dissenting.
I respectfully dissent.
The Internal Revenue Service (“IRS” or “Service”) did not issue the summonses for El Paso’s tax pool analysis to investigate fraud or negligence, nor to determine or discover any information respecting the ac*546tual receipts, disbursements or transactions engaged in by El Paso, the accuracy and completeness of its records, or any other factual information respecting El Paso’s tax liability or the correctness of its return.1 On the contrary, it is evident that the sole purpose of the summonses was to discover the post-return “private thoughts and theories of the taxpayer”2 and “the thinking of [its] accounting analysts and policy makers about tax decisions.”3 Further, it is apparent that the IRS does not seek these theories and opinions of the taxpayer concerning its prior taxes because they may be relevant to the actual correctness of El Paso’s returns or the actual amount of tax owed. The Service asserts no such potential relevance here. Rather, as the majority candidly recognizes, the taxpayer’s theories and opinions are sought merely because they may “focus and concentrate the Service’s energy,” “may be useful to the IRS as a ‘road map’ through a company’s tax return,” and will enhance “the efficiency” of the “IRS audit.” Accordingly, here the bottom line is not figures or facts or even opinions; it is, rather, the convenience of the Service. That will not suffice. United States v. Coopers & Lybrand, 550 F.2d 615, 621 (10th Cir. 1977); United States v. Mates, 487 F.2d 1271, 1275 (8th Cir. 1973) (“The term ‘relevant’ connotes and encompasses more than ‘convenience.’ ... [T]he government failed to sustain its burden of proof by alleging a general need for a ‘road map.’ ”).
A court should not enforce these summonses unless authority for them exists in the following portions of 26 U.S.C. § 7602:
“For the purpose of ascertaining the correctness of any return, .. . [or] determining the liability of any person for any internal revenue tax, ... the Secretary is authorized-—
“(1) To examine any books, papers, records, or other data which may be relevant or material to such inquiry; ...”
The statute requires, first, a showing of potential relevance.4 The matter to which the examined books or data must be potentially relevant is “such inquiry.” It appears that Congress used “inquiry” here in the sense of “question,” to refer to the subject matter inquired about rather than the process or performance of investigation to elicit the correct answer to the question.5 Ac*547cordingly, “such inquiry” or “such question” is properly understood as referring to the question of “the correctness of any return” or “the liability of any person for any internal revenue tax.” Thus, the section means that the books, papers, records or other data which the Service may summon are only those that it seeks because of their potential relevance to the correctness of a return or the liability of any person for a tax, as opposed to those the Service seeks merely because they have the potential of facilitating or making more efficient the process of examination. In this case, the IRS clearly seeks the tax pool analysis for the latter purpose, not the former.
I have no doubt that a congressional desire to promote effective and efficient tax determination and collection played a substantial part in the decision to grant the Service the summons power provided for in sections 7602 and 7603. Nevertheless, what Congress said in essence was not that the Service would have summons power whenever it might facilitate the examination process, but rather that the process of tax determination and collection would be enhanced and facilitated by giving the Service the power to summon records and other data that might be relevant to the correctness of returns or the liability of persons for taxes. After all, for a great many years the Service effectively employed its summons power in routine audits (such as that involved here) of large corporate taxpayers without the necessity of attempting to acquire their “tax pool” analyses. The practice of attempting in cases of this kind to summon the tax pool analysis is almost uniformly recognized as being both a qualitatively significant and a relatively recent expansion of the Service’s prior practice. So far as I am aware, Coopers & Lybrand is the first appellate decision dealing with the question, and the instant case is the first appellate decision upholding the practice. One can hardly argue, then, that Congress must have intended to grant this power to the Service in enacting section 7602 because without it that section would be substantially deficient in providing for efficient and effective methods of ascertaining and collecting corporate taxes. In this connection, it is noteworthy that former Commissioner of Internal Revenue Mortimer Cap-lin, certainly a person fully familiar with the past practice and needs of the Service, has stated that the Service “should limit its requests to only those portions of the [tax pool analysis] workpapers containing factual information relevant to the tax audit.”6
I recognize that this Court, in common with many others, has held that the general test for potential relevance of summonsed books or records is, as stated in Foster v. United States, 265 F.2d 183, 187 (2d Cir. 1959), cert. denied, 360 U.S. 912, 79 S.Ct. 1297, 3 L.Ed.2d 1261 (1960), “whether the inspection sought ‘might have thrown light upon’ the correctness of the taxpayer’s returns.”7 See United States v. Wyatt, 637 F.2d 293, 300 (5th Cir. 1981). Nevertheless, virtually all these cases involve records sought because of their potential to provide factual information relating to the actual receipts, disbursements or transactions giving rise to the taxes in question.8 That the *548taxpayer may not have based its return on this information does not, of course, change either its factual nature or the character of its relevance to the actual correctness of the return.9 Further, courts facing the summons question in different contexts have somewhat narrowed the definition of “relevant” to require “a realistic expectation rather than an idle hope that something may be discovered.” United States v. Harrington, 388 F.2d 520, 524 (2d Cir. 1968); United States v. Wyatt, 637 F.2d at 300-01. Now, the courts are called upon to apply the relevancy test in yet another context, that of a summons of the tax pool analysis for aid in a routine audit where the summons is not sought to procure factual data. In this situation, I believe that “throw light upon” means throw factual light upon.10
The context in which “relevant” appears enhances the propriety of this construction. It is “books, papers, records, or other data” that must be potentially relevant. The word “other” suggests that the books, papers, and records referenced are those whose contents share common general characteristics with or are analogous to “data.” As former Commissioner Caplin has aptly observed respecting section 7602(1):
“Little discussion has taken place on the scope of the word ‘data’ as contained in that Section. What comes to mind in the use of that term are items such as books, records and other factual materials — not opinions, projections, conjectures and other thought processes. Free access to accountants’ tax accrual files, containing information of the second type, would be of obvious help to the IRS as a convenience, roadmap and time-saving device in carrying out its tax audit responsibility. But is it clear that Congress intended that that category of work product should be readily available to the IRS? Is there any evidence—to paraphrase Circuit Judge Sneed in SEC v. Arthur Young & Co., 590 F.2d 785 (CA-9,1979)—that Congress intended Section 7602 to be a ‘conscription bill’ to make the accounting profession ‘an enforcement arm’ of the IRS? It is hard to believe that even the most ardent in the IRS would strain the summons authority to fit the latter mold. As the Supreme Court reiterated in [United States v.] Bisceglia, 420 U.S. 141 [95 S.Ct. 915, 43 L.Ed.2d 88], (1975), the summons power of the IRS ‘is a limited power and should be kept within its proper bounds.’ ” Caplin, note 6 supra, at 199-200.
Further, while United States v. Powell, 379 U.S. 48, 57-58, 85 S.Ct. 248, 255, 13 L.Ed.2d 112, 119 (1964), teaches us that section 7605(b)’s provision that “[n]o tax*549payer shall be subjected to unnecessary examination or investigations” does not impose any character of “probable cause” requirement on the Service, the quoted language nonetheless appears to bespeak a congressional hostility to attenuated claims of relevance or to an expansion of “relevant” to include that which is in essence only for the convenience of the Service.
Of course, I recognize that the tax pool analysis may in some instances throw factual light upon the correctness of the return or the liability for tax. The opinions or knowledge of the taxpayer may be relevant facts in a fraud or negligence penalty case. And, it is not inconceivable that under certain circumstances the tax pool analysis may contain relevant factual data that is not otherwise available. A summons of the tax pool analysis for such a purpose would ordinarily meet the “may be relevant” test. However, under the Powell criteria, such information must not be “already within the Commissioner’s possession,” and must be potentially “relevant to the purpose” of the summons. 379 U.S. at 57-58, 85 S.Ct. at 254r-55. Moreover, as noted, Congress has mandated that “[n]o taxpayer shall be subjected to unnecessary examination or investigation.” 26 U.S.C. § 7605(b). Accordingly, we should not sustain a summons of a tax pool analysis, even if it may Incidentally disclose some relevant data, when the Service’s purpose, as here,11 is not to elicit this data, but is rather to discover the taxpayer’s post-return thought processes, theories, and opinions for use as a convenient road map in the performance of the audit of the return.
My views on the relevance of the tax pool analysis are, I believe, essentially those espoused by former Commissioner Caplin and are supported by the decision in Coopers & Lybrand and, inferentially, by the Matras opinion. I recognize, however, that they are at variance not only with those of my learned colleagues on this panel but also with those of the Second Circuit panel in United States v. Arthur Young & Co., 677 F.2d 211 (2d Cir. 1982). Nevertheless, in Arthur Young a majority of the Court, recognizing the potentially far-ranging adverse consequences of its decision on relevance, felt it necessary to in effect create an auditor’s work product privilege respecting the tax pool analysis. In my view, the Second Circuit’s concerns are better accommodated by refusing to give an overly broad construction to the necessarily somewhat imprecise and flexible criteria embodied in section 7602(1).
Accordingly, I would hold that the IRS made an insufficient showing of potential relevance, and would therefore reverse the judgment below.
I agree with the majority that “El Paso failed to particularize its assertion of the [attorney-client] privilege and prove its case with respect to any specific document,” and with so much of the majority opinion as concerns the denial on this ground of El Paso’s attempt to defeat the summonses by invoking the attorney-client privilege. I also believe that essentially the same reasoning dictates denial of El Paso’s attempt to invoke blanket work product protection against these summonses.12
*550I am concerned, however, with the majority’s holding that El Paso waived any possible attorney-client privilege. Not only is this holding unnecessary to the . result reached, but this case presents a particularly unfortunate context within which to make such a determination of waiver. Because we really have no information whatever respecting any asserted attorney-client communication, determining whether the privilege in regard thereto has been waived is virtually impossible (which is another good reason to deny blanket assertions of the privilege).
The majority bases its waiver holding on the district court’s finding13 that El Paso “discussed ‘some of the information and many of the potential tax liability issues’ in the tax pool analysis with the independent auditors.” (Emphasis added.) The majority then asserts that the auditors’ task “carries” them “into the tax pool analysis and into at least some of the supporting memo-randa,” which are not confidential since “they are created with the knowledge that independent accountants may need access to them.” (Emphasis added).14
But what of the supporting memoranda and items in the tax pool analysis that are not discussed with or shown to the auditors? (The evidence and findings below indicate such items exist, and indeed that a substantial number of the “subject files” fell into this category.) Moreover, even regarding a subject file or tax pool item that has been “discussed” with the auditors, the scope of the waiver would appear to depend on the scope of the discussion. Surely not every discussion relating to a topic included in a file mandates a waiver of the entirety of every item in that file.15 Footnote 18 in United States v. Davis, 636 F.2d 1028, 1043 (5th Cir.), cert. denied, 454 U.S. 862, 102 S.Ct. 320, 70 L.Ed.2d 162 (1981), is far too slender a reed to bear the full weight of such an extensive holding.16 Nor does United States v. Pipkins, 528 F.2d 559, 563 (5th Cir.), cert. denied, 426 U.S. 952, 96 S.Ct. 3177, 49 L.Ed.2d 1191 (1976), provide significant guidance in this setting.17
Moreover, in my view substantial concerns argue in favor of holding that the attorney-client privilege is not waived by confidential disclosures to the client’s independent auditors. Of course, as to information which the client intends the auditors to disclose to others or can anticipate a sub*551stantial possibility that such disclosure may be made in the proper fulfillment of the auditors’ engagement, no such protection should exist. In a practical, operational sense the corporation and its outside auditor regard the information which the latter acquires in the course of its work, especially that within the scope of the attorney-client privilege, as fully confidential except insofar as the nature of the engagement contemplates the actual or likely disclosure of same to third parties. The reality of the matter is that for publicly held companies very little information can properly be held entirely confidential from independent auditors. If we do not take these realities into account in developing and applying theories of waiver of the attorney-client privilege in these situations, there is a substantial danger that the attorney-client privilege for publicly held corporations, so recently reaffirmed in Upjohn Co. v. United States, 449 U.S. 383, 101 S.Ct. 677, 66 L.Ed.2d 584 (1981), will become nothing but an empty theory. This danger is especially acute if courts broadly apply the concept of waiver of the entirety of a communication by the disclosure of any significant portion of it. Though informed by some of the same considerations, the suggestion here made does not create a new privilege or cover anywhere near as wide a range of communications as the rule adopted in Arthur Young & Co. Moreover, the restrictions on waiver hypothesized here are considerably narrower, and more in keeping with the concept of expected confidentiality, than the limited waiver approved in Diversified Industries, Inc. v. Meredith, 572 F.2d 596, 611 (8th Cir. 1977) (en banc), involving material disclosed to a public body.18 And, there is no reason to believe that the courts will be any less able to control abuse of the privilege in this context than they have been in others. See, e.g., United States v. Woodall, 438 F.2d 1317, 1324 (5th Cir. 1970) (en banc), cert. denied, 403 U.S. 933, 91 S.Ct. 2262, 29 L.Ed.2d 712 (1971).
For these reasons I cannot accept the majority’s blanket and unnecessary invocation of waiver of the attorney-client privilege in the context of this case.
The majority, as a part of its affirmance of the district court’s judgment, lifts our previously granted stay of that judgment, in accordance with our normal practice. Of course, no motions for rehearing or for a continuation of the stay have yet been filed, and our mandate will not issue until action on a timely filed motion for rehearing, so the previously granted stay remains in effect at least until then. Accordingly, it is premature at this time to consider whether we should grant a further stay, and I do not understand the majority to prejudge that potential issue. I note, however, that compliance with the summons will moot further review, United States v. Arthur Andersen & Co., 623 F.2d 720 (1st Cir. 1980), cert. denied, 449 U.S. 1021, 101 S.Ct. 588, 66 L.Ed.2d 483 (1980), that it does not appear a further stay will seriously affect the IRS, and that the instant decision, the first by an appellate court upholding an IRS tax pool analysis summons, is essentially contrary to the Tenth Circuit’s holding in Coopers & Lybrand. Under these circumstances a further stay may be proper even if it were determined that the applicant’s chances of success on the merits are not sufficiently strong as to be deemed “probable.” See Ruiz v. Estelle, 650 F.2d 555, 565-66 (5th Cir. 1981); Providence Journal Co. v. FBI, 595 F.2d 889 (1st Cir. 1979).
. I agree with the majority’s analysis of the scope of the district court’s enforcement order.
. United States v. Coopers & Lybrand, 413 F.Supp. 942, 950 (D.Colo.1975), aff’d, 550 F.2d 615, 618 (10th Cir. 1977).
. United States v. Arthur Andersen & Co., 623 F.2d 725, 729 (1st Cir. 1980).
. In the phrase “may be relevant or material,” use of the words “may be” rather than “is” denotes that potential, as distinguished from actual, relevance is the test. Potential relevance is specified “since the Commissioner cannot be certain that the documents are relevant or material until he sees them.” United States v. Noall, 587 F.2d 123, 125 (2d Cir. 1978), cert. denied, 441 U.S. 923, 99 S.Ct. 2031, 60 L.Ed.2d 396 (1979).
However, the words “may be” are not used to define the character of relationship denoted by “relevant,” nor to attenuate or dilute the nexus that “relevant” implies. Rather, the expression “may be” serves merely to inform us that the Service need not establish in advance the existence of the requisite relevance or nexus to a certainty, but only to some lesser degree of likelihood. Moreover, “may be” tells us nothing about the identity of that to which the books and other data must be likely relevant.
The decisions do not make a distinction in this regard between “relevant” and “material.” See, e.g., United States v. Powell, 379 U.S. 48, 57, 85 S.Ct. 248, 255, 13 L.Ed.2d 112, 119 (1964) (Commissioner “must show that the investigation will be conducted pursuant to a legitimate purpose, [and] that the inquiry may be relevant to the purpose, ... ”); United States v. Wyatt, 637 F.2d 293, 300 (5th Cir. 1981) (“must be pursuant to and relevant to a legitimate purpose”; “the test of relevancy”).
. Section 7602(1) does not say relevant “to making such inquiry” or “to performing such inquiry," but simply “to such inquiry.” And, if “inquiry” were used in the sense of the process or performance of investigation, then “necessary or convenient to” would be more appropriate wording than “relevant or material to.” Indeed, to construe “inquiry” here as meaning the performance or process of investigation is necessarily to construe “relevant or material” as being the essential equivalent of- “convenient.” As noted, courts have consistently rejected such a construction. United States v. Coopers & Lybrand, supra; United States v. Matras, supra. See also United States v. Ar*547thur Young & Co., 496 F.Supp. 1152, 1157 (S.D.N.Y.1980), rev’d in part on other grounds, 677 F.2d 211 (2d Cir. 1982).
. Caplin, Should the Service Be Permitted to Reach Accountants’ Tax Accrual Workpapers?, 51 J. Tax. 194, 199 (1979) (emphasis added).
. However, so far as I am aware, this formulation has never been expressly adopted or approved by the Supreme Court.
. As observed in Note, A Balancing Approach to the Discoverability of Accountants’ Tax Liability Workpapers, 60 Wash.U. L.Q. 185, 195-96 (1982):
“Most of the courts construing relevance have dealt with records of the actual transactions that formed the basis of the income tax liability. The few cases in which the IRS has sought background materials that were either conjectural or only tangentially related to records of actual transactions have produced widely divergent results.” (Footnotes omitted.)
For example, in Wyatt this Court pointed out:
*548“The record clearly establishes that a prima facie showing of relevancy was made. Not only did the agent give examples of the types of transactions to which the questions were directed and how such transactions would affect the taxpayer’s income tax liabilities, but detailed the need of the discovery and explanation of reported and unreported transactions.” 637. F.2d at 301 (footnotes omitted).
. Thus, in Noall the Second Circuit approved a summons for a conventional internal audit (not a tax pool analysis) although it was not used in preparing the returns:
“Clearly the purposes of the internal audit include the detection of overstatements or understatements of revenues or expenses .... If the internal auditors have ascertained an understatement of revenues or an overstatement of expenses, this plainly might throw light on the correctness of the returns. ... The Commissioner’s interest lies in whether the tax returns correctly reflected [taxpayer] Bunge’s actual income, not simply whether they were correctly prepared from the books of account and other records used.” 587 F.2d at 126.
Obviously when the Court says “might throw light” it means the “light” thrown by factual data, e.g., an “ascertained ... understatement of revenues.”
In United States v. Euge, 444 U.S. 707, 100 S.Ct. 874, 63 L.Ed.2d 141 (1980), the Court sustained a summons (under clause (3) of section 7602) requiring the taxpayer to appear and execute handwriting exemplars. Obviously, this did not relate to preparation of the returns being examined. Just as obviously, however, the relevance was purely factual — the agent was trying to verify whether the taxpayer was maintaining several bank accounts under aliases to conceal taxable income.
. “Mental impressions, legal analysis, conclusions, and recommendations are generally not relevant.” P. T. & L. Construction Co. v. Comm’r, 63 T.C. 404, 414 (1974).
. During the course of this audit (which is presumably still in progress), the Service served on El Paso some 578 separate “Information Document Requests.” Through August 1981, El Paso had complied with 529 of these; 29 were held for the agent to return or for additional information as to whether the agent in fact wanted the documents; 2 were later complied with; 5 were in some way involved in this proceeding. The IRS has not suggested any particular area for which it wishes to elicit information by its summonses of the tax pool analysis, or any particular need for the tax pool analysis other than as a convenient road map through the entirety of the returns being audited.
. However, I do not think the record here is sufficiently particularized to support the majority’s apparent holding that none of the tax pool analysis and backup memoranda are work product materials. Surely it is not unlikely that, for example, some of the backup materials include memoranda generated in anticipation of possible litigation with the IRS, and that portions of the tax pool analysis are supported by such memoranda. Indeed, such memoranda may relate to ongoing litigation respecting problems in earlier years that are repeated in or affect the years in question. Accordingly, *550while I do not agree with the apparent inference that “if it’s in the tax pool analysis or backup it can’t be work product,” I nevertheless recognize that El Paso cannot achieve work product protection merely by invoking a blanket claim and suggesting purely hypothetical examples.
. I accept this finding, for the reasons stated by the majority.
. The majority may be factually in error here. I believe the record indicates that generally the auditors were not furnished the original or copies of the actual tax pool analysis and its backup “subject” documents. Certain of these items were discussed with the auditors (as indicated by the trial court’s findings), and if they desired further investigation or verification beyond that provided by the discussion, they went to the regular company transaction records or summaries instead of into the tax pool analysis files.
. See, e.g., United States v. Upjohn Co., 600 F.2d 1223, 1227 n. 12 (6th Cir. 1979), rev’d on other grounds, 449 U.S. 383, 101 S.Ct. 677, 66 L.Ed.2d 584 (1981); United States v. Lipshy, 492 F.Supp. 35, 44 (N.D.Tex.1979) (employee communications made to attorney in connection with preparing report for corporate client on corporate wrongdoing; held that disclosure of the report to the IRS and of portions of it to the SEC did not waive the privilege as to “all details underlying the statement”).
. For example, a young man informs his mother that he was at the convenience store which was held up earlier in the evening. At her suggestion he sees Lawyer Jones, and on returning home his mother asks, “Did you tell Lawyer Jones you were at the store when it was held up?” He acknowledges that he did so. Has he thereby waived the privilege as to his entire conversation with Lawyer Jones respecting the occurrence at the store?
. Pipkins involved handwriting samples the client gave a handwriting expert retained by his attorney. This Court rejected a claim of privilege, pointing out that Pipkins had “already voluntarily disclosed his handwriting style to the government,” 528 F.2d at 563, and that “one’s style of handwriting, readily observable by anyone, [is] not subject to the attorney-client privilege,” 528 F.2d at 563 n. 2.
. Under the Meredith limited-waiver approach, of course, the privilege does not extend to the governmental unit to which disclosure was made. Cf. United States v. Miller, 660 F.2d 563 (5th Cir. 1981) (in a criminal tax fraud case the attorney-client privilege was waived with respect to ledger books previously turned over to the IRS during its civil investigation respecting the taxes in issue). The Meredith limited-waiver theory was rejected in Permian Corp. v. United States, 665 F.2d 1214 (D.C.Cir.1981).