with whom PAULINE NEWMAN, Circuit Judge, joins, concurring in the judgment.
I am compelled to concur in the judgment of the court, but I respectfully disagree with the opinion because it profoundly misapprehends Markman v. Westview Instruments, Inc., 517 U.S. 370, 116 S.Ct. 1384, 134 L.Ed.2d 577 (1996). The Supreme Court concluded there that the historical record is insufficiently firm to declare that juries construed patent claims in England when the Seventh Amendment to the Constitution was *1464adopted in 1791. So it decided as a matter of policy that judges, not juries, are better able to perform this task given the complexity of evidence and documentation. This was a perilous decision of last resort. For juries regularly render verdicts in civil eases based on complex forensic and documentary evidence of equal or greater difficulty than seen in patent cases. And the implications for criminal cases under Article III and the Sixth Amendment are even more profound. Increasingly complex cases involving scientific and complicated documentary evidence are • presented to criminal juries which, of course, decide matters of life and liberty, not merely money. Nevertheless, having so ruled, it seems to me the Court would not also have repealed part of the Federal Rules of Civil Procedure and Evidence without so much as a mention that district courts no longer have discretion to admit expert evidence, see Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993), and need not find facts when evidence is disputed in these cases. Nor would it have so excused this court from the normal, historical role of appellate courts to review for reversible error, and installed it as a collegial trial court.
I. STANDARD OF REVIEW
We review the denial of a motion for judgment as a matter of law de novo by reapplying the same standard. See Read Corp. v. Portec, Inc., 970 F.2d 816, 821, 23 USPQ2d 1426, 1431 (Fed.Cir.1992). Under this standard, appellant must show that “the jury’s factual findings, presumed or express, are not supported by substantial evidence or, if they are, that the legal conclusions implied from the jury’s verdict cannot in law be supported by those findings.” Alpex Computer Corp. v. Nintendo Co., 102 F.3d 1214, 1221, 40 USPQ2d 1667, 1672 (Fed.Cir.1996) (quoting Kearns v. Chrysler Corp., 32 F.3d 1541, 1547-48, 31 USPQ2d 1746, 1751 (Fed. Cir.1994)).
Our review of claim construction is controlled by the Supreme Court’s judgment in Markman, 517 U.S. 370, 116 S.Ct. 1384, not the opinion of this court it reviewed, 52 F.3d 967, 34 USPQ2d 1321 (Fed.Cir.1995); the Supreme Court did not adopt this court’s reasoning as its own. 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.1 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 eviden-tiary underpinnings or whether the importance of uniformity is best served by giving these evidentiary questions of meaning to a judge. It would have been a simple matter for the Court to give short shrift to this argument by proclaiming construction purely, solely, and always a matter of law that would never have gone to the jury.
The Supreme Court recognized that in some cases there will be conflicting evidence that has to be resolved — where there are factual determinations that are more than just incident to claim construction — such as the understanding of one skilled in the art at the time the patent application was filed. In these cases, all that Markman stands for is that the judge will do the resolving, not the jury. Wisely, the Supreme Court stopped short of authorizing us to find facts de novo when evidentiary disputes exist as part of the construction of a patent claim and the district court has made these findings without committing clear error. See Fromson v. Anitec Printing Plates, Inc., 132 F.3d 1437, 1446, 45 USPQ2d 1269, 1275 (Fed.Cir.1998) (findings of fact made by a district court to resolve disputes as to the meaning of claim terms are affirmed absent clear error); Eastman Ko*1465dak Co. v. Goodyear Tire & Rubber Co., 114 F.3d 1547, 1555-56, 42 USPQ2d 1737, 1742 (Fed.Cir.1997) (reliance on expert testimony to clarify ambiguous claim term is acceptable and deference is given to the trial court’s credibility determination). Rather, when the judge finds facts or accepts the factual determinations of a jury, those facts are entitled to greater deference than de novo fact finding on appeal.
In Fromson, the trial judge admitted expert testimony to develop the record on the meaning of the word “anodized,” as it was understood by one skilled in the art at the time of the invention. Fromson, 132 F.3d at 1444-45, 45 USPQ2d at 1273-74. Based on that extrinsic evidence, he found “in 1973 no reasonable practitioner of this process would have had the opinion that a non-porous non-adherent oxide coating, as thin as the native 5 nanometer coating found naturally in the environment, of phosphoric oxide, constituted an anodized surface.” Id. at 1444, 45 USPQ2d at 1274. This finding limited the meaning of the word anodized beyond what a standard definition required: “[T]o subject [a metal] to action by making [it] the anode of a cell before coating with a protective or decorative film.” Id., 132 F.3d 1437, 45 USPQ2d at 1273 (quoting Webster’s Third Int’l Dictionary). This finding also conflicted with the alleged infringer’s description of its own oxide coating process as anodization. Nevertheless, the district court construed the word anodized to mean: “an electrolytically formed, adherent, porous aluminum oxide coating sufficiently thick (meaning thicker than native oxide)____” Id. at 1445, 45 USPQ2d at 1274. Because the accused process did not form an oxide coating thicker than native oxide, the district court entered a judgment of noninfringement.
Even had we disagreed with the construction given by the trial court in Fromson, we were not tasked by our standard of review to reexamine de novo each of the days of testimony and volumes of record relating to interpretation of the word anodized. This court did not affirm the trial court’s judgment because it had educed the best construction of the word anodized. It did so because the district court’s construction was properly predicated on a factual finding about what anodized meant to one skilled in the art in 1973, which on the appellate record presented by the parties was not clearly erroneous. The Supreme Court never contemplated in Markman that we feign first-hand experience with the technology, or that we embellish our abilities by construing a claim without the respect due both a trial court’s decision that a factual dispute underlies the meaning of a claim term and its resolution of that dispute.
Provided that no factual findings about disputed terms were necessarily made in the course of construing the claim, our making a claim construction that supersedes that of the district court is not inconsistent with the Court’s opinion. See Serrano v. Telular Corp., 111 F.3d 1578, 42 USPQ2d 1538 (Fed. Cir.1997) (construction of the claims involved no factual disputed between the parties); International Communication Materials, Inc. v. Ricoh Co., 108 F.3d 316, 318-19, 41 USPQ2d 1957, 1958-59 (Fed.Cir.1997); Alpex, 102 F.3d 1214 (using Alpex’s expert testimony against Alpex, there were no factual dispute between the parties); Metaullics Sys. Co. v. Cooper, 100 F.3d 938, 939, 40 USPQ2d 1798, 1799 (Fed.Cir.1996); Insituform Techs., Inc. v. Cat Contracting, Inc., 99 F.3d 1098, 40 USPQ2d 1602 (Fed.Cir.1996) (attorney argument regarding claim construction presented no factual dispute about the technology); 1 S. Childress & M. Davis, Federal Standards of Review: Civil Cases and General Review Principles, §§ 2.13, 2.14 (2d ed.1991). If this court does not believe the claim construction to be erroneous based on an independent review of the legal conclusions and a review of the constituent factual findings for substantial evidence, it must affirm. See Bose Corp. v. Consumers Union of United States, 466 U.S. 485, 514 n. 31, 104 S.Ct. 1949, 1967 n. 31, 80 L.Ed.2d 502 (1984); see also Pullman-Standard v. Swint, 456 U.S. 273, 287-88, 102 S.Ct. 1781, 1789-90, 72 L.Ed.2d 66 (1982).
This court may not, however, independently review the constituent facts or disregard the jury’s findings, absent proof that they lack sufficient evidence that “a. reasonable mind might accept as adequate to support” them. Consolidated Edison Co. v. NLRB, 305 U.S. 197, 229, 59 S.Ct. 206, 217, 83 L.Ed. *1466126 (1938); see Genentech, Inc. v. Wellcome Found. Ltd., 29 F.3d 1555, 1565, 31 USPQ2d 1161, 1169 (Fed.Cir.1994). Nor may we construe a claim in a manner that is inconsistent with our function as a court of review; we cannot divine new interpretations of terms in a claim or make. findings of fact from a record that cannot support them. See Avia Group Int'l Inc. v. L.A. Gear California, Inc., 853 F.2d 1557, 1561, 7 USPQ2d 1548, 1551 (Fed.Cir.1988).
Sometimes it may be necessary for an appellate court to pronounce new legal principles during an appeal. However, even if this court identifies a question of claim construction as one of law — though its resolution is relevant to only the particular litigation or document — it cannot elevate the activity to one of determining legal principles, as is for example, statutory construction. See Markman, 517 U.S. at 391, 116 S.Ct. at 1396 (“[I]ssue preclusion could not be asserted against new and independent infringement defendants even within a given jurisdiction.-...”); In re Freeman, 30 F.3d 1459, 1466, 31 USPQ2d 1444, 1450 (Fed.Cir.1994); Jackson Jordan, Inc. v. Plasser American Corp., 747 F.2d 1567, 1574-75, 224 USPQ 1, 5-6 (Fed.Cir.1984) (second alleged infringer not bound by prior claim constrúction unless it had, inter alia, a full and fair opportunity to litigate the construction in the first infringement action). Thus, regardless of the labels we attach to these questions, without the benefit of a full record from the trial court, it is neither the function of this court nor is it within our capacity as an appellate court to adopt new interpretations, most especially not if the interpretation. under review is of terms informed by conflicting evidence. Such indiscriminate and conclusive review deprives the parties of important substantive and procedural mechanisms provided in the trial courts, where interpretation can be informed by additional discovery and expert testimony, and where it can be checked by appellate review as a matter of right. If claim construction is only a question of law to be -reviewed by this court de novo, then the absence of review as a matter of right over our claim constructions, which may be new and unsupported by legal analysis, or may never have been tested by the adversarial process, would transform this court into a trial court of first and usually last resort.
II. METHOD OF REVIEW
To determine whether á patent has been infringed, a district judge must partake of a two-step analysis: (1) construction of the claims to determine their legal effect .by examining and resolving, inter alia, factual disputes over the meaning and scope of technical words or terms of art used in the patent, see Fromson, 132 F.3d at 1442, 45 USPQ2d at 1272; Markman, 52 F.3d at 999, 34 USPQ2d at 1347 (Newman, J., dissenting); and (2) a comparison of the properly construed claims to the accused device, see Carroll Touch, Inc. v. Electro Mechanical Sys., Inc., 15 F.3d 1573, 1576, 27 USPQ2d 1836, 1839 (Fed.Cir.1993). The first step, construction of the claim, is ultimately a question for the judge. See Markman, 517 U.S. at 388, 116 S.Ct. at 1395 (“So it turns out here, for judges, not juries, are the better suited to find the acquired meaning of patent terms.”) In construing the claims, a judge looks to the claims themselves, the specification, the prosecution history of the patent, see Minnesota Mining and Mfg. Co., v. Johnson & Johnson Orthopaedics, Inc., 976 F.2d 1559, 1566, 24 USPQ2d 1321, 1327 (Fed. Cir.1992), and, if necessary, extrinsic evidence for information or to resolve disputes over the meaning of terms, see Texas Instruments, Inc. v. United States Int’l Trade Comm’n, 988 F.2d 1165, 1171, 26 USPQ2d 1018, 1023 (Fed.Cir.1993). Although Markman, 517 Ú.S. 370, 116 S.Ct. 1384, requires the judge to make the ultimate determination about the meaning and legal consequence of a claim, it does not prohibit him from submitting subsidiary factual disputes to a jury, even if it does not encourage it.
The judge follows the same first step to construe claims containing means-plus-function limitations under 35 U.S.C. § 112(6) (1994). See Alpex, 102 F.3d at 1220, 40 USPQ2d at 1672; Intellicall, Inc. v. Phonometrics, Inc., 952 F.2d 1384, 1388, 21 USPQ2d 1383, 1387 (Fed.Cir.1992). However, to complete the construction of these means-plus-function terms, the judge must look to the structures, materials, or acts *1467disclosed in the patent’s specifications and to their equivalents. Id., 952 F.2d at 1388, 21 USPQ2d at 1387. To determine the scope of such equivalents, the district court must resolve questions of fact by resorting to the expertise of the fact finder. Cf. Markman, 52 F.3d at 977 n. 8, 34 USPQ2d at 1337, n. 8 (expressly declining to reach “the issue of whether a determination of equivalents under § 112, para. 6 is a question of law or fact”); In re Hayes Microcomputer Prods., Inc., 982 F.2d 1527, 1541-43, 25 USPQ2d 1241, 1253 (Fed.Cir.1992) (“This list [of factors that may be considered when determining the scope of a means-plus-function limitation] is not exhaustive and reasonable inferences by the fact finder are appropriate.”); King Instrument Corp. v. Otari Corp., 767 F.2d 853, 862-63, 226 USPQ 402, 408-09 (Fed.Cir. 1985); Palumbo v. Don-Joy Co., 762 F.2d 969, 975-76, 226 USPQ 5, 8 (Fed.Cir.1985) (“Whether that accused device is a § 112 equivalent ... is a question of fact.”); D.M.I., Inc. v. Deere & Co., 755 F.2d 1570, 1575, 225 USPQ 236, 239 (Fed.Cir.1985). For pragmatic reasons, the resolution of this factual determination is often made at the same time the fact finder determines infringement (step two, described below).
The second step of the infringement analysis requires a factual comparison of the claimed invention to. the accused device, which is done by the fact finder. See Winans v. Denmead, 56 U.S. (15 How.) 330, 338, 14 L.Ed. 717 (1853). To prove literal infringement, the patentee must show that the accused device contains every limitation in the asserted claims. See Dolly, Inc. v. Spalding & Evenflo Cos., 16 F.3d 394, 397, 29 USPQ2d 1767, 1769 (Fed.Cir.1994). If tried to a jury, it is the jury’s factual findings on infringement that are reviewed by this court for lack of substantial evidence, see Genentech, 29 F.3d at 1565, 31 USPQ2d at 1168-69, as part of this court’s reapplication of the standard for judgment as a matter of law.
A claim of infringement under the doctrine of equivalents modifies this second step by requiring that the fact finder determine whether differences between particular elements of the accused device and the asserted claims are insubstantial. See Wamer-Jen-kinson Co. v. Hilton Davis Chem. Co., — U.S. -, -, 117 S.Ct. 1040, 1054, 137 L.Ed.2d 146 (1997). However, because they have separate origins, purposes, and applications, determining equivalence under paragraph 112(6) requires an analysis different from that used to determine equivalence under the doctrine of equivalents. See Alpex, 102 F.3d at 1222, 40 USPQ2d at 1673 (“Under § 112, the concern is whether the accused device, which performs the claimed function, has the same or an equivalent structure as the structure described in the specification corresponding to the claim’s means. Under the doctrine of equivalents, on the other hand, the question is whether the accused device is only insubstantially different than the claimed device.”) (internal citations omitted).
After the judge construes the means-plus-funetion limitations identifying structures, materials, or acts described in the patent’s specification, and their equivalents as determined by the fact finder (step one, described above), the judge gives the construed claims to the fact finder, in this case a jury, for a determination of infringement. See D.M.I., 755 F.2d at 1575, 225 USPQ at 239. For literal infringement, the fact finder must determine whether the accused device performs an identical function to the one recited in the means-plus-function clause. See Pennwalt Corp. v. Durand-Wayland, Inc., 833 F.2d 931, 934, 4 USPQ2d 1737, 1739 (Fed.Cir, 1987) (if the identical function is not performed, literal infringement is not possible). If the identical function is performed, the fact finder must then determine whether the accused device utilizes the same structure or materials as described in the specification, or their equivalents.
Just as the fact finder’s infringement analysis differs between equivalence under paragraph 112(6) and the doctrine of equivalents, so too differs the analytical effect of statements made during the prosecution of the patent on construction of the claims. Under paragraph 112(6), a statement made during prosecution may confine the range of equivalent structures, materials, or acts that are directly claimed by the patent. However, in the context of a doctrine of equivalents analysis, the patentee seeks protection beyond *1468that claimed by the patent directly. As such, the judge’s construction of the claims — which includes the interpretation- of claim terms— may not be sufficient to remove from the jury’s consideration all subject matter that was disclaimed during prosecution.
Prosecution history estoppel addresses this problem by excluding equivalents surrendered during prosecution. Under this doctrine, statements made to overcome rejections based, as here, on prior art estop the patentee from extending its right to exclude others from-making, using, or selling subject matter known to be insubstantially different from, or interchangeable with, claimed elements at the time of the alleged infringement. Warner-Jenkinson, — U.S. at-, 117 S.Ct. at 1053. Although both forms of equivalence require the district court to examine the prosecution history as part of its construction of the claims, under the doctrine of equivalents, the judge gives the claim, properly construed to exclude disclaimed subject matter, to the jury and then, where appropriate, also instructs the jury on the possible range of equivalents that it may or may not consider due to prosecution history estoppel.
III. REVIEW
As depicted in figure two of the patent, ante at 1452 [maj. opn], the ’837 patent discloses a dual-stage pump. It claims a first pumping means that pumps fluid through a filtering means “to” a “second pumping means.” It also claims a “means to enable said second pumping means to collect and/or dispense the fluid, or both,” at rates or during periods that are independent of the operation of the first pumping means. As a result of this structure, the pump described in the ’837 patent is able to accumulate fluid for later dispense, dispense fluid immediately, or partially accumulate and partially dispense fluid in precise measurements and at precise times. The preferred embodiment describes a second pumping means where the second pump collects and dispenses the fluid from its own internal reservoir.
Like the pump disclosed in the ’837 patent, Cybor Corporation’s Model 5226 dual-stage pump also allows users to pump or dispense highly accurate amounts of liquid chemicals in increments as small as 0.1 microliter. Both pumps are primarily used for accurately dispensing chemicals — such as photoresist and polimide — onto semiconductor wafers during their fabrication.
Of relevance to this appeal, , in construing the ’837 patent the district judge interpreted the disputed term “to” and submitted the term to the jury for a determination of infringement. The judge also submitted the terms “second pumping means,” “means to enable said second pumping means,” and “or both” to .the jury as means-plus-function terms, along with structures in the ’837 patent’s specification, for a determination of infringement. The judge left the task of considering prosecution history estoppel to the jury.2 The jury found that Cybor’s Model 5226 pump literally infringed claims 1-10, 13-15, and 17-20, and that it infringed claims 11, 12, and 16 under the doctrine of equivalents. It is unclear which combination of components in Cybor’s Model'5226 pump the jury found to infringe claims 1-10,13-15, and 17-20. It may have interpreted “second pumping means” in such a way that it found either Cybor’s second pump or the combination of its second pump and its external reservoir to be equivalent structures to those provided in the specification of the ’837 patent corresponding to the “second pumping means.” Likewise, the jury may have interpreted “means to enable said second pumping means” in such a way that it found either Cybor’s external reservoir or the combination of its external reservoir and its second pump to be equivalent structures to those provided in the specification of the ’837 patent corresponding to the “means to enable said second pumping means.” In total, there are five possible combinations of claim term interpretations and findings of paragraph 112(6) equivalents that would have led the jury to its infringement verdict.
*1469Cybor renewed its earlier motion for judgment as a matter of law pursuant to Rule 50(b) and for a new trial pursuant to Rule 59 of the Federal Rules of Civil Procedure, on the basis that no reasonable juror could find that its pump infringes the ’837 patent. The court denied the motion, finding “that the jury’s infringement verdict is supported by the evidence ... the jury instructions were proper and the verdict was not against the clear weight of the evidence or. contrary to the instructions.” Cybor moved for reconsideration, arguing that this court’s then recently issued Markman, 52 F.3d 967, 34 USPQ2d 1321, required that the judge, and not the jury, interpret all the claims. The district court denied the motion and entered final judgment, from which Cybor appealed.
On appeal, Cybor argues that had the district court properly analyzed the prosecution history of the ’837 patent, it would have excluded reservoir structures like Cybor’s from consideration as an equivalent under paragraph 112(6). Cybor further argues that the judge would have instructed the jury not to consider its reservoir when determining whether Cybor’s second pump has the same or an equivalent structure as the structure in the portions of the ’837 patent’s specification corresponding to the second pumping means. Without its reservoir, Cybor argues that its second pump cannot infringe the “second pumping means” limitation, the “or both” functional limitation, or the filter means “to” second pumping means limitation. Cybor also argues that the district judge erred in failing to identify the corresponding structures in the specification, and in permitting the judge to consider the effect of statements made during prosecution of the ’837 patent.
FAS responds that the district court properly considered and rejected Cybor’s argument about the prosecution history of the ’837 patent, that it properly construed the necessary claims, and that it properly submitted them along with relevant corresponding structures — as described in the specification — to the jury for a determination of infringement. FAS further argues that the district court was not obliged/ either before or after Markman, to submit complete claim constructions to the jury so long as the judge’s ultimate claim construction is supported by the jury’s factual findings, presumed or expressed, regarding disputed claim terms, and so long as the judge later evaluates and sustains the jury’s factual findings.
Claim 1 of the ’837 patent contains three disputed limitations and is representative of how these limitations are used, where they appear throughout the 20 asserted claims:
In a device for filtering and dispensing fluid in a precisely controlled manner, the combination of:
first pumping means;
second pumping means in fluid communication with said first pumping means; and
filtering means between said first and second pumping means, whereby said first pumping means pumps the fluid through said filtering means to said second pumping means;
in which each of said first and second pumping means includes surfaces that contact the fluid, said surfaces being of materials that are non-contaminating to industrial fluids which are viscous and/or high purity and/or sensitive to molecular shear; and
comprising, means to enable said second pumping means to collect and/or dispense the fluid, or both, at rate or during periods of operation, or both which are independent of rates or periods of operation, or both, respectively, of said first pumping means.
‘837 patent, col. 9, lines 45-61 (emphasis and formatting added). The specification of the’ 837 patent further describes the “second pumping means” as having tubing that connects the pump to the second incremental pump advancement means, col. 5, lines 49-52, a diaphragm inside the second pump, col. 6, lines 2-3, and a separate inlet passage and outlet passage, col. 7, lines 10-20.
During prosecution, the inventors further limited the scope by making the following statements to overcome a prior art rejection that the claimed invention was unpatentable over Storkebaum et al., U.S. Patent No. 4,749,476.
Additionally, Storkebaum specifically provides a separate container 12 for collecting *1470the permeate. Obviously, Storkebaum does not- teach the collection of fluid in a second pumping means____ Storkebaum discloses a permeate collecting container 12 that is separate from the conveying pump 13. Nothing in Storkebaum discloses or makes obvious the claimed invention, nor the precise and flexible control provided by the second pump means of Claim 1.
Given the nature and function of the Storkeb-aum container, these statements were made to distinguish the - invention over the prior art. They confine the possible interpretation of terms in the claims under paragraph 112(6) and they may prevent FAS’s desired use of the doctrine of equivalents. See Warner-Jenkinson, — U.S. at- n. 7, 117 S.Ct. at 1051 n. 7. However, in this case the limitation of possible equivalents extends only so far as was necessary to distinguish aspects of the invention’s “second pumping means” from Storkebaum’s external container. The statements do not require that the district judge provide the jury with more specific interpretations of claim terms because the inventors disclaimed the use of a separate, physically unattached reservoir that cannot achieve differential pumping rates between the two pumps by accumulating fluid in a reservoir — the function performed by the ’837 patent’s “means to enable said second pumping means.”
Storkebaum’s separate container collects permeate and vents the fluid; it is designed to feed a conveying pump that is part of a big circulation loop, not a fluid dispensing device. Because the inventors never disclaimed an external reservoir that accumulates fluid to permit different pumping rates and because equivalence under paragraph 112(6) is a question of fact, the district court properly permitted the jury to consider the structure of Cybor’s external reservoir as a possible equivalent, under paragraph 112(6), to structures in the ’837 patent’s specification that define the “second pumping means” and the “means to enable the second pumping means.” It was also proper for the district court to permit the jury to consider Cybor’s reservoir in measuring the substantiality of differences between its device and the invention claimed by the ’837 patent, under the doctrine of equivalents.
A. “Second Pumping Means”
“Means to Enable Said Second Pumping Means”
As to these two disputed limitations, each of the independent claims, except for claim 16, uses the term “second pumping means,” which the court interpreted in means-plus-function language as: “a structure identical to the structure disclosed in the specification of the patent or the equivalent of that structure which performs the function of a fluid aceumulator/dispense pump.” The district court also interpreted the term “means to enable said second pumping means,” which appears in each of the independent claims except claims 10, 16, and 17, in means-plus-function language, as: “a structure identical to the structure disclosed in the specification of the patent, or the equivalent thereof, which allows the second pumping means to [accumulate, dispense, and partially collect and dispense fluid at rates and during periods independent of the first pumping means].”
Cybor argues that these means-plus-function elements- should be limited to the exact structures disclosed in the claims and the specification of the' ’837 patent. However, paragraph 112(6) permits the disclosure of specific structures in the specification, without limiting protection to the disclosed structures. See 35 U.S.C. ¶ 112(6) (“[S]uch claims shall be construed to cover the corresponding structure, materials, or acts described in the specification and equivalents thereto.”); D.M.I., 755 F.2d at 1574, 225 USPQ at 238. Cybor is correct that the reservoir and the second pump in its dual-stage pump are not identical to the structure disclosed in the specification of the ’837 patent; they contain a valve unlike the disclosed structure, and if uncoupled from the attached reservoir, the pump does not permit the partial accumulation and partial dispensing of fluid as contemplated by the term “flow through” in the ’837 patent. However, as explained above, the prosecution history does not require the district court to uncouple the valve or separate the reservoir from the second pump before submitting the terms “second pumping means” and “means to enable said second pumping means” to the jury. *1471The ’837 patent’s preferred embodiment shows the reservoir inside the second pump, but nothing requires such a restriction, and limitations in the specification should not be read into the claims. See Constant v. Advanced Micro-Devices, Inc., 848 F.2d 1560, 1570-71, 7 USPQ2d 1057, 1064 (Fed.Cir. 1988). Finally, such a restriction would go against the clear weight of testimony submitted by FAS’s expert witnesses, Clark, who was qualified as one skilled in the art of fluid dynamics, and Snodgrass and Gibson.
“As in all cases involving assertions of equivalency, wherein the patentee seeks to apply its claims to structures not disclosed by the patentee, the court is required to exercise judgment.” Texas Instruments, Inc. v. United States Int’l. Trade Comm’n, 846 F.2d 1369, 1371, 6 USPQ2d 1886, 1889 (Fed.Cir. 1988). Here, the district court exercised its judgment, made a partial, though legally sufficient interpretation of these limitations, and presented them to the jury for resolution of factual disputes on the way to determining equivalents under paragraph 112(6). Based on these interpretations and the jury’s verdict, we presume3 that the jury found that the combination of one of Cybor’s Model 5016 pumps with the external reservoir it uses in its Model 5226 pump created a structural equivalent to the “second pumpiing means” or the “means to enable said second pumping means” claimed in the ’837 patent and described by the specification under paragraph 112(6). Because we have not been shown that the jury lacked substantial evidence to support this finding of equivalents or infringement, we must affirm.
However, if the court is correct that claim construction is purely and solely a question of law to be reviewed de novo on appeal, it could not affirm the judgment of infringement. Infringement was determined in this case by a jury at the same time that it was asked to construe these claims by resolving factual disputes about their meaning and to determine whether the scope of these claims should be limited by arguments made during prosecution of the patent. Ordinarily, a jury’s verdict of infringement permits us to presume the existence of factual findings necessary to that verdict. But we are in no way entitled to presume factual findings or even to rely on the verdict of infringement where neither the jury, nor the judge on a motion for judgment as a matter of law, articulated a construction of the claims upon which the findings are premised. As explained above, there are five possible combinations of claim term interpretations and findings of paragraph 112(6) equivalents that could have led the jury to its infringement verdict. In light of the possibilities, it is difficult to imagine how the court can affirm the jury’s finding of infringement based upon its own de novo claim construction, which merely presumes a claim construction by the district churt and that admits of no deference to the findings of fact.
The effect of allowing such a combination of presumptions in this case has been to deny Cybor meaningful review of anything more than the district judge’s decision not to apply prosecution history estoppel, since it was the only legal determination that can be located outside of those presumptions. Not surprisingly, this is the decision Cybor contests most strenuously on appeal. The effect has also been to allow this court to presume that it made the same claim construction as the jury, without so much as articulating its own ■construction of these two claim terms, and then to presume that the jury’s infringement finding is still valid. Penultimately, Cybor is left with little more than the ability to make presumptions as to our own claim interpretations in any motion it may file for reconsideration. Worse still is the message the court is sending to district courts: we will affirm a finding of infringement as long as the district court avoids articulating its construction of the claims — which coincidentally leaves us free to arrive at our own constructions — and as long as we can presume that the jury used *1472our constructions to arrive at that finding of infringement.
B. “To”
The second of these three disputed claim limitations, “pumps fluid through said filtering means to said second pumping means,” is the only one that the district court did not analyze as part of a means-plus-function limitation. Cybor argued that the fluid has to be pumped directly to the second pump. FAS maintains that it is enough that the fluid merely reaches the second pump. The district court interpreted the term “to” as pumping “towards and with a destination of’ the second pump and left the determination of infringement to the jury. Absent its prosecution history estoppel argument, there remains little to Cybor’s interpretation of this claim limitation. FAS presented sufficient expert testimony from Gibson and Clark that the fluid must merely be pumped “ultimately” to the “second pumping means.” This evidence is consistent with the expert testimony of Snodgrass and Gibson that the “to” limitation requires only a designed or dedicated pathway between the two pumping means, which Cybor’s device possesses. Because all experts were consistent in their testimony about the meaning of terms to one skilled in the art, there was no occasion here for the court to resolve a factual dispute over the meaning of a term en route to their interpretation. Moreover, because the interpretation of this claim limitation describes Cybor’s Model 5226 dual-stage pump, regardless of whether or not the reservoir is part of the second pumping means, no reasonable juror could use this as a basis for finding noninfringement, and I agree we can affirm this aspect of the district court’s judgment.
C. “Or Both”
The third disputed limitation, “or both” is contained within means-plus-function language, it describes a function of the “means to enable second pumping means,” and it must be a function of Cybor’s device if the fact finder is to consider equivalence under paragraph 112(6). Cybor argues solely that the inventor of the ’837 patent disclaimed external reservoirs such as its own, by distinguishing Storkebaum. Thus the jury should not consider its external reservoir when it determines whether the Model 5226 dual-stage pump contains structural equivalents to each of the relevant limitations disclosed in the ’837 patent. Given the prosecution history analysis and the jury’s presumed finding that Cybor’s reservoir and second pump contain equivalents of all the limitations to the “means to enable said second pumping means,” Cybor’s argument cannot stand. The jury had substantial evidence upon which to find that Cybor’s reservoir and second pump together can collect, dispense, and both collect and dispense fluid, so its finding of infringement is also sufficient.
IV. JUDGMENT AS A MATTER OF LAW
Finally, Cybor argues that Markman, 52 F.3d 967, 34 USPQ2d 1321, requires that the judge determine the meaning of the claim before submitting it to the jury. However, that opinion said that the judge must determine the meaning of a claim, but that the meaning need not be established before the claims are submitted to the jury. Markman, 52 F.3d at 981, 34 USPQ2d at 1331 (pronouncing the meaning of a claim “ordinarily can be accomplished by the court in framing its charge to the jury, but may also be done in the context of dispositive motions such as those seeking judgment as a matter of law.”) The risk of waiting, of course, is that the jury’s deliberative efforts on infringement may be rendered useless if the instruction was improper and the affected factual dispute was material to the verdict. Here, this happens only under the court’s standard of review. Moreover, given the nature and interrelatedness of these disputed limitations and the underlying factual questions — whether the combination of Cybor’s second pump and its reservoir is a paragraph 112(6) equivalent to the “second pumping means” or the “means to enable said second pumping means” — it is difficult to see how the district court could have better instructed the jury, absent the use of detailed and alternative special verdicts. Therefore, I see no reversible error in the district court’s limited claim interpretations or its denial of Cybor’s motion for judgment as a matter of law.
. Rather than bluntly force the square peg of claim construction into the round hole of fact or law, the Court described the questions presented by claim construction in more chary terms: "[b]ut the sounder course, when available, is to classify a mongrel practice," "no clear answers,” "in theory there could be a case," "leaves us doubtful," "fi]n the main," "[w]e accordingly think there is sufficient reason to treat construction of terms of art like many other responsibilities that we ceded to a judge,” "independent reason to allocate,” "treating interpretive issues as purely legal will promote____” Markman, 517 U.S. at 377-91, 116 S.Ct. at 1390-96 (emphasis added). Even a cursory reading of that opinion indicates that the Court meant to determine who should interpret the claims, without mandating a standard of appellate review to be .used under all circumstances.
. Although having a juiy make a legal determination of the effect of statements made - during prosecution of the patent is error, alone this error was harmless because to find infringement, the jury must have arrived at the same legal conclusion that we did on appeal: that the prosecution history does not exclude Cybor’s external reservoir from the jury’s consideration of equivalents under paragraph 112(6).
. Findings of fact are presumed where they would he necessary to support a jury's verdict. See Perkin-Elmer Corp. v. Computervision Corp., 732 F.2d 888, 893, 221 USPQ 669, 673 (Fed.Cir. 1984) (“To facilitate review on a motion for [judgment as a matter of law] and on appeal, it is preferred that a jury be provided with special interrogatories designed to reveal more clearly the findings it made. Absent such interrogatories, the law presumes the existence of findings necessary to support the verdict the jury reached.”); see also Railroad Dynamics, Inc. v. A. Stucki Co., 727 F.2d 1506, 1516, 220 USPQ 929, 939 (Fed.Cir.1984).