These appeals challenge an order enforcing summonses that the Internal Revenue Service issued to Richard Frederick. Frederick is both a lawyer and an accountant, and he both provides legal representation to, and prepares the tax returns of, Randolph and Karin Lenz and their company, KCS Industries, Inc. The IRS is investigating the Lenzes and their company, and the summonses directed Frederick to hand over hundreds of documents that may be germane to the investigation. Frederick balked at handing over all of them, claiming that some were protected by either the attorney-client privilege or the work-product privilege (or both). His refusal precipitated this enforcement proceeding. 26 U.S.C. sec. 7604(b). The district Judge examined the documents in camera and ruled that some were privileged but others were not. The appeals challenge the latter ruling.

As is generally though not always the case when an appeal challenges the application of a legal rule to the facts (sometimes called a ruling on a "mixed question of fact and law," Ornelas v. United States, 517 U.S. 690, 696-97 (1996); Pullman-Standard v. Swint, 456 U.S. 273, 289 n. 19 (1982)), our review of the Judge's ruling on the privilege claims is deferential; we ask not whether the ruling was erroneous but whether it was clearly erroneous, e.g., In re Teranis, 128 F.3d 469, 471 (7th Cir. 1997); Williams v. Commissioner, 1 F.3d 502, 505 (7th Cir. 1993), just as when we are reviewing simple factfindings. Fed. R. Civ. P. 52(a). Whether a particular document is privileged is a fact-specific and case-specific issue, the sort of issue that district Judges are particularly experienced in resolving. It is not the sort of issue that lends itself to governance by a uniform rule that a court of appeals might prescribe and enforce. In these circumstances, a light appellate touch is best.

This is generally the case when the issue on appeal is whether the trial court or jury correctly applied a rule of law to the facts. Such an issue is fact- and case-specific, and so does not lend itself to uniform resolution across different cases, the sort of resolution that requires plenary appellate review. That is why we said in the preceding paragraph that "generally" the standard for reviewing such issues is deferential. Generally is not always, and the exceptions, illustrated by Ornelas, are important. But the only exceptions the Supreme Court has carved are for certain constitutional issues, where the risk of error is thought sufficiently serious to warrant a more searching than normal review for error, and in this circuit, at least, we have been reluctant to recognize exceptions outside the constitutional area. Other courts have not been so austere; it is easy to cite a string of cases in which a court describes the standard of review for a mixed question of non-constitutional law and fact, such as fair use in a copyright case or likelihood of confusion in a trademark case; but we are not aware of any case which explains why such an issue requires plenary review, and we cannot think of any respect bearing on the optimal standard of review in which any of these issues differs from the issue of privilege in this case. The presumption in this circuit is and we hope will remain that the clear-error standard is the proper standard for appellate review of determinations of mixed questions of fact and law. This presumption is a helpful simplification of the law of appellate review, with no downside that we can see; and there is certainly nothing in the circumstances of the present case, or the class of cases that it exemplifies (non-constitutional privilege cases), to rebut it.

Most of the documents in issue were created in connection with Frederick's preparation of the Lenzes' tax returns. They are drafts of the returns (including schedules), worksheets containing the financial data and computations required to fill in the returns, and correspondence relating to the returns. These are the kinds of document that accountants and other preparers generate as an incident to preparing their clients' returns, or that the taxpayers themselves generate if they prepare their own returns, though in the latter case there is unlikely to be correspondence. The materiality of the documents to the IRS's investigation of the Lenzes is not in issue.

There is no common law accountant's or tax preparer's privilege, Couch v. United States, 409 U.S. 322, 335 (1973); United States v. Arthur Young & Co., 465 U.S. 805, 817-19 (1984), and a taxpayer must not be allowed, by hiring a lawyer to do the work that an accountant, or other tax preparer, or the taxpayer himself or herself, normally would do, to obtain greater protection from government investigators than a taxpayer who did not use a lawyer as his tax preparer would be entitled to. United States v. Lawless, 709 F.2d 485, 487-88 (7th Cir. 1983); United States v. Bornstein, 977 F.2d 112, 116-17 (4th Cir. 1992); In re Grand Jury Investigation, 842 F.2d 1223, 1224-25 (11th Cir. 1987); United States v. Davis, 636 F.2d 1028, 1043 (5th Cir. 1981). To rule otherwise would be to impede tax investigations, reward lawyers for doing non-lawyers' work, and create a privileged position for lawyers in competition with other tax preparers--and to do all this without promoting the legitimate aims of the attorney-client and work-product privileges. The attorney-client privilege is intended to encourage people who find themselves involved in actual or potential legal disputes to be candid with any lawyer they retain to advise them. Upjohn Co. v. United States, 449 U.S. 383, 389 (1981). The hope is that this will assist the lawyer in giving the client good advice (which may head off litigation, bring the client's conduct into conformity with law, or dispel legal concerns that are causing the client unnecessary anxiety or inhibiting him from engaging in lawful, socially productive activity) and will also avoid the disruption of the lawyer-client relationship that is brought about when a lawyer is sought to be used as a witness against his client. The work-product privilege is intended to prevent a litigant from taking a free ride on the research and thinking of his opponent's lawyer and to avoid the resulting deterrent to a lawyer's committing his thoughts to paper. United States v. Nobles, 422 U.S. 225, 236-39 (1975); Hickman v. Taylor, 329 U.S. 495, 510-11 (1947); id. at 516 (Jackson, J., Concurring).

Communications from a client that neither reflect the lawyer's thinking nor are made for the purpose of eliciting the lawyer's professional advice or other legal assistance are not privileged. The information that a person furnishes the preparer of his tax return is furnished for the purpose of enabling the preparation of the return, not the preparation of a brief or an opinion letter. Such information therefore is not privileged.

We do not, however, accept the government's argument that there is no issue of privilege here because the information was transmitted to a tax preparer with the expectation of its being relayed to a third party, namely the IRS. It is true that "if the client transmitted the information so that it might be used on the tax return, such a transmission destroys any expectation of confidentiality." United States v. Lawless, supra, 709 F.2d at 487; see also United States v. Windfelder, 790 F.2d 576, 579 (7th Cir. 1986); In re Grand Jury Proceedings, 727 F.2d 1352, 1356 (4th Cir. 1984). That is, the transmittal operates as a waiver of the privilege. But the tax preparer here was also the taxpayers' lawyer, and it cannot be assumed that everything transmitted to him by the taxpayer was intended to assist him in his tax-preparation function and thus might be conveyed to the IRS, rather than in his legal-representation function. Cf. United States v. (Under Seal), 748 F.2d 871, 875-76 (4th Cir. 1984).

We also reject the government's argument that numerical information can never fall within the attorney-client or work-product privilege. Cf. United States v. Schwimmer, 892 F.2d 237, 242 (2d Cir. 1989); United States v. Davis, supra, 636 F.2d at 1043; In re Grand Jury Proceedings, 601 F.2d 162, 171-72 (5th Cir. 1979). Such cases are rare, but they can be imagined. Suppose a lawyer prepared an estimate of his client's damages; the estimate would be numerical, but insofar as it reflected the lawyer's professional assessment of what to ask the jury for it would be attorney work product. Similarly, if the lawyer asked his client how much he had obtained in the theft for which he was being prosecuted and the client answered, "$10,000," the answer would be protected by the attorney-client privilege. But we do not agree with the appellants that the district Judge based his ruling on the erroneous view that numbers can never be privileged. He found no basis for privileging these numbers, remarking, rightly, "It cannot be argued that numbers in the hands of the accountant are different from numbers in the hands of a lawyer."

Besides the information supplied to Frederick by the Lenzes, there are the worksheets, which Frederick prepared and which doubtless reflect some of his own thinking. But the Supreme Court has held that an accountant's worksheets are not privileged, United States v. Arthur Young & Co., supra, 465 U.S. at 817-19, and a lawyer's privilege, as we explained earlier, is no greater when he is doing accountant's work. A complicating factor is that when Frederick was doing these worksheets and filling out the Lenzes' tax returns, he knew that the IRS was investigating the Lenzes and their company, albeit in connection with different tax years, and he was representing them in that investigation. But people who are under investigation and represented by a lawyer have the same duty as anyone else to file tax returns. They should not be permitted, by using a lawyer in lieu of another form of tax preparer, to obtain greater confidentiality than other taxpayers. By using Frederick as their tax preparer, the Lenzes ran the risk that his legal cogitations born out of his legal representation of them would creep into his worksheets and so become discoverable by the government. The Lenzes undoubtedly benefited from having their lawyer do their returns, but they must take the bad with the good; if his legal thinking infects his worksheets, that does not cast the cloak of privilege over the worksheets; they are still accountants' worksheets, unprotected no matter who prepares them.

Put differently, a dual-purpose document--a document prepared for use in preparing tax returns and for use in litigation--is not privileged; otherwise, people in or contemplating litigation would be able to invoke, in effect, an accountant's privilege, provided that they used their lawyer to fill out their tax returns. And likewise if a taxpayer involved in or contemplating litigation sat down with his lawyer (who was also his tax preparer) to discuss both legal strategy and the preparation of his tax returns, and in the course of the Discussion bandied about numbers related to both consultations: the taxpayer could not shield these numbers from the Internal Revenue Service. This would be not because they were numbers, but because, being intended (though that was not the only intention) for use in connection with the preparation of tax returns, they were an unprivileged category of numbers.

The most difficult question presented by this appeal, and one on which we cannot find any precedent, relates to documents, numerical and otherwise, prepared in connection with audits of the taxpayers' returns. An example is a memo from Frederick to a paralegal asking her for the amount that Mr. Lenz and his corporation had paid Frederick for legal services rendered personally to Lenz in 1992. The memo was prepared to help Frederick respond to questions raised in an audit of the Lenzes' and the corporation's tax returns. An audit is both a stage in the determination of tax liability, often leading to the submission of revised tax returns, and a possible antechamber to litigation. When a revenue agent is merely verifying the accuracy of a return, often with the assistance of the taxpayer's accountant, this is accountants' work and it remains such even if the person rendering the assistance is a lawyer rather than an accountant. Throwing the cloak of privilege over this type of audit-related work of the taxpayer's representative would create an accountant's privilege usable only by lawyers. If, however, the taxpayer is accompanied to the audit by a lawyer who is there to deal with issues of statutory interpretation or case law that the revenue agent may have raised in connection with his examination of the taxpayer's return, the lawyer is doing lawyer's work and the attorney-client privilege may attach. But the documents in issue do not, so far as we are able to determine, relate to such representation.

We should consider the possible bearing of a new statute, 26 U.S.C. sec. 7525, which extends the attorney-client privilege to "a federally authorized tax practitioner," that is, a non-lawyer who is nevertheless authorized to practice before the Internal Revenue Service. sec. 7525(a)(3)(A). Non-lawyers (including tax preparers, many of them accountants) have long been allowed to practice before it. 5 U.S.C. sec. 500(c); 31 C.F.R. sec.sec. 10.3, 10.7(c)(viii). The new statute protects communications between a taxpayer and a federally authorized tax practitioner "to the extent the communication would be considered a privileged communication if it were between a taxpayer and an attorney." sec. 7525(a)(1). (It does not protect work product.) Nothing in the new statute suggests that these non-lawyer practitioners are entitled to privilege when they are doing other than lawyers' work; and so the statute would not change our analysis even if it were applicable to this case, which it is not, because it is applicable only to communications made on or after July 22, 1998, the date the statute was enacted. See Note following 26 U.S.C. sec. 7525.

We have looked at all the documents that Frederick argues are privileged. Most are dual-purpose documents, about which no more may be said; some were not even submitted to the district Judge for consideration of whether they might be privileged; in others as well, privilege was waived. We cannot find any clear errors in the district Judge's rulings.

Affirmed.

Diane P. Wood, Circuit Judge, Concurring.

Diane P. Wood, Circuit Judge, Concurring.

Although I concur in the result the court reaches here, I wish to comment on its initial statement about the proper standard of review in this case. Before concluding that "a light appellate touch" is appropriate for reviewing the district court's decisions about whether particular documents were privileged--a Conclusion with which I agree--the court comments that it is "generally" the case that mixed questions of law and fact are reviewed under the clearly erroneous standard. That statement is neither here nor there for this case, since we are all agreed that a deferential standard applies on these facts. But I am concerned that it is too sweeping and could lead to error in cases requiring a more nuanced appellate review of mixed questions.

As the majority notes, the appropriate standard of review for mixed questions of law and fact is not the same in all cases, and thus this is not a question that lends itself to uniform resolution across different cases. Ante at 500. With that much, I am in full agreement. I note here that this statement, coupled with the observation that the character of documents as privileged or not privileged is a highly fact-specific inquiry for which the clearly erroneous standard is proper, is enough to decide the case before us. The majority goes on, however, to speculate that one can discern a "presumption" in the cases decided within this circuit in favor of clear error review for mixed questions of law and fact. With respect, I think this overstates the case considerably.

Some background is important to understand the way in which mixed questions have in fact been treated in this court and elsewhere. Prior to Ornelas v. United States, 517 U.S. 690 (1996), this court occasionally made sweeping statements along the lines of the majority's opinion today. See, e.g., Baldwin v. United States, 60 F.3d 363, 365 (7th Cir. 1995) ("This court . . . has moved decisively to the position that appellate review of determinations of mixed questions of fact and law should be governed by the standard of clear error, and not by the de novo standard.") (internal citations omitted), vacated, 517 U.S. 1231 (1996); see also Mucha v. King, 792 F.2d 602, 605 (7th Cir. 1986) ("[M]ost courts treat legal characterizations (negligence, possession, ratification, principal place of business, etc.) as facts to which the clearly-erroneous standard applies. That is certainly the view of this circuit.") (internal citations omitted); see generally Evan Tsen Lee, Principled Decision Making and the Proper Role of Federal Appellate Courts: The Mixed Questions Conflict, 64 S. Cal. L. Rev. 235, 240-41 (1991) (discussing the Seventh Circuit's approach to mixed questions as "follow[ing] the rule stated in Mucha v. King without regard to the particular type of mixed question."). But in Ornelas, the Supreme Court reversed our use of the clear error standard to review a district court's determination of probable cause and reasonable suspicion, and the Court underscored the point by vacating the judgment in Baldwin, supra, as well, in light of Ornelas. Since then, we have rightly exercised more caution in our choice of the proper standard for mixed questions. See In re Krehl, 86 F.3d 737, 742 (7th Cir. 1996).

The majority concedes that Ornelas is not consistent with its idea of a presumption in favor of clear error review, but it attempts to distinguish Ornelas and the cases in this circuit that have followed its methodology by observing that they typically implicate constitutional rights. It is true that one can find recent examples from this court that recognize the need for review more searching than clear error, in which the legal principle at stake came from the Constitution. So, for example, we explained in one case that "[m]ost mixed questions of law and fact that require the application of constitutional principles to historical fact receive de novo review." Coulter v. Gilmore, 155 F.3d 912, 917 (7th Cir. 1998), citing Ornelas v. United States, 517 U.S. at 697-99 and Thompson v. Keohane, 516 U.S. 99, 111-12 (1995); see also, e.g., United States v. Raszkiewicz, 1999 WL 74692, at *2 (7th Cir. Feb. 18, 1999) (reviewing de novo the mixed question of whether defendant was denied the right to a jury selected from a cross-section of the community); Mahaffey v. Page, 162 F.3d 481, 484 (7th Cir. 1998) (reviewing de novo the mixed question of whether defendant made a prima facie case of a Batson violation). But the source of the law that contributes the legal component to the "mixed" question logically has nothing to do with the depth of appellate review. Examples abound of courts that apply a stricter standard of review than clear error to mixed questions that do not implicate the Constitution. See, e.g., Indest v. Freeman Decorating, Inc., 168 F.3d 795, 1999 WL 98505, at *8 (5th Cir. Feb. 26, 1999) (sexual harassment); Nihon Keizai Shimbun, Inc. v. Comline Bus. Data, Inc., 166 F.3d 65, 72 (2d Cir. 1999) (fair use of a copyrighted work); Jet, Inc. v. Sewage Aeration Sys., 165 F.3d 419, 422 (6th Cir. 1999) (likelihood of confusion in trademark infringement action); Hoover v. Morales, 164 F.3d 221, 224 (5th Cir. 1998) (grant of a preliminary injunction); United States v. Moore, 159 F.3d 1154, 1157 (9th Cir. 1998) (conflict of interest of trial counsel); Brink's Ltd. v. South African Airways, 149 F.3d 127, 132 (2d Cir. 1998) (wilful misconduct under the Warsaw Convention).

In my view, therefore, it is at best unhelpful and at worst misleading to claim that a single standard of review "generally" applies to mixed questions, or that there is some kind of "presumption" in non-constitutional cases. We are required instead to engage in the type of nuanced, policy-driven analysis the Supreme Court outlined in Ornelas. There the Court examined its past practice, the need for appellate courts to ensure a unitary system of law, the fact that the legal principles at issue acquired their content only through application to fact patterns, and the tendency of de novo review to unify precedent. 517 U.S. at 697-99. In other cases, the Court has considered the district court's superior fact-finding capabilities and whether searching appellate review would clarify legal doctrine. Salve Regina College v. Russell, 499 U.S. 225, 233 (1991); Pierce v. Underwood, 487 U.S. 552, 559-60 (1988); Miller v. Fenton, 474 U.S. 104, 114 (1985). Only if we, too, engage in a careful review of these factors when selecting a standard of review for each different type of mixed question can we perform our proper function of expounding the law and assuring its uniform application.

In this case, the clear error standard is the right choice for the reasons I have identified: only the underlying facts about the documents are really at issue (i.e., who said what to whom, was the document written for the purpose of preparing tax returns, in what capacity did Frederick operate). Without flagging it, however, the majority has also made an important legal ruling that affected its review: dual-purpose documents are not privileged. This principle can be expressed without regard to the specific facts of a given case, including this one. Because it was easy enough to dissect this general legal principle out from the facts, it was also easy to apply clear error review to the factual components of the privilege question. Even if the question is theoretically a "mixed" one, clear error review is the appropriate choice whenever the appellate court is able cleanly to separate and focus its attention exclusively on the factual findings that lie behind the mixed question. When the meaning of the law (constitutional or otherwise) cannot be separated from the significance of the facts, however, Ornelas requires us to apply de novo review. This is not a choice that can be made in the abstract, or through presumptions. Only a careful analysis of each case will enable us to make the proper choice.