Cynthia J. Fisher, Plaintiff-Appellee-Cross-Appellant v. Vassar College, Defendant-Appellant-Cross-Appellee

70 F.3d 1420

Opinion

JACOBS, Circuit Judge:

Following a three week bench trial, the United States District Court for the Southern District of New York (Motley, J.) found that Vassar College discriminated against plaintiff Cynthia J. Fisher in 1985 when it denied her tenure as a professor in its biology department (a) by reason of her sex together with her status as a married woman, in violation of Title VII of the Civil Rights Act of 1964 (“Title VII”) and (b) by reason of her age, in violation of the Age Discrimination in Employment Act (“ADEA”). The district court rejected, however, plaintiffs claim of “simple” sex discrimination based solely on her status as a woman. The district court further found and concluded that Vassar’s discrimination against Fisher violated provisions of the Equal Pay Act. Based on these findings, recorded in a detailed opinion, see Fisher v. Vassar College, 852 F.Supp. 1193 (S.D.N.Y.1994), the district court directed Vassar to pay an aggregate money judgment of $626,872.12 together with Fisher’s attorneys’ fees. The judgment also ordered that Fisher be restored to the Vassar faculty for a period of two years, after which time she again would be evaluated by the school for the purpose of determining whether she was qualified for promotion to full professor with tenure.

Vassar challenges the findings relied on by the district court to support judgment against Vassar under Title VII, the ADEA and the Equal Pay Act, and seeks to vacate the district court’s award of attorneys’ fees. In a cross-appeal, Fisher argues that the district court erred in (1) finding that plaintiff had failed to prove a case of “simple” sex discrimination in violation of Title VII, and (2) failing to order her reinstatement with full tenure without further tenure-track service or further tenure evaluation.

The trial focused on two issues: Fisher’s qualifications for tenure and Vassar’s history of tenure decisions involving candidates who were married women. As to her qualifications for tenure, Fisher sponsored testimony by several experts in the field of biology and voluminous documentation comparing her teaching and scholarly record to other Vassar biology professors who had been considered for tenure. As to Vassar’s treatment of married women candidates for tenure, Fisher presented (inter alia) numbers purporting to demonstrate that no married woman on the Vassar faculty had been granted tenure in the “hard” sciences in the three decades preceding plaintiff’s 1985 denial of tenure; the testimony of former Vassar biology professors relating the events of their careers at Vassar and expressing their belief that the Vassar biology department discriminated against married women; and the testimony of a sociologist, describing the discrimination generally experienced by married women in the sciences and opining that Vassar had discriminated against plaintiff. Finally, Fisher served as her own chief witness, recounting her experiences in the Vassar biology department and related anecdotes.

Vassar defended on two fronts: Fisher’s lack of qualifications for a tenured professorship in Vassar’s biology department, and the absence of any record of discrimination by Vassar on the basis of sex or on the basis of sex conjoined with marital status. Most of the testimony offered by Vassar came from current and former Vassar biology professors and administrators, who described the procedures used in the Vassar tenure review process generally and the care with which Vassar proceeded in Fisher’s particular case. Defendant also sought to present its own data concerning Vassar’s school-wide record of promoting married women. The district court refused to admit these statistics into evidence.

The district court’s opinion rejected plaintiffs claim of “simple” sex discrimination, but found that, as a married woman, plaintiff had established a claim of “sex plus” discrimination in violation of Title VII of the Equal Rights Act (“Title VII”). After a detailed critique of the department’s report recommending against tenure for Fisher, the district court concluded that the department’s assessment of her qualifications was biased and that plaintiff had met all the recognized requirements for tenure in the biology department:

The court ... finds that the termination of plaintiffs employment resulted not from any inadequacy of her performance or qualifications or service, but from the pre-textual and bad faith evaluation by Vassar of her qualifications ...

Fisher, 852 F.Supp. at 1218. In arriving at its conclusions, the court credited plaintiffs theory that Vassar had a practice of discriminating against married women employed in the “hard” sciences, took note of plaintiffs eight-year absence from her career to raise a family, and addressed the contentious issue of whether Title VII permits Vassar to weigh that fact against her in its tenure decision:

At trial plaintiff presented evidence in support of her claim that she was discriminated against because of her marital status: 1) evidence distinguishing between departments in the “hard” sciences (including the Mathematics, Physics, Chemistry, Geology, Biology and Computer Science departments) and the departments in the rest of the College; 2) evidence that traditionally the “hard” sciences faculties have been composed of men (married and single) and women (single only); 3) evidence that although married women constituted 52 percent of the women in science nationally, no married women had ever received tenure in the “hard” sciences at Vassar. Plaintiff claims that it was her decision to be a wife, mother and a scientist before seeking tenure at Vassar that resulted in the denial of tenure application.

Part of plaintiffs claim of sex discrimination due to her marital status involves the fact that plaintiff left formal academia for approximately eight years in order to raise a family. In 1965, when plaintiff left academia, marriage and family went hand in hand for women, especially for women of plaintiffs race and class. Prior to the 1970’s Women’s Movement and social advances that have permitted women more freedom in shaping their private and professional lives, women who were of plaintiffs race and class rarely had children out of wedlock. Likewise, women of plaintiffs race and class who were married almost always had children. In other words, raising a family was an assumed aspect of marriage.

Subsequent to her parental leave, plaintiff returned to academia and spent nine years in academia before coming up for tenure. She claims that despite her qualifications and her progress in becoming current in her field, her absence from academia was held against her as an insurmountable barrier.

Fisher, 852 F.Supp. at 1225-26 (footnotes omitted). The court also found in favor of Fisher on her claim of age discrimination in violation of the ADEA, and on her claim of salary discrimination in violation of the Equal Pay Act. The district court rejected plaintiffs claim for compensation based on Vassar’s alleged “retaliation” against her for the filing of this lawsuit.

Following a thorough survey of the record, we reverse the liability judgment against Vassar because our conviction that a mistake has been committed is definite and firm. We further vacate the district court’s order and judgment of attorneys’ fees. We affirm, however, the district court’s conclusion that plaintiff had failed to establish a claim of “simple” sex discrimination. We do not reach Fisher’s cross-appeal concerning the terms of her reappointment to the Vassar faculty.

BACKGROUND

The Candidate’s Career. Cynthia J. Fisher was born in 1932. She was honored in the Westinghouse national science competition as a 17-year-old high school senior, and received a bachelor’s degree in zoology from the University of Wisconsin in 1955. She pursued graduate studies in zoology at Rutgers State University, earning a master’s degree in 1957 and a Ph.D in 1963, and stayed on at Rutgers for two more years doing post-doctoral work.

Plaintiff married Armen G. Fisher on a date not specified in the record; they adopted the first of their two daughters in 1965 and the second in 1968. From 1966 through 1974, Fisher devoted the bulk of her energy to raising her children and did no work outside her home. The district court found that this eight-year break in plaintiffs career was “not a total hiatus”. Fisher, 852 F.Supp. at 1216. In her appellate brief, plaintiff claims that throughout this time she “kept in touch with her field by reading journals and keeping abreast of work in the field of biology.” Brief of Appellee at 3. In 1974 she took a part-time position at Marist College in Poughkeepsie, New York, and returned to teaching.

Vassar hired Fisher as a visiting assistant professor in biology in 1977, at first on a part-time basis and eventually on a full-time basis. She remained at Vassar as an assistant professor and was offered a tenure-track position in 1980. In 1982, plaintiff was reappointed for a three year term with a promise of tenure review in 1985, notwithstanding a mixed departmental evaluation in which two of the then four senior members of the biology department concluded that plaintiff did “not meet the criterion of high quality teaching and barely meets the criterion of research promise”. Joint Exhibits (“JE”) at 197e.

The Tenure Process. Tenure review at Vassar proceeds in several steps. First, the candidate’s credentials are presented to the department. After a review, the department forwards its recommendation to the dean of the college and to the college-wide Faculty Appointments and Salary Committee (the “Appointments Committee”). The dean and the Appointments Committee also receive input on the candidate’s scholarly achievement from outside evaluators. Based on this information, the dean and Appointments Committee make their recommendation to the president of the college who, in turn, submits a recommendation to the Board of Trustees. The candidate has the option of appealing a negative decision to the Faculty Appeals Committee.

Vassar’s review of Fisher’s candidacy took place over the course of the 1984-85 academic year. Fisher submitted a proposed list of outside evaluators in October 1984. On March 29, 1985, the biology department issued a unanimous and unfavorable recommendation. JE at 216-229. The report evaluated plaintiffs qualifications in four categories that the department deemed relevant: scholarship, teaching, service and leadership, and concluded that Fisher failed to exhibit the required “outstanding quality” in any of the four areas. JE at 216. It therefore “recommend[ed] unanimously against promotion.” JE at 229.

The Findings of the Tenure Report. The biology department’s report surveyed plaintiffs numerous publications and grant proposals, but found that plaintiffs record posed

serious questions about her independence as a scholar, about the depth of her mastery of the field in which she is working, and about whether or not she is seriously engaged in an effort that will continue to be productive....

JE at 221.

The report’s analysis of plaintiffs teaching ability included a review of her student evaluations, which (according to the report) reflected “consistent problems with clarity and her ability to illuminate difficult material” but were otherwise generally positive. JE at 223. The report further noted that plaintiff had a “poor” record in supervising independent research. JE at 225. Finally, the report voiced a “serious dissatisfaction [with plaintiffs] failure to introduce a 300-level course of her own design in an area of interest to herself.” JE at 225. The department was “unanimous in the firm belief that Ms. Fisher does not meet the criterion for high quality teaching at Vassar College.” JE at 226.

The report briefly reviewed plaintiff’s service to the department and to the college as a member of the Safety Committee, as an ad-visor to majors in the department, and in helping secure various grants used to purchase additional equipment for the department. The report further noted that plaintiff had served the college-wide community as a member of the Academic Panel and the Library Committee as well as various other committees. Despite these contributions, the report concluded that “[ojverall [the department was] not satisfied by Ms. Fisher’s service to the department and feel that while her service to the college may be adequate it is in no way outstanding ... [and] not of sufficiently high quality for promotion.” JE at 227.

Leadership was the final formal criterion discussed in the report. Here, the department found her to be “a great disappointment.” JE at 227. The report stated:

We find no evidence of leadership in curricular matters either inside or outside of the department, and as best we can judge her interest is just not engaged by the subject. Part of Ms. Fisher’s problem seems to be that she has difficulty in establishing straightforward, open, trusting, collegial relationships with others in the department. Another part of the problem is that she just doesn’t often speak her mind on matters of departmental concern and thus falls short as an intellectually stimulating colleague and contributor to departmental policy-making. Her deferential attitude has been a continuing source of frustration. We also find that we cannot imagine her assuming the responsibilities of the department chairmanship. This is something we believe every tenured member of the department ought to be able to do. Neither do we feel we could nominate her to serve on major faculty committees of the College. None of us would recommend her as director of our introductory biology course.

JE at 227.

As to Fisher’s area of expertise — developmental biology — the Report concluded that her work did not fill a departmental need:

[She] offers to our curriculum little that is unique. Her areas of expertise are already to be found in the tenured ranks of the department. Ms. Fisher’s failure to introduce a[n] advanced level course of her own design exacerbates the situation. In effect she has failed to make a unique curricular contribution and to carve out her own niche as every other member of our faculty has done.

JE at 228. The review of the candidate’s educational and professional history discusses her hiatus from academia:

Ms. Fisher began her scientific career in the usual way and with an acceptable, though not outstanding record of publication. However, beginning in 1965, at the end of her postdoctoral studies, she devoted, by her own report, nine years (1965-1974) entirely to her family and to her personal life. As far as we can tell from materials submitted by her, Ms. Fisher attended no meetings, belonged to no societies, and ceased all communication with colleagues in the scientific community.

JE at 217. 2 The report was signed by the five then-tenured members of the Vassar biology department, three men and two women. Plaintiff was told of the department’s recommendation by department chairman Leathern Mehaffey in a 12-minute meeting on March 29, 1985. In keeping with Vassar policy, the report was confidential and plaintiff did not actually read it until she obtained it through discovery in this lawsuit.

Consideration of the Report. Dean H. Patrick Sullivan met with members of the Appointments Committee on April 19, 1985 to exchange their initial reactions to the biology department report. According to notes taken at that meeting, two of the Committee’s five members expressed an inclination to vote against tenure, one was in favor and two expressed no view. Dean Sullivan voted tentatively against tenure. JE at 290, Transcript (“T”) at 2104-05.

Dean Sullivan and the Committee met with Vassar’s president on April 30, 1985. Again, two Committee members and the dean voted against tenure, one member voted in favor, and two abstained. According to Dean Sullivan’s notes of the meeting, the president made an observation concerning “overkill on dept’s part”. JE at 290. Later, the district court held up this observation as a “virtual smoking gun” evidencing the college’s policy of discrimination. Fisher, 852 F.Supp. at 1229.

On May 3, 1985, the Appointments Committee and Dean Sullivan met with the senior members of the biology department to discuss Fisher’s departmental recommendation, the outside evaluations and student comments. Notes taken at that meeting by one Committee member attributed to the biology department the view that “she is out of date — out of field for 10 years”. JE at 288. This statement too was later cited by the district court as evidence of the college’s discrimination. The evaluations offered by the three outside scholars and considered at this meeting were markedly more positive than the departmental appraisal. For example, a professor at the University of Washington praised plaintiffs scholarly work:

I have never been provided with so much information about a candidate for promotion, nor have I completed a review with as much appreciation for the value of the candidate as a teacher, scientist and scholar, leader and human being as I have gained in reading the information about Dr. Cynthia Fisher. Her enthusiasm, dedication and energy for her profession comes through all of her carefully and articulately written documents.

Letter of Karen A. Holbrook to H. Patrick Sullivan, March 5,1985, JE at 277. See also Letter of Dennis R. Roop to Patrick Sullivan, April 18, 1985, JE at 280; Letter of Paul F. Goetinck to H. Patrick Sullivan, March 13, 1985, JE at 282.

After this meeting, the biology department twice sent letters to Dean Sullivan and other participants in the tenure review process. The letters reaffirmed the department’s negative recommendation, and accused Fisher of distorting her record during the tenure review process. In the letter dated May 6, 1985, the department commented:

We were and continue to be deeply disturbed by the fragmentary information revealed to us during the meeting from documents supplied by Ms. Fisher to you. We hope that we were able to convince you that much of that material was self-serving, distorted, and inaccurate. Now that distortions have been revealed in those documents you discussed with us, we hope you will allow us to assist you in evaluating any other information that you may have before it is accepted for fact and used in your decisions.

JE at 233.

At a final meeting on May 16,1985, all five members of the Appointments Committee voted against tenure, as did Dean Sullivan and the president of the college. 3 Acting upon this recommendation, the Vassar Board of Trustees denied Fisher tenure. Two other candidates in the biology department were up for tenure in the same round of evaluations; one man (Edward Tucker) was denied tenure, and one woman (Pinina Norrod) achieved it.

Fisher appealed to the Faculty Appeals Committee, which in due course rejected her challenge; she then lodged a complaint with Vassar’s affirmative action officer, asserting that she was a victim of discrimination against married women. On July 26, 1985, Fisher filed a claim of discrimination with the New York Division of Human Rights. In this claim, plaintiff charged Vassar “with denying me equal terms, conditions, and privileges of employment and dismissing me because of my age, sex, and marital status, which is a violation of the New York State Human Rights Law.” Supplemental Joint Appendix at 1. Plaintiff left Vassar at the conclusion of her contract in May 1986.

Aftermath and Litigation. Fisher was subsequently hired by Bard College in a halftime position as adjunct faculty. Bard did not renew her contract, leading Fisher to believe that she had been blacklisted by Vassar because she had by then commenced this lawsuit. 4 T. at 951-52. In the fall of 1988, she began study toward a Master’s degree at Adelphi University’s School of Social Work. She graduated in May 1991, passed her State Certification Examination in June of that year, and was soon employed as a social worker at the Poughkeepsie Continuing Treatment Center at a salary of approximately $29,900 a year. During her final year at Vassar, plaintiff had earned $27,000.

Fisher filed her complaint in the Southern District of New York on July 7,1987 and the matter proceeded to trial in June and July of 1993, with the results outlined earlier in this opinion. A June 30, 1994 amendment to the May 16, 1994 opinion spelled out the remedies. 5 In addition to $626,872.12 in damages, 6 plus an award of attorneys’ fees, the district court ordered Fisher

restored to the faculty of the Vassar Biology Department for a period of two years, commencing with the Fall Term of 1994, to the rank of Associate Professor with tenure, at a salary equal to the average salary of an Associate Professor in the Biology Department at Vassar.... At the end of the two year period, Plaintiff shall be evaluated by Defendant for retention and promotion to the Full Professorship track.

Fisher, 852 F.Supp. at 1235. The court further amended its Findings of Fact and Conclusions of Law to address — and reject— broader injunctive relief:

The court finds that the policy of discriminating against married women at Vassar College has changed and, presently, married women are hired and retained in the hard sciences departments. However, this policy change occurred after 1986, the year Plaintiff was denied tenure and after she filed her complaint with the New York State Human Rights Commission. Injunc-tive relief as to the policy under attack here is, therefore, not warranted at this time.

Id. The district court retained jurisdiction over the case in order to ensure compliance with its rulings. Id.

Vassar filed its notice of appeal on July 18, 1994 and Fisher filed her notice of cross-appeal on July 29, 1994. On September 28, 1994, the parties entered into a stipulation fixing attorneys’ fees at a total of $392,205.75. On November 4, 1994, Vassar filed a separate notice of appeal concerning the attorneys’ fees. By stipulation of the parties, the appeal concerning attorneys’ fees was consolidated with Vassar’s main appeal and Fisher’s cross-appeal.

Although Vassar secured a stay of the money judgment and the attorneys’ fee award pending proceedings in this Court, Vassar sought no stay of the portion of the judgment ordering Fisher’s reinstatement as an associate professor. Plaintiff returned to Vassar on September 1, 1994. The college built her a new laboratory and furnished her with an assistant. She spent a portion of the 1994-95 academic year abroad in Grenoble at the University of Joseph Fournier. She is scheduled to return to teaching at Vassar in the fall of 1995.

On appeal, Vassar ascribes clear error to the fact finding on which the district court predicated liability under Title VII, the ADEA and the Equal Pay Act.

DISCUSSION

Page

A. Title VII Claims 1432

1. The McDonnell Douglas test 1432

2. “Sex Plus” Discrimination 1433

(a) Pretextual Decision-Making 1434

(i) Scholarship 1435

(ii) Teaching 1435

(iii) Service 1436

(iv) Leadership 1436

(v) Summary 1437

(b) Discrimination in Fact 1438

(i) Anecdotal Evidence 1438

(ii) Party Admissions 1440

(iii) Statistics 1442

a. Classifying “Hard Sciences” 1444

b. Anecdotes as Statistics 1444

c. School-Wide Tenure Decisions 1445

d. Women Granted Tenure in the “Hard” Sciences 1446

e. Sex to Sex Comparison 1446

f. Marital Status and Parent Status 1447

(iv) Expert Testimony 1447

3. Discrimination Based on Parenthood or Nurture 1448

4. “Simple” Sex Discrimination 1448

B. Age Discrimination Claim 1449

1. Amendment of the Complaint 1449

2. Merits 1449

C. Equal Pay Act Claim 1451

1. Amendment of the Complaint 1451

2. Statute of Limitations 1452

3. Merits 1452

D. Attorneys’ Fees 1453

A. Title VII Claims.

1. The McDonnell Douglas test.

Title VII provides in pertinent part: (1) 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. At the outset, let us make clear what this case is not. Contrary to the district court’s conclusion, see Fisher, 852 F.Supp. at 1229-31, this is not a “mixed-motive” case falling under Price Waterhouse v. Hopkins, 490 U.S. 228, 109 S.Ct. 1775, 104 L.Ed.2d 268 (1989), as modified by the Civil Rights Act of 1991. Had there been direct evidence that Vassar had denied Fisher tenure because she was a married woman, it might have been appropriate to require Vassar to prove that it would have denied Fisher tenure even if it had not considered Fisher’s marital status. See Price Waterhouse, 490 U.S. at 244-45, 109 S.Ct. at 1787-88 (plurality opinion); id. at 259-60, 109 S.Ct. at 1795-96 (White, J., concurring); Ostrowski v. At lantic Mut. Ins. Cos., 968 F.2d 171, 181-83 (2d Cir.1992). But there was no such evidence, as we shall see. Moreover, this is not a disparate impact case falling under Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989), as modified by the Civil Rights Act of 1991, 42 U.S.C. § 2000e-2(k). Fisher did not mount a disparate impact challenge to Vassar’s tenure criteria. See Fisher, 852 F.Supp. at 1226 n. 15.

Thus, Fisher’s only chance of prevailing on her Title VII claim was by using the three-tiered burden-shifting framework enunciated in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). In McDonnell Douglas, the Supreme Court devised a three-tiered burden-shifting framework for applying § 2000e-2. First, the plaintiff is required to present a prima facie case of discrimination which must include allegations (1) that the plaintiff is a member of a protected class, (2) that the plaintiff applied for and was qualified for a job for which the employer was seeking applicants, (3) that despite these qualifications, the employer rejected the plaintiff and (4) that, after this rejection, the job remained open and the employer continued to seek applicants having the plaintiffs qualifications. Id. at 802, 93 S.Ct. at 1824.

If the plaintiff presents a prima facie case, the burden shifts to the employer, who is required to demonstrate “some legitimate, nondiscriminatory reason” for the decision. Id. The employer’s burden here is one of production of evidence rather than one of persuasion. Texas Dep’t of Community Affairs v. Burdine, 450 U.S. 248, 101 S.Ct. 1089, 67 L.Ed.2d 207 (1981). The defendant need only articulate-but need not prove-the existence of a non-discriminatory reason. Id. at 254-56, 101 S.Ct. at 1094-95.

If the defendant carries this burden of production, the plaintiff then assumes the burden to “show that [the employer’s] stated reason for [the plaintiffs] rejection was in fact pretext.” McDonnell Douglas, 411 U.S. at 804, 93 S.Ct. at 1825. “The ultimate burden of persuading the trier of fact that the defendant intentionally discriminated against the plaintiff remains at all times with the plaintiff.” Burdine, 450 U.S. at 253, 101 S.Ct. at 1093. In St. Mary’s Honor Ctr. v. Hicks, - U.S. -, 113 S.Ct. 2742, 125 L.Ed.2d 407 (1993), the Supreme Court held that, as to the third prong of the McDonnell Douglas test, it is not enough for a plaintiff to show that the defendant’s legitimate, nondiscriminatory reason for its employment decision is pretextual; the plaintiff must also prove by a preponderance of the evidence that defendant’s stated reason is “a pretext for discrimination.” St. Mary’s, - U.S. at -, 113 S.Ct. at 2752 (emphasis added). The plaintiff must establish “both that the reason was false, and that discrimination was the real reason.” Id. (emphasis in original).

2. “Sex Plus” Discrimination.

The basis of the discrimination alleged by Fisher under § 2000e-2 is her sex conjoined with her marital status. The Supreme Court has adopted the proposition that sex considered in conjunction with a second characteristic-“sex plus”-can delineate a “protected group” and can therefore serve as the basis for a Title VII suit. See Phillips v. Martin Marietta Corp., 400 U.S. 542, 91 S.Ct. 496, 27 L.Ed.2d 613 (1971). The defendant in Phillips refused to hire women with pre-school-aged children while hiring men who (viewed from the employer’s perspective) had the same problem. In that case, there was no serious question of any general bias against hiring women; the district court had found that 75 to 80 percent of those hired for the position were women. Nevertheless, the Court held that “permitting one hiring policy for women and another for men-each having pre-school-age children”-violated Title VII. Id. at 544, 91 S.Ct. at 498. See also Newport News Shipbldg. & Dry Dock Co. v. EEOC, 462 U.S. 669, 103 S.Ct. 2622, 77 L.Ed.2d 89 (1983) (sex plus pregnancy); United Automobile Workers v. Johnson Controls, 499 U.S. 187, 111 S.Ct. 1196, 113 L.Ed.2d 158 (1991) (sex plus fertility); Price Waterhouse v. Hopkins, 490 U.S. 228, 109 S.Ct. 1775, 104 L.Ed.2d 268 (1989) (sex plus gender stereotypes).

Some courts have specifically accepted the viability of a Title VII claim premised on sex plus marital status. See Bryant v. Interna tional Schs. Servs., Inc., 675 F.2d 562, 573 n. 18 (3d Cir.1982); Sprogis v. United Air Lines, Inc., 444 F.2d 1194, 1197-98 (7th Cir.), cert. denied, 404 U.S. 991, 92 S.Ct. 536, 30 L.Ed.2d 543 (1971); 29 C.F.R. § 1604.4 (discrimination against married women). As a matter of pleading, Fisher satisfied the threshold test of McDonnell Douglas by alleging discrimination against women. In any event, Vassar has not put in issue the sufficiency of plaintiffs prima facie case. However, we note that marital status is not necessarily a bipolar distinction between people who are married and people who are not. The facts of this case entail other categories (e.g., divorced, engaged, seriously involved, never married, etc.) that may become the object of supposed bias; as these splintered categories come into play, marital status may become an unmanageable factor for sex-discrimination analysis.

As to the second prong of the McDonnell Douglas test, Vassar has sustained its burden by relying upon the 14-page recommendation of the biology department. The report offered a “legitimate, nondiscriminatory reason” for dismissing plaintiff-her lack of scholarship, teaching ability, service and leadership. Regardless of Vassar’s record in promoting women, married or otherwise, the school was under no obligation to promote Fisher if she was not qualified for tenure. Because Vassar has satisfied its step-two burden of production, the burden of persuasion shifts back to plaintiff.

The Title VII component of this appeal therefore chiefly concerns the third prong of the McDonnell Douglas test, the plaintiffs burden to demonstrate (a) that the College’s explanation for denial of tenure was false and pretextual and (b) that the real reason for denial was discrimination based on either sex or sex plus marriage. See St. Mary’s, - U.S. at-, 113 S.Ct. at 2752. In reviewing the district court’s decision on these issues, we acknowledge that the district court’s findings of fact should be disturbed only if they are clearly erroneous within the meaning of Fed.R.Civ.P. 52(a). See Anderson v. City of Bessemer City, 470 U.S. 564, 573, 576, 105 S.Ct. 1504, 1512-13, 84 L.Ed.2d 518 (1985); Sumner v. U.S. Postal Serv., 899 F.2d 203, 208 (2d Cir.1990). We will not “upset a district court’s determination simply because we would have reached a different conclusion had we considered the matter in the first instance.” United States v. Juvenile Male # 1, 47 F.3d 68, 71 (2d Cir.1995). Rather, we will “disturb such findings only when we are left with the definite and firm conviction that a mistake has been committed.” Cosgrove v. Sears, Roebuck & Co., 9 F.3d 1033, 1038 (2d Cir.1993) (citations and internal quotations omitted). On at least some of the material issues addressed by the district court, we are left that definite and firm conviction.

(a) Pretextual Decision-Making.

After a thoroughgoing analysis of the Biology Department’s 14-page evaluation of Fisher, the district court concluded that “the termination of plaintiffs employment resulted not from any inadequacy of her performance or qualifications or service, but from the pretextual and bad faith evaluation by Vassar of her qualifications.... ” Fisher, 852 F.Supp. at 1218. On appeal, Vassar contends that the district court paid insufficient deference to the academic standards and values employed in the process that led to the denial of tenure in this case. Vassar cites Lieberman v. Gant, 630 F.2d 60 (2d Cir.1980), for the proposition that district courts should refrain from “infer[ring] discrimination from a comparison among candidates .... ” Id. at 67. The Lieberman court voiced wise caution that the nation’s universities and the federal courts might be heavily burdened if “the many adverse tenure decisions against women or members of a minority group that must be made each year [were] regularly taken to court____” Id. at 62 n. 1. The court noted:

A university’s prerogative to determine for itself on academic grounds who may teach is an important part of our long tradition of academic freedom. Although academic freedom does not include the freedom to discriminate, this important freedom cannot be disregarded in determining the proper role of courts called upon to try allegations of discrimination by universities in teaching appointments. The Congress that brought educational institutions within the purview of Title VII could not have contemplated that the courts would sit as “Super-Tenure Review Committee[s].”

Id. at 67 (citations and internal quotations omitted, alteration in original).

Nevertheless, as the district court observed, Title VII applies to the academic community. As originally enacted, Title VII did exempt from scrutiny the faculty employment practices of educational institutions, but the Equal Opportunity Act of 1972 eliminated that exemption specifically. See Powell v. Syracuse Univ., 580 F.2d 1150, 1154 (2d Cir.), cert. denied, 439 U.S. 984, 99 S.Ct. 576, 58 L.Ed.2d 656 (1978). Congress has, therefore, given federal courts the unsought task of examining the sometimes surprising ways in which colleges and universities award tenure. See Zahorik v. Cornell Univ., 729 F.2d 85, 93 (2d Cir.1984) (“Tenure decisions are not exempt under Title VII ... ”).

According to Vassar, Lieberman stands for the proposition that comparative proof, measuring successful candidates against those rejected for tenure, is inadmissible in a Title VII case. That view was specifically rejected by this court in Gibson v. American Broadcasting Cos., 892 F.2d 1128, 1133 (2d Cir.1989): “In light of the strong remedial purposes expressed in the Civil Rights Act of 1964 and the 1972 Amendments to Title VII ... we could hardly hold that relevant proof was inadmissible.” Therefore, it was appropriate for the district court to consider plaintiffs record and to compare it with the records of other Vassar professors.

The district court found that Vassar’s tenure review of Fisher was conducted using standards that were unclear and unspecified. Fisher, 852 F.Supp. at 1227. However, since it is difficult to conceive of tenure standards that would be objective and quantifiable, the district court properly undertook to examine the enumerated criteria used by Vassar, however intuitive they may be, in order to detect their use as pretext. At the end of that examination, the district court found that the department’s evaluation of Fisher was disingenuous in each category:

(i) Scholarship.

The district court concluded: “Based on the foregoing testimony and evidence, the court finds that [the biology department’s] conclusions concerning Dr. Fisher’s scholarship were made in bad faith, were pretextual, and represented the application of patently discriminatory standards.” Fisher, 852 F.Supp. at 1209. As an example of the department’s bad faith, the district court cited the department’s criticism of plaintiffs productivity during the sabbatical year she took before coming up for tenure review. The department claimed that she had spent little time in the laboratory, thereby raising concerns “about the choices she made in respect to how to spend her leave year.” JE at 221. In fact, as the district court points out, Fisher spent nine months out of her sabbatical year in a laboratory during which time she “collaborated with four different groups, submitted eight grant proposals, six of which ... were funded, published one manuscript, wrote one other, served as a consultant to both the [National Science Foundation] and the [National Institute of Health] and presented papers at national and international meetings.” Fisher, 852 F.Supp. at 1205. 7

(ii) Teaching.

The district court found that the biology department had distorted Fisher’s teaching recommendations by “selectively excluding] favorable ratings”, by “focus[ing] on the two courses in which Dr. Fisher had difficulties” and by “applying different standards to her than were applied to other tenure candidates”. Id. at 1209. Further, the district court observed that “the males tenured while Dr. Fisher was on the faculty were praised for their fine teaching while Dr. Fisher was criticized, although the facts on which the Committee’s determinations were based (student evaluations, Biology Majors Reports and [Student Advisory Committee] reports) revealed that Dr. Fisher’s evaluations were superior to theirs.” Id. at 1211.

(iii) Service.

The district court did not undertake to compare Fisher’s service in the academic community to the service rendered by others similarly situated, nor was the district court required to conduct a comparison of things so amorphous. Instead, after examining the Vassar Governance (the College’s basic policy document), the district court found that service was not a significant factor in the tenure review process, Fisher, 852 F.Supp. at 1213, a finding that we accept. However, the district court’s finding of pretext on this issue is unsubstantiated and conclusory:

It is additionally significant that early evaluations by the Department of Dr. Fisher’s service to the College and the Department were glowing____ The only explanation for the change in attitude towards Dr. Fisher that the court can find is the discrimination manifested in its decisions concerning her culminating in its decision in 1985 not to promote her.

Id. at 1213.

(iv) Leadership.

The Biology Department found plaintiffs leadership skills to be “a great disappointment”. JE at 227. The district court reviewed the National Science Foundation’s favorable reviews of the plaintiff as well as two previous evaluations from the biology department, and concluded that “Dr. Fisher’s collegiality, personal relationships, and service to the College and to its students were more than satisfactory and that the opinions given contrary to this finding by the Biology Department were pretextual.” Fisher, 852 F.Supp. at 1214. This finding is unsustainable. The biology department’s previous reviews of Fisher had been far from uniformly glowing. The department’s split 1982 recommendation concerning Fisher’s reappointment stated:

[S]he seldom takes an active part in discussion of issues in departmental meetings. Some of us believe that she is tentative about taking a position on issues and prefers not to commit herself openly. This represents a manifestation of her personality which may be viewed as a danger signal in respect to how she fulfills her collegial responsibilities.

JE at 197d. Notwithstanding her previous favorable reviews, by the time she came before the department for tenure consideration in 1985 the senior members of the biology department had concluded that she lacked the requisite leadership qualities:

Part of Ms. Fisher’s problem seems to be that she has difficulty in establishing straightforward, open, trusting, collegial relationships with others in the department. Another part of the problem is that she just doesn’t often speak her mind on matters of departmental concern and thus falls short as an intellectually stimulating colleague and contributor to departmental policy-making. Her deferential attitude has been a continuing source of frustration.

JE at 227. Certainly her colleagues were in the best position to judge her “collegiality [and] personal relationships”. The leadership section of the report makes clear that the senior members of the biology department simply did not like Fisher and did not wish to establish a career-long professional association with her. It is arguable that such grounds alone justified the department’s recommendation against tenure. 8

Nevertheless, neither “collegiality” nor any essentially social characteristic is listed in the Vassar Governance as a controlling aspect of tenure review. Therefore, even if the district court’s analysis of plaintiffs leadership qualifications was clearly erroneous, the court’s overall conclusions regarding the overall assessment of the department report still survives appellate review.

(v) Summary.

The district court found that the biology department tenure report on the plaintiff was pretextual. We conclude that this finding is based on sufficient evidence presented to the court. Certainly, the portion of the district court’s opinion that surveys the department’s report does not provoke “the definite and firm conviction that a mistake has been committed.” See Cosgrove, 9 F.3d at 1038 (internal quotations and citations omitted).

We recognize that in some situations [t]he factfinder’s disbelief of the reasons put forward by the defendant (particularly if disbelief is accompanied by a suspicion of mendacity) may, together with the elements of the prima facie ease, suffice to show intentional discrimination. Thus, rejection of the defendant’s proffered reasons[ ] will permit the trier of fact to infer the ultimate fact of intentional discrimination ....

St Mary’s, - U.S. at -, 113 S.Ct. at 2749 (emphasis in original). See also EEOC v. Ethan Allen, Inc., 44 F.3d 116, 120 (2d Cir.1994); DeMarco v. Holy Cross High Sch., 4 F.3d 166, 170 (2d Cir.1993). Nevertheless, while this passage makes clear that “some cases exist where a prima facie case and the disbelief of a pretext could provide a strong enough inference of actual discrimination to permit the fact-finder to find for the plaintiff[,].... we do not think that the Supreme Court meant to say that such a finding would always be permissible.” Woods v. Friction Materials, Inc., 30 F.3d 255, 261 n. 3 (1st Cir.1994) (emphasis in original); see also United States v. Redondo-Lemos, 27 F.3d 439, 442 (9th Cir.1994) (Kozinski, J.); cf. Woroski v. Nashua Corp., 31 F.3d 105, 109-10 (2d Cir.1994).

On these facts, the district court’s finding that Vassar’s “use of the tenure standards and process” was pretextual, Fisher, 852 F.Supp. at 1228, could not itself support a judgment for Fisher. It bears emphasis that the district court did not find Vassar hable simply because it disbelieved Vassar’s proffered reasons for denying Fisher tenure. See id. (“Not only has plaintiff proven by a preponderance of the credible evidence that defendant’s legitimate non-discriminatory reason ... is pretextual, plaintiff has also proven that defendant’s reason is a pretext for discrimination.”) (emphasis added). Rather, the district court felt compelled to comb the record for evidence of discrimination, evidently convinced, as are we, that the finding of pretext here did not alone justify a finding of discrimination. See id. at 1228-29 (Vassar’s unfair application of its tenure standards, together with statistical and anecdotal evidence, adduced to prove that Vassar had intentionally discriminated against Fisher). After all, “the ultimate burden of proof remainfed] at all times with [Fisher] to demonstrate that illegal discrimination actually motivated” Vassar’s decision to deny her tenure. Yellow Freight Sys., Inc. v. Reich, 38 F.3d 76, 85 (2d Cir.1994); see Woroski v. Nashua Corp., 31 F.3d 105, 108 (2d Cir.1994) (once employer articulates a non-discriminatory reason, “the plaintiff has the burden of proving that his age was the real reason for the discharge.”). And an employer may offer a pretextual reason for a personnel decision that is nonetheless nondiscriminatory, and the factfinder may review the record to determine the real reason for the personnel decision. See St. Mary’s, - U.S. at -, 113 S.Ct. at 2755-56.

Doubtless, there are cases in which discriminatory intent is the only probable reason for the employer’s proffer of a pretextual reason to the court. See, e.g., Binder v. Long Island Lighting Co., 57 F.3d 193, 200 (2d Cir.1995). Not so here. The finding of pretext affirmed in this opinion points nowhere, as the district court’s own opinion suggests. See Fisher, 852 F.Supp. at 1228-29. Therefore, our ruling on pretext does not require as a corollary that we affirm the ultimate finding of discrimination. See St. Mary’s, - U.S. at-, 113 S.Ct. at 2754 (“It is not enough ... to disbelieve the employer; the factfinder must believe the plaintiffs explanation of intentional discrimination.”); Saulpaugh v. Monroe Community Hosp., 4 F.3d 134, 142 (2d Cir.1993) (“[A] Title VII plaintiff does not necessarily meet [her] burden of persuasion by convincing the factfinder that the employer’s non-discriminatory explanation is not credible; rather the trier of fact must find that the plaintiff has proven [her] explanation of discriminatory intent by a fair preponderance of the evidence.”), cer t. denied, — U.S. -, 114 S.Ct. 1189, 127 L.Ed.2d 539 (1994). Rather, we must review the record to see if the district court’s ultimate finding of discrimination was clearly erroneous. See St. Mary’s, — U.S. at -, 113 S.Ct. at 2756 (“That the employer’s proffered reason is unpersuasive, or even obviously contrived, does not necessarily establish that the plaintiffs proffered reason of race is correct. That remains a question for the factfinder to answer,” subject to the clearly erroneous standard of appellate review.); id. at -, 113 S.Ct. at 2749 n. 4 (“Even though ... rejection of the defendant’s proffered reasons is enough at law to sustain a finding of discrimination, there must be a finding of discrimination.”) (emphasis added).

(b) Discrimination in Fact.

In arriving at the conclusion that Vassar discriminated against married women, the district court relied on four categories of evidence: anecdotes; perceived admissions made by Vassar officials participating in plaintiffs tenure review; statistics; and expert testimony. We review below the evidence cited by the district court in its opinion as well as other evidence offered by plaintiff at trial.

(i) Anecdotal Evidence.

The district court relied upon the experiences of Karen Friedman and Marlene Palmer, two unhappy former members of the Vassar biology department. Friedman and Palmer joined Vassar in 1977 along with plaintiff and had spouses during some part of their employment at Vassar. Neither was ever considered for tenure.

Friedman testified that she was married when she first came to Vassar and that the senior members of the biology department ignored her. She divorced while at Vassar and noticed that she was subsequently “included in social invitations” by various senior members of the biology department. T. at 1255. Several months after her divorce, Friedman began to date again. She soon “expressed some frustration at the limited pool of available men at Vassar College and expressed a desire to find [a] lasting relationship, and in fact to remarry at some point----” T. at 1256. The invitations to social functions from the senior members of the biology department then “declined and stopped entirely.” T. at 1256.

After Friedman was denied reappointment, she met with Ed Tokay, the acting chairman of the biology department. According to Friedman’s testimony at trial, Tokay told her that she had been denied reappointment, in part, because “there is an unspoken agenda” — which he did not define — at biology department deliberations. T. at 1345. Finally, Friedman testified that, before being considered for reappointment, she was asked to take a leave of absence without pay in order to engage in further research. She stated that her male colleagues were not asked to take similar leaves. Despite her compliance with this request (which ultimately amounted to two years’ absence), she was denied reappointment.

According to the district court, “[e]videnee adduced from Dr. Karen Friedman indicated that there was a policy of discouraging married women who were hired in the sciences as assistant professors from advancing to tenure.” Fisher, 852 F.Supp. at 1215. This conclusion is unsupportable. Nothing in the testimony of Friedman indicates that Vassar discriminated against married women. It is not remarkable that members of the academic community took special pains to offer her social invitations in the months after her divorce, or that these efforts trailed off when she began to date and became otherwise socially occupied. Friedman’s conversation with Tokay, while cryptic, cannot be said to evidence a policy of the college or the department to discriminate on the basis of marital status, or indeed to discriminate on any basis. The “unspoken agenda” remains unknown. Finally, testimony offered by Friedman on cross-examination suggests that her leave was voluntary and undertaken in order to allow her to engage in further research before contract renewal. T. at 1349-50. Ultimately, the biology department recommended against Friedman’s reappointment because the department found her research record to be weak. At the time she was denied reappointment, she was single.

The second anecdote relied upon by the district court concerned Dr. Marlene Palmer. As stated above, Palmer joined the Vassar biology department with the plaintiff in 1977. She left after failing to secure reappointment in 1983. Palmer did not testify at the trial; her account came into evidence by way of a letter written by Palmer to the Vassar Faculty Appeals Committee in November 1985 in support of plaintiffs appeal of her denial of tenure. The letter, which was read into the record during trial, stated in part:

S) Discrimination against Married Women:

Just prior to the 1979-80 school year, I married a man with 8 children, and attitudes toward me in the department (especially that of [Professors] Pat Johnson and Marci Greenwood) took a severe turn for the worse. This marriage made no difference in my work; nevertheless, the line of reasoning seemed to be that married men are stable and responsible but married women (especially those with children) will not give their all to the job. In fact, [Professor] Anita Zorzoli remarked to someone at my wedding that she predicted I would now retire to be a housewife and mother and give up science entirely. Aside from this comment, I can recall nothing specific that was said to me. However, my sense of being discriminated against on this basis was so strong that, when I discovered that I was pregnant in 1981, I voluntarily and secretly had an abortion. I knew they would terminate me at the next contract renewal if I had the baby. I thought I might have a chance if I did not have the baby.

Letter from Marlene (Palmer) Hutt to Janet Knapp, November 20, 1985, at 2, JE at 555. This letter was never offered to prove the matters asserted; it was admitted into evidence by stipulation to show that the college received notice of these accusations and failed nevertheless to conduct a formal investigation. The district court, however, cited this letter — and specifically noted the story of the compelled abortion — to support findings of discrimination. Fisher, 852 F.Supp. at 1215. Admissibility aside, Palmer’s vaporous account does not support the weight placed on it by the district court. Palmer’s “sense of being discriminated against” is not evidence. The fact that this unsubstantiated “sense” alone caused her to have a (apparently no longer) “secret” abortion is sad but hardly significant.

The statements offered by Palmer and Friedman thus do not amount to evidence of discrimination based on marital status. In her brief on appeal, plaintiff cites three other examples of “direct” evidence of discrimination involving herself and various single women who either were or at one time had been professors in the Vassar biology department. None of the women, however, played any official role in plaintiffs tenure review. First, plaintiff testified to her conversation in a parking lot in 1979 with Professor Sue Lumb:

I said, Sue, do I have to be the security blanket around here forever?

And she said, oh, well, if we didn’t promote those nice young men, they wouldn’t stay.

T. at 55. Plaintiff defined a “security blanket” for the benefit of the court:

A security blanket is a person who, in the academic world, is a person who hangs around when needed. If they need someone to teach a course part time, to be— teach one semester and then another, they’re always there to fill in. And it’s a term used particularly when talking about sex discrimination where women who are geographically trapped in an area who cannot move to get a better job are often used as, quote, security blankets, unquote, to fill in as needed. But not to be given rank and position or permanency.

T. at 56-57.

The second incident occurred while plaintiff was lunching with professor Anita Zorzo-li, who was at the time on an unpaid leave with total disability:

I again said I was a little tired of being the security blanket around here. And ■without a second’s hesitation she said, but, my dear, you have your children.

And I said, what relevance does that have to my performance as a biologist?

Well, she said, there was a 10-year hiatus in your career. Actually, it was only eight, but she made it 10.

I said, am I punished for having children and taking care of my family?

And she said, well, my dear, there is always part-time work for you.

And I said, Anita, how would you feel if 30 years ago you were given a part-time job with your credentials?

And again, without a second’s hesitation she reported to me, she said, well, you and I are not in the same class. I did all this research when nobody did research at Vassar, etc. etc.

T. at 66-67. The final incident involved retired biology professor Margaret Wright:

I asked her, what do you have to do to belong around here, Margaret?

And she said, well, I didn’t think you’d really want a full-time position because you have your family.

To which I responded, what do you think I’m working so hard for, a part-time position? And that was the end of that conversation.

T. at 69.

It is remarkable that, in each of these three incidents, the (arguably) discriminatory characterizations were made by plaintiff. She described herself as a security blanket. She suggested she was being punished for having children. She suggested that she did not “belong.” None of the incidents add up to evidence that Vassar had a policy of discriminating against married women. Even under plaintiffs own definition of “security blanket,” that term does not apply solely to married women. Although the comments attributed to Zorzoli are patronizing and tinged with stereotypes, Zorzoli by that time had already begun her disability leave and played no role in the department’s decision making process. Finally, although Wright reportedly made an impertinent assumption, she did not indicate that Wright (let alone Vassar) would disqualify plaintiff from a full-time position simply because she had a family. 9

The district court opinion, which nowhere relies on these three conversations, properly discounts this type of evidence in the context of the ADEA claim. In commenting on another statement made by Professor Lumb the court noted:

Dr. Lumb’s statement ... is not attributable to the Tenure Committee’s decision making process since she was not a member. This statement could be attributed to the sentiment of the Department but on its own does not meet plaintiffs burden.

Fisher, 852 F.Supp. at 1231 n. 25. Neither professors Zorzoli nor Wright were at all involved in plaintiffs tenure review.

In sum, none of the anecdotal evidence offered by plaintiff supports a finding that Vassar had a general policy of discriminating against married women or that Vassar discriminated against plaintiff on the basis of her sex conjoined with her marital status.

(ii) Party Admissions.

The record shows no link between plaintiffs marital status and the decision to deny her tenure. The district court pointed generally to a “persistent fixation of the Biology Department’s senior faculty on a married woman’s pre-Vassar family choices____” Fisher, 852 F.Supp. at 1216. The court cites a handful of invalid examples of what it deems to be this “persistent fixation”.

One example is drawn from the text of the biology department’s report:

[Beginning in 1965, at the end of her postdoctoral studies, she devoted, by her own report, nine years (1965-1974) entirely to her family and to her personal life. As far as we can tell from materials submitted by her, Ms. Fisher attended no meetings, belonged to no societies, and ceased all communication with colleagues in the scientific community.

JE at 217. The district court also cites notes taken during the May 3 meeting of the Ap pointments Committee. According to these notes, the department represented to the Committee that Fisher was “out of date — out of field for 10 years.” JE at 288. The district court found that these notations evidence the department’s

acceptance of a stereotype and bias: that a married woman with an active and ongoing family life cannot be a productive scientist and, therefore, is not one despite much evidence to the contrary.

Fisher, 852 F.Supp. at 1216. The disputed passage of the department report appears in an account of plaintiffs pre-Vassar career. It overstates by one year her withdrawal from academic life, but some reference to this period is to be expected if the readers of the report were to understand why her teaching career started years after the end of her post-doctoral training. The district court read into this tenure-relevant statement an unstated assumption that family life is incompatible with scientific pursuits. This jaundiced reading places a strain on the text, and erroneously assumes that family commitments naturally compel complete withdrawal from the scientific community.

Similarly, the note made at the May 8 meeting of the Appointments Committee— that she had been “out of field for 10 years” — overstates the length of plaintiffs hiatus, but is no more than an acknowledgment of an uneontested fact: plaintiff did spend a long period of time away from academia and biology. Indeed, the district court opinion later recasts its initial interpretation of the “out of field” statement; a footnote offers the following non-discriminatory explanation for this statement: “The court interprets the Department’s statement to the [Appointments Committee] ... as an attempt to convey that she was not up to speed in educational requirements and/or research techniques____” Fisher, 852 F.Supp. at 1231 n. 25. We accept this interpretation, and conclude that the notation does not contribute to an inference of bias.

The district court further relied upon the trial testimony of professor Leath-ern Mehaffey:

Q: Did you join in the thought that her being ten years away from active involvement in biological research represented a huge gap in her knowledge that was presently relevant to her being on tenure track?

A: Yes, I think I did.

May I explain a little bit?

Q: Certainly.

A: Largely because the field moves very quickly. Any field in biology moves very quickly. And ten years of not being actively involved in the field means that you have an awful lot of catching up to do, and that’s what we were worried about here, just that there’s an enormous amount of work to catch up.

T. at 1906. Based on these comments, the district court stated:

Dr. Mehaffey admitted at trial that the fact that the time plaintiff had taken off in order to raise a family was a principal factor in the department’s recommendation to deny her tenure. He felt that this represented a huge gap in her knowledge which she was unable to rectify. This statement was in spite of the fact that she was actively publishing and researching after her hiatus. This is also direct evidence that the Biology Department, perhaps still deeply rooted in the tradition of the hard sciences, was unable to overcome the stereotypical view of women as either scientist or wife — but not both.

Fisher, 852 F.Supp. at 1230 (citations omitted). But Professor Mehaffey never stated that “the time plaintiff had taken off in order to raise a family was a principal factor in the department’s recommendation to deny her tenure.” He merely agreed with the proposition that “her being ten years away from active involvement in biological research represented a huge gap in her knowledge that was presently relevant to her being on tenure track.” A “relevant” factor is not the same thing as a “principal factor.” Furthermore, nothing in Dr. Mehaffey’s statements implies that the “factor” at issue was marital status. Plaintiff claims discrimination by reason of her status as a married woman, not her status as a person who has taken time off from academia. It was clear error for the district court to consider Mehaffey’s statement as evidence of discrimination.

Finally, the district court pointed to what it termed “evidence of a virtual smoking gun.” Fisher, 852 F.Supp. at 1229. This gun and its smoke consist of a notation by Dean Sullivan on a copy of plaintiffs Faculty Record Card. The notation attributes to the president of Vassar the following comment: “overkill on dept’s part.” JE at 290. At trial, the dean stated that his notes were based on his “impression” of what the meeting participants had said and were not “verbatim”. T. at 2108. He offered two possible interpretations of what the “overkill” notation meant to him:

The first possibility would be that what I was recording here was my sense that the president believed that the department’s recommendation was harshly critical, in the light of outside evaluations which were on the whole positive, and student responses, which were on the whole positive. The other possibility ... was that overkill on the department’s part meant that the president felt that the department in what I believe is a 14 page memorandum single spaced, had strained to say the obvious, that the case upon its face lacked merit for promotion to tenure and that the department had simply overkilled, it had gone to extraordinary lengths to state [the obvious],

T. at 2108-09.

The district court interpreted these notes as “direct evidence that the President believed that the Department’s recommendation to deny Dr. Fisher’s application for tenure was not made without bias.” Fisher, 852 F.Supp. at 1229. This interpretation is clearly erroneous, because it places too much weight on the word “overkill.” The “overkill” notation can be construed to reflect the president’s view that the department’s report was vehement, but nothing in the statement or its context implies that the department’s position arose from impermissible bias as opposed to, for example, strong opposition on principled grounds. In the absence of actual evidence of such an impermissible bias, the district court erred in assuming that one existed.

(iii) Statistics.

Statistics may be a part of a plaintiffs effort to establish discrimination under a theory of disparate treatment — the theory pursued by plaintiff before the district court. See Anderson v. Douglas & Lomason Co., 26 F.3d 1277, 1285 (5th Cir.1994), cert. denied, — U.S. -, 115 S.Ct. 1099, 130 L.Ed.2d 1066 (1995). Fisher’s effort to establish that Vassar discriminated against married women leans heavily on statistics. For a number of reasons, the district court’s reliance on this data was clearly erroneous.

Plaintiffs statistical analysis was reproduced in the appendix on appeal. 10 Plaintiffs summary chart lists the names of 59 female professors who were employed by Vassar at or above the rank of visiting assistant professor for two years or more in the departments of biology, chemistry, mathematics, physics, geology and psychology at some point during the years 1956 through 1985. JE at 484. Of the 59 people, 19 already had tenure in 1956. The chart divides the remaining 40 professors into two periods: 'those who achieved tenure or left in the years 1956 through 1971; and those who achieved tenure or left in the years 1972 through 1985. For each time period the names are divided into two marital categories: single and non-single. Within each category, the chart indicates whether each professor was either “Promoted” (which means granted tenure) or “Terminated or left” (which includes professors who left Vassar for any reason whatsoever).

On the 1956-1971 chart, eleven names are categorized as being single. Six of these are listed as having been “Promoted” while five are listed as “Terminated or left”. Four women are categorized as non-single, and all of these are placed under the heading “Terminated or left.” On the 1972-1985 chart, 14 names are categorized as being single. Eight of these are listed as “Promoted” while six were listed as “Terminated or left”. Eleven women are categorized as non-single, and all but one of these are listed as “Terminated or left”. Plaintiffs own name appears there. Thus, of a total of 15 non-single women listed during the 30 year period, only one (a psychology professor) was granted tenure. At trial, plaintiff argued (and the district court agreed) that this woman as well as the other psychology professors should be excluded from the list because psychology is not one of the “hard” sciences. In sum, plaintiffs statistics suggested that, of nine non-single women employed as professors at Vassar in biology, chemistry, math (which later included computer science), physics and geology between 1972 and 1985, none were awarded tenure.

Based on its reading of this statistical analysis the district court found as follows:

In the 30 years prior to Dr. Fisher’s tenure review, no married woman ever achieved tenure in the hard sciences (defined as Mathematics, Physics, Chemistry, Geology, Biology and Computer Science). This information was based on the Vassar College Directory of Administration/StaffiFaculty which is published annually by Vassar College and lists each person on the faculty with that person’s rank, and in parentheses lists that person’s spouse, if any. Such entries for the directory are voluntarily obtained from the professor himself or herself on a form solicited annually by the College Publications Office. Also consulted were faculty biography cards in the Vassar library where there are brief personal biographical files on each faculty member. Between 1956 and 1985, out of 18 single women in the hard sciences department who received at least two year contracts at the Assistant Professor rank, ten were tenured and eight were terminated or left. From 1972 on, six out of nine were tenured. From 1956 to 1985, among nine women who were apparently married (by listing a male parenthetical name in the College Directory, or otherwise checked out as married from the biographical files of the College) all nine were terminated or left. Of the eight women on the Biology faculty for a period of two years in the rank of assistant professor from 1972 to 1985, three single women were tenured and the five married women left the faculty.

[T]he court finds that never, in at least the 30 year period from 1956 to 1986 has Vassar advanced, through the four steps to tenure as stated in the Governance, any woman in the six “hard” sciences (Biology, Chemistry, Geology, Mathematics, Physics and Computer Science) whose marital status was at any time married between hiring as assistant professor and the tenure review, and that all women tenured in the hard science departments during this period were single at all times from appointment as assistant professor up to the grant of tenure, including all eight tenured from 1972 on, and including the three tenured in Biology. Moreover, the appropriate and available pool of applicants for positions in the Science Departments and particularly the Biology Department contained a large percentage of married women who would be satisfactory faculty if employed. Accordingly, this record shows an affirmative intent to deny tenure to married women in the “hard” sciences.

Fisher, 852 F.Supp. at 1215, 1218 (citations omitted).

We conclude that reliance on these statistics to support a finding of discrimination is clear error. For the most part the statistics do show that a number of departments at Vassar employ few female professors who are married; but close analysis discloses no evidence of discrimination based on sex (or sex plus something else). Plaintiffs statistical case is built on gerrymandered data and a series of statistical fallacies. A spurious correlation between the marital status of women and their promotion in the “hard sciences” at Vassar is created (a) by purposefully defining which sciences are hard and which are not; (b) by treating anecdotes as statistics; (e) by disregarding the pattern of school-wide tenure decisions; (d) by overlooking women who did achieve tenure in hard sciences; (e) by comparing married women to single women rather than by comparing women of each status with men simi larly situated; and (f) by conflating marital status with status as a parent.

a. Classifying “Hard Sciences”. In 1983, the psychology department awarded tenure to Professor Gwen J. Broude, a married woman. This fact should have been reflected in the district court’s analysis of plaintiffs statistics. At the time of plaintiffs tenure review, the Division III classification of Vassar’s administrative structure was composed of the following departments: mathematics, physics, chemistry, geology, biology and computer science and psychology. Yet the district court (as plaintiff requested) excluded consideration of the Psychology Department from its statistical review. The district court acknowledged this discrepancy, but gave no reason for it. Dean Sullivan, who served from 1978 through 1988, testified at trial that psychology was a hard science:

Q: And would you say that the psychology department at Vassar College is part of the hard sciences?

A: Yes, I would.

Q: Why do you say that?

A: Why do I say that? Psychology at Vassar, as well as psychology throughout the nation is an academic field, as opposed to as a clinical field, as an academic field is based on the same experimental methodology that biology and chemistry, physics are based on.

It especially holds close paradigms and methods with biology. As there is great variety in the biological discipline, there’s also a great variety in psychology. But our majors are required to take laboratory courses in experimental methodology, they are required to take courses in statistics, they are required to explore the primary scientific evidence, they are required to do experimental and other sorts of studies as part of their undergraduate education.

It is very much an experimental science. Many of our subjects are people, but nevertheless, an experimental science.

Q: And on that basis, how would you classify psychology as part of hard sciences?

A: Well, if one needs to use hard or soft,

I would say hard.

T. at 2217-18. Neither at trial nor in her brief has plaintiff refuted Dean Sullivan’s testimony that psychology employs the same scientific method used in other hard science disciplines and is therefore a hard science. We can find no principled reason for rejecting Vassar’s institutional grouping of psychology among the sciences characterized as “hard”.

b. Anecdotes as Statistics. Plaintiffs “statistics” contain numerous errors in the categorization of the marital status of the various professors. At trial, plaintiff did not even pretend that her analysis of tenure decisions compared the experience of married women as compared with that of single women. Rather, according to plaintiffs counsel, the statistical chart differentiated on the basis of whether or not an individual woman was “perceived” as married:

Now, if a woman wants to be married and teach at Vassar College and conceal the fact or not make a public demonstration of it, it may be that somebody slipped through our statistics. But since discrimination is based on perception to a large degree, it’s our contention that no woman perceived to be married for 30 years has even been tenured at Vassar in the hard sciences.

Now, maybe there was someone who secretly had a husband at home or maybe there was a married woman. Our statistics do not go into that at all. Our statistics are only the perception.

T. at 2227. This reliance on perception masks the unreliability of the sources used to compile plaintiffs numbers and her unwillingness to consult rehable sources of data. Moreover, the perception of marital status does not square with plaintiffs legal argument. At trial and on appeal, plaintiff has argued that she was discriminated against on the basis of sex plus marital status, not on the basis of sex plus perceived marital status. We doubt that such a claim would be cognizable under Title VII.

In compiling her statistics, plaintiff relied in large part on the Vassar College directory. A professor who listed a spouse in the directory was counted as married. Testimony at trial, however, indicated that the directory was “notoriously unreliable.” T. at 2334. Had plaintiff sought accurate information on marital status, she would have consulted the College Benefits Office. T. at 2293-94.

Plaintiffs methodology invited error. For example, Susan Cormier, a biology professor who was not granted tenure, is listed as married by plaintiff yet Cormier was never legally married. T. at 2283. Karen Kato Friedman, listed as married by plaintiff, was in fact divorced when she left Vassar. T. at 2282-2283. The Vassar directory never listed her as having a spouse. Shuler Richardson, married at the time of her tenure review, was single when she joined the Vassar faculty. T. at 2283. Pinina Norrod, who was granted tenure in the biology department the same year plaintiff was denied tenure, was married at the time she was hired by Vassar. T. at 2283. By the time she was considered for tenure, she had entered a long-term relationship with a man she eventually married. Patricia Johnson, one of the tenured members of the biology department at the time of plaintiffs tenure review, was a divorced woman with two children. T. at 2283. M.R.C. Greenwood, another tenured member of the biology department at the time it recommended against granting tenure to plaintiff, had been married and was divorced shortly before or immediately after she joined the Vassar faculty. T. at 1828.

Some of the women listed by plaintiff as having been “Terminated or left”, departed the Vassar campus for reasons completely unrelated to denial of promotion or tenure. For example, Christine Fahlund, a biology professor, apparently left Vassar in 1976 to relocate with her husband, who had taken a new job in another area of the country. T. at 2279. She was never denied reappointment. Susan West, listed among the non-single women who were terminated or left, was never employed as a professor and was not a scientist. She worked at Vassar for ten years as a nursery school teacher and was listed as an assistant professor for purely administrative reasons. T. at 2262, 2267.

c. School-Wide Tenure Decisions. At trial, Vassar sought to introduce the testimony of Professor Orley Ashenfelter concerning tenure granted to married women on a school-wide basis. T. at 2067. According to the offer of proof, this testimony would have shown that nearly one out of six married women professors hired by Vassar between 1977 and 1984 eventually achieved tenure by the end of the 1983-84 academic year. JE at 743. Of the 23 married women faculty members considered for tenure between 1972 and 1984, 18 succeeded. JE at 690-91. This strongly suggests that, college-wide, there was no discrimination against married women in the tenure review process. JE at 690-91. The district court, however, excluded this testimony on the ground “that this case involves a claim that women in the hard sciences as defined by the plaintiff have been discriminated against.” T. at 2067.

Review of a district court’s eviden-tiary rulings is for abuse of discretion. Eagleston v. Guido, 41 F.3d 865, 873 (2d Cir. 1994); United States v. Jamil, 707 F.2d 638, 642 (2d Cir.1983). Even under this lenient standard, however, the court’s failure to receive this expert testimony cannot be upheld. Evidence of the treatment of married women throughout the college was highly relevant to a proper assessment of plaintiffs claim. Tenure decisions at Vassar are considered at two levels: first, departments recommend professors for or against tenure; then the Appointments Committee and the president prepare a final recommendation for the Board of Trustees. There is no intermediate stage in which the departments deemed “hard” sciences participate collectively in the tenure review of scientists.

Given the small size of the “hard” sciences sample — plaintiff’s summary chart for the years 1972-1985 shows a population of 11 “non-single” women — it is questionable whether any statistics limited to the “hard” sciences carries any persuasive weight as a matter of law. See, e.g., Pace v. Southern Railway Sys., 701 F.2d 1383, 1389 (11th Cir.) (twelve employment decisions over twelve years too few to be meaningful), cert. denied, 464 U.S. 1018, 104 S.Ct. 549, 78 L.Ed.2d 724 (1983). At the very least, we hold that it was an abuse of discretion for the district court to fail to consider the university-wide statistics. The court’s decision to base its analysis sole ly on the statistics related to the “hard sciences” was error.

d. Women Granted Tenure in the “Hard” Sciences. Plaintiffs statistics included all women who were “Terminated or left” while plaintiff was at Vassar. However, the statistics do not take into account what happened to women who were hired while Fisher was at Vassar, were not scheduled for tenure review until after 1986, and stayed on after plaintiff left Vassar in 1986. Thus, the only tenure-track women hired while plaintiff was at Vassar who were included in plaintiffs “statistical” chart were those who left the college, and the most promising and successful of plaintiffs married female colleagues were excluded from the statistical model. 11

There were six married female Vassar professors in the natural sciences who were in tenure-track positions at the time Fisher was denied tenure. Most did quite well. (1) Janet K. Andrews was hired as a member of the psychology department in 1979. She was married at the time of her reappointment in 1983 and at the time she was awarded tenure in 1988. T. at 2331, 2332; JE 668, 781. Had she been properly placed in plaintiffs statistical summary chart, she would have been placed in the married/promoted category. (2) Debra Meloy Elmegreen was hired as a member of the Physies/Astronomy Department in 1984. She was married at the time of hire, contract extension and during her 1990 grant of tenure. T. 2318, 2328; JE at 786. Had she been included in plaintiffs statistical summary chart she too would have been placed in the married/promoted category. (3) Janet M. Gray was hired as a visiting member of the psychology department in 1981. She was married when she was hired, when her contract was extended, when she was reappointed and when she was granted tenure in 1990. T. at 2217-2220, 2318, 2334; JE at 782. Had she been included in plaintiffs chart she too would have fallen into the married/promoted category. (4) Amy G. Halberstadt was hired as a member of the psychology department in 1981. She was married at the time of her reappointment in 1985 but had been divorced by the time she was denied tenure in 1988. T. at 2336; JE at 675, 789. Had she been included in plaintiffs chart she presumably would have fallen into the married/terminated or left column. (5) Nancy M. Ide was hired as a member of the computer science department in 1982. She was married at the time of hire, contract extension and during her 1989 grant of tenure. T. at 2318, 2334; JE at 784. Had she been included on plaintiffs statistical summary chart she would have been placed under the married/promoted category. (6) Finally, Mary Louise Serafine was hired as a member of the psychology department in 1983. She was married at the time of her reappointment in 1987 but resigned before she could be reviewed for tenure. Had she been included in plaintiffs chart she presumably would have fallen into the married/terminated or left column. T. at 2335; JE at 785.

In sum, plaintiffs statistics excluded reference to four married women employed by Vassar during plaintiffs employment who subsequently received tenure at Vassar in a hard science department. The district court erred in failing to factor these individuals into its analysis of Vassar’s treatment of married women.

e. Sex to Sex Comparison. Plaintiffs “statistical” model focused solely on Vassar’s treatment of women. There is nothing to show how Vassar treated married men or unmarried men. Absent this sex-to-sex comparison on a statistically sound basis, plaintiffs statistics are meaningless. To establish that Vassar discriminated on the basis of sex plus marital status, plaintiff must show that married men were treated differently from married women. See Bryant v. International Sch. Servs., Inc., 675 F.2d 562, 575 (3d Cir.1982) (Higginbotham, J.) (to prevail, plaintiff must show “that married males, in circumstances similar to appellants, received better, or even different treatment”). One cannot expect a plaintiff to compare statistically significant groups of men and women who have taken a prolonged hiatus from academia, but Fisher failed to present any evidence to show that married males who were up for tenure received “better, or even different treatment.” See id 12

If Vassar was as unlikely to promote married men as it was to promote married women, then the only thing one could say is that Vassar discriminated against married people. But marital status alone is not a ground for bringing a suit under Title VII. Absent a legally and statistically meaningful comparison between men and woman, similarly situated, plaintiffs statistics carry no weight. 13

f. Marital Status and Parent Status. The district court notes: “Part of plaintiffs claim of sex discrimination due to her marital status involves the fact that plaintiff left formal academia for approximately eight years in order to raise a family.” Fisher, 852 F.Supp. at 1226. Ultimately, it is unclear exactly what allegations constitute plaintiffs claim. Did Vassar discriminate on the basis of sex plus marital status or on the basis of sex plus children, or sex plus a hiatus to raise children, or some combination of these theories? What is clear is that at no point did plaintiff offer statistics reflecting the treatment of women (or men) who had children.

All told, the statistics adopted by the district court are merely organized anecdotes. They cannot be credited or relied upon to support a result.

(iv) Expert Testimony.

Plaintiff presented expert testimony concerning the prejudice against married women in science. At trial Dr. Rita Simon, a sociologist, opined:

Drawing upon data from a great many universities and thousands of responses from men and women in the hard sciences, and examining the performance of Cynthia Fisher, versus the other people who were coming up for tenure at the same time as she did, I concluded that Vassar was discriminating on the basis of marital status.

T. at 966. The district court concluded that this testimony

established that in academia, married women are more productive than single women in many fields. Particularly in the “hard sciences”, they are as productive as men, but are often not given the same rewards as either men or single women in terms of professional rank.

Fisher, 852 F.Supp. at 1216. Accordingly the district court concluded that Vassal’ has “consistently shown prejudice toward its married female faculty in the hard sciences,” notwithstanding Vassar’s protest that it “advances the cause of women.” Id.

As the finder of fact, the district court was under no obligation to accept the word of Dr. Simon simply because she had been qualified as an expert. See United States v. Williams, 583 F.2d 1194, 1200 n. 13 (2d Cir.1978) (trier of fact entitled to reject expert testimony), cert. denied, 439 U.S. 1117, 99 S.Ct. 1025, 59 L.Ed.2d 77 (1979). As it happens, Dr. Simon’s conclusion concerning discrimination at Vassar was based on nothing more than (1) her own review of the statistics presented at trial by the plaintiff, T. at 958-62, and (2) Dr. Simon’s own nationwide study of female science professors. T. at 956-57. Dr. Simon’s own study arguably demonstrates that, nationwide, married women professors in the hard sciences have been more productive than other groups of professors. This broad-gauge observation has no probative bearing on the treatment of married women at Vassar, as to which Dr. Simon relied entirely on plaintiffs statistics — statistics that are insubstantial, immaterial and riven by fallacies. Dr. Simon’s conclusion was, therefore, not based on credible evidence. The district court’s adoption of Dr. Simon’s conclusion was clearly erroneous.

All told, plaintiffs evidence does not amount to proof sufficient to support a finding of liability. The district court’s conclusion that Vassar college discriminated against plaintiff on the basis of her status as a married woman was clearly erroneous. We therefore reverse the finding of liability under Title VII.

S. Discrimination Based on Parenthood or Nurture.

The evidence supports an inference that Fisher’s eight-year absence from academia hurt her chance for tenure. Fisher treats this hiatus as a proxy for having children and family and suggests that a sex-based animus thereby underlies this aspect of Vassar’s tenure process. Vassar argues (in its reply brief) that the only way a claim of “sex plus” discrimination could be predicated on the detrimental effects of prolonged professional inactivity would be by comparing (a) the tenure experience of women who took extended leaves of absence from their work (regardless of the reason), with (b) the tenure experience of men who had also taken long leaves of absence. Reply Brief at 13.

We agree. In making tenure decisions, it is perfectly reasonable to consider as a factor the candidate’s prolonged absence from academia. The law does not prevent employers from considering such things. Cf. Troupe v. May Dep’t Stores, 20 F.3d 734, 738 (7th Cir.1994) (an employer is “require[d] ... to ignore an employee’s pregnancy, but ... not her absence from work, unless the employer overlooks the comparable absences of non-pregnant employees”). Moreover, the choice of remaining at home for an extended period following the birth of a child “is not the inevitable consequence of a medical condition related to pregnancy”. Maganuco v. Leyden Community High Sch. Dist. 212, 939 F.2d 440, 444-45 (7th Cir.1991); see also Schafer v. Board of Pub. Educ., 903 F.2d 243, 247 (3d Cir.1990) (policy granting leave of absence for childbearing to female employees but not to male employees may serve as the basis for a Title VII claim). Compare Brief of Amici Curiae Equal Rights Advocates and NOW Legal Defense and Education Fund, at 2 (suggesting that only women have family responsibilities). A policy may discriminate between those employees who take off long periods of time in order to raise children and those who either do not have children or are able to raise them without an appreciable career interruption. That is not inherently sex specific and does not give rise to a claim under Title VII.

k. “Simple” Sex Discrimination.

The district court concluded that plaintiff failed to set forth even a prima facie Title VII claim based on “simple” sex discrimination. In reaching this result, the court noted that several women sat on the departmental committee that recommended against tenure and that, at the same time Fisher was denied tenure, another woman (Pinina Norrod) was granted tenure and a man (Edward Tucker) was denied tenure. We agree with plaintiff that the district court erred in basing her dismissal of plaintiffs claim of simple sex discrimination on this rationale.

The participation of women in plaintiff’s tenure review does not establish she was shielded from sex discrimination. Similarly, Vassar’s grant of tenure to another female biology professor at the same time Vassar rejected the plaintiff does not exclude the possibility that plaintiff was subjected to discrimination based on her sex. See Lam v. University of Hawaii, 40 F.3d 1551, 1561 (9th Cir.1994) (favorable treatment of one member of a protected class does not rule out the possibility that another member of the same class suffered discrimination); Gutzwiller v. Fenik, 860 F.2d 1317, 1320-21 (6th Cir.1988) (tenure position denied one female faculty member in favor of another). Nevertheless, at no point did plaintiff present any credible evidence that Vassar’s decision to deny her tenure was in any way based on her sex. 14 Thus, whether or not Fisher established a prima facie Title VII case based on “simple” sex discrimination, she ultimately failed to carry her burden of persuasion. The reasons set forth above, any findings made by the district court to the contrary, are clearly erroneous. The district court properly rejected plaintiffs claim of discrimination based on plaintiffs status as a woman.

B. Age Discrimination Claim.

The district court held Vassar liable for age discrimination, and relied on that violation of the ADEA to double Fisher’s actual damages. 29 U.S.C. §§ 216, 626. In its appeal, Vassar argues that the district court improperly allowed plaintiff to amend her complaint on the last day of trial to include this claim and that the finding of discrimination under the ADEA was clearly erroneous.

1. Amendment of the Complaint.

At the close of plaintiffs case, the district court permitted Fisher to amend her complaint to include a claim for discrimination under the ADEA. Plaintiffs initial complaint made no mention of the ADEA and did not allege that Vassar had discriminated against her on the basis of her age. 15 Three times prior to the commencement of the trial, plaintiff amended her complaint; none of these amendments mentioned age discrimination or invoked the ADEA. When the issue surfaced at one pretrial stage of the case, Judge Cedarbaum (who had initially been assigned to the case) denied an application to add an ADEA claim. At a colloquy on the first day of the trial (by then before Judge Motley), plaintiffs attorney took note of that ruling and conceded that age was “not in the case.” 16 On the last day of the trial, Judge Motley sua sponte suggested that plaintiffs attorney amend the complaint to include the ADEA claim. T. at 2390. The district court then heard further evidence from the plaintiff concerning this claim. At no point did Vassar claim prejudice by reason of this amendment or seek a continuance.

Fed.R.Civ.P. 15(b) allows a party to amend its pleadings to conform to the proof received into evidence. The decision of whether to allow such an amendment is left to the discretion of the district court judge. See Grand Light & Supply Co., Inc. v. Honeywell, Inc., 771 F.2d 672, 680 (2d Cir.1985). When such a motion is made after the start of trial, however, it should be granted only “if the party against whom the amendment is offered will not be prejudiced by the amendment”. Hillburn by Hillburn v. Maher, 795 F.2d 252, 264 (2d Cir.1986), cert. denied, 479 U.S. 1046, 107 S.Ct. 910, 93 L.Ed.2d 859 (1987). Ordinarily, the assertion of a new claim for liability on the last day of trial will be prejudicial. Nevertheless, absent an objection or a request for a continuance, it was not an abuse of discretion for the district court’s decision to allow the amendment.

2. Merits.

ADEA claims are subject to the same burden-shifting applicable to Title VII claims, as outlined in McDonnell Douglas and St. Mary’s. In order to establish a prima facie case of age discrimination, Fisher was required to show (1) that she was at least 40 years old but less than 70, (2) that she was otherwise qualified for the job, and (3) that she was discharged “under circumstances giving rise to an inference of age discrimination.” See Stetson v. NYNEX Serv. Co., 995 F.2d 355, 359 (2d Cir.1993). On appeal, defendant claims that plaintiffs ADEA claim does not meet the third of these prima facie requirements. In the alternative, defendant argues that, even if plaintiff could be said to have articulated a prima facie case of age discrimination, she has not ultimately satisfied her burden under the third step of McDonnell Douglas: she has not established that the reason she was denied tenure was discrimination based on her age. Stanojev v. Ebasco Servs., Inc., 643 F.2d 914, 920 (2d Cir.1981).

A plaintiff may establish a claim of age discrimination through “direct evidence, such as statements by the employer that age was the reason for the discharge, or through proof of circumstances from which an inference of age discrimination may be drawn.” Haskell v. Kaman Corp., 743 F.2d 113, 119 (2d Cir.1984). If direct evidence is not available, “a plaintiff may through statistical evidence establish a pattern or practice of discharging or failing to promote older employees, from which an inference of age discrimination may be drawn.” Id. at 119.

The district court’s analysis of plaintiffs ADEA claim consisted of the following:

[Plaintiff] was 53 years old at the time of her tenure denial, she was well qualified for the position and the statistical evidence on faculty members ages at tenure review showed that all other tenured faculty who were equally or less qualified than Dr. Fisher were at least nine years younger than Dr. Fisher when they were tenured.

That some of the tenured faculty members were over 40 when they were promoted does not undercut this finding.

Fisher, 852 F.Supp. at 1230 (footnote omitted).

We uphold the district court’s findings that plaintiff satisfied the first two requirements of an ADEA prima facie case. Fisher was clearly within the protected age group and, given that we have upheld the district court’s assessment of plaintiffs credentials under the clearly erroneous standard, we proceed on the basis that plaintiff was otherwise qualified for the job. Furthermore, we agree with the district court that the mere fact that other faculty members were over 40 when they were granted tenure in the biology department does not foreclose a claim by plaintiff under the ADEA. See Maxfield v. Sinclair International, 766 F.2d 788, 792 (3d Cir.1985), cert. denied, 474 U.S. 1057, 106 S.Ct. 796, 88 L.Ed.2d 773 (1986). Although three of the other eight biology professors were over 40 years of age at the time they were awarded tenure by Vassar, all three were significantly younger than was the plaintiff, who was 53 at the time of her tenure denial. Thus, we conclude that plaintiff sufficiently set forth a prima facie claim of age discrimination.

Nevertheless, we find that the evidence presented to the district court was ultimately insufficient to support plaintiffs ADEA claim. As was the case with plaintiffs Title VII claims, resolution of this issue turns on the application of step three of the McDonnell Douglas burden shifting test. Even if we accept the district court’s conclusion that plaintiff was qualified to receive tenure in the Vassar biology department, we reject the court’s conclusion that she was denied tenure because of impermissible discrimination based on age.

The district court cited two items of direct evidence. First, responding to questions posed by the court at trial, Fisher testified that in 1979 Professor Lumb told her she “was too old to ever become tenured faculty.” T. at 2392. Second, the district court eites to the comment at the May 3 meeting of the Appointments Committee that the Biology Department believed plaintiff was “out of date — out of field for 10 years.” JE at 288. However, the district court specifically rejected the significance of this evidence:

[T]hese statements do not establish, by themselves, direct evidence of an illegitimate factor. The court interprets the Department’s statement to the [Appointments Committee] not as direct evidence that age played a part in its decision but rather as an attempt to convey that she was not up to speed in educational requirements and/or research techniques even though this is contrary to the evidence. Furthermore, Dr. Lumb’s statement that Dr. Fisher was too old to be considered for tenure is not attributable to the Tenure Committee’s decision making process since she was not a member. This statement could be attributed to the sentiment of the Department but on its own does not meet plaintiffs burden.

Fisher, 852 F.Supp. at 1231 n. 25. To support the requisite “inference of age discrimination,” the district court relied entirely upon a listing of the six faculty members who were already tenured in 1985, the one faculty member who received tenure in 1985, the first faculty member to be tenured after 1985, and Fisher herself, together with the age of each person at the time that person was considered for tenure:

Fisher 53

Norrod 44

Mehaffey 39

Hemmes 38

Suter 38

Schlessman 35

Williams 33

Johnson 44

Greenwood 48

Fisher, 852 F.Supp. at 1219.

This table does not amount to proof sufficient to support a finding of discrimination under the ADEA. Nothing in plaintiffs list indicates that Vassar had a policy or practice of denying tenure to older candidates. Plaintiff has not shown that Vassar discriminates on the basis of age; she has only shown that she was the oldest candidate in the biology department to have been considered for tenure. Since few candidates for tenure have had an eight year break in their careers, it is unsurprising (and not at all sinister) that most of the candidates to be considered for tenure in the biology department were in their late 30s and early 40s.

We would be reluctant in any event to draw conclusions from a sample of only eight faculty members out of a College-wide pool in excess of 100 faculty members who were eligible for tenure between 1972 and 1985. This circuit and others “have eonsis-tently rejected similar statistical samples as too small to be meaningful.” See Haskell, 743 F.2d at 121 (ten terminations over eleven years too few to be statistically reliable); see also Pace v. Southern Railway Sys., 701 F.2d 1383, 1389 (11th Cir.) (twelve employment decisions over twelve years too few to be meaningful), cert. denied, 464 U.S. 1018, 104 S.Ct. 549, 78 L.Ed.2d 724 (1983); Coble v. Hot Springs Sch. Dist. No. 6, 682 F.2d 721, 723 (8th Cir.1982) (15 employment decisions over eight years “too small to support any inference of discriminatory pattern or practice”); Turner v. Texas Instruments, Inc., 555 F.2d 1251, 1257 (5th Cir.1977) (eight employment decisions over two years too few to be significant). The district court’s conclusion that the evidence presented by plaintiff amounted to sufficient proof to establish a claim under the ADEA was clearly erroneous. 17

C. Equal Pay Act Claim.

Vassar challenges the district court’s award of damages under the Equal Pay Act on the grounds that (a) the district court abused its discretion in suggesting that plaintiff amend her complaint to include the Equal Pay Act claim after the conclusion of plaintiffs presentation of her case; (b) the claim was barred by the statute of limitations; and (c) plaintiff failed to prove her Equal Pay Act claim.

1. Amendment of the Complaint.

Plaintiffs claim under the Equal Pay Act was added at the conclusion of plaintiffs case. T. at 1357-59. Defendant objected to this amendment. T. at 1357. However, although the district court offered Vassar the opportunity to brief the issue at the end of trial, T. at 1359, Vassar never submitted such a brief and never argued that Vassar would be prejudiced by such an amendment.

As noted above, the decision of whether to allow such an amendment is left to the discretion of the district judge. See Grand Light & Supply Co., 771 F.2d at 680. As in the case of the ADEA claim, defendant waived any objection it might have had to the amendment in this case by failing to timely file a brief in opposition to the amendment. The district court’s decision to allow plaintiff to amend her complaint to include the Equal Pay Act claim was not an abuse of discretion.

2. Statute of Limitations.

Appellant urges for the first time on appeal that plaintiffs Equal Pay Act claim is barred by the statute of limitations. Whatever the merits of this argument, the statute of limitations is an affirmative defense and can be waived. See Professional Firefighters Ass’n v. City of Clayton, 759 F.Supp. 1408, 1415 (E.D.Mo.1991). The defense cannot be raised for the first time on appeal. Davis v. Bryan, 810 F.2d 42, 44 (2d Cir.1987). The amended complaint (which included the Equal Pay Act claim) was filed on September 24, 1993, two months after trial and nearly eight months before the court issued its opinion. Vassar never interposed the statute of limitations by way of answer or motion. Since this issue was never brought to the district court’s attention, notwithstanding reasonable opportunity to do so, defendant has waived this defense.

S. Merits.

The district court assessed damages under the Equal Pay Act based upon a comparison of plaintiffs salary with the salary of Professor Robert Suter, a man hired as a visiting assistant professor at approximately the same time as plaintiff. The court limited damages to the three year period from July 1,1983 through June 30,1986 for a total sum of $8,265. The court then doubled this sum to $16,530 pursuant to the liquidated damages provision of the Equal Pay Act.

Under the Equal Pay Act,

[n]o employer ... shall discriminate ... between employees on the basis of sex by paying wages to employees ... at a rate less than the rate at which [the employer] pays wages to employees of the opposite sex in such establishment for equal work on jobs the performance of which requires equal skill, effort, and responsibility, and which are performed under similar working conditions, except where such payment is made pursuant to (i) a seniority system; (ii) a merit system; (iii) a system which measures earnings by quantity or quality of production; or (iv) a differential based on any other factor other than sex.

29 U.S.C. § 206(d)(1). In order to sustain a claim under the Equal Pay Act, a plaintiff must demonstrate that the jobs at issue require “equal effort, skill and responsibility, ... and are performed under similar working conditions,” McKee v. McDonnell Douglas Technical Servs., 700 F.2d 260, 262 (5th Cir.1983), and that any difference in pay is attributable to sex. The defendant then bears the burden of demonstrating that one of four stated exceptions is applicable. See Winkes v. Brown Univ., 747 F.2d 792, 793 (1st Cir.1984).

In assessing the Equal Pay Act claim, the district court found that Fisher was paid less than some of her male colleagues with equal or lesser credentials. Fisher, 852 F.Supp. at 1232. The court further found that her duties were substantially similar to the duties of her higher paid male colleagues, noting that “the job of an assistant professor in the biology department is essentially the same with regard to skills required, effort involved and responsibility afforded to the instructor regardless of the specific courses being taught by that instructor.” Fisher, 852 F.Supp. at 1232. Finally, the district court found that Vassar had “failed to present evidence that would show that the pay disparity was based on factors other than sex. Fisher, 852 F.Supp. at 1232.

Plaintiff argued that her salary should never have been lower than Suter’s. Yet, plaintiff never introduced evidence establishing that she and Suter performed equivalent work. Indeed at one point, she acknowledged that he had responsibilities she did not share. T. at 2388.

Ultimately, the available data does not support the district court’s ruling that Vassar paid junior female professors less, as a group, than junior male professors. Four pages of charts listing professors’ salaries in the Vassar biology department are relevant to this inquiry. JE at 514-517. The first compares starting salaries of four people who started in the biology department in 1977: Fisher, Palmer (female), Friedman (female) and Suter (male). All we learn from this chart is that Suter, the male, earned less than one of the women (Friedman) and more than the other two (Palmer and Fisher). The second chart compares Fisher’s salary to salaries paid to four of the men who entered the biology department during the 1970s. The chart demonstrates that, of those five professors, plaintiff is the only one not to have received more than the minimum required pay during her first year at Vassar. One foundational problem is that we do not know who else was hired in the biology department in this period, or what starting salary they earned. Certainly the chart excludes other women hired in the 1970s, such as Friedman who (as the first chart shows) earned more than the minimum salary during her first year.

The third chart shows that plaintiffs salary was consistently lower than Suter’s throughout her time at Vassar. Plaintiff consistently earned less than Suter. However, Suter was placed on tenure track in 1978, three years before Fisher, and remained ahead of Fisher. Thus, Suter was ultimately granted tenure in 1983, some two years before Fisher was denied tenure. Since Suter was more senior than plaintiff, it cannot be deemed unreasonable for Vassar to pay him more money. In her testimony before the district court, plaintiff acknowledges that tenure-track professors were entitled to receive a higher starting salary than non-tenured professors. T. at 517. Under the clear terms of the Equal Pay Act, there is no liability and no discrimination if salary discrepancies can be explained through either seniority or merit. 29 U.S.C. § 206(d)(1).

The final chart compares the salaries of seven biologists, four women and three men, during the first year of each professor’s full-time Vassar contract. It shows that the two individuals who received the largest premium over the minimum starting salary were both women, and that the two individuals who failed to receive any premium over the minimum salary were also women. This chart, as well as the three preceding, fails to demonstrate that Vassar violated the Equal Pay Act. The district court’s finding that Vassar discriminated on the basis of sex in setting its pay scales for junior professors in the biology department is without factual basis and is therefore reversed.

D. Attorneys’ Fees.

Only a “prevailing party” may recover attorneys’ fees and costs in a civil rights action. See Clarke v. Frank, 960 F.2d 1146, 1152 (2d Cir.1992); 42 U.S.C. § 2000e-5(k). Our reversal of the district court on the merits of plaintiffs suit therefore would automatically require that we vacate the award of attorneys’ fees and costs. In her sur-reply brief, however, plaintiff argues that we should reject Vassar’s request to vacate the award of attorneys’ fees and costs, regardless of whether or not we uphold the district court’s finding on liability.

After the district court found that the plaintiff was “entitled to recover attorneys’ fees and costs” and announced that “[a] separate hearing [would] be held on the matter,” Fisher, 852 F.Supp. at 1234, the parties entered into a stipulation fixing the attorneys’ fees. The document was signed by counsel, and was endorsed by the district court on September 28,1994. The award of damages, together with these funds, were placed in an escrow account payable “only upon the completion and exhaustion of any appeals”. JA at 361.

Win or lose, plaintiff argues, this stipulation bars any challenge to the award of attorneys’ fees and costs. In support of this argument, plaintiff cites to Cohen v. Virginia Electric & Power Co., 788 F.2d 247, 249 (4th Cir.1986):

Cohen now attempts to escape the effect of the consent judgment by arguing that he was only consenting to the amount of fees, and that he was doing so because the court had already ordered that fees would be awarded. Nevertheless, the plain language of the consent order makes clear that the parties are acquiescing in the award of attorneys fees. There is no indication on the fact of the order that Cohen was reserving any rights to appeal the judgment.

Plaintiff argues that we should follow the lead of Cohen and allow her recovery of her attorneys’ fees notwithstanding our decision on the substantive claims.

We need not determine whether Cohen represents the law of this Circuit. The consent signed in Cohen was factually distinguishable from the stipulation signed in this case. The stipulation in this case recites that

said amounts [are] to be paid out of the Escrow Fund ... only upon the completion and exhaustion of any appeals from the said Order and Judgment dated June 30, 1994 [the liability judgment], and the Judgment for Attorneys’ fees and costs to plaintiff as entered herein; or the final determination of any remand, retrial, or rehearing of any provision in said Order and Judgment filed June 30, 1994, or the Judgment to be filed herein; or of any appeal therefrom. Defendant has filed a Notice of Appeal on or about July 13, 1994 with the Clerk, United States District Court, Southern District, New York, appealing each and every part of said Order and Judgment, dated June 30, 1994, to the United States Circuit Court, Second Circuit. Defendant intends to file a Notice of Appeal with the Clerk, United States Court, Southern District of New York, appealing each and every part of the Judgment to be entered herein awarding attorneys’ fees and costs to plaintiff, to the United States Circuit Court, Second Circuit. Defendant intends to request consolidation of its appeals from the two said separate judgments.

JA at 361. In thus reserving the right to challenge the award of attorneys’ fees on appeal, the stipulation in this case did exactly what the Cohen court found the stipulation before it had failed to do. Because plaintiff is not a “prevailing party,” we vacate the district court’s award of attorneys’ fees and costs.

CONCLUSION

In summary, we vacate the judgments of the district court because we find that the district court’s findings of liability on the sex discrimination claim, the age discrimination claim and the Equal Pay Act claim were clearly erroneous and because, absent success on one of these claims, plaintiff is no longer entitled to attorneys’ fees. We affirm the district court’s rejection of plaintiff’s claim for “simple” sex discrimination. We have no need to reach plaintiffs cross-appeal challenging the terms of her reinstatement at Vassar. We direct the district court to dismiss plaintiffs lawsuit.

2 . Testimony at trial suggested that plaintiff herself invited attention to her hiatus. H. Patrick Sullivan, who served as dean of the college during plaintiff's tenure review process, testified as follows:

Q: What is your understanding with regard to that statement [referring to the department’s reference to plaintiff's hiatus] as expressed on the first paragraph of page 2 as to who injected that as a factor in the tenure approval process?

A: I think the telling phrase is ‘‘by her own report.” My sense of that is that it was a matter which Ms. Fisher in particular had cited in speaking of her accomplishments.

Transcript at 2131.

3 . The district court made the erroneous finding that “the College-wide faculty committee gave a positive recommendation for tenure.” 852 F.Supp. at 1230.

4 . As further evidence of Vassar's animosity toward her, plaintiff points to the difficulty she had securing the transfer of her laboratory from Vassar to Bard.

5 . In its original May 16, 1994 opinion, the district court indicated that it was "unclear to the court whether Dr. Fisher still seeks reinstatement due to her change in career”. 852 F.Supp. at 1234. In a hearing before the district court, plaintiff indicated that she still sought reinstatement as an associate professor.

6 . The court found Vassar to be liable for $256,-653 in back pay as the difference between the income actually earned by plaintiff between July 1, 1986 and June 30, 1994 and the average salary of an associate professor in the biology department at Vassar. Because the court found defendant acted wilfully, and because Fisher elected to receive her recoveiy under the ADEA, the court doubled this figure. Plaintiff's total recovery under the ADEA was $513,306. The district court found damages under the Equal Pay Act in the amount of $16,530. The court further ordered Vassar to make contributions on Fisher's behalf to the Teacher's Annuity Funds ($79,969.12) and Social Security ($17,067).

7 . At least some of the district court’s findings regarding plaintiff’s scholarship were unfounded. For example, the district court criticized the College for "improperly ignoring]” the appraisals of the outside evaluators. 852 F.Supp. at 1205. However, Vassar’s policy is to prepare department reports without considering such evaluations. The evaluations were, however, consulted by the Faculty and Appointments Committee, the dean and the president prior to the final vote on plaintiff's tenure.

8 . Of course, social judgments are sometimes influenced by impermissible sexual stereotypes. See Price Waterhouse v. Hopkins, 490 U.S. 228, 109 S.Ct. 1775, 104 L.Ed.2d 268 (1989).

9 . Of course, plaintiff's claim is not that she was dismissed for having a family but solely that she was dismissed for being a married woman.

10 . At no point during the trial did plaintiff offer an expert to establish the technical validity of her statistical model. She did, however, offer the testimony of two "fact checkers”. T. at 1291— 1308.

11 . We acknowledge that the district court amended its initial opinion to include the finding that Vassar's policy of discriminating against married women in the hard sciences changed after 1986. Fisher, 852 F.Supp. at 1235. Such a finding does not, however render irrelevant the evidence concerning the professors employed by Vassar during plaintiff's career.

12 . In an instance in which the complainant establishes by evidence that there are no married males who were up for tenure and that it is unlikely that there would be any, then it may be that the complainant would be able to prevail by providing some other evidence of discrimination. However, since that situation is not presented, we need not reach that issue.

13 . Plaintiffs brief on appeal argues that the district court did compare Vassar’s treatment of married women faculty members to married men when it compared plaintiff’s qualifications to the qualifications of the four married men who were in the biology department while plaintiff was at Vassar and who obtained tenure prior to 1987. Brief of Appellee at 33-34. While the court did compare the academic and teaching records of these four men to the court’s assessment of plaintiff’s record, the plaintiff never in its statistical analysis differentiated between men who were married and men who were not. Thus, despite plaintiff's objections, her statistical analysis cannot be said to have compared Vassar’s record of granting tenure to married men with its record of granting tenure to married women.

14 . In its amicus brief, the Equal Employment Opportunity Commission argues that the district court erred when it separated its analysis of sex discrimination from sex plus marriage discrimination. This is a valid criticism. Sex plus marriage discrimination is simply a sub-categoiy of sex discrimination. To discriminate on the basis of sex plus marriage is to discriminate on the basis of sex.

15 . Plaintiff's original complaint to the State of New York Department of Human Rights did note that she "charge[d] the respondent with denying [her] equal terms, conditions, and privileges of employment and dismissing [her] because of [her] age, sex, and marital status....” Supplemental Joint Appendix at 1 (emphasis added).

16 .Judge Cedarbaum’s ruling excluding both the ADEA claim and the EPA claim was apparently based on her conclusion that plaintiff's initial complaint filed with the New York Division of Human Rights did not encompass these claims and that they were, therefore, barred. Judge Motley sua sponte revisited this decision.

17 . Because we conclude that the district court erred in finding Vassar liable on the ADEA claim, we have no need to consider appellee’s additional contention on appeal that the district court erred in finding that Vassar's violation of the ADEA was willful.