Opinion for the Court filed by Circuit Judge HARRY T. EDWARDS.
Separate statement, concurring in part and dissenting in part, filed by Circuit Judge BORK.
HARRY T. EDWARDS, Circuit Judge:This decision marks the end of another inconclusive chapter in a course of litigation that has already been unduly prolonged. In 1977-1978, the appellants filed first a request and then a suit under the Freedom of Information Act, 5 U.S.C. § 552 (1976 & Supp. V 1981) (“FOIA”), seeking disclosure by the CIA of the names of the institutions and individuals who, in the late 1950’s and early 1960’s, had conducted secret research for the CIA under its so-called MKULTRA program. The grisly details of the situation that gave rise to the appellants’ request, and the history of the agency’s responses to their demand, are recounted in our prior decision in the case, reported at 642 F.2d 562 (D.C.Cir.1980). For present purposes, we need only highlight the critical events associated with this litigation.
The CIA refused to reveal most of the information that was initially requested by the appellants. In defense of its recalcitrance, the agency invoked exemptions 31 and 62 of the FOIA; however, agency officials expressly renounced reliance on exemption l.3 After affording the CIA sever*97al opportunities to present evidence in support of its positions, the District Court rejected both of the arguments advanced by the agency. On appeal, a panel of this court upheld the District Court’s disposition of the CIA’s exemption 6 argument. The panel concluded, however, that the trial court’s analysis of the agency’s exemption 3 defense was hampered by the unavailability of a coherent definition of an “intelligence source” within the meaning of the National Security Act.4 Accordingly, the court formulated a definition of that crucial phrase and remanded the case for reconsideration of the agency’s argument in light of its holding. Sims v. CIA [Sims I], 642 F.2d 562, 571-72 (D.C.Cir.1980).
On remand, the District Court invited the CIA to submit evidence that would indicate which of the institutions and individuals whose identities the agency sought to withhold fell within the prescribed definition of “intelligence source.” The CIA’s responses were neither thorough nor prompt.5 Eventually, however, the trial court did succeed in assembling a significant body of relevant information. After examining and analyzing the parties’ various submissions, the court issued two opinions separating the individuals and institutions into various categories, determining whether the members of each group satisfied the definition of “intelligence source,” and ordering disclosure of the identities of those who failed to qualify. Memorandum Opinion, Dec. 22, 1981, reprinted in Appendix (“App.”) 27-34; Memorandum Opinion, June 15, 1982, reprinted in App. 35-40.
Almost all of the District Court’s various rulings were judicious and proper. One aspect of its analysis, however, was flawed; the court misconceived the level of generality at which the definition of “intelligence source” should apply. We are obliged, therefore, to reverse that portion of the District Court’s judgment and remand the case for further proceedings.
I.
Sims I defined an “intelligence source” as:
a person or institution that provides, has provided, or has been engaged to provide the CIA with information of a kind the Agency needs to perform its intelligence function effectively, yet could not reasonably expect to obtain without guaranteeing the confidentiality of those who provide it.
642 F.2d at 571. It is beyond dispute that the foregoing definition constitutes both the law of the case and the law of this circuit. The only reason for the reversal of the District Court’s decision in Sims I was to allow the trial court on remand to reconsider, on the basis of this definition, its evaluation of the CIA’s invocation of exemption 3. See id. at 571-72. Clearly, therefore, the definition itself is part of the holding of the Sims I decision.
Moreover, since the issuance of Sims I, the court has had an opportunity once again to consider the meaning of “intelligence source”; in Holy Spirit Association for the Unification of World Christianity v. CIA, 686 F.2d 838, 844 (D.C.Cir.1980), other portions of the decision vacated and remanded as moot, 455 U.S. 997, 102 S.Ct. 1626, 71 L.Ed.2d 858 (1982), a second panel expressly reaffirmed the definition enunciated in Sims I.6 Thus, the argument advanced by the agency on this appeal — that the last clause of that definition should be regarded as mere dictum, the validity of which is subject to reevaluation by the present panel at this juncture — is either disingenuous or obtuse.
II.
Unfortunately, the District Court on remand failed to adhere to the instructions in *98Sims I regarding the manner in which the definition of “intelligence source” should be applied. In Sims I, it was made clear that the first thing the trial court should do is define “the class or ‘kind’ of information involved.” 642 F.2d at 571. Such a determination, it was indicated, would make it possible for the court then to assess “the likelihood that disclosure would undermine CIA access to information of that kind.” Id. (emphasis added). In Holy Spirit Ass’n (issued prior to the District Court’s decisions on remand), the court again emphasized that application of the definition of “intelligence source” requires an identification of the “type” of information that was provided by a particular source. 686 F.2d at 844. After such a determination has been made, the trial court can and should consider whether the agency could reasonably expect to obtain information of that type without guaranteeing its providers confidentiality. See id.
We do not mean to suggest that the District Court simply ignored or totally misconstrued the holding of Sims I. Some comments in the District Court’s memorandum opinions do reflect an attempt first to identify the types of information involved in the MKULTRA program and then to assess the ease with which the agency could obtain data of those types without pledging its sources secrecy. Thus, at one point, the court observed:
[F]or many of the subprojects, even assuming that a guarantee was given, it is doubtful at best whether it was necessary. Examples include Subproject No. 104, research on the ecological relations involved in the deterioration of petroleum products; No. 92, an exploration of the usefulness of mechanization in foreign language training; No. 112, a study of current problems in vocational guidance; No. Ill, a study of the measurement of motivation; No. 96, a study of the psychology of personal constructs as it applies to foreign cultures; No. 123, a study of emerging ethnic images in three countries; No. 84, a study of high motivation by development of specific interpersonal relationships; and No. 85, a study of blood groupings as a means of identifying individuals. It seems clear that such research, which goes on constantly at many places, could have been done without a guarantee of confidentiality. At least, the defendant has provided no evidence to rebut this inference.
Dec. 22 Opinion at 5-6, reprinted in App. 31-32. Unfortunately, the court’s attention to questions of this order was deflected by its interest in whether the agency had, in fact, promised confidentiality to individual researchers. The court’s concern over whether such specific promises were made seems to have derived from its belief that “[a] necessary prerequisite to the satisfaction of [the third] portion of the definition is that the CIA have in fact made a guarantee of the confidentiality of the source; obviously if no guarantee was given, and the information was nonetheless obtained, then, under the Court of Appeals’ definition, no guarantee was necessary.” Id. at 3, reprinted in App. 29.
Proof that the CIA did or did not make promises of secrecy (either express or tacit) to specific informants would undoubtedly be helpful to a court when deciding whether the agency could expect to obtain information of the kind they provided without guaranteeing confidentiality to sources of it. Evidence that data was obtained despite the fact that the agency failed or refused to pledge secrecy to a source would certainly give rise to an inference that, in general, the agency is capable of securing data of that type without providing guarantees of confidentiality. Conversely, a pledge of secrecy made by the CIA would be probative at least of the agency’s belief that the source desired anonymity. But, in neither case could such evidence be dispositive of the question whether a given informant qualifies as an “intelligence source.” A particular researcher may have been atypically unconcerned with his reputation or safety; that he was willing to provide information without a promise of confidentiality does not mean that, in general, the agency could obtain information of the kind he provided without guaranteeing confiden*99tiality to sources of it. Conversely, an informant may have been unreasonably and atypically leery of providing the agency with innocuous information; if the agency readily and openly could have obtained, from other sources, data of the sort he provided, he would not constitute an “intelligence source.” 7 In short, it was error for the District Court to rule that those six researchers who apparently did request confidentiality automatically qualified as “intelligence sources” and that those who did not request such a pledge automatically did not qualify.8
The dissent objects to the foregoing explication of the Sims I definition on the grounds of both “precedent and common sense.” With regard to “common sense,” we think that the marginal risk that a few persons, who provided the agency with information it readily could have obtained elsewhere but who nevertheless requested a promise of confidentiality, might be surprised to learn that their identities could not be shielded under exemption 3 is more than offset by two considerations. First, there is the serious potential for widespread evasion of the letter and spirit of the FOIA that would be created by the rule advocated by the dissent. See note 7 supra. Second,, there is the fact that, if revelation of the identity of a source of information would in any way impair national security, the agency can easily justify withholding his name by invoking exemption 1 of the Act.
The dissent’s reliance upon “precedent” warrants a more extended response. The dissent makes much of a sentence used by a panel of this court in Holy Spirit Ass’n, pointing to the fact that information contained in three of the documents at issue in that case had been obtained from “persons who gave [the] information with the understanding that it would be kept in confidence.” 636 F.2d at 844. The dissent concludes that the foregoing reference establishes as the law of this circuit that “an informant-solicited promise of secrecy ... automatically qualifpes] the informant as an intelligence source.” We find this suggestion to be quite extraordinary.
For two reasons, the language from Holy Spirit Ass’n cannot possibly support the weight the dissent would place upon it. First, the panel in Holy Spirit Ass’n never intimated that proof that the agency promised a source secrecy conclusively establishes that the promisee was an “intelligence source.” Indeed, the court did not even purport to consider the issue. Furthermore, the most natural reading of the language in question suggests the contrary: the panel in Holy Spirit Ass’n recognized, as do we, that the fact that a pledge was made is probative of the fact that information of the type provided could not have been obtained without guaranteeing its sources confidentiality; in the absence of any evidence in the record in Holy Spirit Ass’n capable of undermining the foregoing inference, the panel saw no reason to overturn the district court’s decision that the documents were shielded by exemption 3. The fact that the panel did not elaborate upon its reasoning in more detail is readily explainable by the posture in which the case appeared before the court. Holy Spirit Ass’n was briefed and argued before issuance of the decision in Sims I. Consequently, neither the district court nor the parties on appeal had the benefit of our reformulated definition of an “intelligence source” when considering the applicability *100of exemption 3. Faced with the task of deciding an issue that, on one hand, had not been addressed by the litigants or the trial judge and, on the other hand, seemed clear-cut on the facts of the case, the panel not surprisingly was somewhat curt in its ratio decidendi.
Second, the sentence highlighted by the dissent is sandwiched between two more extensive discussions in the Holy Spirit Ass’n opinion emphasizing the importance of identifying the type of information at issue and whether the agency could obtain information of that type without promising secrecy to providers of it. In the preceding paragraph, the panel quotes the Sims I definition in full and reiterates the principle that “the appropriate focus is on ‘the practical necessity of secrecy.’ ” 636 F.2d at 844. Clearly, the existence of “an informant-solicited promise of secrecy” does not conclusively establish “the practical necessity of secrecy.” And in the succeeding paragraph, the panel in Holy Spirit Ass’n expressly grounds its ruling that four other documents were properly withheld on its judgment that “the information contained in those documents is of the type that the Agency could not obtain without a guarantee of confidentiality.” Id. In short, if we are bound by precedent on this issue, we are bound to reject rather than accept a rule that proof that an informant requested and was granted a promise of secrecy is determinative of his status as an intelligence source.
On remand, therefore, the District Court should undertake the two-step analysis prescribed in Sims I. The court should begin by defining the type or types of information provided by the various researchers. Because the nature of the subprojects varied considerably,9 this may require delineation of several categories of data.10 The court should then determine whether the CIA could reasonably expect to obtain information of each of those kinds without pledging its providers confidentiality.11 In deciding the latter question, the court may rely upon evidence in the record indicating that specific researchers did or did not in fact request and receive guarantees of confidentiality; but, for the aforementioned reasons, such evidence cannot be controlling.
III.
Two other issues raised by the parties, regarding the meaning of Sims I, merit brief attention. First, the District Court on the first remand of this case correctly concluded that the definition of “intelligence source” is satisfied only when the sources of information of a particular sort would be unwilling to provide it to the CIA in the absence of confidentiality. In a situation where it may be concluded that the CIA could reasonably expect to obtain certain kinds of information from persons without guaranteeing them confidentiality, but the agency, in order to avoid adverse public reaction, wishes to maintain secrecy, the “intelligence source” exemption would not apply.
Second, the agency’s professed fear that it will be able to invoke exemption 3 in the future only by presenting elaborate *101proof of the circumspection of its sources is unfounded. Much of the information obtained by the CIA obviously could only be gathered through some kind of covert activity. There is no question that the agency in general could not reasonably expect to obtain data of that type without guaranteeing secrecy to those who provide it. Thus, the agency will be able successfully to resist a request for disclosure of the identity of a source of such information merely by informing the court of the nature of the material involved.12 It is only in cases like the present, where a great deal of the information is not self-evidently sensitive, where the reasons why its sources would desire confidentiality are not obvious, and where the agency’s desire for secrecy seems to derive principally from fear of a public outcry resulting from revelation of the details of its past conduct,13 that the CIA will be obliged to adduce extrinsic evidence in order to demonstrate its entitlement to the statutory exemption.
Conclusion
For the foregoing reasons, the District Court’s determination regarding which of the individual researchers satisfy the “need-for-confidentiality” portion of the definition of “intelligence source” promulgated in Sims I is reversed. All other aspects of the court’s decision are affirmed. The case is remanded for expeditious reconsideration of the researchers’ statuses.
So ordered.
.Exemption 3 authorizes withholding of “matters that are ... specifically exempted from disclosure by statute (other than [5 U.S.C. § 552b]), provided that such statute (A) requires that the matters be withheld from the public in such a manner as to leave no discretion on the issue, or (B) establishes particular criteria for withholding or refers to particular types of matters to be withheld.” 5 U.S.C. § 552(b)(3) (1976). In invoking this provision, the CIA relied on the well-established rule that § 102(d)(3) of the National Security Act of 1947, 50 U.S.C. § 403(d)(3) (1976), which provides that “the Director of Central Intelligence shall be responsible for protecting intelligence sources and methods from unauthorized disclosure,” meets the statutory requirements. See, e.g., Gardels v. CIA, 689 F.2d 1100, 1103 (D.C. Cir.1982).
. Exemption 6 authorizes withholding of “personnel and medical files and similar files the disclosure of which would constitute a clearly unwarranted invasion of personal privacy.” 5 U.S.C. § 552(b)(6) (1976).
. Exemption 1 authorizes withholding of “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) (1976).
. See note 1 supra.
. The CIA never complied with the District Court’s repeated suggestions that, in order to obtain some evidence of the status of the individual researchers, the agency should contact them and ascertain their understanding of the terms of their past relationship with the CIA.
.See also McGehee v. CIA, 697 F.2d 1095, 1112 n. 78 (D.C.Cir.1983) (reiterating our commitment to the definition).
. A further reason for not automatically allowing the CIA to shield the identities of informants who request anonymity is the possibility of collusion between the agency and its sources. A rule that any informant who asks for a guarantee of confidentiality qualifies as an “intelligence source” (regardless of the nature of the information he provides) would make it easy for the agency to seek shelter in exemption 3 by simply “suggesting” to all of the persons with whom it deals that they sign a form expressing their desire for secrecy and then memorializing its accession to their “requests.”
. Our rejection of a per se rule that an informant qualifies as an “intelligence source” if and only if the agency can show that he himself requested and was promised confidentiality is reinforced by our sensitivity to the agency’s argument that keeping records of such pledges would be administratively burdensome.
. The projects ranged from the drafting of a manual on sleight-of-hand tricks to various experimental programs “involving the surreptitious administration of LSD to unwitting non-volunteer subjects ‘at all social levels, high and low, native American and foreign.’ ” See Final Report of the Select Comm, to Study Governmental Operations With Respect to Intelligence Activities, S.Rep. No. 755, 94th Cong., 2d Sess., Bk. I at 393 (1976); Defendant’s Answers to Plaintiffs’ Interrogatories of Nov. 13, 1980, at 10, reprinted in App. 73.
. In view of the fact that the MKULTRA program is now declassified, it would seem practicable and appropriate for the court on remand to request both parties to submit, on the public record, memoranda proposing alternative schemes for categorizing the information at issue in the case. Such a procedure would be consistent with the line of decisions in this circuit requiring maximum adversarial testing of claims of FOIA exemptions. See, e.g., Founding Church of Scientology v. National Security Agency, 610 F.2d 824, 832-33 (D.C. Cir.1979); Philiippi v. CIA, 546 F.2d 1009, 1013 (D.C.Cir.1976).
.The spirit in which the court should make this determination is described in Sims I, 642 F.2d at 571-72.
. Cf. Holy Spirit Ass'n v. CIA, 636 F.2d at 844 (“The index with respect to ... four documents does not specifically mention confidentiality, but we are satisfied from the descriptions that the information contained in those documents is of the type that the Agency could not obtain without a guarantee of confidentiality.” (footnote omitted)).
. The dissent takes issue with this characterization of the most plausible explanation for the agency’s desire for secrecy. The dissent argues that there already has been “substantial ‘public outcry’ about MKULTRA,” as if to suggest that the agency has little to fear from fuller revelation of the nature and purpose of its “experiments.” Our prior decision in this case flatly refutes this suggestion. As noted in Sims I, “the CIA destroyed most of its substantive records pertaining to the project in 1973.” 642 F.2d at 564. Indeed, it never has been seriously disputed that, for the most part, only “general descriptions” of the MKULTRA program are in the public domain. Id. The horrifying features of some of those descriptions, see note 9 supra, suggest that the agency has much to fear from the kind of fuller disclosures of its past behavior that discussions with the participating researchers might make possible.
The dissent goes on to suggest that there are “many reasons more plausible than fear of public outcry for the CIA’s resistance to releasing the names of the researchers.” However, the conduct of the agency before the District Court casts considerable doubt on each of the motives the dissent offers. The dissent argues that the CIA may wish (1) not to “embarrass” researchers, (2) to protect researchers from harm, or (3) to keep its promises to researchers. These possibilities are plainly belied by the fact that the agency repeatedly refused to contact the researchers to ascertain their understanding of their prior relationship with the agency or their wishes regarding further concealment of their participation in the program. See Sims I, 642 F.2d at 565-66; note 5 supra. And the agency’s express renunciation of reliance on exemption 1 renders implausible the suggestion that it believes that “revelation of the identities of the researchers could aid foreign intelligence operations in discovering sensitive information.”