BREYER, Circuit Judge.

The appellant, Ann Seidman, claims that Brown University unlawfully discriminated against her because of her sex when — some seven years ago — it refused to make her the Henry R. Luce Professor of the Comparative Study of Development. She now appeals a district court judgment in Brown's favor, 613 F. Supp. 971. Because this appeal depends heavily upon matters of fact, we have read the 2200-page record with care. We conclude that we must remand the case to the district court for further proceedings, though we shall narrow the focus of those proceedings.

I

This case arises under a 1978 consent decree entered in a Title VII class action brought against Brown University in 1975. (The decree is reprinted in the appendix to Lamphere v. Brown University, 491 F. Supp. 232, 238-46 (D.R.I. 1980), aff'd, 685 F.2d 743 (1st Cir. 1982).) Although the decree provides for certain special procedures and special 'burden of proof' rules in respect to faculty hiring at Brown, the present case is, in a sense, typical of many that arise under Title VII. That is to say, the appellant points to factual circumstances that, if unexplained, strongly suggest she was treated differently, and prejudicially, because of her sex. The university offers a neutral, nondiscriminatory explanation of these circumstances. But, the appellant claims that this explanation is but a 'pretext' that hides the university's true, discriminatory motive. The basic argument in this case concerns the last point: Does Brown's explanation for its conduct amount to a pretext masking sex discrimination? Or (in terms of the decree's burden-of-proof rules) has Brown 'clearly and convincingly' proved it does not?

One can best understand the argument by starting with a rough outline of the key events. In June 1978 Brown announced that the Luce Foundation would give it $250,000 over five years to establish the Henry R. Luce Professorship in the Comparative Study of Development. The stated object of the grant was "to further interdisciplinary research and teaching on problems of social and economic development." Brown expected to appoint "an economist or another social scientist concerned with economic analysis who is willing and able to work with [other social scientists] . . . toward an integration of economic and institutional sociological analysis."

Brown began the search for candidates by issuing an announcement called a "personnel vacancy authorization" (PVA), which said, among other things, that the "Luce professor will hold a joint appointment in the Departments of Sociology and Economics" and that the "professorship is a permanent, tenured position." (The relevant text of the PVA is reprinted in Appendix A.) Brown also appointed a search committee consisting of three professors from the Sociology Department, one professor from the Economics Department, and two alternates (one from Sociology and one from Economics). The search committee considered 37 applicants, placed nine on a short list, and interviewed five (four men and one woman). At a meeting on December 18, 1978, the search committee decided to recommend that Brown offer the job first to Pranab Bardhan, a professor of economics at the University of California; then (should Bardhan decline) to Paul Streeten, an adviser to the World Bank who had served as director of development-related institutes at Sussex and Oxford universities; and then (should Streeten decline) to the appellant, who was a (nontenured) visiting professor at Brown. The search committee qualified these recommendations in two important ways. First, the offer, if extended to Streeten, would consist of a chair in the Sociology Department alone, not in both Sociology and Economics. Second, the offer, if extended to Seidman, would be in Sociology only, and would consist not of a permanent, tenured professorship, but rather of a three- to five-year, nontenured appointment.

Normal procedure required the search committee to report to the relevant academic department or departments, whose tenured members would then pass upon the committee's recommendation. If approved by the department, the recommendation would be presented by the university provost to a university-wide appointments committee called the Committee on Faculty Reappointment and Tenure (CONFRAT). If CONFRAT approved, the recommendation would go to the university president, and, with his approval, to the university's trustees. These procedures were followed in respect to Bardhan's application. Both Economics and Sociology recommended his appointment, and Brown offered him the job; but in early January of 1979 he declined the offer.

There were two further developments in early January. First, the president of Brown decided that the Luce chair need not be a joint appointment in both Sociology and Economics, as had been originally indicated in the PVA. (The PVA was not formally amended to reflect this change.) Second, the university provost informed the chairman of the search committee, Professor Dietrich Rueschemeyer, that the Luce professorship must provide tenure; hence the committee could not recommend offering Seidman a three- to five-year, nontenured appointment as Luce professor.

In light of this new information, Rueschemeyer met on January 22 with Professors Peter Evans and Robert Marsh, evidently the only two other members of the search committee who were in town at the time. Rueschemeyer and Evans agreed that Seidman should remain the committee's third choice for the Luce chair; Marsh said something that, to the other two, sounded like 'grudging acquiescence.' (Marsh would later stress the "grudging.") Whether this gathering was an official meeting of the search committee remains a subject of dispute, in part because the original search plan suggests that Evans and Marsh were to share a spot on the committee, serving alternately. See infra pp. 541-42.

Later that day, Rueschemeyer presided at a meeting of the Sociology Department's tenured faculty. The department reached a consensus decision to recommend Streeten for the job and then, if he declined, Seidman. The Sociology faculty forwarded this recommendation to the provost for presentation to CONFRAT. The provost presented Streeten's recommendation; it received the relevant approvals; and Brown offered the job to Streeten. But, he turned it down.

The provost then did not present the Seidman recommendation to CONFRAT. Rather, he sent it back to the Sociology Department. He said that Brown's procedures required a formal vote on the recommendation and that the department had acted by informal consensus, not by vote. Consequently, the department held another meeting on March 16, 1979. At that meeting the department decided not to recommend Seidman for the chair, but rather to reopen the search. The panel of outside scholars who conducted the new search did not include Seidman on their short list, and Brown eventually appointed a man to the professorship.

The suspicious circumstances to which Seidman points consist of the events surrounding the decision to reopen the search. Why did the provost send her name back to the Sociology Department for a formal vote? The department had not voted formally on either Bardhan or Streeten. Further, what happened at the March 16 meeting of the Sociology faculty? That is to say, why did the department decide to reopen the search, reversing its January 22 recommendation of Seidman? It cannot have done so simply because the Economics Department refused to endorse Seidman, for it knew this in January; indeed, Economics had refused to endorse Streeten, but Sociology approved him anyway. These questions are easy to ask but difficult to answer; and that very difficulty, Seidman suggests, is sufficient under the special burden-of-proof rules to make out a violation of the consent decree.

Brown, in fact, provides lengthy, detailed answers to these questions. See infra pp. 540-42. But, for the moment, we shall put those answers to one side. At this point, we simply note that the district court did not decide whether Brown proved its explanation a true one — that it was not a pretext masking sex discrimination. Instead, the district court noted that Brown's original announcement of the position's availability — the PVA — described a joint appointment in both Sociology and Economics. The Economics Department, however, refused to endorse either Streeten or Seidman for tenure (for reasons having nothing to do with sex). In the district court's view, since the consent decree requires Brown's appointments to meet the terms of a PVA, Brown could no longer make a valid appointment once Bardhan — the only candidate who 'qualified' under the PVA's terms — rejected Brown's offer. According to the court, Seidman's claim must fail because "[w]hatever discrimination [she] suffered took place after the determination that she was not qualified for the position in accord with the terms of the PVA."

Seidman appeals this decision.

II

This case is special because it arises not directly under Title VII, but rather under a consent decree entered in a Title VII case. The decree has several features: it encourages Brown to hire more woman faculty members, setting "goals and timetables" for Brown's doing so; it contains promises to compensate past and then-present woman faculty members and outlines procedures for doing so; it requires Brown to create detailed, specific criteria for hiring and promotion, setting forth detailed hiring and promotion procedures; and it creates a university 'compliance' committee called the Affirmative Action Monitoring Committee, which has power to "implement" and "enforce" the decree, and to hear (at least as an initial matter) complaints that Brown has not lived up to its substantive and procedural promises in respect to hiring and promotions. See Decree, reprinted in 491 F. Supp. at 238-46. The Monitoring Committee is composed of two members selected by the original Title VII plaintiffs, two selected by the university's faculty, and one joint selection. Decree ¶ 2(L), reprinted in 491 F. Supp. At 241-42. Before turning to the main issues in this case, we must consider the burden-of-proof rules contained in the decree and the relation it creates between the Monitoring Committee and the courts.

A. The standard for measuring discrimination. In an ordinary Title VII sex discrimination case, once the plaintiff makes out a prima facie case of discrimination, the employer must "articulate" a legitimate, nondiscriminatory reason for its actions. If the employer does so, the plaintiff must then show, by a preponderance of the evidence, that the asserted reason is a mere pretext for unlawful behavior. See, e.g., Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 252-53, 101 S.Ct. 1089, 1093-94, 67 L.Ed.2d 207 (1981). The burden of persuasion remains on the plaintiff throughout. See id. 450 U.S. at 254-56, 101 S.Ct. at 1094-95.

The provision of the consent decree applicable here, however, changes the traditional standards in two ways: it imposes the burden of persuasion on the employer (even on the ultimate issue of pretext), and it requires the employer to carry that burden by "clear and convincing evidence." To be specific, the decree provides:

In the event that a faculty position is not filled after the prescribed search procedure, . . . [Brown] shall demonstrate . . . by clear and convincing evidence that the decision not to fill the position was nondiscriminatory as to sex . . . .

Decree ¶ F(4), reprinted in 491 F. Supp. at 240. "Clear and convincing evidence" is defined as "an amount of evidence which is less than 'beyond a reasonable doubt' but more than 'a preponderance of the evidence.'" Decree glossary, reprinted in 491 F. Supp. at 246. The decree further provides that the prescribed burden-of-proof rules shall apply both before the Affirmative Action Monitoring Committee and upon court review.

The decree's language raises at least two concerns. The first one, noted by the Monitoring Committee, is a problem of construction, namely, how to interpret the decree to avert undue problems of proof. Concerned that a literal interpretation of the decree would impose on Brown the near-impossible task of "proving a negative" (that it did not discriminate based on sex), the Monitoring Committee rephrased the decree's substantive standard as follows:

the University must show that [its] action can be plausibly explained without reference to considerations of sex, and it must provide clear and convincing evidence that such an explanation is the correct one.

That is to say, the decree requires Brown 1) to articulate a legitimate, nondiscriminatory reason for its actions, and 2) to show by clear and convincing evidence that this reason is not a pretext. The parties basically agree that this interpretation is reasonable, and so do we.

The second concern — this one noted by the district court — involves the relationship between the consent decree and Title VII. The court expressed doubt about whether a consent decree can validly alter the standards that courts usually apply in Title VII actions. This concern likely reflects uncertainty about the limits on a court's power to adopt and apply a consent decree that provides for 'affirmative action' remedies arguably broader than those that a court could order at the close of a litigated Title VII suit. The Supreme Court has recently held, however, that the particular provision of Title VII that limits a court's power to "order" relief does not affect the validity of a consent decree. See Local Number 93, International Ass'n of Firefighters v. City of Cleveland, ___ U.S. ___, 106 S.Ct. 3063, 92 L.Ed.2d 405 (1986) (construing Civil Rights Act of 1964, Title VII, § 706(g), 42 U.S.C. § 2000e-5(g)). Compare Firefighters Local Union No. 1784 v. Stotts, 467 U.S. 561, 104 S.Ct. 2576, 81 L.Ed.2d 483 (1984) (Title VII's limits on remedial authority do constrain court's power to modify a consent decree over a party's objection). To be sure, consent decrees remain subject to scrutiny under § 703 of Title VII — the provision that defines liability for unlawful employment practices, as opposed to the court's power to remedy them. See City of Cleveland, ___ U.S. at ___ n. 8, 106 S.Ct. at 3073 n. 8; 42 U.S.C. § 2000e-2. But the particular form of 'affirmative action' present here — a modification of burden-of-proof rules in favor of the members of a protected class — does not seem so drastic a remedy as to amount to an "absolute bar to the advancement" of others, so long as the rules are applied sensibly so as not to impose an unreasonably difficult burden on Brown. United Steelworkers of America v. Weber, 443 U.S. 193, 208, 99 S.Ct. 2721, 2729, 61 L.Ed.2d 480 (1979) (construing § 703). Since the Monitoring Committee's interpretation offers a sensible construction, since the modified burden-of-proof rules do not plainly exceed the bounds of "reasonable [sex]-conscious relief," City of Cleveland, ___ U.S. at ___, 106 S.Ct. at 3073, and particularly since no party to this case argues that they do, we think it proper to apply the decree's rules in considering Professor Seidman's claim.

B. The scope of court review. Initially, Seidman brought her claim to Brown's Affirmative Action Monitoring Committee. The committee found that Brown had failed to carry its burden of proof under the decree. Brown then moved for de novo consideration in federal district court. It pointed to the following provision of the decree:

[I]n the event that any woman is dissatisfied with a decision of the [Monitoring] Committee which directly affects her, she may seek de novo consideration by appropriate motion to the Court. Likewise, de novo judicial consideration of Committee decisions with respect to implementation or compliance with the provisions of this Decree may be initiated by any class member. The University may also seek de novo consideration of any decision of the Committee by appropriate motion to the Court.

Decree ¶ 2(L), reprinted in 491 F. Supp. at 242.

Seidman argues that this provision does not allow the district court to take new evidence on the question of discrimination. She believes the court can reconsider the Monitoring Committee's decision only on the basis of the committee's own record. She notes that the relevant language of the decree — "de novo consideration" — is narrower than that contained in a different section of the decree (concerning review of compensation awards), which speaks of a "de novo hearing on all the issues," Decree ¶ 2(M)(2)(h), reprinted in 491 F. Supp. at 244. And, she points to a Supreme Court case in which a statutory provision for "de novo determination" was held not to require the trial court to rehear live testimony. See United States v. Raddatz, 447 U.S. 667, 673-76, 100 S.Ct. 2406, 2411-12, 65 L.Ed.2d 424 (1980). Finally, she says it makes sense to pay particular attention to the Monitoring Committee's work, since its members have special familiarity with the university and its tenure process.

The district court rejected Seidman's interpretation of "de novo consideration." It believed the decree required a full judicial trial. And, we conclude that the court was correct.

For one thing, the decree's words "de novo" are typically a signal to start from scratch. Black's Law Dictionary 392 (5th ed. 1979) ("Anew; afresh; a second time."). Seidman correctly notes that the Supreme Court gave the words "de novo determination" a special meaning in Raddatz, supra; but that case has little relevance here. Raddatz involved difficult constitutional questions about the authority of federal magistrates, and its reading of "de novo" turned on the particular legislative history of the Federal Magistrates Act. See 447 U.S. at 673-76, 100 S.Ct. at 2411-12 (House Report explicitly states that "use of the words 'de novo determination' is not intended to require the judge to actually conduct a new hearing").

For another thing, the decree's provision for court review applies to employer and would-be employee alike. If we adopted the appellant's interpretation of "de novo consideration," a disappointed faculty applicant who invoked the review provision would be entitled to less than the full court review normally available to an ordinary Title VII plaintiff; the appellant offers no convincing reason to believe the decree's authors intended such a result (or could lawfully have prescribed it).

Finally, practical considerations do not argue for less than full court review. The Monitoring Committee's institutional expertise lies not so much in its ability to develop a full and fair factual record as in its ability to investigate informally while bringing to bear a judgment informed by experience with university affairs. Any comparative expertise that the court possesses lies in its ability to develop an evidentiary record and to examine that record with a disinterested eye. Seidman's interpretation of the decree's language (that the court should review a factual record made by the committee) risks reversing these comparative institutional advantages. It also would threaten a campus invasion of legalistic procedure, for the Monitoring Committee might feel compelled to follow courtroom rules of evidence and to generate a formal 'record' comprehensive enough to make it suitable for later use in court. There is no reason to think, in the university context, that the forced use of such procedure is desirable — or that the parties in fact desired it when they framed the decree.

All these considerations favor the district court's interpretation of "de novo," which nevertheless leaves the court free to read, and to benefit from, the Monitoring Committee's opinions.

III

As we noted at the end of Part I, the district court did not decide whether or not Brown discriminated on the basis of sex. It rejected Professor Seidman's claim for a different reason. Brown's original notice about the job opening, the PVA, stated that the Luce professor would hold a joint appointment in both the Sociology and Economics Departments. The Economics Department, for reasons not disputed here, refused to endorse either Streeten or Seidman for tenure in Economics. That fact, in the district court's view, was sufficient to invalidate any offer to either candidate, notwithstanding their endorsement by the Sociology Department. Apparently, the district court thought that, since neither Streeten nor Seidman could obtain approval from Economics, and since Bardhan had declined Brown's offer, Brown had to reopen the search. Thus, Seidman's sex discrimination claim must fail, either because she simply was not qualified, or because Brown, in any event and regardless of motive, had to reopen the search.

We cannot accept the court's conclusion because we disagree with its premise, namely, that once Brown deviated from the PVA, it could no longer validly continue the original search process. Having learned that the leading candidate for the Luce chair had declined its offer, Brown responded in a way that kept both the remaining candidates in the running: it relaxed the requirement that the Luce professor be approved for tenure in Economics as well as Sociology. This deviation from the original plan did not hurt Seidman; it helped her (and Streeten as well). She did not complain about it then, and does not now. We have found nothing in the decree that required Seidman or anyone else to challenge the deviation or that required Brown to halt the original search in the absence of such a challenge. To the contrary, the decree says that Monitoring Committee review " may be initiated at the request of the woman faculty member or prospective woman faculty member involved within sixty days after the final decision was known or after she should have had knowledge of the same, or by any member of the Monitoring Committee." Decree ¶ F(5), reprinted in 491 F. Supp. at 240 (emphasis added). This provision suggests that the university and the committee may overlook 'violations' that do not prompt any complaints.

At the same time, the 'violation' in question is neither obviously serious, nor obviously material, in light of the decree's purposes and objectives. Nothing in the decree commands unbending adherence to the PVA at all costs, nor does each and every deviation — however slight or well-meaning — require Brown to start the lengthy search process over again. See infra Part IV(A). We think the particular deviation at issue, in the absence of anyone's complaining about it, did not automatically render the offer to Streeten unlawful — nor would it have invalidated an offer to Seidman. And, if Brown could lawfully have appointed Seidman, its decision to reopen the search (instead of making her an offer) caused her harm. If sex discrimination led to that decision, Brown unlawfully harmed Professor Seidman. Thus, there is no way to avoid the basic issue in this case: did Brown's decision to reopen the search embody sex discrimination?

IV

We turn now to this basic issue. Professor Seidman says that Brown violated both procedural and substantive aspects of the decree, and that the district court's failure to make the relevant findings requires a remand. Brown, in essence, contends that the record so strongly supports its position that no remand is necessary. Although we agree with parts of Brown's argument, we conclude with Professor Seidman that a remand is necessary in respect to the ultimate issue of pretext.

A. The procedural claim. Professor Seidman's procedural claim arises out of the decree's requirement that the university demonstrate that "the search procedures described in Exhibit E [of the decree] were complied with." Decree ¶ F(4), reprinted in 491 F. Supp. at 240. The relevant language of Exhibit E requires a "written hiring plan . . . approved by the EEO officer before any search takes place." (The entire exhibit is set forth in Appendix B to this opinion.) In this case, the "written hiring plan" includes the PVA. See infra Appendix A. In appellant's view, the university violated the decree's requirements by departing from the PVA.

We have already discussed one deviation from the PVA, namely, the decision to proceed with the original search after the only candidate who qualified for a joint appointment in Economics and Sociology rejected Brown's offer. See supra Part III. This departure worked to Seidman's advantage, not to her detriment, and she does not rely on it as a basis for relief.

Rather, Seidman says that Brown departed from the PVA in a different and important way. She claims that the Sociology Department's express reasons for deciding on March 16 to reopen the search (reasons documented in a March 23 memo from department chairman Alden Speare to the provost) are inconsistent with the terms of the PVA. This deviation, Seidman says, amounts to a 'procedural' violation of Exhibit E of the decree and itself entitles her to relief.

We do not agree. If Seidman means that Sociology's stated reasons were not its true reasons, she is arguing 'pretext' — a 'substantive' question that we shall discuss later. See infra Part IV(B). If she means that the stated reasons — though the true reasons for reopening — nevertheless demonstrate an independent 'procedural' violation of the decree, she is wrong.

To begin with, Exhibit E does not require that the PVA contain a comprehensive list of job criteria, nor does it forbid Brown from revising or refining selection criteria as the search proceeds. The language of the exhibit does not mandate so rigid an approach, see infra Appendix B, and neither does common sense. Filling academic jobs, like many other jobs, often requires selectors to winnow a pool of applicants while simultaneously revising or refining the original selection criteria in light of the applications received — all in an effort to find the candidate of greatest value to the institution. To insist upon beginning the lengthy search process again as the price of any such revision is to create an inflexible, impractical appointments process — one that, as far as we can tell, the decree does not envisage.

Moreover, the appellant overstates the "deviation" between the PVA and the reasons given in the March 23 memo. The memo says Sociology reopened the search because 1) Seidman would not "provide the type of bridge between sociology and economics which was envisioned"; 2) Seidman's methodology was not "distinctive" in respect to that of some sociologists already tenured at Brown; 3) Seidman would not "provide the desired level of recognition for Brown in the area of Development Studies"; and 4) Seidman had not previously held a tenured appointment at "a major university." The first reason falls within the PVA's statement that the Luce chair was created "to further interdisciplinary research and teaching on problems of social and economic development." Brown's decision not to require tenure in the Economics Department did not make the candidate's interdisciplinary skills irrelevant; and understanding of economics and an ability to work with economists remained important.

The other three reasons given in the March 23 memo, although not explicitly listed as valid criteria by the PVA, are nonetheless consistent with the sort of criteria Brown's academic departments ordinarily use in evaluating candidates for tenure. And, nothing in the PVA or the decree purports to preempt reliance on such standard criteria. The criteria ordinarily applied by Brown's Sociology Department — which appear in a document entered in evidence — include "scholarly work at levels of excellence meriting national and international recognition"; "professional service to the Department, to the University, or to the wider community"; and the "program needs of the Department." Joint Appendix vol. PE, at 272-73. The stated reasons for reopening the search lie within the broad scope of these standard criteria.

Thus, we see no independent 'procedural' ground for relief here. Evidence that Brown changed the criteria it applied to Seidman as the selection process progressed, or that it applied different criteria to Seidman than to other candidates, may shed light on the 'substantive' issue of pretext. Yet, if the stated reasons for reopening the search are Brown's true reasons, the mere fact that they do not explicitly appear in the original PVA does not, without more, make the reopening unlawful.

B. The 'substantive' issues and the question of 'pretext.' Our discussion of Professor Seidman's substantive claims brings us back to the factual questions we posed in Part I. See supra p. 535. Seidman has pointed to two suspicious sets of circumstances. Why, she asks, did the provost return her name to the Sociology Department for a formal vote when he had allowed the names of Bardhan and Streeten to proceed to CONFRAT without a formal vote? And what happened, she wonders, at the March 16 Sociology Department meeting at which the faculty decided to reopen the search despite its January 22 consensus to recommend her?

Brown provides lengthy, detailed answers to these questions. It sought to prove the absence of sex discrimination, and the absence of 'pretext,' through evidence and argument that we roughly summarize as follows:

Brown says its faculty, from the outset, noted a striking difference in the qualifications of Bardhan and Streeten, on the one hand, and Seidman, on the other. Bardhan and Streeten had both held tenured positions at "major" universities, published extensively in important journals, and achieved worldwide scholarly recognition. Seidman, while an impressive teacher, had not held comparable faculty positions, lacked a comparable record of publication, and was not well known as a scholar in her field.

Brown adds that, in light of these differences, both the members of the search committee and the tenured members of the Sociology Department were divided from the outset about the merits of appointing Seidman to the Luce chair. Professor Hanson (the Economics Department representative on the committee) made clear that Bardhan was the only one of the committee's three choices whom Economics would approve for an appointment in that department; he added that Brown's economists were significantly more likely to work together with Streeten than with Seidman. Two other committee members, Professor Sidney Goldstein, a population studies sociologist (and former department chairman), and Professor March, an institutional sociologist, opposed giving Seidman full tenure as Luce professor. Seidman's proponents included only Professor Rueschemeyer, the committee's chairman (and the Sociology Department chairman when the Luce search began), and Professor Evans, a younger sociologist, who had been away during many of the committee's meetings. The committee's December 1978 decision to recommend Seidman as its third choice — and only for a three- to five-year, nontenured appointment — represented a compromise. And, Brown says, it is far from clear that the committee would have recommended Seidman at all had it realized in December 1978 that a term appointment was not an acceptable option.

Brown goes on to say that the Sociology Department's January 22 decision to approve Seidman for a fully tenured appointment (should Streeten decline) reflects a greater degree of uncertainty, perhaps confusion, than one might at first suspect. Bardhan's refusal of Brown's first offer forced the search committee to face the possible reality of a Seidman offer. Rueschemeyer then found out, in the week preceding January 22, that the committee could recommend only a tenured appointment; it could not recommend a three- to five-year appointment to the Luce chair. On the morning of January 22, Rueschemeyer and Evans talked to March about offering Seidman (as second choice to Streeten) a fully tenured Luce chair. According to March, Rueschemeyer and Evans put "undue pressure" on him to reverse his opposition to Seidman; March said only that he would go along with whatever the department decided. Rueschemeyer and Evans did not talk to Goldstein, who was away; and Hanson, the economist on the search committee, who was also out of town that morning, had previously said he felt the matter was now Sociology's concern. Later that day, Rueschemeyer presided over a meeting of the tenured members of the Sociology Department. Marsh left the meeting early; the remaining department members then informally acquiesced in a decision to recommend Seidman as third choice for the Luce chair, perhaps without fully realizing the extent of Marsh's and Goldstein's opposition or the significance of Hanson's noncommittal stance.

Brown then points to further events that explain the provost's remand to the department. After the January 22 department meeting, Rueschemeyer wrote a memo to the provost, Maurice Glicksman, the language of which suggested a degree of committee unanimity that did not exist. March found out about the memo on March 14. He and Goldstein (together with Evans) went to Glicksman and explained the divisions of opinion. Glicksman asked what the department vote had been; then, upon learning that the department had taken no formal vote, he sent the matter back for a vote to be taken. The simple reason he sent Seidman's recommendation back for a formal vote (though he had not sent Streeten's) is that, in Seidman's case, he found out about the failure to take the vote and about the faculty divisions; he consequently realized the need for a formal vote to be taken.

Finally, Brown says that the tenured Sociology faculty — at its subsequent March 16 meeting — discussed Seidman's qualifications, debated whether her reputation was adequate for appointment to the Luce chair, took full note of the divisions among the search committee members, and eventually voted, by a 7-2 margin, to reopen the search. This result, in Brown's view, might reflect a change in the persons who attended (Goldstein was back from sabbatical; Rueschemeyer was away); it might reflect certain new information (Professor Barbara Anderson had telephoned several of Seidman's professional 'sources,' seeking added information); it might reflect the faculty's explicit recognition (or decision) that the Luce chair was to be more important than an ordinary tenured position; it might reflect the additional faculty time spent considering Seidman's credentials; it might reflect a fuller and more accurate understanding of the extent to which the search committee members held differing views. But, says Brown, there is nothing in the testimony of those present, nor in the notes made soon after the event, that suggests the decision reflected invidious discrimination based on sex.

The district court, however, did not decide whether Brown's asserted explanation for reopening the search was genuine or pretextual. And, by asking us to affirm the court's judgment nevertheless, Brown, in effect, is arguing that the district court would commit clear error were it to find on remand that Brown has not clearly and convincingly proved its explanation was a true one. With one exception, we are not now willing to say this is so.

The exception concerns Provost Glicksman. On the basis of our reading of the lengthy record, we are willing to say now as a matter of law that Brown has shown (by clear and convincing evidence) that his decision to return the Seidman recommendation to the Sociology Department did not reflect sex discrimination. The reason that Brown offers, at least in this respect, is not a pretext. It is conceded that Goldstein and Marsh leaned fairly strongly against giving the fully tenured Luce chair to Seidman; Hanson was neutral because he felt the matter was up to Sociology. Rueschemeyer and Evans were in favor. Goldstein was away when the department discussed the matter on January 22, and Marsh missed part of the meeting. Rueschemeyer, in forwarding the department's recommendation to the provost, explained none of this. Instead, he simply wrote: "This recommendation has been discussed with, and has the support of, the Selection Committee, the Department of Sociology, and the Tenure Committee of Sociology." Marsh read the memo containing this sentence sometime in March. (There is no evidence to the contrary.) He then went, with Goldstein and Evans, to Glicksman, and explained that the Sociology faculty was in fact divided — that any impression of unanimity that the memo provides was misleading. At that point Glicksman asked what the vote had been; and he discovered no formal vote had been taken. Rather than send the recommendation on to CONFRAT, he sent the recommendation back to the department for a vote.

The obvious, and given, explanation for Glicksman's action is both that university rules required a vote, and that there was reason to believe some degree of confusion had infected the earlier faculty deliberations. A formal vote would require each member of the faculty to face up to and decide the question, thereby allowing Glicksman and CONFRAT to gauge the degree to which the earlier consensus decision really reflected solid departmental support for Seidman. The obvious, and given, explanation for Glicksman's having failed to send Streeten's recommendation back for a formal vote is that no one called the lack of a formal vote to his attention (evidently because the faculty was not divided, and there was no particular need to find out just where each member stood). There is no significant evidence (indeed, we could find no evidence at all) suggesting that these given explanations were not the real ones. And, on the basis of our reading of the record, we think a fact finder who found the reasons were pretextual would be clearly wrong. That is to say, a finding that sex discrimination motivated Glicksman's decision would not be "plausible in light of the record viewed in its entirety." Anderson v. City of Bessemer City, 470 U.S. 564, 574, 105 S.Ct. 1504, 1512, 84 L.Ed.2d 518 (1985).

However, we are not now willing to decide — without an initial district court decision — the key factual question about the Sociology Department's decision on March 16 to reopen the search. Brown has presented witnesses, testimony, and documents offering a legitimate, nondiscriminatory explanation of this event. Professor Seidman has presented evidence of her own, including one witness, Professor Evans, who testified that Professor March opposed Seidman because of "attitudes toward women, with the sex of the candidate." The district court, which has heard the witnesses and can better evaluate their credibility, should decide in the first instance whether or not Brown has shown that its explanation is not a pretext. In doing so, the court should decide whether Brown has articulated legitimate reasons for reopening the search, and has proved clearly and convincingly that those reasons are the true ones. At the same time, it must take care not to interpret the decree in a way that would impose an unreasonably difficult burden of proof upon Brown. See supra pp. 536-37. The district court may decide this case on the existing record should it believe it appropriate to do so.

The judgment of the district court is vacated, and this case is remanded for further proceedings consistent with this opinion.

APPENDIX A

PERSONNEL VACANCY AUTHORIZATION

APPENDIX B

EXHIBIT E

HIRING