* This is the fourth time we have considered an appeal arising out of the FOIA dispute between Davis and the FBI. In 1986, Davis submitted a FOIA request for all audiotapes recorded during an FBI criminal investigation known as "BRILAB." That investigation, conducted during 1979-80, concerned bribery and racketeering activities among organized crime figures, politicians, and labor unions in Louisiana. The investigation led to the indictment of five individuals, two of whom were ultimately convicted—including reputed Mafia boss Carlos Marcello.2 Portions of more than 130 BRILAB tape recordings were played at the defendants' 1981 trial. Davis sought the tapes as background for a book he subsequently published in 1989. See JOHN H. DAVIS, MAFIA KINGFISH: CARLOS MARCELLO AND THE ASSASSINATION OF JOHN F. KENNEDY (McGraw-Hill 1989).

After the government refused to release the tapes, Davis brought suit pursuant to FOIA. See 5 U.S.C. § 552(a)(4)(B). The government contended that each tape was properly withheld under one or more statutory exemptions, but the district court concluded that material "unconditionally revealed in open court . . . enter[s] the public domain beyond recall for all time" and therefore cannot be withheld under FOIA. Davis v. Dep't of Justice, No. 88-0130, Order at 3 (D.D.C. May 6, 1991). Although the government argued that it was no longer possible to determine which of a "play list" of 163 taped excerpts had actually been played in the courtroom, the district court held that the government bore the burden of showing that the tapes had not entered the public record and must "suffer the consequences of the impasse." Id. at 4. The court ordered release of all the tapes.

On appeal, this court reversed. See Davis v. Dep't of Justice, 968 F.2d 1276, 1282 (D.C.Cir.1992) ("Davis I"). We held that, while the ultimate burden of persuasion remains on the government, "a party who asserts that material is publicly available carries the burden of production on that issue." Davis I, 968 F.2d at 1279 (emphasis omitted). We then remanded to give Davis an opportunity to show that the tapes he sought, or portions of them, were played at the trial. Id. at 1282.

In an effort to meet his burden under Davis I, Davis produced docket entries and transcripts from the Marcello trial. In response, the FBI released 157 of the 163 tapes and said it would have released another tape but could not find it. The FBI continued to withhold the five remaining tapes on the basis of FOIA Exemption 7(C), which permits an agency to withhold otherwise disclosable records if they were "compiled for law enforcement purposes" and their release "could reasonably be expected to constitute an unwarranted invasion of personal privacy." See 5 U.S.C. § 552(b)(7)(C). The district court sustained the FBI's actions. See Davis v. Dep't of Justice, No. 88-0130, Order at 15 (D.D.C. Oct. 16, 1997).

Davis appealed a second time. In Davis II, we upheld the district court's determination that the FBI's search for the missing tape was adequate. See Davis v. Dep't of Justice, 1998 WL 545422, at *1 (D.C.Cir. July 31, 1998) ("Davis II"). We again remanded, however, this time for the court to determine "whether any of the five tapes withheld in their entirety . . . contains material that can be segregated and disclosed without unwarrantably impinging upon anyone's privacy." Id.

On remand, the FBI determined that it could release one of the five tapes because the principal speaker on the tape had died. But the Bureau concluded that the remaining four tapes were wholly subject to Exemption 7(C), because it could not determine whether the speakers on those tapes were living or dead. See Decl. of Scott A. Hodes ¶¶ 5, 7 (Nov. 24, 1998). Citing an FBI affidavit, see id. ¶ 7, the district court held that the "defendant has made adequate efforts to establish that the speakers on these tapes are not deceased." Davis v. Dep't of Justice, No. 88-0130, Order at 2 (D.D.C. Sept. 15, 2000).

Once again, Davis appealed. In Davis III, we summarily reversed the district court and again remanded the case. See Davis v. Dep't of Justice, 2001 WL 1488882, at *1 (D.C.Cir. Oct.17, 2001) ("Davis III"). "The FBI's affidavit," we held, was "insufficient to determine the extent of the Bureau's efforts to ascertain whether putative beneficiaries of Exemption 7(C) are alive or dead." Id. As a consequence, we were "unable to say `whether the Government reasonably balanced the interests in personal privacy against the public interest in release of the information at issue.'" Id. (quoting Schrecker v. Dep't of Justice, 254 F.3d 162, 167 (D.C.Cir.2001) ("Schrecker I")). Citing our recent opinion in Schrecker v. Department of Justice, we remanded so that "the FBI may document what sources it consulted, and the district court can decide in the first instance whether the government `did all it should have done, and whether it may withhold the disputed information pursuant to Exemption 7(C).'" Id. (quoting Schrecker I, 254 F.3d at 167).

Following our remand order, the FBI filed two more affidavits, spelling out the steps it took to determine whether the speakers were dead or alive, and declaring that those steps did not establish that the speakers on the tapes were deceased. See Second Decl. of Scott A. Hodes (Feb. 26, 2002); Third Decl. of Scott A. Hodes (July 11, 2002). We detail those steps in Part II.A below. The FBI's filings make clear that there are only "two speakers on the audiotapes at issue." Def.'s Mot. for Recons. at 3 (citing Fourth Decl. of Scott A. Hodes ¶ 4 (Aug. 7, 2002)). According to the government, the four tapes come from the FBI's undercover investigation of a "prominent individual," and the speakers are that "prominent individual and the undercover informant." Appellee's Br. 11 (citing affidavits).

In July 2002, unsatisfied with the government's efforts, the district court ordered the FBI to advise each of the two speakers, "by first class mail[,] . . . of defendant's obligation pursuant to this Order to [release the tapes] unless the speaker objects thereto in writing within 30 days." Davis v. Dep't of Justice, No. 88-0130, Order at 1 (D.D.C. July 23, 2002). More than a year later, after this court issued a subsequent opinion in the Schrecker case, see Schrecker II, 349 F.3d at 657, the government asked the district court to reconsider that order. On August 31, 2004, the court granted the motion to reconsider, "relieve[d] the government from undertaking the additional tasks mandated" in its July 2002 order, and granted summary judgment in favor of the FBI. Davis v. Dep't of Justice, No. 88-0130, Order at 1 (D.D.C. Aug 31, 2004).

Davis then filed his fourth notice of appeal, challenging both the district court's grant of summary judgment dismissing his FOIA complaint, and an earlier order denying his motion for an award of attorney's fees. We consider the former in Part II and the latter in Part III. We "review de novo a decision granting summary judgment to an agency claiming to have complied with FOIA." Schrecker II, 349 F.3d at 661-62. We also review de novo a district court's attorney's fees determination, like the one at issue here, that "rests on an interpretation of the statutory terms that define eligibility for an award." Edmonds v. FBI, 417 F.3d 1319, 1322 (D.C.Cir.2005) (internal quotation marks omitted).

FOIA Exemption 7(C) exempts law enforcement records from release "to the extent that" release "could reasonably be expected to constitute an unwarranted invasion of personal privacy." 5 U.S.C. § 552(b)(7)(C). In deciding whether the release of particular information constitutes an "unwarranted" invasion of privacy, an agency must balance the privacy interest at stake against the public interest in disclosure. See Department of Justice v. Reporters Comm. for Freedom of the Press, 489 U.S. 749, 777, 109 S. Ct. 1468, 103 L. Ed. 2d 774 (1989); Schrecker I, 254 F.3d at 166. We have recognized "that the privacy interest in nondisclosure of identifying information may be diminished where the individual is deceased." Schrecker II, 349 F.3d at 661. Indeed, the "fact of death, . . . while not requiring the release of information, is a relevant factor to be taken into account in the balancing decision whether to release information." Id. (quoting Schrecker I, 254 F.3d at 166). Consequently, "without confirmation that the Government took certain basic steps to ascertain whether an individual was dead or alive, we are unable to say whether the Government reasonably balanced the interests in personal privacy against the public interest in release of the information at issue." Schrecker I, 254 F.3d at 167.

The government's obligation in this regard is to "ma[k]e a reasonable effort to ascertain life status." Schrecker II, 349 F.3d at 662. Its "efforts must be assessed in light of the accessibility of the relevant information." Id. As we said in Schrecker II, there "`would be a question whether the Bureau's invocation of the privacy interest represented a reasonable response to the FOIA request . . . if the Bureau has, or has ready access to, data bases that could resolve the issue.'" Id. (quoting Summers, 140 F.3d at 1085 (Williams, J., concurring)) (emphasis added in Schrecker II). In short, "the proper inquiry is whether the Government has made reasonable use of the information readily available to it, and whether there exist reasonable alternative methods that the Government failed to employ." Schrecker II, 349 F.3d at 662.

* The government's affidavits and pleadings declare that the FBI took three steps to determine whether the two speakers on the tapes were deceased. The Bureau reports that, because those individuals' deaths could not "be ascertained" by these methods, "the four tapes were withheld in full pursuant to exemption b(7)(C)." Second Hodes Decl. ¶ 8. Of course, "[t]he failure to discover the information sought is not conclusive evidence that the agency has failed to make a reasonable effort." Schrecker II, 349 F.3d at 662. Here, however, the government's own declarations provide that evidence.3

1. The government describes the first method it employed in the following paragraph from its principal affidavit:

The FBI has institutional knowledge of the death of certain individuals from the processing of prior FOIA requests or internal records. The FBI relies on this institutional knowledge, as well as Who Was Who, a book of famous individuals [who have died].

Second Hodes Decl ¶ 5. From this description, it appears that the government's first step involved resort to two different sources: institutional knowledge and Who Was Who.

If the FBI truly used its "institutional knowledge" to determine whether the speakers were dead or alive, this first step might well be reasonable. But the Bureau's method of accessing that knowledge is so constrained as to render it effectively useless. Although the affidavit could be read as suggesting that the FBI uses its "internal records" to determine an individual's status, the same affidavit indicates that the Bureau did not search any records that were not themselves "responsive" to Davis' FOIA request—that is, it did not search any records other than the audiotapes themselves. See Second Hodes Decl. ¶¶ 6, 7.4 Needless to say, the tapes themselves disclose nothing on this point, other than that the speakers were alive when they were speaking sometime during 1979-80.

The only other piece of "institutional knowledge" mentioned in the FBI's description is knowledge gained from "the processing of prior FOIA requests." Second Hodes Decl. ¶ 5. We have no idea what that means. On its face, however, its utility must depend upon there having been a prior FOIA request involving the same individuals. The FBI does not suggest that there had ever been such a request.

If the FBI's reference to its institutional knowledge means anything more than what we have just described, we cannot determine that from the affidavit the Bureau filed. What we said of an earlier affidavit in this case, one that made a similar reference to the FBI's "institutional knowledge," remains equally true regarding this affidavit's treatment of that subject: "The FBI's affidavit is insufficient to determine the extent of the Bureau's efforts to ascertain whether putative beneficiaries of Exemption 7(C) are alive or dead." Davis III, 2001 WL 1488882, at *1. Indeed, with respect to the FBI's reliance on its institutional knowledge, it appears that the Bureau has "been completely passive on the issue, taking death into account only if the fact has happened to swim into [its agents'] line of vision." Summers, 140 F.3d at 1085 (Williams, J., concurring); see also Schrecker I, 254 F.3d at 167 (reversing and remanding the district court's judgment regarding the applicability of Exemption 7(C) because the FBI's affidavit was too vague to determine if the agency had taken "certain basic steps to ascertain whether an individual was dead or alive").

The other source mentioned in the FBI's affidavit is Who Was Who, a multi-volume set of books published periodically by Marquis Who's Who, LLC. Each new volume "includes the biographies of the most prominent and noteworthy people who have died since the publication of the previous edition."5 It is a select company: of the more than 7.2 million Americans who died during 2000-02,6 for example, no more than 4000 are portrayed in the Who Was Who volume covering that period. See 14 WHO WAS WHO IN AMERICA, 2000-02 (2002). How one earns a place in it is unclear, although Marquis reports that most of the entries were originally listed with the subjects' permission in its sister Who's Who in America publication, and that many of the biographies "have been scrutinized and revised by relatives or legal representatives of the deceased Biographee." Id. at vi. All of this suggests both considerable self-selection and considerable lag time.

The government describes one of the two speakers at issue here as a "prominent individual" and the other as an "undercover informant." Appellee's Br. 11. The latter seems unlikely to qualify for the distinction of a Who Was Who entry, and we have no way of knowing whether the former was prominent enough to qualify. Accordingly, we cannot conclude that the FBI's first method—reference to its institutional knowledge and to Who Was Who —was reasonably calculated to determine whether the speakers on the tapes were living or dead.

2. The government describes the second method it employed as follows:

When birth dates are provided in responsive records, and these dates indicate the individual would be over 100 years of age, the name and/or any other identifiers will be released. Although the FBI is aware that many individuals live to be older than 100 years of age, . . . the FBI has consistently relied upon the 100-year rule in all of its FOIA processing.

Second Hodes Decl. ¶ 6 (emphasis added). The key to the utility of the FBI's 100-year rule is the clause that we have italicized. As explained above, when the FBI refers to "responsive records," it means those records—and only those records— actually sought in the FOIA request. In this case, the only responsive records were the audiotapes, and "there were no birth dates on these tapes." Second Hodes Decl. ¶ 6. Therefore, the affidavit concludes, since "no birth dates were provided in the responsive records, the FBI did not assume death of the individuals speaking on these tapes." Id.

The reasonableness of this second method obviously depends upon the probability that the responsive records will contain the individual's birth date—as might well be the case if the records sought by the FOIA requester were FBI investigative reports or personnel files. But unless the FBI has tape recorded a birthday party, it seems highly unlikely that the participants in an audiotaped conversation would have announced their ages or dates of birth.7 Accordingly, this second method was also destined to fail, as it did.

3. The third method the FBI used was the following:

If a social security number is revealed on the responsive records, the FBI, in its administrative discretion, may check the Social Security Death Index (SSDI)—a database maintained by a third party on the Internet. This website is maintained by a private individual, and the FBI cannot verify or vouch for the accuracy of this index, which the website purchases from the Social Security Administration.

Second Hodes Decl. ¶ 7 (emphasis added). Once again, the rub is that the FBI will not even check the Social Security Death Index unless the speaker's social security number is revealed on responsive records. See id. (declaring that the "FBI does not research third-party names internally to discover social security" numbers). As expected, the FBI reports that, because "[t]his case concerns audio tapes," and "[a]s no social security numbers are on the tapes at issue, this website [the SSDI] was not checked." Needless to say, no one announces his or her social security number in ordinary conversation—not even at a birthday party. Accordingly, the Bureau again utilized a method that could not help but fail in the circumstances of this case.

4. As we have discussed, none of the three methods used by the FBI had any likelihood of discovering whether the two individuals, whose conversations were captured by the audiotapes in question, were living or dead. Indeed, if the FBI limits itself to those methods in the future, it is doubtful that it will ever be able to discover the status of a speaker on an audiotape. Although futility alone may not render the FBI's efforts unreasonable, it surely is an important factor in the equation.

The other factor we must consider is "whether the Government has made reasonable use of the information readily available to it, and whether there exist reasonable alternative methods that the Government failed to employ." Schrecker II, 349 F.3d at 662. As we have said, the question is whether "the Bureau has, or has ready access to, data bases that could resolve the issue." Id. (internal quotation marks omitted). We now consider some of the alternatives the FBI failed to employ.

Turning first to the FBI's internal records, we cannot conclude that the government "made reasonable use" of its own information in this case. The flaw in the government's 100-year rule and Social Security Death Index methodologies—at least as far as audiotapes are concerned—is that the Bureau refused to look anywhere but in the tapes themselves to discover the speakers' birth dates or social security numbers. This meant that, even if those personal identifiers were present in other FBI records, the FBI would not have found them. In this case, it is not unlikely that such other records do exist. In an FBI undercover operation like BRILAB, for example, one would expect to find investigative reports that list the various players (witnesses, informants, subjects, targets) and identifying information about them.8 The FBI's Electronic Surveillance (ELSUR) indices also may well contain the necessary personal identifiers for the two individuals, whose voices were captured by electronic surveillance.9

The FBI has neither denied that it has such records nor suggested that it would be difficult to access them in this case. To the contrary, yet another FBI affidavit explains that the Bureau's Central Records System (CRS) contains "all pertinent information which it has acquired in the course of fulfilling its mandated law enforcement responsibilities," and makes clear that much of that system can be searched by an individual's name. Fourth Hodes Decl. ¶ 7. For example, with respect to FBI Headquarters:

Communications directed to FBIHQ from the various field offices . . . are filed in the pertinent case files and indexed to the names of individuals, groups, or organizations which are listed in the case caption(s) or title(s) as a subject(s), a suspect(s), or as a victim(s). Therefore, for example, a search made in this index to locate records concerning a particular individual would be made by searching the name of that individual in the index.

Id. ¶ 8 (emphasis added). And similarly, with respect to FBI field offices:

Access to the CRS files at FBI field divisions is also afforded by the General Indices (automated and manual), which are likewise arranged in alphabetical order, and consist of an index on various subjects, including the names of individuals and organizations. Searches made in the General Indices to locate records concerning a particular subject, such as John Doe, are made by searching the subject requested in the index. Indexing functions have been automated by field divisions.

Id. ¶ 9 (emphasis added).

It is plain, then, that the FBI could have searched its files by the names of the two speakers—one the subject of a criminal investigation, the other an informant—to determine whether records in those files disclose their dates of birth or social security numbers (or even their deaths). But the FBI did not do so. See Second Hodes Decl. ¶¶ 6, 7. The Bureau gave no reason at all for not searching its records for the speakers' birth dates, and only one reason for not searching for their social security numbers:

The FBI does not research third-party names internally to discover social security [numbers] because to do so would violate these third parties' privacy rights. The records that may possess these individual[s'] social security numbers were created for law enforcement purposes; they were not created for the purpose of ascertaining whether individuals contained in FBI records are still alive.

Second Hodes Decl. ¶ 7. We expressly rejected this precise rationale (asserted by the same FBI declarant) in Schrecker II, saying: "We fail to see how the purpose for which an internal record was created bears on whether searching the record for an individual's social security number would violate that individual's privacy." 349 F.3d at 664.

Turning from the FBI's own databases to those it "has ready access to," id. at 662 (internal quotation marks omitted), we have to ask why the FBI limited itself to Who Was Who. The fact that the Bureau uses such an outside source indicates that there is no bar to its doing so. And apparently the biographies of the decedents contained in those books make it possible to determine that the "John Doe" for whom the FBI is searching is the same John Doe whose death is there reported. See infra Part II.B.

But if that is so, one has to ask why—in the age of the Internet—the FBI restricts itself to a dead-tree source with a considerable time lag between death and publication, with limited utility for the FBI's purpose, and with entries restricted to a small fraction of even the "prominent and noteworthy"? Why, in short, doesn't the FBI just Google the two names? Surely, in the Internet age, a "reasonable alternative" for finding out whether a prominent person is dead is to use Google (or any other search engine) to find a report of that person's death.10 Moreover, while finding a death notice for the second speaker—the informant—may be harder (assuming that he was not prominent), Googling also provides ready access to hundreds of websites collecting obituaries from all over the country, any one of which might resolve that speaker's status as well. See, e.g., http://www.legacy.com (hosting the obituary sites of more than 275 newspapers, including three Louisiana papers); http://www. obituarycentral.com (containing a directory of links to online obituaries and death notices in every state).

We do not suggest that the FBI must use one, or any, of the search methods outlined above. But when the only search methods the FBI did employ were plainly fated to reach a dead end (in a manner of speaking), and when there appear to be reasonable alternatives that the government failed to consider, there is a serious "question whether the Bureau's invocation of the privacy interest represented a reasonable response to the FOIA request." Schrecker II, 349 F.3d at 662 (internal quotation marks omitted). This a question that has not yet been answered, and that the district court must address on remand.

The government does not dispute that the steps the FBI took to determine whether the speakers on the audiotapes were dead could not reasonably be expected to answer that question. Nor does it contend that resort to any of the alternatives identified above would be burdensome on the facts of this case. Instead, its argument is simply that "the steps taken by the FBI here were the same taken by the FBI in Schrecker II," and that because we affirmed the grant of summary judgment for the government there, we must do so here as well. Appellee's Br. 9; see Oral Arg. Tape at 29:15-34:12 (confirming the government's view that the FBI does not have to search for the speakers' birth dates or social security numbers, even if it knows they can quickly be found by a name search, because "there is no obligation under Schrecker to conduct additional searches" of nonresponsive records). That is a serious misreading of Schrecker II.

It is true that the three methods the FBI employed to determine whether individuals mentioned in responsive records in Schrecker II were dead are the same as those it employed in this case. See Schrecker II, 349 F.3d at 660. It is also true that we affirmed summary judgment in that case, holding, in particular, that the Bureau did not have to examine nonresponsive internal records for the mentioned individuals' birth dates or social security numbers. Id. at 663-65. But Schrecker II did not purport to affirm a set of search methodologies as per se sufficient to satisfy the "reasonable efforts" standard. Id. at 662. To the contrary, we noted that the "`adequacy of an agency's search is measured by a standard of reasonableness, and is dependent upon the circumstances of the case.'" Id. at 663 (quoting Truitt v. Dep't of State, 897 F.2d 540, 542 (D.C.Cir.1990))(additional quotation marks omitted). And we expressly "cautioned . . . that it would be inappropriate for the court to mandate `a bright-line set of steps for an agency to take in this situation,'" because FOIA requires "`both systemic and case-specific exercises of discretion and administrative judgment and expertise.'" Id. at 662 (quoting Johnson v. Executive Office for U.S. Attorneys, 310 F.3d 771, 776 (D.C.Cir.2002)).

Our determination in Schrecker II, that in that case the FBI had reasonably decided to examine only responsive records for birth dates and social security numbers, is distinguishable from this case in two important respects. First, Schrecker involved a FOIA request not for audiotapes, but for documents. Although there is virtually no chance that a speaker will announce those personal identifiers during an oral conversation, there is a reasonable probability that they will be contained in responsive documents. Indeed, in Schrecker II itself, the FBI found the birth dates and social security numbers of at least some of the mentioned individuals in a search of responsive pages. 349 F.3d at 663, 664.

Second, at issue in Schrecker II were 113 names appearing in over 24,000 responsive documents, making a search of nonresponsive documents for personal identifiers "unduly burdensome." Id. at 664. That burden was magnified by the fact that the 113 names were not those of the subjects of an FBI investigation, but rather were merely mentioned in documents relating to that investigation. This meant that "any name-based search would likely encounter . . . duplication" of common names, "making verification difficult or impossible." Id. It also meant that the prospects of success were "dubious." Id. Under those circumstances, we held that to "require the Government to shoulder such a potentially onerous task . . . goes well beyond the `reasonable effort' demanded in this context." Id.

But Davis' FOIA request entails no such burden. Here, there are only two names and only four responsive records. Those two names belong to the prominent subject of a major FBI investigation, and to the FBI's own undercover informant. Given the Bureau's knowledge of their biographies, even if those individuals have common names, verification that a name-based search has produced records that relate to them would be neither "difficult [n]or impossible." Id. While there may be thousands of John Does, there are unlikely to be thousands connected with the BRILAB investigation. And given the fact that the FBI maintains name-specific records regarding its subjects and informants, see Fourth Hodes Decl. ¶¶ 8, 9 (quoted supra Part II.A.4); cases cited supra notes 8 & 9, here a name-based search would not have "dubious prospects of success." Schrecker II, 349 F.3d at 664.

Indeed, the request in this case looks less like the one at issue in Schrecker II and more like one that Schrecker II expressly distinguished. The latter involved another FOIA request by Ellen Schrecker, for documents relating to Joseph Fischetti, a Chicago-based organized-crime figure. As we explained, "Fischetti was the sole subject of that FOIA request, not one of a multitude of third parties appearing in responsive documents." Id. at 664 (emphasis in original). The request in Schrecker II was different, we said, because "[w]hile it may be reasonable to pursue internal research to determine whether a single subject is the same individual shown by the [Social Security Death Index] to be deceased, . . . it would be unduly burdensome to require the Government [to] do so for the large number of third parties appearing in documents responsive to Schrecker's request." Id. Davis' request, involving only two names in four audiotapes, is far closer to the Fischetti request than to the one we dismissed in Schrecker II.11

To repeat what we said at the beginning of this subpart, there is no "bright-line set of steps for an agency to take in this situation." Id. at 662 (internal quotation marks omitted). Rather, the "adequacy of an agency's search is measured by a standard of reasonableness, and is dependent upon the circumstances of the case." Id. at 663 (internal quotation marks omitted). In determining whether an agency's search is reasonable, a court must consider the likelihood that it will yield the sought-after information, the existence of readily available alternatives, and the burden of employing those alternatives. In this case, the methodology employed by the agency was extremely unlikely to produce the needed information, and it appears—although we do not know for certain—that there are readily available alternatives that would not impose an undue burden on the government. We remand to permit the agency an opportunity to evaluate the alternatives, and either to conduct a further search or to explain satisfactorily why it should not be required to do so.

Finally, we address Davis' request for an award of attorney's fees, which the district court denied on the authority of this circuit's decision in Oil, Chemical & Atomic Workers International Union, AFL-CIO v. Department of Energy, 288 F.3d 452 (D.C.Cir.2002) ("OCAW"). See Davis, Order at 2 (July 23, 2002). The district court's denial was correct. OCAW forecloses a decision in Davis' favor.

FOIA provides that a district court "may assess against the United States reasonable attorney fees . . . reasonably incurred in any case under this section in which the complainant has substantially prevailed." 5 U.S.C. § 552(a)(4)(E) (emphasis added). In Buckhannon Board & Care Home, Inc. v. West Virginia Department of Health & Human Resources, 532 U.S. 598, 121 S. Ct. 1835, 149 L. Ed. 2d 855 (2001), the Supreme Court construed the attorney's fees provision of two other statutes that permit courts to award fees to a "prevailing party."12 The Court rejected the plaintiffs' contention, which it characterized as the "catalyst theory," that "a plaintiff is a `prevailing party' if it achieves the desired result because the lawsuit brought about a voluntary change in the defendant's conduct." Buckhannon, 532 U.S. at 601, 121 S. Ct. 1835. Rather, the Court ruled that, for a litigant to be a "prevailing party," there must have been a "judicially sanctioned change in the legal relationship of the parties." Id. at 605, 121 S. Ct. 1835. "[E]nforceable judgments on the merits and court-ordered consent decrees," the Court said, suffice to create such a change. Id. at 604, 121 S. Ct. 1835.

In OCAW, this circuit extended the holding of Buckhannon to the fee-shifting provision of FOIA. 288 F.3d at 454-57. The OCAW court concluded that "the `substantially prevail' language in FOIA [is] the functional equivalent of the `prevailing party' language found in" the statutes interpreted in Buckhannon. Id. at 455-56. It "therefore h[e]ld that in order for plaintiffs in FOIA actions to become eligible for an award of attorney's fees, they must have `been awarded some relief by [a] court,' either in a judgment on the merits or in a court-ordered consent decree." Id. at 456-57 (quoting Buckhannon, 532 U.S. at 603, 121 S. Ct. 1835) (emphasis added).

Davis' problem is that, although to date he has received a total of 158 tapes from the government, none were produced as the result of a "judgment on the merits" or a "court-ordered consent decree." Id. at 457. It is true that, in 1991, he did secure such a judgment from the district court, directing the FBI to release 163 tapes. But that judgment was reversed by this court in Davis I, and the government's subsequent release of 157 of those tapes— after the FBI determined that they had been played at the trial—was not made pursuant to any judgment or order. Similarly, although Davis II remanded for the district court to determine whether any of the remaining tapes contained material that could be segregated, Davis II, 1998 WL 545422, at *1, such a remand is insufficient to satisfy the OCAW test.13 The FBI's subsequent release of an additional tape (after determining that the speaker was dead) similarly was not pursuant to a judgment or order.

Davis appears to recognize the futility of his effort to distinguish OCAW, as his brief devotes considerably more pages to arguing that OCAW was wrongly decided than to arguing that it can be distinguished. See Appellant's Br. 24-28. The former is an argument that we cannot entertain, because "[o]ne three-judge panel . . . does not have the authority to overrule another three-judge panel of the court." LaShawn A. v. Barry, 87 F.3d 1389, 1395 (D.C.Cir.1996) (en banc). We therefore conclude that the district court correctly determined that it could not grant Davis' request for attorney's fees.

For the foregoing reasons, we affirm the district court's denial of attorney's fees, but we reverse its grant of summary judgment dismissing Davis' FOIA complaint. The case is remanded with directions that the FBI evaluate alternative methods for determining whether the speakers on the requested audiotapes are dead, and that thereafter the district court determine whether the FBI's chosen course is reasonable.

So ordered.