delivered the opinion of the Court.

In No. 83-1075, we granted certiorari to decide whether § 102(d)(3) of the National Security Act of 1947, as incorporated in Exemption 3 of the Freedom of Information Act, •exempts from disclosure only those sources of intelligence information to which the Central Intelligence Agency had to guarantee confidentiality in order to obtain the information. In No. 83-1249, the cross-petition, we granted certiorari to decide whether the Freedom of Information Act requires the Agency to disclose the institutional affiliations of persons whose identities are exempt from disclosure as “intelligence sources.”

1 — 4

Between 1953 and 1966, the Central Intelligence Agency financed a wide-ranging project, code-named MKULTRA, concerned with “the research and development of chemical, biological, and radiological materials capable of employment in clandestine operations to control human behavior.”1 The *162program consisted of some 149 subprojects which the Agency-contracted out to various universities, research foundations, and similar institutions. At least 80 institutions and 185 private researchers participated. Because the Agency funded MKULTRA indirectly, many of the participating individuals were unaware that they were dealing with the Agency.

MKULTRA was established to counter perceived Soviet and Chinese advances in brainwashing and interrogation techniques. Over the years the program included various medical and psychological experiments, some of which led to untoward results.2 These aspects of MKULTRA surfaced publicly during the 1970’s and became the subject of executive and congressional investigations.3

On August 22, 1977, John C. Sims, an attorney, and Sidney M. Wolfe, M.D., the director of the Public Citi*163zen Health Research Group,4 filed a request with the Central Intelligence Agency seeking certain information about MKULTRA. Respondents invoked the Freedom of Information Act (FOIA), 5 U. S. C. § 552. Specifically, respondents sought the grant proposals and contracts awarded under the MKULTRA program and the names of the institutions and individuals that had performed research.5

Pursuant to respondents’ request, the Agency made available to respondents all of the MKULTRA grant proposals and contracts. Citing Exemption 3 of the FOIA, 5 U. S. C. § 552(b)(3)(B),6 however, the Agency declined to disclose the names of all individual researchers and 21 institutions.7 Exemption 3 provides that an agency need not disclose “matters that are . . . specifically exempted from disclosure by statute . . . provided that such statute . . . refers to par*164ticular types of matters to be withheld.” Ibid. The Agency relied on § 102(d)(3) of the National Security Act of 1947, 61 Stat. 498, 50 U. S. C. § 403(d)(3), which states that

“the Director of Central Intelligence shall be responsible for protecting intelligence sources and methods from unauthorized disclosure.”

Dissatisfied with the Agency’s limited disclosure, respondents filed suit under the FOIA, 5 U. S. C. § 552(a)(4)(B), in the United States District Court for the District of Columbia. That court ordered disclosure of the withheld names, holding that the MKULTRA researchers and affiliated institutions were not “intelligence sources” within the meaning of § 102(d)(3). 479 F. Supp. 84 (1979).

On appeal, the United States Court of Appeals concluded, as had the District Court, that § 102(d)(3) qualifies as a withholding statute under Exemption 3 of the FOIA. The court held, however, that the District Court’s analysis of that statute under the FOIA lacked a coherent definition of “intelligence sources.” Accordingly, it remanded the case for reconsideration in light of the following definition:

“[A]n ‘intelligence source’ is 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.” 206 U. S. App. D. C. 157, 166, 642 F. 2d 562, 571 (1980).

On remand, the District Court applied this definition and ordered the Agency to disclose the names of 47 researchers and the institutions with which they had been affiliated. The court rejected respondents’ contention that the MKULTRA research was not needed to perform the Agency’s intelligence function, explaining that

“[i]n view of the agency’s concern that potential foreign enemies could be engaged in similar research and the *165desire to take effective counter-measures, . . . [the Agency] could reasonably determine that this research was needed for its intelligence function.” App. to Pet. for Cert. in No. 83-1075, pp. 22a-23a.

The court then turned to the question whether the Agency could show, as the Court of Appeals’ definition requires, that it could not reasonably have expected to obtain the information supplied by the MKULTRA sources without guaranteeing confidentiality to them. The court concluded that the Agency’s policy of considering its relationships with MKULTRA researchers as confidential was not sufficient to satisfy the Court of Appeals’ definition because “the chief desire for confidentiality was on the part of the CIA.” Id., at 24a. The court recognized that some of the researchers had sought, and received, express guarantees of confidentiality from the Agency, and as to those held that their identities need not be disclosed. The court also exempted other researchers from disclosure on the ground that their work for the Agency, apart from MKULTRA, required that their identities remain secret in order not to compromise the Agency’s intelligence networks in foreign countries. Id., at 26a-27a, 30a-31a. Finally, the court held that there was no need to disclose the institutional affiliations of the individual researchers whose identities were exempt from disclosure; this withholding was justified by the need to eliminate the unnecessary risk that such intelligence sources would be identified indirectly. Id., at 27a, 34a.

Both the Agency and respondents appealed. The Court of Appeals affirmed that part of the District Court’s judgment exempting from disclosure the institutional affiliations of individual researchers found to be intelligence sources. However, it reversed the District Court’s ruling with respect to which individual researchers satisfied “the need-for-confidentiality” aspect of its formulation of exempt “intelligence sources.” 228 U. S. App. D. C. 269, 275, 709 F. 2d 95, 101 (1983).

*166At the outset, the court rejected the suggestion that it reconsider the definition of “intelligence sources.” Id., at 271, 709 F. 2d, at 97. The court then criticized the District Court for focusing its inquiry on whether the Agency had in fact promised confidentiality to individual researchers. The court held that the District Court’s decision automatically to exempt from disclosure those researchers to whom confidentiality had been promised was erroneous; it directed the District Court on remand to focus its inquiry on whether the Agency offered sufficient proof that it needed to cloak its efforts in confidentiality in order to obtain the type of information provided by the researcher. Only upon such a showing would the individual qualify as an “intelligence source” exempt from disclosure under the FOIA.8

We granted certiorari, 465 U. S. 1078 (1984) and 467 U. S. 1240 (1984). We now reverse in part and affirm in part.

t — I

No. 83-1075

A

The mandate of the FOIA calls for broad disclosure of Government records.9 Congress recognized, however, that *167public disclosure is not always in the public interest and thus provided that agency records may be withheld from disclosure under any of the nine exemptions defined in 5 U. S. C. § 552(b). Under Exemption 3 disclosure need not be made as to information “specifically exempted from disclosure by statute” if the statute affords the agency no discretion on disclosure, § 552(b)(3)(A), establishes particular criteria for withholding the information, or refers to the particular types of material to be withheld, § 552(b)(3)(B).

The question in No. 83-1075 is twofold: first, does § 102(d)(3) of the National Security Act of 1947 constitute a statutory exemption to disclosure within the meaning of Exemption 3; and second, are the MKULTRA researchers included within § 102(d)(3)’s protection of “intelligence sources.”

B

Congress has made the Director of Central Intelligence “responsible for protecting intelligence sources and methods from unauthorized disclosure.” 50 U. S. C. §403(d)(3). As part of its postwar reorganization of the national defense system, Congress chartered the Agency with the responsibility of coordinating intelligence activities relating to national security.10 In order to carry out its mission, the Agency was expressly entrusted with protecting the heart of all intelligence operations — “sources and methods.”

Section 102(d)(3) of the National Security Act of 1947, which calls for the Director of Central Intelligence to protect “intelligence sources and methods,” clearly “refers to particular types of matters,” 5 U. S. C. § 552(b)(3)(B), and thus qualifies as a withholding statute under Exemption 3. The “plain meaning” of the relevant statutory provisions is sufficient to resolve the question, see, e. g., Garcia v. United *168States, 469 U. S. 70, 75 (1984); United States v. Weber Aircraft Corp., 465 U. S. 792, 798 (1984). Moreover, the legislative history of the FOIA confirms that Congress intended § 102(d)(3) to be a withholding statute under Exemption 3.11 Indeed, this is the uniform view among other federal courts.12

Our conclusion that § 102(d)(3) qualifies as a withholding statute under Exemption 3 is only the first step of the inquiry. Agency records are protected under § 102(d)(3) only to the extent they contain “intelligence sources and methods” or if disclosure would reveal otherwise protected information.

C

Respondents contend that the Court of Appeals’ definition of “intelligence sources,” focusing on the need to guarantee confidentiality in order to obtain the type of information desired, draws the proper line with respect to intelligence sources deserving exemption from the FOIA. The plain meaning of the statutory language, as well as the legislative history of the National Security Act, however, indicates that Congress vested in the Director of Central Intelligence very *169broad authority to protect all sources of intelligence information from disclosure. The Court of Appeals’ narrowing of this authority not only contravenes the express intention of Congress, but also overlooks the practical necessities of modem intelligence gathering — the very reason Congress entrusted this Agency with sweeping power to protect its “intelligence sources and methods.”

We begin with the language of § 102(d)(3). Baldrige v. Shapiro, 455 U. S. 345, 356 (1982); Steadman v. SEC, 450 U. S. 91, 97 (1981). Section 102(d)(3) specifically authorizes the Director of Central Intelligence to protect “intelligence sources and methods” from disclosure. Plainly the broad sweep of this statutory language comports with the nature of the Agency’s unique responsibilities. To keep informed of other nations’ activities bearing on our national security the Agency must rely on a host of sources. At the same time, the Director must have the authority to shield those Agency activities and sources from any disclosures that would unnecessarily compromise the Agency’s efforts.

The “plain meaning” of § 102(d)(3) may not be squared with any limiting definition that goes beyond the requirement that the information fall within the Agency’s mandate to conduct foreign intelligence. Section 102(d)(3) does not state, as the Court of Appeals’ view suggests, that the Director of Central Intelligence is authorized to protect intelligence sources only if such protection is needed to obtain information that otherwise could not be acquired. Nor did Congress state that only confidential or nonpublic intelligence sources are protected.13 Section 102(d)(3) contains no such limiting language. Congress simply and pointedly protected all sources *170of intelligence that provide, or are engaged to provide, information the Agency needs to perform its statutory duties with respect to foreign intelligence. The plain statutory language is not to be ignored. Weber Aircraft Corp., supra, at 798.

The legislative history of § 102(d)(3) also makes clear that Congress intended to give the Director of Central Intelligence broad power to protect the secrecy and integrity of the intelligence process. The reasons are too obvious to call for enlarged discussion; without such protections the Agency would be virtually impotent.

Enacted shortly after World War II, § 102(d)(3) of the National Security Act of 1947 established the Agency and empowered it, among other things, “to correlate and evaluate intelligence relating to the national security.” 50 U. S. C. § 403(d)(3). The tragedy of Pearl Harbor and the reported deficiencies in American intelligence during the course of the war convinced the Congress that the country’s ability to gather and analyze intelligence, in peacetime as well as in war, must be improved. See, e. g., H. R. Rep. No. 961, 80th Cong., 1st Sess., 3-4 (1947); S. Rep. No. 239, 80th Cong., 1st Sess., 2 (1947).

Congress knew quite well that the Agency would gather intelligence from almost an infinite variety of diverse sources. Indeed, one of the primary reasons for creating the Agency was Congress’ recognition that our Government would have to shepherd and analyze a “mass of information” in order to safeguard national security in the postwar world. See ibid. Witnesses with broad experience in the intelligence field testified before Congress concerning the practical realities of intelligence work. Fleet Admiral Nimitz, for example, explained that “intelligence is a composite of authenticated and evaluated information covering not only the armed forces establishment of a possible enemy, but also his industrial capacity, racial traits, religious beliefs, and other related aspects.” National Defense Establishment: *171Hearings on S. 758 before the Senate Committee on Armed Services, 80th Cong., 1st Sess., 132 (1947) (Senate Hearings). General Vandenberg, then the Director of the Central Intelligence Group, the Agency’s immediate predecessor, emphasized that “foreign intelligence [gathering] consists of securing all possible data pertaining to foreign governments or the national defense and security of the United States.” Id., at 497.14

Witnesses spoke of the extraordinary diversity of intelligence sources. Allen Dulles, for example, the Agency’s third Director, shattered the myth of the classic “secret agent” as the typical intelligence source, and explained that “American businessmen and American professors and Americans of all types and descriptions who travel around the world are one of the greatest repositories of intelligence that we have.” National Security Act of 1947: Hearings on H. R. 2319 before the House Committee on Expenditures in the Executive Departments, 80th Cong., 1st Sess., 22 (1947) (Closed House Hearings).15 In a similar vein, General Van-denberg spoke of “the great open sources of information upon which roughly 80 percent of intelligence should be based,” and identified such sources as “books, magazines, technical and scientific surveys, photographs, commercial analyses, newspapers, and radio broadcasts, and general information from *172people with knowledge of affairs abroad.” Senate Hearings, at 492.

Congress was also well aware of the importance of secrecy in the intelligence field. Both General Vandenberg and Allen Dulles testified about the grim consequences facing intelligence sources whose identities became known. See Closed House Hearings, at 10-11, 20. Moreover, Dulles explained that even American citizens who freely supply intelligence information “close up like a clam” unless they can hold the Government “responsible to keep the complete security of the information they turn over.” Id., at 22.16 Congress was plainly alert to the need for maintaining confidentiality — both Houses went into executive session to consider the legislation creating the Agency — a rare practice for congressional sessions. See n. 15, supra.

Against this background highlighting the requirements of effective intelligence operations, Congress expressly made the Director of Central Intelligence responsible for “protecting intelligence sources and methods from unauthorized disclosure.” This language stemmed from President Truman’s Directive of January 22, 1946, 11 Fed. Reg. 1337, in which he established the National Intelligence Authority and the Central Intelligence Group, the Agency’s predecessors. These institutions were charged with “assuring] the most effective accomplishment of the intelligence mission related to the national security,” ibid., and accordingly made “responsible *173for fully protecting intelligence sources and methods,” id., at 1339. The fact that the mandate of § 102(d)(3) derives from this Presidential Directive reinforces our reading of the legislative history that Congress gave the Agency broad power to control the disclosure of intelligence sources.

Ill

A

Applying the definition of “intelligence sources” fashioned by the Congress in § 102(d)(3), we hold that the Director of Central Intelligence was well within his statutory authority to withhold the names of the MKULTRA researchers from disclosure under the FOIA. The District Court specifically ruled that the Agency “could reasonably determine that this research was needed for its intelligence function,”17 and the Court of Appeals did not question this ruling. Indeed, the record shows that the MKULTRA research was related to the Agency’s intelligence-gathering function in part because it revealed information about the ability of foreign governments to use drugs and other biological, chemical, or physical agents in warfare or intelligence operations against adversaries. During the height of the cold war period, the Agency was concerned, not without reason, that other countries were charting new advances in brainwashing and interrogation techniques.18

Consistent with its responsibility to maintain national security, the Agency reasonably determined that major research *174efforts were necessary in order to keep informed of our potential adversaries’ perceived threat. We thus conclude that MKULTRA researchers are “intelligence sources” within the broad meaning of § 102(d)(3) because these persons provided, or were engaged to provide, information the Agency needs to fulfill its statutory obligations with respect to foreign intelligence.

Respondents’ belated effort to question the Agency’s authority to engage scientists and academic researchers as intelligence sources must fail. The legislative history of § 102(d)(3) indicates that Congress was well aware that the Agency would call on a wide range and variety of sources to provide intelligence. Moreover, the record developed in this case confirms the obvious importance of scientists and other researchers as American intelligence sources. Notable examples include those scientists and researchers who pioneered the use of radar during World War II as well as the group which took part in the secret development of nuclear weapons in the Manhattan Project. See App. 43; App. to Pet. for Cert. in No. 83-1075, p. 88a.19

B

The Court of Appeals narrowed the Director’s authority under § 102(d)(3) to withhold only those “intelligence sources” who supplied the Agency with information unattainable without guaranteeing confidentiality. That crabbed reading of the statute contravenes the express language of § 102(d)(3), the statute’s legislative history, and the harsh realities of the present day. The dangerous consequences of that narrowing of the statute suggest why Congress chose to vest the *175Director of Central Intelligence with the broad discretion to safeguard the Agency’s sources and methods of operation.

The Court of Appeals underestimated the importance of providing intelligence sources with an assurance of confidentiality that is as absolute as possible. Under the court’s approach, the Agency would be forced to disclose a source whenever a court determines, after the fact, that the Agency could have obtained the kind of information supplied without promising confidentiality.20 This forced disclosure of the identities of its intelligence sources could well have a devastating impact on the Agency’s ability to carry out its mission. “The Government has a compelling interest in protecting both the secrecy of information important to our national security and the appearance of confidentiality so essential to the effective operation of our foreign intelligence service.” Snepp v. United States, 444 U. S. 507, 509, n. 3 (1980) (per curiam). See Haig v. Agee, 453 U. S. 280, 307 (1981). If potentially valuable intelligence sources come to think that the Agency will be unable to maintain the confidentiality of its relationship to them, many could well refuse to supply information to the Agency in the first place.

Even a small chance that some court will order disclosure of a source’s identity could well impair intelligence gathering and cause sources to “close up like a clam.” To induce some sources to cooperate, the Government must tender as absolute an assurance of confidentiality as it possibly can. “The continued availability of [intelligence] sources depends upon the CIA’s ability to guarantee the security of information *176that might compromise them and even endanger [their] personal safety.” Snepp v. United States, supra, at 512.

We seriously doubt whether a potential intelligence source will rest assured knowing that judges, who have little or no background in the delicate business of intelligence gathering, will order his identity revealed only after examining the facts of the case to determine whether the Agency actually needed to promise confidentiality in order to obtain the information. An intelligence source will “not be concerned with the underlying rationale for disclosure of” his cooperation if it was secured “under assurances of confidentiality.” Baldrige v. Shapiro, 455 U. S., at 361. Moreover, a court’s decision whether an intelligence source will be harmed if his identity is revealed will often require complex political, historical, and psychological judgments. See, e. g., Fitzgibbon v. CIA, 578 F. Supp. 704 (DC 1983). There is no reason for a potential intelligence source, whose welfare and safety may be at stake, to have great confidence in the ability of judges to make those judgments correctly.

The Court of Appeals also failed to recognize that when Congress protected “intelligence sources” from disclosure, it was not simply protecting sources of secret intelligence information. As noted above, Congress was well aware that secret agents as depicted in novels and the media are not the typical intelligence source; many important sources provide intelligence information that members of the public could also obtain. Under the Court of Appeals’ approach, the Agency could not withhold the identity of a source of intelligence if that information is also publicly available. This analysis ignores the realities of intelligence work, which often involves seemingly innocuous sources as well as unsuspecting individuals who provide valuable intelligence information.

Disclosure of the subject matter of the Agency’s research efforts and inquiries may compromise the Agency’s ability to gather intelligence as much as disclosure of the identities of intelligence sources. A foreign government can learn a great deal about the Agency’s activities by knowing the *177public sources of information that interest the Agency. The inquiries pursued by the Agency can often tell our adversaries something that is of value to them. See 228 U. S. App. D. C., at 277, 709 F. 2d, at 103 (Bork, J., concurring in part and dissenting in part). For example, disclosure of the fact that the Agency subscribes to an obscure but publicly available Eastern European technical journal could thwart the Agency’s efforts to exploit its value as a source of intelligence information. Similarly, had foreign governments learned the Agency was using certain public journals and ongoing open research projects in its MKULTRA research of “brainwashing” and possible countermeasures, they might have been able to infer both the general nature of the project and the general scope that the Agency’s inquiry was taking.21

C

The “statutory mandate” of § 102(d)(3) is clear: Congress gave the Director wide-ranging authority to “protec[t] intelligence sources and methods from unauthorized disclosure.” Snepp v. United States, supra, at 509, n. 3. An intelligence source provides, or is engaged to provide, information the Agency needs to fulfill its statutory obligations. The record establishes that the MKULTRA researchers did in fact provide the Agency with information related to the Agency’s intelligence function. We therefore hold that the Director was authorized to withhold the identities of these researchers from disclosure under the FOIA.

M <

No. 83-1249

The cross-petition, No. 83-1249, calls for decision on whether the District Court and the Court of Appeals eor-*178rectly ruled that the Director of Central Intelligence need not disclose the institutional affiliations of the MKULTRA researchers previously held to be “intelligence sources.” Our conclusion that the MKULTRA researchers are protected from disclosure under § 102(d)(3) renders unnecessary any extended discussion of this discrete issue.

In exercising the authority granted by Congress in § 102(d)(3), the Director must, of course, do more than simply withhold the names of intelligence sources. Such withholding, standing alone, does not carry out the mandate of Congress. Foreign intelligence services have an interest in knowing what is being studied and researched by our agencies dealing with national security and by whom it is being done. Foreign intelligence services have both the capacity to gather and analyze any information that is in the public domain and the substantial expertise in deducing the identities of intelligence sources from seemingly unimportant details.

In this context, the very nature of the intelligence apparatus of any country is to try to find out the concerns of others; bits and pieces of data “may aid in piecing together bits of other information even when the individual piece is not of obvious importance in itself.” Halperin v. CIA, 203 U. S. App. D. C. 110, 116, 629 F. 2d 144, 150 (1980). Thus,

“ ‘[w]hat may seem trivial to the uninformed, may appear of great moment to one who has a broad view of the scene and may put the questioned item of information in its proper context.’” Halkin v. Helms, 194 U. S. App. D. C. 82, 90, 598 F. 2d 1, 9 (1978), quoting United States v. Marchetti, 466 F. 2d 1309, 1318 (CA4), cert. denied, 409 U. S. 1063 (1972).

Accordingly, the Director, in exercising his authority under § 102(d)(3), has power to withhold superficially innocuous information on the ground that it might enable an observer to discover the identity of an intelligence source. See, e. g., *179Gardels v. CIA, 223 U. S. App. D. C. 88, 91-92, 689 F. 2d 1100, 1103-1104 (1982); Halperin v. CIA, supra, at 113, 629 F. 2d, at 147.

Here the Director concluded that disclosure of the institutional affiliations of the MKULTRA researchers could lead to identifying the researchers themselves and thus the disclosure posed an unacceptable risk of revealing protected “intelligence sources.”22 The decisions of the Director, who must of course be familiar with “the whole picture,” as judges are not, are worthy of great deference given the magnitude of the national security interests and potential risks at stake. It is conceivable that the mere explanation of why information must be withheld can convey valuable information to a foreign intelligence agency.

The District Court, in a ruling affirmed by the Court of Appeals, permitted the Director to withhold the institutional affiliations of the researchers whose identities were exempt from disclosure on the ground that disclosure of “the identities of the institutions . . . might lead to the indirect disclosure of” individual researchers. App. to Pet. for Cert. in No. 83-1075, p. 27a. This conclusion is supported by the record.23 The Director reasonably concluded that an ob*180server who is knowledgeable about a particular intelligence research project, like MKULTRA, could, upon learning that research was performed at a certain institution, often deduce the identities of the individual researchers who are protected “intelligence sources.” The FOIA does not require disclosure under such circumstances.

Respondents contend that because the Agency has already revealed the names of many of the institutions at which MKULTRA research was performed, the Agency is somehow estopped from withholding the names of others. This suggestion overlooks the political realities of intelligence operations in which, among other things, our Government may choose to release information deliberately to “send a message” to allies or adversaries.24 Congress did not mandate the withholding of information that may reveal the identity of an intelligence source; it made the Director of Central Intelligence responsible only for protecting against unauthorized disclosures.

The national interest sometimes makes it advisable, or even imperative, to disclose information that may lead to the identity of intelligence sources. And it is the responsibility of the Director of Central Intelligence, not that of the judiciary, to weigh the variety of complex and subtle factors in determining whether disclosure of information may lead to an unacceptable risk of compromising the Agency’s intelligence-gathering process. Here Admiral Turner, as Director, decided that the benefits of disclosing the identities of institutions that had no objection to disclosure outweighed the costs *181of doing so. But Congress, in § 102(d)(3), entrusted this discretionary authority to the Director, and the fact that Admiral Turner made that determination in 1978 does not bind his successors to make the same determination, in a different context, with respect to institutions requesting that their identities not be disclosed. See, e. g., Salisbury v. United States, 223 U. S. App. D. C. 243, 248, 690 F. 2d 966, 971 (1982).

V

We hold that the Director of Central Intelligence properly invoked § 102(d)(3) of the National Security Act of 1947 to withhold disclosure of the identities of the individual MKULTRA researchers as protected “intelligence sources.” We also hold that the FOIA does not require the Director to disclose the institutional affiliations of the exempt researchers in light of the record which supports the Agency’s determination that such disclosure would lead to an unacceptable risk of disclosing the sources’ identities.

Accordingly, we reverse that part of the judgment of the Court of Appeals regarding the disclosure of the individual researchers and affirm that part of the judgment pertaining to disclosure of the researchers’ institutional affiliations.

It is so ordered.

with whom Justice Brennan joins, concurring in the result.

To give meaning to the term “intelligence source” as it is used in § 102(d)(3) of the National Security Act of 1947, the Court today correctly concludes that the very narrow definition offered by the Court of Appeals is incorrect.1 That the *182Court of Appeals erred does not, however, compel the conclusion that the Agency’s sweeping alternative definition is in fact the correct one.2 The Court nonetheless simply adopts wholesale the Agency’s definition of “intelligence source.” That definition is mandated neither by the language or legislative history of any congressional Act, nor by legitimate policy considerations, and it in fact thwarts congressional efforts to balance the public’s interest in information and the Government’s need for secrecy. I therefore decline to join the opinion of the Court.

I

The Freedom of Information Act (FOIA or Act) established a broad mandate for disclosure of governmental information by requiring that all materials be made public “unless explicitly allowed to be kept secret by one of the exemptions . . . .” S. Rep. No. 813, 89th Cong., 1st Sess., 10 (1965). The Act requires courts to review de novo agency claims of exemption, and it places on the agency the burden of defending its withholding of information. 5 U. S. C. § 552(a)(4)(B). Congress, it is clear, sought to assure that the Government would not operate behind a véil of secrecy, and it narrowly tailored the exceptions to the fundamental goal of disclosure.

Two of these few exceptions are at issue in this case. The first, on which the Court focuses, is Exemption 3, which exempts information “specifically exempted from disclosure by statute,” if the statute affords the agency no discretion on disclosure, § 552(b)(3)(A), establishes particular criteria for withholding the information, § 552(b)(3)(B), or refers to the particular types of material to be withheld, ibid. The Court *183quite rightly identifies § 102(d)(3) of the National Security Act as a statutory exemption of the kind to which Exemption 3 refers; that section places with the Director of Central Intelligence the responsibility for “protecting intelligence sources and methods from unauthorized disclosure.”

A second exemption, known as Exemption 1, 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). This latter Exemption gives to the Executive Branch the authority to define material that will not be disclosed, subject of course to congressional amendment of the Exemption. Agency decisions to withhold are subject to de novo review in the courts, which must ascertain whether documents are correctly classified, both substantively and procedurally.

Exemption 1 is the keystone of a congressional scheme that balances deference to the Executive’s interest in maintaining secrecy with continued judicial and congressional oversight. In the past, Congress has taken affirmative steps to make clear the importance of this oversight. See n. 5, infra. Exemption 1 allows the Government to protect from the scrutiny of this Nation’s enemies classes of information that warrant protection, as long as the Government proceeds through a publicly issued, congressionally scrutinized, and judicially enforced executive order. See Hearing on Executive Order on Security Classification before the Subcommittee of the Committee on Government Operations of the House of Representatives, 97th Cong., 2d Sess. (1982) (Hearing).

Exemption 1 thus plays a crucial role in the protection of Central Intelligence Agency information. That the Court does not mention this Exemption even once, in the course of its lengthy analysis on the policy reasons for broadly inter*184preting the “intelligence source” provision, is extraordinary. By focusing myopically on the single statutory provision on which the Agency has chosen to rely in asserting its secrecy right, the Court rewards the Agency’s decision not to invoke Exemption 1 in these cases.3 Of course, the Agency may fairly assert any possible ground for decision, and it has no duty to select that which is narrowest. But the Court, intent to assure that important information is protected, today plays into the Agency’s hands by stretching the “intelligence source” exception beyond its natural limit; it does so while simply ignoring the fact that the information sought could properly have been withheld on other grounds — on which the Agency chose not to rely. The cost of acceding to the Agency’s litigation strategy, rather than undertaking a thorough analysis of the entire statutory scheme, is to mangle, seriously, a carefully crafted statutory scheme.

J — l HH

I turn, then, to consider in light of this statutory framework the Court’s analysis of Exemption 3. After concluding that Exemption 3 incorporates § 102(d)(3) as a withholding provision, the Court sets out to define the term “intelligence source.” First, it looks to the “plain meaning” of the phrase and concludes that an “intelligence source” is self-evidently the same as an “information source.” Ante, at 169-170. Second, the Court looks to the legislative history. Pulling *185together pieces of testimony from congressional hearings on the need to establish a centralized agency to gather information, it concludes that Congress knew that the Agency would collect information from diverse sources, and that “Congress was plainly alert to the need for maintaining confidentiality” so as not to lose covert sources of information. Ante, at 172; see also Brief for Petitioners in No. 83-1075, pp. 18-21. Third, the Court chastises the Court of Appeals for adopting a “crabbed” reading of the statute and explains how, as a policy matter, the “forced disclosure of the identities of its intelligence sources could well have a devastating impact on the Agency’s ability to carry out its mission.” Ante, at 175; see also Brief for Petitioners in No. 83-1075, p. 31. The Court offers examples of highly sensitive information that, under the lower court’s reading, might be disclosed. See ante, at 176-177; see also Brief for Petitioners in No. 83-1075, pp. 34-37.

Before this Court, the Agency argued against the lower court’s definition of “intelligence source,” substituted its own sweeping offering, and then recounted a litany of national security nightmares that would surely befall this Nation under any lesser standard; today the Court simply buys this analysis. But the Court thereby ignores several important facts. First, the holding today is not compelled by the language of the statute, nor by the legislative history on which the Court relies. Second, the Court of Appeals’ definition is not the sole alternative to the one adopted by the Court today. Third, as noted, supra, other broad exemptions to FOIA exist, and a holding that this Exemption 3 exception does not apply here would in no way pose the risk of broad disclosure the Agency suggests. The Court’s reliance on the Nation’s national security interests is simply misplaced given that the “intelligence source” exemption in the National Security Act is far from the Agency’s exclusive, or most potent, resource for keeping probing eyes from secret documents. In its haste to adopt the Agency’s sweeping defini*186tion, the Court completely bypasses a considerably more rational definition that comports at least as well with the statutory language and legislative history, and that maintains the congressionally imposed limits on the Agency’s exercise of discretion in this area.

To my mind, the phrase “intelligence source” refers only to sources who provide information either on an express or implied promise of confidentiality, and the exemption protects such information and material that would lead to disclosure of such information. This reading is amply supported by the language of the statute and its history.

First, I find reliance on “plain meaning” wholly inappropriate. The heart of the issue is whether the term “intelligence source” connotes that which is confidential or clandestine, and the answer is far from obvious. The term is readily susceptible of many interpretations, and in the past the Government itself has defined the term far less broadly than it now does before this Court. In testimony before the House Subcommittee on Government Operations on President Reagan’s Exemption 1 Executive Order, Steven Garfinkel, Director of the Information Security Oversight Office, explained that the term “intelligence source” is narrow and does not encompass even all confidential sources of information:

“[CJertain of these sources are not ‘intelligence sources.’ They are not involved in intelligence agencies or in intelligence work. They happen to be sources of information received by these agencies in confidence.” Hearing, at 204.

The current administration’s definition of the term “intelligence source” as used in its Executive Order does not, of course, control our interpretation of a longstanding statute. But the fact that the same administration has read the phrase in different ways for different purposes certainly undercuts the Court’s argument that the phrase has any single and readily apparent definition.

*187“[PJlain meaning, like beauty, is sometimes in the eye of the beholder,” Florida Power & Light Co. v. Lorion, 470 U. S. 729, 737 (1985), and in an instance such as this one, in which the term at issue carries with it more than one plausible meaning, it is simply inappropriate to select a single reading and label it the “plain meaning.” The Court, like the Government, argues that the statute does not say “confidential source,” as it might were its scope limited to sources who have received an implied or express promise of confidentiality. See ante, at 169, and n. 13; Brief for Petitioners in No. 83-1075, p. 16. However, the statute also does not say “information source” as it might were it meant to define the class of material that the Court identifies. I therefore reject the Court’s basic premise that the language at issue necessarily has but a single, obvious interpretation.

Nor does the legislative history suggest anything other than a congressional desire to protect those individuals who might either be harmed or silenced should their identities or assistance become known. The congressional hearings quoted by the Court, and by the Government in its brief, focus on Congress’ concern about the “deadly peril” faced by intelligence sources if their identities were revealed, and about the possibility that those sources would “ ‘close up like a clam’” without protection. See ante, at 172; Brief for Petitioners in No. 83-1075, p. 20. These concerns are fully addressed by preventing disclosure of the identities of sources who might face peril, or cease providing information, if their identities were known, and of other information that might lead an observer to identify such sources. That, to my mind, is the start and finish of the exemption for an “intelligence source” — one who contributes information on an implicit understanding or explicit assurance of confidentiality, as well as information that could lead to such a source.4

*188This reading of the “intelligence source” language also fits comfortably within the statutory scheme as a whole, as the Court’s reading does not. I focus, at the outset, on the recent history of FOIA Exemption 1 and particularly on the way in which recent events reflect Congress’ ongoing effort to constrain agency discretion of the kind endorsed today. The scope of Exemption 1 is defined by the Executive, and its breadth therefore quite naturally fluctuates over time. For example, at the time this FOIA action was begun, Executive Order 12065, promulgated by President Carter, was in effect. That Order established three levels of secrecy — top secret, secret, and confidential — the lowest of which, “confidential,” was “applied to information, the unauthorized disclosure of which reasonably could be expected to cause identifiable damage to the national security.” 3 CFR 191 (1979).

The Order also listed categories of information that could be considered for classification, including “military plans, weapons, or operations,” “foreign government information,” and “intelligence activities [and] sources.” Id., at 193. As it is now, nondisclosure premised on Exemption 1 was subject to judicial review. A court reviewing an Agency claim to withholding under Exemption 1 was required to determine de novo whether the document was properly classified and whether it substantively met the criteria in the Executive Order. If the claim was that the document or information in it contained military plans, for example, a court was required to determine whether the document was classified, whether it in fact contained such, information and whether disclosure of the document reasonably could be expected to cause at least identifiable damage to national security. The burden was on the Agency to make this showing. At one time, this *189Court believed that the Judiciary was not qualified to undertake this task. See EPA v. Mink, 410 U. S. 73 (1973), discussed in n. 5, infra. Congress, however, disagreed, overruling both a decision of this Court and a Presidential veto to make clear that precisely this sort of judicial role is essential if the balance that Congress believed ought to be struck between disclosure and national security is to be struck in practice.5

Today’s decision enables the Agency to avoid making the showing required under the carefully crafted balance embodied in Exemption 1 and thereby thwarts Congress’ effort to limit the Agency’s discretion. The Court identifies two categories of information — the identity of individuals or entities, whether or not confidential, that contribute material related *190to Agency information gathering, and material that might enable an observer to discover the identity of such a “source” — and rules that all such information is per se subject to withholding as long as it is related to the Agency’s “intelligence function.” The Agency need not even assert that disclosure will conceivably affect national security, much less that it reasonably could be expected to cause at least identifiable damage. It need not classify the information, much less demonstrate that it has properly been classified. Similarly, no court may review whether the source had, or would have had, any interest in confidentiality, or whether disclosure of the information would have any effect on national security. No court may consider whether the information is properly classified, or whether it fits the categories of the Executive Order. By choosing to litigate under Exemption 3, and by receiving this Court’s blessing, the Agency has cleverly evaded all these carefully imposed congressional requirements.6

If the class thus freed from judicial review were carefully defined, this result conceivably could make sense. It could *191mean that Congress had decided to slice out from all the Agency’s possible documents a class of material that may always be protected, no matter what the scope of the existing executive order. But the class that the Court defines is boundless. It is difficult to conceive of anything the Central Intelligence Agency might have within its many files that might not disclose or enable an observer to discover something about where the Agency gathers information. Indeed, even newspapers and public libraries, road maps and telephone books appear to fall within the definition adopted by the Court today. The result is to cast an irrebuttable presumption of secrecy over an expansive array of information in Agency files, whether or not disclosure would be detrimental to national security, and to rid the Agency of the burden of making individualized showings of compliance with an executive order. Perhaps the Court believes all Agency documents should be susceptible to withholding in this way. But Congress, it must be recalled, expressed strong disagreement by passing, and then amending, Exemption 1. In light of the Court’s ruling, the Agency may nonetheless circumvent the procedure Congress has developed and thereby undermine this explicit effort to keep from the Agency broad and unreviewable discretion over an expansive class of information.

HH I — I

The Court today reads its own concerns into the single phrase, “intelligence source.” To justify its expansive reading of these two words in the National Security Act the Court explains that the Agency must be wary, protect itself, and not allow observers to learn either of its information resources or of the topics of its interest. “Disclosure of the subject matter of the Agency’s research efforts and inquiries may compromise the Agency’s ability to gather intelligence as much as disclosure of the identities of intelligence sources,” ante, at 176, the Court observes, and the “intelligence source” *192exemption must bear the weight of that concern as well. That the Court points to no legislator or witness before Congress who expressed a concern for protecting such information through this provision is irrelevant to the Court. That each of the examples the Court offers of material that might disclose a topic of interest, and that should not be disclosed, could be protected through other existing statutory provisions, is of no moment.7 That the public already knows all about the MKULTRA project at issue in this case, except for the names of the researchers, and therefore that the Court’s concern about disclosure of the Agency’s “topics of interest” argument is not appropriate to this case, is of no consequence. And finally, that the Agency now has virtually unlimited discretion to label certain information “secret,” in contravention of Congress’ explicit efforts to confine the Agency’s discretion both substantively and procedurally, is of no importance. Instead, simply because the Court can think of information that it believes should not be disclosed, and that might otherwise not fall within this exemption, the Court undertakes the task of interpreting the exemption to cover that information. I cannot imagine the canon of statutory construction upon which this reasoning is based.

*193Congress gave to the Agency considerable discretion to decide for itself whether the topics of its interest should remain secret, and through Exemption 1 it provided the Executive with the means to protect such information. If the Agency decides to classify the identities of nonconfidential contributors of information so as not to reveal the subject matter or kinds of interests it is pursuing, it may seek an Exemption 1 right to withhold. Under Congress’ scheme, that is properly a decision for the Executive. It is not a decision for this Court. Congress has elsewhere identified particular types of information that it believes may be withheld regardless of the existence of an executive order, such as the identities of Agency employees, or, recently, the contents of Agency operational files. See 50 U. S. C. §403g (exempting from disclosure requirements the organization, functions, names, official titles, salaries, or numbers of personnel employed by the Agency); Central Intelligence Agency Information Act, Pub. L. 98-477, § 701(a), 98 Stat. 2209, 50 U. S. C. § 431 (1982 ed., Supp. III) (exempting the Agency’s operational files from disclosure under FOIA). Each of these categorical exemptions reflects a congressional judgment that as to certain information, the public interest will always tip in favor of nondisclosure. In these cases, we have absolutely no indication that Congress has ever determined that the broad range of information that will hereinafter be enshrouded in secrecy should be inherently and necessarily confidential. Nevertheless, today the Court reaches out to substitute its own policy judgments for those of Congress.

IV

To my mind, the language and legislative history of § 102(d)(3), along with the policy concerns expressed by the Agency, support only an exemption for sources who provide information based on an implicit or explicit promise of confidentiality and information leading to disclosure of such sources. That reading of the “intelligence source” exemption poses no threat that sources will “clam up” for fear of *194exposure, while at the same time it avoids an injection into the statutory scheme of the additional concerns of the Members of this Court. The Court of Appeals, however, ordered the release of even more material than I believe should be disclosed. Accordingly, I would reverse and remand this case for reconsideration in light of what I deem to be the proper definition of the term “intelligence source.”