OPINION OF THE COURT
SEITZ, Chief Judge.
The United States appeals from an order of the district court quashing a grand jury subpoena issued to Sun Company, Inc. The subpoena sought various questionnaires and memoranda compiled by Sun Company and the law firm of Pepper, Hamilton & Scheetz during an internal corporate investigation of various foreign transactions. Althbugh Sun has contested our jurisdiction to consider this appeal, we recently settled this question in In the Matter of Grand Jury Empanelled February 14, 1978 (Colucci), 597 F.2d 851 (3d Cir. 1979). Pursuant to 18 U.S.C. § 3731,1 the United States Attorney has certified that this appeal “is not taken for the purpose of delay and that the evidence is a substantial proof of a fact material in the proceeding.” The district court having received this certification, we are not required by section 3731 to evaluate independently the substantiality or the materiality of the contested material. Compare 18 U.S.C. § 3731 with 28 U.S.C. § 1292(b). Cf. United States v. Comiskey, 460 F.2d 1293, 1297-98 (7th Cir. 1972) (United States Attorney need not allege any specific facts in support of his affidavit). We accordingly have jurisdiction to consider this appeal under 18 U.S.C. § 3731 or, alternatively, under 28 U.S.C. § 1291. See Colucci, supra, at 854-858.
I
Although the facts before us are largely uncontested, the record has been impounded to preserve the confidentiality of the grand jury’s inquiry. We therefore will refrain from identifying some individuals by name.
[1227]*1227In January 1976 Sun Company, Inc., (Sun) began an investigation into possible illegal payments made by some of its employees in connection with Sun’s foreign operations. The corporation’s Audit Committee, a standing committee of the Board of Directors, supervised the investigation and, on July 20, 1976, reported that no significant violations had occurred.
Late in 1976, however, information submitted by a Sun employee led to a reopening of the investigation. On January 25, 1977, Samuel K. White, Sun’s vice-president and general counsel, retained Pepper, Hamilton & Scheetz (Pepper) to advise Sun of its legal obligations in regard to certain payments uncovered during the internal audit. Two days later, the Audit Committee asked Pepper to assist and advise it in the conduct of the investigation itself. The full Board of Directors later ratified Pepper’s retention by the Audit Committee and authorized funds for the investigation.
On February 11, 1977, H. Robert Shar-baugh, the chairman of Sun’s Board of Directors, sent a covering letter, a questionnaire, and a return envelope to each of 1,877 managerial employees of Sun and its majority-owned subsidiaries. The letter explained the purpose of the investigation and asked the employees to complete the questionnaires and to mail them directly to Pepper. The questionnaire itself contained ten questions probing the employee’s knowledge of any suspicious transactions. The instructions accompanying the questionnaire explained that an employee should answer “yes” if he knew of payments like those described in the question, “no” if he didn’t, and “conference” if he was uncertain. Neither the questionnaire nor the instructions requested any further elaboration on the answers.
On February 16, 1977, Pepper began to follow up questionnaires containing responses of “yes” or “conference.” By September, Pepper had conducted 265 telephone interviews and 90 personal interviews. No interviews were conducted in the presence of anyone except representatives of Pepper and Sun, and, when necessary, an interpreter. The interviews were not transcribed. Instead, the Pepper attorneys reduced their notes and recollections concerning the interviews to memoranda, always within ten days of the actual interview. These memoranda have remained in Pepper’s files, and have been released only to members of Sun’s Board of Directors.
On September 21, 1977, the Audit Committee filed its report, which discussed a number of transactions deemed “questionable.” One of the selected transactions involved the renegotiation of a contract with an entity of a foreign government. Sun had paid a citizen of that country a total of $235,000 for services rendered during the renegotiation of the contract. The Audit Committee found reason to believe that some of the money paid to the foreign representative may have been passed on to governmental officials as an inducement to renegotiate the contract.
The Audit Committee made a number of recommendations in its report, including the amendment of corporate tax returns and the filing of a form 8-K with the Securities and Exchange Commission. On October 27, 1977, Sun did, in fact, file an 8-K with the SEC, disclosing all the questionable payments noted in the report, including the payment of the $235,000 to the foreign representative. This filing prompted an article in the Philadelphia Inquirer and, in turn, an investigation by the United States Attorney for the Eastern District of Pennsylvania.
In a letter dated December 20, 1977, the United States Attorney asked Sun to turn over, inter alia, “all documents referring, in any way to questionable foreign payments made by Sun, its officers or employees.” Sun responded by releasing a number of documents including the Audit Committee report itself. After examining the report, the United States Attorney narrowed the inquiry, requesting all documentation of the renegotiation of the foreign contract and the payment of the $235,000. This request specifically included “all interview memo-randa, questionnaires, statements, or other recorded recollections of these events [1228]*1228. Although Sun released a number of documents pertaining to the affair, it refused to release the interview memoranda or the questionnaires, claiming protection under the attorney-client privilege and the work-product doctrine. On March 21, 1978, Sun received a grand jury subpoena requesting these documents. That request later was amended to allow redaction of those portions of the documents that “would disclose the mental impressions, conclusions, opinions, or legal theories of an attorney.” Sun moved to quash the subpoena and the district court granted that motion on June 1, 1978. Although the district judge did not file a written opinion, he indicated at oral argument some of the bases for his conclusions.
Sun has asserted, and the government has not contested, that only thirteen persons provided the Audit Committee with any information regarding the targeted transaction. Eleven of the thirteen were employees of Sun at the time they were interviewed. One of the employees interviewed is now deceased. Except for the deceased employee and the foreign representative himself, who is neither a Sun employee nor a resident of this country, all of the interviewees could be reached by grand jury subpoenas. Although Sun has offered to ensure the appearance of its employees before the grand jury, the government has made no attempt to summon any of the interviewees.
II
Sun claims that the documents sought in the subpoena are protected by the attorney-client privilege or the work-product doctrine or both. As Sun concedes, the attorney-client privilege cannot protect any communications made by either of the two interviewees who were not employed by Sun. Sun asserts, however, that the work-product doctrine protects all the summoned documents and that this protection is absolute. We therefore consider first the more inclusive of the two alleged shields: the work-product doctrine.
A
The work-product doctrine, recognized initially in Hickman v. Taylor, 329 U.S. 495, 67 S.Ct. 385, 91 L.Ed. 451 (1947), protects from discovery materials prepared or collected by an attorney “in the course of preparation for possible litigation.” Id. at 505, 67 S.Ct. at 391. See also Fed.R.Civ.P. 26(b)(3). This doctrine applies in criminal, as well as in civil, litigation. See United States v. Nobies, 422 U.S. 225, 236, 95 S.Ct. 2160, 45 L.Ed.2d 141 (1975). Moreover, the government has conceded in this appeal that the doctrine applies to documents sought by a grand jury. See In re Grand Jury Proceedings (Duffy), 473 F.2d 840 (8th Cir. 1973); In re Grand Jury Investigation (Sturgis), 412 F.Supp. 943 (E.D.Pa.1976).
We must decide three questions in regard to the work-product doctrine. First, were these materials collected or prepared in preparation for possible litigation so as to qualify as “work product”? Second, if they are entitled to protection as work product, is the protection afforded them absolute or qualified? Third, if the documents are entitled to only qualified protection, has the government made an adequate showing to overcome that protection?
The government asserts that the subpoenaed materials are not entitled to protection as work product because they were not collected or prepared “in the course of preparation for possible litigation.” Hickman v. Taylor, supra, 329 U.S. at 505, 67 S.Ct. at 391. At the close of oral argument, the district court stated that it would be “difficult” to hold that the investigation was not conducted “in contemplation of litigation.” Although this statement, by itself j may be ambiguous, the district court subsequently quashed the subpoena as to all the requested documents. Because Sun based its motion on the attorney-client privilege and the work-product doctrine, and because Sun conceded that the attorney-client privilege did not apply to some of the documents, the conclusion is inescapable that the district court made a factual finding that the work-[1229]*1229product doctrine applied. Cf. Milliner v. Government of the Virgin Islands, 593 F.2d 532 (3d Cir. 1979) (oral findings and conclusions held sufficient to “indicate the basis of the trial judge’s decision and provide an adequate basis for appellate review”), at 534.
Indisputably, the work-product doctrine extends to material prepared or collected before litigation actually commences. On the other hand, some possibility of litigation must exist. Courts and commentators have offered a variety of formulas for the necessary nexus between the creation of the material and the prospect of litigation. See, e.g., Home Insurance Co. v. Ballenger Corp., 74 F.R.D. 93, 101 (N.D.Ga.1977) (must be a “substantial probability that litigation will occur and that commencement of such litigation is imminent”); In re Grand Jury Investigation (Sturgis), supra, at 948 (threat of litigation must be “real and imminent”); Stix Products, Inc. v. United Merchants & Manufacturers, Inc., 47 F.R.D. 334, 337 (S.D.N.Y.1969) (prospect of litigation must be “identifiable”); 4 Moore’s Federal Practice ¶ 26.63[2.-1], at 26-349 (1970) (litigation must “reasonably have been anticipated or apprehended”). Professors Miller and Wright offer an attractive formulation based on the purpose of the work-product doctrine itself:
Prudent parties anticipate litigation, and begin preparation prior to the time suit is formally commenced. Thus the test should be whether, in light of the nature of the document and the factual situation in the particular case, the document can fairly be said to have been prepared or obtained because of the prospect of litigation.
8 Wright & Miller, Federal Practice and Procedure: Civil § 2024, at 198 (1970) (emphasis added; footnote omitted).
Several circumstances in this case convince us that, under any of these tests, the district court reasonably could have found that the prospect of litigation was real enough to mandate work-product protection for the questionnaires and the interview memoranda. First, the investigation concerned suspected criminal violations. When Pepper entered the case, the Audit Committee already had uncovered enough evidence to support “a suspicion of illicit payments having been made.” See Excerpts of Minutes of Audit Committee, January 27, 1977. If further investigation confirmed that suspicion, litigation of some sort was almost inevitable. The most obvious possibilities included criminal prosecutions, derivative suits, securities litigation, or even litigation by Sun to recover the illegal payments. Moreover, the potential for litigation was immeasurably intensified by Sun’s legal obligations to report any wrongdoing to its stockholders and to various governmental agencies. Between 1973 and 1977, the SEC alone commenced 31 actions for injunctions against companies that had allegedly engaged in transactions similar to those suspected by the Audit Committee. See Addendum to Brief of Sun Company, Inc.
The prospect of litigation in this case was sufficiently strong to distinguish it from Abel Investment Co. v. United States, 53 F.R.D. 485 (D.Neb.1971), and Peterson v. United States, 52 F.R.D. 317 (S.D.Ill.1971). Those cases dealt with the discoverability of internal IRS memoranda prepared during the investigatory and settlement phases of a tax audit. In Peterson the district court rejected the government’s naked assertion that the memoranda were trial preparation materials: “[ljitigation cannot be anticipated in every such ease when relatively few result in litigation.” 52 F.R.D. at 321. See also Abel Investment Co. v. United States, supra, at 489-90 (quoting Peterson).
The government’s other arguments against work-product protection are without merit. The questionnaires and memo-randa at issue in this case clearly differ from the statutorily required accident reports held discoverable in Goosman v. A. Duie Pyle, Inc., 320 F.2d 45 (4th Cir. 1963). Furthermore, we perceive no reason to distinguish between Pepper’s role as a legal advisor and its role as an investigator. The attorney in Hickman acted in a similar dual capacity when he interviewed witnesses. [1230]*1230Under these circumstances we conclude that the district court did not err in holding that Pepper was acting in contemplation of litigation and that the work-product doctrine applies to the questionnaires and the interview memoranda at issue. But see Diversified Industries, Inc. v. Meredith, 572 F.2d 596, 611 n.4 (8th Cir. 1977) (affirming district court’s refusal to grant work-product protection on similar facts).
Sun concedes that the work-product protection afforded the questionnaires is qualified and can be overcome by a showing of good cause by the government. As to the interview memoranda, however, Sun asserts that the work-product protection is absolute and that no showing of need by the government can justify enforcement of the subpoena. Sun attempts to draw this rule directly from Hickman v. Taylor.
In Hickman, the Supreme Court dealt with two types of work product. The defendant’s attorney had taken written statements from a number of witnesses at the scene of a tug-boat accident. The Court held that plaintiff’s “naked, general demand” for these written statements was insufficient to overcome work-product protection. The burden was on plaintiff “to establish adequate reasons to justify production” of those statements. 329 U.S. at 512, 67 S.Ct. at 394.
Plaintiff also had sought to discover the content of oral interviews with witnesses, some of which interviews had been summarized in memoranda prepared by the attorney. The Court called for greater protection of this information than it had afforded the written statements:
. [A]s to oral statements made by witnesses to [defendant’s attorney], whether presently in the form of his mental impressions or memoranda, we do not believe that any showing of necessity can be made under the circumstances of this case so as to justify production. Under ordinary conditions, forcing an attorney to repeat or write out all that witnesses have told him and to deliver the account to his adversary gives rise to grave dangers of inaccuracy and untrustworthiness. No legitimate purpose is served by such production. The practice forces the attorney to testify as to what he remembers or what he saw fit to write down regarding witnesses’ remarks. Such testimony could not qualify as evidence; and to use it for impeachment or corroborative purposes would make the attorney much less an officer of the court and much more an ordinary witness. The standards of the profession would thereby suffer.
329 U.S. at 512-13, 67 S.Ct. at 394 (emphasis added). Citing this excerpt, Sun asserts that the Supreme Court intended to afford the attorney’s interview memoranda absolute protection from discovery.
Initially, we note that a few sentences after the quoted passage the Hickman Court implied that the protection was less than absolute: “[i]f there should be a rare situation justifying production of these matters, petitioner’s case is not of that type.” Id. Nevertheless, at least one court has read Hickman as calling for absolute protection of such interview memoranda. In In re Grand Jury Investigation (Sturgis), 412 F.Supp. 943, 949 (E.D.Pa.1976), the court stated that such memoranda “are so much a product of the lawyer’s thinking and so little probative of the witness’s actual words that they are absolutely protected from disclosure.” The Court of Appeals for the Eighth Circuit also has indicated that such memoranda are “absolutely, rather than conditionally, protected.” In re Grand Jury Proceedings (Duffy), supra, at 848. In the same opinion, however, the Court of Appeals implied that the protection was less than absolute: “[w]e do not believe that the attorney of a prospective criminal defendant, absent unusual circumstances, should be required to produce a summary of a witness’s statement . . .” Id. at 848-49.
Apparently, our court already has decided this issue adversely to Sun’s position. In In re Natta, 410 F.2d 187 (3d Cir.), cert. denied, 369 U.S. 836, 90 S.Ct. 95, 24 [1231]*1231L.Ed.2d 87 (1969), this court considered the level of protection to be afforded an attorney’s memoranda containing “analyses or assessments of [the client’s] position with respect to the various parties” in the litigation. The documents contained pure legal opinion. See id. at 193. Such material, often called “opinion work product,” is the most sacrosanct of all forms of work product. See, e.g., Fed.R.Civ.P. 26(b)(3) (a court must “protect against disclosure of the mental impressions, conclusions, opinions, or legal theories of an attorney”); Note, Protection of Opinion Work Product Under the Federal Rules of Civil Procedure, 64 Va.L.Rev. 333, 334 (1978). Despite the sensitive nature of the materials sought in Natta, we did not believe that they were absolutely protected:
Hickman grants to attorney’s work product a qualified immunity from discovery . . The work product immunity in Hickman is definitely limited since a showing of good cause may justify production of documents which otherwise might be protected as work product.
410 F.2d at 192-93. If pure opinion work product is only entitled to qualified protection, we believe that the memoranda at issue here are perforce entitled to no greater protection.
Other courts and commentators also have declined to interpret Hickman as clothing interview memoranda with absolute immunity from discovery. See, e.g., Harper & Row Publishers, Inc. v. Decker, 423 F.2d 487 (7th Cir. 1970), aff’d by an equally divided court, 400 U.S. 348, 91 S.Ct. 479, 27 L.Ed.2d 433 (1971); Xerox Corp. v. International Business Machines Corp., 64 F.R.D. 367, 377-81 (S.D.N.Y.1974); 4 Moore’s Federal Practice ¶ 26.63[8], at 26-394, and ¶ 26.64[4], at 26-442 (1970); Note, 64 Va.L.Rev., supra, at 339.
In declining to afford the interview memoranda absolute protection, we do not hold that they are to be treated identically to the questionnaires. As we noted earlier, quoting Professors Wright and Miller, a demand for discovery of work product must be considered “in light of the nature of the document.” 8 Wright & Miller, supra, at 198. Memoranda summarizing oral interviews present several unique and well-documented problems to the court which considers their discoverability. First, they may indirectly reveal the attorney’s mental processes, his opinion work product. See, e. g., Hickman v. Taylor, supra, 329 U.S. at 513, 67 S.Ct. 385; In re Grand Jury Investigation (Sturgis), supra, at 949. Second, their reliability as accurate reflections of the witness’s statements is a function of many factors, including the conditions of the interview, the contemporaneousness of the writing, and the editorial discretion of the attorney. See, e.g., Hickman v. Taylor, supra, 329 U.S. at 512-13, 67 S.Ct. 385. Third, discovery and use of such material creates a danger of converting the attorney from advocate to witness. See, e.g., United States v. Nobles, supra, 422 U.S. at 252-53, 95 S.Ct. 2160 (White, J., concurring); Hickman v. Taylor, supra, 329 U.S. at 513, 67 S.Ct. 385. Finally, the information contained in such memoranda generally is of limited utility, especially where the witness himself is readily available to the opposing party. See, e. g., Hickman v. Taylor, supra, at 513, 67 S.Ct. 385. Although this list is not exhaustive, it does reflect many of the special considerations that must shape any ruling on the discoverability of interview memoranda like those at issue in this case. The result, we believe, is exactly that contemplated in Hickman; such documents will be discoverable only in a “rare situation.” Id.
We now must consider whether the government has demonstrated sufficient necessity or “good cause” to overcome the work-product protection of any of the materials sought in the subpoena.
As Sun virtually concedes, the government has shown good cause to discover the questionnaire and interview memoranda of Sun’s deceased employee. The decedent’s earlier recollection of the transaction at issue undoubtedly is relevant to the grand jury’s investigation. Moreover, [1232]*1232discovery of this material evokes few, if any, of the concerns enumerated in the prior section. Although the interview memoranda conceivably could contain opinion work product, the government has offered to minimize the intrusion by allowing Pepper and Sun to delete such material from the factual recitation. Because the decedent will not testify, the possibility that the memoranda will be admitted into evidence, and the concurrent possibility that an attorney will become a witness, are remote. Finally, although the memoranda might contain inaccuracies, that possibility must be weighed against the stark inability of the government to secure the information from any more reliable source. Under these circumstances we believe that the goi ernment has demonstrated sufficient necessity to overcome the protection afforded the deceased interviewee’s questionnaire and interview memoranda.
We are not persuaded, however, that the government has demonstrated a similar quantum of necessity in regard to the remaining materials. The government advances three possible bases for a finding of good cause. First, it suggests various reasons why the remaining interviewees might be unavailable or unwilling to testify before the grand jury. But because the government has made absolutely no effort to secure their testimony, the interviewees’ unavailability is purely conjectural. All the living interviewees, with the exception of the non-resident alien, can be reached by grand jury subpoenas. Even as to the nonresident alien, we believe that the government should make a reasonable effort to secure his testimony before attempting to invade an attorney’s files.
Second, the government argues that the grand jury is investigating not only questionable transactions, but also the possibility of a corporate cover-up of those transactions. This necessitates, according to the government, a determination of “what was stated when to whom.” Conceivably, the fact that a witness made a particular statement to an investigating attorney at a particular time might indicate a cover-up and might have relevance apart from the bare facts related in the statement. The government conceded at oral argument, however, that it never filed an affidavit with the district court or otherwise presented a substantiated claim that the grand jury suspected an illegal cover-up. We have only the government’s naked assertion that it might discover such a cover-up if granted access to these materials. Given the sensitive nature of the documents subpoenaed, we do not believe that this general, unsubstantiated allegation is sufficient to overcome the protection afforded by the work-product doctrine.
Finally, the government asserts that the grand jury needs the subpoenaed materials to assess the credibility of the various interviewees. This alleged basis for disclosure requires us once again to distinguish between the questionnaires and the interview memoranda. The questionnaires, having been filled out and signed by the employees themselves, are closely analogous to the written statements considered in Hickman v. Taylor:
Such written statements and documents might, under certain circumstances, be admissible in evidence or give clues as to the existence or location of relevant facts. Or they might be useful for purposes of impeachment or corroboration. And production might be justified where the witnesses are no longer available or can be reached only with difficulty. Were production of written statements and documents to be precluded under such circumstances, the liberal ideals of the deposition-discovery portions of the Federal Rules of Civil Procedure would be stripped of much of their meaning.
329 U.S. at 511-12, 67 S.Ct. at 394 (emphasis added). Similarly, were we to deny the grand jury access to written statements under such circumstances, we might impinge on the grand jury’s “ ‘right to every man’s evidence,’ ” see Branzburg v. Hayes, 408 U.S. 665, 688, 92 S.Ct. 2646, 33 L.Ed.2d 626 (1972) (quoting J. Wigmore, Evidence § 2192 (3d ed. 1940)), a right valued at least as highly as the “liberal ideals” underlying [1233]*1233civil discovery. We conclude that the government may be entitled to discover an employee’s questionnaire to impeach or corroborate the testimony of that employee before the grand jury. To date, however, the grand jury has summoned no witnesses at all. We would act both prematurely and overbroadly were we to order production of all the questionnaires at this stage. We believe it advisable to allow the district court, after the issuance of new subpoenas, to consider the scope and timing of any such disclosure.
We do not believe, however, that the desire to impeach or corroborate a witness’s testimony, by itself, would ever overcome the protection afforded the interview mem-oranda. This rule is implicit in Hickman’s heightened protection of such material. See 329 U.S. at 512-13, 67 S.Ct. 385. As noted earlier, the testimonial quality of an attorney’s memorandum raises grave concerns not raised by a witness’s written statement.
In sum, we hold that all the questionnaires and interview memoranda are protected by the work-product doctrine. We believe that the government has demonstrated good cause to overcome that protection as to the questionnaire and interview memoranda generated by Sun’s deceased employee. We do not believe, however, that the government has yet demonstrated good cause to overcome the protection as to any of the other questionnaires or interview memoranda.
B
Because we cannot affirm the district court’s entire order solely on the basis of the work-product doctrine, we are forced to confront Sun’s alternative ground for quashing the subpoena. Sun asserts that the attorney-client privilege protects any questionnaire or interview memorandum derived from one of its employees. If Sun is correct, this privilege would prevent the government from obtaining the deceased employee’s materials, which we have held to be outside work-product protection. Additionally, the privilege would bar discovery of any other employee-generated materials which the grand jury might need in the future.
Perhaps the most commonly cited formulation of the attorney-client privilege is that offered by Judge Wyzanski in United States v. United Shoe Machinery Corp., 89 F.Supp. 357, 358-59 (D.Mass.1950):
The privilege applies only if (1) the asserted holder of the privilege is or sought to become a client; (2) the person to whom the communication was made (a) is a member of the bar of a court, or his subordinate and (b) in connection with this communication is acting as a lawyer; (3) the communication relates to a fact of which the attorney was informed (a) by his client (b) without the presence of strangers (c) for the purpose of securing primarily either (i) an opinion on law or (ii) legal services or (iii) assistance in some legal proceeding, and not (d) for the purpose of committing a crime or tort; and (4) the privilege has been (a) claimed and (b) not waived by the client.
The government advances several reasons why the attorney-client privilege should’not apply to the questionnaires and interview memoranda. Most significantly, it argues that the individual employees who communicated with Pepper did not hold positions in the corporation that allowed them to speak for the “client.” This argument forces us to confront an issue that has troubled courts for a number of years: the extent to which the attorney-client privilege protects communications to outside corporate counsel2 by employees at various levels of the corporate hierarchy.
Although this issue began to receive attention more than 20 years ago, see Simon, The Attorney-Client Privilege as Applied to [1234]*1234Corporations, 65 Yale L.J. 953 (1956), the seminal federal case was not decided until 1962. In City of Philadelphia v. Westinghouse Electric Corp., 210 F.Supp. 483 (E.D.Pa.), mandamus and prohibition denied sub nom. General Electric Co. v. Kirkpatrick, 312 F.2d 742 (3d Cir. 1962), cert. denied, 372 U.S. 943, 83 S.Ct. 937, 9 L.Ed.2d 969 (1963), Judge Kirkpatrick recognized that, although corporations were entitled to invoke the attorney-client privilege, not every communication between corporate counsel and an employee merits the protection of the privilege. In striking the balance between confidentiality and disclosure, he held that
if the employee making the communication, of whatever rank he may be, is in a position to control or even to take a substantial part in a decision about any action which the corporation may take upon the advice of the attorney, or if he is an authorized member of a body or group which has that authority, then, in effect, he is (or personifies) the corporation when he makes his disclosure to the lawyer and the privilege would apply.
Id. at 485.
This test, which focuses on the ability of the communicating employee to take discretionary action upon the attorney’s advice, has been called the “control-group test.” Since 1962, the issue confronted by Judge Kirkpatrick has received considerable attention in judicial opinions and legal commentaries. See notes 3-6, infra, and accompanying text. A clear majority of the federal courts adhere to the control-group test.3
In 1970, the Court of Appeals for the Seventh Circuit concluded that the control-group test was too restrictive of the privilege and chose instead to focus on the subject matter of the communication between the attorney and the employee. In Harper & Row Publishers, Inc. v. Decker, 423 F.2d 487, 491-92 (7th Cir. 1970), affd by an equally divided court, 400 U.S. 348, 91 S.Ct. 479, 27 L.Ed.2d 433 (1971), that court held that
an employee of a corporation, though not a member of its control group, is sufficiently identified with the corporation so that his communication to the corporation’s attorney is privileged where the employee makes the communication at the direction of his superiors in the corporation and where the subject matter upon which the attorney’s advice is sought by the corporation and dealt with in the communication is the performance by the employee of the duties of his employment.
A few district courts outside the Seventh Circuit have credited this approach;4
More recently, the Court of Appeals for the Eighth Circuit, sitting en banc, agreed that the control-group test was too restrictive but concluded that the pure subject-matter test announced in Harper & Row was too broad. In Diversified Industries, Inc. v. Meredith, 572 F.2d 596 (8th Cir. 1977), it chose a modified subject-matter test first proposed in Judge Weinstein’s treatise on the Federal Rules of Evidence. See 2 Weinstein’s Evidence ¶ 503(b)[04] (1975). That test would protect an employee’s communication with corporate counsel if
(1) the communication was made for the purpose of securing legal advice; (2) the employee making the communication did so at the direction of his corporate superi- or; (3) the superior made the request so [1235]*1235that the corporation could secure legal advice; (4) the subject matter of the communication is within the scope of the employee’s corporate duties; and (5) the communication is not disseminated beyond those persons who, because of the corporate structure, need to know its contents.
572 F.2d at 609.5 As noted earlier, legal commentators have seized upon the issue and have themselves proposed a host of alternative tests.6
Confronted with such an array of possibilities, we feel compelled to examine certain basic principles. First, as all courts and commentators seem to agree, the attorney-client privilege exists to foster disclosure and communication between the attorney and the client. See 8 Wigmore on Evidence § 2291, at 545 (McNaughton rev. 1961). Nevertheless, because the privilege obstructs the search for the truth and because its benefits are, at best, “indirect and speculative,” it must be “strictly confined within the narrowest possible limits consistent with the logic of its principle.” Id. at 554. Cf. Herbert v. Lando, - U.S. -, -, 99 S.Ct. 1635, 60 L.Ed.2d 115 (1979) (“Evidentiary privileges in litigation are not favored . . . .”). Moreover, although the need for a rule of predictable application has been questioned, especially in the corporate context, see Note, The Attorney-Client Privilege: Fixed Rules, Balancing, and Constitutional Entitlement, 91 Harv.L.Rev. 464 (1977), we agree with the majority view that the incentive to confide is at least partially dependent upon the client’s ability to predict that the communication will be held in confidence. See, e.g., 2 Weinstein’s Evidence, supra, at p. 503-44; Note, 84 Harv.L.Rev., supra note 3, at 426-27, 430.
With these principles in mind, we believe that the control-group test offers a suitable starting place for our analysis. That test, adopted by a majority of the federal courts, draws as bright a line as any of the proposed approaches. More important, there seems to be a consensus that this test affords a corporation the bare minimum of protection. Although many have argued that the test’s protection is inadequate, no one yet has criticized it as overly generous. But see Radiant Burners, Inc. v. American Gas Ass’n, 207 F.Supp. 771 (privilege is not available at all to corporate clients), aff’d on reconsideration, 209 F.Supp. 321 (N.D.Ill.1962), rev’d, 320 F.2d 314 (7th Cir.) (en banc), cert. denied, 375 U.S. 929, 84 S.Ct. 330, 11 L.Ed.2d 262 (1963). As even one critic of the control-group test has noted, “[t]he idea at the core of the control group cases is a sound one: to restrict the application of the privilege so that it shelters only those communications that it is socially desirable to protect.” Kobak, supra note 6, at 365-66. We believe that it is socially desirable to protect, at a minimum, communications made by a person who has the authority to take part in a decision about any action to be taken in response to the solicited advice. Whether the privilege should be enlarged beyond that point should depend upon whether a broader rule would serve the policy of full communication underlying the privilege itself.7
[1236]*1236Criticism of the control-group test usually begins with the observation that a corporate client is different from an individual client. An individual can tell the attorney the relevant facts and then act on the attorney’s advice. In the corporate setting, however, the people who know the relevant facts and the people who make the decisions are seldom the same. If an attorney is going to give sound advice to the control group, he must secure information from outside the control group. The attorney’s need for this information allegedly dictates the conclusion that the communications should be privileged. See, e.g., Diversified Industries, Inc. v. Meredith, supra, at 608-09; Kobak, supra note 6, at 368; Note, 69 Mich.L.Rev., supra note 6, at 374.
Although we agree that an attorney often needs to secure information from lower-echelon employees, we are not convinced that extension of the corporation’s attorney-client privilege would enhance his or her ability to secure that information. The “confidentiality”' offered to non-control-group employees would be quite illusory from their standpoint. Because they have no control over the privilege itself, their communications remain confidential only in the sense that they are not released to outsiders, and only as long as the corporate control group desires to assert the privilege. If the employees had engaged in questionable activity, the corporation clearly would have the power to waive the privilege and to turn the employees’ statements over to law enforcement officials. See, e. g., Diversified Industries, Inc. v. Meredith, supra, at 611 n.5. Privilege or no privilege, lower-level employees would confide in corporate counsel at their own risk. Conversely, where no questionable activity is involved, non-control-group employees have little reason not to relate information to corporate counsel, especially where a superior has instructed them to do so. In short, we do not believe that extension of the corporation’s privilege against disclosure would significantly add to an attorney’s ability to obtain information from employees outside the control group.
Other critics of the control-group test have focused on the possibility that an attorney will be less willing to ferret out relevant information if it will not be privileged. In Diversified, for example, the Eighth Circuit expressed concern about confronting corporate counsel with a dilemma: “ ‘If he interviews employees [outside the control group], their communications to him will not be privileged. If, on the other hand, he interviews only [the control group], he may find it extremely difficult, if not impossible, to determine what happened.’ ” 572 F.2d at 609, quoting Wein-schel, Corporate Employee Interviews and the Attorney-Client Privilege, 12 B.C. Ind. & Comm.L.Rev. 873, 876 (1970) (emphasis in original). We believe, however, that the Supreme Court adequately dealt with this dilemma in Hickman v. Taylor, 329 U.S. 495, 67 S.Ct. 385, 91 L.Ed. 451 (1947). There, the Court considered the detrimental effect of allowing an opponent to discover information collected by an attorney in the course of an investigation:
Were such materials open to opposing counsel on mere demand, much of what is now put down in writing would remain unwritten. An attorney’s thoughts, heretofore inviolate, would not be his own. Inefficiency, unfairness and sharp practices would inevitably develop in the giving of legal advice and in the preparation of cases for trial. The effect on the legal profession would be demoralizing. And [1237]*1237the interests of the clients and the cause of justice would be poorly served.
Id. at 511, 67 S.Ct. at 894. In balancing these concerns against the liberal policies of discovery, the Court concluded that the protection afforded work product should be qualified, not absolute. That doctrine, of course, only applies when the attorney acts in anticipation of litigation. Nevertheless, where there is no prospect of litigation, corporate counsel has little reason to be apprehensive about the unprivileged nature of his investigation. Moreover, we believe that an attorney should and would resolve any remaining apprehension in favor of a complete investigation.
Some courts and commentators have suggested that the control-group test discourages a corporation from conducting internal investigations. See, e.g., Diversified Industries, Inc. v. Meredith, supra, at 609; Report of the Committee on the Federal Courts of the New York County Lawyers Association 7-10 (April 1970), quoted in 2 Weinstein’s Evidence, supra, at p. 503-12 n.1. The short answer to this argument is that they have little choice. We do not doubt that the ability to conduct a confidential investigation would make “compliance with the complex laws governing corporate activity” more palatable, see Report of the Committee on Federal Courts, supra; we do doubt, however, that a corporation would risk civil or criminal liability under those complex laws by foregoing introspection. In our opinion, the potential costs of undetected noncompliance are themselves high enough to ensure that corporate officials will authorize investigations regardless of an inability to keep such investigations completely confidential. See Note, 84 Harv.L.Rev., supra note 3, at 431-32. Moreover, having been alerted to the applicable rule, corporate officials will attempt to adapt their dealings with counsel to maximize both the confidentiality and the utility of their communications. See e. g., Brown & Hyman, The Scope of the Attorney-Client Privilege in Corporate Decision Making, 26 Bus.Law. 1145, 1156-57 (1971). By adopting the majority approach, we would move the federal courts one step closer to a uniform rule, thereby facilitating such adaptation.
Perhaps, as Judge Heaney has argued, application of the control-group test will result in less frequent use of attorneys as corporate sleuths. See Diversified Industries, Inc. v. Meredith, supra, at 606 (Hea-ney, J., concurring and dissenting from panel opinion). A broader test undoubtedly would encourage the employment of outside counsel to conduct internal corporate investigations. Judge Heaney felt that this incentive was desirable because attorneys “are not only professionally trained but are obligated by their code of ethics to make a thorough and complete report.” Id. Even assuming the desirability of such an incentive, however, we do not believe that application of the- narrower control-group test will significantly reduce it. As one commentator has noted, attorneys and corporations are “ineluctable bedfellows.” Note, 91 Harv.L.Rev., supra, at 474. Furthermore, in appropriate cases, an attorney will be able to offer the corporation the limited confidentiality afforded by the work-product doctrine.
In sum, we find no persuasive reason to deviate from the approach taken by the majority of the federal courts. We believe that the control-group test is both broad enough and flexible enough to accommodate the needs of a corporate client Having reached this conclusion, we have no difficulty applying it to the facts of this case. Sun has conceded that none of the interviewed employees were members of the relevant control group, as defined by the ability to take part in a decision about action to be taken in response to Pepper’s advice on this matter. None of the contested documents, therefore, is protected by the attorney-client privilege. Because we conclude that the attorney-client privilege does not apply in this case, we need not reach the government’s claim that Sun waived the privilege.
Ill
The judgment of the district court will be reversed insofar as it quashed that portion [1238]*1238of the subpoena seeking the questionnaire and interview memoranda derived from Sun’s deceased employee. In all other respects the judgment of the district court will be affirmed in accordance with this opinion. The parties will bear their own costs.