LEVAL, Circuit Judge:
This appeal concerns the proper interpretation of Federal Rule of Civil Procedure 26(b)(3) (“the Rule”), which grants limited protection against discovery to doeu-[1195]*1195ments and materials prepared “in anticipation of litigation.”1 Specifically, we must address whether a study prepared for an attorney assessing the likely result of an expected litigation is ineligible for protection under the Rule if the primary or ultimate purpose of making the study was to assess the desirability of a business transaction, which, if undertaken, would give rise to the litigation. We hold that a document created because of anticipated litigation, which tends to reveal mental impressions, conclusions, opinions or theories concerning the litigation, does not lose work-product protection merely because it is intended to assist in che making of a business decision influenced by the likely outcome of the anticipated litigation. Where a document was created because of anticipated litigation, and would not have been prepared in substantially similar form but for the prospect of that litigation, it falls within Rule 26(b)(3).
The district court ruled that the document sought by the IRS in this case did not fall within the scope of Rule 26(b)(3) and ordered its production. Because we cannot determine whether the district court used the correct standard in reaching its decision, we vacate the judgment and remand for reconsideration.
Background
Sequa Corporation is an aerospace manufacturer with annual revenues of nearly $2 billion. Prior to 1989, Atlantic Research Corporation (“ARC”) and Chromalloy Gas Turbine Corporation (“Chromalloy”) were wholly-owned Sequa subsidiaries. Appellant Monroe Adlman is an attorney and Vice President for Taxes at Sequa.
In the spring of 1989, Sequa contemplated merging Chromalloy and ARC. The contemplated merger was expected to produce an enormous loss and tax refund, which Adlman expected would be challenged by the IRS and would result in litigation. Adlman asked Paul Sheahen, an accountant and lawyer at Arthur Andersen & Co. (“Arthur Andersen”), to evaluate the tax implications of the proposed restructuring. Sheahen did so and set forth his study in a memorandum (the “Memorandum”). He submitted the Memorandum in draft form to Adlman in August 1989. After further consultation, on September 5, 1989, Sheahen sent Adlman the final version. The Memorandum was a 58-page detailed legal analysis of likely IRS challenges to the reorganization and the resulting tax refund claim; it contained discussion of statutory provisions, IRS regulations, legislative history, and prior judicial and IRS rulings relevant to the claim. It proposed possible legal theories or strategies for Sequa to adopt in response, recommended preferred methods of structuring the transaction, and made predictions about the likely outcome of litigation.
Sequa decided to go ahead with the restructuring, which was completed in December 1989 in essentially the form recommended by Arthur Andersen. Sequa sold 93% of its stock in ARC to Chromalloy for $167.4 million, and the remaining 7% to Bankers Trust for $12.6 million. The reorganization resulted in a $289 million loss. Se-qua claimed the loss on its 1989 return and carried it back to offset 1986 capital gains, thereby generating a claim for a refund of $35 million.
In an ensuing audit of Sequa’s 1986-1989 tax returns, the IRS requested a number of documents concerning the restructuring transaction. Sequa acknowledged the existence of the Memorandum, but cited work-product privilege as grounds for declining to [1196]*1196produce it.2 On September 23,1993, the IRS served a summons on Adlman for production of the Memorandum.
When Adlman declined to comply, the IRS instituted an action in the United States District Court for the Southern District of New York to enforce the subpoena. Adlman defended on the grounds that the Memorandum was protected by both the attorney-client and work-product privileges. The district court (Knapp, J.) in its first decision rejected Adlman’s claim that the Memorandum was protected by attorney-client privilege, finding that Adlman had not consulted Arthur Andersen in order to obtain assistance in furnishing legal advice to Sequa. United States v. Adlman, M-18-304, 1994 WL 191869, at *2 (S.D.N.Y. May 16, 1994). It rejected Adlman’s claim of work-product privilege because the Memorandum was prepared for litigation based on actions or events that had not yet occurred at the time of its creation. Id. at *3. The court granted the IRS’s petition to enforce the summons.
On appeal, we affirmed denial of Adlman’s claim of attorney-client privilege. United States v. Adlman, 68 F.3d 1495 (2d Cir.1995). We vacated the district court’s enforcement order, however, because the district court had evaluated Adlman’s claim of work-product privilege under the wrong standard. Although the non-occurrence of events giving rise to litigation prior to preparation of the documents is a factor to be considered, we explained, it does not necessarily preclude application of work-product privilege. See id. at 1501. For example, where a party faces the choice of whether to engage in a particular course of conduct virtually certain to result in litigation and prepares documents analyzing whether to engage in the conduct based on its assessment of the likely result of the anticipated litigation, we concluded that the preparatory documents should receive protection under Rule 26(b)(3). Id. We therefore remanded for reconsideration whether the Memorandum was protected work product. Id.
On remand, Adlman argued that the Memorandum was protected by Rule 26(b)(3) because it included legal opinions prepared in reasonable anticipation of litigation. Litigation was virtually certain to result from the reorganization and Sequa’s consequent claim of tax losses. Sequa’s tax returns had been surveyed or audited annually for at least 30 years. In addition, the size of the capital loss to be generated by the proposed restructuring would result in a refund so large that the Commissioner of Internal Revenue would be required by federal law to submit a report to the Joint Congressional Committee on Taxation. See 26 U.S.C.A. § 6405(a). Finally, Sequa’s tax treatment of the restructuring was based on an interpretation of the tax code without a case or IRS ruling directly on point. In light of the circumstances of the transaction, Adlman asserted there was “no doubt that Sequa would end up in litigation with the IRS.” Sequa’s accountant at Arthur Andersen concurred, opining that “any corporate tax executive would have realistically predicted that this capital loss would be disputed by the IRS” because of the “unprecedented and creative nature of the reorganization, the fact that Sequa was continually under close scrutiny by the IRS and the size of the refund resulting from the capital loss.”
The district court again rejected the claim of work-product privilege, concluding that the Memorandum was not prepared in anticipation of litigation. M-18-304, 1996 WL 84502, at *1 (S.D.N.Y. Feb. 27, 1996). Adl-man appeals.
Discussion
The work-product doctrine, codified for the federal courts in Fed.R.Civ.P. 26(b)(3), is intended to preserve a zone of privacy in which a lawyer can prepare and develop legal theories and strategy “with an eye toward litigation,” free from unnecessary intrusion by his adversaries. Hickman v. Taylor, 329 U.S. 495, 510-11, 67 S.Ct. 385, 393-94, 91 L.Ed. 451 (1947). Analysis of one’s case “in antiei-[1197]*1197pation of litigation” is a classic example of work product, see NLRB v. Sears, Roebuck & Co., 421 U.S. 132, 154, 95 S.Ct. 1504, 1518, 44 L.Ed.2d 29 (1975), and receives heightened protection under Fed.R.Civ.P. 26(b)(3).
This case involves a question of first impression in this circuit: whether Rule 26(b)(3) is inapplicable to a litigation analysis prepared by a party or its representative in order to inform a business decision which turns on the party’s assessment of the likely outcome of litigation expected to result from the transaction. Answering that question requires that we determine the proper interpretation of Rule 26(b)(3)’s requirement that documents be prepared “in anticipation of litigation” in order to qualify for work-product protection.
I.
In Hickman v. Taylor, 329 U.S. 495, 67 S.Ct. 385, 91 L.Ed. 451 (1947), the Supreme Court held that notes taken by the defendant’s attorney during interviews with witnesses to the event that eventually gave rise to the lawsuit in the ease were not discoverable by the plaintiff. 329 U.S. at 510, 67 S.Ct. at 393. As the Court explained,
In performing his various duties, ... it is essential that a lawyer work with a certain degree of privacy, free from unnecessary intrusion by opposing parties and their counsel. Proper preparation of a client’s ease demands that he ... prepare his legal theories and plan his strategy without undue and needless interference.
Id. at 510-11, 67 S.Ct. at 393.
Were the attorney’s work accessible to an adversary, the Hickman court cautioned, “much of what is now put down in writing would remain unwritten” for fear that the attorney’s work would redound to the benefit of the opposing party. Id. at 511, 67 S.Ct. at 393-94. Legal advice might be marred by “inefficiency, unfairness and sharp practices,” and the “effect on the legal profession would be demoralizing.” Id. at 511, 67 S.Ct. at 394. Neither the interests of clients nor the cause of justice would be served, the court observed, if work product were freely discoverable. Id.
The Supreme Court has reaffirmed the “strong public policy” underlying the work-produet privilege in the decades since Hickman. See United States v. Nobles, 422 U.S. 225, 236, 95 S.Ct. 2160, 2169, 45 L.Ed.2d 141 (1975); Upjohn Co. v. United States, 449 U.S. 383, 398, 101 S.Ct. 677, 687, 66 L.Ed.2d 584 (1981). It has also made clear that documents that “tend[] to reveal the attorney’s mental process” — described by commentators as “opinion work product,” see Special Project, The Work Product Doctrine, 68 Cornell L.Rev. 760, 817 (1983) — receive special protection not accorded to factual material. Upjohn, 449 U.S. at 399, 101 S.Ct. at 687. Special treatment for opinion work product is justified because, “[a]t its core, the work-product doctrine shelters the mental processes of the attorney, providing a privileged area within which he can analyze and prepare his client’s case.” Nobles, 422 U.S. at 238, 95 S.Ct. at 2170.
Rule 26(b)(3) codifies the principles articulated in Hickman. The Rule states that documents “prepared in anticipation of litigation or for trial” are discoverable only upon a showing of substantial need of the materials and inability, without undue hardship, to obtain their substantial equivalent elsewhere. Even where this showing has been made, however, the Rule provides that the court “shall protect against disclosure of the mental impressions, conclusions, opinions, or legal theories of an attorney or other representative of a party concerning the litigation.”
II.
The first problem we face is to determine the meaning of the phrase prepared “in anticipation of litigation.” The phrase has never been interpreted by our circuit; furthermore, courts and commentators have expressed a range of views as to its meaning. It is universally agreed that a document whose purpose is to assist in preparation for litigation is within the scope of the Rule and thus eligible to receive protection if the other conditions of protection prescribed by the Rule are met. The issue is less clear, however, as to documents which, although prepared because of expected litigation, are intended to inform a business decision influenced by [1198]*1198the prospects of the litigation. The formulation applied by some courts in determining whether documents are protected by work-product privilege is whether they are prepared “primarily or exclusively to assist in litigation” — a formulation that would potentially exclude documents containing analysis of expected litigation, if their primary, ultimate, or exclusive purpose is to assist in making the business decision. Others ask whether the documents were prepared “because of’ existing or expected litigation — a formulation that would include such documents, despite the fact that their purpose is not to “assist in” litigation. Because we believe that protection of documents of this type is more consistent with both the literal terms and the purposes of the Rule, we adopt the latter formulation.
1. “Primarily to assist in” litigation.
The “primarily to assist in litigation” formulation is exemplified by a line of cases from the United States Court of Appeals for the Fifth Circuit. In United States v. Davis, 636 F.2d 1028 (5th Cir.), cert. denied, 454 U.S. 862, 102 S.Ct. 320, 70 L.Ed.2d 162 (1981), the Fifth Circuit denied protection to documents made in the course of preparation of a tax return. This result was well justified as there was no showing whatsoever of anticipation of litigation. In what might be characterized as a dictum, or in any event a statement going far beyond the issues raised in the ease, the court asserted that the Rule applies only if the “primary motivating purpose behind the creation of the document was to aid in possible future litigation.” Id. at 1040.
Then, in United States v. El Paso Co., 682 F.2d 530 (5th Cir.1982), cert. denied, 466 U.S. 944, 104 S.Ct. 1927, 80 L.Ed.2d 473 (1984), a large public corporation sought to shield documents that analyzed prospective liabilities that might result from litigation with the IRS over its tax returns. The documents were prepared not to assist in litigation but to establish and justify appropriate reserves in El Paso’s financial statements. Treating the Davis dictum as law, the Fifth Circuit held that because the “primary motivating force [behind the preparation of the documents was] not to ready El Paso for litigation” but rather “to bring its financial books into conformity with generally accepted auditing principles,” id. at 543, and because the documents’ liability analysis was “only a means to a business end,” id., the documents were not prepared “in anticipation of litigation” within the meaning of the Rule and enjoyed no work-product protection. The El Paso requirement that the document be prepared to aid in litigation was then applied by a Fifth Circuit judge writing for the Temporary Emergency Court of Appeals in United States v. Gulf Oil Corp., 760 F.2d 292, 296-97 (Temp.Emer.Ct.App.1985).3
We believe that a requirement that documents be produced primarily or exclusively to assist in litigation in order to be protected is at odds with the text and the policies of the Rule. Nowhere does Rule 26(b)(3) state that a document must have been prepared to aid in the conduct of litigation in order to constitute work product, much less primarily or exclusively to aid in litigation. Preparing a document “in anticipation of litigation” is sufficient.
The text of Rule 26(b)(3) does not limit its protection to materials prepared to assist at trial. To the contrary, the text of the Rule clearly sweeps more broadly. It expressly states that work-product privilege applies not only to documents “prepared ... for trial” but also to those prepared “in anticipation of litigation.” If the drafters of the Rule intended to limit its protection to documents made to assist in preparation for litigation, this would have been adequately conveyed by the phrase “prepared ... for trial.” The fact that documents prepared “in anticipation of litigation” were also included confirms that the drafters considered this to be a different, and broader category. Nothing in the Rule states or suggests that documents prepared “in anticipation of litigation” with the pur[1199]*1199pose of assisting in the making of a business decision do not fall within its scope.
In addition, the Rule takes pains to grant special protection to the type of materials at issue in this ease — documents setting forth legal analysis. While the Rule generally withholds protection for documents prepared in anticipation of litigation if the adverse party shows “substantial need” for their disclosure and inability to obtain their equivalent by other means, even where the party seeking disclosure has made such a showing the Rule directs that “the court shall protect against disclosure of the mental impressions, conclusions, opinions, or legal theories of ... [a party or its representative] concerning the litigation.” Fed.R.Civ.P. 26(b)(3) (emphasis added). As the Advisory Committee notes indicate, Rule 26(b)(3) is intended to ratify the principles that “each side’s informal evaluation of its case should be protected, that each side should be encouraged to prepare independently, and that one side should not automatically have the benefit of the detailed preparatory work of the other side.” Where the Rule has explicitly established a special level of protection against disclosure for documents revealing an attorney’s (or other representative’s) opinions and legal theories concerning litigation, it would oddly undermine its purposes if such documents were excluded from protection merely because they were prepared to assist in the making of a business decision expected to result in the litigation.
Admittedly, there are fragmentary references in the caption to the Rule and in its commentary that can be read to lend support to a contrary interpretation. The caption, for example, refers to “Trial Preparation,” and the Advisory Committee Notes make occasional reference to “trial preparation materials.” We attach small importance to those references. Given that the text of the Rule (and of the commentary) expressly goes beyond documents “prepared ... for trial” to encompass also those documents “prepared in anticipation of litigation,” we cannot read the references in the caption and commentary as overriding the text of the Rule. See United States v. Minker, 350 U.S. 179, 185, 76 S.Ct. 281, 285, 100 L.Ed. 185 (1956); Whitehouse v. United States District Court, 53 F.3d 1349, 1358 n. 12 (1st Cir.1995).
In addition to the plain language of the Rule, the policies underlying the work-product doctrine suggest strongly that work-product protection should not be denied to a document that analyzes expected litigation merely because it is prepared to assist in a business decision. Framing the inquiry as whether the primary or exclusive purpose of the document was to assist in litigation threatens to deny protection to documents that implicate key concerns underlying the work-product doctrine.
The problem is aptly illustrated by several hypothetical fact situations likely to recur:
(i) A company contemplating a transaction recognizes that the transaction will result in litigation; whether to undertake the transaction and, if so, how to proceed with the transaction, may well be influenced by the company’s evaluation of the likelihood of success in litigation. Thus, a memorandum may be prepared in expectation of litigation with the primary purpose of helping the company decide whether to undertake the contemplated transaction. An example would be a publisher contemplating publication of a book where the publisher has received a threat of suit from a competitor purporting to hold exclusive publication rights. The publisher commissions its attorneys to prepare an evaluation of the likelihood of success in the litigation, which includes the attorneys’ evaluation of various legal strategies that might be pursued. If the publisher decides to go ahead with publication and is sued, under the “primarily to assist in litigation” formulation the study will likely be disclosed to the opposing lawyers because its principal purpose was not to assist in litigation but to inform the business decision whether to publish. We can see no reason under the words or policies of the Rule why such a document should not be protected. See United States v. Adlman, 68 F.3d at 1501.
(ii) A company is engaged in, or contemplates, some kind of partnership, merger, joint undertaking, or business association with another company; the other company reasonably requests that the company fur[1200]*1200nish a candid assessment by the company’s attorneys of its likelihood of success in existing litigations. For instance, the company’s bank may request such a report from the company’s attorneys concerning its likelihood of success in an important litigation to inform its lending policy toward the company. Or a securities underwriter contemplating a public offering of the company’s securities may wish to see such a study to decide whether to go ahead with the offering without waiting for the termination of the litigation. Such a study would be created to inform the judgment of the business associate concerning its business decisions. No part of its purpose would be to aid in the conduct of the litigation. Nonetheless it would reveal the attorneys’ most intimate strategies and assessments concerning the litigation. We can see no reason why, under the Rule, the litigation adversary should have access to it. But under the Fifth Circuit’s “to assist” test, it would likely be discoverable by the litigation adversary.
(in) A business entity prepares financial statements to assist its executives, stockholders, prospective investors, business partners, and others in evaluating future courses of action. Financial statements include reserves for projected litigation. The company’s independent auditor requests a memorandum prepared by the company’s attorneys estimating the likelihood of success in litigation and an accompanying analysis of the company’s legal strategies and options to assist it in estimating what should be reserved for litigation losses.
In each scenario, the company involved would require legal analysis that falls squarely within Hickman’s area of primary concern — analysis that candidly discusses the attorney’s litigation strategies, appraisal of likelihood of success, and perhaps the feasibility of reasonable settlement. The interpretation of Rule 26(b)(3) advocated by the IRS imposes an untenable choice upon a company in these circumstances. If the company declines to make such analysis or scrimps on candor and completeness to avoid prejudicing its litigation prospects, it subjects itself and its co-venturers to ill-informed de-cisionmaking. On the other hand, a study reflecting the company’s litigation strategy and its assessment of its strengths and weaknesses cannot be turned over to litigation adversaries without serious prejudice to the company’s prospects in the litigation. Cf. Hickman, 329 U.S. at 516, 67 S.Ct. at 396 (Jackson, J., concurring) (“Discovery was hardly intended to enable a learned profession to perform its functions either without wits or on wits borrowed from the adversary.”).
We perceive nothing in the policies underlying the work-product doctrine or the text of the Rule itself that would justify subjecting a litigant to this array of undesirable choices. The protection of the Rule should be accorded to such studies in these circumstances. See United States v. Adlman, 68 F.3d at 1501 (“[T]here is no rule that bars application of work-product protection to documents created prior to the event giving rise to litigation.”). We see no basis for adopting a test under which an attorney’s assessment of the likely outcome of litigation is freely available to his litigation adversary merely because the document was created for a business purpose rather than for litigation assistance. The fact that a document’s purpose is business-related appears irrelevant to the question whether it should be protected under Rule 26(b)(3).4
[1201]*1201We note that in Delaney, Migdail & Young, Chartered v. IRS, 826 F.2d 124 (D.C.Cir.1987), the IRS successfully argued against the very position it here advocates. The D.C. Circuit sustained the IRS’s claim of work-product privilege in circumstances where the claim would have failed under the test applied by the Fifth Circuit and advocated by the IRS on this appeal. The documents sought in Delaney were prepared by IRS attorneys for a business purpose — to help the IRS decide whether to adopt a proposed system of statistical sampling for its corporate audit program for large accounts. However, the study was prepared because of expected litigation which would result from adoption of the program; it analyzed expected legal challenges to the use of the proposed program, potential defenses available to the agency, and the likely outcome. Based on the preparatory study, the IRS concluded that the proposed statistical sampling program presented an acceptable legal risk and authorized it. The court refused discovery with the observation that the party requesting discovery was
seeking the agency’s attorneys’ assessment of the program’s legal vulnerabilities in order to make sure it does not miss anything in crafting its legal case against the program. This is precisely the type of discovery the [Supreme] Court refused to permit in Hickman v. Taylor.
Id. at 127 (emphasis added) (footnote omitted). The Seventh Circuit has also considered and rejected the contention that documents automatically fall outside the scope of the work-produet doctrine when they are prepared for purposes other than assistance in litigation. See In re Special September 1978 Grand Jury (II), 640 F.2d 49, 61-62 (7th Cir.1980).5
Similarly, several district courts have rejected the contention that Rule 26(b)(3) does not apply to documents that aré not prepared for the primary or exclusive purpose of assisting in that litigation. See Vanguard Sav. and Loan Ass’n v. Banks, No. CIV. 93-CV-4627, 1995 WL 555871, at *4 (E.D.Pa.1995) (letters from defendant’s attorneys giving the attorneys’ opinion as to the likely outcome of pending litigation protected from discovery even though the letters “were created primarily for the business purpose of satisfying the regulatory examination required by the Pennsylvania Department of Banking”); American Optical Corp. v. Medtronic, 56 F.R.D. 426, 431 (D.Mass.1972) (memoranda prepared by a lawyer in the course of counseling a client whether to accept a license or challenge the validity of a competitor’s patent in court protected work product); Sylgab Steel & Wire Corp. v. Imoco-Gateway Corp., 62 F.R.D. 454, 456 (N.D.Ill.1974) (opinion letters analyzing whether proposed products violated patents is protected work product), aff'd, 534 F.2d 330 (7th Cir.1976); Chemcentral/Grand Rapids Corp. v. EPA, No. 91 C 4380, 1992 WL 281322, at *5 (N.D.Ill. Oct. 6, 1992) (EPA documents analyzing likely legal challenges to proposed toxic waste cleanup plans are protected work product); In re Woolworth Corp. Sec. Class Action Litig., No. 94 CIV. 2217 (RO), 1996 WL 306576, at *3 (S.D.N.Y. June 7, 1996) (company’s business purpose for creating material would not preclude application of Rule 26(b)(3)).
The few commentators to have specifically addressed whether the work-product doctrine should apply to documents analyzing [1202]*1202anticipated litigation, but prepared to assist in a business decision rather than to assist in the conduct of the litigation, have generally concluded that protection is desirable. One argues that the work-produet doctrine is intended to protect a lawyer’s (or other representative’s) personal evaluation of his or her case, and to ensure that adversaries have the opportunity to memorialize their mental impressions, strategies, and ideas free from the concern that their litigation opponents might gain access to the material. Special Project, The Work Product Doctrine, 68 Cornell L.Rev. at 784-86. “The fact that the materials serve other functions apart from litigation does not mean that they should not be protected by work-product immunity if they reveal directly or indirectly the mental impressions or opinions of the attorney who prepared them.” Note, The Work Product Doctrine: Why Have an Ordinary Course of Business Exception?, 1988 Colum.Bus. L.Rev. 587, 604. Allowing discovery of this type of material has also been characterized as an “intolerable intrusion on the [settlement] bargaining process ... [which] allow[s] one party to take advantage of the other’s assessment of his prospects for victory and an acceptable settlement figure.” Edward H. Cooper, Work Product of the Rulesmakers, 53 Minn.L.Rev. 1269, 1283 (1969). Under the standard advocated by the IRS, documents assessing the strengths and weaknesses of one’s ease, or the likelihood of settlement and its expected cost, would be unprotected if prepared for a business purpose rather than to assist in litigation. This result is unwarranted.
2. Prepared “because of’ litigation.
The formulation of the work-product rule used by the Wright & Miller treatise, and cited by the Third, Fourth, Seventh, Eighth and D.C. Circuits, is that documents should be deemed prepared “in anticipation of litigation,” and thus within the scope of the Rule, if “in light of the nature of the document and the factual situation in the particular ease, the document can fairly be said to have been prepared or obtained because of the prospect of litigation.” Charles Alan Wright, Arthur R. Miller, and Richard L. Marcus, 8 Federal Practice & Procedure § 2024, at 343 (1994) (emphasis added). See In re Grand Jury Proceedings, 604 F.2d 798, 803 (3d Cir.1979); National Union Fire Ins. Co. v. Murray Sheet Metal Co., Inc., 967 F.2d 980, 984 (4th Cir.1992); Binks Mfg. Co. v. National Presto Indus., Inc., 709 F.2d 1109, 1118-19 (7th Cir.1983); Simon v. G.D. Searle & Co., 816 F.2d 397, 401 (8th Cir.), cert. denied, 484 U.S. 917, 108 S.Ct. 268, 98 L.Ed.2d 225 (1987); Senate of Puerto Rico v. United States Dep’t of Justice, 823 F.2d 574, 586 n. 42 (D.C.Cir.1987).
The Wright & Miller “because of’ formulation accords with the plain language of Rule 26(b)(3) and the purposes underlying the work-produet doctrine. Where a document is created because of the prospect of litigation, analyzing the likely outcome of that litigation, it does not lose protection under this formulation merely because it is created in order to assist with a business decision.
Conversely, it should be emphasized that the “because of’ formulation that we adopt here withholds protection from documents that are prepared in the ordinary course of business or that would have been created in essentially similar form irrespective of the litigation. It is well established that work-product privilege does not apply to such documents. See Fed.R.Civ.P. 26(b)(3), Advisory Committee’s note (“Materials assembled in the ordinary course of business ... are not under the qualified immunity provided by this subdivision.”); see, e.g., National Union Fire, 967 F.2d at 984. Even if such documents might also help in preparation for litigation, they do not qualify for protection because it could not fairly be said that they were created “because of’ actual or impending litigation. See Wright & Miller § 2024, at 346 (“even though litigation is already in' prospect, there is no work-product immunity for documents prepared in the regular course of business rather than for purposes of the litigation”).
Furthermore, although a finding under this test that a document is prepared because of the prospect of litigation warrants application of Rule 26(b)(3), this does not necessarily mean that the document will be protected against discovery. Rather, it means that a [1203]*1203document is eligible for work-produet privilege. The district court can then assess whether the party seeking discovery has made an adequate showing of substantial need for the document and an inability to obtain its contents elsewhere without undue hardship. The district court can order production of the portions of the document for which a litigant has made an adequate showing. The court can focus its attention on whether the document or any portion is the type of material that should be disclosed, while retaining the authority to protect against disclosure of the mental impressions, strategies, and analyses of the party or its representative concerning the litigation.
In short, we find that the Wright & Miller “because of’ test appropriately focuses on both what should be eligible for the Rule’s protection and what should not. We believe this is the proper test to determine whether a document was prepared “in anticipation of litigation” and is thus eligible for protection depending on the further findings required by the Rule.
III.
We cannot determine from the district court’s opinion what test it followed in concluding that the Memorandum was ineligible for protection.
There are indications that the district court may have followed the “primarily to assist in litigation” test, which we here reject. At one point the opinion asserted that “the court must determine whether the party seeking [work-product protection] has shown that the ‘documents were prepared principally or exclusively to assist in anticipated or ongoing litigation.’ United States v. Construction Products Research Inc., [73 F.3d 464, 473 (2d Cir.), cert. denied, — U.S.-, 117 S.Ct. 294, 136 L.Ed.2d 213 (1996)].” (emphasis added by district court).6 Then, in stating its conclusion, the court said, “The primary purpose [of the Memorandum] was not to prepare for litigation; the primary purpose was to decide whether or not to go through with a multimillion dollar transaction.” These passages suggest the district court may have employed the test we reject.
On the other hand, the tenor of the discussion in the court’s opinion suggests it may have focussed properly on the question whether the Memorandum studying the tax implications of the contemplated restructuring would have been prepared in substantially similar form regardless whether litigation was contemplated, and thus was not prepared “because of’ the expected litigation.
We remand with instructions to the district court to reconsider the issue under the Wright & Miller test of whether “the document can fairly be said to have been pre[1204]*1204pared ... because of the prospect of litigation.” Wright & Miller, § 2024 at 343. There is little doubt under the evidence that Sequa had the prospect of litigation in mind when it directed the preparation of the Memorandum by Arthur Andersen. Whether it can fairly be said that the Memorandum was prepared because of that expected litigation really turns on whether it would have been prepared irrespective of the expected litigation with the IRS.
If the district court concludes that substantially the same Memorandum would have been prepared in any event — as part of the ordinary course of business of undertaking the restructuring — then the court should conclude the Memorandum was not prepared because of the expected litigation and should adhere to its prior ruling denying the protection of the Rule.
On the other hand, if the court finds the Memorandum would not have been prepared but for Sequa’s anticipation of litigation with the IRS over the losses generated by the restructuring, then judgment should be entered in favor of Sequa.
The IRS contends that even if the Memorandum qualifies as work product, it has made a sufficient showing of substantial need and unavailability so as to overcome the qualified protection accorded by Rule 26(b)(3). We disagree. The Memorandum falls within the most protected category of work product — that which shows the “mental impressions, conclusions, opinions or legal theories of an attorney or other representative.” Fed.R.Civ.P. 26(b)(3). The Rule makes clear that a showing of “substantial need” sufficient to compel disclosure of other work product is not necessarily sufficient to compel disclosure of such materials. We need not decide whether such opinion work product is ever discoverable upon a showing of necessity and unavailability by other means. See Duplan Corp. v. Moulinage et Retorderie de Chavanoz, 509 F.2d 730, 734 (4th Cir.1974) (opinion work product never discoverable), cert. denied, 420 U.S. 997, 95 S.Ct. 1438, 43 L.Ed.2d 680 (1975). Contra Holmgren v. State Farm Mut. Auto. Ins. Co., 976 F.2d 573, 577 (9th Cir.1992). The Rule is clear that, at a minimum, such material is to be protected unless a highly persuasive showing is made. See Upjohn, 449 U.S. at 402, 101 S.Ct. at 689. The IRS has failed to meet that high standard.
The IRS claims necessity for the Memorandum on the ground that it will provide insight into Sequa’s subjective motivation for engaging in corporate restructuring and is thus relevant to determining whether the transaction was motivated by a legitimate business purpose. See, e.g., Kirchman v. Commissioner, 862 F.2d 1486, 1492 (11th Cir.1989); Rice’s Toyota World, Inc. v. Commissioner, 752 F.2d 89, 91-92 (4th Cir.1985). In camera review of the Memorandum shows that it does not reflect the motives of Sequa’s executives, but rather the legal analysis of its accountants. While the Memorandum unquestionably presupposes a desire to achieve a favorable tax result, such a desire is in no way incompatible with the existence of a “legitimate non-tax business reason” for its choice. Newman v. Commissioner, 902 F.2d 159, 163 (2d Cir.1990); see also Frank Lyon Co. v. United States, 435 U.S. 561, 580, 98 S.Ct. 1291, 1302, 55 L.Ed.2d 550 (1978) (“The fact that favorable tax consequences were taken into account ... on entering into [a] transaction is no reason for disallowing those consequences” where the transaction has a legitimate business purpose.). The Memorandum, being the technical and legal analysis of outside accountants, and not the reflections of decisionmakers at Sequa, simply does not address or reflect Sequa’s business reason for the proposed restructuring.
Moreover, the IRS has made no showing, beyond bare assertion, that the relevant information in the Memorandum is unavailable by other means. This falls far short of the heightened showing mandated by Upjohn.'7
[1205]*1205In short, the enforceability of the IRS summons for the Memorandum will turn on whether it (or substantially the same document) would have been prepared irrespective of the anticipated litigation and therefore was not prepared because of it.
Conclusion
The order enforcing the IRS summons is vacated, and the matter is remanded to the district court for further findings under the standard prescribed in this ruling.