Opinion for the Court filed by Chief Judge WALD.
Dissenting opinion filed by Circuit Judge STARR.
WALD, Chief Judge:
The Department of Defense (“DOD”) petitions here for a writ of mandamus directing District Judge Louis F. Oberdorfer to revoke the appointment of a special master in this Freedom of Information Act (“FOIA”) case. Because we conclude that Judge Oberdorfer did not abuse his discretion in appointing the master, we deny the petition.
I. Background
This petition arises from a lawsuit brought under FOIA1 by the Washington Post and staff writer R. Scott Armstrong seeking to compel production of documents relating to attempts to rescue United States hostages in Iran in 1980. DOD withheld in whole or in part approximately 2,000 documents comprising approximately 14,000 pages. It premised its withholding of almost all of these documents on FOIA’s “national security” exemption. That exemption covers matters that are “(A) specifically authorized under criteria established by an Executive order to be kept secret in the interest of national defense or foreign policy and (B) are in fact properly classified pursuant to such Executive order.” 5 U.S.C. § 552(b)(1). Because it deemed highly sensitive much of the information in the withheld documents, DOD submitted to the district court not only an unclassified document index, but also a classified index providing in greater detail document descriptions and DOD’s justifications for withholding.2
On June 30, 1987, following the submission of these public and in camera document indexes, Judge Oberdorfer announced that he was considering the appointment of a special master to review the documents in question. The Post and Armstrong endorsed the idea and suggested several potential appointees. DOD opposed such an appointment. It proposed instead to prepare a random sample of the withheld documents, based upon which the district court could judge in camera the applicability of the claimed exemptions.
[234] On December 14,1987, Judge Oberdorfer announced that he would appoint a special master. In a three-page notice to counsel, he rejected DOD's random sampling proposal. Judge Oberdorfer explained:
it is difficult to clear the Court’s clerks and other staff to enable them to perform their usual function with respect to a process such as sampling. The defendant presumably contemplates that its personnel would select the samples, if for no other reason, because they have the clearance necessary for access to the original classified documents. However, the integrity of sampling by the government has been authoritatively questioned. See Lame v. United States Dept. of Justice, 654 F.2d 917, 928 n. 11 (3d Cir.1981), cited with approval in Weisberg [v. United States Dept. of Justice ], 745 F.2d [1476] at 1490 [D.C.Cir.1984]. Nor is a random sampling particularly appropriate for the circumstances here. Compare Meeropol v. Meese, 790 F.2d 942, 956-59 (D.C.Cir.1986).
The best solution, Judge Oberdorfer’s notice concluded, was to appoint as master “a person who holds, or has recently held, sufficient clearance to have access to the documents,” and to charge that party with “developing the sample and summarizing to the Court the arguments that each party has made, or could make with respect to the exemptions claimed by defendant.” Under this arrangement,
[t]he Court’s Article III role would be preserved, and indeed enhanced beyond that performed by the Court in an in camera review: adjudication of the exemptions claimed would be based on the samples selected, an appraisal of the method used by the special master to select the samples, and consideration of the arguments pro and con summarized for the Court by the master together with any additional arguments suggested by the parties. The master will make no recommendation.
Judge Oberdorfer also announced his tentative intention to appoint as master Kenneth C. Bass III, a Washington, D.C. attorney who was previously counsel for intelligence policy in the Department of Justice and who in that capacity held top-secret security clearance.
On January 15,1988, the judge issued an order making Bass’ appointment official. He reasoned in the order that there was an indisputable need for careful in camera consideration of such highly sensitive documents, because “the Court and the plaintiff will be denied the truth-testing benefits of an adversary proceeding.” Given the practical difficulty of the judge or his law clerk perusing all 2,000 documents, and the judge’s conclusion based upon his several years experience with the case that random sampling would not be appropriate in this case, a “necessity [existed] of obtaining the services of an independent person who is skilled and experienced in national security matters (including the control of sensitive documents).” The order charged Bass with selecting a “representative sample” of the withheld documents and with summarizing the arguments for and against granting each claimed exemption.3 Four weeks later, DOD filed this petition for a writ of mandamus, to which plaintiff Armstrong has responded by supporting the reference to a master.4 We now turn to the propriety of granting the requested writ.
II. Discussion
The writ of mandamus is “an extraordinary remedy, to be reserved for extraordinary situations.” Gulfstream Aerospace Corp. v. Mayacamas Corp., — U.S. -,-, 108 S.Ct. 1133, 1143, 99 L.Ed.2d [235]*235296 (1988). Historically, the writ has been used in the federal courts “only ‘to confine an inferior court to a lawful exercise of its prescribed jurisdiction or to compel it to exercise its authority when it is its duty to do so.’ ” Will v. United States, 389 U.S. 90, 95, 88 S.Ct. 269, 273, 19 L.Ed.2d 305 (1967) (quoting Roche v. Evaporated Milk Association, 319 U.S. 21, 26, 63 S.Ct. 938, 941, 87 L.Ed. 1185 (1943)). The Supreme Court has offered several reasons for thus limiting the invocation of mandamus to “exceptional cases where there is clear abuse of discretion or ‘usurpation of judicial power.’ ” La Buy v. Howes Leather Co., 352 U.S. 249, 257, 77 S.Ct. 309, 314, 1 L.Ed.2d 290 (1957) (quoting Bankers Life & Casualty Co. v. Holland, 346 U.S. 379, 383, 74 S.Ct. 145, 148, 98 L.Ed. 106 (1953)). First, mandamus actions have “the unfortunate consequence of making the [district court] judge a litigant, obliged to obtain personal counsel or to leave his defense to one of the litigants [appearing before him] in the underlying case.” Second, “particularly in an era of crowded dockets, it is in the interest of the fair and prompt administration of justice to discourage piecemeal litigation.” Third, Congress, ever since the Judiciary Act of 1789, has signalled its desire that appellate review be postponed until a trial court has rendered a final judgment. Issuing a writ of mandamus “in anything less than an extraordinary situation would run the real risk of defeating” the policies Congress sought to further. Kerr v. United States District Court, 426 U.S. 394, 402-03, 96 S.Ct. 2119, 2123-24, 48 L.Ed.2d 725 (1976) (describing mandamus remedy as “a drastic one”) (citations omitted).
In light of the extraordinary nature of the requested form of relief, the burden falls upon DOD here to show that its right to issuance of the writ is “clear and indisputable.” Kerr, 426 U.S. at 403, 96 S.Ct. at 2124. DOD’s argument that it can show such a clear and indisputable right here turns on its assertion that although Federal Rule of Civil Procedure 53(b) requires an “exceptional condition” to exist before a judge can appoint a master, no such condition exists in this case. The rule states:
A reference to a master shall be the exception and not the rule. In actions to be tried by a jury, a reference shall be made only when the issues are complicated; in actions to be tried without a jury, save in matters of account and of difficult computation of damages, a reference shall be made only upon a showing that some exceptional condition requires it.
Fed.R.Civ.P. 53(b).
Although Rule 53(b) does impose significant limits on when a trial judge can refer matters to a master, we cannot agree that Judge Oberdorfer abused his discretion in making the appointment in this case. Our circuit has recognized the propriety under Rule 53(b) of designating masters in certain exceptional FOIA cases. In Vaughn v. Rosen, 484 F.2d 820 (D.C.Cir.1973), cert. denied, 415 U.S. 977, 94 S.Ct. 1564, 39 L.Ed.2d 873 (1974), we surveyed the options available to a judge overseeing a FOIA lawsuit. In cases where a judge, after considering the option of himself reviewing indexed documents, concludes that dealing with “the raw material of an FOIA lawsuit may still be extremely burdensome,” we wrote,
it is within the discretion of the trial court to designate a special master to examine documents and evaluate an agency’s contention of exemption. This special master would not act as an advocate; he would, however, assist the adversary process by assuming much of the burden of examining and evaluating voluminous documents that currently falls on the trial judge.
484 F.2d at 828. See also Meeropol v. Meese, 790 F.2d 942, 961 (D.C.Cir.1986) (noting in FOIA case that “the decision whether to appoint a master lies within the discretion of the trial court,” and therefore refusing to appoint a master where the trial court had opted not to).5
[236] In our view, this case features a very special set of circumstances which meet the "exceptional condition” test of Rule 53. It involves a FOIA claim with respect to which the judge has no access to impartial expert witnesses or other features of the adversary process in order to assist him in making his decision about disclosure. It also involves a massive collection of 14,000 pages which must be sifted through. As Judge Oberdorfer’s response to the mandamus petition pointed out, he had only recently used his law clerks for several months to assist him in a “much less technical and sensitive FOIA case,” see Nishnic v. United States Department of Justice, 671 F.Supp. 771 (D.D.C.), aff'd, 828 F.2d 844 (D.C.Cir.1987); Nishnic v. United States Department of Justice, 671 F.Supp. 776 (D.D.C.1987), and presumably would have done so again, except for the fact that to do so in this case would have required that they obtain special security clearances, a time-consuming process in itself. The judge also made a judgment based upon several years experience on the case that random sampling would not be an appropriate technique in this instance. He therefore resorted to appointment of a master to ensure a prompt, thorough, and independent look at all the documents. Apart from our assumption that Rule 53 itself clearly envisions considerable discretion in the district court in deciding when an “exceptional condition” exists, we find the special combination of circumstances here fits well within that term. See also Fed.R.Civ.P. 53 advisory committee note (“such masters may prove useful when some special expertise is desired”).
Moreover, in both his notice to counsel and his order appointing the master, Judge Oberdorfer carefully cabined the master’s authority. He expressly forbade the master from making any recommendations, charging him instead with the more limited task of developing a representative sample and summarizing each party’s arguments or potential arguments. In this respect, this case is not at all akin to those few in which courts have issued writs of mandamus revoking masters’ appointments as outside of a trial court’s discretion. On each such occasion, the trial court’s error was in delegating ultimate decisional authority to the master.
Thus, in the leading case of La Buy v. Howes Leather Co., 352 U.S. 249, 77 S.Ct. 309, 1 L.Ed.2d 290 (1957), relied on here by DOD, the Supreme Court overturned a district court’s wholesale reference to a master of two full civil antitrust trials. The Court, rejecting the district judge’s argument that calendar congestion and issue complexity constituted “exceptional circumstances,” noted that the use of masters is “ ‘to aid judges in the performance of specific judicial duties as they may arise in the progress of a cause,’ and not to displace the court.” 352 U.S. at 256, 77 S.Ct. at 313 (citations omitted). It wrote that the litigants were entitled to “trial before a regular, experienced trial judge rather than before a temporary substitute appointed on an ad hoc basis.” Id. at 259, 77 S.Ct. at 315. See also In Re: United States, 816 F.2d 1083, 1089 (6th Cir.1987) (district judge’s reference to master of “disposi-tive[] pretrial motions” was error, even though need for pretrial management “may be persuasive as to the reference of discovery matters”); In Re: Armco, Inc., 770 F.2d 103, 105 (8th Cir.1985) (upholding reference to master in complex environmental case insofar as he was granted broad authority to supervise pretrial matters, but issuing writ revoking master’s [237] authority “to preside at trial on the merits of the case”); Jack Walters & Sons Corp. v. Morton Building, Inc., 737 F.2d 698, 711-13 (7th Cir.1984) (criticizing in dicta reference to master of power to make recommendations on motion for summary judgment where “the opinion [the master] wrote was indistinguishable in form from a judicial opinion; and the district judge approved that opinion rather than write his own”); Bennerson v, Joseph, 583 F.2d 633, 642 (3d Cir.1978) (district judge violated La Buy principle in referring to master entire case on validity of land conveyances).
This case involves no such untoward or irretrievable delegation of judicial authority to decide the merits. See Will, 389 U.S. at 104 n. 14, 105, 88 S.Ct. at 278 n. 14, 279 (noting that the La Buy decision “specifically relied upon evidence in the record which showed a pattern of improper reference of cases to special masters by the district judge,” and thus La Buy’s approval of the revocation of a master “is simply inapposite where there is no showing of a persistent disregard of the federal rules”); see also Vickers Motors, Inc. v. Wellford, 502 F.2d 967, 968 (6th Cir.1974) (mandamus should not issue unless order of reference has “deprive[d] ‘the parties of a trial before the court on the basic issues involved in the litigation’ ”) (citations omitted). Indeed, Judge Oberdorfer specifically promised the parties that they would have the opportunity to challenge before him the conclusions reached in the master’s report. See Transcript of December 21, 1987 Hearing at 12 (reprinted as Attachment C to Petition for a Writ of Mandamus).
Despite the weight of this precedent and the rather unremarkable nature of the master’s duties in this case, DOD makes two arguments why we should grant its petition for mandamus relief. First, DOD contends, the district judge’s decision to appoint a master was grounded on his faulty legal premise that “the integrity of sampling by the government has been authoritatively questioned.” DOD observes that several decisions of this circuit have specifically approved of random sampling by the government. See Petition for a Writ of Mandamus at 12-13 (hereinafter “DOD Petition”) (citing, inter alia, Meeropol v. Meese, 790 F.2d 942, 958 (D.C.1986); Weisberg v. United States Department of Justice, 745 F.2d 1476, 1490 (D.C.Cir.1984); and Ash Grove Cement Co. v. Federal Trade Commission, 511 F.2d 815, 817 (D.C.Cir.1975)). With the district court’s erroneous premise exposed, DOD argues, no “exceptional condition” remains to support the master’s appointment.
We cannot agree. The cases offered by DOD approving of government sampling all involved the technique of random sampling. Yet in the notice to counsel in which he announced his plans to appoint a master, Judge Oberdorfer specifically said that he did not consider random sampling appropriate in this case, deciding instead that the alternative technique of representative sampling was required. See supra. The former method involves a mechanical selection process designed simply to present a decisionmaker (here, the district judge) with a sample of more manageable size.6 In the latter method, however, the authority charged with preparing the sample scrutinizes an entire document pool, and includes in the sample items which in its judgment are typical of broader classes of documents within the pool. The sifting process of representative sampling, when properly executed, has the advantage of assuring that the documents in the sample do indeed typify the broader pool. By contrast, random sampling, like the 1% sampling strategy used in Weisberg, although appropriate in many cases, may overlook distinctive documents or accidentally shield patently unexceptional documents from FOIA release.7 A trial judge familiar with [238] the case clearly has discretion to choose which method is best.
Representative sampling does, however, inevitably involve the exercise of considerable discretion on the part of the party charged with preparing the representative sample. Thus, if representative sampling is deemed necessary as it was here, there is judicial authority for the judge’s additional authority to select a nonparty to conduct the sampling. For example, in Lame v. United States Department of Justice, 654 F.2d 917 (3d Cir.1981), the Third Circuit described as an “obvious danger” that “the government may choose to submit information on ‘representative’ documents for which its claims of exemption fare the most persuasive.” Id. at 928 n. 11.8 In no case of which we are aware has this or any other court suggested that the trial judge lacks discretion to require a neutral selection of representative documents. We cannot agree therefore with DOD’s argument that the reference to a master in this case rested upon a faulty legal foundation. In so ruling we in no way mean to accuse the government of “skullduggery,” as Judge Starr suggests in arguing that we should grant this extraordinary writ, see Diss. op. at 242, but merely to reaffirm the broad discretion over trial-management tactics with which a district judge is vested.
DOD finally argues that the fact that highly sensitive classified materials claimed under FOIA’s national security exemption are involved here counsels against appointment of a master. We disagree. Judge Oberdorfer’s decision to appoint a master “who holds, or has recently held, sufficient clearance to have access to the documents,” shows commendable sensitivity to the importance of confining the number of persons privy to the documents in question. The alternatives would have been charging the district judge’s law clerks with reviewing the documents (thereby expanding the pool of persons with access to the documents, and potentially delaying this litigation by months while the clerks obtained the requisite clearance), using random sampling (a sorting procedure which is not necessarily appropriate in all cases), and relying on DOD to carry out its own representative screening (a strategy which has obvious drawbacks). See Ray v. Turner, 587 F.2d 1187, 1194 (D.C.Cir.1978) (“[w]hether and how to conduct an in camera examination of the documents rests in the sound discretion of the court, in national security cases as in all other cases”) (footnote omitted).
Judge Starr, however, would grant DOD’s petition for mandamus and thereby revoke Judge Oberdorfer’s appointment of a master. Yet granting mandamus here would expand considerably this extraordinary writ, using it to second-guess trial judges in situations where they have not exceeded their “prescribed jurisdiction” or failed to exercise their required authority. See Will, supra. Moreover, Judge Starr’s argument that the trial judge was compelled to consider alternatives to the appointment of a master overlooks the fact that Judge Oberdorfer did consider, but ultimately ruled out as inefficacious or otherwise problematic, a number of alternatives, including those we enumerate in the preceding paragraph. Having found serious flaws in these alternatives, he was well within his discretion to conclude that Rule 53 required recourse to a master. In urging deployment of a mandamus writ against Judge Oberdorfer, Judge Starr would have us conclude that the trial judge —after four years’ experience on this case, and armed with instructive experience on other FOIA litigation to which he alluded in recounting his decision to appoint the master — abused his discretion in adjudging this trial-management technique more effective than others. Nothing in Rule 53, or in the daunting standards governing mandamus, supports such manifest appellate court in[239] terference with Judge Oberdorfer’s trial management.9
Conclusion
We conclude finally that although the facts of this case make it sui generis, our decision to deny the requested mandamus writ is embraced by a clear limiting principle similar to that at work in those cases upholding appointment of masters to conduct unusual pretrial discovery. See note 5, supra. That principle as applied to this case is as follows: where a massive number of classified documents exists such that the judge and his law clerk simply cannot examine them all, and where the judge has reasonably concluded that alternative methods of document review are infected with serious problems, appointment of a master to structure the judge’s review of these documents is appropriate so long as the judge retains decisional authority over the issue in question. In this particular FOIA case, Judge Oberdorfer had no access to impartial expert evidence or even the benefit of the traditional adversary process to illuminate the nature of the documents in question. He therefore determined, based upon years of experience on this case, that perusal of a random sample of the documents would not provide a sufficient basis upon which to make his decision. He was then faced with the choice of reading all 14,000 pages of classified documents himself, or waiting for his law clerks to be security-cleared for access to them. Instead, he chose a security-cleared expert in the field as a master to make the initial selection of the representative documents for his examination. In our view, this combination of factors clearly met the requirements of Rule 53. We therefore deny DOD’s petition for a writ of mandamus.
It is so ordered.