Supreme Court of the United States

Transcription

1 No IN THE Supreme Court of the United States TEVA PHARMACEUTICALS USA, INC., ET AL., Petitioners, v. SANDOZ, INC., ET AL., Respondents. On Writ of Certiorari to the United States Court of Appeals for the Federal Circuit REPLY BRIEF FOR PETITIONERS JAY P. LEFKOWITZ, P.C. JOHN C. O QUINN JASON M. WILCOX KIRKLAND & ELLIS LLP 655 Fifteenth Street, N.W. Washington, DC WILLIAM M. JAY Counsel of Record WILLIAM G. JAMES, II GOODWIN PROCTER LLP 901 New York Ave., N.W. Washington, DC wjay@goodwinprocter.com (202) (Additional counsel listed on inside cover) DAVID M. HASHMALL ELIZABETH J. HOLLAND STEVEN J. BERNSTEIN GOODWIN PROCTER LLP 620 Eighth Ave. New York, NY September 10, 2014

2 ALAN M. DERSHOWITZ 1575 Massachusetts Ave. Cambridge, MA DARYL L. WIESEN HENRY C. DINGER JOHN C. ENGLANDER NICHOLAS K. MITROKOSTAS TODD MARABELLA JAIME A. SANTOS GOODWIN PROCTER LLP Exchange Place 53 State Street Boston, MA 02109

3 RULE 29.6 STATEMENT The disclosure statement included in the opening brief remains accurate. i

4 TABLE OF CONTENTS Page RULE 29.6 STATEMENT... i TABLE OF AUTHORITIES... iv I. Defining The Scientific Techniques And Terminology In A Patent May Require District Courts To Find Classic Adjudicative Facts A. Except For The Federal Circuit, Appellate Courts Defer To District Courts Findings On The Fact Part Of A Mixed Question Of Law And Fact B. There Is No Claim-Construction Exception To Rule Markman Does Not Turn Claim- Construction Facts Into Law No Pre-Markman Decision Even Considers The Standard Of Review The Facts Underlying Claim Construction Are Adjudicative Facts That Require Evidence There Is No Policy Justification For Ignoring Rule ii

5 II. The Federal Circuit s Decision Rests On Its Failure To Defer A. Factfinding Appropriately Considers Both Technology And Terminology B. The Federal Circuit Rejected The Presumed Meaning Derived From The Reference To SEC C. The Federal Circuit Rejected The Correct Scientific Understanding Of Figure D. The Federal Circuit Rejected The Skilled Artisan s Understanding Of The Mistake In Prosecution History CONCLUSION iii

6 CASES: TABLE OF AUTHORITIES Page(S) Blonder-Tongue Labs., Inc. v. Univ. of Ill. Found., 402 U.S. 313 (1971) Brown v. Piper, 91 U.S. 37 (1875)... 9 Coupe v. Royer, 155 U.S. 565 (1895)... 7 Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 (1993) Dennison Mfg. Co v. Panduit Corp., 475 U.S. 809 (1986) (per curiam)... 2 Elbex Video, Ltd. v. Sensormatic Elecs. Corp., 508 F.3d 1366 (Fed. Cir. 2007) EON Corp. IP Holdings, LLC v. FLO TV Inc., 2014 WL (D. Del. Mar. 4, 2014) Gen. Elec. Co. v. Joiner, 522 U.S. 136 (1997)... 8 Graham v. John Deere Co., 383 U.S. 1 (1966)... 2, 15 Great N. Ry. Co. v. Merchants Elevator Co., 259 U.S. 285 (1922)... 9, 15 Harries v. Air King Prods. Co., 183 F.2d 158 (2d Cir. 1950)... 4, 6, 13 Hoechst Celanese Corp. v. BP Chems. Ltd., 78 F.3d 1575 (Fed. Cir. 1996) In re Bear Creek Techs., Inc. ( 722) Patent Litig., 858 F. Supp. 2d 1375 (J.P.M.L. 2012) iv

7 In re Mahurkar Double Lumen Hemodialysis Catheter Patent Litig., 831 F. Supp (N.D. Ill. 1993), aff d on other grounds, 71 F.3d 1573 (Fed. Cir. 1995)... 1 Intertrust Techs. Corp. v. Microsoft Corp., 275 F. Supp. 2d 1031 (N.D. Cal. 2003)... 6 Knowles v. Mirzayance, 556 U.S. 111 (2009)... 3 Lighting Ballast Control LLC v. Philips Elecs. N. Am. Corp., 744 F.3d 1272 (Fed. Cir. 2014) (en banc)... 3, 11, 12 Loom Co. v. Higgins, 105 U.S. 580 (1881)... 4, 7 Maine v. Taylor, 477 U.S. 131 (1986)... 9 Markman v. Westview Instruments, Inc., 517 U.S. 370 (1996)... passim Miller v. Fenton, 474 U.S. 104 (1985)... 3 Nautilus, Inc. v. Biosig Instruments, Inc., 134 S. Ct (2014)... 4, 10 Patterson v. McLean Credit Union, 491 U.S. 164 (1989) Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc) Pullman-Standard v. Swint, 456 U.S. 273 (1982)... 2 Thompson v. Keohane, 516 U.S. 99 (1995)... 3 Tilghman v. Proctor, 102 U.S. 707 (1880)... 7 v

8 TSC Indus., Inc. v. Northway, Inc., 426 U.S. 438 (1976)... 8, 10 United Carbon Co. v. Binney & Smith Co., 317 U.S. 228 (1942)... 7 United States v. Kubrick, 444 U.S. 111 (1979)... 8 United States v. Navarro, 90 F.3d 1245 (7th Cir. 1996)... 3 United States v. Pelletier, 469 F.3d 194 (1st Cir. 2006)... 3 Viskase Corp. v. Am. Nat l Can Co., 261 F.3d 1316 (Fed. Cir. 2001) STATUTES: 28 U.S.C U.S.C. 299(b) RULES: Fed. R. Civ. P. 52(a)... passim Fed. R. Civ. P. 52 advisory committee s note Fed. R. Evid. 201(b)... 9 Fed. R. Evid. 201 advisory committee s note... 8 OTHER AUTHORITIES: Peter S. Menell, et al., Patent Case Management Guide (Fed. Judicial Ctr. 2009) S. Rep. No (2007) vi

9 REPLY BRIEF FOR PETITIONERS A district court s task is to read a patent as a skilled artisan would. What seems clear to a judge may read otherwise to a skilled designer. In re Mahurkar Double Lumen Hemodialysis Catheter Patent Litig., 831 F. Supp. 1354, 1359 (N.D. Ill. 1993) (Easterbrook, J.), aff d on other grounds, 71 F.3d 1573 (Fed. Cir. 1995). And discerning the skilled designer s perspective will often require the generalist judge to take evidence and to resolve factual disputes between the parties. In any other context, in any other court of appeals, the district court s resolution of a scientific disagreement or dispute over specialized terminology would be presumed correct. Nothing in patent law justifies the Federal Circuit s refusal to follow that same principle. De novo review has produced too many cases like this one, in which the Federal Circuit second-guesses the district court on fundamentally factual questions. This Court should end that experiment. I. Defining The Scientific Techniques And Terminology In A Patent May Require District Courts To Find Classic Adjudicative Facts. A. Except For The Federal Circuit, Appellate Courts Defer To District Courts Findings On The Fact Part Of A Mixed Question Of Law And Fact. The Federal Circuit, respondents, and their amici all argue that because the end product of claim con-

10 2 struction is reviewed de novo, supporting facts must be reviewed de novo too. But facts do not dissolve into law that way. To the contrary, Rule 52(a)(6) applies equally to subsidiary facts and ultimate facts. Pullman-Standard v. Swint, 456 U.S. 273, 287 (1982). In every other area of law, including the interpretation of other written instruments, the facts supporting a legal conclusion are reviewed deferentially. Patent law is no different. At least once before, this Court sharply cautioned the Federal Circuit to follow this rule and defer to subsidiary factual findings. Whether a patent is obvious in light of the prior art is a legal question. But a court can answer that legal question only against the background set by factual inquiries that overlap with those in claim construction: what is the level of ordinary skill in the art and what would a skilled artisan know of the prior art? Graham v. John Deere Co., 383 U.S. 1, 17 (1966) (emphasis added). When the Federal Circuit reversed a holding of obviousness without explaining why the underlying factual findings were clearly erroneous, this Court readily corrected that error: even though the ultimate question of obviousness is not a question of fact subject to Rule 52(a), the subsidiary determinations of the District Court, at the least, ought to be subject to the Rule. Dennison Mfg. Co v. Panduit Corp., 475 U.S. 809, 811 (1986) (per curiam) (emphasis added). Respondents offer no response to this point; they do not even cite Graham or Dennison, which directly refute their contention that patent law permits factual findings only when the ultimate question is factual. See Resp. Br

11 3 Even in the standard-of-review case on which respondents chiefly rely, Miller v. Fenton, 474 U.S. 104 (1985), this Court carefully reiterated that facts receive deferential review. In Miller, the Court considered whether the habeas statute s presumption of correctness for factual issues applies to a state court s conclusion that a confession was voluntary. While the Court concluded that the ultimate issue of voluntariness is not a factual finding, it carefully noted that within the voluntariness question lie subsidiary questions of historical fact, which may well involve the resolution of conflicting testimony. Id. at 117. Those questions do not lose their factual character just because they are subsidiary, and a trial court s answers are reviewed deferentially, not de novo. See id.; see, e.g., United States v. Pelletier, 469 F.3d 194, 201 (1st Cir. 2006) (deferring to findings of historical fact, which in turn compelled the legal conclusion that statements were voluntary); United States v. Navarro, 90 F.3d 1245, 1258 (7th Cir. 1996) (same). That is simply standard practice. In a host of contexts, the Court has held that while an ultimate issue may be a mixed question subject to de novo review, the appellate court must accept the facts found by the district court, absent clear error. See, e.g., Knowles v. Mirzayance, 556 U.S. 111, 126 (2009) (ineffective assistance of counsel); Thompson v. Keohane, 516 U.S. 99, 112 (1995) (custody); Pet. Br. 20. The Federal Circuit bemoaned the supposed difficulty of [d]isentangling arguably factual aspects of claim construction. Lighting Ballast Control LLC v. Philips Elecs. N. Am. Corp., 744 F.3d 1272, 1284 (Fed. Cir. 2014) (en banc). But appellate courts regu-

12 4 larly separate law from facts. They even did so in patent cases before the Federal Circuit was created. Harries v. Air King Prods. Co., 183 F.2d 158, 164 (2d Cir. 1950) (L. Hand, J.); Pet. Br. 45. There is no reason why the Federal Circuit cannot do the same. B. There Is No Claim-Construction Exception To Rule 52. Respondents various attempts to remove patentrelated factfinding from the scope of Rule 52 are unavailing. First and foremost, patents are scientific documents written for scientists persons skilled in the relevant art. Courts accordingly must interpret a patent through the eyes of a skilled artisan. Nautilus, Inc. v. Biosig Instruments, Inc., 134 S. Ct. 2120, 2128 (2014). Respondents key analogy between patents and statutes founders on that basic principle. Statutes are written for the same general audience, including judges. Patents and many other written instruments are not; each is written for an interpretive community that shares a mass of previous knowledge and nomenclature that is not written out in the patent, but is just as much a part of the patent as if it were. Loom Co. v. Higgins, 105 U.S. 580, (1881). That interpretive community does not include the judge. Before construing such a document, the judge must learn what the document s intended readership already knows. This Court recognized in Markman v. Westview Instruments, Inc., 517 U.S. 370 (1996), that that part of interpretation is factual in nature. And the judge acquires those facts not through free-

13 5 wheeling legislative factfinding, but through the evidentiary process. 1. Markman Does Not Turn Claim- Construction Facts Into Law. Markman did not hold that there were no subsidiary factual issues in claim construction, as respondents assert. Resp. Br. 21 (citing Markman, 517 U.S. at ). If all aspects of claim construction were purely legal, there would have been no colorable claim to a jury determination. What made the judge/jury question complex was that claim construction does include facts the evidentiary underpinnings to which the Court referred, 517 U.S. at 390. That is why the Court called it a mongrel practice. Id. at 378. The Court ultimately concluded that the task contained more elements better suited to a court than a jury; either a judge or a jury could competently evaluate testimony, and the judge was far better positioned to evaluate the testimony in relation to the overall structure of the patent. Id. at 390. But the Court did not decree that the mongrel practice would from that day on be purebred. To the contrary, the Court explicitly recognized that credibility judgments have to be made about the experts who testify in patent cases. 517 U.S. at 389. Sometimes factfinders might even choose between experts based on credibility. Id. The Court did not anticipate many cases like that, another reason why claim construction goes to the judge and not the jury. Id. But the Court did not signal that there would henceforth be no cases like that that the deeply rooted tradition of trial-court factfinding

14 6 would be replaced with de novo review by the Federal Circuit even of credibility determinations. Rather, Markman left them to be reviewed on appeal as district-court credibility judgments are in every other context. Indeed, given respondents repeated references (at 2, 3, 43, 44, 45) to the notion that experts are paid and therefore less than credible, it is surprising that respondents do not acknowledge district courts expertise at making exactly that sort of credibility determination, subject to clear-error review No Pre-Markman Decision Even Considers The Standard Of Review. As for the pre-markman decisions respondents cite (at 31-32), nearly all of which predate Rule 52, none even considers the standard of review, much less endorses de novo review of facts. U.S. Br. 20 n.2. That is why Judge Hand was able to write in 1950 after all of respondents cases that how the art understood [a claim] term was plainly a question of fact; and unless the finding was clearly erroneous, we are to take this definition as controlling. Harries, 183 F.2d at 164. Indeed, in none of these cases did the Court indicate its disagreement with the trial court s assessment of the evidence supporting its claimconstruction decision, let alone reverse any subsidiary determinations based on the evidence. In one example, the Court reversed the trial court s con- 1 See, e.g., Intertrust Techs. Corp. v. Microsoft Corp., 275 F. Supp. 2d 1031, 1044 (N.D. Cal. 2003) (finding not credible an opinion created for this litigation ).

15 7 struction of a disputed claim not because it disagreed with the court s assessment of the evidence, but because of a legal error: the court s claim construction was not in conformity with the self-imposed limitations contained in the claims. Coupe v. Royer, 155 U.S. 565, (1895). In others, the Court concluded that the meaning of the disputed claim terms was readily determinable solely by reference to the claim language and specification. Loom Co., 105 U.S. at 590 ( [T]he meaning of this claim is quite obvious. If any explanation of it is needed, it can be readily derived from the body of the specification. ); Tilghman v. Proctor, 102 U.S. 707, (1880) ( The true construction of this claim is to be sought by comparing it... with the context of the specification... ). Looking at the record is not de novo review, and relying on testimony that was in no wise impeached or contradicted, United Carbon Co. v. Binney & Smith Co., 317 U.S. 228, 233 (1942), is hardly the same as second-guessing the trial court when the evidence is contradicted. 3. The Facts Underlying Claim Construction Are Adjudicative Facts That Require Evidence. After their initial stab at suggesting that Markman resolved this issue, respondents retreat to the notion that the type of facts necessary to construe a patent are invariably legislative facts which require no testimony at all and which appellate courts can weigh, re-weigh, ignore, or find for themselves. The Federal Circuit has never itself advanced that novel justification. And it is fundamentally flawed. Scientific facts that inform the interpretation of a particular document, and thus bear on a dispute between

16 8 two parties, are classic adjudicative facts, which must be established through the introduction of evidence, Fed. R. Evid. 201 advisory committee s note (1972), and reviewed under Rule 52(a). Legislative (or non-evidence ) facts, by contrast, are used in the formulation of a legal principle or ruling by a judge or court or in the enactment of a legislative body. Id. The construction of a single patent claim is not a legislative endeavor, and the scientific facts that bear on claim construction are adjudicative facts reviewed under Rule 52(a)(6). First, respondents contend (at 26-27) that because the skilled artisan is hypothetical, figuring out what she would know and understand must be a legal question. Respondents cite no authority for that proposition, and it is a red herring. The law often assumes away gaps in individuals knowledge by hypothesizing a fully-informed individual. But what that hypothetical individual would have known or concluded at a particular time is a question of fact. See TSC Indus., Inc. v. Northway, Inc., 426 U.S. 438, 450 (1976) (what inferences a reasonable shareholder would draw from a given set of facts is for the trier of fact ); see also United States v. Kubrick, 444 U.S. 111, (1979) (treating medical standard of care as question of historical fact). Second, and more fundamentally, the legislative fact concept does not encompass the type of historical scientific evidence, drawn from the time of the patent application, that sheds light on the meaning of the particular patent-in-suit. [J]udges are not scientists, Gen. Elec. Co. v. Joiner, 522 U.S. 136, 148 (1997) (Breyer, J., concurring), and they are not

17 9 equipped to take in and resolve a scientific dispute on the basis of their own experience. When a text has been written by and for specialists, establishing the usage among those specialists presents a predicate question of fact and not legislative fact. Great N. Ry. Co. v. Merchants Elevator Co., 259 U.S. 285, 293 (1922); Pet. Br Respondents try to distinguish those holdings away by pointing out that, in the tariff context, such findings were made by an administrative agency with nationwide jurisdiction. That argument confuses cause and effect: because this Court held that the specialized meaning was a question of fact, that question was one for the factfinder (the agency). And the Court noted that the specialized meaning of a term is likewise a factual question under basic contract law, see id. at 292 & n.1 (citing cases), which of course owes nothing to the idea of a single nationwide factfinder. That rule does not change when the predicate facts are scientific facts. See, e.g., Maine v. Taylor, 477 U.S. 131, 146 (1986) (even in constitutional case, mandating deferential review of question whether scientifically accepted techniques existed). Respondents suggest (at 28) that Brown v. Piper, 91 U.S. 37 (1875), lets courts garner scientific background knowledge from wherever they wish, but they overlook what this Court said: before taking notice of a fact not in evidence, [c]are must be taken that the requisite notoriety exists, and [e]very reasonable doubt upon the subject should be resolved promptly in the negative. Id. at 43; see Fed. R. Evid. 201(b); Pet. Br. 24. If federal appellate courts truly could consider whatever contested scientific

18 10 facts they wished, unconstrained by evidence, that would thoroughly transform litigation in a host of areas from products liability to environmental law. But no other federal appellate court has asserted such power, and with good reason. The historical nature of claim construction provides even clearer confirmation that the necessary scientific grounding consists of adjudicative facts that may be unique to each patent dispute. Patents are interpreted by referring not to present-day scientific understanding, but to what a skilled artisan would have known when the patent was filed. Nautilus, 134 S. Ct. at That is a fact-sensitive matter, Pet. Br. 31; cf. TSC Indus., 426 U.S. at 450, that requires inquiry into the state of scientific knowledge at a specific time, to resolve a single debate over a single patent. That is adjudicative factfinding, and it must follow the rules of evidence except for facts that are so firmly established as to have attained the status of scientific law, which properly are subject to judicial notice. Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 592 n.11 (1993). The contested scientific facts in this case plainly do not meet that standard. 4. There Is No Policy Justification For Ignoring Rule 52. Respondents final submission, echoed by their amici, is that review should be de novo, essentially notwithstanding Rule 52(a). Once the Court con- 2 E.g., EON Corp. IP Holdings, LLC v. FLO TV Inc., 2014 WL , at *6 n.9 (D. Del. Mar. 4, 2014) (requiring expert evidence to resolve the dispute about the capabilities of a computer in 1991 ).

19 11 cludes that fact[s], under the Rule, include scientific facts necessary to understand the technology and terminology in a patent, these policy arguments are beside the point. 3 But they are also incorrect on their own terms, because they ignore the current rule s deleterious effect on accuracy, predictability, cost, and settlement, and the sole value they claim this rule serves uniformity is overstated. First, the current no-deference rule heightens uncertainty and discourages settlement by holding out the possibility of an appellate do-over. ABA Br ; AIPLA Br ; Menell Br. 17. Respondents (at 42) invoke the unresponsive point that earlier claim construction promotes settlement. But claim construction can be done early even if it involves factual submissions. Respondents are looking at the wrong variable: what impedes settlement after claim construction is that de novo review reduces the district court s claim construction to just an opening bid. See Lighting Ballast, 744 F.3d at 1313 (O Malley, J., dis- 3 The closest thing to a legal argument in these passages of respondents brief is the assertion (at 41) that Congress has endorsed the current no-deference regime through sheer inaction. But respondents citations show only that Congress believed that the Federal Circuit reverses too many claim constructions hardly a ringing endorsement of de novo review. See S. Rep. No , at (2007). Furthermore, this Court has regularly noted the danger of placing undue reliance on the notion that congressional failure to act equals congressional ratification. Patterson v. McLean Credit Union, 491 U.S. 164, 175 n.1 (1989). Just because Congress has not acted to reverse a nonstatutory Federal Circuit precedent from 1995, one that this Court has never endorsed, does not mean that Congress agrees with it quite unlike Congress s acquiescence in the venerable doctrine of equivalents laid down by this Court, see Resp. Br. 41.

20 12 senting). What would promote settlement would be to make district-court constructions meaningful. Second, and relatedly, appellate second-guessing of the facts leads to inaccurate, unpredictable decisionmaking. Pet. Br , 24. Judge O Malley, a former district judge now on the Federal Circuit, speaks from experience when she says that [the Federal Circuit] is given free rein to interpret claim terms, but lacks the resources to do it right. Lighting Ballast, 744 F.3d at 1313 (dissenting opinion). The Federal Circuit may have technical advisers on staff and a few judges with training in a scientific field, but it does not and cannot call on true skilled artisans from every conceivable art. A college degree in the art usually will not suffice, and a doctorate may not either. Cf. J.A. 914a-15a (ordinary skill in the art in this case required a Ph.D. and extensive experience in polymer synthesis). And even if the Federal Circuit could obtain direct access to scientists skilled in every conceivable art, it cannot subject their input to the extended, meaningful adversarial testing that the district-court setting allows. Treating district courts findings with the respect Rule 52(a) requires will encourage district courts to conduct just such testing, and to explain their conclusions in a way that the Federal Circuit can review (deferentially) and fellow district courts can follow (because of their persuasive value). Menell Br ; Pet. Br. 48. Third, amici again miss the mark when they suggest that expert evidence in claim construction is the problem. Google Br. 11, 19. This Court has entertained such evidence for over a century, the Federal Circuit accepts it, and trial courts welcome it. The

21 13 question, rather, is whether findings of scientific fact in patent cases should be treated on appeal like all other findings, or ignored at the Federal Circuit s pleasure. Fourth, respondents fundamentally misconceive the Federal Circuit s role in promoting uniformity. Having a single patent appeals court put a stop to forum-shopping based on different legal rules in different circuits, such as different standards for holding a patent obvious. Pet. Br Respondents identify no comparable history suggesting the Federal Circuit was meant to drum facts out of claim construction in the name of intra-patent uniformity. 4 The creation of a single patent appeals court did not redefine law and fact, or make an exception to Rule 52(a). Congress consciously did not create a specialized trial court; it left patent factfinding to be done by district judges as they have always done. If facts underlying claim construction were governed by Rule 52(a) before the Federal Circuit was created, see Harries, 183 F.2d at 164, they are still governed by Rule 52(a). Federal courts certainly can work toward uniformity while respecting Rule 52(a), using a number of different tools. Issue preclusion is one. Pet. Br. 47. Another is the multidistrict-litigation statute, 28 U.S.C. 1407, which permits coordination of just such pretrial matters as claim construction. 5 Per- 4 Forum-shopping based on how circuits view the law is considerably easier than forum-shopping based on how individual judges might find facts relevant to one patent. 5 Intel notes (at 15) that a recent amendment to the Patent Act restricts consolidat[ion] for trial. 35 U.S.C. 299(b). But nothing restricts the multidistrict-litigation panel from order-

22 14 suasion is a third: district courts can follow each other s reasoned opinions opinions which they will have greater incentive to write if their factual findings are respected on appeal. See, e.g., Peter S. Menell, et al., Patent Case Management Guide (Fed. Judicial Ctr. 2009) (citing cases). And the backstop is clear-error review. In the limited circumstances where claim construction entails factual questions and these tools do not lead to consistent answers, patent law, like every other area of law, simply tolerates a degree of variability. Infringement is a factual question. So is the doctrine of equivalents. So are various patentinvalidity doctrines. Thus, because proof may vary from one action to another, identical devices may be found to infringe the same patent in one case but not in the next. Or the same patent may be found not invalid today, invalid tomorrow. Pet. Br That is the consequence of according each new defendant one fair chance to present [its] evidence and arguments on the [factual] claim. Blonder-Tongue Labs., Inc. v. Univ. of Ill. Found., 402 U.S. 313, 329 (1971). Facts cannot be squeezed completely out of claim construction in the name of predictability. Ultimately, absolute Procrustean uniformity is not desirable, especially not at the expense of accuracy. And as the Federal Circuit s decision here illustrates, appellate courts unconstrained by the record are simply more likely to make scientific mistakes. ing pretrial coordination. In re Bear Creek Techs., Inc. ( 722) Patent Litig., 858 F. Supp. 2d 1375, (J.P.M.L. 2012).

23 II. 15 The Federal Circuit s Decision Rests On Its Failure To Defer. This Court could answer the question presented simply by disapproving the Federal Circuit s rule that there are no factual issues in claim construction, and remanding to apply the ordinary fact/law distinction set out in Rule 52(a). See Pet. i. But if this Court chooses to use this case to illustrate how that fact/law distinction applies in the patent context, the outcome is clear. The Federal Circuit considered three aspects of the patent: the use of SEC; Figure 1; and the prosecution history. Each is the subject of an express factual finding by the district court. And in each respect the Federal Circuit s decision cannot be reconciled with that finding. This Court therefore should reverse that decision, either outright or with instructions to construe the patent in accordance with the district court s findings, which respondents have never challenged. A. Factfinding Appropriately Considers Both Technology And Terminology District courts are engaged in factfinding when they seek to understand the science or technology discussed in the patent, cf. Graham, 383 U.S. at 17, or determine the accepted meaning of particular scientific or technical terminology at the relevant time. Courts turn to the legal enterprise of claim construction when they apply that factfinding to the claim language and determine the extent to which the historical facts inform the meaning of the claims. Cf. Great Northern, 259 U.S. at 292 ( Where... the peculiar meaning of words... is proved by evidence, the

24 16 function of construction is necessarily preceded by the determination of the matter of fact. ). Some cases will require no factfinding, because the disputed term either is an ordinary word with no technical meaning, or is defined in the patent itself. In other cases, the court will find as a fact the commonly accepted meaning of a term among skilled artisans at the relevant time, but then decline to apply that commonly accepted meaning as a matter of law. For instance, courts presume that a patent claim should be construed, if possible, not to exclude the embodiments of the invention recited in the specification. 6 Applying that principle to reject the commonly accepted meaning of a term involves a legal conclusion. See U.S. Br. 22. Respondents contend (at 43-44) that fact evidence can pertain only to what is true in the world generally, and that relating scientific facts to the patent is exclusively the job of the court. But some testimony that respondents (and the government to some degree, see U.S. Br. 22, 31) identify as document interpretation is really just an expert s attempt to select the scientific facts that are relevant to the task at hand. A scientific term may have two accepted meanings in the field e.g., one associated with hardware and one with software. An expert would be addressing matters of fact if she testifies about both meanings. If the expert looks at the patent, sees that it involves software, not hardware, and testifies about the definition relevant to software, respondents would protest that her testimony is no 6 See, e.g., Hoechst Celanese Corp. v. BP Chems. Ltd., 78 F.3d 1575, (Fed. Cir. 1996) (applying this rule to reject the commonly accepted meaning).

25 17 longer separate and apart from the patent. But her testimony does not become any less factual just because she addresses the definition that is relevant in context. The test, therefore, is not whether the expert refers to the patent or instead answers entirely hypothetical questions. Nor does it depend on whether the testimony was live, Pet. Br , particularly where (as here) respondents forwent the opportunity to submit live testimony at trial after the district court denied their summary-judgment motions. Rather, what matters is whether the expert opines on, and the district court makes findings on, judgments about the science or technology embodied in the patent, or the scientific or technical terminology used in the patent. Those points are factual, not legal. b. The government s attempt to draw a line between two key issues the import of Figure 1 to a skilled artisan, and the import of the reference to SEC to a skilled artisan is fundamentally unpersuasive. Even amici recognize that a rule treating one of those issues as factual and the other as legal cannot be correct. Intel Br The district court properly made factual findings on both subjects, which speak directly to the understanding of scientific technique and terminology necessary to understand the patent correctly. Dr. Grant did not impermissibly construe the claim in explaining what Figure 1 of the patent would and would not convey to a skilled artisan, because his testimony related to the common scientific understanding of that type of graph. Nor did he impermissibly construe the

26 18 claim in explaining what a reference to SEC would convey to a skilled artisan, because his testimony related to the common scientific understanding of how average molecular weight would most likely be measured using SEC. 7 B. The Federal Circuit Rejected The Presumed Meaning Derived From The Reference To SEC. The district court found that, when certain predicate facts are present, the presumed meaning of [average molecular weight] to a skilled artisan would be peak average. Pet. App. 43a (emphasis added); see id. at 44a. Those predicate facts are (1) the use of SEC to find average molecular weight, and (2) the absence of any further calculations. See id. at 43a, 125a; J.A. 138a. No calculations are needed to derive peak average, and none appear in the patent. That factual finding speaks to both scientific technique and terminology: a scientist s knowledge of technique affects the words she would use to describe the technique s output. And that finding, standing alone, is enough to compel reversal: the panel gave no weight to the presumed meaning found by the district court. Instead, the panel treated all three possible measurements of average molecular weight as equally likely, because weight average and number average can be obtained from the chromatogram after some calculations. Pet. App. 10a. That understanding is not compatible with the district court s factual finding. 7 The section headings in Dr. Grant s declaration (Resp. Br. 49) are irrelevant. What matters is whether the district court s findings are factual in substance, not in labeling.

27 19 C. The Federal Circuit Rejected The Correct Scientific Understanding Of Figure 1. As the government correctly states (at 31-32), the Federal Circuit impermissibly engaged in its own factfinding with respect to Figure 1. The Federal Circuit concluded that Figure 1 refuted Teva s peakaverage interpretation, because the peaks of the curves in Figure 1 do not correspond to 7.7 kda. Pet. App. 10a. Respondents insist that that is a legal conclusion, but they cannot escape that the appellate panel was stating a factual premise that if 7.7 kda was the peak average, then the peaks of Figure 1 should correspond to 7.7 kda and that the district court had clearly found that that premise is wrong. That finding was not legal interpretation: it relied on scientific facts, and was grounded in the patent only insofar as the patent discloses what Figure 1 portrays. Pet. App. 141a, 144a. Assessing whether Figure 1 is consistent with Teva s construction requires an understanding of what a chromatogram is, what Figure 1 is, and why they are different. That understanding is factual, even if Figure 1 s legal relevance turns on it. Respondents offer a long digression (at 62-64) on whether Dr. Grant s estimate (J.A. 310a) of weight average was closer to 7.7 kda than the peak of the Figure 1 curve was. 8 Respondents sought unsuccessfully to impeach Dr. Grant with that argument, see 08-cv-7611 ECF No. 141, at 9-10, and they have not 8 Notably, respondents expert calculated the weight average in Figure 1 as 12.9 kda, J.A. 381a, which, if true, would defeat respondents argument about relative proximity to 7.7 kda.

28 20 challenged the district court s finding as clearly erroneous. The sample s average molecular weight (as used in the patent) was 7.7 kda. Whether the weight average was close to 7.7 is irrelevant to whether the peak of the chromatogram, and hence the peak average, was at 7.7. The panel thought that because the peak of Figure 1 does not fall at 7.7 kda, Figure 1 essentially rules out Teva s interpretation (peak average). See Pet. App. 10a. That conclusion flies directly in the face of the District Court s factfinding. D. The Federal Circuit Rejected The Skilled Artisan s Understanding Of The Mistake In Prosecution History. Respondents and the court of appeals begin rather than end with the prosecution history. Yet the Federal Circuit has repeatedly cautioned that prosecution history lacks the clarity of the specification and thus is less useful for claim construction purposes. Phillips v. AWH Corp., 415 F.3d 1303, 1317 (Fed. Cir. 2005) (en banc). Mistakes sometimes happen during prosecution, often enough that the Federal Circuit has a number of precedents on how to deal with them. 9 And those precedents start with whether it [would] have been apparent to a person experienced in this technology that an error was made, or 9 That is why, even if a statement to the PTO can be read as voluntarily surrendering part of the scope of a claim, the Federal Circuit will not read it that way unless the disavowal was both clear and unmistakable, not ambiguous. Elbex Video, Ltd. v. Sensormatic Elecs. Corp., 508 F.3d 1366, 1371 (Fed. Cir. 2007) (citation omitted).

29 21 at least that the information should not be relied upon. Viskase Corp. v. Am. Nat l Can Co., 261 F.3d 1316, 1322 (Fed. Cir. 2001). Here the district court found that it would. See Pet. App. 52a. Indeed, Sandoz s expert conceded that it would. J.A. 157a. The Federal Circuit never grappled with that finding The relevant question is whether a skilled artisan would regard two prosecution-history statements as irreconcilable. To accept respondents argument that [t]he skilled artisan would have believed that the applicant s disclaiming statements were not a mere mistake, Sandoz C.A. Reply Br. 11 (citation omitted) the Federal Circuit had to substitute its own view for the district court s. In this Court, respondents act as though the patent became invalid the moment that Teva used the term weight average, but even on respondents theory that cannot be right: a statement that average molecular weight has a definite meaning cannot make a claim indefinite. Rather, respondents must argue that the claim became indefinite when Teva made a second statement to the PTO, a statement that respondents say is irreconcilable with the first. Whether the statements are irreconcilable to a skilled artisan rests on a factual question. A skilled artisan would read the prosecution history as a whole. If an earlier statement is transparently incorrect as a scientific matter, and a later statement 10 Respondents suggest (at 52-54) that the Federal Circuit considered this finding, but rejected it as legally irrelevant. The passage respondents quote merely summarizes the parties arguments. See Pet. App. 8a. Nowhere does the court actually address its precedent on prosecution-history errors.

30 22 is consistent with the best reading of the specification, the skilled artisan would read the later statement as a correction, not as invalidating an entire family of patents. Assessing the premise that a skilled artisan would recognize that the first statement is an obvious scientific error is just as factual as understanding chromatography or a graph. The district court found that a skilled artisan would grasp the obvious error and, therefore, see no conflict. The Federal Circuit made the factual assertion that the two definitions cannot be reconciled, Pet. App. 9a, ignoring the finding that a skilled artisan would reconcile them by deeming the earlier one an obvious error and the later one a necessary correction. At a minimum, if this Court corrects the standard of review, this issue requires reconsideration with proper deference to the factual record. 2. Respondents assert (at 54) that the weight average statement has legal significance even if it is wrong, because if Teva had not told the PTO the correct measure was weight average, Teva would not have obtained its patent. But Teva obtained the only remaining patent-in-suit, the 808 patent, nearly four years before that statement. See Pet. App. 137a; J.A. 1322a. Respondents argued throughout the district-court proceedings that prosecution history of subsequent patents was irrelevant even to the interpretation of the 808 patent. See Pet. App. 52a-53a. Now they take the opposite position: that even if the 808 patent was definite when it issued, subsequent prosecution history can render it indefinite. No case from this Court or the Federal Circuit supports this

31 23 springing-indefiniteness proposition. Subsequent prosecution history is at most an interpretive aid a tiebreaker, not a game-changer that can overcome other persuasive indicia of meaning. Even if the prosecution history supported respondents position, the weight it receives is far from clear now that the other patents are out of the case. That question is more appropriate for the Federal Circuit. 11 * * * * What did average molecular weight mean to an expert in polypeptide synthesis in 1994? No other court of appeals would regard that question as a legal one, especially if the testifying experts disagreed over it. In every other circuit, the district court s resolution of the expert dispute would get the full measure of appellate deference. In the Federal Circuit, it gets none. Nothing about patent litigation justifies that departure. To the contrary, the bedrock reasons for deference apply fully in patent cases: appellate factfinding not only multipl[ies] appeals by encouraging appellate retrial of the facts, it undermine[s] the legitimacy of the district courts in the eyes of litigants. Fed. R. Civ. P. 52 advisory committee s note (1985). This Court should restore district courts and the Federal Circuit to their proper roles. 11 Respondents claim (Br. 65) that there is an urgent need for this Court to resolve claim construction without remanding is without merit. The Federal Circuit can handle remanded cases swiftly. See Akamai Techs., Inc. v. Limelight Networks, Inc., No , ECF Nos. 297, 303 (Fed. Cir.) (argument about nine weeks after remand). Respondents have not obtained the FDA approvals necessary to launch a product in any event.

32 24 CONCLUSION The judgment of the court of appeals should be reversed. Respectfully submitted. JAY P. LEFKOWITZ, P.C. JOHN C. O QUINN JASON M. WILCOX KIRKLAND & ELLIS LLP 655 Fifteenth Street, N.W. Washington, DC ALAN M. DERSHOWITZ 1575 Massachusetts Ave. Cambridge, MA WILLIAM M. JAY Counsel of Record WILLIAM G. JAMES, II GOODWIN PROCTER LLP 901 New York Ave., N.W. Washington, DC wjay@goodwinprocter.com (202) DAVID M. HASHMALL ELIZABETH J. HOLLAND STEVEN J. BERNSTEIN GOODWIN PROCTER LLP 620 Eighth Ave. New York, NY DARYL L. WIESEN HENRY C. DINGER JOHN C. ENGLANDER NICHOLAS K. MITROKOSTAS TODD MARABELLA JAIME A. SANTOS GOODWIN PROCTER LLP Exchange Place 53 State Street Boston, MA Counsel for Petitioners September 10, 2014