision. Under the accepted approach to causation that I have discussed, sex is a cause for the employment decision whenever, either by itself or in combination with other factors, it made a difference to the decision. Discrimination need not be the sole cause in order for liability to arise, but merely a necessary element of the set of factors that caused the decision, i. e., a but-for cause. See McDonald v. Santa Fe Trail Transportation Co., 427 U. S. 273, 282, n. 10 (1976). The plurality seems to say that since we know the words “because of” do not mean “solely because of,” they must not mean “because of” at all. This does not follow, as a matter of either semantics or logic.

The plurality’s reliance on the “bona fide occupational qualification” (BFOQ) provisions of Title VII, 42 U. S. C. §2000e-2(e), is particularly inapt. The BFOQ provisions allow an employer, in certain cases, to make an employment decision of which it is conceded that sex is the cause. That sex may be the legitimate cause of an employment decision where gender is a BFOQ is consistent with the opposite com*285mand that a decision caused by sex in any other case justifies the imposition of Title VII liability. This principle does not support, however, the novel assertion that a violation has occurred where sex made no difference to the outcome.

The most confusing aspect of the plurality’s analysis of causation and liability is its internal inconsistency. The plurality begins by saying: “When ... an employer considers both gender and legitimate factors at the time of making a decision, that decision was ‘because of’ sex and the other, legitimate considerations — even if we may say later, in the context of litigation, that the decision would have been the same if gender had not been taken into account.” Ante, at 241. Yet it goes on to state that “an employer shall not be liable if it can prove that, even if it had not taken gender into account, it would have come to the same decision.” Ante, at 242.

Given the language of the statute, these statements cannot both be true. Title VII unambiguously states that an employer who makes decisions “because of” sex has violated the statute. The plurality’s first statement therefore appears to indicate that an employer who considers illegitimate reasons when making a decision is a violator. But the opinion then tells us that the employer who shows that the same decision would have been made absent consideration of sex is not a violator. If the second statement is to be reconciled with the language of Title VII, it must be that a decision that would have been the same absent consideration of sex was not made “because of” sex. In other words, there is no violation of the statute absent but-for causation. The plurality’s description of the “same decision” test it adopts supports this view. The opinion states that “[a] court that finds for a plaintiff under this standard has effectively concluded that an illegitimate motive was a ‘but-for’ cause of the employment decision,” ante, at 249, and that this “is not an imposition of liability ‘where sex made no difference to the outcome,”’ ante, at 246, n. 11.

*286The plurality attempts to reconcile its internal inconsistency on the causation issue by describing the employer’s showing as an “affirmative defense.” This is nothing more than a label, and one not found in the language or legislative history of Title VII. Section 703(a)(1) is the statutory basis of the cause of action, and the Court is obligated to explain how its disparate-treatment decisions are consistent with the terms of § 703(a)(1), not with general themes of legislative history or with other parts of the statute that are plainly in-apposite. While the test ultimately adopted by the plurality may not be inconsistent with the terms of § 703(a)(1), see infra, at 292, the same cannot be said of the plurality’s reasoning with respect to causation. As Justice O’Connor describes it, the plurality “reads the causation requirement out of the statute, and then replaces it with an ‘affirmative defense.’” Ante, at 276. Labels aside, the import of today’s decision is not that Title VII liability can arise without but-for causation, but that in certain cases it is not the plaintiff who must prove the presence of causation, but the defendant who must prove its absence.

II

We established the order of proof for individual Title VII disparate-treatment cases in McDonnell Douglas Corp. v. Green, 411 J. S. 792 (1973), and reaffirmed this allocation in Texas Dept. of Community Affairs v. Burdine, 450 U. S. 248 (1981). Under Burdine, once the plaintiff presents a prima facie case, an inference of discrimination arises. The employer must rebut the inference by articulating a legitimate nondiscriminatory reason for its action. The final burden of persuasion, however, belongs to the plaintiff. Burdine makes clear that the “ultimate burden of persuading the trier of fact that the defendant intentionally discriminated against the plaintiff remains at all times with the plaintiff.” Id., at 253. See also Board of Trustees of Keene State College v. *287Sweeney, 439 U. S. 24, 29 (1978) (Stevens, J., dissenting).3 I would adhere to this established evidentiary framework, which provides the appropriate standard for this and other individual disparate-treatment cases. Today’s creation of a new set of rules for “mixed-motives” cases is not mandated by the statute itself. The Court’s attempt at refinement provides limited practical benefits at the cost of confusion and complexity, with the attendant risk that the trier of fact will misapprehend the controlling legal principles and reach an incorrect decision.

In view of the plurality’s treatment of Burdine and our other disparate-treatment cases, it is important first to state why those cases are dispositive here. The plurality tries to reconcile its approach with Burdine by announcing that it applies only to a “pretext” case, which it defines as a case in which the plaintiff attempts to prove that the employer’s proffered explanation is itself false. Ante, at 245-247, and n. 11. This ignores the language of Burdine, which states that a plaintiff may succeed in meeting her ultimate burden of persuasion “either directly by persuading the court that a discriminatory reason more likely motivated the employer or indirectly by showing that the employer’s proffered explanation is unworthy of credence.” 450 U. S., at 256 (emphasis added). Under the first of these two alternative methods, a plaintiff meets her burden if she can “persuade the court that the employment decision more likely than not was motivated by a discriminatory reason.” United States Postal Service Bd. of Governors v. Aikens, 460 U. S. 711, 717-718 (1983) *288(Blackmun, J., concurring). The plurality makes no attempt to address this aspect of our cases.

Our opinions make plain that Burdine applies to all individual disparate-treatment cases, whether the plaintiff offers direct proof that discrimination motivated the employer’s actions or chooses the indirect method of showing that the employer’s proffered justification is false, that is to say, a pretext. See Aikens, supra, at 714, n. 3 (“As in any lawsuit, the plaintiff may prove his case by direct or circumstantial evidence”). The plurality is mistaken in suggesting that the plaintiff in a so-called “mixed-motives” case will be disadvantaged by having to “squeeze her proof into Burdine’& framework.” Ante, at 247. As we acknowledged in McDonnell Douglas, “[t]he facts necessarily will vary in Title VII cases,” and the specification of the prima facie case set forth there “is not necessarily applicable in every respect to differing factual situations.” 411 U. S., at 802, n. 13. The framework was “never intended to be rigid, mechanized, or ritualistic.” Aikens, supra, at 715. Burdine compels the employer to come forward with its explanation of the decision and permits the plaintiff to offer evidence under either of the logical methods for proof of discrimination. This is hardly a framework that confines the plaintiff; still less is it a justification for saying that the ultimate burden of proof must be on the employer in a mixed-motives case. Burdine provides an orderly and adequate way to place both inferential and direct proof before the factfinder for a determination whether intentional discrimination has caused the employment decision. Regardless of the character of the evidence presented, we have consistently held that the ultimate burden “remains at all times with the plaintiff.” Burdine, supra, at 253.

Aikens illustrates the point. There, the evidence showed that the plaintiff, a black man, was far more qualified than any of the white applicants promoted ahead of him. More important, the testimony showed that “the person responsible for the promotion decisions at issue had made numerous *289derogatory comments about blacks in general and Aikens in particular.” 460 U. S., at 713-714, n. 2. Yet the Court in Aikens reiterated that the case was to be tried under the proof scheme of Burdine. Justice Brennan and Justice Blackmun concurred to stress that the plaintiff could prevail under the Burdine scheme in either of two ways, one of which was directly to persuade the court that the employment decision was motivated by discrimination. 460 U. S., at 718. Aikens leaves no doubt that the so-called “pretext” framework of Burdine has been considered to provide a flexible means of addressing all individual disparate-treatment claims.

Downplaying the novelty of its opinion, the plurality claims to have followed a “well-worn path” from our prior cases. The path may be well worn, but it is in the wrong forest. The plurality again relies on Title VH’s BFOQ provisions, under which an employer bears the burden of justifying the use of a sex-based employment qualification. See Dothard v. Rawlinson, 433 U. S. 321, 332-337 (1977). In the BFOQ context this is a sensible, indeed necessary, allocation of the burden, for there by definition sex is the but-for cause of the employment decision and the only question remaining is how the employer can justify it. The same is true of the plurality’s citations to Pregnancy Discrimination Act cases, ante, at 248. In such cases there is no question that pregnancy was the cause of the disputed action. The Pregnancy Discrimination Act and BFOQ cases tell us nothing about the case where the employer claims not that a sex-based decision was justified, but that the decision was not sex-based at all.

Closer analogies to the plurality’s new approach are found in Mt. Healthy City Bd. of Ed. v. Doyle, 429 U. S. 274 (1977), and NRLB v. Transportation Management Corp., 462 U. S. 393 (1983), but these cases were decided in different contexts. Mt. Healthy was a First Amendment case involving the firing of a teacher, and Transportation Management involved review of the NLRB’s interpretation of the National Labor Re*290lations Act. The Transportation Management decision was based on the deference that the Court traditionally accords NLRB interpretations of the statutes it administers. See 462 U. S., at 402-403. Neither case therefore tells us why the established Burdine framework should not continue to govern the order of proof under Title VII.

In contrast to the plurality, Justice O’Connor acknowledges that the approach adopted today is a “departure from the McDonnell Douglas standard.” Ante, at 262. Although her reasons for supporting this departure are not without force, they are not dispositive. As Justice O’Connor states, the most that can be said with respect to the Title VII itself is that “nothing in the language, history, or purpose of Title VII prohibits adoption” of the new approach. Ante, at 269 (emphasis added). Justice O’Connor also relies on analogies from the common law of torts, other types of Title VII litigation, and our equal protection cases. These analogies demonstrate that shifts in the burden of proof are not unprecedented in the law of torts or employment discrimination. Nonetheless, I believe continued adherence to the Burdine framework is more consistent with the statutory mandate. Congress’ manifest concern with preventing imposition of liability in cases where discriminatory animus did not actually cause an adverse action, see ante, at 262 (opinion of O’Connor, J.), suggests to me that an affirmative showing of causation should be required. And the most relevant portion of the legislative history supports just this view. See n. 3, supra. The limited benefits that are likely to be produced by today’s innovation come at the sacrifice of clarity and practical application.

The potential benefits of the new approach, in my view, are overstated. First, the Court makes clear that the Price Waterhouse scheme is applicable only in those cases where the plaintiff has produced direct and substantial proof that an impermissible motive was relied upon in making the decision at issue. The burden shift properly will be found to apply in *291only a limited number of employment discrimination cases. The application of the new scheme, furthermore, will make a difference only in a smaller subset of cases. The practical importance of the burden of proof is the “risk of nonpersuasion,” and the new system will make a difference only where the evidence is so evenly balanced that the factfinder cannot say that either side’s explanation of the case is “more likely” true. This category will not include cases in which the allocation of the' burden of proof will be dispositive because of a complete lack of evidence on the causation issue. Cf. Summers v. Tice, 33 Cal. 2d 80, 199 P. 2d 1 (1948) (allocation of burden dispositive because no evidence of which of two negligently fired shots hit plaintiff). Rather, Price Water-house will apply only to cases in which there is substantial evidence of reliance on an impermissible motive, as well as evidence from the employer that legitimate reasons supported its action.

Although the Price Waterhouse system is not for every case, almost every plaintiff is certain to ask for a Price Waterhouse instruction, perhaps on the basis of “stray remarks” or other evidence of discriminatory animus. Trial and appellate courts will therefore be saddled with the task of developing standards for determining when to apply the burden shift. One of their new tasks will be the generation of a jurisprudence of the meaning of “substantial factor.” Courts will also be required to make the often subtle and difficult distinction between “direct” and “indirect” or “circumstantial” evidence. Lower courts long have had difficulty applying McDonnell Douglas and Burdine. Addition of a second burden-shifting mechanism, the application of which itself depends on assessment of credibility and a determination whether evidence is sufficiently direct and substantial, is not likely to lend clarity to the process. The presence of an existing burden-shifting mechanism distinguishes the individual disparate-treatment case from the tort, class-action discrimination, and equal protection cases on which *292Justice O’Connor relies. The distinction makes Justice White’s assertions that one “need look only to” Mt. Healthy and Transportation Management to resolve this case, and that our Title VII cases in this area are “inapposite,” ante, at 258-260, at best hard to understand.

Confusion in the application of dual burden-shifting mechanisms will be most acute in cases brought under 42 U. S. C. § 1981 or the Age Discrimination in Employment Act (ADEA), where courts borrow the Title VII order of proof for the conduct of jury trials. See, e. g., Note, The Age Discrimination in Employment Act of 1967 and Trial by Jury: Proposals for Change, 73 Va. L. Rev. 601 (1987) (noting high reversal rate caused by use of Title VII burden shifting in a jury setting). Perhaps such cases in the future will require a bifurcated trial, with the jury retiring first to make the credibility findings necessary to determine whether the plaintiff has proved that an impermissible factor played a substantial part in the decision, and later hearing evidence on the “same decision” or “pretext” issues. Alternatively, perhaps the trial judge will have the unenviable task of formulating a single instruction for the jury on all of the various burdens potentially involved in the cáse.

I do not believe the minor refinement in Title VII procedures accomplished by today’s holding can justify the difficulties that will accompany it. Rather, I “remain confident that the McDonnell Douglas framework permits the plaintiff meriting relief to demonstrate intentional discrimination.” Bur-dine, 450 U. S., at 258. Although the employer does not bear the burden of persuasion under Burdine, it must offer clear and reasonably specific reasons for the contested decision, and has every incentive to persuade the trier of fact that the decision was lawful. Ibid. Further, the suggestion that the employer should bear the burden of persuasion due to superior access to evidence has little force in the Title VII context, where the liberal discovery rules available to all litigants are supplemented by EEOC investigatory files. Ibid. *293In sum, the Burdine framework provides a “sensible, orderly-way to evaluate the evidence in light of common experience as it bears on the critical question of discrimination,” Aikens, 460 U. S., at 715, and it should continue to govern the order of proof in Title VII disparate-treatment cases.4

Ill

The ultimate question in every individual disparate-treatment case is whether discrimination caused the particular decision at issue. Some of the plurality’s comments with respect to the District Court’s findings in this case, however, are potentially misleading. As the plurality notes, the District Court based its liability determination on expert evidence that some evaluations of respondent Hopkins were based on unconscious sex stereotypes,5 and on the fact that *294Price Waterhouse failed to disclaim reliance on these comments when it conducted the partnership review. The District Court also based liability on Price Waterhouse’s failure to “make partners sensitive to the dangers [of stereotyping], to discourage comments tainted by sexism, or to investigate comments to determine whether they were influenced by stereotypes.” 618 F. Supp. 1109, 1119 (DC 1985).

Although the District Court’s version of Title VII liability is improper under any of today’s opinions, I think it important to stress that Title VII creates no independent cause of action for sex stereotyping. Evidence of use by decision-makers of sex stereotypes is, of course, quite relevant to the question of discriminatory intent. The ultimate question, however, is whether discrimination caused the plaintiff’s harm. Our cases do not support the suggestion that failure to “disclaim reliance” on stereotypical comments itself violates Title VII. Neither do they support creation of a “duty to sensitize.” As the dissenting judge in the Court of Appeals observed, acceptance of such theories would turn Title VII “from a prohibition of discriminatory conduct into an engine for rooting out sexist thoughts.” 263 U. S. App. D. C. 321, 340, 825 F. 2d 458, 477 (1987) (Williams, J., dissenting).

Employment discrimination claims require factfinders to make difficult and sensitive decisions. Sometimes this may mean that no finding of discrimination is justified even though a qualified employee is passed over by a less than admirable employer. In other cases, Title VII’s protections properly extend to plaintiffs who are by no means model employees. As Justice Brennan notes, ante, at 258, courts do not sit to determine whether litigants are nice. In this *295case, Hopkins plainly presented a strong case both other own professional qualifications and of the presence of discrimination in Price Waterhouse’s partnership process. Had the District Court found on this record that sex discrimination caused the adverse decision, I doubt it would have been reversible error. Cf. Aikens, supra, at 714, n. 2. That decision was for the finder of fact, however, and the District Court made plain that sex discrimination was not a but-for cause of the decision to place Hopkins’ partnership candidacy on hold. Attempts to evade tough decisions by erecting novel theories of liability or multitiered systems of shifting burdens are misguided.

IV

The language of Title VII and our well-considered precedents require this plaintiff to establish that the decision to place her candidacy on hold was made “because of” sex. Here the District Court found that the “comments of the individual partners and the expert evidence of Dr. Fiske do not prove an intentional discriminatory motive or purpose,” 618 F. Supp., at 1118, and that “[b]ecause plaintiff has considerable problems dealing with staff and peers, the Court cannot say that she would have been elected to partnership if the Policy Board’s decision had not been tainted by sexually based evaluations,” id., at 1120. Hopkins thus failed to meet the requisite standard of proof after a full trial. I would remand the case for entry of judgment in favor of Price Waterhouse.

Justice Scalia

delivered the opinion of the Court.

This case presents the question whether workplace harassment can violate Title VII’s prohibition against “diserimina-t[ion]... because of... sex,” 42 U. S. C. §2000e-2(a)(I), when the harasser and the harassed employee are of the same sex.

I

The District Court having granted summary judgment for respondents, we must assume the facts to be as alleged by petitioner Joseph Oncale. The precise details are irrelevant *77to the legal point we must decide, and in the interest of both brevity and dignity we shall describe them only generally. In late October 1991, Oncale was working for respondent Sundowner Offshore Services, Inc., on a Chevron U. S. A., Inc., oil platform in the Gulf of Mexico. He was employed as a roustabout on an eight-man crew which included respondents John Lyons, Danny Pippen, and Brandon Johnson. Lyons, the crane operator, and Pippen, the driller, had supervisory authority, App. 41, 77, 43. On several occasions, Oncale was forcibly subjected to sex-related, humiliating actions against him by Lyons, Pippen, and Johnson in the presence of the rest of the crew. Pippen and Lyons also physically assaulted Oncale in a sexual manner, and Lyons threatened him with rape.

Oneale’s complaints to supervisory personnel produced no remedial action; in fact, the company’s Safety Compliance Clerk, Valent Hohen, told Oncale that Lyons and Pippen “picked [on] him all the time too,” and called him a name suggesting homosexuality. Id., at 77. Oncale eventually quit — asking that his pink slip reflect that he “voluntarily left due to sexual harassment and verbal abuse.” Id., at 79. When asked at his deposition why he left Sundowner, Oncale stated: “I felt that if I didn’t leave my job, that I would be raped or forced to have sex.” Id., at 71.

Oncale filed a complaint against Sundowner in the United States District Court for the Eastern District of Louisiana, alleging that he was discriminated against in his employment because of his sex. Relying on the Fifth Circuit’s decision in Garcia v. Elf Atochem North America, 28 F. 3d 446, 451-452 (1994), the District Court held that “Mr. Oncale, a male, has no cause of action under Title VII for harassment by male co-workers.” App. 106. On appeal, a panel of the Fifth Circuit concluded that Garcia was binding Circuit precedent, and affirmed. 83 F. 3d 118 (1996). We granted certiorari. 520 U. S. 1263 (1997).

*78H-4 H-i

Title VII of the Civil Rights Act of 1964 provides, in relevant part, that “[i]t shall be an unlawful employment practice for an employer ... to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual’s race, color, religion, sex, or national origin.” 78 Stat. 255, as amended, 42 U. S. C. § 2000e-2(a)(1). We have held that this not only covers “terms” and “conditions” in the narrow contractual sense, but “evinces a congressional intent to strike at the entire spectrum of disparate treatment of men and women in employment.” Meritor Savings Bank, FSB v. Vinson, 477 U. S. 57, 64 (1986) (citations and internal quotation marks omitted). “When the workplace is permeated with discriminatory intimidation, ridicule, and insult that is sufficiently severe or pervasive to alter the conditions of the victim’s employment and create an abusive working environment, Title VII is violated.” Harris v. Forklift Systems, Inc., 510 U. S. 17, 21 (1993) (citations and internal quotation marks omitted).

Title VII’s prohibition of discrimination “because of . . . sex” protects men as well as women, Newport News Shipbuilding & Dry Dock Co. v. EEOC, 462 U. S. 669, 682 (1983), and in the related context of racial discrimination in the workplace we have rejected any conclusive presumption that an employer will not discriminate against members of his own race. “Because of the many facets of human motivation, it would be unwise to presume as a matter of law that human beings of one definable group will not discriminate against other members of their group.” Castaneda v. Partida, 430 U. S. 482, 499 (1977). See also id., at 515-516, n. 6 (Powell, J., joined by Burger, C. J., and Rehnquist, J., dissenting). In Johnson v. Transportation Agency, Santa Clara Cty., 480 U. S. 616 (1987), a male employee claimed that his employer discriminated against him because of his sex when it preferred a female employee for promotion. Al*79though we ultimately rejected the claim on other grounds, we did not consider it significant that the supervisor who made that decision was also a man. See id., at 624-625. If our precedents leave any doubt on the question, we hold today that nothing in Title VII necessarily bars a claim of discrimination “because of... sex” merely because the plaintiff and the defendant (or the person charged with acting on behalf of the defendant) are of the same sex.

Courts have had little trouble with that principle in cases like Johnson, where an employee claims to have been passed over for a job or promotion. But when the issue arises in the context of a “hostile environment” sexual harassment claim, the state and federal courts have taken a bewildering variety of stances. Some, like the Fifth Circuit in this ease, have held that same-sex sexual harassment claims are never cognizable under Title VII. See also, e. g., Goluszek v. H. P. Smith, 697 F. Supp. 1452 (ND Ill. 1988). Other decisions say that such claims are actionable only if the plaintiff can prove that the harasser is homosexual (and thus presumably motivated by sexual desire). Compare McWilliams v. Fairfax County Board of Supervisors, 72 F. 3d 1191 (CA4 1996), with Wrightson v. Pizza Hut of America, 99 F. 3d 138 (CA4 1996). Still others suggest that workplace harassment that is sexual in content is always actionable, regardless of the harasser’s sex, sexual orientation, or motivations. See Doe v. Belleville, 119 F. 3d 563 (CA7 1997).

We see no justification in the statutory language or our precedents for a categorical rule excluding same-sex harassment claims from the coverage of Title VII. As some courts have observed, male-on-male sexual harassment in the workplace was assuredly not the principal evil Congress was concerned with when it enacted Title VII. But statutory prohibitions often go beyond the principal evil to cover reasonably comparable evils, and it is ultimately the provisions of our laws rather than the principal concerns of our legislators by which we are governed. Title VII prohibits “discrimina-*80t[ion].. . because of. .. sex” in the “terms” or “conditions” of employment. Our holding that this includes sexual harassment must extend to sexual harassment of any kind that meets the statutory requirements.

Respondents and their amici contend that recognizing liability for same-sex harassment will transform Title VII into a general civility code for the American workplace. But that risk is no greater for same-sex than for opposite-sex harassment, and is adequately met by careful attention to the requirements of the statute. Title VII does not prohibit all verbal or physical harassment in the workplace; it is directed only at “discrimination] . . . because of . . . sex.” We have never held that workplace harassment, even harassment between men and women, is automatically discrimination because of sex merely because the words used have sexual content or connotations. “The critical issue, Title VII’s text indicates, is whether members of one sex are exposed to disadvantageous terms or conditions of employment to which members of the other sex are not exposed.” Harris, supra, at 25 (Ginsburg, J., concurring).

Courts and juries have found the inference of discrimination easy to draw in most male-female sexual harassment situations, because the challenged conduct typically involves explicit or implicit proposals of sexual activity; it is reasonable to assume those proposals would not have been made to someone of the same sex. The same chain of inference would be available to a plaintiff alleging same-sex harassment, if there were credible evidence that the harasser was homosexual. But harassing conduct need not be motivated by sexual desire to support an inference of discrimination on the basis of sex. A trier of fact might reasonably find such discrimination, for example, if a female victim is harassed in such sex-specific and derogatory terms by another woman as to make it clear that the harasser is motivated by general hostility to the presence of women in the workplace. A same-sex harassment plaintiff may also, of course, offer di*81reet comparative evidence about how the alleged harasser treated members of both sexes in a mixed-sex workplace. Whatever evidentiary route the plaintiff chooses to follow, he or she must always prove that the conduct at issue was not merely tinged with offensive sexual connotations, but actually constituted “discrimination] ... because of... sex.”

And there is another requirement that prevents Title VII from expanding into a general civility code: As we emphasized in Meritor and Harris, the statute does not reach genuine but innocuous differences in the ways men and women routinely interact with members of the same sex and of the opposite sex. The prohibition of harassment on the basis of sex requires neither asexuality nor androgyny in the workplace; it forbids only behavior so objectively offensive as to alter the “conditions” of the victim’s employment. “Conduct that is not severe or pervasive enough to create an objectively hostile or abusive work environment — an environment that a reasonable person would find hostile or abusive — is beyond Title VII’s purview.” Harris, 510 U. S., at 21, citing Meritor, 477 U. S., at 67. We have always regarded that requirement as crucial, and as sufficient to ensure that courts and juries do not mistake ordinary socializing in the workplace — such as male-on-male horseplay or intersexual flirtation — for discriminatory “conditions of employment.”

We have emphasized, moreover, that the objective severity of harassment should be judged from the perspective of a reasonable person in the plaintiff’s position, considering “all the circumstances.” Harris, supra, at 23. In same-sex (as in all) harassment cases, that inquiry requires careful consideration of the social context in which particular behavior occurs and is experienced by its target. A professional football player’s working environment is not severely or pervasively abusive, for example, if the coach smacks him on the buttocks as he heads onto the field — even if the same behavior would reasonably be experienced as abusive by the coach’s secretary (male or female) back at the office. The *82real social impact of workplace behavior often depends on a constellation of surrounding circumstances, expectations, and relationships which are not fully captured by a simple recitation of the words used or the physical acts performed. Common sense, and an appropriate sensitivity to social context, will enable courts and juries to distinguish between simple teasing or roughhousing among members of the same sex, and conduct which a reasonable person in the plaintiff’s position would find severely hostile or abusive.

III

Because we conclude that sex discrimination consisting of same-sex sexual harassment is actionable under Title VII, the judgment of the Court of Appeals for the Fifth Circuit is reversed, and the ease is remanded for further proceedings consistent with this opinion.

It is so ordered.

Justice Thomas,

concurring.

I concur because the Court stresses that in every sexual harassment ease, the plaintiff must plead and ultimately prove Title VII’s statutory requirement that there be discrimination “because of... sex.”

*1077TASHIMA, Circuit Judge:

Plaintiff Darlene Jespersen, a bartender at Harrah’s Casino in Reno, Nevada, brought this Title VII action alleging that her employer’s policy requiring that certain female employees wear makeup discriminates against her on the basis of sex. The district court granted summary judgment for Harrah’s, holding that its policy did not constitute sex discrimination because it imposed equal burdens on both sexes. We have jurisdiction pursuant to 28 U.S.C. § 1291, and we affirm.

I.

The following facts are undisputed. Darlene Jespersen was a bartender at the sports bar in Harrah’s Casino in Reno, Nevada, for nearly 20 years. She was an outstanding employee. Over the years, Jespersen’s supervisors commented that she was “highly effective,” that her attitude was “very positive,” and that she made a “positive impression” on Harrah’s guests. Harrah’s customers repeatedly praised Jespersen on employee feedback forms, writing that Jespersen’s excellent service and good attitude enhanced their experience at the sports bar and encouraged them to come back.

Throughout the 1980s and '90s Harrah’s encouraged its female beverage servers to wear makeup, but wearing makeup was not a formal requirement. Although Jes-persen never cared for makeup, she tried wearing it for a short period of time in the 1980s. But she found that wearing makeup made her feel sick, degraded, exposed, and violated. Jespersen felt that wearing makeup “forced her to be feminine” and to become “dolled up” like a sexual object, and that wearing makeup actually interfered with her ability to be an effective bartender (which sometimes required her to deal with unruly, intoxicated guests) because it “took away [her] credibility as an individual and as a person.” After a few weeks, Jespersen stopped wearing makeup because it was so harmful to her dignity and her effectiveness behind the bar that she could no longer do her job. Harrah’s did not object to Jespersen’s choice not to wear makeup and Jespersen continued to work at the sports bar and receive positive performance reviews for over a decade.

In February 2000, Harrah’s implemented its “Beverage Department Image Transformation” program at 20 Harrah’s locations, including its casino in Reno. The goal of the program was to create a “brand standard of excellence” throughout Har-rah’s operations, with an emphasis on guest service positions. The program imposed specific “appearance standards” on each of its employees in guest services, including heightened requirements for beverage servers. All beverage servers were required to be “well groomed, appealing to the eye, be firm and body toned, and be comfortable with maintaining this look while wearing the specified uniform.” In addition to these general appearance standards applicable to both sexes, there were gender-specific standards for male and female beverage servers. Female beverage servers were required to wear stockings and colored nail polish, and they were required to wear their hair “teased, curled, or styled.” Male beverage servers were prohibited from wearing makeup or colored nail polish, and they were required to maintain short haircuts and neatly trimmed fingernails.1

*1078Harrah’s called its new appearance standards the “Personal Best” program. In order to enforce the “Personal Best” standards, Harrah’s required each beverage service employee to attend “Personal Best Image Training” prior to his or her final uniform fitting. At the training, “Personal Best Image Facilitators” instructed Har-rah’s employees on how to adhere to the standards of the program and tested their proficiency. At the conclusion of the training, two photographs (one portrait and one full body) were taken of the employee looking his or her “Personal Best.” Each employee’s “Personal Best” photographs were placed in his or her file and distributed to his or her supervisor. The supervisors used the “Personal Best” photographs as an “appearance measurement” tool, holding each employee accountable to look his or her “Personal Best” on a daily basis. Jespersen acknowledged receipt of the policy and committed to adhere to the appearance standards for her position as a beverage bartender in March 2000.

Shortly thereafter, however, the “Personal Best” standards were amended such that in addition to the existing appearance standards, all female beverage servers (including beverage bartenders) were required to wear makeup.2 As before, male beverage servers were prohibited from wearing makeup. Because of her objection to wearing makeup, Jespersen refused to comply with the new policy. In July 2000, Harrah’s told Jespersen that the makeup requirement was mandatory for female beverage service employees and gave her 30 days to apply for a position that did not require makeup to be worn. At the expiration of the 30-day period, Jespersen had not applied for another job, and she was terminated.

After exhausting her administrative remedies with the Equal Employment Opportunity Commission, Jespersen brought this action alleging that Harrah’s makeup requirement for female beverage servers constituted disparate treatment sex discrimination in violation of 42 U.S.C. § 2000e-2(a) (“Title VII”). The district *1079court granted Harrah’s motion for summary judgment, holding that the “Personal Best” policy did not run afoul of Title VII because (1) it did not discriminate against Jespersen on the basis of “immutable characteristics” associated with her sex, and (2) it imposed equal burdens on both sexes. Jespersen timely appealed from the judgment.

II.

We review the grant of summary judgment de novo. United States v. City of Tacoma, 332 F.3d 574, 578 (9th Cir.2003). Summary judgment is proper where no genuine issues of material fact remain in dispute, such that the moving party is entitled to judgment as a matter of law. Fed.R.Civ.P. 56(c). The moving party has the initial burden of informing the court of the basis for its motion and identifying the absence of any genuine issue of material fact. Celotex Corp. v. Catrett, 477 U.S. 317, 323, 106 S.Ct. 2548, 91 L.Ed.2d 265 (1986). Where the moving party is not the party bearing the burden of proof at trial, it can meet its initial burden simply by “identifying those portions of ‘the pleadings, depositions, answers to interrogatories, and admissions on file, together with the affidavits, if any,’ which it believes demonstrate the absence of a genuine issue of material fact.” Id. (quoting Fed. R.Civ.P. 56(c)); see also Nissan Fire & Marine Ins. Co. v. Fritz Cos., 210 F.3d 1099, 1105 (9th Cir.2000). Once the moving party has met its initial burden, the non-moving party must produce some evidence showing that there remains a genuine issue of material fact for trial. Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248-49, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986). An issue is “genuine” only where there is sufficient evidence for a reasonable fact finder to find for the non-moving party. Far Out Prods., Inc. v. Oskar, 247 F.3d 986, 992 (9th Cir.2001) (citing Anderson, 477 U.S. at 248-49, 106 S.Ct. 2505). We view the evidence in the light most favorable to the non-moving party. Id,

III.

Title VII prohibits employers from discriminating against “any individual with respect to ... compensation, terms, conditions, or privileges of employment, because of such individual’s ... sex.” 42 U.S.C. § '2000e-2(a)(l). In order to prevail on a Title VII disparate treatment sex discrimination claim, an employee need only establish that, but for his or her sex, he or she would have been treated differently. UAW v. Johnson Controls, Inc., 499 U.S. 187, 200, 111 S.Ct. 1196, 113 L.Ed.2d 158 (1991)' (citing Los Angeles Dep’t of Water & Power v. Manhart, 435 U.S. 702, 711, 98 S.Ct. 1370, 55 L.Ed.2d 657 (1978)). Although the employee must prove that the employer acted intentionally, the intent need not have been malevolent. Id. at 199, 111 S.Ct. 1196 (“Whether an employment practice involves disparate treatment through explicit facial discrimination does not depend on why the employer discriminates but rather on the explicit terms of the discrimination.”).3

Pursuant to the “Personal Best” program, women are required to wear makeup, while men are prohibited from doing so. Women are required to wear their hair “teased, curled, or styled” each day, whereas men are only required to maintain short haircuts. We must decide whether these standards are discriminatory; whether they are “based on a policy *1080which on its face applies less favorably to one gender .... ” Gerdom v. Continental Airlines, Inc., 692 F.2d 602, 608 (9th Cir.1982). If so, then Harrah’s would have discriminated against Jespersen “because of ... sex.” 42 U.S.C. § 2000e-2(a)(l); see id.

We have previously held that grooming and appearance standards that apply differently to women and men do not constitute discrimination on the basis of sex. In Baker v. Cal. Land Title Co., 507 F.2d 895 (9th Cir.1974), employees challenged their employer’s rule banning men, but not women, from having long hair. Id. at 896. We concluded that grooming and dress standards were entirely outside the purview of Title VII because Congress intended that Title VII only prohibit discrimination based on “immutable characteristics” associated with a worker’s sex. Id. at 897 (“Since race, national origin and color represent immutable characteristics, logic dictates that sex is used in the same sense rather than to indicate personal modes of dress or cosmetic effects.”); see also Fountain v. Safeway Stores Inc., 555 F.2d 753, 755 (9th Cir.1977) (“It is clear that regulations promulgated by employers which require male employees to conform to different grooming and dress standards than female employees is not sex discrimination within the meaning of Title VII.”). Because grooming and dress standards regulated “mutable” characteristics such as hair length, we reasoned, employers that made compliance with such standards a condition of employment discriminated on the basis of their employees’ appearance, not their sex.

Our later cases recognized, however, that an employer’s imposition of more stringent appearance standards on one sex than the other constitutes sex discrimination even where the appearance standards regulate only “mutable” characteristics such as weight. Gerdom, 692 F.2d at 605-06. In Frank v. United Airlines, Inc., 216 F.3d 845 (9th Cir.2000) (en banc), a class of female flight attendants challenged their employer’s weight restrictions as a violation of Title VII because women were held to more strict weight limitations than were men. The employer insisted that all employees maintain a weight that corresponded to the “desirable” weight for their height as determined by an insurance company table, but women were required to maintain the weight corresponding to women of “medium” build, whereas men were permitted to maintain the weight corresponding to men of “large” build. Id. at 848. Citing Fountain, the employer argued that because the weight restrictions were mere “appearance” standards, they were not subject to Title VII. Id. at 854. We rejected the employer’s argument, holding that “[a] sex-differentiated appearance standard that imposes unequal burdens on men and women is disparate treatment that must be justified as a BFOQ.” Id. at 855; see also Carroll v. Talman Fed. Sav. & Loan Ass’n, 604 F.2d 1028, 1032 (7th Cir.1979) (holding that employer’s policy requiring female employees to wear uniforms but permitting male employees to wear “appropriate business attire” of their choosing was sex discrimination in violation of Title VII). Although employers are free to adopt different appearance standards for each sex, they may not adopt standards that impose a greater burden on one sex than the other. Frank, 216 F.3d at 855.

Although in Frank we characterized the weight standards at issue as “appearance standards,” id., we have, as yet, had no occasion to apply the “unequal burdens” test to gender-differentiated dress and grooming requirements. In Frank and Gerdom, we were called upon only to compare the relative burdens of different weight limitations imposed on male and female employees. In those cases our task *1081was simple because it was apparent from the face of the policies at issue that female flight attendants were subject to a more onerous standard than were males. See Frank, 216 F.3d at 854; Gerdom, 692 F.2d at 608.

In order to evaluate the relative burdens the “Personal Best” policy imposes, we must assess the actual impact that it has on both male and female employees. In doing so we must weigh the cost and time necessary for employees of each sex to comply with the policy; Harrah’s contends that the burden of the makeup requirement must be evaluated with reference to all of the requirements of the policy, including those that burden men only, such as the requirement that men maintain short haircuts and neatly trimmed nails. Jespersen contends that the only meaningful appearance standard against which the makeup requirement can be measured is the corresponding “no makeup” requirement for men. We agree with Harrah’s approach. Because employers are permitted to apply different appearance standards to each sex so long as those standards are equal, our task in applying the “unequal burdens” test to grooming and dress requirements must sometimes involve weighing the relative burdens that particular requirements impose on workers of one sex against the distinct requirements imposed on workers of the-other sex.4

Jespersen contends that the makeup requirement imposes “innumerable” tangible burdens on women that men do not share because cosmetics can cost hundreds of dollars per year and putting on makeup requires a significant investment in time. There is, however, no evidence in the record in support of this contention. Jesper-sen cites to academic literature discussing the cost and time burdens of cosmetics generally, but she presents no evidence as to the cost or time burdens that must be borne by female bartenders in order to comply with the makeup requirement. Even if we were to .take judicial notice of the fact that the application of makeup requires some expenditure of time and money, Jespersen would still have the burden of producing some evidence that the burdens associated with the makeup requirement are greater than the burdens the “Personal Best” policy imposes on male bartenders, and exceed whatever “burden” is associated with ordinary good-grooming standards. Because there is no evidence in the record from which we can assess the burdens that the “Personal Best” policy imposes on -male bartenders either, Jespersen’s claim fails for that reason alone.

Jespersen cites United States v. Seschil-lie, 310 F.3d 1208, 1212 (9th Cir.2002), for the proposition that “a jury can make determinations requiring simple common sense without specific supporting evidence.” But Seschillie involved the entirely different question of whether jurors in a criminal case could draw common-sense inferences from the evidence without the aid of expert testimony. Id. It cannot be construed as relieving Jespersen of her burden of production at the summary judgment stage in a civil case. As the non-moving party that bore the ultimate burden of proof at trial, Jespersen had the burden of producing admissible evidence that the “Personal Best” appearance stan*1082dard imposes a greater burden on female beverage servers than it does on male beverage servers. See Anderson, ill U.S. at 248, 106 S.Ct. 2505. She has not met that burden.

Jespersen also contends that even if Harrah’s makeup requirement survives the “unequal burdens” test, that test should be invalidated in light of the Supreme Court’s decision in Price Waterhouse v. Hopkins, 490 U.S. 228, 109 S.Ct. 1775, 104 L.Ed.2d 268 (1989). In Price Waterhouse, the Supreme Court held that an employer may not force its employees to conform to the sex stereotype associated with their gender as a condition of employment. Id. at 250-51, 109 S.Ct. 1775. When evaluating a female associate’s candidacy for partnership in an accounting firm, decision makers referred to her as “macho” and suggested that she “overcompensated for being a woman” by behaving aggressively in the workplace. Id. at 235, 109 S.Ct. 1775. The associate was advised that her partnership chances would be improved if she learned to behave more femininely, wear makeup, have her hair styled, and wear jewelry. Id. Noting that “we are beyond the day when an employer could evaluate employees by assuming or insisting that they matched the stereotype associated with their group,” the Court held that the employer’s discrimination against the associate because of her failure to conform to a traditional, feminine gender stereotype was sex discrimination in violation of Title VII. Id. at 251, 109 S.Ct. 1775.

Following Price Waterhouse, we have held that sexual harassment of an employee because of that employee’s failure to conform to commonly-accepted gender stereotypes is sex discrimination in violation of Title VII. In Nichols v. Azteca Restaurant Enter., Inc., 256 F.3d 864 (9th Cir.2001), a male waiter at a restaurant sued his employer under Title VII for sexual harassment. The waiter contended that he was harassed because he failed to conform his behavior to a traditionally male stereotype. Id. at 874. Noting that Price Waterhouse “sets a rule that bars discrimination on the basis of sex stereotypes,” we concluded that the harassment and abuse was actionable under Title VII because the waiter was systematically abused for failing to act “as a man should act” and for walking and carrying his tray “like a woman.” Id. at 874-75. Similarly, in Rene v. MGM Grand Hotel, Inc., 305 F.3d 1061 (9th Cir.2002) (en banc), we held that a man stated a claim for sexual harassment under Title VII where he alleged that he was the victim of assaults “of a sexual nature” by his co-workers because of stereotypical assumptions. Id. at 1068.

Although Price Waterhouse held that Title VII bans discrimination against an employee on the basis of that employee’s failure to dress and behave according to the stereotype corresponding with her gender, it did not address the specific question of whether an employer can impose sex-differentiated appearance and grooming standards on its male and female employees. Nor have our subsequent cases invalidated the “unequal burdens” test as a means of assessing whether sex-differentiated appearance standards discriminate on the basis of sex. Although the precise issue was not before us, we declined to apply Price Waterhouse to grooming and appearance standards cases when we rendered our decision in Nichols, 256 F.3d at 875 n. 7 (“Our decision does not imply that there is any violation of Title VII occasioned by reasonable regulations that require male and female employees to conform to different dress and grooming standards.”). And while a plurality of judges in Rene endorsed an independent claim for gender-stereotyping sexual harassment, such a claim is distinct from the claim Jespersen advances here. She has presented no evidence that she or *1083any other employee has been sexually harassed as a result of the “Personal Best” policy. In short, although we have applied the reasoning of Price Waterhouse to sexual harassment cases, we have not done so in the context of appearance and grooming standards cases, and we decline to do so here. We thus disagree with the dissent’s assertion that “Jespersen has articulated a classic case of Price Waterhouse discrimination .... ” Dissent at 1084.

Finally, we note that we are, in any event, bound to follow our en banc decision in Frank, in which we adopted the unequal burdens test. Price Waterhouse predates Frank by more than a decade and, presumably, the Frank en banc court was aware of it when it adopted the unequal burdens test. Thus, Price Waterhouse does not qualify as an “intervening decision” which could serve as a basis for overruling Frank. See EEOC v. Luce, Forward, Hamilton & Scripps, 345 F.3d 742, 744 n. 1 (9th Cir.2003) (en banc) (explaining that “[a] three-judge panel can overrule a prior decision of this court [only] when an intervening Supreme Court decision undermines ah existing precedent of the Ninth Circuit, and both cases are closely on point”) (internal quotation marks and citations omitted).

IV.

We hold that under the “unequal burdens” test, which is this Circuit’s test for evaluating whether an employer’s sex-differentiated appearance standards constitute sex discrimination in violation of Title VII, Jespersen failed to introduce evidence raising a triable issue of fact as to whether Harrah’s “Personal Best” policy imposes unequal burdens on male and female employees.

The judgment of the district court is AFFIRMED.

THOMAS, Circuit Judge,

dissenting.

I respectfully dissent.

Harrah’s required Darlene Jespersen to wear makeup to work. She refused because the cost — measured in time, money, and personal dignity — was too high. Har-rah’s fired her. The majority holds that Jespersen failed to raise a triable issue of fact as to whether Harrah’s policy imposes unequal burdens on men and women. In fact, Jespersen easily satisfied her burden. A reasonable factfinder could determine that Harrah’s acted because of Jespersen’s sex under not just one theory, but two. First, Harrah’s fired Jespersen because of her failure to conform to sex stereotypes, which is discrimination based on sex and is therefore impermissible under Title VII. Second, Jespersen created a triable issue of fact as to whether the policy imposed unequal burdens on men and women, because the policy imposes a requirement on women that is not only time-consuming and expensive, but burdensome for its requirement that women conform to outdated and impermissible sex stereotypes.

I

Darlene Jespersen was fired from her position as a bartender in a sports bar at Harrah’s Casino. There is no question as to why she was fired: because she would not or could not comply with Harrah’s stringent company policy requiring female beverage servers to wear foundation, blush, mascara, and lip color, and to ensure that lip color is on at all times. There is also no question that her performance was not only competent; it was spectacular. She was consistently given glowing recommendations by numerous customers and supervisors, despite the fact that she did not wear makeup.

The Harrah’s policy is far more stringent than simply asking female employees to wear some makeup. The policy essen*1084tially requires women to wear a uniform of makeup including at least mascara, blush, lipstick, and foundation. In fact, according to the “image consultant” who helped implement the policy, wearing makeup “completes” the “uniformed look” of women beverage servers.

All employees at Harrah’s were given “image consultations” as part of the “Personal Best” policy, which for women included a makeover that would result in them being “properly made-up.” The post-makeover photographs are used as an “appearance measurement tool,” and each employee is held “accountable” to the photograph “on a daily basis.” Thus, while men are held accountable to look as clean, have their hair as neat, and have their clothes as tidy and fitted as in their photo, women are held accountable to do all these things as well as be “properly made up,” as they are in the post-makeover photo.

II

Under Harrah’s “Personal Best” policy, Jespersen was required to wear makeup and thus conform to a sex stereotype; when she refused, Harrah’s fired her. When an employer takes an adverse employment action against a plaintiff based on the plaintiffs failure to conform to sex stereotypes, the employer has acted because of sex. Price Waterhouse v. Hopkins, 490 U.S. 228, 251, 109 S.Ct. 1775, 104 L.Ed.2d 268 (1989) (“As for the legal relevance of sex stereotyping, we are beyond the day when an employer could evaluate employees by assuming or insisting that they matched the stereotype associated with their group, for ... Congress intended to strike at the entire spectrum of disparate treatment of men and women resulting from sex stereotypes.”) (emphasis added); see also Smith v. City of Salem, 369 F.3d 912 (6th Cir.2004) (holding, based on Price Waterhouse, that the suspension of a pre-operative transsexual employee based on his gender non-conforming appearance and behavior is actionable under Title VII); Nichols v. Azteca Rest. Enters., Inc., 256 F.3d 864 (9th Cir.2001) (holding, based on Price Waterhouse, that harassment of a male employee for failure to act masculine enough is actionable under Title VII). Jespersen has articulated a classic case of Price Waterhouse discrimination and has tendered sufficient undisputed, material facts to avoid summary judgment.

The majority attempts to distinguish this case from Price Waterhouse and Nichols because this is not a sexual harassment case. But neither was Price Waterhouse, in which the adverse employment action taken against the plaintiff was that she was denied partnership. 490 U.S. at 233, 109 S.Ct. 1775. Even if it were, that would not matter. The question of whether an action is “because of sex” is separate from the question of whether the action constitutes an adverse employment action actionable under Title VII:

Title VII does not prohibit all verbal or physical harassment in the workplace; it is directed only at ‘discrimination] ... because of ... sex.’ We have never held that workplace harassment, even harassment between men and women, is automatically discrimination because of sex merely because the words used have sexual content or connotations.

Oncale v. Sundowner Offshore Servs., Inc., 523 U.S. 75, 80, 118 S.Ct. 998, 140 L.Ed.2d 201 (1998) (quoting 42 U.S.C. § 2000e-2(a)(1)) (alterations in original). There is no grounding whatsoever in Title VII for the notion that harassing an employee because he or she fails to conform to a sex stereotype is illegitimate, while firing them for the same reason is acceptable.

The majority also suggests that Price Waterhouse only applies in certain contexts and did not address sex-differentiated appearance and grooming standards. *1085In Price Waterhouse v. Hopkins, the plaintiff was denied partnership at a prestigious accounting firm where she had excelled because she didn’t act femininely enough, and was specifically faulted for not wearing makeup. 490 U.S. at 235, 109 S.Ct. 1775. Jespersen was fired from a job she also excelled at, for exactly the same reason. The distinction created by the majority opinion leaves men and women in service industries, who are more likely to be subject to policies like the Harrah’s “Personal Best” policy, without the protection that white-collar professionals receive.

Title VII does not make exceptions for particular industries, and we should not write them in. Pervasive discrimination often persists within an industry with exceptional tenacity, and the force of law is sometimes required to overcome it. See Gerdom, 692 F.2d at 606-07 (describing the history of litigation over gender discrimination in the airline industry). When a company acts to enforce sexual stereotypes through grooming standards, it is not immune from Price Waterhouse liability; to the contrary, such actions fall precisely within the heartland of Price Water-house.

Ill

Even if Price Waterhouse did not apply in the grooming and appearance context, Harrah’s was not entitled to summary judgment, for Jespersen created an issue of material fact as to whether the Harrah’s policy is a grooming standard that imposes unequal burdens on men and women, in violation of Title VII. Frank v. United Airlines, Inc., 216 F.3d 845, 855 (9th Cir.2000); Gerdom v. Continental Airlines, Inc., 692 F.2d 602, 606 (9th Cir.1982); see also Carroll v. Talman Federal Sav. & Loan Ass’n of Chicago, 604 F.2d 1028, 1032 (7th Cir.1979), cited with approval in Frank, 216 F.3d at 855.

The majority opinion’s holding that the burdens imposed by all of Harrah’s appearance policy requirements must be compared to each other does not follow from prior caselaw permitting employers to maintain sex-differentiated appearance standards that do not impose unequal burdens. Under the majority opinion’s methodology, a sex-differentiated appearance requirement that unfairly burdens women, such as a requirement that women meet more stringent weight limits than men, Frank, 216 F.3d at 855, could be permissible if the employer unfairly burdened men via another sex-differentiated appearance requirement, for instance, by requiring men to wear contacts but permitting women to wear glasses, id.

Rather than permit all sorts of sex discrimination as long as it “balances out” for both genders, I would instead compare individual sex-differentiated appearance requirements that correspond to each other, given that Title VII prohibits employers from taking adverse employment actions “because of ... sex.” 42 U.S.C. § 2000e-2(a)(l). Harrah’s hair length requirement and ponytail prohibition for men should be compared to the requirement that women wear their hair “teased, curled, or styled” every day and that their hair be “worn down” at all times. Similarly, Harrah’s requirement that men keep their hands and fingernails clean and trimmed and not wear colored nail polish should be compared with the rule allowing women to have longer nails, although not of “exotic length,” and allowing them to wear clear, white, pink, or red nail polish. Finally, the requirement that women wear makeup and lip color at all times should be compared to the prohibition on makeup for men. If the makeup requirement for women is compared to the clean face requirement for men, there can be no dispute that Jespersen created an issue of material fact as to whether the burdens *1086are unequal. “A rule which applies only to women, with no counterpart applicable to men, may not be the basis for depriving a female employee who is otherwise qualified of her right to continued employment.” Gerdom, 692 F.2d at 606 (quoting Allen v. Lovejoy, 553 F.2d 522, 524 (6th Cir.1977)).

Furthermore, the majority neglects burdens other than time and money that are imposed by the policy. The sex-stereotyping inherent in certain appearance standards is a burden that falls more heavily on one sex than the other. Thus, we have recognized that the unequal burdens test does not permit sex-differentiated appearance standards that denigrate one gender based on sex stereotypes. See Gerdom, 692 F.2d at 606 (quoting Carroll, 604 F.2d at 1032-33) (“In Carroll, which involved a requirement to wear uniforms, the court noted that while there is nothing offensive about uniforms per se, requiring only female employees to wear them is ‘disparate treatment ... demeaning to women ... based on offensive stereotypes prohibited by Title VII.’ ”).

Jespersen testified very compellingly to the burdens she personally felt in complying with the makeup policy, explaining that it required her to conform with a feminine stereotype that she felt had nothing to do with making drinks. Given her stellar customer and supervisor evaluations, Jesper-sen is obviously not alone in this analysis. Sex-differentiated appearance standards stemming from stereotypes that women are unfit for work, fulfill a different role in the workplace, or are incapable of exercising professional judgment systematically impose a burden on women, converting such stereotypes into a stubborn reality. See Nevada Dep’t of Hum. Res. v. Hibbs, 538 U.S. 721, 736, 123 S.Ct. 1972, 155 L.Ed.2d 953 (2003) (noting that “mutually reinforcing stereotypes create[ ] a self-fulfilling cycle of discrimination”); see also Carroll, 604 F.2d at 1032-33, cited in Gerdom, 692 F.2d at 606.

This is not to say that all gender-differentiated appearance requirements are prohibited; what violates Title VII are those that rest upon a message of gender subordination. The distinction is apparent in the history of our caselaw on grooming and appearance standards under Title VII. When early challenges to requirements that men keep their hair short arose in the federal courts, those requirements stemmed not from gender subordination, but from fear of a youth counterculture. See Willingham v. Macon Tel. Pub’g Co., 507 F.2d 1084, 1087, 1092 (5th Cir.1975) (quoting Dodge v. Giant Food, Inc., 488 F.2d 1333, 1337 (D.C.Cir.1973)) (“Neither sex is elevated by these regulations to an appreciably higher occupational level than the other.”). Similarly, we have held that requiring men to wear neckties is permissible under Title VII, Fountain v. Safeway Stores Inc., 555 F.2d 753, 755 (9th Cir.1977). However, we have held that requiring women to wear contacts while men may wear glasses, Frank, 216 F.3d at 855, maintain a lower relative weight than men, id., or wear uniforms while men wear appropriate business attire, id. (citing Carroll, 604 F.2d at 1032), is impermissible under Title VII.

Finally, even if all appearance requirements for men are compared to all appearance requirements for women, and even if the burdens engendered by sex-stereotyping are neglected, a reasonable jury could easily conclude that having to wear approximately as much makeup as one was wearing post-makeover, in addition to teasing, curling, or styling one’s hair every day, constitutes more of a burden than having to keep one’s hair short and cut one’s fingernails. All of these activities are ones with which factfinders have everyday familiarity, and “summary judg*1087ment will not lie ... if the evidence is such that a reasonable jury could return a verdict for the nonmoving party.” Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986).

IV

A reasonable factfinder could conclude that the Harrah’s makeup requirement imposes an unequal burden on women, that Jespersen was fired for failure to conform to a sex stereotype, or both. Darlene Jespersen should be allowed to present her case to a jury.

Therefore, I respectfully dissent.

OPINION

McKEOWN, Circuit Judge:

For years, Washington faced problems common to a number of states in their women’s prisons: sexual abuse and misconduct by prison guards, breaches of inmate privacy, and security gaps. A primary driver, according to prison authorities, was the lack of female corree-*982tional officers to oversee female offenders and administer sensitive tasks, such as observing inmates showering and dressing and performing the pat (or “pat-down”) and strip searches that are stitched into the fabric of day-to-day-prison life. After long wrestling with this gender gap, the state undertook a comprehensive assessment and ultimately designated a limited number of female-only correctional positions — specifically, 110 positions to patrol housing units, prison grounds, and work sites. The prison guards’ union, Teamsters Local No. 117 (“Teamsters” or the “Union”), challenged this practice, though it acknowledges the legitimacy of 50 of the female-only designations. This case juxtaposes the prison’s penological interests against male correctional officers who claim the staffing policy discriminates against them on the basis of sex in violation of Title YII of the Civil Rights Act of 1964. 42 U.S.C. § 2000e.

We conclude that the Washington Department of Corrections’ (the “Department” or the “state”) individualized, well-researched decision to designate discrete sex-based correctional officer categories was justified because sex is a bona-fide occupational qualification (“BFOQ”) for those positions. The Union’s thin eviden-tiary submissions — coupled with expert claims that were largely unsubstantiated or missed the point — failed to raise a material factual issue. Indeed, the startling statement by one of the Union’s experts underscores the legitimacy of the state’s efforts to combat sexual abuse: “Sexual abuse is present in all areas of our society ... [FJemale inmates must be taught as part of the rehabilitation process to deal with all abusive staff: males and females ...” The Department was well-justified in concluding that rampant abuse should not be an accepted part of prison life and taking steps to protect the welfare of inmates under its care. We affirm the district court’s grant of summary judgment in favor of the Department.

Background

The Department runs two women’s prisons. The Washington Corrections Center for Women in Gig Harbor has a capacity of 738 inmates, although it is often overcrowded. That prison runs the gamut from minimum security facilities to housing for violent offenders and those with mental health issues. It also houses Washington’s death row for female prisoners. The second facility is Mission Creek Corrections Center for Women in Belfair, a smaller minimum-security prison that houses around 300 inmates.

For decades, men dominated the ranks of prison guards, though neither party has provided precise figures. Facing a shortage of femále guards in the late 1980s, state prison administrators began allowing male guards to perform random, clothed body searches — commonly known as pat searches — of the female inmates at Washington Corrections Center. Female inmates challenged these cross-gender searches as unconstitutional. The district court granted an injunction and halted the practice. Sitting en banc, we affirmed, concluding that cross-gender body searches inflict unnecessary and wanton pain on female inmates, many of whom have suffered a history of sexual abuse before incarceration, and, therefore, violate the Eighth Amendment.1 Jordan v. Gardner, 986 F.2d 1521, 1531 (9th Cir.1993) (en banc). Under both Jordan and a later-enacted Washington law, female correctional officers must perform all non-emer*983gency pat searches of female inmates. Wash. Rev.Code § 9.94A.631(2) (2012).

In the years following Jordan, the Department struggled with the challenges posed by having an overwhelmingly male workforce. In 1998, it asked the Washington Human Rights Commission (the “Commission”) for an opinion on proposed correctional assignments reserved exclusively for female officers. The Commission did not favor the Department’s approach at that time.2

In 2008, Congress passed the Prison Rape Elimination Act, which included findings that, based upon experts’ conservative estimates, 13% of prisoners had been sexually assaulted while in prison. See 42 U.S.C. § 15601. The legislation also noted that many instances of abuse go unreported and prison personnel were inadequately trained to deal with these issues. See id. §§ 15601-09. Under the Act, the Department received a $1 million grant to hire two full-time employees to investigate sexual misconduct allegations in prisons.

In the years that followed, the Department fielded widespread allegations of sexual abuse in its women’s prisons. State officials, for example, substantiated 46 instances of misconduct in a single two-and-a-half-year stretch. In the aftermath, in 2007, female inmates brought a class action in state court alleging misconduct at the Washington Corrections Center. The complaint detailed incidents where guards assaulted and fondled female inmates and forced them to perform oral sex and masturbate in the presence of male officers. Complaint, Jane Doe v. Clarke, No. 07-2-01513-0, Dkt. No. 4 (Thurston Co.Super.Ct. July 31, 2007).

Within a week of the filing of that lawsuit, the Department hired a consultant to investigate sexual activity and misconduct. After a four-month internal investigation, the consultant detailed the facts in a 240-plus-page report. The investigation included interviews with 72 “Jane Doe” inmates, who alleged that they faced sexual advances and harassment from prison guards. Among the lurid details, male guards twice impregnated inmates and smuggled contraband in exchange for sexual favors.

The Department also hired two additional consultants to review prison practices. Marianne McNabb, of the Social Research Institute based in Olympia, Washington, wrote:

Cross-sex supervision is currently one of the most significant issues facing the administration of women’s prisons. Today in many states, over 50 percent of the custody force in prisons for women are men. The fact that so many women in prison have experienced sexual abuse by men makes them different from male prisoners who do not share that history and therefore do not experience the same level of anxiety or.violation as do women, when under the custody or supervision of an officer of the opposite sex.

McNabb noted that several jurisdictions, including Idaho and Michigan, “have established sex-specific posts in female institutions” in response to these dynamics. Her report concluded, ‘While this may *984seem to be a solution for many of the concerns identified, this practice is generally not fully understood or accepted by staff and has faced some legal challenges.”

Donald Kelchner, superintendent of the Pennsylvania Department of Corrections, urged the Department to adopt a host of reforms, including guard assignments reserved specifically for women. In particular, Kelchner recommended that the state ensure any double-staffed housing units have at least one female guard. Kelchner concluded, “It is more desirable in an institution housing females to have a higher number of female staff, to work with and supervise the inmates.”

Following the expert recommendations, the Department in January 2008 implemented an array of reforms to “reduce prison sexual assaults and related behavior.” Those efforts included aggressive recruitment of female prison guards; pre-hiring psychological testing; training programs to enhance “gender awareness”; and the installation of privacy curtains, security cameras, and restricted access entry cards.

Then, in May 2008, prison administrators again requested guidance from the Commission on the Department’s proposed 110 female-only guard post assignments at the two prisons. The Department submitted a tailored request for each post, explaining the job responsibilities and why the positions needed a female officer. The state told the Commission that “[increasing the number of female staff will reduce the risk of sexual misconduct, reduce allegations of sexual misconduct, and protect male staff exposed to vulnerable situations” and unfounded complaints of abuse. The state also emphasized the privacy requirements of female inmates and the operational need to have female officers on hand to perform necessary searches and other tasks. The requested staffing changes, according to the state, would “ensure the security of the prisons, safety of incarcerated offenders, and protection of the privacy and dignity of female offenders.”

After touring the prisons, interviewing administrators, and collecting detailed documentation, the Commission in February 2009 approved the Department’s request for all 110 positions. The Commission offered Teamsters the chance to provide input but none was forthcoming. The Commission determined that, with the then-existing staff makeup at the prisons, the state was “unable to ensure a proper balance between security considerations and the privacy rights of offenders” and that there were no reasonable alternatives to sex-based staffing.

The class action settled soon after. As part of the settlement agreement, the Department agreed to enforce a “zero tolerance” policy regarding sexual misconduct, not to rehire five male correctional officers accused of abuse, and to submit regular reports on staff misconduct in women’s prisons. The settlement also included an undisclosed payout to abused prisoners. Stipulation and Proposed Order, Jane Doe v. Clarke, No. 07-2-01513-0, Dkt. No. 170 (Thurston Co.Super.Ct. Aug. 6, 2010).

The Department’s reprieve from the courtroom did not last long. In September 2011, Teamsters, which represents some 6,000 state correctional workers, filed this federal lawsuit, alleging that the sex-based staffing policy implemented in 2009 violates the civil rights of male prison guards.3

*985At the conclusion of discovery, the district court granted summary judgment for the state. Teamsters Local Union No. 117 v. Wash. Dep’t of Corr., No. C11-5760 BHS, 2013 WL 1412335 (W.D.Wash. Apr. 8, 2013). The district court first found that the Union had failed to demonstrate the type of “cognizable injury” required to trigger Title VII liability. The court noted that the record developed by Teamsters included only “hypothetical evidence” of the damages its members would face, thus entitling the state to summary judgment. Id. at *4. Alternatively, the district court also granted summary judgment for the state on the question of sex discrimination. As an initial matter, the district court ruled that judicial.deference to state pris- . on officials was warranted. Id. The court concluded that, although there may have been factual questions on whether female guards were needed inside the housing units to prevent sexual assaults, the staffing policy was justified as a BFOQ to protect the privacy of inmates for each job category. Id. at *5-9.

Analysis

I. Standing

This case is a cautionary tale about the threshold importance of standing. The state argues, for the first time on appeal, and after receiving Teamsters’ opening brief, that the Union lacks standing because it produced no evidence that any of its members suffered concrete injury. The issue is not, as Teamsters urges, whether the question was addressed by the district court; rather, as a jurisdictional matter, “a challenge to constitutional standing is one which we are required to consider” apart-from whether it was argued or addressed below. Laub v. U.S. Dep’t of Interior, 342 F.3d 1080, 1085 (9th Cir.2003) (internal quotation mark and citation omitted).

The standing inquiry is governed by the familiar elements of injury-in-fact, traceability, and redressability. “To establish Article III standing, an injury must be concrete, particularized, and' actual or imminent; fairly- traceable to the challenged action; and redressable by a favorable ruling.” Clapper v. Amnesty Int’l USA — U.S.-, 133 S.Ct. 1138, 1147, 185 L.Ed.2d 264 (2013) (internal quotation marks and citation omitted). For associational standing, Teamsters must show in addition that “its members would otherwise have standing to sue in their own right.” Associated Gen. Contractors of Am., San Diego Chapter v. Cal. Dep’t of Transp., 713 F.3d 1187, 1194 (9th Cir.2013). In other words, the Union “must show that a member suffers an injury-in-fact that is traceable to the defendant and likely to be redressed by a favorable decision.” Id.

The complaint posited varied theories of injury: male guards were forced to transfer jobs and prisons; suffered lost earnings, including overtime pay; were laid off; and experienced “loss of status, diminished sense of self-worth, anxiety, emotional distress, embarrassment, humiliation, mental anguish, and other related damages.” Surprisingly, proof of these general allegations did not materialize as evidence. The Union’s submissions on summary judgment are thin, at best, in terms of identifying one or more specific members who suffered injury. Ironically, it is the testimony the Union elicited from the state that provides the strongest support for the Teamsters’ constitutional standing. Throughout discovery, the state did not dispute the general allegations that its staffing policy resulted in the transfer of male guards and lost overtime opportunities. While no evidence linked a specific officer with a discrete wage loss, Superintendent Doug Cole indicated that six male officers had been displaced from their reg*986ular shifts — mentioning two by name — and agreed with the Union’s lawyer that, with respect to male correctional officers, the staffing changes would result in “some reduction in their overtime opportunity ...” Every male correctional officer who was displaced from his regular shift, however, was offered a position on a different shift.

As the state points out, when a challenge to standing is raised at summary judgment, a plaintiff organization must “submit competent evidence, not mere allegations, to demonstrate that at least one of its members had standing.” Associated General, 713 F.3d at 1194. This is a settled proposition, though curiously the state never moved for summary judgment on standing nor contested the Union’s standing allegations. Nonetheless, a party is not excused from establishing standing simply because the opposing party did not tumble to the issue until the appeals stage. See Gest v. Bradbury, 443 F.3d 1177, 1181 (9th Cir.2006) (“The[] elements of standing must be supported in the same way as any other matter for which a plaintiff bears the burden of proof, ie., with the manner and degree of evidence required at the successive stages of the litigation.”).

In response to the state’s motion to dismiss on appeal, Teamsters moved to supplement the record with affidavits from five Union members and an administrator. Why this evidence surfaced only on appeal is a mystery. Ordinarily, we do not allow parties to supplement the record on appeal absent “extraordinary circumstances.” United States v. Boulware, 558 F.3d 971, 975-76 (9th Cir.2009). Here, however, we accept the affidavits for the limited purpose of confirming the job-related harms that the Department acknowledged in general terms during discovery. Doing so is “in the interests of justice and efficiency,” Ouachita Watch League v. Jacobs, 463 F.3d 1163, 1170 (11th Cir.2006) (internal citation omitted), since the hiring policy has been on the books since 2009 and has been the subject of multiple proceedings; a remand or dismissal on procedural grounds would merely prolong resolution of the underlying issues. In the affidavits, male correctional officers assert that they have suffered precisely the types of harm that the state acknowledged in discovery— most importantly, lost overtime.

Although the Union hardly made a slam-dunk showing of prospective harm, the record as supplemented on appeal reflects the bare minimum necessary to satisfy the threshold requirement of standing. The Department’s motion to dismiss the appeal on standing grounds is denied.

II. Title VII and the Bona Fide Occupational Qualification

Title VII of the Civil Rights Act of 1964 prohibits employment practices that discriminate on the basis of race, color, religion, sex, or national origin. 42 U.S.C. § 2000e-2. Nevertheless, a facially discriminatory employment practice, such as the sex-based hiring practice we have here, may pass legal muster if sex is a bona fide occupational qualification or BFOQ. That narrow exception — found in § 2000e-2(e)(l) — provides:

[I]t shall not be an unlawful employment practice for an employer to hire and employ employees ... on the basis of ... sex ... where ... sex ... is a bona fide occupational qualification reasonably necessary to the normal operation of that particular business or enterprise.

The Supreme Court has emphasized that “[t]he BFOQ defense is written narrowly, and this Court has read it narrowly.” UAW v. Johnson Controls, Inc., 499 U.S. 187, 201, 111 S.Ct. 1196, 113 L.Ed.2d 158 (1991). The BFOQ defense applies to “special situations” where employment discrimination is based upon “objective, verifiable requirements” that “con*987cern job-related skills and aptitudes.” Id. An “occupational qualification” means a “qualification[ ] that affect[s] an employee’s ability to do the job.” Id.

Under our precedent, the BFOQ defense “may be invoked only when the essence of the business operation would be undermined by hiring individuals of both sexes.” Breiner v. Nev. Dep’t of Corr., 610 F.3d 1202, 1210 (9th Cir.2010) (emphasis in original) (internal quotation marks and citation omitted). To justify discrimination under the BFOQ exception, an employer must show, by a preponderance of the evidence, that: (1) the “job qualification justifying the discrimination is reasonably necessary to the essence of its business”; and (2) that “sex is a legitimate proxy for determining” whether a correctional officer has the necessary job qualifications. Ambat v. City & Cty. of San Francisco, 757 F.3d 1017, 1025 (9th Cir.2014) (quoting Breiner, 610 F.3d at 1210).

In light of these demanding legal standards, BFOQs are few and far between. In many industries, it is difficult to imagine any jobs that would qualify as BFOQs. However, the “unique context of prison employment,” id. at 1028, is one area where courts have found sex-based classifications justified. The Supreme Court directly addressed the prison environment in just one case, Dothard v. Rawlinson, 433 U.S. 321, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977). The Court held that, in the context of a maximum-security facility “where violence is the order of the day” and sex offenders were interspersed with other prisoners, a female guard’s sex may “undermine her capacity to provide the security that is the essence of a correctional counselor’s responsibility.” Id. at 335-36, 97 S.Ct. 2720. Referencing Dothard, the Court in Johnson Controls explained that “[s]ex discrimination was tolerated because sex was related to the guard’s ability to do the job — maintaining prison security.” 499 U.S. at 202, 111 S.Ct. 1196.

When justified under the circumstances, we and other circuits similarly have upheld sex-based correctional officer assignments in women’s prisons. See Robino v. Iranon, 145 F.3d 1109, 1110 (9th Cir.1998) (per curiam) (BFOQ designation of six correctional officer positions at Hawaii women’s prison); Everson v. Mich. Dep’t of Corr., 391 F.3d 737, 749-50 (6th Cir.2004) (BFOQ designation of 250 correctional officer positions at Michigan women’s prisons); Tharp v. Iowa Dep’t of Corr., 68 F.3d 223, 224 (8th Cir.1995) (BFOQ designation of all correctional officer positions in women’s residential unit within a mixed-gender minimum security prison); cf. Torres v. Wisc. Dep’t of Health and Soc. Servs., 859 F.2d 1523, 1532 (7th Cir.1988) (en banc) (noting that prison officials are not required to provide “objective evidence, either from empirical studies or otherwise,” and remanding the denial of a BFOQ designation for evaluation “on the basis of the totality of the circumstances contained in the entire record.”).

Although limited gender discrimination may be permissible in the prison employment context, prison administrators do not get a free pass. The Department must have an objective “basis in fact” for “its belief that gender discrimination is ‘reasonably necessary’ — not merely reasonable or convenient — to the normal operation of its business.” Everson, 391 F.3d at 748 (citing W. Air Lines, Inc. v. Criswell, 472 U.S. 400, 414, 105 S.Ct. 2743, 86 L.Ed.2d 321 (1985)). This means prison administrators “seeking to justify a BFOQ must show a high correlation between sex and ability to perform job functions.” Breiner, 610 F.3d at 1213 (internal quotation marks and citation omitted). Speculation about gender roles is insufficient — the evidence must demonstrate that prison ad*988ministrators had a “concrete, logical basis for concluding that gender restrictions are reasonably necessary” and that alternatives to sex discrimination have been “reasonably considered and refuted.” Ambat, 757 F.3d at 1028 (internal quotation marks and citation omitted).

An additional significant factor is at play: deference to prison officials. “Judgments by prison administrators that are the product of a reasoned decision-making process, based on available information and expertise, are entitled to some deference.” Breiner, 610 F.3d at 1212 n. 6 (internal quotation marks omitted); see also Robino, 145 F.3d at 1110 (holding that, where Hawaii prison administrators appointed a task force to review prison policies, their “professional judgment is entitled to deference”).

Although we have not offered up a cookbook for a “reasoned decision-making process,” cases that have invoked the deference principle point to undertakings that address systemic issues, consider outside views and data, and weigh reasonable alternatives. See Robino, 145 F.3d at 1111 (deferring to Hawaii prison administrators who directed a “specially appointed task force” to study prison problems); Everson, 391 F.3d at 741-45 (deferring to Michigan prison administrators where they conducted three studies, one pursuant to a settlement with the Department of Justice). To be sure, although studies and empirical data are indicia of a deliberative approach, we have emphasized that “the decision-making process supporting a discriminatory policy” need not “take any particular form.” Ambat, 757 F.3d at 1026. Deference is a threshold legal determination.

The Department’s exhaustive process fits well within the rubric of “reasoned decision making” and is entitled to deference. After the Jane Doe prisoner class action was filed in 2007, the Department did not rush headlong into sex-based staffing. Instead, it hired experts, consulted with other states, reviewed relevant case-law, documented scores of sexual misconduct allegations and investigated many more, and sought advice from the Human Rights Commission. Drawing on its decades of experience, the state did not view sex-based staffing as a panacea, instead proposing a package of reforms that included measures such as applicant psychological testing, sex-awareness training, and security cameras.

Teamsters argues that the Department implemented sex-based staffing “during a time of Departmental crisis” and in a “panic” that was little more than a “desperate attempt” to settle the state court class action. The Union’s characterization begs the question: If sordid details of sexual abuse and constitutional violations do not inspire a “crisis” and feelings of “panic,” then what does? The state shouldn’t be demonized for kicking into gear to find a remedy for its long-running challenges. In any event, our inquiry does not turn on the subjective state of mind of the Department’s leadership. The Department undertook a rigorous review of its staffing policies to address the issues raised in the report and the class action.

The Department’s thorough, thoughtful approach stands in stark contrast to the sheriff in Ambat, who rejected out of hand alternatives to discrimination — such as pre-hiring screening, surveillance cameras, and training — and declined to order an internal investigation or hire outside consultants. See 757 F.3d at 1022, 1026. The sheriff did not consult deputies directly responsible for prisoner supervision or other jurisdictions with similar policies, and no internal review documented the extent of misconduct. Id.

Ambat instructs that “[djetermining whether a corrections official is entitled to deference is a fact-intensive and case-spe*989cific inquiry” that is “generally within the discretion of the. district court.” Id. at 1026. The district court found that the Department’s process merited deference, and we see no reason to conclude otherwise. Accordingly, we give “some deference,” Rabino, 145 F.3d at 1110, to Washington’s prison administrators, although we remain mindful of the antidiscrimi-nation mandate of Title VII.

III. The Prison Policy and BFOQ Requirements

In 2009 the Department determined that designating 110 female-only guard positions at the two prisons would substantially improve prison security, protect the privacy of female inmates, and prevent sexual assaults. Teamsters challenges approximately sixty of those positions, which fall into four general categories: medium- and high-security housing units (18 positions); programs and activities supervisors (3 positions); work crew supervisors (6 positions); and relief posts to replace female guards who are on breaks or absent from work (32 positions).4

The Union paints the Department staffing policy as “broad and overreaching” — a “blunderbuss approach to the issue.” The record demonstrates the opposite. Instead of a blanket ban on male prison personnel, the Department crafted the staffing needs to fit each specific facility and guard post. It targeted only guard assignments that require direct, day-today interaction with inmates and entail sensitive job responsibilities such as conducting pat and strip searches and observing inmates while they shower and use the restroom.

As the Union’s expert acknowledged, “[n]o remedy is perfect nor perfectly effective.” We couldn’t agree more. This reality underscores the rationale for deference to prison administrators and the hazard of nitpicking the state’s thoughtful response to deep-rooted problems in its women’s prisons.

A. The Department’s Policy Rationales Are Reasonably Necessary to the Essence of Prison Administration

At issue on appeal is whether the state established as a matter of law that sex-based restrictions are “a bona fide occupational qualification reasonably necessary” to normal prison operations. Under the well-worn standard of Federal Rule of Civil Procedure 56, we affirm the district court’s grant of summary judgment because there is “no genuine dispute' as to any material fact and the [state] is entitled to judgment as a matter of law.” Fed. R.Civ.P. 56(a). Teamsters failed to produce “specific facts showing there is a genuine issue for trial” to survive summary judgment. Celotex Corp. v. Catrett, 477 U.S. 317, 324, 106 S.Ct. 2548, 91 L.Ed.2d 265 (1986) (quoting Fed.R.Civ.P. 56(e)).

Curiously, Teamsters did not offer specific testimony from its members, former guards, or others with actual knowledge of on-the-block operations at the prisons. Although Teamsters offered up the reports of two experts, in the end their testimony does not create a material factual issue. For starters, the reports generally fail to address the specific posts at issue. Instead, they rehash alternatives to sex-based staffing that were exhausted and rejected by prison administrators or serve up proposals without any evidence of efficacy or practicality. Finally, the Union’s experts opine on the interpretation of so*990cial science research that is not central to the state policy in any event.

Although the sexual assaults that spawned the Jane Doe class action permeate this lawsuit, the state did not justify its BFOQ positions solely as a means to prevent sexual assaults. Instead, it identified several intertwined reasons for designating the female-only positions. In the initial request to the Human Rights Commission, the Department cited the need to “enhance the security of the prisons, safety of staff and offenders, and to protect the privacy and dignity of female offenders.” The Commission, in turn, concluded that absent the BFOQ designations, the prison is “unable to ensure a proper balance between security considerations and privacy rights of offenders” and endorsed the female job assignments “for the explicit purpose of ensuring privacy rights of female offenders.”

Not surprisingly, under our precedent all of these interrelated objectives go to the heart of prison operations. In Ambat, we held that, as a matter of law, “protecting female inmates from sexual misconduct by male deputies, maintaining jail security, [and] protecting inmate privacy” were all reasonably necessary to the essence of prison administration. 757 F.3d at 1027-28. The same holds true here.

Security, of course, is the paramount concern of prison administrators. As the Supreme Court has noted: “The essence of a correctional counselor’s job is to maintain prison security.” Dothard, 433 U.S. at 335, 97 S.Ct. 2720; see also Everson, 391 F.3d at 753 (“Unquestionably, the security of the prisons relates to the essence of [prison business].”). That maxim is no less true today. Security concerns are necessarily intertwined with prison programs and objectives.

Inmate privacy encompasses the inmate’s “interest in not being viewed unclothed by members of the opposite sex”— , an interest that “survives incarceration” despite prisoners’ diminished privacy expectations. See Robino, 145 F.3d at 1111. In the same vein, inmates have a privacy interest in having non-emergency strip and pat searches — a pervasive fact of prison life — performed by guards of the same sex. See Jordan, 986 F.2d at 1524; Tharp, 68 F.3d at 226.

Preventing sexual assaults is also a legitimate prison objective. First and foremost, prison administrators have a high interest in shielding inmates from abusive and inherently coercive encounters. Indeed, even allegations of sexual misconduct can destabilize prison life: they can breed mistrust and damage morale among officers and prisoners; drain prison resources; and undercut the effectiveness of male officers with the looming threat of a career-ending accusation. See Robino, 145 F.3d at 1111 (discussing damage to prison morale caused by allegations of male staff sexual misconduct); Everson, 391 F.3d at 753 (“[Allegations of sexual abuse, whether true or not, create a ‘poisoned atmosphere’ that breeds misconduct on the part of inmates and guards.”).

Amazingly, one of the Union’s experts offered the following view:

Female inmates cannot be shielded from the world in which we live. If they are to reintegrate into society, they have to be taught how to deal with abusive staff, male or female. They have to be taught what constitutes a healthy interaction and what does not. They cannot learn those skills if they are sheltered from contact with males in a position of authority. ¶ Sexual abuse is present in all areas of our society: in schools, (at all levels), business, government, military and families. Just as females have to be taught how to deal with those abuses in the larger society, female inmates must be taught .as part of the rehabilitation *991process how to deal with all abusive staff: males and females, custody staff and civilian staff.

To state something so obvious we never imagined it would need to be written: we reject any suggestion that female prisoners would benefit from being subjected to abusive prison guards as “part of the rehabilitation process” so that they may better “reintegrate into society.” See, e.g., Prison Rape Elimination Act, 42 U.S.C. § 15601(11) (“Victims of prison rape suffer severe physical and psychological effects that hinder their ability to integrate into the community and maintain stable employment upon their release from prison.”).

We have little difficulty holding that the state’s reasons for adopting the BFOQ designations — improving security, protecting inmate privacy, and preventing sexual assaults — are each reasonably necessary to the essence of operating Washington’s women’s prisons. That conclusion does not end the analysis, however. The state also must demonstrate that sex is a “legitimate proxy” to achieve one or more of these goals, meaning that there is a “high correlation between sex and ability to perform job functions.” Breiner, 610 F.3d at 1213 (quoting Johnson Controls, 499 U.S. at 202, 111 S.Ct. 1196). In addition, the state must show that alternatives to the sex-based classification were “reasonably considered and refuted.” Ambat, 757 F.3d. at 1028.

Before addressing these remaining requirements in the context of specific positions, we consider Teamsters’ overarching arguments that the staffing policy is based on stereotypes and that the state failed to consider nondiscriminatory alternatives.

Teamsters argues at length that the state policy is based on an impermissible stereotype that male guards are more likely to commit sexual misconduct than their female counterparts. This stereotyping argument misses the mark. To begin, the Union acknowledged that the policy was adopted in the face of documented allegations of abuse.5 The Department also did not rest on assumptions; it provided objective legal and operational justifications for why only women can perform particular job functions, like observing inmates unclothed and conducting non-emergency searches.

We also reject Teamsters’ argument that the Department could simply have hired new executives or reconfigured prison layouts. As our discussion of the day-to-day realities of the positions at issue demonstrates, neither of those alternatives actually addresses the specific operational challenges of maintaining prison security, preserving inmates’ privacy, and stopping abuse.

B. Sex is an Objective, Verifiable Job Qualification for the Designated Positions

We conclude that sex is an objective, verifiable job qualification for the posts designated as female-only by the Department and that the Department appropriately considered reasonable alternatives.

*9921. Housing Units

The staffing restriction with the “largest impact,” according to the Union, involves 18 positions at the medium- and high-security housing units at Washington Corrections Center.6 The housing units have two guards on duty on each shift. Unlike other states, the Department did not ban male guards entirely; rather, the staffing policy requires at least one female guard per shift, an approach recommended by one of the state’s consultants.

In the housing units, correctional officers “must conduct pat and strip searches of female offenders entering and leaving the facility” as well as frequent random and suspicion-based searches within the housing units. In the segregation and mental illness units, inmates are strip searched every time they enter or leave their cells. Except in emergency circumstances, male guards cannot legally perform any of these searches. Jordan, 986 F.2d at 1523; see also Wash. Rev.Code § 9.94A.63K2).

Beyond searches, officers in the housing units also “may encounter female offenders in varying states of undress while showering, toileting, and dressing.” Guards must collect urine samples from inmates, and a failure, to “observ[e] the offenders during the entire process of urinalysis collection significantly impacts the reliability of the test results ...” According to the state, “[m]ale staff cannot observe female offenders when they are engaged in these activities.”

Given these operational needs, there is no reasonable substitute for having female guards inside housings units, according to the Department. Notably, temporarily removing a female guard from another part of the prison to cover in a housing unit “creates a gap for dealing with privacy issues at the post vacated.” At best, that solution fixes one problem but creates another.

The evidence Teamsters puts forward to counter the Department’s justifications is entirely inapposite. One of its experts points out that sexual assault is not a severe problem in medium- and high-security housing because “as the level of security increases, the opportunity for sexual assault decreases.” This may be true, but it fails to acknowledge that the staffing decisions were designed to protect inmate privacy, which is “essential to the operation of a corrections facility and has been recognized as justifying facially discriminatory policies in other contexts.” Ambat, 757 F.3d at 1028. The Union’s other expert quarrels with citations to social science regarding female inmates’ privacy needs and matters relating to sexual relationships between inmates and guards. This testimony again fails to raise any genuine dispute of material fact as to the Department’s reasoned determination that the realities of operating Washington’s women’s prisons necessitate designating these specific positions as female-only.

2. Programs and Activities

Programs and activities officers directly supervise inmate activities such as educational and religious classes, gym, crafts, and visitation hours. During these programs and activities, inmates are searched at random and if suspected of hiding contraband. Guards must collect urine samples from inmates and at times relieve housing unit officers, which requires “room checks” where they may “encounter female offenders in varying states of undress .... ” These guards also supervise *993visitation hours, after which 50% of inmates are pat searched and 50% are strip searched. To fulfill these job functions, the state designated three programs and activities positions as female-only.

The Union’s proposed alternative to designating these positions as female-only is a return to the system employed for the last two decades: dispatching female officers as rovers — or “response and movement” guards, in prison lingo — who could be paged when needed for searches.7 The Union offers no data, expert testimony, or other evidence to support the efficacy of this approach. Instead, undisputed evidence established that the rover system was rife with problems, to say the least. During this era, prison administrators “shuttle[d] women staff from location to location throughout the prisons to perform essential security procedures, leaving other areas of the prison without appropriate staffing.” Wait times for searches lasted an hour or more. With female guards stretched thin, inmates went unsupervised showering, using the restroom, or dressing — raising security and safety risks. Superintendent Eldon Vail testified that, before the BFOQ positions were implemented, the prison functioned “in the broadest sense” but the shortage of female guards restricted the prison’s ability to deploy unannounced, random pat searches, an important tool in preventing the flow of contraband.

In light of this checkered history, the Union’s conclusory assertion that the Department successfully “managed [privacy and search] issues for at least two decades” rings hollow. FTC v. Publ’g Clearing House, Inc., 104 F.3d 1168, 1171 (9th Cir.1997) (“[CJonclusory, self-serving statements in appellate briefs ... are insufficient to create a genuine issue of material fact.”). We will not displace prison administrators’ experience and expertise in favor of an alternative that boils down to the “same old, same old.” Cf. Torres, 859 F.2d at 1529 (“[PJrison administrators always have been expected to innovate and experiment.”) (citing Turner v. Safley, 482 U.S. 78, 107, 107 S.Ct. 2254, 96 L.Ed.2d 64 (1987) (prison administrators must be allowed “to adopt innovative solutions to the intractable problems of prison administration”)).

3. Work Crews

Work crew officers escort groups of ten prisoners to off-site work locations and supervise their workdays. Searches are again part and parcel of the job — comprising 70% of day-to-day responsibilities. Strip searches are required each time an inmate leaves and reenters the prison grounds. Before the 2009 staffing policy, female officers had to be “pulled from somewhere else in the facility” to conduct these searches, which can “creat[e] [] a staff shortage in another area of the facility” and pose “a potential security risk,” according to the Human Rights Commission. During the workday, officers also must accompany female inmates as they use the restroom. The Department concluded that, because of these job responsibilities, it needed female officers alongside work crews. The Department therefore designated six positions as female-only.

Nonetheless, with respect to work crews, the Union argues that the Department should merely station female guards at prison entry and exit points. If the need for a search arises “while work is in progress, this would constitute an ‘emergent’ search which is not prohibited for a male officer as a matter of law, policy or *994contract.” The Union produced no evidence or legal support for its emergency-search proposal. Even if the Department could disingenuously label every work-site search as an emergency, the state’s interest is broader than merely avoiding illegal searches. Having male officers conduct pat searches under any non-emergency circumstances is undesirable and harmful to prisoner privacy and security.

Staging female officers at entry and exit points also ignores the state’s interest in preserving security during work assignments. The record showed that at least two inmates escaped from public bathrooms while on work crews, when they were not watched by male guards and no female guards were on hand. The Union does not explain, much less provide evidence for, how its alternative proposal would address concerns about on-the-job observation.

4. Relief Posts

Officers in the 32 relief positions substitute for female guards in female-only positions when they have a regular day off, are on vacation, or are out sick. The relief officers perform the job responsibilities described above in housing units and elsewhere. As the Human Rights Commission put it, the relief positions “alleviate under-staffing of female officers, because a BFOQ position needs to be relieved by a BFOQ position.” In other words, if only male officers are available to fill in for BFOQ positions, it undermines the documented need of making those positions female-only in the first place.

According to the Union, 32 relief positions is too many, so the issue “must be reserved for trial because the Court cannot assess whether the relief sought was excessive without conducting a careful analysis of all such positions. To survive summary judgment, however, the Union “may not merely state that it will discredit the moving party’s evidence at trial and proceed in the hope that something can be developed at trial in the way of evidence to support its claim.” T.W. Elec. Serv. v. Pac. Elec. Contractors Ass’n, 809 F.2d 626, 630 (9th Cir.1987). Argument without evidence is hollow rhetoric that cannot defeat summary judgment.

Conclusion

We affirm the district court’s grant of summary judgment for the state. The Washington Department’s creation of a narrow category of female-only job assignments is a “bona fide occupational qualification reasonably necessary to the normal operation” of the women’s prisons.8

AFFIRMED.