Opinion for the court filed by Circuit Judge WALD.
WALD, Circuit Judge:Between December, 1991 and August, 1992, appellant Robert Burka filed several Freedom of Information Act (“FOIA” or “Act”) requests with the Department of Health and Human Services (“HHS” or “agency”) seeking information related to smoking behavior. The agency disclosed some of the documents Burka sought, but withheld data from a comprehensive study of how community-based interventions affect smoking habits. Burka then filed suit in federal court challenging the agency’s decision to withhold the research material. On remand from an earlier round in this court, the district court granted the agency’s motion for summary judgment, ruling that the materials sought by Burka were exempt from FOIA disclosure under Exemption 5 of the Act as “confidential research information.” Burka now seeks reversal of the district court’s decision. As explained below, Exemption 5 allows an agency to refuse disclosure of records “not available by law to a party ... in litigation with the agency.” We find that the material sought by Burka— computer tapes recording survey responses about smoking habits and attitudes, as well as paper questionnaires on the same topic— is not generally protected in civil discovery because of its nature, and so conclude it does not fall under the protection of Exemption 5. Accordingly, we reverse the grant of summary judgment to HHS and remand to the district court for further proceedings.
I. Background
Before considering whether the agency properly withheld the the information requested, we explain exactly what material Burka sought from HHS, and how the agency and district court responded.
A. The “COMMIT” Study
1. Study Goals, Design, and Implementation
The data sought by Burka was compiled during an extensive study of smoking behavior conducted from 1986 to 1993 by the National Cancer Institute (“NCI”), a division of HHS’ National Institutes of Health. NCI *511initiated this project, called the Community Intervention Trial for Smoking Cessation Study (“COMMIT” or “study”), in order to assess the effectiveness of community-based anti-smoking efforts. COMMIT represented a significant investment of $45 million by NCI, see Editor’s Foreword, 11 Int’l Q. Community Health Educ. 169 (1991), reprinted in App. at 189, which described it as “the most comprehensive planning, implementation and evaluation effort to reduce smoking through community approaches attempted to date.” Lawrence Wallack & Russell Seiandra, Media Advocacy and Public Education in the Community Intervention Trial to Reduce Heavy Smoking (COMMIT), 11 Int’l Q. Community Health Educ. 205, 206 (1991), reprinted in App. at 221, 222.
COMMIT was designed to test two related hypotheses: first, that smoking behavior is significantly shaped by social circumstances and pressures, and second, that using community sites such as schools, workplaces, and health care offices as the locus for anti-smoking efforts will induce people to alter their smoking behavior. Lichtenstein, et al., Introduction to the Community Intervention Trial for Smoking Cessation (COMMIT), 11 Int’l Q. Community Health Educ. 173, 175 (1991), reprinted in App. at 191, 193; Thompson, et al., Principles of Community Organization and Partnership for Smoking Cessation in the Community Intervention Trial for Smoking Cessation (COMMIT), 11 Int’l Q. Community Health Educ. 187, 188-89 (1991), reprinted in App. at 204, 205-06; Thompson, et al., Community mobilization for smoking cessation: lessons learned from COMMIT, 8 Health Promotion Int’l 69, 70 (1993), reprinted in App. at 363, 364. In eleven different locations across the United States and Canada, the COMMIT team paired communities with similar demographic profiles, mobility and migration patterns, urbanization rates, numbers of work sites, estimated smoking rates, and access to various channels for community-based intervention efforts. COMMIT Research Group, Community Intervention Trial for Smoking Cessation (COMMIT): Summary of Design and Intervention, 83 J. Nat’l Cancer Inst. 1620, 1621 (1991), reprinted in App. at 178, 179. One community in each pair was randomly selected to participate in a concentrated smoking cessation campaign, and the other community was used as a control for comparison purposes. In each of the eleven communities targeted for intervention, COMMIT researchers conducted 4 years of smoking cessation efforts at local schools, workplaces, and health care provider offices, as well as through the mass media. Id. at 1620-21, reprinted in App. at 178-79.
To measure the effectiveness of these cessation programs, COMMIT researchers developed survey tools to measure smoking behavior both before and after the intervention efforts. Prior to launching the cessation programs, the researchers conducted two “before” polls of smoking habits in each of the eleven pairs of communities (22 communities total) to obtain baseline data. In the first survey, conducted in 1988, COMMIT researchers asked adults about their demographic characteristics, smoking status, current and past smoking behavior, use of tobacco products other than cigarettes, attitudes and knowledge about smoking, and other health-related information. First Decl. of Dr. Edward J. Sondik ¶ 17, reprinted in App. at 120,125 [hereinafter “First Sondik Decl.”]. In the second “before” survey, two years later, the COMMIT team asked approximately 8,000 ninth-grade students to complete a lengthy written questionnaire regarding their current and past smoking habits, as well as a range of other topics related to influences which might affect their smoking behavior.1 Id. ¶¶ 19-20. The researchers then engaged in various intervention efforts at a range of different community locations, and surveyed the same populations afterwards to determine how the cessation efforts had affected *512their smoking habits.2 Memorandum & Order, No. 92-2636, at 4 (D.D.C. Feb. 10, 1995), reprinted in App. at 10, 13 [hereinafter “1995 Mem. & Order”].
Rather than coordinating the collection and analysis of this data itself, NCI contracted out this task to Information Management Systems (“IMS”). See NC1/IMS Contract, reprinted in App. at 406-13; Plaintiffs Statement of Material Facts Not in Dispute ¶4, reprinted in App. at 446, 447; Defendant’s Response to Plaintiffs Statement of Material Facts ¶ 4, reprinted in App. at 451. According to NCI and IMS, two other firms actually conducted the adult phone survey and created computer tapes of the results, and then turned over these tapes to IMS. First Sondik Deck ¶ 21; First Deck of Janis A. Beach, ¶¶ 3-4 reprinted in App. at 84, 85 [hereinafter “First Beach Deck”]. Various research institutions collaborating on the COMMIT project collected the paper questionnaires and turned them over to IMS as well. Id. IMS transferred these results to computer tapes, which it still has in its possession, id., and transmitted the paper questionnaires to NIH, where they are currently stored in a federal records center, First Sondik Deck ¶ 23; First Beach Deck ¶ 11.
2. Publications by COMMIT Researchers
One of the central goals of the COMMIT project is the publication of research results. For example, members of the COMMIT team have already published several articles based solely on the baseline data — one concerning the methodology developed by the team for measuring community-wide behavior changes, see Gail, et al., Aspects of Statistical Design for the Community Intervention Trial for Smoking Cessation (COMMIT), 13 Controlled Clin. Tr. 6 (1992), reprinted in App. at 327, and several others discussing the smoking rates in various test communities and across various demographic categories, see, e.g., Mattson, et ah, Evaluation Plan for the Community Intervention Trial for Smoking Cessation (COMMIT), 11 Int’l Q. Community Health Educ. 271 (1991), reprinted in App. at 284; Sorensen et ah, Promoting Smoking Control Through Work-sites in the Community Intervention Trial for Smoking Cessation (COMMIT), 11 Int’l Q. Community Health Educ. 239 (1991), reprinted in App. at 254. Still another article discusses the purchasing habits of teen smokers, see Cummings, et al., Where teenagers get their cigarettes: a survey of the purchasing habits of 13-16 year olds in 12 U.S. communities, 1 Tobacco Control 264 (1992), reprinted in App. at 323. The research team has also published other articles which compare the baseline and follow-up data and draw conclusions about the use of various resources for quitting smoking, as well as other factors which might affect the likelihood that a particular segment of the population will quit smoking, see, e.g., Pomrehn, et al., Enhancing Resources for Smoking Cessation Through Community Intervention: COMMIT as a Prototype, 11 Int’l Q. Community Health Educ. 259 (1991), reprinted in App. at 273; Royce, et al., Smoking Cessation Factors among African Americans and Whites, 83 Am. J. Pub. Health 220 (1993), reprinted in App. at 356. The agency has stated that the COMMIT team plans to publish a total of 11 additional reports, see Br. for Appellee at 33 n.12, although when asked at oral argument, counsel for the government could not provide a specific time line for publication. Nor has the agency provided us with information about the topics of these proposed articles, other than to assert that they will “disseminate the results of the project.” Id. at 32; see also First Sondik Deck ¶ 29 (reports will present “the results and conclusions from the project”).
B. Burka’s FOIA Requests
After filing two FOIA requests in 1991 and 1992 seeking information relating to cigarette advertising,3 on August 31, 1992, Burka filed *513his third and final FOIA request, seeking “all questionnaires and data tapes relating to the National Cancer Institute 1990 survey of ninth-grade students undertaken as part of the [COMMIT] study.” Burka FOIA Request of 8/31/92, reprinted in App. at 65-66. The agency refused to give him this material. It acknowledged that while it had the written questionnaires in its possession, it was withholding them under Exemptions 4 and 5 of FOIA; as to the tapes, the agency stated only that they were in possession of IMS. HHS Response Letter of 9/4/92, reprinted in App. at 67-68.
After the agency denied his administrative appeal, see Second Burka Deck ¶ 3, reprinted in App. at 99-100, Burka filed suit in district court. Abandoning his quest for other materials related to cigarette advertising, he challenged the agency’s decision to withhold only three specific items: the written questionnaires completed by ninth-graders in 1990; the computer data tape containing the results of these questionnaires; and the computer data tape containing the results of the 1988 adult phone survey. Compl. for Injunctive Relief ¶ 9, reprinted in App. at 31, 34. The government continued to oppose disclosure, arguing that the data tapes were not “agency records” covered by FOIA, and that even if they were, both the tapes and written questionnaires were protected under Exemptions 4, 5, and 6.
Adopting the government’s position in part, the district court first noted that an agency document is covered by Exemption 5 if not “routinely discoverable by an agency’s opponent upon a showing of relevance in a civil action.” Memorandum, No. 92-2636 at 8 (D.D.C. Dec. 13, 1993) [hereinafter “1993 Mem.”]. The court further stated that defendants had cited “ample authority” for the proposition that confidential research material may be protected in the context of civil discovery. The remaining question, the court found, was whether or not the specific information sought by Burka — the question-names and tapes — would be fairly categorized as “confidential research material” if the agency attempted to shield it in civil discovery proceedings. It concluded that the agency had identified two interests which would trigger protection of the material in discovery:!) interference with ongoing scientific research, which could occur if the data was released prior to the follow-up surveys, and individuals then modified their behavior or survey answers in response to this information; and 2) harm to “the COMMIT researchers’ interest in having their research published in prestigious, peer-reviewed journals,” which the agency claimed would result if any study data was disclosed prior to release of the remaining articles slated for publication. Id. at 8-9. Accordingly, the court granted summary judgment for the agency, ruling that although the questionnaires and data tapes were “agency records” for purposes of FOIA, they could be withheld under Exemption 5.
*514Burka appealed that decision to our court, and before oral argument, filed a motion to supplement the record with twelve additional publications allegedly containing data related to the COMMIT study. 1995 Mem. & Order at 4r-5. This court remanded the case to the district court for consideration of Burka’s supplemental request. Order, No. 94-5003, 1994 WL 315403 (D.C.Cir. June 15, 1994) (per curiam). On remand, the district court granted the motion to supplement the record and proceeded to resolve the case on the merits. Leaving undisturbed its conclusion that the requested information constituted “agency records” subject to FOIA, the court considered whether the material was exempt from disclosure. It first determined that since the COMMIT research had been concluded during the time between the district court’s initial ruling and the remand, the government’s interest in avoiding a skewing of the research results was no longer relevant. 1995 Mem. & Order at 4. The court then turned to the government’s interest in publishing the COMMIT results in prestigious journals, finding that even though some of the COMMIT data had been made public via the additional articles included in the motion to supplement, this partial disclosure did not vitiate the researchers’ interest in keeping confidential the remainder of the data until they had published the rest of the proposed articles. It concluded that “[tjhose interests [in publication] still render the data ‘confidential research information’ within the meaning of Rule 26(e)(7) and hence, justify exempting the data from disclosure under FOIA exemption 5.” Id. at 8.
Burka now seeks reversal of the district court’s grant of summary judgment for the government. He advances several arguments in support of his claim that Exemption 5 is inapplicable. While appearing to maintain, at times, that Exemption 5 incorporates only a limited number of privileges (deliberative material, attorney/client communications, work product, and confidential commercial information), Burka acknowledges elsewhere that the provision encompasses all well-recognized civil discovery protections. He claims, however, that the rules of civil discovery recognize neither a blanket protection of confidential research information, nor a more tailored protection which might be invoked to prevent premature disclosure of research data if disclosure would harm the researcher’s opportunities for publication. Alternatively, he argues that even if confidential research data may be shielded in discovery to protect a researcher’s publication opportunities, the government has not adequately explained how disclosure of the COMMIT tapes and questionnaires would harm that interest and thus has not demonstrated that this material would be protected in discovery. Not surprisingly, the government takes a contrary position on both points, asserting that confidential research data is a category of information protected in civil discovery proceedings, and that the specific records sought by Burka would be eligible for such protection and may therefore be withheld in the FOIA context, as well.
II. Analysis
Having detailed the type of information sought by appellant Burka and the tortuous legal path his FOIA request has taken, we reach the question of whether the agency appropriately withheld the COMMIT questionnaires and data tapes — a query answered in the affirmative by the district court in granting summary judgment for HHS. It is well established that summary judgment is appropriate where, viewing the facts in the light most favorable to the non-moving party, no genuine issue of material fact remains. Fed.R.Civ.P. 56(c); Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 249, 106 S.Ct. 2505, 2510-11, 91 L.Ed.2d 202 (1986). We review de novo a district court’s grant of summary judgment for an agency in a FOIA case. Nation Magazine v. U.S. Customs Service, 71 F.3d 885, 889 (D.C.Cir.1995) (citing Steinberg v. United States Dep’t of Justice, 23 F.3d 548, 551 (D.C.Cir.1994)), though “mindful that the ‘burden is on the agency’ to show that requested material falls within a FOIA exemption.” Petroleum Information Corp. v. U.S. Dep’t of Interior, 976 F.2d 1429, 1433 (D.C.Cir.1992) (quoting 5 U.S.C. § 552(a)(4)(B) and citing Anderson, 477 U.S. at 254, 106 S.Ct. at 2513).
*515A. COMMIT data tapes are “agency records” for purposes of FOIA
To qualify as an “agency record” subject to FOIA disclosure rules, “the agency must ‘either create or obtain’ the requested materials,” and “the agency must be in control of [them] at the time the FOIA request is made.” Dep’t of Justice v. Tax Analysts, 492 U.S. 136, 144, 109 S.Ct. 2841, 2848, 106 L.Ed.2d 112 (1989) (quoting Forsham v. Harris, 445 U.S. 169, 182, 100 S.Ct. 977, 985, 63 L.Ed.2d 293 (1980)). HHS suggests that the COMMIT data tapes 4 do not satisfy these requirements because they were neither created by agency employees, nor are they currently located on agency property. The district court disposed of this argument in its 1993 ruling, and we now affirm its persuasive analysis on this point. See 1993 Mem. at 3-7. While the mere fact that the firms which created the 1988 data tapes, as well as IMS, which created the 1990 tapes and now possesses both sets of tapes, received federal funds to finance the research would not be sufficient to conclude the data were “created or obtained” by the agency, see Forsham, 445 U.S. at 177, 100 S.Ct. at 982, the extensive supervision and control exercised by the agency over collection and analysis of the data indicates that these firms acted on behalf of HHS in creating the tapes. As to the second prong of the Tax Analysts test, this circuit has identified four factors relevant to a determination of whether an agency exercises sufficient control over a document to render it an “agency record”: “(1) the intent of the document’s creator to retain or relinquish control over the records; (2) the ability of the agency to use and dispose of the record as it sees fit; (3) the extent to which agency personnel have read or relied upon the document; and (4) the degree to which the document was integrated into the agency’s record system or files.” Tax Analysts v. Dep’t of Justice, 845 F.2d 1060, 1069 (D.C.Cir.1988) (citation omitted), aff'd on other grounds, 492 U.S. 136, 109 S.Ct. 2841, 106 L.Ed.2d 112 (1989). NCI ordered creation of the materials, plans to take physical possession of the tapes at the conclusion of the project, has indicated it will disclose the information after its publication schedule is completed and prohibited IMS from making any independent disclosures, and has read and relied significantly on the information in writing articles and developing agency policies. As did the district court, we find these actions sufficient to conclude that HHS had constructive control of the tapes at the time Burka filed his FOIA requests.5 Since Burka has adequately demonstrated that both prongs of the Tax Analysts test are satisfied here, the tapes must be treated as “agency records” for purposes of FOIA.
B. Scope of Exemption Five
Because FOIA establishes a strong presumption in favor of disclosure, see S. Rep. No. 813, 89th Cong., 1st Sess. 3 (1965), requested material must be disclosed unless it falls squarely within one of the nine exemptions carved out in the Act. United States Dep’t of Justice v. Julian, 486 U.S. 1, 8, 108 S.Ct. 1606, 1611, 100 L.Ed.2d 1 (1988) (citing United States v. Weber Aircraft Corp., 465 U.S. 792, 104 S.Ct. 1488, 79 L.Ed.2d 814 (1984)); see also Petroleum Info. Corp., 976 F.2d at 1433 (summary judgment in favor of FOIA plaintiff appropriate if agency seeks to protect information that falls outside proffered exemption). Thus, HHS may withhold the agency records sought by Burka only if they fall under an applicable exemption. The agency contended, and the district court agreed, that Exemption 5, which removes from FOIA’s reach “inter-agency or intraagency memorandums or letters which would not be available by law to a party ... in litigation with the agency,” 5 U.S.C. § 552(b)(5) (1994), would fill that bill. In support of this position, HHS draws our attention to Federal Rule of Civil Procedure 26(c)(7), which allows a court to issue a protective order in civil discovery mandating that “a trade secret or other confidential research, development, or commercial infor*516motion not be revealed or be revealed only in a designated way.” Fed. R. Crv. P. 26(c)(7) (emphasis added).6 The agency argues that the type of information sought by Burka would be shielded under this provision as “confidential research information,” and thus the material is not “routinely discoverable.” Accordingly, the agency contends, the COMMIT tapes and questionnaires may be withheld under Exemption 5. We now examine this claim.
As indicated by its language, the parameters of Exemption 5 are determined by reference to the protections available to litigants in civil discovery; if material is not “available” in discovery, it may be withheld from FOIA requesters. FOIA’s legislative history expressly mentions three types of material which fall within the exemption’s protective scope: predecisional deliberative memoranda, see, e.g., EPA v. Mink, 410 U.S. 73, 87-92, 93 S.Ct. 827, 836-38, 35 L.Ed.2d 119 (1973); Petroleum Info. Corp., 976 F.2d at 1434; Coastal States Gas Corp. v. Dep’t of Energy, 617 F.2d 854, 866 (D.C.Cir.1980); Mead Data Central, Inc. v. Dep’t of Air Force, 575 F.2d 932 (D.C.Cir.1978); Montrose Chem. Corp. v. Train, 491 F.2d 63 (D.C.Cir.1974); attorney-client communications, see, e.g., Schlefer v. United States, 702 F.2d 233 (D.C.Cir.1983); and attorney work product, see, e.g., FTC v. Grolier, Inc., 462 U.S. 19, 103 S.Ct. 2209, 76 L.Ed.2d 387 (1983); NLRB v. Sears, Roebuck & Co., 421 U.S. 132, 154, 95 S.Ct. 1504, 1518, 44 L.Ed.2d 29 (1975); Coastal States Gas Corp., 617 F.2d at 864-65. Yet as the government reminds us (and Burka concedes, as he must), Exemption 5 is not confined to these three privileges. Instead, it incorporates other generally recognized civil discovery protections. If a document requested through FOIA “would be ‘routinely’ or ‘normally’ disclosed [in civil discovery] upon a showing of relevance,” it must also be disclosed under FOIA7; conversely, information which is routinely protected in discovery falls within the reach of Exemption 5. Grolier, 462 U.S. at 26, 103 S.Ct. at 2213-14.
Deciding which of these two categories better describes information that is sometimes subject to a protective order, but in other circumstances disclosed, is not an easy task.8 Precedent suggests that Exemption 5 applies to any material mentioned in FOIA’s legislative history, as well as information which is not routinely disclosed in discovery because it has been accorded a “well-settled” or “well-recognized” protection. In considering whether the “Machin privilege,” which protects confidential statements made during air crash safety investigations from discovery, was incorporated in Exemption 5, the Supreme Court explained in Weber:
It is one thing to say that recognition under Exemption 5 of a novel privilege, or one that has found less than universal acceptance, might not fall within Exemption 5 if not discussed in its legislative history. It is quite another to say that the Machin privilege, which has been well-settled for some two decades, need be viewed with the same degree of skepticism.
465 U.S. at 801, 104 S.Ct. at 1493. The court concluded that “since the Machin privilege is well recognized in the case law as precluding routine disclosure of the statements, the *517statements are covered by Exemption 5.”9 Id. at 799, 104 S.Ct. at 1492 (emphasis added). Cf. Martin, 819 F.2d at 1185 (Exemption 5 incorporates “all civil discovery rules,” rather than only those related to the executive deliberative process privilege).
Both parties agree that in several contexts, courts have exercised their authority under Rule 26(c)(7) to shield from discovery confidential research information — including raw data developed by researchers. See Br. for Appellant at 31; Br. for Appellee at 18; see also infra n. 13 (listing eases). Based on these cases, the agency would have us conclude that courts have generally recognized a protection for this type of information — a protection sufficient to bring all confidential research information within Exemption 5. We do not think, however, that these discovery decisions, taken alone, would justify a decision to extend protection to this entire category of information in the FOIA context. The decision to limit or deny discovery by means of a Rule 26 protective order rests on a balancing of various factors: the requester’s need for the information from this particular source, its relevance to the litigation at hand, the burden of producing the sought-after material; and the harm which disclosure would cause to the party seeking to protect the information. Federal Open Market Comm. of the Fed. Reserve Sys. v. Merrill, 443 U.S. 340, 362-63, 99 S.Ct. 2800, 2813-14, 61 L.Ed.2d 587 (1979); see also Wright et al., Federal Practice and Procedure § 2043 (1994) (explaining balancing test for issuance of Rule 26 protective order). Yet several of these criteria are not pertinent in the FOIA context; relevance and need mean nought, since by statute a FOIA requester need not disclose either, and any burdensomeness of a request is accounted for in other provisions of the FOIA statute. See Merrill, 443 U.S. at 362, 99 S.Ct. at 2813. Moreover, under this standard it will almost always be possible to identify some combination of circumstances under which requested information may appropriately be the subject of a protective order. Even though a court might generally permit discovery of a particular document, it may deny discovery in a case where the information is not particularly relevant or necessary to the requester’s ability to present his case. Thus, if we simply look at whether the agency can identify any eases extending a protective order to a certain category of information, the answer will usually be “yes.” If Exemption 5 is triggered every time an agency can point to such cases, the exception carved out of FOIA’s overarching policy of disclosure will quickly swallow up the rule.10 We are inclined to the view that to justify nondisclosure under Exemption 5, an agency must show that the type of material it seeks to withhold is generally protected in civil discovery for reasons similar to those asserted by the agency in the FOIA context.
This approach accords with the Supreme Court’s assessment of the relationship between Rule 26(c) and Exemption 5. In Merrill, the Court faced the question of whether financial reports generated by the *518Federal Reserve Board could be temporarily withheld to avoid potentially deleterious effects on the price of government bonds. It noted that Rule 26(c)(7) permits courts to protect confidential commercial information from discovery.11 Yet it was not enough, the Court determined, for the agency to simply demonstrate that the reports it sought to withhold fell within this category. The Federal Reserve was also required to show, on remand, that the specific interest which justifies issuance of a protective order for this type of material — harm to the financial position of the entity which holds the information — was implicated by the FOIA request. See Merrill, 443 U.S. at 363, 99 S.Ct. at 2813-14 (“the sensitivity of the commercial secrets involved, and the harm that would be inflicted upon the Government by premature disclosure, should continue to serve as relevant criteria in determining the applicability of this Exemption 5 privilege”). Likewise, we find it insufficient for HHS merely to categorize the COMMIT data as “confidential research information.” To prevail, it must persuade us that the interest which it identifies as warranting protection in civil discovery — the researchers’ publication opportunities — is both a generally recognized ground for shielding the material from discovery12 and actually implicated by Burka’s request.
Our interpretation of the scope of Exemption 5 squares with the general structure of Rule 26. The rule allows litigants to obtain discovery “regarding any matter, not privileged, which is relevant to the subject matter involved in the pending action.” Fed. R.Civ.P. 26(b)(1). Privileged information, then, is presumptively not discoverable. In addition, the rule shields work product prepared in anticipation of litigation from disclosure, unless the party seeking discovery demonstrates “substantial need” for the materials, see Fed.R.Civ.P. 26(b)(3), as well as protecting from discovery expert witnesses not expected to testify at trial, except where a party can show “exceptional circumstances” which make it “impracticable” to obtain information on the subject by other means, see Fed.R.Civ.P. 26(b)(4). Material which falls outside these categories is presumptively discoverable, although a court may issue a protective order limiting or denying discovery “for good cause shown,” to protect from “annoyance, embarrassment, oppression, or under burden or expense.” Fed.R.Civ.P. 26(c). Such an order may, inter alia, deny discovery completely; limit the conditions, time, place, or topics of discovery; or limit the manner in which a trade secret or other confidential research, development, or commercial information may be revealed. Id. Translating this two-tiered arrangement into the FOIA context means that merely identifying information as falling within a privileged category — deliberative material, attorney/client communications, work product— may be sufficient to withhold the material from a FOIA requester. When dealing with information withheld on the grounds that it may be subject to a protective order, though, something more is required: just as the target of a discovery request must show “good cause” under Rule 26(c) to escape disclosure, the agency claiming exemption from FOIA must identify an interest similar to one which courts have found sufficient to justify the “good cause” standard in discovery proceedings.
*519C. Applying this Standard to HHS’ NonDisclosure of the COMMIT Data
The asserted government interest in nondisclosure of the COMMIT data is that the COMMIT researchers and NCI want to publish their results in prestigious, peer-reviewed journals. By doing so, they can present their findings in a forum which commands professional respect and enhances their credibility, and by inference strengthens the weight of any agency actions or policy recommendations based on those findings. Unfortunately, the government has identified only one case in which a federal court has granted a protective order based on that kind of interest, and our own research has unearthed no others.
Although we doubt initially that one federal case is sufficient to support the agency’s claim that the type of material sought by Burka is covered by Exemption 5, we naturally look to that decision for the persuasiveness of its reasoning. In Dow Chemical Company v. Allen, the Seventh Circuit considered whether research notes, reports, working papers, and raw data relating to ongoing animal toxicity studies were discoverable by Dow, which planned to use the information at hearings on the possible cancellation of its license for production of an herbicide. 672 F.2d 1262 (7th Cir.1982). The district court declined to enforce a subpoena issued against the researcher, on the grounds that the probative value of the information was quite limited at that stage of the case; Dow’s need for the material was not overwhelming; and the subpoena was excessively burdensome. Id. at 1270-73. In affirming this ruling, the Seventh Circuit examined each of these interests. It agreed that the research material lacked probative value because it could not be used by Dow to demonstrate the safety of the herbicide product at issue, and was not of great need to Dow because neither the researcher’s testimony nor evidence from his studies would be introduced at the cancellation hearing. Id. at 1272. On the other side of the protective order balancing test, the court identified several ways in which discovery would impose a significant burden on the researcher. Citing affidavits unchallenged by Doio, the court found that immediate public access to the requested material would deprive researchers of both the opportunity to have their results published by prestigious peer-reviewed journals, as well as the professional benefits accompanying this achievement. Id. at 1273-74. It also determined that the researchers’ interest in academic freedom would be chilled by having to disclose all of their underlying raw data and notes. Id. at 1276-77. Since the harm to these interests that might be caused by immediate disclosure of the research material outweighed Dow’s minimal relevance and need for the material, the court concluded that the district court had not erred in denying discovery.
The Seventh Circuit revisited this issue in Deitchman v. E.R. Squibb & Sons, Inc., where it vacated and remanded a district court’s order quashing a subpoena for research data on the connection between the drug diethylstilbestrol (“DES”) and cancer. 740 F.2d 556 (7th Cir.1984). Balancing the need for disclosure against any harm it might impose, the court found the burden of disclosure significant. The information sought — a registry of patients suffering from vaginal cancer — was a critical tool for epidemiological research, and since it was obtained from patients with a promise of strict confidentiality, the continued viability of the registry could be severely compromised by disclosure. Id. at 559-60. This concern, the court determined, could be sufficiently addressed by conducting discovery in a manner which preserved confidentiality. Id. at 560. The second interest asserted by the researcher in his effort to defeat the subpoena was similar to that discussed in Dow: he wished “to divulge to the public the results of his studies only in his own time and way.” Id. Without exploring this issue, the court tersely acknowledged that it had previously recognized an interest in protecting research material from premature disclosure: “We agree arguendo that the [research materials] may enjoy a qualified privilege and are not to be pried into unnecessarily.” Id. at 560-61. On the other side of the scale, the Seventh Circuit found the company’s need for the data more compelling than had the district court; without discovery, it could not review the study protocol for any flaws or biases. Id. at 563. *520Although remanding for issuance of a protective order that addressed the need for patient confidentiality and reimbursement of discovery expenses, the court indicated that discovery of material which did not reveal patient identities or the researchers’ as-yet unpublished conclusions (as compared with factual information regarding, inter alia, study protocols, questionnaires, and statistical analyses) was appropriate. Id. at 564-65.
The Second Circuit has also considered whether a researcher’s interest in her data merited a protective order, but declined to conclusively rule on this issue. In re American Tobacco Company, 880 F.2d 1520 (2d Cir.1989). The district court there had ruled that under state law, a researcher might have a cognizable interest in preventing disclosure of research data to protect her interest in publishing those results, although the existence of such a privilege was far from clear; the only state court case which could be said to have recognized the privilege focused its attention on the burden of the contested subpoena, and thus “it [was] possible that the court regarded the scholar’s interest in his research data as merely a factor to be taken into account in weighing the burdens of production.” Id. at 1528. The court of appeals eventually determined that even if such a privilege existed, there was no possibility that disclosure would preempt publication, since the data requested had been published several years prior to the discovery battle at issue. In light of the significant need demonstrated by the tobacco companies, the privilege — even if one existed — would be overridden. Id. at 1530; see also In re Grand Jury Subpoena Dated January 4, 1981, 750 F.2d 223 (2d Cir.1984) (declining to decide if qualified “scholar’s privilege” should be recognized because factual record insufficiently developed).13
In several other cases, courts have not only determined that any interest in protecting research material from disclosure is outweighed by the opposing side’s need for the information, but have gone farther, eschewing recognition of such an interest altogether. In Wilkinson v. FBI, a California district court declined to adopt either a “researcher’s privilege” or a broader “archival privilege.” 111 F.R.D. 432 (C.D.Cal.1986). While acknowledging that the Seventh Circuit had been more amenable to this theory in Dow, the court stated that
the researcher’s privilege is by no means universally accepted by other courts. The *521other eases that have considered the privilege have uniformly declined to adopt it. Some have been generally sympathetic to the privilege, yet have nevertheless held that it could not be applied on the facts of the ease. Still others appeal’ to be hostile to the proposed privilege and have declined to adopt it in the first instance. Therefore, ... the cases considering a researcher’s privilege are not only distinguishable from the case at hand, but are simply too tenuous to justify a major extension of their rationale to create a wholly novel archival privilege.
Id. at 442 (citations omitted). Similarly, another district court, while quashing a subpoena issued for production of a researcher’s study protocols and notes on the grounds that the plaintiff failed to demonstrate the relevance of the material, rejected the researcher’s claim of privilege: “experts or researchers do not have any federal statutory, ease law or common law privilege which protects against their having to involuntarily share their expertise with the parties in the litigation. Nor does [state] law provide privilege protection for academic or scientific researchers.” Anker v. G.D. Searle & Co., 126 F.R.D. 515, 519 (M.D.N.C.1989). See also In re Snyder, 115 F.R.D. 211, 213 (D.Ariz.1987) (quashing subpoena ordering researcher to give deposition testimony and disclose underlying research data because researcher had already been required to respond to four prior subpoenas in related litigation seeking similar information, but ruling that “there is no general academic privilege protecting [the researcher’s information]”); Wright v. Jeep Corp., 547 F.Supp. 871, 876 (E.D.Mich.1982) (reversing magistrate judge’s order quashing subpoena of researcher and rejecting researcher’s privilege as unsupported in ease law and contrary to general rule of disclosure).
At this juncture in the development of the law, we cannot say that there is an established or well-settled practice of protecting research data in the realm of civil discovery on the grounds that disclosure would harm a researcher’s publication prospects. This is not to say that there never will be such a well-established practice; we would not presume to say how many cases are sufficient to transform an occasional grant into a generally recognized protection. But as of now, however, the precedent does not rise to the level of a routinely accepted dispensation shielding data from discovery in order to preserve a researcher’s opportunities for publication. Because therefore the agency has not defeated Burka’s claim that the data he seeks is “routinely available” in civil litigation, it may not withhold the information pursuant to Exemption 5.14
* * 5¡! * * *
In order to withhold the tapes and questionnaires sought by Burka, the agency bears the responsibility of demonstrating that this type of information is shielded from discovery for the same reason that it asks us to protect the information from FOIA disclosure requirements — because disclosure would diminish the researchers’ ability to publish their results in prestigious journals. It has not done so. Since the agency has not satisfied its burden of showing that the questionnaires and data tapes sought by Burka fall within the scope of Exemption 5, we reverse the district court’s grant of summary judgment for HHS and remand for further proceedings consistent with this opinion.
So ordered.
. The written questionnaire asked students about their current and past smoking habits, where they purchased cigarettes, whether price increases would affect their purchase of cigarettes, their use of other tobacco products, the smoking habits of their friends, their exposure to tobacco advertising, the presence of anti-smoking programs and policies in their schools, their knowledge and attitudes about smoking, and the attitudes towards smoking of their parents, friends, and teachers. See First Sondik Decl. ¶ 20; COMMIT Youth Survey, reprinted in App. at 147-67.
. Although Burka’s initial FOIA requests might be interpreted as including these follow-up responses, in his final request to the agency and in his briefs to the district court and this court, he clarified that he now seeks only the baseline data collected prior to the COMMIT intervention efforts. Thus, the follow-up responses are not at issue in this appeal.
. In December of 1991, Burka submitted a FOIA request for "[a]Il documents constituting, referring or relating to cigarette/tobacco advertising *513targeted at specific age groups of actual or potential consumers of tobacco products and the effects of such advertising on brand recognition, loyalty and market share among different age groups.” Burka FOIA Request of 12/19/91, reprinted in App. at 42-44. The agency released a number of documents and notified Burka that it was withholding other responsive materials under FOIA Exemptions 5 and 6, but did not specifically mention the existence of the COMMIT data tapes or questionnaires. HHS Response Letters of 1/7/92, 3/2/92, and 3/13/92, reprinted in App. at 45-52. In the meantime, an HHS publication printed an article summarizing some of the baseline COMMIT data on brand selection among adults and teenagers, several other articles regarding the effect of cigarette advertising on children appeared in other publications, and the Surgeon General made public remarks on the same topic. See, e.g., CDC Morbidity and Mortality Weekly Report, reprinted in App. 348-55; Glenn Ruffenach, Study Says Teen-Agers’ Smoking Habits Seem to Be Linked to Heavy Advertising, Wall St. J., Mar. 13, 1992, at B8, reprinted in App. at 347; Suggested Remarks for Antonia Novello, M.D., Surgeon General of the Public Health Service, at the AMA Press Conference on "Old Joe" Advertising, Mar. 9, 1992, reprinted in App. at 343-46. Burka then filed a second, more specific FOIA request on March 25, 1992 referencing these articles and the COMMIT study, and on April 10, 1992, appealed the denial of his first request. Burka FOIA Request of 3/25/92, reprinted in App. at 53-57; Burka Appeal of Denial of FOIA Request, reprinted in App. at 58-62. The agency, in turn, provided several additional documents related to "Old Joe Camel” advertising, upheld denial of a relevant memorandum pursuant to Exemption 5, and stated that there were no further responsive records. HHS Response Letter of July 8, 1992, reprinted in App. at 63-64. Several months later, it also released copies of the COMMIT contracts and protocol. First Burka Deck ¶ 13, reprinted in App. at 91, 95.
. HHS does not contest Burka's position that the COMMIT questionnaires are agency records for purposes of FOIA.
. In fact, it is not clear from the government's brief whether it contests on appeal Burka's claim that the agency exercised constructive control over the records at the time appellant's FOIA request was filed.
.In relevant part, Rule 26(c) provides:
Upon motion by a party or by the person from whom discovery is sought, ... for good cause shown, the court in which the action is pending ... may make any order which justice requires to protect a party or person from annoyance, embarrassment, oppression, or undue burden or expense, including one or more of the following:
... (7) that a trade secret or other confidential research, development, or commercial information not be revealed or be revealed only in a designated way.
Fed. R. Civ. P. 26(c).
. For example, presentence investigation reports requested by the prisoner himself are routinely available in discovery to the subject of the report and thus not within Exemption 5’s purview. Julian, 486 U.S. at 12-13, 108 S.Ct. at 1613-14.
. Cf. Julian, 486 U.S. at 11, 108 S.Ct. at 1612-13 (Exemption 5 is a "somewhat Delphic provision”); Martin v. Office of Special Counsel, 819 F.2d 1181, 1184 (D.C.Cir.1987) ("the exact relationship between ordinary civil discovery and Exemption (b)(5), particularly the application of discovery privileges under the exemption, has bedeviled the courts since the Act’s inception”).
. Grolier, decided the year before Weber, offers some additional insight on when a civil discovery protection is "well-settled.” 462 U.S. at 26, 103 S.Ct. at 2213-14. There, the Supreme Court ruled that Exemption 5 protected work-product materials even after the termination of litigation, concluding that material is not "routinely available” if a majority of the circuits have categorized it as privileged in the context of discovery: "all of the courts of appeals that had decided the issue ... [and] an overwhelming majority of the federal district courts reporting decisions ... were in accord with that view.” Id. While not tabulating the “yeas” and "nays,” the court cited cases from five circuits, six district courts, and several state courts in accord with its holding, and no cases to the contrary.
. For a similar point, see Anderson v. Dep't of Health & Human Services, 907 F.2d 936, 945-46 (10th Cir.1990). There, the court rejected a claim that Exemption 4 of FOIA, which protects "trade secrets and commercial or confidential information” provided to an agency by a private party, extends to all materials subject to protective order under Rule 26(c)(7). Recognizing that courts may issue protective orders imposing varied restrictions, for a wide range of reasons, the Tenth Circuit concluding that shielding all material covered by some type of protective order from FOIA disclosure would be an overly broad reading of the exemption: "It cannot be argued seriously that a Rule 26(c)(7) order merely limiting production of certain documents to a specified place renders those materials privileged for FOIA purposes.” Id. at 946.
. In deciding that Exemption 5 included confidential commercial information which would be protected in discovery, the Meirill Court also relied on the oblique reference in a House Report to the need to withhold information generated in the process of awarding a contract. See Merrill, 443 U.S. at 357-59, 99 S.Ct. at 2810-12. The Court subsequently made clear, however, that if a civil discovery protection is not a novel one (i.e., if it is "well-recognized”), it may be incorporated in Exemption 5 even if not mentioned in FOIA’s legislative history. See Weber, 465 U.S. at 800-01, 104 S.Ct. at 1493-94.
. In Merrill, only one reason (harm to an entity’s competitive financial position) was advanced for protecting confidential commercial information from discovery. By contrast, other case law regarding confidential research information suggests several reasons to shield this material from disclosure, e.g., patient confidentiality, continued viability of a research project, and researchers' publication opportunities. To give force to the Court’s insistence in Weber and Grolier that a protection be generally accepted in civil discovery before it is incorporated into Exemption 5, we think that an agency must show not only that the interest asserted in support of non-disclosure has been relied upon to justify issuance of a protective order, but that protection based on that interest is more than a "novel” occurrence.
. The government cites other cases which grant or affirm issuance of a protective order for confidential research information, but do so on grounds other than the nature of the information. See Deford v. Schmid Products Co., 120 F.R.D. 648, 654 (D.Md.1987) (permitting discovery of research data because researcher failed to demonstrate that disclosure of information would cause harm); Harris v. Upjohn, 115 F.R.D. 191, 192 (S.D.Ill.1987) (ordering production of reports regarding adverse drug reactions but allowing redaction of the names of patients and physicians from these reports, on grounds that patients will participate in research only if their identities remain confidential); Farnsworth v. Procter & Gamble Co., 101 F.R.D. 355, 356-58 (N.D.Ga.1984) (granting protective order to keep confidential the names of patients participating in study), aff'd, 758 F.2d 1545 (11th Cir.1985); Rogers v. Proctor & Gamble Co., No. 83-0487, 1984 WL 3658 (E.D. Mo. May 4, 1984) (research data has little probative value; premature disclosure would make completion of study difficult; and data is so "inconclusive, unconfirmed, and unreliable” that "[t]o allow the jury to hear [the] unfinished work would run the risk of misleading the jury”). One of the cited cases, Plough, Inc. v. National Academy of Sciences, does acknowledge a researcher's interest in protecting from discoveiy "the analyses, opinions and conclusions drawn ... from [her] data." 530 A.2d 1152, 1156 (D.C.1987). This case, however, involved internal deliberations of a scientific research committee conducted under a promise of confidentiality, not underlying research data, and the court's decision rested on the limited relevance and necessity of the documents, coupled with the committee's interest in protecting the confidentiality of these evaluations, not on the researcher's interest in publishing her results. Another state court case cited by the agency, In re RJ. Reynolds Tobacco Co., mentions a researcher's publication interest, as well as an interest in academic freedom, but the court bases its decision to deny discovery primarily on the excessive burden which discovery would impose on the researchers. 136 Misc.2d 282, 518 N.Y.S.2d 729, 732-34 (Sup.Ct.1987). Moreover, the Second Circuit, in a ruling ordering release of the same information sought by the plaintiffs in Reynolds, questioned whether the court in Reynolds did intend to recognize a state researcher's privilege and declined to establish one itself. See In re American Tobacco Co., 880 F.2d at 1526-29. Because these decisions turn on interests other than the one asserted by HHS here, we do not find them directly on point.
. Additionally, it is not clear to us that the agency has successfully demonstrated that disclosure of the results would hamper its researchers' interests. The information sought by Burka relates only to the baseline data, not to the effectiveness of the smoking cessation intervention efforts. The agency asserts that it plans to publish additional articles, but has not indicated how disclosure of the baseline information alone — ■ data already the subject of several articles published before conclusion of the study — will affect its ability to publish any remaining pieces.