No. 471, Docket 79-3070.United States Court of Appeals, Second Circuit.Argued October 16, 1979.
Decided February 25, 1980.
Thomas D. Barr, New York City (Cravath, Swaine Moore, New York City, of counsel), for petitioner-defendant.
John J. Powers, III, Washington, D.C. (Dept. of Justice, John H. Shenefield, Asst. Atty. Gen., Bruce E. Fein, Washington, D. C., and Robert J. Staal, Don A. Resnikoff, Mark W. Gaffney, John P. Hannigan, Dept. Of Justice, New York City, of counsel), for respondent-plaintiff.
Before MULLIGAN, VAN GRAAFEILAND and MESKILL, Circuit Judges.
MULLIGAN, Circuit Judge:
 More than a decade ago, on January 17, 1969, the United States of America, by its attorneys, acting under the direction of the Attorney General, filed a complaint in the United States District Court for the Southern District of New York which alleged that International Business Machines Corporation (IBM), commencing in or about 1961, had monopolized and attempted to monopolize the market for general purpose
electronic digital computers in violation of Section 2 of the Sherman Act (15 U.S.C. § 2). In addition to injunctive relief the Government sought such “divorcement, divestiture and reorganization” of IBM as might be appropriate to restore competitive conditions. On January 26, 1972, Hon. David N. Edelstein, Chief Judge of the Southern District, assumed control of the case. After extensive pretrial discovery, the bench trial was commenced on May 19, 1975. Counsel for the Government estimated that its case would take two to three months and IBM’s counsel predicted that its defense would take six to eight months.
 These estimates in fact proved to be grossly erroneous. The Government’s direct case lasted for almost three years, ending on April 26, 1978. IBM’s defense began on that date and it continues as of this writing. Eleven years have elapsed since the filing of the initial complaint, pre-trial depositions commenced some eight years ago, and more than four and one-half years of trial time have been consumed. A mammoth record of trial transcript and exhibits has been assembled. To the best of our knowledge no litigation has taken so much time and involved such expense.
 On July 19, 1979, IBM filed an application requesting Chief Judge Edelstein to recuse himself on the grounds that he has a “personal bias and prejudice against IBM and in favor of plaintiff, that his impartiality in this action may reasonably be questioned, that he has a bent of mind that will prevent impartiality of judgment, and that his bias and prejudice could not have come from any source other than an extrajudicial source.” IBM therefore urged his recusal under 28 U.S.C. §§ 144, 455 and the due process clause of the Fifth Amendment. In addition IBM argued that resumption of the trial before a new judge would be inappropriate until the record has been “purged of the effects of the Chief Judge’s bias.”
 On September 11, 1979, Chief Judge Edelstein filed a written opinion in which he denied the request for recusal as both untimely and legally insufficient. In his opinion he stated, “This court has no interest in the outcome of this case other than the interest of every judicial officer that the truth be discovered and the law correctly applied. IBM identifies no personal prejudice and bias on the part of this court because none exists.” On September 12, 1979 the trial judge denied IBM’s oral application for a stay of all proceedings for 30 days so that IBM might file a Petition for a Writ of Mandamus. On September 13, 1979 IBM filed in this court a Petition for a Writ of Mandamus pursuant to 28 U.S.C. § 1651 and Fed.R.App.Pr. 21 together with a motion for a stay of all proceedings pending our determination of the merits of the petition. Oral argument on the motion for a stay was heard on September 14, 1979 before this panel and it was denied from the bench. Argument on the merits of the petition was heard in this court on October 16, 1979. As of that time more than 90,000 pages of testimony had been transcribed, almost 9,000 documents had been received into evidence, several hundred witnesses deposed, and some seventy trial witnesses called. In the argument counsel for IBM suggested that the case might take five more years of trial time to complete, putting the case into 1984, an appropriately Orwellian denouement. Because the litigation has lasted so tortuously long and is costing taxpayers of the United States as well as the stockholders of IBM untold millions of dollars, to say nothing of the continuing toll of time and effort made upon the federal judiciary, we have requested the
parties to conduct settlement discussions. Conferences have been held, as reported in the press, and this court has been kept apprised by the parties of their progress. Although we have no way of predicting the ultimate outcome of this prolonged litigation, it does seem that its settlement would be in the best interests of all those concerned. Yet it seems fair to state that while there is reason to be somewhat optimistic so long as the parties are meeting, there is little reason to expect a settlement in the immediate future and no reason to further delay this opinion.
 It is clear to this panel that the issuance of the extraordinary writ of mandamus to remove Chief Judge Edelstein from the trial of the case on the ground of personal bias, as set forth in the petition here, is not the appropriate vehicle to relieve IBM of the burdens it sets forth in its papers. IBM complains in gist of erroneous rulings and judicial mismanagement of the trial. Our task here is not to decide this case on its merits or to provide a critique of its conduct but rather to determine whether the petitioner has met the burden of establishing that Chief Judge Edelstein is unfit to continue by reason of his personal bias against IBM. We hold that the burden has not been satisfied.
 The United States makes the threshold argument that mandamus does not lie here because under Will v. Calvert Fire Ins. Co., 437 U.S. 655, 665-66, 98 S.Ct. 2552, 2559, 57 L.Ed.2d 504 (1978), where a matter is committed to the discretion of the district court, a petitioner cannot meet the “clear and indisputable” right test which is a condition precedent to the issuance of the extraordinary relief sought in mandamus. The Government argues that the findings of untimeliness and insufficiency made below are discretionary and therefore preclude the writ which IBM would have us issue. We do not agree. Calvert Fire Insurance was not a case involving recusal but rather dealt with the determination of a trial judge to defer federal proceedings until the termination of concurrent state litigation. This the Court found was a matter within the discretion of the district court. “Where a matter is committed to the discretion of a district court, it cannot be said that a litigant’s right to a particular result is `clear and indisputable.'” Id.
 The question here is not whether the trial judge has abused his discretion but whether he could exercise any discretion because of a personal, extrajudicial bias which precludes dispassionate judgment. This presents a distinguishable issue. This court has long since taken the position that there are “few situations more appropriate for mandamus than a judge’s clearly wrongful refusal to disqualify himself.” Rosen v. Sugarman, 357 F.2d 794, 797 (1966). We do not read Calvert Fire Insurance Co. to overrul sub silentio this well established rule, followed by the great majority of other circuits as well. A claim of
personal bias and prejudice strikes at the integrity of the judicial process, and it would be intolerable to hold that the disclaimer of prejudice by the very jurist who is accused of harboring it should itself terminate the inquiry until an ultimate appeal on the merits. We conclude that we have the power to issue the writ but only if the petitioner has satisfied the burden of establishing that its right is “clear and indisputable.”
 On July 19, 1979 five distinguished members of the Board of Directors of IBM (William T. Coleman, Jr., Carla Anderson Hills, John N. Irwin II, Wiliam W. Scranton and Irving S. Shapiro), acting as a special committee of the Board, filed an affidavit stating their belief, as noted above, that Chief Judge Edelstein “has a personal bias and prejudice against IBM and in favor of plaintiff, that his impartiality in this action may reasonably be questioned, that he has a bent of mind that will prevent impartiality of judgment, and that his bias and prejudice could not have come from any source other than an extrajudicial source.” The committee requested his recusal and the assignment of another judge to hear the case. The affidavit was filed pursuant to the Fifth Amendment of the United States Constitution, and Sections 144 and 455 of Title 28, United States Code. The action of the committee was based upon the affidavits of IBM trial counsel and certain IBM witnesses and the appendices attached thereto which include some 2500 pages of charges. The Government has responded with its own appendix of approximately 550 pages and IBM has replied with another appendix of 250 pages. In substance, IBM’s charge of bias and prejudice is based upon the following allegations:
1) The Chief Judge has become an advocate for the plaintiff, disrupting the direct testimony of IBM’s witnesses, cross-examining IBM witnesses himself, shielding plaintiff’s witnesses from cross examination, and routinely abusing IBM’s witnesses and counsel.
2) His rulings on questions of evidence, procedure and discovery have been inconsistent and the general conduct of the trial has been in favor of the Government and against IBM.
3) He has expressed an antipathy toward IBM counsel which manifests an attitude that they can do no right.
4) He has deliberately attempted to create a record that cannot be subject to full and adequate appellate review by refusing to file papers submitted by IBM, refusing to hand down timely rulings in matters of importance to IBM, and secretly making alterations in and deletions from the trial transcript.
 IBM’s petition recognizes that the alleged prejudice of the trial judge must be extrajudicial, that it must arise by virtue of some factor which creates partiality arising outside of the events which occur in the trial itself. The Supreme Court has held, “The alleged bias and prejudice to be disqualifying must stem from an extrajudicial source and result in an opinion on the merits on some basis other than what the judge has learned from his participation in the case.” United States v. Grinnell Corp., 384 U.S. 563, 583, 86 S.Ct. 1698, 1710, 16 L.Ed.2d 778 (1966). Section 144 in haec verbs requires that the party’s affidavit of prejudice state that the judge before whom the matter is pending have a “personal bias or prejudice either against him or in favor of any adverse party.” The mandamus petition is also based on section 455. Section
455(b) lists five instances which mandate the recusal of a judge. The only relevant subdivision is section 455(b)(1), which requires recusal where the judge “has a personal bias or prejudice concerning a party . . .” This language is taken directly from section 144. We agree with the Fifth Circuit’s holding in Davis v. Board of School Commissioners, 517 F.2d 1044, 1052 (5th Cir. 1975), cert. denied, 425 U.S. 944, 96 S.Ct. 1685, 48 L.Ed.2d 188 (1976), that, “[c]onstruing §§ 144 and 455 in pari materia we believe the test is the same under both. . The determination should be made on the basis of conduct extrajudicial in nature as distinguished from conduct within a judicial context.”
 IBM has not shown and does not purport to establish or identify any personal connection, relationship or extrajudicial incident which accounts for the alleged personal animus of the trial judge. IBM’s claim of prejudice is based completely on Chief Judge Edelstein’s conduct and rulings in the case at hand. These we have repeatedly held form no basis for a finding of extrajudicial bias. Thus in King v. United States, 576 F.2d 432, 437 (2d Cir. 1978), we stated:
“The grounds urged for disqualification are for the most part rulings made by [the trial judge] during the trial or statements made by him in the course of his judicial duties. . . . Nothing of this kind, what the judge has learned from or done in the proceedings before him, is any basis for disqualification; to be sufficient for disqualification the alleged bias
or prejudice must be from an extrajudicial source.”
 See also United States v. Bernstein, 533 F.2d 775, 784-785 (2d Cir.), cert. denied, 429 U.S. 998, 97 S.Ct. 523, 50 L.Ed.2d 608 (1976); United States v. Sclafani, 487 F.2d 245, 255 (2d Cir.), cert. denied, 414 U.S. 1023, 94 S.Ct. 445, 38 L.Ed.2d 313 (1973).
 IBM urges that disqualification is also required under section 455(a). Section 455(a) provides for recusal where the judge’s “impartiality might reasonably be questioned.” This section sets up an objective standard for recusal, creating the socalled “appearance of justice” rule. Thus it may well be, for example, that if a judge’s first cousin is a party to a case and no disqualification arises under section 455(b)(5) since he is not within the third degree of kinship, reasonable men might well question his impartiality where a close personal relationship exists between the two. This is an instance posed in Wright, Miller Cooper, Federal Practice Procedure § 3549 (1975). However, we cannot agree that adverse rulings by a judge can per se create the appearance of bias under section 455(a). A trial judge must be free to make rulings on the merits without the apprehension that if he makes a disproportionate number in favor of one litigant, he may have created the impression of bias. Judicial independence cannot be subservient to a statistical study of the calls he has made during the contest. As Mr. Justice Frankfurter noted in Wilkerson v. McCarthy, 336 U.S. 53, 65, 69 S.Ct. 413, 419, 93 L.Ed. 497 (1949) (Frankfurter, J., concurring), “A timid judge, like a biased judge, is intrinsically a lawless judge.” We conclude that under section 455(a) the bias to be established must be extrajudicial and not based upon in-court rulings.
 A reading of the cases supporting the general proposition that the bias which requires recusal must be personal and cannot rest upon trial rulings or conduct, reveals two practical considerations which have influenced the courts. The first is quite obvious. As the Supreme Court noted in Ex parte American Steel Barrel Co., 230 U.S. 35, 44, 33 S.Ct. 1007, 1010, 57 L.Ed.2d 1379 (1913), “[The recusal statute] was never intended to enable a litigant to oust a judge for adverse rulings made, for such rulings are reviewable otherwise. . . .” This argument was repeated in Berger v. United States, 255 U.S. 22, 31, 41 S.Ct. 230, 232, 65 L.Ed. 481 (1921).
 IBM claims that 86% of the some 10,000 oral motions made and 74 out of 79 written motions have been decided against the petitioner and in favor of the Government. The United States counters that some 70% of these motions were made by IBM, which allegedly has adopted an increasingly hostile and uncooperative trial strategy, making unnecessary or frivolous objections for the purpose of creating the very statistics upon which it now depends. IBM would have us contrast the favorable rulings made in the Government’s favor in its presentation with those which it claims were made inconsistently against IBM in its defense. The Government, asserting that IBM’s statistics are flawed, claims that the IBM appendices refer only to a small fraction of the motions made and that no pattern of bias emerges. IBM has made further computations which refute those of the Government and allegedly re-establish that a disproportionate number of rulings have been adverse to its interest.
 It seems evident that statistics alone, no matter how computed, cannot establish extrajudicial bias. There is no authority for, and no logic in, assuming that either party to a litigation is entitled to a certain percentage of favorable decisions. The inquiry to be at all meaningful would necessarily require this court to examine each and every ruling to determine whether it was, initially, legally valid. If we determined that some adverse rulings were correctly made, obviously they could not be tainted by bias. Even if they were deemed to be incorrect, it of course does not follow that they were motivated by personal bias. We would next have to ask whether the error could be attributed to the judge’s misunderstanding of the facts or the law. The exercise would require this court to become intimately familiar with a 90,000 page trial transcript and to examine thousands of underlying documents and exhibits. That in turn would require us to become familiar with the highly technical and complicated engineering issues involved. There is no precedent for such an undertaking on this scale at this interim stage of the litigation. The exercise is futile in any event, since it is impossible from the record before us to ascribe extrajudicial animus to an evidentiary ruling. There is no ruling or comment by the judge which gives any clue or inkling of extrajudicial bias against IBM. If material legal or factual error has been committed it can be dealt with on plenary appeal.
 This brings us to the second policy consideration underlying the rule that the bias necessary for recusal must be extrajudicial and not based upon what the judge has learned in this case. Chief Judge Edelstein is the sole finder of the fact here. His role is not that of a passive observer. His obligation is to determine the facts in a field which is exceedingly complex and technical. His function was well described by Judge Frank in In re J. P. Linahan, Inc., 138 F.2d 650, 653-54 (2d Cir. 1943):
“But, just because his fact-finding is based on his estimates of the witnesses, of their reliability as reporters of what they saw and heard, it is his duty, while listening to and watching them, to form attitudes towards them. He must do his best to ascertain their motives, their biases, their dominating passions and interests, for only so can he judge of the accuracy of their narrations. He must also shrewdly observe the stratagems of the opposing lawyers, perceive their efforts to sway him by appeals to his predilections. He must cannily penetrate through the surface of their remarks to their real purposes and motives. He has an official obligation to become prejudiced in that sense. Impartiality is not gullibility. Disinterestedness does not make child-like innocence. If the judge did not form judgments of the actors in those court-house dramas called trials, he could never render decisions.”
 The point of this comment is perhaps best illustrated by IBM’s complaint that Chief Judge Edelstein’s treatment of its witnesses and its counsel has been marked by asperity, incivility and hostility. We have examined the affidavits of the nine IBM witnesses, all of whom state that his conduct demonstrates an unidentified personal bias. Much of this is not discernible from the record. Many complain of his “stares,” “glares” and “scowls.” One urges that the Chief Judge “would often glower at me, scowl and then turn and make bold notes on a pad of paper, as if threatening me with the contents.”
 There is no question but that these witnesses have felt that the judge’s conduct was intimidating. However, the burden of
IBM is to establish clearly and convincingly that his attitude can only be attributed to his personal prejudice. As Judge Frank indicated, supra, the trial judge has the obligation to form judgments as to the veracity of the witnesses before him. His asperity and incivility may well be due to his feeling that a witness is not forthright, that he is trying to protect a position, or is otherwise attempting to obfuscate the fact finding process. IBM complains, for example, that Chief Judge Edelstein has asked its witnesses too many questions, has required them to draw meaningless charts and diagrams, to answer questions “yes” or “no” when they are not susceptible to such simplistic solutions. He has, we are assured, interrupted plaintiff’s 52 witnesses “only 846 times and has interrupted IBM’s first 19 witnesses over 1200 times.” Accepting all of these contentions at face value we do not find them to be of the stuff upon which one can sensibly premise extrajudicial bias.
 What may be a simple technical issue to the expert witness or even to IBM counsel, who have been given technical training in preparation for this litigation, may well be arcane to the jurist. His questioning, his interruption, his insistence on clarification may well be prompted by his struggle to determine the truth in a field in which he is not sophisticated. His asperity may well be prompted by a feeling that the witnesses for IBM (three are long time employees) are dissembling. We do not know and cannot on this record determine whether his conduct has been guided by what he has learned during the trial, in which case his reaction is licit, or whether it is due to a personal prejudice which is clearly impermissible. The point is that IBM has not met its burden of showing that the Chief Judge is personally biased against the petitioner.
 There is of course another factor at play here — the seemingly interminable length of the trial. The IBM witnesses did not begin to testify until almost three years of trial time had elapsed. Even the most stoic might well lose patience in these circumstances. The judge’s allegedly hostile attitude to IBM witnesses as compared to Government witnesses may be due to the natural factor of fatigue in a case of this difficulty and duration. At any rate, while the duration of trial is certainly the responsibility of the trial judge, the marathon here is as wearing on the Government as it is upon IBM and cannot constitute evidence of extrajudicial bias against IBM.
 Our reading of the record does disclose that on several occasions Chief Judge Edelstein has expressed his dissatisfaction with counsel for IBM. There have been exchanges in the courtroom and in robing room conferences which indicate that he has, whether justifiably or not, reached the conclusion that he has been “baited” by counsel by their persistence in raising points which he believed had already been determined by previous rulings. The issues here
are hotly contested, much is at stake and charges and countercharges of intransigence have continued to be made in the papers before us. Although we cannot condone intemperate behavior on the part of a trial judge, we have observed following a trial of only nine days, “Judges, while expected to possess more than the average amount of self-restraint, are still only human. They do not possess limitless ability, once passion is aroused, to resist provocation.” United States v. Weiss, 491 F.2d 460, 468 (2d Cir.), cert. denied, 419 U.S. 833, 95 S.Ct. 58, 42 L.Ed.2d 59 (1974); see Smith v. Danyo, 585 F.2d 83, 87 (3rd Cir. 1978); Pfizer, Inc. v. Lord, 456 F.2d 532, 538, 542 (8th Cir.), cert. denied, 406 U.S. 976, 92 S.Ct. 2411, 32 L.Ed.2d 676 (1972); Rosen v. Sugarman, supra, 357 F.2d at 798, 800; Refior v. Lansing Drop Forge Co., 124 F.2d 440, 444-45 (6th Cir.), cert. denied, 316 U.S. 671, 62 S.Ct. 1047, 86 L.Ed. 1746 (1942).
 We have examined the record and are persuaded that while occasional flare-ups toward counsel have undoubtedly occurred, there is no indication that this is other than sporadic. Such isolated instances are undoubtedly endemic to a trial of this dimension, and do not provide any basis for finding personal prejudice against IBM, as distinct from its counsel. See United States v. Carignan, 600 F.2d 762, 764 (9th Cir. 1979); Davis v. Board of School Commissioners, supra, 517 F.2d at 1050-52.
 In sum we conclude that the rulings and conduct of the trial judge complained of here are legally insufficient to warrant refusal under Title 28, §§ 144, 455 or under the due process clause of the Fifth Amendment of the United States Constitution.
 The court below, as we have indicated, found that the motion to recuse under §§ 144, 455 was not only legally insufficient, but also untimely. Section 144 explicitly requires that the motion must be timely made. While section 455 has no such requirement, we have without explication previously assumed that timeliness was also requisite under the amended statute. United States v. Daley, 564 F.2d 645, 651 (2d Cir. 1977), cert. denied, 435 U.S. 933, 98 S.Ct. 1508, 55 L.Ed.2d 530 (1978). But see SCA Servs., Inc. v. Morgan, 557 F.2d 110, 117 (7th Cir. 1977). We see no reason why there should not be a timeliness requirement here even without an explicit statutory provision. See Ex parte American Steel Barrel Co., 230 U.S. 35, 46, 33 S.Ct. 1007, 1011, 57 L.Ed.2d 1379 (1913).
 In any event IBM has not argued that a section 455 motion need not be made in a timely manner. Its argument is that it did not become ultimately persuaded of Chief Judge Edelstein’s personal bias until June 25, 1979, when he refused to quash a deposition subpoena served upon Frank T. Cary,
IBM’s Chairman, to review and produce “virtually all computer-related documents in the files of the IBM Corporation for the years 1974 to some unspecified time in the future when Mr. Cary’s trial testimony would be complete.” IBM has submitted affidavits estimating that compliance would cost $1 billion and require 62,000 man-years of effort just for the documents already in existence. IBM contends that full compliance with the order is not possible “within the lifetimes of any of the participants in this trial.” The breadth of this order, IBM argues, enabled it only at that late point to sensibly construe his prior rulings and conduct as not simply erroneous but explainable only as a product of extrajudicial bias.
 We are persuaded that the refusal to quash the Cary subpoena is not evidence of extrajudicial bias nor does it provide an excuse for a recusal motion which is patently made far too late in this litigation. While facially the subpoena is excessively broad, the underlying circumstances make it far less draconian than suggested by IBM. The record discloses, that counsel for IBM had announced that Mr. Cary was about to testify to post 1974 events in the industry, which events would render the Government’s case obsolete. He was expected to testify broadly about all aspects of the case and to give 500 pages of narrative, which IBM refused to define in any detail. The Government had had no opportunity to examine Mr. Cary since 1974. Its position was that it could not properly cross-examine him with respect to post 1974 developments unless it had access to documentation.
 Moreover, when the subpoena was issued, the Government offered to narrow its scope, to limit its requests and negotiate its terms depending upon the scope of Cary’s proposed testimony. Rather than negotiate, IBM proceeded with its motion to quash in toto. In his opinion Chief Judge Edelstein did invite the parties to enter into an agreement whereby compliance could proceed in stages and “might for instance be limited to particular files within Cary’s possession, custody or control, or in any other reasonable manner.” Prior to the argument of this appeal no negotiations had taken place, each side accusing the other of intransigence.
 The Government argues that IBM’s refusal to negotiate the Cary subpoena was part of a strategy by IBM counsel to maximize its burdens and to multiply the rulings against it in order somehow to build a case for recusal on appeal. We attribute no sinister motives to IBM but by the same token we cannot attribute evil or prejudicial conduct to the trial judge. We are, in short, unconvinced by IBM’s argument that the refusal to quash the Cary subpoena in July, 1979 suddenly triggered the comprehension that for the prior seven years the trial judge had harbored some unidentified, out of court animus against IBM. What was not clear to those who were actively engaged in this combat ab initio is hardly as clear as a mountain lake in springtime to this court faced with a cold record covering some seven years.
 What is clear is that both sides have had liberal and perhaps overly liberal discovery both before and during trial. Before trial IBM deposed all of the Government’s witnesses. Between 1975 and 1977 during trial IBM deposed or redeposed 24 Government trial witnesses. IBM has taken some 160 depositions of Government agencies and non-parties. Government agencies have produced some 26 million pages of documents to IBM, which has copied some 890,000 pages. The Government has further answered some 500,000 Requests for Admission made by IBM.
 Until a decision on the merits has been made, the material issues fixed, the wheat hopefully separated from the chaff, we cannot make any determination as to the propriety of any ruling in this monstrous record, much less divine its motivation. A major practical reason for the timeliness requirement is that the granting of a motion to recuse necessarily results in a waste of the judicial resources which have already been invested in the proceeding. United States v. Daley, supra, 564 F.2d at 651 Rosen v. Sugarman, supra, 357 F.2d at 798.
We have found no case, and none exists, where a recusal would result in the waste involved here. During the past decade the investment of judicial time and energy as well as that of the parties to this litigation has been immense. IBM is not unconscious of this factor. It therefore urges the assignment of a new judge who would examine the record and purge those parts which reveal extrajudicial bias. We are told that a “large part” of the record could thus be salvaged. The issue is not simply how long this would take but rather whether we could reasonably expect a judge to accomplish the task at all. More than 70 live witnesses have appeared. Chief Judge Edelstein has undoubtedly made determinations of their credibility based upon their conduct and demeanor, influential factors that are not disclosed in a printed record. Moreover, the trial judge’s exposure to the relevant product market has been intense. The difficulties of learning from a cold record the technical issues here involved are enormous. IBM has even argued on this appeal that the Chief Judge’s handling of the case has made it virtually unreviewable. The labors of Sisyphus pale by comparison to those that would be imposed upon a new judge, and in the end, the attribution of extrajudicial bias would require extrasensory perception.
 In his closing remarks in oral argument counsel for IBM stated that if this case comes up on appeal before this court five years from now and we reverse and send it back for a new trial, “that would be a catastrophe. That is the worst thing I can possibly imagine.” Terminating this trial at this point before judgment and starting anew would be equally catastrophic.
 We cannot prognosticate the outcome of this litigation or what will happen on appellate review. Our role here, we repeat, was to determine whether IBM has established clearly and indisputably that Chief Judge Edelstein has a personal bias and prejudice against IBM and in favor of the United States. For the foregoing reasons we conclude that the petitioner has not met its burden and that in any event its motion for recusal is untimely.
 The petition for the writ of mandamus is denied.
The Government has not raised this issue and we will assume that § 455 is applicable.
(1) Where he has a personal bias or prejudice concerning a party, or personal knowledge of disputed evidentiary facts concerning the proceeding;
(2) Where in private practice he served as lawyer in the matter in controversy, or a lawyer, with whom he previously practiced law served during such association as a lawyer concerning the matter, or the judge or such lawyer has been a material witness concerning it;
(3) Where he has served in governmental employment and in such capacity participated as counsel, adviser or material witness concerning the proceeding or expressed an opinion concerning the merits of the particular case in controversy;
(4) He knows that he, individually or as a fiduciary, or his spouse or minor child residing in his household, has a financial interest in the subject matter in controversy or in a party to the proceeding, or any other interest that could be substantially affected by the outcome of the proceeding;
(5) He or his spouse, or a person within the third degree of relationship to either of them, or the spouse of such a person:
(i) Is a party to the proceeding, or an officer, director, or trustee of a party;
(ii) Is acting as a lawyer in the proceeding;
(iii) Is known by the judge to have an interest that could be substantially affected by the outcome of the proceeding;
(iv) Is to the judge’s knowledge likely to be a material witness in the proceeding.
IBM’s major reliance for its position that in court conduct and rulings may be relevant to establish extrajudicial prejudice is a dictum in this court’s opinion in Wolfson v. Palmieri, 396 F.2d 121, 124 (2d Cir. 1968). While we agree that conduct in the course of a trial might be relevant to indicate a bias that can only be explained as a personal prejudice against a party, the fact is that no case in this circuit has ever found such bias on the basis of a trial court’s rulings or conduct. This reticence, as we explain below, is well justified. See Wright, Miller
Cooper, supra, § 3549, at 269 (Supp. 1978).
163 U.S.App. D.C. at 377, 502 F.2d at 379 (MacKinnon, J., dissenting); United States v. Conforte, 457 F.Supp. 641, 657-58 n. 12 (D.Nev. 1978); Lazofsky v. Sommerset Bus Co., Inc., 389 F.Supp. 1041, 1044 (E.D.N.Y. 1975).
1.4 to 1 (first 26 Government witnesses) 3.4 to 1 (last 26 witnesses) 3.7 to 1 (last 10 witnesses) 4.9 to 1 (last 5 witnesses)
IBM attacks the Government’s use of these statistics, claiming that they relate “only to the number of objections made and ruled upon by the Chief Judge. If, instead, the total numbers of objections made by each side — the only statistics conceivably relevant to plaintiff’s hypothesis — are used, the result is quite different.” According to IBM, the ratios are:
1.24 to 1 (first 26 witnesses) 1.96 to 1 (last 26 witnesses) 1.85 to 1 (last 10 witnesses) 2.34 to 1 (last 5 witnesses)
While the computers have obviously been running overtime, the relevance of the statistics they have spawned to the issue of judicial bias is highly questionable.
Apparently the practice of the judge has been to read and correct the daily transcript before releasing it to the parties. While IBM characterizes this practice as “secret tampering,” obviously they have been able to check the official transcript with their own, and in three instances of which they complain, the record was as a result corrected. We have further examined IBM’s complaint that the trial judge impermissibly struck or limited the testimony of its witness Spain, but the record reveals that he also struck parts of Spain’s cross-examination by the Government. We again stress that whether testimony has been improperly limited can be determined upon an appeal. The incidents complained of, particularly when viewed against a trial of this length, do not in our judgment rise to the level of suggesting personal bias on the judge’s part.