Concept
Opinion Work Product
Work product, privilege, and protective orders
Opinion Work Product
Rule 26(b)(3)(B) is the last limit on a requester who has cleared Rule 26(b)(3)(A). Under subparagraph (A), a party that shows substantial need for work-product materials and an inability to obtain their substantial equivalent without undue hardship may reach them.1 Subparagraph (B) then qualifies the resulting order: the court “must protect against disclosure of the mental impressions, conclusions, opinions, or legal theories of a party's attorney or other representative concerning the litigation.”2
The limit follows from what a need showing proves. A requester who shows need claims that particular information bears on its case and cannot be had elsewhere without undue hardship, and that claim concerns information. It gives no reason to show the requester how opposing counsel sized the information up. Lawyers who expected to surrender their evaluations whenever an adversary proved need for the facts would stop writing the evaluations down.3 The rule therefore divides work product into a part that a sufficient showing can unlock and a part that the court must protect even after the showing is made.
This unit asks four questions about that division. What puts a document in the opinion tier instead of the fact tier? How firm is the protection once a document lands there? How can a court produce the factual part of a mixed document without exposing the rest? And when does a lawyer's choice and arrangement of ordinary documents become protected thinking? In a withholding dispute, the answers supply the last step of the sequence this chapter has built: the floor that remains after the requester has used up the exception for need and hardship.
Fact Versus Opinion
Rule 26(b)(3)(B) splits work product into two tiers. Opinion work product, also called core work product, consists of the mental impressions, conclusions, opinions, and legal theories of a party's attorney or other representative, and the court must protect it even when it orders other work product produced.4 Fact work product, also called ordinary work product, is the rest, including the results of a factual investigation.5 A requester reaches the fact tier by showing need and hardship. The opinion tier is “generally afforded near absolute protection from discovery” and is discoverable only in rare and exceptional circumstances.6 The next subsection tests how firm that protection is. This one teaches the sorting, because a document placed in the wrong tier is either surrendered on a showing it should have defeated or withheld under a label it does not deserve.
The fact tier is reachable because facts rarely exist in only one place. Where work product consists of witness statements, the “other means” is interviewing the witness directly, so hardship usually turns on whether the witness is dead, beyond the court's reach, hostile, or memory-impaired.7 Hickman itself arose from interviews of this kind: counsel privately interviewed survivors of a towing accident, took their signed statements with an eye toward the anticipated litigation, and made memoranda of what other persons told him, and the adversary demanded copies of the statements and, for oral ones, the exact provisions of what was said.8
A lawyer's evaluation of those statements raises a different concern. Hickman rested on the point that proper preparation requires a lawyer to assemble information, sift the relevant from the irrelevant facts, prepare legal theories, and plan strategy “without undue and needless interference.”9 A showing of need for the statements gives no reason to expose how counsel weighed them.
That reasoning puts a lawyer's notes of a witness interview near the top of the opinion tier. Forcing an attorney to disclose notes and memoranda of witnesses' oral statements is particularly disfavored, the Court said in Upjohn Co. v. United States, because it tends to reveal the attorney's mental processes: “what he saw fit to write down regarding witnesses' remarks.”10 To the extent such notes do not reveal client communications, they reveal the attorneys' mental processes in evaluating those communications, and they cannot be disclosed simply on a showing of substantial need and inability to obtain the equivalent without undue hardship.11 The Eighth Circuit classifies interview notes the same way. In In re Green Grand Jury Proceedings, it treated an attorney's notes of witness interviews, and the attorney's recollections of conversations with the client, as opinion work product, because an attorney taking notes tends to record the facts the attorney considers legally significant. Intangible work product that consists of counsel's thoughts and recollections is often eligible for the same special protection.12
One scenario shows both tiers at work. A defendant's insurer takes a nonparty eyewitness's statement and has it transcribed. The transcript is work product, but it records only what the witness said, so it sits in the fact tier. The plaintiff can reach it by showing substantial need and undue hardship, a showing that will probably fail if the plaintiff can simply interview the witness. Defense counsel then reads the transcript and writes a memo to the file: “X appears to be lying for the following three reasons.” The memo records counsel's conclusion and reasoning, so it is opinion work product, and on the treatise's account of this example, no showing by the plaintiff entitles him to the memo.13 The next subsection explains why courts disagree about whether that immunity is truly absolute.
The opposite pole is a document that reports facts without displaying any evaluation. In United States ex rel. Bibby v. Wells Fargo Bank, N.A., the court held that two disputed bullet points were fact work product, because they were short statements about the “result of a factual investigation.” Wells Fargo still won a protective order, because the relators had not met their heavy burden of showing substantial need and undue hardship and could still discover the underlying facts.14 In Coburn Group v. Whitecap Advisors LLC, one Whitecap participant e-mailed another information he had gathered about Whitecap's dealings with Coburn, in order to answer requests from attorneys already representing Whitecap in the pending suit. The court's in camera review showed that the e-mail was work product, and because it disclosed no attorney's mental impressions, conclusions, opinions, or theories, it was not opinion work product.15 Set beside the credibility memo and the interview notes, the two decisions show what moved the answer. The opinion-tier documents displayed what counsel found legally significant or how counsel evaluated it, while the bullet points and the e-mail reported facts without showing an attorney's judgment.
The label follows content, not format. A lawyer's interview notes are not always core work product, because purely factual material embedded in them may not deserve the “super-protection” afforded to a lawyer's mental impressions.16 The setting matters too. An interview that encourages wide-ranging disclosure, without any nascent focus chosen by the lawyer, is more likely to produce fact work product, while a litigation-related investigation elicits facts that reflect a focus the lawyer chose.17 After an exhaustive in camera review in In re Intel Corp. Microprocessor Antitrust Litigation, the special master found that interview materials held both core and non-core work product but that the facts could easily be separated from the core.18 The D.C. Circuit likewise remanded a memorandum in United States v. Deloitte LLP for in camera review, because beside counsel's thoughts and analyses it might hold facts and non-attorney analyses not so intertwined with the legal analysis as to warrant protection. The redaction mechanics follow later in this unit.
Heightened protection also has an outer boundary. Almost every step a litigant takes supports some vague inference about strategy, and treating each such inference as opinion work product would erase the line between the tiers. The First Circuit therefore held in In re San Juan Dupont Plaza Hotel that heightened protection is not triggered unless disclosure creates a “real, nonspeculative danger of revealing the lawyer's thoughts.”19 The same opinion draws a second boundary. An answer to a complaint reflects counsel's choices about which averments to deny and which defenses to assert, yet no one treats a filed pleading as opinion work product, because counsel drafts it knowing it will be served on the other parties and has no justifiable expectation that its message will stay private.20
Allegations of wrongdoing test the opinion tier from another side. Courts have held that the work-product doctrine, like the attorney-client privilege, does not limit discovery into matters where a law firm allegedly acted criminally, and they have allowed discovery of opinion work product where the government made a prima facie case that the attorney suborned perjury.21 A client's wrongdoing is different, because the crime-fraud exception can strip the client's protection while leaving the lawyer's intact. The Eighth Circuit reasoned in Green that an attorney's own work-product interest protects the attorney's opinions and thought processes for the benefit of all of the attorney's clients. An attorney who is unaware of the client's wrongdoing may therefore still assert the protection for opinion work product, though the attorney may be required to disclose ordinary work product.22 The fact-opinion line thus decides what a lawyer who did nothing wrong must still hand over.
Step three decides most disputes. A document that reports what happened lands in the fact tier, and a document that shows what counsel thinks about what happened lands in the opinion tier. The next subsections show how firmly the second tier is protected and how courts separate the two within a single document.
Mandatory Protection of Mental Impressions
The rule's own words mark the difference between the tiers. Under subparagraph (A), protected materials “may be discovered” once the showing is made, which leaves the judge room to weigh it.31 Under subparagraph (B), the court “must protect” counsel's mental impressions, conclusions, opinions, and legal theories, and “those materials” points back to the work product that (A) makes discoverable, so the command travels with every order issued under (A).32 The federal rule does not flatly forbid disclosure, as some state rules do. It directs the court to guard against disclosure, which recognizes the highest protection for counsel's impressions but lets the court order a response when the information is necessary to the requester and can be provided in a way that protects those impressions.33
The mandatory verb does work that a balancing standard could not do. A showing of need and hardship proves something about information: the case cannot be prepared without it, and no other source supplies it. That proof gives the court no reason to let the requester also see how opposing counsel weighed the information, because the requester can draw its own conclusions once it holds the facts. Timing sharpens the point. Counsel decides whether to commit an evaluation to paper long before anyone moves to compel, and protection that depended on a later weighing of need would leave counsel guessing, the uncertainty that leads lawyers to write less down. Hickman drew the same connection: if a lawyer's materials were open to opposing counsel on mere demand, much of what is now put down in writing would remain unwritten, and an attorney's thoughts, heretofore inviolate, would not be his own.34 The Court added that it did not mean all written materials obtained by an adversary's counsel with an eye toward litigation are necessarily free from discovery in all cases.35
Hickman gives the strict view real support. It held that an attempt, without purported necessity or justification, to secure counsel's written statements, private memoranda, and personal recollections falls outside the arena of discovery.36 As to counsel's recollections of witnesses' oral statements, the Court did “not believe that any showing of necessity can be made under the circumstances of this case,” though it added that if a rare situation justifying production should arise, that case was not of that type.37 The opinion stops short of an absolute rule, and so does Upjohn.
In Upjohn, the Government sought notes and memoranda of witness interviews, and the magistrate had applied the ordinary standard of need and hardship. The Court declined to say that such material is always protected, but it held that the magistrate's standard was too lenient: a far stronger showing of necessity and unavailability by other means would be required to compel disclosure.38 The Government had stressed that the interviewees were scattered across the globe and that Upjohn forbade its employees to answer questions it considered irrelevant. The Court answered that Hickman's language about unavailable witnesses did not reach oral statements preserved in counsel's mental impressions or memoranda.39 Because Upjohn fixed only a floor, each position below claims some support from it, and none can say the Supreme Court has settled the ceiling.
The strictest courts treat the protection as absolute, holding that “no showing of relevance, substantial need or undue hardship should justify compelled disclosure of an attorney's mental impressions, conclusions, opinions or legal theories,” though that view is not universally accepted.40 The Fourth Circuit took that position in Duplan Corp. v. Moulinage et Retorderie de Chavanoz, treating opinion work product as never discoverable.41 The textual argument sits on the face of the rule: (A) says may, and (B) says must. State courts applying counterparts to the federal rule sometimes speak the same way. The Georgia Supreme Court said in St. Simons Waterfront, LLC v. Hunter, Maclean, Exley & Dunn that absolute protection still attaches to mental impressions after the requester shows need and hardship.
Courts that read the rule as absolute would hold that no showing compels disclosure of an evaluation like that one. Courts that stop short of an absolute rule still make the opinion tier extremely hard to reach. The Second Circuit declined in United States v. Adlman to say whether opinion work product is ever discoverable, noted the conflict between Duplan and Holmgren, and held that at a minimum such material is protected unless a highly persuasive showing is made.42 The IRS wanted a memorandum on the theory that it would reveal Sequa's motive for a corporate restructuring, but in camera review showed that the memorandum reflected the legal analysis of Sequa's accountants and not the motives of its executives.43 The Third Circuit held in In re Cendant Corp. Securities Litigation that opinion work product is not absolutely immune but requires a heightened showing of extraordinary circumstances, so that it is generally afforded near absolute protection and discoverable only in rare and exceptional circumstances.44
The Ninth Circuit's decision in Holmgren v. State Farm Mutual Automobile Insurance Co. shows what a granted exception looks like. A claimant who sued an insurer for bad faith in settling a claim obtained an order compelling an adjuster's handwritten memoranda, which valued the claimant's losses and fixed a range of potential liability.45 The court agreed that the memoranda were opinion work product but held that discovery is still permitted when mental impressions are the pivotal issue in the current litigation and the need for the material is compelling.46 Both conditions were met. In a bad-faith case, the strategy, mental impressions, and opinions of the insurer's agents about handling the claim are directly at issue, and a plaintiff can establish compelling need for the claim file's evaluation of the claim's value unless the same information is available elsewhere.47
The court met the textual argument for absolute protection on its own ground. The advisory committee's 1970 note says the rule conforms to the holdings of the cases viewed in light of their facts, and those cases, beginning with Hickman, left room for a rare situation justifying production.48 The result also follows from the purpose of the tier. The protection keeps an opponent from borrowing or probing counsel's preparation, but a bad-faith claimant is trying to prove what the insurer actually thought of the claim, which is the conduct sued on. The IRS in Adlman asked for similar access and failed because of what each requester had to prove. The IRS wanted a window into corporate motive that the accountants' legal analysis did not supply, while the Holmgren claimant wanted the very evaluation its claim put in issue. Adlman cited Holmgren as contrary authority to Duplan without choosing between them.49
In practice the opinion tier is rarely reached, because requesters usually fail the ordinary showing first. One court rejected a requester's claimed need for a nonparty's confidential sales data because similar information was available from the actual defendants in the pending litigation.50 An essay question that gives a requester who has plainly shown need and hardship and asks whether a lawyer's memorandum must be produced calls for the rule text first: the court must protect counsel's thinking. The answer then states the disagreement without claiming the Supreme Court has settled it, and it turns to the facts that matter, namely whether the lawyer's or representative's own thinking is itself the issue in the case and whether the requester's need for it is compelling. A requester who wants only counsel's evaluation of witnesses, claims, or strategy in an ordinary case gets nowhere under any approach. A requester who claims that the opposing party's evaluation was made in bad faith has the strongest available argument for the exception.
Redaction and In Camera Review
Rule 26(b)(3)(B) tells the court to “protect against disclosure,” which governs how production is carried out. It does not require a court to withhold a whole document when only part of it records counsel's thinking.54 Redaction and in camera review are the two tools that let a court order the facts while shielding the thinking, provided the facts can be lifted out of the document without carrying the thinking along.
Real documents hold both. An interview memorandum records what a witness said and what the lawyer made of it. If a court could only produce a document whole or withhold it whole, it would have to deny facts the requester has proved it needs or expose the strategy the rule requires it to protect. Striking the evaluation and producing the rest gives each side what its interest justifies.
The coal-contract illustration from earlier in this chapter shows the technique. Irwin seeks the notes Mauldin's counsel took in interviewing a witness named Hersey. The notes were prepared in anticipation of trial and so carry qualified protection, but Irwin might show that Hersey's testimony is crucial and cannot otherwise be obtained, which is the showing Rule 26(b)(3)(A)(ii) demands.55 Even then, counsel's notes on the admissibility of the evidence, on Hersey's credibility, and on the other evidence in the case are opinion work product, so the court should allow Mauldin's counsel to produce a redacted version of the notes with all of that evaluative material stricken out.56 Irwin receives the record of what Hersey said, which it needed, and not the lawyer's view of how far to believe it, which the need showing never entitled it to see.
Redaction presupposes that someone has read the document closely enough to know where the facts end and the evaluation begins, and neither side is well placed to do that. The requester has never seen the document, and the party defending its file has every reason to call the whole thing strategy. The judge can read it without either handicap, and an in camera inspection, which is a private review by the judge, lets the court do so without showing the contents to the adversary. Courts use the device to classify as well as to redact. In Coburn, in camera review showed that an e-mail was work product but not opinion work product, so its protection depended on the ordinary showing of need.57 In Adlman, in camera review showed that the memorandum the IRS wanted reflected the legal analysis of Sequa's accountants and not the motives of Sequa's executives.58 The Georgia Supreme Court adds in St. Simons Waterfront that where otherwise discoverable materials contain mental impressions, the trial court may need to review them in camera so that those portions can be redacted before production. In Deloitte, the D.C. Circuit vacated a ruling that treated a memorandum as purely work product because the record gave the court no sufficient foundation for it, and remanded so the district court could decide whether a partial or redacted version could be disclosed.
The most detailed illustration of line-by-line sorting comes from Intel. After an exhaustive in camera review, the special master concluded that the interview materials contained both core and non-core work product, but that the core was not so intertwined with fact information that the whole had to be treated as core, because the facts could easily be separated.59 His itemized rulings sorted entries by whether they showed the attorney's own judgment, listing as examples any indication that something was important, such as a star or asterisk, and any conclusion that a follow-up interview was necessary.60 A star beside a remark says nothing about what the witness said and a good deal about what the lawyer thought mattered, which puts it on the evaluative side of the line even in a document that is otherwise ordinary fact.
Redaction works only when the facts can be separated from the thought. If a lawyer's mental impressions are so thoroughly intertwined with factual information that an entire memorandum should be treated as opinion work product, it cannot be produced at all, as the special master in Intel recognized.61 Selection can create the same problem even when the commentary can be stricken, because the lawyer's choice of what to include can itself reveal counsel's view. As the next subsection shows, selecting and ordering a few documents out of thousands can reveal important aspects of counsel's understanding of the case.62 A memorandum that evaluates prospective trial witnesses and also quotes verbatim a handful of their statements shows the difficulty. One court might let counsel excise the evaluations and produce the rest once need is shown. Another might refuse even that, because striking the commentary does not erase the choice: a reader who sees five statements quoted from fifty witnesses can infer which ones the lawyer regards as dangerous.
A fact pattern that gives a requester with a strong showing of need and describes counsel's notes of a witness interview is usually testing step four. The usual answer is a redacted production, and it changes only when the facts recorded were chosen in a way that itself shows counsel's theory of the case.
Selection and Compilation
Rule 26(b)(3)(B) requires a court that orders discovery of work product to protect counsel's mental impressions, conclusions, opinions, and legal theories.70 That command is easy to apply to a memorandum that says what the lawyer thinks. It is harder when the lawyer wrote no evaluation and only chose documents from a large pile and arranged them. Each document is ordinary evidence. What the doctrine protects is the lawyer's selection and distillation, so the question is whether a demand for the collection is a demand for the thinking behind it.71
James Julian, Inc. v. Raytheon Co. protected one such compilation. To prepare Julian's principals, officers, and employees for deposition by the Raytheon defendants, plaintiff's counsel assembled a binder of selected documents obtained through discovery, public records obtained through a Freedom of Information Act request, and documents Julian prepared during the project.72 Julian did not resist the defendants' obtaining the documents, many of which had already been turned over. It objected to producing the binder itself, because the selection and ordering reflected counsel's opinions, mental impressions, conclusions, or legal theories.73 The court agreed, at least in the first instance: “In selecting and ordering a few documents out of thousands counsel could not help but reveal important aspects of his understanding of the case.”74 It added that in a case with extensive document discovery, the process of selection and distillation is often more critical than pure legal research.75
The reasoning turns on what a subset communicates. Someone who holds all of the thousands of documents learns nothing about counsel's view from owning them, while someone who also holds the few that counsel chose can infer which ones counsel treats as central, which look damaging, and which a witness must be ready to explain. The binder therefore exposes what the doctrine shields, counsel's sorting of the relevant from the irrelevant, even though every page in it is available elsewhere. The defendants' fallback argument, that using the binder to prepare witnesses waived whatever protection it carried, raises a waiver question that this chapter takes up later.76
The protection has a boundary. SEC v. Collins & Aikman summarized Second Circuit law by calling compilation protection a “narrow exception” aimed at requests made with “the precise goal of learning what the opposing attorney's thinking or strategy may be.” Equity favors rejecting protection for a compilation of documents that are otherwise unavailable or beyond reasonable access. The decisions protecting compilations tended to involve collections assembled to help witnesses, which expose counsel's thinking about preparation and strategy, while the decisions rejecting protection involved documents not otherwise available or very large collections.77 Those limits point to the question that decides these disputes: would handing over the collection expose thinking that the requester cannot otherwise see?
Collins & Aikman began when a defendant, Stockman, requested the documents supporting particular factual allegations in the SEC's complaint. The SEC had produced an omnibus collection, yet its lawyers had already sorted documents into about 175 file folders keyed to specific factual contentions, and the SEC refused to hand over the folders on the ground that the compilation would reveal its attorneys' mental impressions. The court held that selecting documents to support the factual allegations of a complaint does not produce core work product, because core protection reaches a compilation only if it is organized by legal theory or strategy.78 Rule 11 requires evidentiary support for the factual contentions in a pleading, so documents supporting those contentions show no more than the complaint already revealed. The court also rejected the SEC's theory that every document an attorney reviews is core work product, since that theory leaves nothing to surround the core. Even if some protection applied, the defendant had shown substantial need and undue hardship, because a page-by-page manual review of ten million pages would be strikingly expensive, and making a party search ten million pages for documents its adversary had already identified would be patently inequitable.79
The two decisions reach opposite results on facts that look alike, since in each a lawyer had chosen documents from a larger set. The difference lies in what the choice revealed and what the requester could do without it. The Julian binder was built to prepare the party's own witnesses, so its contents showed which documents counsel thought a witness would have to face, and nothing in the lawsuit would ever display that. Julian did not object to the defendants' receiving the documents, so the selection was the only thing in dispute. The SEC's folders tracked allegations the SEC had already published in its complaint, so the selection displayed no strategy beyond the pleading, and the defendant could not practicably rebuild it from the volume produced. A demand for a lawyer's choice is therefore a demand for mental impressions when the choice reveals thinking that would otherwise stay hidden, and it seeks little that the rule protects when the choice repeats what the lawyer has already disclosed.
The same inquiry explains a second limit, which concerns timing. In In re San Juan Dupont Plaza Hotel, the First Circuit reviewed a case-management protocol that required a party to identify in advance the documents it planned to use as exhibits at depositions.80 The court accepted that selecting deposition exhibits is often more crucial than legal research and that a glimpse of the selection gives insight into opposing counsel's understanding of the case.81 It still held that the lists were not opinion work product, because once the deposition begins the questioner's document selection, and the stratagems it shows, become obvious to everyone present, so pre-identification only moves up the schedule of a disclosure that would occur anyway.82 The court explained that the discovery rules do not compel disclosure of matters that would otherwise stay private and instead accelerate disclosure of matters that would probably come out in due course.83 It treated the lists as ordinary work product, which safeguarded whatever legitimate privacy concern remained, and it found that the warranted protection was clearly outweighed by the demands of managing the case.84 A lawyer's private notes assessing the likely testimony of prospective witnesses stand differently, because the lawyer can expect that no opponent will ever see them, while no lawyer can expect the opposition to remain ignorant of the exhibits chosen for use at a deposition.85
Step two is where exam facts set the trap. A stem that says only that counsel chose or organized documents has not yet told you whether the choice reveals anything, so the analysis depends on how the collection was built and who will see it. A collection organized by legal theory or strategy, or assembled to prepare witnesses, points toward protection, while one that follows the facts alleged in the complaint, or that must be shown at the deposition anyway, points away.90 Classification then sets the price of the requester's showing: Collins & Aikman tied core status to organization by legal theory or strategy, and the First Circuit treated deposition exhibit lists as ordinary work product that case-management needs can outweigh.91
Two later topics test how durable these protections are. Rule 26(b)(3) is itself expressly subject to Rule 26(b)(4), and the next unit turns to the testifying experts that provision governs.92 Waiver, taken up afterward, asks whether disclosure to an outsider ends work-product protection, and one dispute previews the question. A company that gave its auditor reports its lawyers had prepared conceded that the attorney-client privilege was waived and claimed work product alone, while acknowledging that the auditor's own audit work and discussions with company personnel were discoverable.93 Its declarants said the reports were provided on the understanding that they were prepared by counsel and privileged, that the auditor would keep them confidential, and that there would be no further disclosure.94 One court held that the disclosure did not waive work-product protection, reasoning that the “adversary” the doctrine contemplates is not necessarily a litigation adversary and that a corporation's auditor is not such an adversary.95 The Federal Circuit has explained the premise: voluntary disclosure waives the attorney-client privilege because it is inconsistent with a confidential attorney-client relationship, but it does not necessarily waive work-product protection, because it does not necessarily undercut the adversary process.96