COFFIN, Chief Judge.

This is the second appeal in a sex discrimination case from the trial court's ruling against plaintiff-appellant on the merits. The basic facts of the case appear in Blizard v. Fielding, 572 F.2d 13 (1st Cir. 1978). The core of appellant's complaint is that she was not promoted to the position of Deputy Commissioner of Public Health of the State of Massachusetts because of her sex and that she was retaliated against when she complained of this discrimination to state and federal authorities. Appellant now challenges the trial judge's refusal to recuse himself under 28 U.S.C. § 455(a) from hearing the case on remand. Appellant argues further that a new trial was required by our first opinion in this case and raises for a second time several evidentiary points not decided in the first appeal. We affirm the judgment below.

This case was tried fully to the court. In an unpublished opinion, the trial court ruled for the defendants, finding that any adverse employment decisions by the defendants were not sex-based and were grounded upon legitimate, nonpretextual justifications. The court found further that any adverse changes in plaintiff's work assignments and working conditions were not in retaliation for plaintiff's assertion of sex discrimination claims before state and federal agencies. The court did not, however, apply the analytical framework of McDonnell Douglas v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973), to the evidence in the case, failing to determine whether or not plaintiff had made out a prima facie case of sex discrimination. We remanded for application of the McDonnell Douglas standards.

On remand, plaintiff-appellant moved that the trial judge, the Honorable Andrew A. Caffrey, disqualify himself from further consideration of the case because "his impartiality might reasonably be questioned." 28 U.S.C. § 455(a). Appellant also moved for a new trial. In an unpublished opinion, the district court denied both motions and held that although plaintiff-appellant had made out a prima facie case under McDonnell Douglas, defendants had carried their burden of demonstrating legitimate, nonpretextual reasons for their employment decisions. The district court went on to reiterate and more fully explain its original holding that the changes in appellant's work assignments could not fairly be characterized as retaliatory because appellant substantially failed to perform her assigned duties and therefore received assignments of lesser significance, or none at all. Finally, it held that no retaliatory motive prompted the change in appellant's office location and the temporary deprivation of a secretary, both problems being a part of a restructuring necessitated by physical and fiscal restraints on department resources.

The Section 455(a) Recusal Issue

Appellant bifurcates her attack on the district judge's resolution of the recusal issue into an effort to show that he applied the wrong legal standards and a lengthy attempt to demonstrate that application of the correct legal standards would require recusal. We are not convinced by either argument.

The court began its consideration by setting forth our interpretation of the new 28 U.S.C. § 455(a), as propounded in United States v. Cepeda Penes, 577 F.2d 754 (1st Cir. 1978) and United States v. Cowden, 545 F.2d 257 (1st Cir. 1976). In essence, section 455(a) allows a judge to disqualify himself if a reasonable man would have factual grounds to doubt the impartiality of the court. Such a standard allows recusal when objective appearances provide a factual basis to doubt impartiality, even though the judge himself may subjectively be confident of his ability to be evenhanded. As the court recognized, Cepeda Penes also stands for the proposition that a judge's conduct during prior judicial involvement in a case (as opposed to extrajudicial knowledge of the parties or evidence) can conceivably provide a factual basis for doubting impartiality. Finally, the new section 455(a) was expressly intended to do away with the so-called "duty to sit" doctrine, which required a judge to hear a case unless a clear demonstration of extrajudicial bias or prejudice was made. Compare In re Union Leader Corp., 292 F.2d 381, 391 (1st Cir. 1961) with H.Rep. No. 93-1453, 93d Cong., 2d Sess., reprinted in, U.S. Code Cong. Admin.News, pp. 6351, 6355.

Appellant does not quibble with the judge's thorough explanation of the ground rules laid out in section 455(a). Rather, appellant takes issue minutely with his application of those rules, arguing that the language of his opinion demonstrates that he used a different, and erroneous legal standard. First, appellant objects to the trial court's vehement denial of extrajudicial knowledge of appellant. Contrary to the inference drawn by appellant, we do not read this statement as basing the final decision solely on the lack of extrajudicial knowledge of appellant. The trial court's opinion does not state that a finding of bias is precluded because there is no demonstration of extrajudicial knowledge of plaintiff. It does not state that knowledge and attitudes acquired solely in judicial proceedings can never be a basis for disqualification. Rather, appellant herself raised the issue of extrajudicial knowledge in her brief supporting the motion to recuse, and the court was careful to cover that point as one of several raised.

Second, appellant argues that the judge's recitation of record support for all of his less-than-complimentary findings of fact demonstrates a subjective, self-justifying approach to the section 455(a) issue where an objective evaluation of the appearance created by his prior words and acts was called for. We disagree. As a finder of fact in a Title VII case, a trial judge must give his candid evaluation of plaintiff's case. If a case is weak to the point of frivolousness, it is appropriate to say so. A reasonable man evaluating the possible partiality of the judge-as-fact-finder would rationally look to the record to see if the judge's candid evaluation is fully supported. We see nothing "subjective" about objectively evaluating whether or not the content and tone of earlier findings have a solid basis in the record.

Appellant's final argument concerning legal standards faults the court's quotation of Union Leader, to the effect that "[t]here is as much obligation upon a judge not to recuse himself when there is no occasion as there is for him to do so when there is." In re Union Leader, supra, 292 F.2d at 391. Appellant finds in this quotation a clear invocation of the "duty to sit" doctrine. We agree that this quotation is drawn from a case applying the rigorous standards for disqualification found in 28 U.S.C. § 144. We do not agree, however, with appellant's conclusion that the use of this quotation demonstrates the application of an erroneous legal standard. Appellant reasons that because section 455(a) allows recusal if there are reasonable factual grounds to doubt a judge's impartiality, the burden is on the trial court to show impartiality beyond a reasonable doubt. Appellant goes so far as to imply an analogy between the language of the standard and its corollary in criminal procedure: whereas the "duty to sit" doctrine created a presumption against disqualification, the "reasonable doubt" standard creates a presumption of disqualification.

We do not think that the impartiality of every federal judge must be demonstrated beyond a reasonable doubt in the usual sense of that phrase. A trial judge must hear cases unless some reasonable factual basis to doubt the impartiality or fairness of the tribunal is shown by some kind of probative evidence. See, e.g., Fredonia Broadcasting Corp. v. RCA Corp., 569 F.2d 251 (5th Cir. 1978) (trial judge's former law clerk appears as advocate in trial on remand of case originally heard while clerk worked for trial judge; appearance of unfairness requires recusal). In this sense, i. e., that judges hear cases unless there is some reason not to, the "duty to sit" remains. It was in this sense that the court used the quote from Union Leader, after thoroughly reviewing all of appellant's arguments.

We turn next to the specific evidence of partiality cited by appellant. It seems clear that our review is limited to determining whether the trial court's evaluation of that evidence amounted to an abuse of discretion. Mayberry v. Maroney, 558 F.2d 1159, 1162 (3d Cir. 1977); Davis v. Board of Schl. Comm'rs, 517 F.2d 1044, 1052 (5th Cir.), cert. denied, 425 U.S. 944, 96 S.Ct. 1685, 48 L.Ed.2d 188 (1976); H.Rep. No. 93-1453, supra, U.S. Code Cong. Admin.News, p. 6355. We cannot say there was any such abuse here.

Appellant relies in chief upon three less than complimentary characterizations of appellant and her case included in the district court's first opinion. Even reading words out of context, as appellant presents them to us, we cannot fault the district court's conclusion that they show no factual support for a reasonable inference of partiality. The record supports the conclusions reached by the district court so fully that we cannot find any inference of partiality in the court's choice to express its decision in no uncertain terms. Perhaps more important, appellant's quibble with its choice of words focuses solely on the two concluding paragraphs of a lengthy opinion that thoroughly disposes of appellant's claims in a fair and neutral evaluation of the evidence. If a concluding paragraph using colorful language to drive home a point proves an entire opinion biased, then few, if any judicial opinions pass muster under section 455(a). We do not hold that words alone can never provide a factual basis to doubt impartiality; but these words, while enough to fuel an argument, do not suffice.

Appellant's other evidence of partiality boils down to sniping at the court's evidentiary rulings, occasional interrogation of appellant on the stand, and alleged hostility toward appellant. We have reviewed the trial transcript and find nothing to support appellant's arguments. The judge initially allowed appellant to ramble in her testimony and to answer the unasked. It soon became clear, however, that appellant had confused her role as a witness with her attorney's role as an advocate. At that point, it was not improper to admonish appellant to confine her answers. Only once did the judge interrogate appellant at any length, and his attempts to get a straight answer demonstrate, at worst, incredulity and not hostility. The right of a trial court, sitting as a trier of fact, to interrogate witnesses to clarify the record needs no citation. And, of course, adverse evidentiary rulings do not, by themselves, demonstrate bias. United States v. Schwartz, 535 F.2d 160, 163 (2d Cir. 1976).

In sum, the denial of the motion to recuse was not an abuse of discretion.

The Motion for a New Trial

Appellant argues that the trial court could not comply with the mandate of our remand of this case without a complete retrial. Appellant argues that the opinion in Blizard v. Fielding, supra, finds such fault with the original consideration of this case that retrial was unavoidable. We do not agree.

Our first opinion in this case expressed concern that failure to apply the McDonnell Douglas analysis might have short-circuited some important factual considerations. We suggested that the existence of a prima facie case might affect the district court's evaluation of the proffered justification for not promoting appellant. We also left it to the district court to consider the importance of its evidentiary rulings and its findings on retaliation in light of the prima facie case issue. We do not think that any of these concerns required more than an exercise of the district court's discretion in determining whether or not a new trial was needed.

As the Supreme Court has recently explained, "[t]he method suggested in McDonnell Douglas for pursuing [the issue of discriminatory action] was never intended to be rigid, mechanized, or ritualistic. Rather, it is merely a sensible, orderly way to evaluate the evidence in light of common experience as it bears on the critical question of discrimination." Furnco Constr. Corp. v. Waters, 438 U.S. 567, 577, 98 S.Ct. 2943, 2949, 57 L.Ed.2d 957 (1978). The district court's second opinion demonstrates just such an orderly evaluation of the evidence, finding that appellant had indeed shown a prima facie case. More important, the issue that was fully joined at trial — defendants-appellees' legitimate reasons for adverse action — could be fairly resolved on the record in defendants' favor, thereby negating any inference founded upon the existence of a prima facie case.

Appellant argues that McDonnell Douglas requires not just a method of analysis but also a method of receiving and responding to proof at trial. We do not think this contention can stand in the light of Furnco. The first step in McDonnell Douglas analysis, consideration of a prima facie case, has ended with all agreeing that the basis for an inference of discrimination exists. Defendants went forward with the second step, showing legitimate reasons for their actions. The district court found that there was no evidence showing the proffered reasons were pretextual, and plaintiff-appellant has not convinced us that any further evidence exists on the issue. The third step is thus completed. We see no reason to rehear the same evidence.

This analysis also disposes of appellant's argument based upon Corley v. Jackson Police Dep't, 566 F.2d 994 (5th Cir. 1978). In Corley, the district court failed to take the third step by failing to consider evidence that discharged black officers were disciplined for taking bribes while white officers were not, i. e., that the legitimate reason was a pretext. Nor do we find appellant's analogies to the confused legal standards and burdens of proof applied by the district court in Corley to be apt. The only "confusion" that existed in this case was the trial court's extraneous consideration of the age discrimination claim that appellant filed with the Massachusetts Commission Against Discrimination (MCAD) and pleaded in federal court but did not press at trial. The district court's second opinion ended all concern about such surplusage.

Finally, no new trial was needed to hear evidence excluded at the first trial. As noted below, we believe the trial court's evidentiary rulings were within its discretion. Moreover, insofar as our first opinion asked the trial court to reassess its evidentiary rulings in the light of its overall reconsideration of the case, a full offer of proof was available in the record of the first trial, making a new hearing unnecessary to assess the value of possible additional evidence.

Evidentiary Issues

Appellant argues that the trial court committed reversible error by excluding testimony on two issues, the alleged sex bias of a prior Commissioner of Public Health and statistical evidence concerning the number of women in high positions in the Department. Appellant also attacks as clearly erroneous the district court's conclusions on the issue of retaliation.

The first question deserves short shrift. Appellant sued Commissioner Bicknell and his replacement, Commissioner Fielding, alleging continuing discrimination from the time that Bicknell first refused to appoint appellant as his Deputy. Appellant attempted to introduce evidence that Bicknell's predecessor, Commissioner Frechette, had rebuffed a recommendation that he promote appellant by saying: "Oh, no. She's a woman." The trial court ruled that the comment of a prior administrator who admittedly had no control over Bicknell had no probative value concerning Bicknell's activities. Were we squarely faced with the issue, we might find such a ruling within the trial court's discretion, but we are not. The trial court later admitted the comment into evidence as it related to another recommendation for another position. Even if Frechette's comment was relevant to his successor's motivations, a proposition we doubt considering the substantial evidence of Bicknell's independence of thought about administrative organization and appointment of substantial numbers, of women to high positions, we can see no error in refusing to hear it repeatedly.

Appellant's second point has some merit, but does not present reversible error. Appellant offered to show that three MCAD investigators visited the Department some time in 1972 and spent three days counting the number of female employees in each of the salary levels of the Department hierarchy. They found a higher percentage of female employees in the lower ranks than in high, policymaking positions. Because the ultimate issue in this case was whether or not the defendants' proffered justifications were pretextual, statistical proof of a pattern of discrimination could have been highly relevant. See McDonnell Douglas, supra, 411 U.S. at 805, 93 S.Ct. 1817. But to be relevant, evidence must have some probative value. Appellant's offer ignored the fact that at the time of the study Bicknell, whose alleged discriminatory animus was at the core of the case and was the main focus of appellant's evidence, had appointed only three out of thousands of Department employees. More important, appellant's offer was in no way related to the available pool of qualified female applicants for the positions filled predominantly by males. See Hazelwood Schl. Dist. v. United States, 433 U.S. 299, 97 S.Ct. 2736, 53 L.Ed.2d 768 (1977). Finally, even if the trial court abused its discretion in excluding such questionable evidence, we would not find the error harmful. The record is replete with uncontradicted evidence that Bicknell appointed large numbers of women to high level positions on his staff. The numbers game simply would not work here.

Appellant's final challenge paints a picture of a massive scheme starting in 1972, when she was first denied promotion, to paper the record against her and provide fallacious justifications for reprisals. She argues that the trial court was clearly erroneous in failing to pierce the paper wall. The trial judge, however, made both specific and general findings concerning the inadequacy of appellant's performance and appellees' reasonable, nondiscriminatory response thereto. Those findings are not clearly erroneous.

In support of the trial court's general findings, the record reveals that appellant had reached a transition stage in her career, concluding a number of substantial assignments, contemporaneous with the appointment of a new Commissioner. She was due for new tasks, and her new superior sought to fathom her abilities and her proper role in his new administration by giving her significant new tasks. Appellant's approach to meetings with him was to begin each time with the topic of her right to the position of Deputy Commissioner, the number two position. As it turned out, the new Commissioner found it more efficient to have a number of department heads reporting directly to him rather than employing an administrative chief of staff.

Undaunted by the Commissioner's right to organize his administration as he saw fit, appellant continued her demands. The trial judge specifically found, on an adequate record, that appellant responded to a request for a description of her former and present responsibilities by stapling together a mass of brochures of organizations with which she was associated. On this record, the trial court found that, as a new Commissioner, Bicknell had two reasons not to appoint appellant: he did not wish to fill the position at all and appellant stymied his attempts to discern her qualifications.

Nevertheless, appellant was given work to do. The trial court correctly found that appellant, who is an attorney, failed to conclude an assignment to compile a manual of laws relevant to the Department's operation, submitting an interim report that left important questions undecided and unanswered. She also failed to complete legal assignments relating to the Hill-Burton Program in Massachusetts and to the extent of the legal authority of the Commissioner. We find no support in the record for appellant's assertion that these assignments were meaningless or represented a demotion.

Appellant faults the trial court's focus upon a 1972 weekly report to the Commissioner in which appellant explained that she was spending much of her time working on her discrimination claims. Appellant argues, and we agree, that she had a right to spend the business hours necessary to assist MCAD and the EEOC in processing her complaints. We do not believe, however, that the trial judge gauged her performance from 1972 until 1977 based upon one weekly report. Rather, the record reveals and the trial court found that in a period of eight to nine months appellant made virtually no progress on her chief assignment, a review of pending bills on national health insurance. Someone else did the job. Finally, when offered another assignment commensurate with her responsibilities, appellant declined on the ground that she would be going on vacation in three weeks and needed to train a new secretary.

On the question of retaliation by depriving appellant of support staff, the record reveals that appellant refused to use a newly instituted dictaphone/typing pool system, although the necessary equipment was provided her. Nor are we moved by appellant's complaint that her office was shifted out of the mainstream of department affairs. The trial court found that there was no causal relation between moving appellant's office and her filing of civil rights claims. The record reveals that appellant was first moved into a suite housing many of the lawyers in the department, a logical shift. Unfortunately, some of her fellow attorneys became involved in defending appellant's suit against the Department. There was uncontroverted testimony that appellant's use of the office suite put her within arm's length of written work of those attorneys. The appearance, if not the fact of impropriety involved in that proximity was the cause of the second office change, not discrimination.

Affirmed.