dissenting,
with whom RADER, Chief Judge, REYNA and WALLACH, Circuit Judges, join.District judges, both parties in this case, and the majority of intellectual property lawyers and academics around the country will no doubt be surprised by today’s majority opinion — and for good reason. The majority opinion is surprising because it refuses to acknowledge what experience has shown us and what even a cursory reading of the Supreme Court’s decision in Markman v. Westview Instruments, Inc., 517 U.S. 370, 116 S.Ct. 1384, 134 L.Ed.2d 577 (1996), confirms: construing the claims of a patent at times requires district courts to resolve questions of fact. And, it puts itself at odds with binding congressional and Supreme Court authority when it refuses to abide by the requirements of Rule 52(a)(6) of the Federal Rules of Civil Procedure, which expressly instructs that, on appeal, all “findings of fact ... must not be set aside unless clearly erroneous.” It is also surprising because, having, for the third time, invited a broad swath of the intellectual property community to express opinions regarding the merits of Cybor Corp. v. FAS Technologies, Inc., 138 F.3d 1448 (Fed.Cir.1998) (en banc), we now premise our refusal to change its holding on principles of stare decisis — that, and a professed inability to come up with a workable alternative to de novo review.1
Criticism of and debate over Cybor have been widespread since it issued — not only among legal scholars and patent practitioners, but also among members of this court. Despite this fact, the majority suggests, for the first time in the ongoing debate over it, that Cybor is too firmly established in our case law to be rethought. In fact, it appears that some members of today’s 6-4 majority believe the pull of stare decisis is so strong that it prevents them from acting on their long-term convictions that Cybor was wrongly decided. No reasoned application of stare decisis principles supports that conclusion.
To the extent the majority is motivated not just by a resistance to change, but by concern over what standard we should change to, those concerns can be allayed by reference to Rule 52(a)(6) of the Federal Rules of Civil Procedure, the Supreme Court’s case law governing that rule, and a realistic assessment of what the claim construction process entails.
Because principles of stare decisis do not justify retention of the rule of Cybor and the appropriate standard of review is dictated by Rule 52(a), I respectfully dissent.
I.
In Cybor, this court held that claim construction, “including any allegedly fact-based questions relating to claim construction,” presents “a purely legal question” subject to de novo review. Cybor, 138 F.3d at 1456. We reached that conclusion even though, in Markman, the Supreme Court repeatedly acknowledged the factual component of claim construction. There, the Court: (1) labeled claim construction as a “mongrel practice,” (2) suggested that construing a patent’s claims “falls somewhere between a pristine legal standard and a simple historical fact,” (3) indicated that “there could be a case in which a simple credibility judgment would suffice to choose between experts whose testimony was equally consistent with a patent’s internal logic,” (4) discussed the need “to ascertain whether an expert’s proposed *1297definition fully comports with the specification and claims,” and (5) described claim construction’s “evidentiary underpinnings.” Markman, 517 U.S. at 378, 388-90, 116 S.Ct. 1384 (citation and internal quotation marks omitted). Despite being urged to do so by both parties, the Patent and Trademark Office, and multiple amici, the majority refuses to overturn Cybor.2 The majority rests its judgment primarily on the principles of stare decisis. It asserts that our fifteen years of experience with Cybor teach that our continued de novo review of all claim construction determinations is needed to assure greater “reliability of outcome” and “interjurisdictional uniformity.” Maj. Op. at 1280, 1281.
Considerations of stare decisis, however, do not justify adhering to precedent that misapprehends the Supreme Court’s guidance, contravenes the Federal Rules of Civil Procedure, and adds considerable uncertainty and expense to patent litigation.
II.
Stare decisis is an important part of our jurisprudence, and departing from our precedent is not something we should do lightly. The doctrine “promotes the evenhanded, predictable, and consistent development of legal principles, fosters reliance on judicial decisions, and contributes to the actual and perceived integrity of the judicial process.” Payne v. Tennessee, 501 U.S. 808, 827, 111 S.Ct. 2597, 115 L.Ed.2d 720 (1991). It also serves to guard against “arbitrary discretion.” Hubbard v. United States, 514 U.S. 695, 711, 115 S.Ct. 1754, 131 L.Ed.2d 779 (1995) (citations and internal quotation marks omitted).
“Stare decisis is not an inexorable command!, however]; rather it ‘is a principle of policy and not a mechanical formula of adherence to the latest decision.’ ” Payne, 501 U.S. at 828, 111 S.Ct. 2597 (quoting Helvering v. Hallock, 309 U.S. 106, 119, 60 S.Ct. 444, 84 L.Ed. 604 (1940)). Its force varies from case to case, moreover — carrying the most weight where reliance interests are at stake, but the least weight where the departure from precedent would not change substantive rights and would “not affect the way in which parties order their affairs.” Pearson v. Callahan, 555 U.S. 223, 233, 129 S.Ct. 808, 172 L.Ed.2d 565 (2009); see also Payne, 501 U.S. at 828, 111 S.Ct. 2597. “Revisiting precedent is particularly appropriate where ... a departure would not upset expectations ... and experience has pointed up the precedent’s shortcomings.” Pearson, 555 U.S. at 233, 129 S.Ct. 808. The Supreme Court has noted that departing from precedent especially is appropriate “when governing decisions ... are badly reasoned.” Payne, 501 U.S. at 827, 111 S.Ct. 2597 (“[W]hen governing decisions are unworkable or are badly reasoned, ‘this Court has never felt constrained to follow precedent.’” (quoting Smith v. Allwright, 321 U.S. 649, 665, 64 S.Ct. 757, 88 L.Ed. 987 (1944))); see also Helvering, 309 U.S. at 119, 60 S.Ct. 444 (cautioning against blindly applying stare decisis when adhering to precedent would “involve! ] collision with a prior doctrine more embracing in its scope, intrinsically sounder, and verified by experience”).
Consistent with this Supreme Court guidance, we have explained that stare decisis does not stand in the way of abro*1298gating our case law — even entire bodies of it — in at least three circumstances: when we conclude our case law (1) was wrongly decided, see, e.g., Wilson v. United States, 917 F.2d 529, 536 (Fed.Cir.1990) (en banc); (2) is at odds with congressional directives, see, e.g., Akamai Technologies, Inc. v. Limelight Networks, Inc., 692 F.3d 1301, 1318 (Fed.Cir.2012) (en banc); or (3) has had negative consequences, see, e.g., Therasense, Inc. v. Becton, Dickinson & Co., 649 F.3d 1276, 1288 (Fed.Cir.2011) (en banc). With these principles in mind, this court has not hesitated to revisit its own precedent. See, e.g., Cardiac Pacemakers, Inc. v. St. Jude Medical, Inc., 576 F.3d 1348 (Fed.Cir.2009) (en banc); In re Seagate Tech., LLC, 497 F.3d 1360 (Fed.Cir.2007) (en banc); Fisher v. United States, 402 F.3d 1167 (Fed.Cir.2005) (en banc); Knorr-Bremse Systeme Fuer Nutzfahrzeuge GmbH v. Dana Corp., 383 F.3d 1337 (Fed.Cir.2004) (en banc). Indeed, we have said that it is “ ‘[t]he province and obligation of the en banc court ... to review the current validity of challenged prior decisions.’ ” Robert Bosch, LLC v. Pylon Mfg. Corp., 719 F.3d 1305, 1316 (Fed.Cir.2013) (en banc) (quoting United States v. Aguon, 851 F.2d 1158, 1167 n. 5 (9th Cir.1988) (en banc)). And, we have made clear that this includes overturning precedent set by this court en banc when appropriate. See Nobelpharma USA, Inc. v. Implant Innovations, Inc., 141 F.3d 1059, 1068 (Fed.Cir.1998) (overruling Atari, Inc. v. JS & A Grp., Inc., 747 F.2d 1422 (Fed.Cir.1984) (en banc), by “changing] our precedent and holding] that whether conduct in procuring or enforcing a patent is sufficient to strip a patentee of its immunity from the antitrust laws is to be decided as a question of Federal Circuit law”).
Thus, both Supreme Court case law and our own teach that it is in cases like this one that stare decisis is weakest.
III.
Reversing Cybor will not “upset settled expectations on anyone’s part.” Pearson, 555 U.S. at 233, 129 S.Ct. 808. The one thing clear about Cybor is that no one in the legal community — except perhaps the members of the majority — has come to believe that either the wisdom or vitality of Cybor is settled. Whether one urges the retention of the holding in Cybor (as do some amici) or urges its revision (as do the parties, the Patent and Trademark Office, and the rest of the amici), it is hard to dispute that tumult has surrounded Cybor since it was decided. During its short life, Cybor repeatedly has been criticized as poorly reasoned. That criticism has come from members of this court, from district court judges, and from academics and practitioners across the country.
Our internal debate over Cybor has been heated, and has not abated over time. There were several ardent detractors from the rule announced in Cybor at the time it was announced. See, e.g., Cybor, 138 F.3d at 1478, 1480 (Newman, J., additional views) (“By continuing the fiction that there are no facts to be found in claim interpretations, we confound rather than ease the litigation process.... However, the Supreme Court has relieved us of adherence to this fiction, by its recognition of the factual component of claim interpretation.”), id. at 1463 (Mayer, C.J., concurring in the judgment) (stating that the Cybor majority opinion “profoundly misapprehends” the Supreme Court’s decision in Markman); id. at 1473 (Rader, J., dissenting from the pronouncements on claim interpretation in the en banc opinion, concurring in the judgment, and joining part IV of the en banc opinion). Even some of the less vocal critics who concurred in the result in Cybor expressed hesitation regarding the wisdom of either *1299the rule established or the legitimacy of its underpinnings. See id. at 1463 (Plager, J., concurring) (“Whether this approach to patent litigation will in the long run prove beneficial remains to be seen.”); see also id. at 1463 (Bryson, J., concurring) (“[W]e approach the legal issue of claim construction recognizing that with respect to certain aspects of the task, the district court may be better situated than we are, and that as to those aspects we should be cautious about substituting our judgment for that of the district court.”).
Since Cybor, our internal debate has continued. In Phillips v. AWH Corp., 415 F.3d 1303 (Fed.Cir.2005) (en banc), the order granting rehearing en banc asked the parties to address whether it is “appropriate for this court to afford any deference to any aspects of trial court claim construction rulings.” Id. at 1328. Despite receiving considerable input from the parties and amici, the Phillips majority, without explanation, “decided not to address that issue at this time.” Id. In dissent, however, Judge Mayer levied a pointed criticism of Cybor, (1) discussing “the absurdity[ ] of this court’s persistence in adhering to the falsehood that claim construction is a matter of law devoid of any factual component,” (2) stating that, “[i]n our quest to elevate our importance, we have ... disregarded our role as an appellate court ... undermining] the legitimacy of the process, if not the integrity of the institution,” and (3) observing that “we are obligated by Rule 52(a) to review the factual findings of the district court that underlie the determination of claim construction for clear error.” Id. at 1330, 1332 (Mayer, J., dissenting).
We have revisited the question multiple times since then: (1) in 2006 in Amgen Inc. v. Hoechst Marion Roussel, Inc., 469 F.3d 1039 (Fed.Cir.2006); (2) in 2011 in Retractable Technologies, Inc. v. Becton, Dickinson & Co., 659 F.3d 1369 (Fed.Cir.2011); and (3) even as recently as a year ago in Highmark, Inc. v. Allcare Health Management Systems, Inc., 701 F.3d 1351 (Fed.Cir.2012), where questions of claim construction were not even at issue. See, e.g., Amgen, 469 F.3d at 1043 (Newman, J., dissenting from denial of reh’g en banc) (“The Federal Circuit’s position that patent interpretation requires more rigorous appellate review than other fact/law issues has not withstood the test of experience. It is time to reopen the question and to rethink, en banc, the optimum approach to accuracy, consistency, and predictability in the resolution of patent disputes.... ”); id. at 1046 n. 3 (Moore, J., dissenting from denial of reh’g en banc) (highlighting the problems Cybor has caused for district courts attempting to construe patent claims); id. at 1045 (Gajarsa, Linn, & Dyk, JJ., concurring in denial of reh’g en banc) (noting that the concurrence “should not be read as ... an unqualified endorsement of the en banc decision in Cybor”)-, Retractable Techs., 659 F.3d at 1373 (Moore, J., dissenting from denial of reh’g en banc) (“The Supreme Court held that claim construction was a ‘mongrel practice.’ As such it is clearly a mixed question of law and fact and deference should be given to the factual parts.... [W]e must acknowledge the factual underpinnings of this analysis and there should be deference.” (citation omitted)); Highmark, 701 F.3d at 1362 (Moore, J., dissenting from denial of reh’g en banc) (citing to Cybor and stressing that “[w]e need to avoid the temptation to label everything legal and usurp the province of the fact finder with our manufactured de novo review”).
Notably, not once during this internal dialogue over the rule promulgated in Cy-bor did anyone contend that stare decisis alone should put an end to our debate. Two members of the current majority have been among the harshest critics of Cy-*1300bor — contending on multiple, and even recent, occasions that it was poorly reasoned, impractical, and should be reversed. A third conceded that Cybor’s rule may be too broad and perhaps should not apply where, as here, the trial court was forced to resort to extrinsic evidence to assess the meaning of claim terms. In none of their discussions of Cybor was concern regarding stare decisis raised. It certainly was never exalted to the hard stop on further consideration of Cybor’s merits that the majority now finds it to be.
And, the debate over Cybor has not all been internal to our court. The external debate has been both consistent and widespread. See, e.g., Amicus Br. of United States, Retractable Techs., Inc. v. Becton, Dickinson & Co., No. 11-1154, 2012 WL 5940288, at *20-21 (U.S. Nov. 28, 2012) (setting out the Solicitor General’s observation that (1) “some claim-construction decisions will depend on a district court’s resolution of factual questions,” (2) this court’s “decision in Cybor does not identify any reason that such factual findings should not be given the deference ordinarily required by Federal Rule of Civil procedure 52(a),” and (3) “appellate courts routinely defer to factual findings made by district courts and juries”). District judges have opposed de novo review, describing it as ill conceived and illogical. See, e.g., Amgen, Inc. v. Hoechst Marion Roussel, Inc., 339 F.Supp.2d 202, 226 n. 23 (D.Mass.2004) (describing the “conundrum” our claim construction jurisprudence has created by “discouraging resort to extrinsic evidence while at the same time urging courts to begin claim construction by considering the plain and customary meaning of a term as understood by one skilled in the art ”); Judge James F. Holderman & Halley Guren, The Patent Litigation Predicament in the United States, Univ. Ill. J.L., Tech. & Pol’y 1, 6-7, 14-15 (2007) (noting that “claim construction involves many of what one would consider to be factual determinations,” stressing that the Supreme Court in Markman “said nothing ... about the de novo standard of review,” and calling for a more deferential review of district court claim constructions); The Honorable William G. Young & Professor R. Carl Moy, Panel Discussion, High Technology Law in the Twenty-First Century: Second Annual High Technology Law Conference, 21 Suffolk Transnat’l L.Rev. 13, 19 (1997) (statements of the Honorable William G. Young).
As have practitioners. See, e.g., Frederick L. Whitmer, Claim Construction in Patent Cases: A Question of Law?, 2 No. 6 Landslide 14, 16-17 (2010) (criticizing our court’s interpretation of the Supreme Court’s guidance in Markman and calling for recognition of “the constituent factual component of claim construction decision making”); Donald R. Dunner & Howard A. Kwon, Cybor Corp v. FAS Technologies: The Final Say on Appellate Review of Claim Construction?, 80 J. Pat. & Trademark Off. Soc’y 481, 492 (1998) (“[N]ot-withstanding its decision that claim construction was an issue for the judge and not the jury, the Court in Markman II seemed to consider the issue a mixed question of law and fact — a characterization that would resist straightforward application of the de novo standard.”); Luke L. Dauchot, The Federal Circuit’s De Novo Review of Patent Claim Construction: A Need for a More Balanced Approach, 18 Am. Bar Ass’n Sec. Pub. I.P.L. 1, 4 (1999) (“A proper approach recognizes that patent claim interpretation is ‘a mongrel practice’ and delegates the fact-finding process to trial courts.... ”).
Academics have been particularly harsh in their criticism of Cybor and have suggested that we reverse it. See, e.g., J. Jonas Anderson & Peter S. Menell, Infor*1301mal Deference: An Historical, Empirical, and Normative Analysis of Patent Claim Construction, 108 Nw. U.L.Rev. (forthcoming), Sept. 9, 2013 manuscript at *57-59 (arguing that Cybor “misapprehends” Supreme Court precedent, “deprive[s] the district court of critical evidence bearing on claim meaning,” and “undermines the appellate process” by leaving “[t]he parties, the public, and the appellate court” with an “anemic record — typically limited to the intrinsic evidence”); Eileen M. Herlihy, Appellate Review of Patent Claim Construction: Should the Federal Circuit Be Its Own Lexicographer in Matters Related to the Seventh Amendment?, 15 Mich. Telecomm. Tech. L.Rev. 469, 515 (2009) (“A de novo standard of review ... runs contrary to the repeated and consistent word choices made by the Court indicating that the Court considers claim construction to be a mixed issue of fact and law.”); Kimberly A. Moore, Markman Eight Years Later: Is Claim Construction More Predictable?, 9 Lewis & Clark L.Rev. 231, 231 (2005) (observing the “concern among the bench and bar that the Federal Circuit’s de novo review of district court claim construction decisions ... ha[s] caused considerable unpredictability”); John R. Lane & Christine A. Pepe, Living Before, Through, and With Markman: Claim Construction as a Matter of Law, 1 Buff. Intell. Prop. L.J. 59, 71 (2001) (“In Markman II, the Supreme Court did concede that there are factual underpinnings to claim construction determinations, raising the logical question of whether de novo review is the appropriate standard.” (footnote omitted)).
In short, the only expectation about Cybor that appears “settled” is the expectation that one day this court might recognize that Cybor is inconsistent with Supreme Court precedent, the Federal Rules of Civil Procedure, and the practical realities involved in the claim construction process, and would reverse it.
Parties do not make claim drafting decisions based on the standard of review we apply to trial court claim constructions. Nor could they given the panel-dependent nature of our own determinations. See Donald R. Dunner, A Retrospective of the Federal Circuit’s First 25 Years, 17 Fed. Cir. B.J. 127, 130 (2007) (noting that many believe “that Federal Circuit predictability is not what it should be and that its decisions are often panel-dependent and result-oriented”); R. Polk Wagner & Lee Petherbridge, Is the Federal Circuit Succeeding? An Empirical Assessment of Judicial Performance, 152 U. Pa. L.Rev. 1105, 1112 (2004) (“Our findings ... indicate that claim construction at the Federal Circuit is panel dependent.”). It is difficult to accept the proposition that our claim construction jurisprudence is a measure against which litigants make important business or innovation decisions. Claim construction disputes are very fact specific — patents do not follow a formulaic structure, or even contain oft repeated language. Claims are drafted, redrafted, and amended in ways intended to reflect and capture particular inventions in a particular field, to avoid very specific prior art, and to respond to the rejections of the unique patent examiner involved in the application process. It is rare that any two claims we review contain the same phrasing, and even more rare that the context in which the phrasing is used would not alter the meaning of even almost identical words.3 Compare Baldwin *1302Graphic Sys., Inc. v. Siebert, Inc., 512 F.3d 1338, 1342-43 (Fed.Cir.2008) (finding that, as a general rule, the words “an” or “a” in a patent claim carry the meaning of “one or more”), with TiVo, Inc. v. EchoStar Commc’ns Corp., 516 F.3d 1290, 1303 (Fed.Cir.2008) (finding that “whether ‘a’ or ‘an’ is treated as singular or plural depends heavily on the context of its use” and concluding that “claims and written description in this case make clear that the singular meaning applies”). Combining the uniqueness of each claim term to be reviewed with the variations in rationale employed by the divergent members of this court, provides little practical guidance regarding how any claim construction dispute might be resolved in this forum— and certainly not the uniform reliability of outcome with which the majority now credits our jurisprudence in this area.
The fact that we have been engaging in a flawed practice for too long does not, alone, create the type of settled expectations stare decisis is meant to protect. Because settled expectations will not be disrupted and no substantive rights will be reordered, stare decisis simply does not stand in the way of this court addressing the merits of Cybor and acknowledging that the rule of law pronounced therein is an incorrect one.
IV.
It is also clear that stare decisis does not stand in the way of overturning Cybor because Cybor is predicated on a mischar-acterization of the Supreme Court’s guidance in Markman and ignores the claim construction process we have ordered district courts to employ. In short, it need not be followed because its premises are wrong. See Wilson, 917 F.2d at 536 (overruling precedent that misconstrued congressional intent).
As noted above, Cybor misapprehends the Supreme Court’s decision in Mark-man, ignoring numerous instances where the Court acknowledged that claim construction can present factual questions. The Supreme Court did not base its conclusion on the fact that a patent is a legal instrument whose construction presents a pure question of law. If it had, there would have been no need for the Court to conduct such a thorough analysis of whether the Seventh Amendment required a jury to resolve issues of claim construction. That question would have needed no discussion if claim construction were purely an issue of law because juries have never been tasked with resolving purely legal questions. See Markman, 517 U.S. at 376-84, 116 S.Ct. 1384; see also Cybor, 138 F.3d at 1464 (Mayer, C.J., concurring in the judgment) (“Though it could have done so easily, the Court chose not to accept our formulation of claim construction: as a pure question of law to be decided de novo in all cases on appeal. If it had, there would have been no need for its extensive exegesis about the Seventh Amendment and whether juries must construe claims that have evidentiary underpinnings or whether the importance of uniformity is best served by giving these evidentiary questions of meaning to a judge.” (footnote omitted)).
While Cybor dismissed Markman’s discussion of the factual aspects of claim construction as mere “prefatory comments,” 138 F.3d at 1455, and insisted that, under Markman, claim construction is a completely legal exercise subject to de novo review, id. at 1456, that conclusion does not flow from Markman. There, the Supreme Court not only acknowledged claim *1303construction’s factual aspects, it also said nothing to suggest that a de novo standard of review would be appropriate. See Retractable Techs., 659 F.3d at 1373 (Moore, J., dissenting from denial of reh’g en banc) (“The Supreme Court held that claim construction was a ‘mongrel practice.’ As such it is clearly a mixed question of law and fact and deference should be given to the factual parts.” (citation omitted)). Markman’s holding was limited to the Court’s determination “that the construction of a patent, including terms of art within its claims, is exclusively within the province of the court.” Markman, 517 U.S. at 372, 116 S.Ct. 1384. There are many circumstances in which trial judges act as triers of both fact and law; in all of those, deference to the factual components of that decision-making is undoubtedly due. “Stating that something is better decided by the judge is not the same as saying it is a matter of law.” Highmark, 701 F.3d at 1362 (Moore, J., dissenting from denial of reh’g en banc).4 And even saying something is a matter of law does not answer the question of the standard of review an appellate court should apply. See Pierce v. Underwood, 487 U.S. 552, 560, 562, 108 S.Ct. 2541, 101 L.Ed.2d 490 (1988) (observing that, “[i]n some cases, such as the present one, the attorney’s fee determination will involve a judgment ultimately based on a purely legal issue governing the litigation,” but concluding that “sound judicial administration counsels[] deferential review of a district court’s decision regarding attorney’s fees” despite its legal character).
Those amici who find great significance in the Supreme Court’s citation to Miller v. Fenton, 474 U.S. 104, 106 S.Ct. 445, 88 L.Ed.2d 405 (1985), in Markman miss the mark. That citation does not, as those amici claim, decide the fact/law question or the question of the appropriate level of appellate review of claim construction determinations. In Miller, the Supreme Court concluded that the ultimate question of whether a confession was sufficiently voluntary to comport with due process, while a mixed question of fact and law, was subject to independent federal review. As the Court noted in Markman, it had concluded in Miller that, where a question “falls somewhere between a pristine legal standard and a simple historical fact,” the conclusion as to which judicial actor is best positioned to decide a question at times turns on the sound administration of justice, rather than a pure fact/law distinction. 517 U.S. at 388, 116 S.Ct. 1384. Though, in Miller, the Court decided that the sound administration of justice supported the conclusion that the ultimate constitutional question of whether a confession was voluntary should be reserved for federal, rather than state, courts, Miller says nothing about the standard of *1304review one federal tribunal should apply to the inquiries of another, or how the sound administration of justice would divvy up the responsibility of claim construction as between the trial and appellate courts.
In fact, in Miller itself, the Court concluded that a presumption of correctness still must be afforded to all “subsidiary factual questions” decided by the state courts. 474 U.S. at 112, 106 S.Ct. 445. And, the Court was careful to explain that its determination of what the sound administration of justice called for vis-á-vis the federal and state courts was reached in the absence of congressional directives to the contrary.
In Markman, the Supreme Court said that judicial efficiencies supported allocation of claim construction determinations to the court rather than the jury. It did not say that “subsidiary factual determinations” made by trial courts ceased to be subject to the deference congressionally mandated by the Federal Rules of Civil Procedure, however. And, it did not say that it was this court and only this court to which the question should be allocated. Indeed, in Cooter & Gell v. Hartmarx Corp., 496 U.S. 384, 110 S.Ct. 2447, 110 L.Ed.2d 359 (1990), decided five years after Markman, the Court addressed the question of how the sound administration of justice can impact the standard of review of questions that involve both factual and legal components. There, while the Court acknowledged that some purely legal inquiries are involved in determinations pursuant to Rule 11 of the Federal Rules of Civil Procedure, it found that the entire determination must be reviewed by the courts of appeals under an abuse of discretion standard. Id. at 403-04, 110 S.Ct. 2447. Returning to the same type of inquiry employed in Miller, the Supreme Court explained that where, as in a Rule 11 inquiry, the line between fact and law is difficult to divine and the trier of fact needs flexibility to decide unique facts that resist generalization, it is the trial judge who is the judicial actor best suited to decide the question. Id. In such instances, the Court found that the sound administration of justice to which it harkened in Miller and again in Markman mandated a fully deferential standard of review.
It is notable that at least one Supreme Court Justice on the Court when Mark-man was decided believes that, if Mark-man can be said to have decided the standard of review to be applied to claim construction determinations, it decided that question very differently than we did in Cybor and than we continue to do today. In Gasperini v. Center for Humanities, Inc., 518 U.S. 415, 116 S.Ct. 2211, 135 L.Ed.2d 659 (1996), dissenting from the judgment there, Justice Stevens described the Court’s decision in Markman as one of three that term in which courts of appeals were “assigned ... the task of independently reviewing similarly mixed questions of law and fact,” and described the nature of that review as one in which appellate courts are required “to construe all record inferences in favor of the fact-finder’s decision and then to determine whether, on the facts found below, the legal standard has been met.” Id. at 442-43, 116 S.Ct. 2211 (Stevens, J., dissenting).
Markman’s citation to Miller, accordingly, lends no support to the notion that Markman somehow dictated the result in Cybor. It only helped explain why the court, rather than the jury, was chosen as the appropriate decision maker. Cybor was not compelled by the Supreme Court’s guidance; as explained in section V below, it is actually a wide departure from it.
Cybor also ignores the realities of the claim construction process. As our en banc court in Phillips observed:
*1305[B]ecause extrinsic evidence can help educate the court regarding the field of the invention and can help the court determine what a person of ordinary skill in the art would understand claim terms to mean, it is permissible for the district court in its sound discretion to admit and use such evidence. In exercising that discretion, and in weighing all the evidence bearing on claim construction, the court should keep in mind the flaws inherent in each type of evidence and assess that evidence accordingly.
415 F.3d at 1319 (emphases added). Cy-bor cannot be squared with this court’s own well-respected description of the very claim construction process to which it purports to apply.
The majority concedes that claims are to be interpreted from the perspective of one of skill in the art at the time of the invention unless it appears from the surrounding record — the specification and prosecution history — that the patentee acted as his own lexicographer to provide a contrary meaning. Maj. Op. at 1289 (citing Phillips, 415 F.3d at 1314). It concludes, however, that all we need to put ourselves into the shoes of a skilled artisan are the patent documents and, perhaps (though not necessarily), some explanation regarding the technology at issue and a dictionary or treatise. It believes we do not need to hear from experts regarding the state of the known science or art at the time of the invention, the commonly understood meaning, if any, of the particular terms or phrases employed, the level of education and skill one reading such a patent would have, or whether there are particular treatises or dictionaries to which a skilled artisan would have turned at the time. See id. at 1284. And, it believes that the conclusions it gleans from the patent documents, including the entirety of the prosecution history, expert descriptions of the technology, and dictionaries are all legal conclusions; that no finding made by any judicial officer in the process of claim construction constitutes a subsidiary factual one. See id. at 1284-85.
The majority justifies these conclusions by analogizing the claim construction process to the interpretation of statutes, where courts routinely consider contemporaneous dictionaries or even the testimony of historians to help determine the meaning of words and phrases therein. See id. The analogy is not a sound one, however.
Statutes are duly enacted laws of broad applicability. Their interpretation by an appellate court is binding on all who would be impacted by that statute in that circuit, whether parties to the original action or not. They are drafted by those with congressional authority to enact such laws and are to be given a meaning common to all. Patents are drafted ex parte, are revised in a closed-door examination process, their terms are, as noted before, unique to the invention at issue, and are assertable only against individual infringers in private actions. The two are simply not of the same ilk. See Markman v. Westview Instruments, Inc., 52 F.3d 967, 998 n. 8 (Fed.Cir. 1995) (Mayer, J., concurring in the judgment) (“Patents cannot be baby statutes .... ”); Amgen, 469 F.3d at 1040-41 (Michel, C.J., dissenting from denial of reh’g en banc) (observing that, in statutory interpretation, a judge construes terms from the perspective of a skilled legal artisan looking at the words only, not from the perspective of a different individual — one skilled in the relevant field of technology in light of the intrinsic and extrinsic record).
The parties agree that there were disputed factual questions in this case that required examination of extrinsic evidence. In the proceedings before the district court and again on appeal, the parties *1306disputed whether the claim term “voltage source means” should be treated as a means-plus-function limitation under 35 U.S.C. § 112, ¶ 6. Lighting Ballast Control LLC v. Philips Elecs. N. Am. Corp., 498 Fed.Appx. 986, 989-90 (Fed.Cir.2013). The use of the word “means” triggers the presumption that the limitation is a means-plus-function term, but that presumption “may be rebutted if the claim itself recites sufficient structure for performing the function.” Id. at 990. The parties focused on whether “voltage source means” denoted a particular structure to those of skill in the art (i.e., whether the term had a specific meaning used by those of skill in the art to describe a defined structure or specific class of structures). See id. If skilled artisans understood “voltage source means” to refer to a defined structure, it would not be considered a means-plus-function limitation. Id. The specification and prosecution history, however, did not resolve the question. Thus, it became necessary and appropriate to look outside the intrinsic record and to consider the testimony of Lighting Ballast’s expert, Dr. Roberts. See id.; see also Rembrandt Data Techs., LP v. AOL, LLC, 641 F.3d 1331, 1341 (Fed.Cir.2011) (“When determining whether a claim term recites sufficient structure, we examine whether it has an understood meaning in the art.”). When situations like this arise, it is appropriate — and sometimes necessary — to make findings based on extrinsic evidence that relate to the meaning of a disputed term. Resolution of these fact-intensive disputes is an area where district courts’ expertise deserves the deference that Rule 52(a)(6) requires.
Both parties, the PTO, and most amici agree that there are factual components to claim construction. Even among the amici that favor retaining Cybor’s de novo review of all aspects of claim construction, most readily identified factual questions that arise during claim construction. Microsoft Corp., for instance, advocated retaining Cybor, but nonetheless listed numerous factual questions that it concedes could arise during claim construction and would require the court to:
determin[e] the field[s] of the invention and the knowledge of a person having ordinary skill in the art; determin[e] the art-accepted meanings of terms used in an issued claim and also used in the specification and/or prior art; deter-min[e] the date of the invention and/or the effective filing date of the patent application; determin[e] whether a proposed construction would exclude all embodiments in the specification or, conversely, whether any embodiment supports the construed issued claim; identifying explicit or implicit definitions in the specification; [and] determin[e] the disclosure of cited prior art references (which are part of the ‘intrinsic evidence’ for claim construction) asserted as invalidating prior art and/or distinguished in the prosecution history.
Amicus Br. at 4-5. Similarly, the Austin Intellectual Property Law Association observed that “district courts are charged with taking evidence of specialized meanings in ... patent interpretation.” Amicus Br. at 8. Likewise, the brief filed by Cisco Systems, Inc. et al. acknowledged that a case could arise where “a question of meaning peculiar to a trade or profession [could] turn[ ] on the resolution of contested questions of historical fact.” Amicus Br. at 24-26 (citation and internal quotation marks omitted).
In sum, it is hard to understand how either the majority in Cybor or the majority here can dispute that claim construction sometimes requires a district court to resolve contested factual issues. Cybor is, thus, based on a faulty premise — that claim construction is a purely legal exer*1307cise. This reveals deep flaws in Cybor’s reasoning, justifying a departure from it. See Payne, 501 U.S. at 827, 111. S.Ct. 2597 (permitting departure from decisions that prove “unworkable or are badly reasoned”).
V.
Stare decisis also must give way because, by refusing to acknowledge the factual component of claim construction, Cy-bor contravenes the clear directives of Federal Rule of Civil Procedure 52(a)(6). When a district court makes findings of fact — as claim construction sometimes .requires — Rule 52(a)(6) provides clear instructions to this court: “Findings of fact, whether based on oral or other evidence, must not be set aside unless clearly erroneous .... ” The rule is clear on its face, and decisions interpreting it show that it makes no exception with regard to fact-finding in the claim construction context. As the Supreme Court has observed, “Rule 52(a) broadly requires that findings of fact not be set aside unless clearly erroneous. It does not make exceptions or purport to exclude certain categories of factual findings from the obligation of a court of appeals to accept a district court’s findings unless clearly erroneous.” Pullman-Standard v. Swint, 456 U.S. 273, 287, 102 S.Ct. 1781, 72 L.Ed.2d 66 (1982); see also Bose Corp. v. Consumers Union of U.S., Inc., 466 U.S. 485, 498, 104 S.Ct. 1949, 80 L.Ed.2d 502 (1984) (“We have repeatedly held that ... Rule [52(a) ] means what it says.”). Thus, there is direct conflict between Cybor — which expressly calls for de novo review of “any allegedly fact-based questions relating to claim construction,” 138 F.3d at 1456 — and Rule 52(a)(6)— which requires deference to all fact-findings that are not clearly erroneous. See Amicus Br. of United States at 9-13 (noting that “[ajppellate courts must defer to a trial court’s factual findings under Rule 52(a)” and that, “[g]iven the clear command of Rule 52(a), no justification exists to treat claim construction any differently”).
The law governing obviousness confirms Rule 52(a)’s broad applicability in patent disputes. Obviousness presents a question of law subject to de novo review, but it involves a number of subsidiary fact-findings. As the Supreme Court observed:
While the ultimate question of patent validity is one of law, ... the § 103 .condition ... lends itself to several basic factual inquiries. Under § 103, the scope and content of the prior art are to be determined; differences between the prior art and the claims at issue are to be ascertained; and the level of ordinary skill in the pertinent are resolved.
Graham v. John Deere Co. of Kansas City, 383 U.S. 1, 17, 86 S.Ct. 684, 15 L.Ed.2d 545 (1966). According to the Court, “[t]his description of the obviousness inquiry makes it clear that whether or not the ultimate question of obviousness is 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, 106 S.Ct. 1578, 89 L.Ed.2d 817 (1986); see also Panduit Corp. v. Dennison Mfg. Co., 810 F.2d 1561, 1569 (Fed. Cir.1987) (“Rule 52(a) is applicable to all findings on the four inquiries listed in Graham: scope and content of prior art; differences between prior art and claimed invention; level of skill; and objective evidence .... ”). Importantly, one of the key fact questions in an obviousness inquiry is what a prior art reference teaches — often, what is claimed and described in a previously issued patent. See Graham, 383 U.S. at 17, 86 S.Ct. 684. And, all findings regarding the scope and content of the prior art are subject to clear error review. *1308See Pcmduit, 810 F.2d at 1569. That we trust jurors to define the scope of patent claims in this context, but are less than comfortable allowing trial judges to do the same when considering the asserted patent claims is at least anomalous. Cybor is thus out of step with our other jurisprudence that faithfully applies Rule 52(a) in patent cases.
This conflict between Cybor and the Federal Rules of Civil Procedure means our case law must fall. As this court has observed,
The Federal Rules of Civil Procedure were promulgated by the Supreme Court pursuant to statutory authority and were implicitly adopted by Congress after transmission to Congress in their proposed form. See 28 U.S.C. §§ 2071-2074. In light of this statutory promulgation scheme, the Supreme Court has held that the Federal Rules of Civil Procedure are deemed to have “the force [and effect] of a federal statute.”
Bright v. United States, 603 F.3d 1273, 1279 (Fed.Cir.2010) (alteration in original) (quoting Sibbach v. Wilson & Co., 312 U.S. 1, 13, 61 S.Ct. 422, 85 L.Ed. 479 (1941)).“ Our cases dealing with the application of stare decisis where statutory interpretation is at issue thus provide useful guidance.
We often have held that stare decisis does not prevent our court from overturning its precedent when we conclude our prior jurisprudence runs contrary to what we believe are a statute’s directives. See, e.g., Akamai Techs., Inc. v. Limelight Networks, Inc., 692 F.3d 1301, 1306 (Fed.Cir.2012) (en banc) (overruling multiple decisions of this court where “we held that in order for a party to be liable for induced infringement, some other single entity must be liable for direct infringement”); Wilson v. United States, 917 F.2d 529, 536 (Fed.Cir.1990) (en banc) (overturning our earlier decision in Ulmet v. United States, 822 F.2d 1079 (Fed.Cir.1987), saying: “We have revisited the legislative history of [10 U.S.C.] § 1163(d) in this case. Our examination has brought to light that the legislative history of the sanctuary provision demands a different result from that reached in Ulmet.”)] Cardiac Pacemakers, Inc. v. St. Jude Med., Inc., 576 F.3d 1348 (Fed.Cir.2009) (en banc) (overruling prior case law because we believed that case law did not properly interpret 35 U.S.C. § 271(f)). In all of these instances, we concluded that stare decisis did not provide a basis for adhering to precedent that ran counter to the dictates of a statute, as properly interpreted.
We should bring our case law in line with the directives of Rule 52(a)(6), as we are required to do, and as we have done with respect to numerous statutory commands in the past. See Amicus Br. of United States at 9-13 (urging the court to overturn Cybor because it runs counter to Rule 52(a)’s clear commands); Amicus Br. of Am. Bar Ass’n at 12-13 (same); Amicus Br. of Prof. Peter Menell at 17-20 (In light of Rule 52(a)’s commands, “the Federal Circuit must defer to trial judges’ factual determinations in claim construction rulings.”); Amicus Br. of Am. Intellectual Prop. Law Ass’n at 4-6, 6 n.6 (“There is no reason for the review of patent claim construction, where the trial court makes constituent determinations of fact, to be any different from review of other ultimate issues of law that have factual underpinnings.”); Amicus Br. of Fed. Cir. Bar Ass’n at 7; Amicus Br. of Intellectual Prop. Owners Ass’n at 7 (stressing that Rule 52(a) requires deference to district courts’ findings of claim construction facts); Amicus Br. of Conn. Intellectual Prop. Law Ass’n at 12 (faulting Cybor for “say[ing] that patent cases have their very own Fed.R.Civ.P. 52(a)(6),” even though *1309there is “no legitimate reason to treat patent cases differently from other cases”); Amicus Br. of Fed’n Internationale Des Conseils en Propriete Intellectuelle at 12 (noting that de novo review of findings of claim construction facts “violates Federal Rule of Civil Procedure 52(a)(6)”).
The majority discounts concerns about the dictates of Rule 52(a)(6) by citing the Supreme Court’s statement in Pullman-Standard v. Swint that Rule 52(a) does not provide a clear formula for distinguishing fact from law. Maj. Op. at 1290 (citing Pullman-Standard, 456 U.S. at 288, 102 S.Ct. 1781). As the Court made clear in Pullman-Standard itself when reversing the Fifth Circuit’s refusal to give deference to a trial court’s factual inquiry, the fact that our inquiry might be a difficult one does not excuse the failure to undertake it. See Pullman-Standard, 456 U.S. at 288-90, 102 S.Ct. 1781. Indeed, the Supreme Court has on numerous occasions charged the courts of appeals with drawing distinctions between subsidiary or “historical facts” and the ultimate legal conclusion regarding the import of those facts, and with adjusting their standard of review accordingly. See Cooper Indus., Inc. v. Leatherman Tool Grp., Inc., 532 U.S. 424, 440 n. 14, 121 S.Ct. 1678, 149 L.Ed.2d 674 (2001) (“While we have determined that the Court of Appeals must review the District Court’s application of the Gore test [set out in BMW of North America, Inc. v. Gore, 517 U.S. 559, 116 S.Ct. 1589, 134 L.Ed.2d 809 (1996),] de novo, it of course remains true that the Court of Appeals should defer to the District Court’s findings of fact unless they are clearly erroneous.”); Ornelas v. United States, 517 U.S. 690, 699, 116 S.Ct. 1657, 134 L.Ed.2d 911 (1996) (holding that “determinations of reasonable suspicion and probable cause should be reviewed de novo on appeal” but “hastening] to point out that a reviewing court should take care both to review findings of historical fact only for clear error and to give due weight to inferences drawn from those facts by resident judges and local law enforcement officers”); Thompson v. Keohane, 516 U.S. 99, 110-12, 116 S.Ct. 457, 133 L.Ed.2d 383 (1995) (concluding that there are two distinct inquiries— one purely factual and another a mixed question — involved in “[t]he ultimate ‘in custody’ determination for Miranda purposes,” with deference to findings on all factual components due). As the PTO explains, “Congress gave no indication in the patent laws that it intended to displace the fundamental principle of appellate review for clear error.” Amicus Br. of United States at 12.5
VI.
The “undesired consequences” flowing from this court’s claim construction jurisprudence also justify departing from the law set out in Cybor. Cybor, 138 F.3d at *13101481 (Newman, J., additional views); see also Pearson, 555 U.S. at 233, 129 S.Ct. 808 (stating that revisiting ease law is “particularly appropriate” where experience has revealed its shortcomings). By refusing to acknowledge the factual component of claim construction, Cybor has made the claim construction process less transparent, accurate, predictable, and efficient, thereby imposing high “social costs.” See Anderson & Menell, supra, at *60-61; Whitmer, supra, at 16 (lamenting the “high reversal rate” with respect to claim construction that is the “consequence of the Cybor uncertainty principle”); Lane & Pepe, supra, at 71-73 (examining the uncertainty that results from de novo review).
A.
The primary interests furthered by stare decisis- — a doctrine rooted in policy- — • support departing from Cybor, not adhering to it. See Helvering, 309 U.S. at 119, 60 S.Ct. 444 (observing that stare decisis “is a principle of policy”). Preserving the stability of the law and protecting the public’s ability to “rel[y] on judicial decisions” are the central interests furthered by stare decisis. Payne, 501 U.S. at 827, 111 S.Ct. 2597. By withholding deference to district courts’ findings of claim construction facts, however, the interests of stability and predictability are disserved. See Highmark, 701 F.3d at 1362 (Moore, J., dissenting from denial of reh’g en banc) (“When we convert factual issues, or mixed questions of law and fact, into legal ones for our de novo review, we undermine the uniformity and predictability goals this court was designed to advance.”); see also Amicus Br. of Prof. Peter Menell at 15 (observing that “[t]he[ ] effects [of de novo review of claim construction determinations] continue to cast doubt on the predictability of patent litigation, discourage settlements following claim construction trial, delay resolution of patent disputes, and run up the overall costs of patent litigation”). Indeed, our resistance to changing Cybor is directly contrary to the purposes of Rule 52(a)(6): to promote stability in the judicial system by (1) avoiding undermining the legitimacy of district courts and (2) preventing unnecessary appeals by discouraging appellate retrial of factual issues. See Fed.R.Civ.P. 52 advisory committee’s note (1985).
Under the Cybor regime, a district court can construe a claim term, and an entire trial can follow premised on that construction. When the district court’s judgment is appealed, however, we review every aspect of its claim construction de novo, leaving us largely free to reinterpret claims— both upsetting parties’ expectations and undoing a tremendous amount of parties’ and district courts’ work in the process. See Cybor, 138 F.3d at 1476 (Rader, J., dissenting from the pronouncements on claim interpretation in the en banc opinion) (“To get a certain claim interpretation, parties must go past the district court’s Markman I proceeding, past the entirety of discovery, past the entire trial on the merits, past post-trial motions, past briefing and argument to the Federal Circuit— indeed past every step in the entire course of federal litigation, except Supreme Court review.”). Once here, moreover, as noted earlier, “[c]ommentators have observed that claim construction appeals are ‘panel dependent’ which leads to frustrating and unpredictable results for both the litigants and the trial court.” Retractable Techs., 659 F.3d at 1370 (Moore, J., dissenting from denial of reh’g en banc) (citations omitted). And, while the majority says that it is “no longer true” that there is a high reversal rate with respect to claim constructions by district courts, Majority Opinion at 1290, that is not what trial judges, litigants, and academics contend. As Professor Peter Menell says in his ami-*1311cus brief before this court: “Although we document a significant drop in the claim construction reversal rate since the Phillips decision, there still remains a high reversal rate compared to other areas of federal practice.”6 Amicus Br. at 15; see also Amicus Br. of Ass’n of Bar of N.Y. (“The high reversal rate of the district court claim construction, documented in numerous studies, is universally acknowledged. It is not an overstatement to conclude that the reversal rate has had a detrimental effect on the parties, the court, and the credibility of the patent system generally.” (footnote omitted)). Departing from Cybor and reviewing claim construction findings for clear error would introduce greater stability and less expense, and would afford the appropriate respect for district courts’ factual determinations — respect that Rule 52(a)(6) demands. As a consequence, this case presents an instance where overturning this court’s precedent will lead to greater stability and predictability, not less.
B.
Refusing to acknowledge that claim construction has a factual component effectively “deprives th[is] court, and the parties, of the accumulated progress and experience of the trial, including the findings of the trial judge, and leaves us on appeal with an expurgated record and generally inferior basis of decision.” Cybor, 138 F.3d at 1481 (Newman, J., additional views). By affording zero deference to any aspect of a district court’s claim construction, we ignore the reality that we lack the tools that district courts have available to resolve factual disputes fairly and accurately. As Judge Rader observed in dissenting in part in Cybor,
the trial judge enjoys a potentially superior position to engage in claim interpretation. For the complex case where the claim language and specification do not summarily dispose of claim construction issues, the trial court has tools to acquire and evaluate evidence that this court lacks. Trial judges can spend hundreds of hours reading and rereading all kinds of source material, receiving tutorials on technology from leading scientists, formally questioning technical experts and testing their understanding against that of various experts, examining on site the operation of the principles of the claimed invention, and deliberating over the meaning of the claim language. If district judges are not satisfied with the proofs proffered by the parties, they are not bound to a prepared record but may compel additional presentations or even employ their own court-appointed expert.
138 F.3d at 1478.
The Supreme Court has explained that “[t]he trial judge’s major role is the determination of fact, and with experience in fulfilling that role comes expertise.” Anderson v. Bessemer City, 470 U.S. 564, 574, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985). The Court also reminds us that “deferential review of mixed questions of law and fact is warranted when it appears that the district court is better positioned than the appellate court to decide the issue in question or that probing appellate scrutiny will *1312not contribute to the clarity of legal doctrine.” Salve Regina Coll. v. Russell, 499 U.S. 225, 233, 111 S.Ct. 1217, 113 L.Ed.2d 190 (1991); see also Cooter, 496 U.S. at 403, 110 S.Ct. 2447 (calling for deference to the decisions of “ ‘the judicial actor ... better positioned than another to decide the issue in question’ ” (alteration in original) (quoting Miller, 474 U.S. at 114, 106 S.Ct. 445)). District courts should be encouraged to resolve the factual questions bearing on claim construction and to develop a thorough record setting out their findings and the evidence supporting their conclusions. When they do, we overstep the bounds of our duty under Rule 52(a)(6) by duplicating, or ignoring, rather than deferring to that process. Anderson, 470 U.S. at 573, 105 S.Ct. 1504; cf. Highmark, 701 F.3d at 1365 (Reyna, J., dissenting) (urging that we “respectf ] the enduring balance between the trial judge and the appellate panel in carrying out their distinct responsibilities” by applying clear error review to trial court findings).
The concurrence downplays the extent to which we usurp the trial court’s function by adherence to Cybor by arguing both that claim construction rarely involves credibility determinations and that we are “quite as able” as district courts — or “sometimes better” able — to review the relevant documents in the record, such as the patent’s prosecution history. Concurrence at 1293. And the majority echoes these themes, contending that claim construction does not present questions of fact because it does not turn on credibility determinations and that leaving these questions to de novo review by our court assures greater correctness of result. The Supreme Court has made clear, however, that this narrow view of the trial court’s factfinding function is an inaccurate one. The district court’s expertise is “not limited to the superiority of the trial judge’s position to make determinations of credibility,” but instead extends to all factual determinations. Anderson, 470 U.S. at 574, 105 S.Ct. 1504. These determinations include findings “based on physical or documentary evidence or inference from other facts.” Id. The Supreme Court has explained that Rule 52(a) requires deference to these findings, as well as those that turn on witness credibility. Id. Indeed, the Court has rejected the concurrence’s reasoning with respect to Rule 52(a)(6), not only in its case law, but also through its rulemaking. In 1985, Rule 52(a) was amended, in part, because
[s]ome courts of appeal have stated that when a trial court’s findings do not rest on demeanor evidence and evaluation of a witnesses] credibility, there is no reason to defer to the trial court’s findings and the appellate court more readily can find them to be clearly erroneous. See, e.g., Marcum v. United States, 621 F.2d 142, 144-45 (5th Cir.1980). Others go further, holding that appellate review may be had without application of the “clearly erroneous” test since the appellate court is in as good a position as the trial court to review a purely documentary record.
Fed.R.Civ.P. 52 advisory committee’s note (1985) (collecting cases). The Advisory Committee continued:
The principal argument advanced in favor of a more searching appellate review of findings by the district court based solely on documentary evidence is that the rationale of Rule 52(a) does not apply when the findings do not rest on the trial court’s assessment of credibility of the witnesses but on an evaluation of documentary proof and the drawing of inferences from it, thus eliminating the need for any special deference to the trial court’s findings. These considerations are outweighed by the public interest in the stability and judicial *1313economy that would be promoted by recognizing that the trial court, not the appellate tribunal, should be the finder of facts. To permit courts of appeals to share more actively in the fact-finding function would tend to undermine the legitimacy of the district courts in the eyes of litigants, multiply appeals by encouraging appellate retrial of some factual issues, and needlessly reallocate judicial authority.
Id.
District court judges are provided training in all aspects of their duties, including claim construction in patent litigation under Phillips. They then employ that training repeatedly over the years, analyzing patents, their written descriptions, and prosecution histories, receiving testimony from inventors and experts, listening to tutorials on the relevant science, and probing counsel during hearings that sometimes last days. In this case, the trial court conducted a three-day evidentiary hearing. Because Cybor allows us to ignore these fact-intensive inquiries by its insistence on de novo review, it not only undermines the authority of district judges, it compromises the decision-making process on appeal. Our court is given free rein to interpret claim terms, but lacks the resources to do it right. See Dunner & Kwon, supra, at 497 (noting that “the Federal Circuit, by function and design, is ill-equipped to engage in the evidentiary evaluations relevant to claim construction that are the staple of district court judges”).
C.
Cybor also creates greater incentives for losing parties to appeal, thus discouraging settlements and increasing the length and cost of litigation. As Judge Rader observed in dissenting from the court’s pronouncements on claim interpretation .in Cybor, “unfettered review authority” undercuts certainty and discourages settlement. 138 F.3d at 1475. It is not until “the parties know the meaning of the claims [that] they can predict with some reliability the likelihood of a favorable judgment, factor in the economies of infringement, and arrive at a settlement to save the costs of litigation.” Id. But under Cybor, “the trial court’s early claim interpretation provides no early certainty at all, but only opens the bidding. The meaning of a claim term is not certain (and the parties are not prepared to settle) until nearly the last step of the process — decision by the Court of Appeals for the Federal Circuit.” Id. at 1476; see also Amicus Br. of Prof. Peter Menell at 3 (lamenting that Cybor “discourage[s] settlements following claim construction and trial, delay[s] resolution of patent disputes, and run[s] up the overall costs of patent litigation”); Amicus Br. of Am. Intellectual Prop. Ass’n at 8 (“Cybor thus fosters wasteful, expensive litigation and discourages timely settlement. That result unnecessarily ties up courts and increases expense to litigants.”); Amicus Br. of Am. Bar Ass’n at 10-11 (observing that Cybor discourages settlement); Amicus Br. of Ass’n of Bar of N.Y. at 15-16 (same); Amicus Br. of Conn. Intellectual Prop. L. Ass’n at 13 (“Even when a case goes to trial, the losing party has very little incentive to settle disputes, since there is a significant chance that at least some material part of the trial court’s decision will be reversed on appeal.”); Amicus Br. of Fed’n Internationale Des Conseils en Pro-priete Intellectuelle at 11 (same); Amicus Br. of Paul R. Michel at 4 (same).7
*1314D.
Contrary to the majority’s claim, moreover, Cybor does not unqualifiedly promote uniformity or predictability of outcome in the patent system. As noted previously, the claim construction issues presented in patent cases are mostly fact and case specific. A claim construction decision in a given case will provide little guidance on the words used in different patents. Their resolution will do no more than declare the boundaries of a patent as between the parties in suit. See Blonder-Tongue Labs., Inc. v. Univ. of Ill. Found., 402 U.S. 313, 329-30, 91 S.Ct. 1434, 28 L.Ed.2d 788 (1971) (observing that “[s]ome litigants— those who never appeared in a prior action-may not be collaterally estopped without litigating [an] issue”). And, there is no guarantee that panels of this court will construe like claims in a like manner, even when in the same patent. Compare CVI/Beta Ventures, Inc. v. Custom Optical Frames, Inc., 92 F.3d 1203 (Fed.Cir.1996) (affirming the determination that “greater than 3% elasticity” did not require “complete recovery after a strain of greater than 3%” within the meaning of claim 1 of U.S. Patent No. 4,896,955), with CVI/Beta Ventures, Inc. v. Tura LP, 112 F.3d 1146, 1158 (Fed.Cir.1997) (construing “greater than 3% elasticity” in claim 1 of U.S. Patent No. 4,896,955 as requiring complete recovery after being subjected to stress); see also Kimberly A. Moore, Are District Court Judges Equipped to Resolve Patent Cases?, 15 Harv. J.L. & Tech. 1, 18-21 (2001) (observing that “[t]he CVI/Beta cases create doubt about whether the Federal Circuit serves as a test of ‘accuracy’ of district court construction”).
In fact, our case law expressly holds that we are not bound by claim constructions we adopt on appeal from the grant or denial of a preliminary injunction when considering the same claims again upon the final judgment.8 See Glaxo Grp. Ltd. *1315v. Apotex, Inc., 376 F.3d 1339, 1346 (Fed.Cir.2004) (“An appellate court’s preliminary injunction opinion has no conclusive bearing at the trial on the merits and is not binding on a subsequent panel.”) (citing Univ. of Tex. v. Camenisch, 451 U.S. 390, 395, 101 S.Ct. 1830, 68 L.Ed.2d 175 (1981)); Transonic Sys., Inc. v. Non-Invasive Med. Techs. Corp., 75 Fed.Appx. 765, 774 (Fed.Cir.2003) (“We have consistently followed the Supreme Court’s precedent by holding that a claim construction reached during an appeal from a grant of a preliminary injunction is tentative and is not binding on the district court in subsequent proceedings.”). We, thus, do not even have an internal structure that unerringly assures uniformity.
To the limited extent uniformity might be served by de novo review, moreover, any marginal benefit from that increased uniformity is more than offset by the decreased certainty caused by making district court decisions more vulnerable to reversal. See Kelly Casey Mullally, Legal (Un)Certainty, Legal Process, and Patent Law, 43 Loy. L.A. L.Rev. 1109, 1149-50 (2010) (examining how de novo review increases one kind of certainty at the cost of “makfing] district court judgments less certain” by “increasing] the probability that the lower court’s decision will be reversed”). And, as the PTO points out, “even if some marginal decrease occurred in this Court’s ability to ensure perfect uniformity in the interpretation of patent claims, that decrease would not provide a reason to ignore the clear mandate of Rule 52(a).” Amicus Br. of United States at 12.
We are not alone in the belief that Cy-bor does little to promote the uniformity with which the majority is now concerned. Indeed, “[i]n the government’s view, recalibrating the standard of review to reflect the trial court’s ‘institutional advantage’ in considering certain types of evidence in the claim-construction process, while preserving this Court’s ability to give de novo review to the trial court’s ultimate construction, would promote ‘in-terjurisdictional uniformity.’ ” Amicus Br. of United States at 12-13 (quoting Markman, 517 U.S. at 391, 116 S.Ct. 1384). As several amici explain, there are numerous other ways to improve uniformity of claim construction scope and interpretation, including improvements to the patent prosecution process, use of post-grant review procedures, or even consolidation of cases addressing the same patents before a single trial judge through the already well-established multidistrict litigation practice. See, e.g., Amicus Br. of Prof. Peter Menell at 22-24. And, as the American Bar Association notes, it is more likely that uniformity will be served by greater reliance on the claim construction decisions of the skilled fact finders— the district court judges — than by adhering to Cybor’s de novo standard of review. Amicus Br. at 13.
Our case law teaches that stare decisis is not an obstacle when our law causes such negative consequences. The recent decision by our en banc court in Therasense, Inc. v. Becton, Dickinson & Co., 649 F.3d 1276 (Fed.Cir.2011) (en banc), is a clear example of this court’s willingness to change our law where, as here, experience proves our past decisions were unwise. In Therasense, this court made drastic changes to the law with the aim of making claims of inequitable conduct more difficult to prove. See id. at 1290-91. In explaining why we did so, we noted that, over the years, we had “embraced ... reduced *1316standards for intent and materiality to foster full disclosure to the PTO.” Id. at 1288. But, “[tjhis focus on encouraging disclosure had numerous unforeseen and unintended consequences.” Id. Given the negative effects of our precedent, we wholly abrogated our decisions in Orthopedic Equipment Co. v. All Orthopedic Appliances, Inc., 707 F.2d 1376 (Fed.Cir.1983), Driscoll v. Cebalo, 731 F.2d 878 (Fed.Cir.1984), and American Hoist & Derrick Co. v. Sowa & Sons, Inc., 725 F.2d 1350 (Fed.Cir.1984), and required a greater showing to demonstrate inequitable conduct. Even the dissent in Therasense had no problem with abrogating our body of case law on inequitable conduct, disputing only what new test should be adopted in its stead. See Therasense, 649 F.3d at 1302 (Bryson, J., dissenting) (urging adoption of a new standard, but one that differed from that proposed by the majority).
Likewise, in Knorr-Bremse Systeme Fuer Nutzfahrzeuge GmbH v. Dana Corp., 383 F.3d 1337 (Fed.Cir.2004) (en banc), we overturned case law because we felt it had problematic effects. There, we observed that “implementation of [our prior] precedent has resulted in inappropriate burdens on the attorney-client relationship.” 383 F.3d at 1343. Looking at the full range of consequences flowing from our ease law, we concluded that “the conceptual underpinnings of this precedent have significantly diminished in force.” Id. at 1344 (citation and internal quotation marks omitted). So, we changed the law. Id. (“The adverse inference that an opinion was or would have been unfavorable, flowing from the infringer’s failure to obtain or produce an exculpatory opinion of counsel, is no longer warranted. Precedent authorizing such inference is overruled.”).
Thus, we have made clear that stare decisis does not prevent our court from changing our law where, as here, there are compelling reasons to do so.
VIL
In short, while Markman instructs us that claim construction presents a question for the court to resolve, it also instructs us that claim construction is a “mongrel practice,” presenting a mixed question of law and fact. While we agree that the ultimate question of claim meaning should remain subject to de novo review, claim construction often requires district courts to resolve underlying issues of disputed fact. These include, among others: whether a claim term had a specialized meaning among those skilled in the art at the time; what texts, including treatises and dictionaries, demonstrate about how a person of skill in the art would interpret a claim term, and which contemporaneous tests are most relevant; whether to credit one expert’s testimony over another’s regarding issues bearing on claim construction; who qualifies as a person of ordinary skill in the art; what is the relevant field of invention; what prior art is relevant; what a person of skill in the art would glean from that prior art; and what inferences can be fairly drawn from the prosecution history, including whether a disclaimer of claim scope has occurred.9 When a district court makes fact-findings needed to resolve claim construction dis*1317putes, Rule 52(a) requires us to defer to those findings unless they are clearly erroneous.
Cybor ignores both the realities of claim construction and Rule 52(a)’s demands. It is time we acknowledge the limitations of our appellate function and our obligation to comply with the Federal Rules of Civil Procedure, and give trial judges the deference their expertise and efforts deserve. Stare decisis is no bar to our doing so. Nor is concern about the fact that employing the proper standard of review in this context will not always be easy. For all these reasons, I respectfully dissent.
. We invited and received input regarding the standard of review to be applied to claim construction in Cybor itself, in Phillips v. AWH Corp., 415 F.3d 1303 (Fed.Cir.2005) (en banc), and now in this case.
. The majority describes three views espoused by the parties and amici, giving substantially more attention to the one that is consistent with the result the majority reaches. Careful review of the materials submitted to the court, and of the many academic and legal writings regarding Cybor since its issuance, show that a substantial portion of the legal community to have considered the issue believes Cybor was wrongly decided and flies in the face of Rule 52 of the Federal Rules of Civil Procedure.
. There are, of course, some common patent terms that have been given universal meanings, or been characterized as open-ended, rather than exclusive. These are terms like "comprising,” "consisting of,” and "consisting essentially of.” The meanings of most of *1302these transitional terms were common to the patent drafting art well before this Circuit was formed. And, litigants and district courts are well aware of these conventions.
. This mistake is one repeated in some of the amicus briefs that support retention of Cybor, stating that Cybor must be retained so as to avoid having to submit claim construction issues to the jury. See Amicus Br. of Microsoft Corp. at 4-5; Amicus Br. of Intellectual Prop. Inst, of William Mitchell Coll, of Law at 10-12. But, the Supreme Court made clear in Markman that it had institutional efficiency reasons for taking claim construction away from the jury, unhampered as it was by Seventh Amendment concerns; the decision to give claim construction to trial judges did not turn on a fact/law distinction. See Markman, 517 U.S. at 384 n. 10, 116 S.Ct. 1384 (“Because we conclude that our precedent supports classifying the question as one for the court, we need not decide either the extent to which the Seventh Amendment can be said to have crystallized a law/fact distinction ... or whether post-1791 precedent classifying an issue as one of fact would trigger the protections of the Seventh Amendment if (unlike this case) there were no more specific reason for decision.” (citations omitted)). Because the views of these amici are based on this legally flawed premise, undue reliance on them is misplaced.
. The scope of this court’s obligation to abide by the dictates of Rule 52(a)(6) is currently before the Supreme Court. In Highmark Inc. v. Allcare Health Management Systems, Inc., 687 F.3d 1300 (Fed.Cir.2012), cert. granted, - U.S. -, 134 S.Ct. 48, 186 L.Ed.2d 962 (2013), the question presented is "[w]hether a district court’s exceptional-case finding under 35 U.S.C. § 285, based on its judgment that a suit is objectively baseless, is entitled to deference.” Petition for Writ of Certiorari, Highmark Inc. v. Allcare Health Mgmt. Sys., Inc., 2013 WL 1209137, at *i (U.S. Mar. 25, 2013) (No. 12-1163). The petitioner asserts that deference must be given to all aspects of a district court’s § 285 determinations because, among other reasons, there are subsidiary findings of fact which Rule 52(a)(6) demands be reviewed for clear error. See id. at *19-20. If the Supreme Court premises its holding in High-mark entirely or even partially on the dictates of Rule 52(a)(6), such a ruling would make clinging to Cybor for no reason other than a resistance to change completely untenable.
. The majority is incorrect that “every amicus brief that complains about high reversal rates relies on data that are seven to ten or more years old.” Maj. Op. at 1290. Professor Me-nell’s amicus brief to this court describes his recent research with Professor Jonas Anderson, which reveals that de novo review of claim construction continues to contribute to “alarming levels of appellate reversals.” Amicus Br. at 13-14; see also Anderson & Menell, supra, at *6 (examining this court’s claim construction jurisprudence from 2000 through 2011).
. The majority cites data showing that a declining percentage of cases proceed to trial or are appealed. See Maj. Op. at 1290-92. According to the majority, these trends show *1314that “the Cybor review procedure assists in resolving litigation before full trial or extensive discovery,” thereby facilitating settlement and reducing litigation costs. Id. at 1291-92. Nothing suggests that these declines can be attributed to this court’s de novo review of claim construction, however. Declining trial and appeal rates can easily be attributable to other factors, including (1) the availability of parallel proceedings at the PTO and ITC where decisions in those tribunals might moot further activity before the district courts, or even prevent district court judgments from becoming final, (2) increased resort to and availability of sophisticated alternative dispute resolution mechanisms, including the increased involvement of retired district court judges with patent litigation experience in such procedures, (3) improved case management practices by trial judges who have become more practiced at handling patent litigation and who now often have the benefit of detailed local rules governing the same, (4) the fact that, once this court provided clear guidance regarding claim construction in Phillips, trial courts were given a better road-map for undertaking the exercise of claim construction, and (5) the increased experience and expertise of trial courts that itself may be fostering settlements. The majority reads far too much regarding the wisdom of Cybor into these general statistics.
. It is curious that, when reserving the right to change our own claim constructions at later points in a single case, we justify that position on grounds that the greater fulsomeness of the record at the final judgment stage better informs our claim construction analysis. Transonic Sys., Inc. v. Non-Invasive Med. Techs. Corp., 75 Fed.Appx. 765, 774 (Fed.Cir. 2003) ("A district court therefore is at liberty to change the construction of a claim term as the record in a case evolves after a preliminary injunction appeal.”); Jack Guttman, Inc. v. Kopykake Enters., Inc., 302 F.3d 1352, 1361 (Fed.Cir.2002) ("District courts may engage in a rolling claim construction, in which the court revisits and alters its interpretation of the terms as its understanding of the technology evolves. This is particularly true where the issues involved are complex, either due to the nature of the technology or because the meaning of the claims is unclear from the intrinsic evidence.” (citation omitted)). If the *1315trial record is effectively meaningless to the claim construction inquiry as we now hold, what more could we know about claim construction later in a case than we knew when we first visited it?
. Notably, a district court’s factual determinations, even those about the historical meaning of a claim term, will not resolve the legal question of what construction is to be afforded a claim term. This court would be free to conclude that a claim term has a different meaning than its historically common one based on the four corners of the patent itself, or on application of legal doctrines applicable to claim construction such as claim differentiation, meanings we have subscribed to common terms (e.g., "comprising”), or the concept of an inventor being permitted to act as his own lexicographer.