Full opinion text
BISSELL, Circuit Judge, filed an opinion, in which MARKEY, Chief Judge, FRIEDMAN, RICH, DAVIS, NIES and ARCHER, Circuit Judges, joined; BENNETT, Senior Circuit Judge, filed a dissent-in-part, in which COWEN, Senior Circuit Judge, SMITH and NEWMAN, Circuit Judges, joined. NIES, Circuit Judge, filed additional views; NEWMAN, Circuit Judge, filed a commentary. BISSELL, Circuit Judge. This appeal and cross-appeal are from a judgment of the United States District Court for the Northern District of Georgia, 225 USPQ 558 (N.D.Ga.1984). The district court found that Durand-Wayland’s accused devices do not infringe any claim, literally or under the doctrine of equivalents. Unable to view that finding as clearly erroneous under Fed.R.Civ.P. 52(a), we affirm the judgment of noninfringement and vacate, as moot, that part of the judgment concerning the validity of Pennwalt’s patent. BACKGROUND Pennwalt sued Durand-Wayland for infringing claims 1, 2, 10 and 18 (claims-at-issue) of its U.S. Patent No. 4,106,628 (the ’628 patent) on an invention of Aaron J. Warkentin and George A. Mills, entitled “Sorter for Fruit and the Like.” Following a nonjury trial on the issues of patent infringement and validity, the district court, on March 22, 1984, issued an opinion concluding that (1) the claims-at-issue were not anticipated by the prior art, (2) the 628 patent had not run afoul of the “on sale” bar of 35 U.S.C. § 102(b) (1982), and (3) the accused devices did not infringe any of the claims-at-issue, either literally or under the doctrine of equivalents. In an unpublished supplemental order of April 26, 1984, the district court concluded as a matter of law that the claims-at-issue would not have been obvious in light of the prior art. Pennwalt appeals from the district court’s holding of noninfringement, both literally and under the doctrine of equivalents, and its award of costs against Penn-walt. Durand-Wayland appeals from the district court’s validity holdings and its denial of Durand-Wayland’s request for recovery of its attorney fees. ISSUE Our disposition of this appeal requires resolution of the single question of whether the district court’s finding of no infringement was clearly erroneous. OPINION The ’628 patent claims a sorter. The principal object of the invention is to provide a rapid means for sorting items, such as fruit, by color, weight, or a combination of these two characteristics. The sorter recited in claims 1 and 2 conveys items along a track having an electronic-weighing device that produces an electrical signal proportional to the weight of the item, along with signal comparison means, clock means, position indicating means, and discharge means, each of which performs specified functions. The specification describes the details of a “hard-wired” network consisting of discrete electrical components which perform each step of the claims, e.g., by comparing the signals from the weighing device to reference signals and sending an appropriate signal at the proper time to discharge the item into the container corresponding to its weight. The combined sorter of claims 10 and 18 is a multifunctional apparatus whereby the item is conveyed across the weighing device and also carried past an optical scanner that produces an electrical signal proportional to the color of the item. The signals from the weighing device and color sensor are combined and an appropriate signal is sent at the proper time to discharge the item into the container corresponding to its color and weight. Durand-Wayland manufactures and sells two different types of sorting machines. The first accused device, the “Microsizer,” sorts by weight only and employs software labeled either Version 2 or Version 5. The second accused device employs software labeled Version 6 and sorts by both color and weight through the use of the “Micro-sizer” in conjunction with a color detection apparatus called a “Microsorter.” A more complete description of the claimed invention and of the accused devices is set forth in the district court’s opinion. Pennwalt, 225 USPQ at 559-60. The claims-at-issue are set forth in the district court’s opinion. Id. at 564-65. I. Literal Infringement Pennwalt asserts on appeal that all limitations set forth in claims 1 and 2 and some limitations set forth in claims 10 and 18 can be read literally on the accused devices. Pennwalt contends that the district court erred in interpreting the claims by going beyond the means-plus-function language of a claim limitation and comparing the structure in the accused devices with the structure disclosed in the specification. Such comparison allegedly resulted in the court’s reading nonexistent structural limitations into the claims. Pennwalt relies on the statement in Graver Tank & Mfg. Co. v. Linde Air Prods. Co., 339 U.S. 605, 607, 70 S.Ct. 854, 855, 94 L.Ed. 1097 (1950): “If accused matter falls clearly within the claim, infringement is made out and that is the end of it.” In view of the literal breadth of means-plus-function language in the claims, that “test” for literal infringement would encompass any means that performed the function of a claim element. 35 U.S.C. § 112 (1982). This is not the “test.” The Graver Tank statement predated the inclusion in the 1952 Act of the provision specifically permitting “means” limitations, section 112, paragraph 6. See P. Federico, Commentary on the New Patent Law, 35 USCA 1, 25 (1954). As Judge Rich, one of the drafters of the statute, stated in a 1952 address explaining the import of section 112, paragraph 6: If you adopt this practice, that element or step is to be construed — shall be construed (it is mandatory) — to cover the corresponding structure, material or acts described in the specification and equivalents thereof. Address before the New York Patent Law Association (November 6, 1952), reprinted in R. Calvert, The Encyclopedia of Patent Practice and Invention Management 17, “Section 112 — Means Claims” (1964). Thus, section 112, paragraph 6, rules out the possibility that any and every means which performs the function specified in the claim literally satisfies that limitation. While encompassing equivalents of those disclosed in the specification, the provision, nevertheless, acts as a restriction on the literal satisfaction of a claim limitation. Data Line Corp. v. Micro Technologies, Inc., 813 F.2d 1196, 1201, 1 USPQ2d 2052, 2055 (Fed.Cir.1987). If the required function is not performed exactly in the accused device, it must be borne in mind that section 112, paragraph 6, equivalency is not involved. Section 112, paragraph 6, plays no role in determining whether an equivalent function is performed by the accused device under the doctrine of equivalents. Thus, it was not legal error (as Pennwalt asserts) for the district court to have made a comparison between Durand-Wayland’s structure and the structure disclosed in the specification for performing a particular function. The statute means exactly what it says: To determine whether a claim limitation is met literally, where expressed as a means for performing a stated function, the court must compare the accused structure with the disclosed structure, and must find equivalent structure as well as identity of claimed function for that structure. Palumbo v. Don-Joy Co., 762 F.2d 969, 975, 226 USPQ 5, 8 (Fed.Cir.1985); D.M.I., Inc. v. Deere & Co., 755 F.2d 1570, 1575, 225 USPQ 236, 239 (Fed.Cir. 1985); Radio Steel & Mfg. Co. v. MTD Prods., Inc., 731 F.2d 840, 848, 221 USPQ 657, 663 (Fed.Cir.1984). Where the issue is raised, it is part of the ultimate burden of proof of the patent owner to establish, with respect to a claim limitation in means-plus-function form, that the structure in the accused device which performs that function is the same as or an equivalent of the structure disclosed in the specification. In essence, Pennwalt erroneously argues that, if an accused structure performs the function required by the claim, it is per se structurally equivalent. That view entirely eliminates the section 112, paragraph 6, restriction and has been rejected in the above-cited precedent of this court. We need not, and do not, determine whether the district court correctly found no equivalency in structure because the district court also found that the accused devices, in any event, did not perform the same functions specified in the claim. For example, the district court found that the accused devices had no position indicating means which tracked the location of the item being sorted. That finding negates the possibility of finding literal infringement. II. Infringement under the Doctrine of Equivalents Under the doctrine of equivalents, infringement may be found (but not necessarily ) if an accused device performs substantially the same overall function or work, in substantially the same way, to obtain substantially the same overall result as the claimed invention. Perkin-Elmer Corp. v. Computervision Corp., 732 F.2d 888, 901-02, 221 USPQ 669, 679 (Fed.Cir.), cert. denied, 469 U.S. 857, 105 S.Ct. 187, 83 L.Ed.2d 120 (1984); Graver Tank, 339 U.S. at 608, 70 S.Ct. at 856. That formulation, however, does not mean one can ignore claim limitations. As this court recently stated in Perkin-Elmer Corp. v. Westinghouse Elec. Corp., 822 F.2d 1528, 3 USPQ2d 1321 (Fed.Cir.1987): One must start with the claim, and though a “non-pioneer” invention may be entitled to some range of equivalents, a court may not, under the guise of applying the doctrine of equivalents, erase a plethora of meaningful structural and functional limitations of the claim on which the public is entitled to rely in avoiding infringement.... Though the doctrine of equivalents is designed to do equity, and to relieve an inventor from a semantic strait jacket when equity requires, it is not designed to permit wholesale redrafting of a claim to cover nonequivalent devices, i.e., to permit a claim expansion that would encompass more than an insubstantial change. (Citations omitted.) ... [I]n applying the doctrine of equivalents, each limitation must be viewed in the context of the entire claim.... “It is ... well settled that each element of a claim is material and essential, and that in order for a court to find infringement, the plaintiff must show the presence of every element or its substantial equivalent in the accused device.” Lemelson v. United States, 752 F.2d 1538, 1551, 224 USPQ 526, 533 (Fed.Cir.1985) (footnote omitted). To be a “substantial equivalent,” the element substituted in the accused device for the element set forth in the claim must not be such as would substantially change the way in which the function of the claimed invention is performed. Id. at 1532-33, 3 USPQ2d at 1324-25. Pennwalt argues that the “accused machines simply do in a computer what the patent illustrates doing with hard-wired circuitry,” and asserts that “this alone is insufficient to escape infringement,” citing Decca Ltd. v. United States, 544 F.2d 1070, 1080-81, 210 Ct.Cl. 546, 563-64, 191 USPQ 439, 447 (1976). If Pennwalt was correct that the accused devices differ only in substituting a computer for hard-wired circuitry, it might have a stronger position for arguing that the accused devices infringe the claims. The claim limitations, however, require the performance of certain specified functions. Theoretically, a microprocessor could be programmed to perform those functions. However, the district court found that the microprocessor in the accused devices was not so programmed. After a full trial, the district court made findings that certain functions of the claimed inventions were “missing” from the accused devices and those which were performed were “substantially different.” Pennwalt, 225 USPQ at 572. The district court observed that “because the ‘Microsizer’ uses different elements and different operations (on the elements it does use) than the elements and operations disclosed in the patent-in-suit to achieve the desired results, infringement can only be found if the different elements and operations are the legal equivalents of those disclosed in the patent-in-suit.” Id. It is clear from this that the district court correctly relied on an element-by-element comparison to conclude that there was no infringement under the doctrine of equivalents, because the accused devices did not perform substantially the same functions as the Penn-walt invention. For example, the district court found in part: The machine described in the patent-in-suit uses shift registers that respond to “clock pulses” in order to indicate the various positions of the items to be sorted before each item is discharged. The “Microsizer” does not have any “indicating means” to determine positions of the items to be sorted since the microprocessor stores weight and color data, not the positions of the items to be sorted. After a piece of fruit has been analyzed by the “Microsorter” and while it is in transit from the optical detection means to the weight scale, the color value determined by the “Microsorter” is sorted in a color value queue. A color value queue pointer (which changes in value) points to the location of the data corresponding to the next piece of fruit to reach the weight scale. A weight value queue pointer (which also changes in value) is used to correspond to the number of cups between the weight scale and the drop location. The microprocessor software utilizes a random access memory that stores the digital numbers which resulted from the conversion of the analog signals generated by the “Microsorter” and the weight scale, and the queue pointers (under clock control) point to the memory location that has the data about a piece of fruit. The data is never “shifted” around, but rather is just stored in memory until the software routines call for data to be utilized in subsequent portions of the program(s). Thus, the “Microsizer” has neither a “first position indicating means” nor a “second position indicating means.” The machine described in the patent-in-suit produces signals that indicate where the fruit is, i.e. track the progression of each cup. The “Microsizer” does not. Id. at 569. Pennwalt argues that the district court erred as a matter of law in interpreting the claims which led to these findings. Penn-walt asserts that the district court looked to the specification and compared the accused devices with the preferred embodiment. As indicated, under the section 112, paragraph 6, analysis, such comparison was entirely correct, but it is readily apparent that the court did not so limit its infringement analysis. The court also looked for equivalent functions. It is not the province of this court to determine de novo what is the preponderance of the evidence underlying the ultimate finding of no infringement or the subsidiary facts supporting that finding. Indeed, Fed.R.Civ.P. 52(a) prohibits our undertaking the role of the fact finder, which is the domain of the trial court. Anderson v. City of Bessemer City, North Carolina, 470 U.S. 564, 573-74, 105 S.Ct. 1504, 1511-12, 84 L.Ed.2d 518 (1985); Graver Tank, 339 U.S. at 611, 70 S.Ct. at 857. As stated in United States v. United States Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 541, 92 L.Ed. 746 (1948): “A finding is ‘clearly erroneous’ when, although there is evidence to support it, the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been committed.” Here, if the district court was not clearly erroneous in finding that even a single function required by a claim or an equivalent function is not performed by the Du-rand-Wayland sorters, the court’s finding of no infringement must be upheld. Perkin-Elmer v. Westinghouse, 822 F.2d at 1531, 3 USPQ2d at 1323. Pennwalt asks us to reweigh the testimony, urging that the testimony of the court-appointed expert, Dr. Vacroux, is so strongly supportive of the contrary finding of infringement under the doctrine of equivalents that the district court’s finding of no infringement cannot stand. Penn-walt, however, misconstrues the effect of Dr. Vacroux’s testimony. Dr. Vacroux was a technical, not a legal, expert. He was not expected to, and did not, analyze infringement under a legal standard. He was not concerned, for example, with the scope of equivalents to which the claims were entitled in view of the prosecution history. In this light, the reports and testimony of Dr. Vacroux fully support the district court’s finding of no infringement, rather than Pennwalt’s position. Pennwalt relies on the conclusory statement in Dr. Vacroux’s initial written report in which he characterized the accused devices as “functionally equivalent” to the “invention.” However, Dr. Vacroux also stated therein that “the new designs could be totally equivalent to those they replaced, although the components and methods used differed significantly.” (Emphasis added.) Further, when Dr. Vacroux was called to testify, and, in response to detailed questioning by the court, Dr. Va-croux testified that the accused devices performed “some" of the same type of operations but “in a different way” from the claimed invention. Specifically, Dr. Va-croux answered a question by the court as follows: Dr. Vacroux: They [the accused and claimed devices] are not equivalent, but functionally the internal operations I would say internally are functionally equivalent because they perform some of the same type of operations. They compare, they shift, they add so we do have the same type of functions internally but done in a different way. [Emphasis added.] Dr. Yacroux’s statement that the accused devices performed only “some of the same type of operations” supports the court’s finding that some were not. It does not conflict with any of his other testimony or with his written report. Thus, on the issue of whether the accused devices performed each of the functional limitations of the claim or its equivalent and, thus, operated in substantially the same way, the district court followed its expert precisely. The court cannot be faulted for finding no infringement where the expert’s testimony as to facts negates a finding of equivalency under the correct legal standard. See Perkin-Elmer v. Westinghouse, 822 F.2d at 1531-32, 3 USPQ2d at 1323-24. With respect to the scope of equivalent functions, the court correctly limited the claims, based upon review of the prosecution history, stating that “the claims of the patent should be read in light of the careful phraseology [of functions] chosen by the inventors” because “the machine disclosed in the patent-in-suit was carefully described so that it did not read on the prior art.” Pennwalt, 225 USPQ at 570. As the court correctly noted, the invention was not a pioneer, but an improvement in a crowded art. The claims are “broad” with respect to what type of product can be sorted, i.e., “items” and, thus, sorters of all types of “items” fall within the relevant prior art. The claims are narrow, however, with respect to how the claimed sorter operates. Originally, the claims contained no position indicating means element with its associated functional limitations. The addition of that element was crucial to patentability. A device that does not satisfy this limitation at least equivalently does not function in • substantially the same way as the claimed invention. Perkin-Elmer v. Westinghouse, 822 F.2d at 1532, 3 USPQ2d at 1324; Chemical Eng’g Corp. v. Essef Indus., Inc., 795 F.2d 1565, 1572-73, 230 USPQ 385, 390-91 (Fed.Cir.1986). The trial court found that the accused devices do not have any position indicating means to determine positions of the items to be sorted. Specifically with respect to claims 10 and 18, the court correctly held that “the microprocessor stores weight and color data, not the positions of the items to be sorted.” Pennwalt, 225 USPQ at 569. Since each of the claims-at-issue requires a position indicating means and the same analysis applies to each, we set forth only the relevant language of claim 10: first position indicating means responsive to a signal from said clock means and said signal from said second comparison means for continuously indicating the position of an item to be sorted while the item is in transit between said optical detection means and said electronic weighing means, second position indicating means responsive to the signal from said clock means, the signal from said first comparison means and said first position indicating means for generating a signal continuously indicative of the position of an item to be sorted after said item has been weighed. The testimony of Dr. Alford, Durand-Wayland’s expert, was that the accused machine had no component which satisfied either of the above limitations defining position indicating means, and Pennwalt has admitted that the accused machines do not sort by keeping track of the physical location of an item in transit, continuously or otherwise, as required by each of these limitations. Pennwalt argues that there is a way to find out where an item is physically located on the track in the accused machine. Its witness, Dr. Moore, testified “you could find the location of a particular fruit as it moves from the scale to the drop by counting the distance from the stored value for that fruit back to the place the pointer is indicating at the start of the queue.” (Emphasis added.) Dr. Alford, Durand-Way-land’s expert, admitted this was possible. Thus, Pennwalt asserts that the accused devices have “position indicating means.” One need not explain the technology to understand the inadequacy of Dr. Moore’s testimony. As Dr. Moore himself indicates, the accused machine simply does not do what he explains “could” be done. It is admitted that the physical tracking of fruit is not part of the way in which the Durand-Wayland sorter works, in contrast to the claimed sorter which requires some means for “continuously indicating the position of an item to be sorted.” While a microprocessor theoretically could be programmed to perform that function, the evidence led the court to a finding that the Durand-Wayland machines performed a substantially different function from that which each of the claims requires. With respect to the other limitations of claim 10, Pennwalt admits that the asserted “position indicating means” of Durand-Wayland does not meet the limitation that the means must be “responsive to ... said signal from said second comparison means” because no comparison is made on the Du-rand-Wayland machine before the discharge point. However, Pennwalt contends that Durand-Wayland “has merely changed the position of an operable element, but the operation and results achieved are the same as those claimed.” Pennwalt’s analysis is flawed in significant respects. First, the claim requires that the “position indicating means” must be responsive to certain specified signals. Thus, finding some combination of components in the accused device that might also be labeled a “position indicating means” is a meaningless exercise when such combination is not responsive to the specified signal. See Polaroid Corp. v. Eastman Kodak Co., 789 F.2d 1556, 1570, 229 USPQ 561, 572 (Fed.Cir.1986) (argument “based on truncation of the claim language” rejected). As stab ed in Panduit Corp. v. Dennison Mfg. Co., 810 F.2d 1561, 1576, 1 USPQ2d 1593, 1603, (Fed.Cir.), cert. denied, — U.S. —, 107 S.Ct. 2187, 95 L.Ed.2d 843 (1987): In interpreting the claims, the district court committed fundamental legal error when it analyzed each by a single word description of one part of the claimed tie. In patent law, a word (“teeth”; “hinge”; “ledge”) means nothing outside the claim and the description in the specification. A disregard of claim limitations, as here, would render claim examination in the PTO meaningless. If, without basis in the record, courts may so rewrite claims, the entire statutory-regulatory structure that governs the drafting, submission, examination, allowance, and enforceability of claims would crumble. Second, the district court correctly rejected Pennwalt’s assertion that the memory component of the Durand-Wayland sorter which stores information as to weight and color of an item performed substantially the same functions as claimed for the position indicating means. The district court found that a memory function is not the same or substantially the same as the function of “continuously indicating” where an item is physically located in a sorter. On this point the record is indisputable that before the words “continuously indicating” were added as an additional limitation, the claim was unpatentable in view of prior art which, like the accused machines, stores the information with respect to sorting criteria in memories, but did not “continuously” track the location. See, e.g., U.S. Patent No. 3,289,832 issued to Ramsey. Thus, the facts here do not involve later-developed computer technology which should be deemed within the scope of the claims to avoid the pirating of an invention. On the contrary, the inventors could not obtain a patent with claims in which the functions were described more broadly. Having secured claims only by including very specific functional limitations, Penn-walt now seeks to avoid those very limitations under the doctrine of equivalents. This it cannot do. Graham v. John Deere Co., 383 U.S. 1, 33-34, 86 S.Ct. 684, 701-02, 15 L.Ed.2d 545 (1965); Chemical Eng’g Corp., 795 F.2d at 1572 n. 8, 230 USPQ at 391 n. 8; see also Exhibit Supply Co. v. Ace Patents Corp., 315 U.S. 126, 136, 62 S.Ct. 513, 518, 86 L.Ed. 736 (1942); Coleco Indus., Inc. v. United States Int’l Trade Comm’n, 573 F.2d 1247, 1257-58, 65 C.C.P.A. 105, 116-17, 197 USPQ 472, 479-80 (1978). Simply put, the memory components of the Durand-Wayland sorter were not programmed to perform the same or an equivalent function of physically tracking the items to be sorted from the scanner to the scale or from the scale to its appropriate discharge point as required by the claims. Contrary to Pennwalt’s arguments, the district court did not disregard the need to consider a range of equivalent functions under the doctrine of equivalents. Rather, upon evaluation of the evidence, the court concluded, as a fact, that no component in the Durand-Wayland devices performed a function within the permissible range of equivalents for the function of the first position indicating means. That function is required by all of the claims-at-issue. No means in the accused devices performs that function and thus there could be no literal infringement. No means with an equivalent function was substituted in the accused devices and thus there can be no infringement under the doctrine of equivalents. The district court’s finding of no infringement is not clearly erroneous. Attorney Fees Durand-Wayland has not shown that the district court’s refusal to find the case exceptional and award it attorney fees was based on clearly erroneous findings. Reactive Metals & Alloys Corp. v. ESM, Inc., 769 F.2d 1578, 1582-83, 226 USPQ 821, 824 (Fed.Cir.1985). CONCLUSION We affirm the judgment based on the finding of no infringement, the award of costs, and the denial of attorney fees to Durand-Wayland. There being no indication that Durand-Wayland’s cross-appeal on validity extends beyond the litigated claims or the accused devices found to be noninfringing, we dismiss the cross-appeal as moot and vacate that part of the judgment concerning the validity of Pennwalt’s patent. AFFIRMED-IN-PART AND VACATED-IN-PART. . The doctrine of equivalents is limited in that the doctrine will not extend (1) to cover an accused device in the prior art, and (2) to allow the patentee to recapture through equivalence certain coverage given up during prosecution. Loctite Corp. v. Ultraseal Ltd., 781 F.2d 861, 870, 228 USPQ 90, 96 (Fed.Cir.1985). . That a patent applicant narrows his claim to secure a patent does not always mean that prosecution history estoppel completely prohibits the patentee from recapturing some of what was originally claimed. The amount of coverage retained depends on the circumstances of each case. See, e.g., Loctite Corp. v. Ultraseal Ltd., 781 F.2d 861, 870, 228 USPQ 90, 96 (Fed.Cir.1985); Bayer Aktienegesellschaft v. Duphar Int'l Research, 738 F.2d 1237, 1243, 222 USPQ 649, 653 (Fed.Cir.1984); Hughes Aircraft Co. v. United States, 717 F.2d 1351, 1362, 219 USPQ 473, 481 (Fed.Cir.1983).
BENNETT, Senior Circuit Judge, dissenting in part, with whom COWEN, Senior Circuit Judge, and EDWARD S. SMITH and PAULINE NEWMAN, Circuit Judges, join. I agree with the majority that the district court’s finding of no literal infringement of the ’628 patent claims has not been shown to be clearly erroneous. However, I believe that the district court failed to consider the doctrine of equivalents issue in accord with the precedents of both this court and the Supreme Court. The in banc majority in this case, in affirming the district court, does not follow the precedents of this court. While the court sitting or hearing a case in banc can overrule its prior decisions, see Capital Elec. Co. v. United States, 729 F.2d 743, 746 (Fed.Cir.1984), it must do so openly and affirmatively, not sub silentio. The majority apparently, curiously, prefers not to acknowledge expressly the significant number of our prior decisions being overruled by its opinion. Therefore, I respectfully dissent from the majority’s discussion of and conclusion on the issue of infringement under the doctrine of equivalents. The majority opinion contravenes Supreme Court precedents, which do not leave the majority free to rewrite the doctrine of equivalents without regard for stare decisis principles. In so doing, the majority has made shortsighted policy choices. The majority has contrived an analytical framework for the doctrine of equivalents that is little more than a redundant literal infringement inquiry, which renders the doctrine of equivalents so unduly restrictive and inflexible as to end its usefulness as judicial doctrine. As this court is confronted in the future with different factual settings in the varied and increasingly complex technologies that comprise this court’s patent cases, the court will be forced to admit (or ignore) the full extent of the ties with which it has bound itself today. The majority facially retains the historical test set forth in Graver Tank & Mfg. Co. v. Linde Air Prods. Co., 339 U.S. 605, 70 S.Ct. 854, 94 L.Ed. 1097, 86 USPQ 328 (1950), for infringement under the doctrine of equivalents by stating that infringement in such instances may be found if an accused device performs substantially the same overall function or work, in substantially the same way, to obtain substantially the same overall result as the claimed invention. Graver Tank, 339 U.S. at 608, 70 S.Ct. at 856, 85 USPQ at 330. But in practical effect, the majority has eviscerated the underlying rationale of the Graver Tank test by requiring, under the doctrine of equivalents, an exact equivalent for each element of the claimed invention. The majority in fact commends the district court for undertaking the proper doctrine of equivalents determination, which the majority describes as an element-by-element comparison of the accused device and the patent-in-suit. However, the purported “element-by-element comparison” was never the extent of the doctrine of equivalents analysis under our here-ignored precedents which also required that the analysis be undertaken in light of the entirety of the accused device and entirety of the patent-in-suit. See, e.g., Martin v. Barber, 755 F.2d 1564, 1568, 225 USPQ 233, 235 (Fed.Cir.1985); Carman Indus., Inc. v. Wahl, 724 F.2d 932, 942, 220 USPQ 481, 488-89 (Fed.Cir.1983); Hughes Aircraft Co. v. United States, 717 F.2d 1351, 219 USPQ 473 (Fed.Cir.1983). Even the district court in this case recognized that our precedents require more. Unlike the limited analysis attributed to it by the majority, once the district court determined that all the elements in the patent-in-suit or equivalents of those elements (as determined under 35 U.S.C. § 112-6) were not present in the accused device, it attempted to go beyond the element-by-element comparison to determine whether the “different elements and operations” used in the accused device were “the legal equivalents of those disclosed in the patent-in-suit.” 225 USPQ at 572. Although in my opinion the district court did not completely understand the significance of the requirement to view the “claimed invention as a whole” and therefore erred in making the proper inquiry, the district court nevertheless was attempting to comply with our precedents that undeniably required application of the doctrine of equivalents to the invention as a whole. Hughes, 717 F.2d at 1364, 219 USPQ at 482. This required the fact finder to decide by viewing the accused device as a whole whether it and the claimed invention operate in substantially the same way and have substantially the same function and result as the claimed invention. Martin, 755 F.2d at 1568, 225 USPQ at 235. The majority’s departure from our recent precedents is illustrated by an examination of Hughes Aircraft Co. v. United States, 717 F.2d 1351, 219 USPQ 473 (Fed.Cir.1983), which is clearly being overruled by the majority, despite not even being mentioned in its opinion. In Hughes, this court squarely addressed the issue of infringement under the doctrine of equivalents of claims containing means-plus-function limitations. The invention in Hughes was an “apparatus” (a space satellite) and the representative claim contained multiple limitations, some of which were written in means-plus-function form, including (1) means for providing an indication of the spin angle to a location external to the satellite; (2) means for receiving control signals from the external source; and (3) means for applying fluid to a fluid expulsion means within a fixed time period after the receipt of a control signal from the external location. Id. at 1364, 219 USPQ at 483. The alleged infringing spacecraft replaced the three above-enumerated steps with an on-board computer which did not transmit the data to an external source, receive a control signal from that external source, or respond to the received signal as expressly required by the claim limitations. The Hughes trial court had treated the subject of equivalency in its finding: There is no obvious or exact equivalent of plaintiffs means for providing an indication of the ISA to an external location in either of the SKYNET II, NATO II or DSCS II systems. Nor is there an obvious or exact equivalent of the means for pulsing the precession jet within a fixed time period after the receipt of a control signal in any of the accused store and execute systems. Id. at 1363, 219 USPQ at 482 (quoting district court opinion). This court, in reversing the trial judge, concluded that: However the phrase “obvious and exact equivalents” may be defined, it was effectively and improperly applied here as a substitute for literal infringement, the absence of which was conceded. The failure to apply the doctrine of equivalents to the claimed invention as a whole, and the accompanying demand for “obvious and exact” equivalents of two elements the presence of which would have effectively produced literal infringement, was error. Id. at 1364, 219 USPQ at 482 (emphasis added). The Hughes decision held that the accused spacecraft with the position computed and retained on-board was the equivalent of the claimed satellite with its position calculated on the ground since the devices performed substantially the same function in substantially the same way to reach the same result, and that this conclusion was not estopped by the prior art or prosecution history. This conclusion was reached in full recognition of the failure of the accused spacecraft to meet separately three limitations of the claims-in-issue. Hughes stated: That argument is clearly effective against an allegation of literal infringement, for if the ... spacecraft did send an ISA position to the ground, literal infringement of that element of the claims would be clear. Williams controls his satellite, and the government controls its S/E spacecraft, from the ground. That Williams does so in “real time” and the government does so in a delayed reaction made possible by the advent of computers does not establish that the S/E spacecraft do not perform the same function in substantially the same way to obtain the same result. Id. at 1365, 219 USPQ at 483 (emphasis in original). Despite any protestations to the contrary, it should be beyond dispute that infringement could not have been found in Hughes under the analysis announced by the majority today. In fact, the majority’s current analysis under the doctrine of equivalents amounts to nothing more than the search for “obvious and exact equivalents” that this court denounced in Hughes. This is aptly illustrated by the majority’s reliance on the Durand microprocessor not being programmed to perform each of the functions required by the claim limitations, even though it was theoretically possible to so program the microprocessor, as grounds for affirmance of the finding of no infringement under the doctrine of equivalents. Although the majority asserts Penn-walt’s case might be stronger if the accused devices only substituted a computer for hard-wired circuitry, the district court found that “the machine described in the patent-in-suit and the ‘Microsizer’ are virtually identical except for the control system that conducts the sorting operation.” 225 USPQ at 569 (emphasis in original). The majority does not criticize that finding. Furthermore, the district court’s finding that the Microsizer performed different functions than the patent-in-suit was solely part of the district court’s analysis leading to its finding that there was no literal infringement, id., and can be characterized as nothing more than a recognition of the inherent differences in hard-wired circuitry and microprocessors. If, as under the district court’s and the majority’s analyses, the use of software and a computer or microprocessor to perform the function of hard-wired circuitry is enough to find that a device employing a microprocessor performs different functions and operations, then no device that employs computers or microprocessors could ever under the doctrine of equivalents infringe a prior patent on a device that employs hard-wired circuitry. The extent of the doctrine of equivalents inquiry cannot be confined, as it was by the district court and the majority, to the mere change to the use of a computer or microprocessor. As stated by the Supreme Court in Graver Tank, 339 U.S. at 609, 70 S.Ct. at 856, 85 USPQ at 331, it is important to know whether those “reasonably skilled in the art would have known of the interchangeability of an ingredient not contained in the patent with one that was,” see infra, but the district court added an erroneous criterion: whether those with ordinary skill in the art would know how to substitute them. Pennwalt, 225 USPQ at 572. In my opinion, the district court misconstrued the inquiry. If the inquiry is to determine whether one item can be substituted for another, and those skilled in the art know that it can, then the actual interchangeability of the two items is not affected by whether those same persons could, without further study, substitute them. See Atlas Powder Co. v. E.I. Du Pont De Nemours & Co., 750 F.2d 1569, 1580, 224 USPQ 409, 416-17 (Fed.Cir.1984); Lockheed Aircraft Corp. v. United States, 553 F.2d 69, 82, 213 Ct.Cl. 395, 420, 193 USPQ 449, 461 (1977). The literal infringement inquiry involving section 112-6 and the infringement inquiry under the doctrine of equivalents are distinct, even when limitations written in means-plus-function language are present, and must be made separately when, as in the present case, each is properly at issue. The majority apparently agrees that the two infringement analyses should be distinct, but nevertheless completely blurs, if not eliminates, the distinctions between them. For, if section 112-6 “plays no role in determining whether an equivalent function is performed by the accused device under the doctrine of equivalents,” it is curious that the majority relies on a lengthy quote from the district court’s discussion of why the accused device does not literally infringe the patent-in-suit to illustrate why the district court’s element-by-element literal infringement comparison yielded no equivalent for the position indicating means. Thus, while the majority’s subsequent discussion does indeed support its conclusion that the district court’s finding of no literal infringement was correct, it is completely inappropriate as part of the infringement analysis under the doctrine of equivalents. The enactment of section 112-6 did not eliminate the applicability of the doctrine of equivalents in situations where there are means-plus-function claims involved. The majority nevertheless abandons the additional and separate equivalence analysis heretofore necessary, and fails to require the district court properly to determine if the accused device contained an equivalent of the position indicating means under the doctrine of equivalents, after having viewed the claimed invention and the accused device as a whole. Furthermore, the court-appointed expert’s testimony does not support either the majority’s conclusions or its analysis. In fact, the majority should have declined the invitation to reweigh the testimony and certainly should not have taken the additional step of recasting the court-appointed expert’s testimony to support its position since the job of evaluating the testimony of expert witnesses and witnesses in general is peculiarly that of the trier of fact. See, e.g., Inwood Labs., Inc. v. Ives Labs., Inc., 456 U.S. 844, 856, 102 S.Ct. 2182, 2189, 72 L.Ed.2d 606 (1982); Gould v. Quigg, 822 F.2d 1074, 1077-78, 3 USPQ2d 1302, 1304 (Fed.Cir.1987). The district court described Dr. Vacroux’s testimony as “unbiased, well-qualified and scholarly,” but nevertheless declined to adopt it. 225 USPQ at 572. As the majority points out, Dr. Vacroux was a technical, not legal, expert and he did not analyze the infringement issue under a legal standard. Yet, despite the district court’s rejection of testimony which was not directed toward the legal issues in the case, the majority curiously determines that the same testimony fully supports its finding of no infringement. Thus, it would appear that the district court was operating under a different view of what was needed for infringement under the doctrine and that the expert’s testimony must not have supported a finding of noninfringement under what the district court perceived the legal standard to be prior to this decision. Since Dr. Vacroux’s testimony was, at best, adequate to support a finding of no literal infringement, but completely inconclusive on the issue of infringement under the doctrine of equivalents, I prefer the district court’s decision not to adopt Dr. Vacroux’s testimony to the majority’s decision to recast it. The majority goes on to conclude that Pennwalt is estopped by the prosecution history from gaining coverage under the doctrine of equivalents of that which it gave up through amendment of its claims during prosecution of the ’628 patent. First, the majority holds that since the original claims were unpatentable without the “position indicating means” element, Penn-walt cannot now argue that “crucial” element is immaterial or deny that a device without that element does not function in substantially the same way as the claimed invention. However, in my view, the amendment to include the term “position indicating means” does not end the infringement inquiry here since both the accused device and the claimed invention “indicate position” in equivalent manners. The majority errs in interpreting the actual meaning of the term, as used by the pat-entee. The district court held that “the microprocessor stores weight and color data, not the positions of the items to be sorted.” 225 USPQ at 569. The majority upholds this finding as not clearly erroneous and, to the limited extent of the finding, I agree. However, the conclusion that the accused device does not indicate position does not end the analysis because it is clear that the claimed invention, just like the accused device, does not store or directly indicate the position of the items to be sorted. In order to apply the claims, they must first “be construed ‘in connection with the other parts of the patent instrument and with the circumstances surrounding the inception of the patent application.’ ” Fromson v. Advance Offset Plate, Inc., 720 F.2d 1565, 1569, 219 USPQ 1137, 1140 (Fed.Cir.1983) (quoting Autogiro Co. v. United States, 181 Ct.Cl. 55, 62, 384 F.2d 391, 397, 155 USPQ 697, 702 (1967)). The claimed invention of the ’628 patent employs shift registers to shift sequentially weight and color data corresponding to a specific item from one memory register to an adjoining register as that item moves along the conveyor belt. The data is placed and shifted across shift registers containing the same number of registers as there are conveyor locations between one detection means (for weight or color) and either the next detection means or the discharge point. Thus, the shifted data in the claimed invention represents weight and color data, not position, and could only be used to indicate position in the literal sense by physically counting where in the shift register is the data for a given item. Just as in the accused device, although it could be done, it is not. Both devices do, however, continually store the data corresponding to a particular item until it has reached the proper drop location corresponding to its color and weight characteristics. The claimed invention does this through the use of shift registers as described above while the accused device stores the data in a queue and uses pointers which move synchronously with the movement of the conveyor cups. The real question to be answered here, one which the district court and the majority never reached, is whether the queues and pointers perform in substantially the same way the functions of the claimed position indicating means when the latter are interpreted in light of the specification. The majority also holds that since Pennwalt added the limitation “continuously indicating” to the claims containing the position indicating means element, it cannot use the doctrine of equivalents to avoid that limitation. But whether the patent claims were amended by adding “continuously indicating” in order to distinguish from the noncontinuous methods employed in prior art, such as the Ramsey patent, is not relevant here since the Durand device also performs continuously the “position indicating” function. Finally, the majority’s treatment of the limitation that the first position indicating means be “responsive to ... said signal from said second comparison means” is the clearest illustration of the shortcomings of the majority’s approach to the doctrine of equivalents. Pennwalt concedes that the Durand devices do not literally infringe claims 10 and 18 because that limitation is not literally met. In the claimed invention, the color grade from the color head is compared to the reference values and converted into data form, which then is stored directly in the first position indicating means while the item is in transit to the weight scale. In the Durand device, the color grade is stored directly and is not compared to the reference values until the item has reached the weight scale. The majority rejected as significantly flawed Pennwalt’s assertion that Durand’s mere change of the position of an operable element was not enough to avoid infringement under the doctrine of equivalents. However, the only apparent flaw noted by the majority is that the claim requires that the position indicating means be responsive to certain signals and that without being responsive to the specified signal, any other combination of elements cannot be deemed to contain an equivalent of the functional limitations describing the position indicating means. In my view, a comparison to the color reference values occurring in the Durand device after the item reaches the weight scale is an equivalent function whether or not its first position indicating means is responsive to a signal from a comparison means, when the accused device is viewed in its entirety and compared to the claimed invention as a whole. To preclude the possibility of finding equivalent functions and therefore preclude finding infringement under the doctrine of equivalents simply because the accused device does not perform its functions in the same order as the claimed invention reduces the doctrine of equivalents, in practical effect, to nothing more than the test for literal infringement. As discussed above, the majority is overruling this court’s recent teachings that require consideration of the claimed invention as a whole when making the doctrine of equivalents analysis. The overruling of our precedents is, of course, permissible since the court sua sponte sits in banc. However, even when sitting in banc, this court can only overrule its own prior decisions, not those of the Supreme Court. I believe that the majority’s opinion departs significantly from the clear teachings of the Supreme Court, specifically those in Graver Tank & Mfg. Co. v. Linde Air Prods. Co., 339 U.S. 605, 70 S.Ct. 854, 94 L.Ed. 1097, 85 USPQ 328 (1950), and, in fact, adopts the position and the spirit of the dissenting opinions rejected in Graver Tank. In Graver Tank, the patent claims at issue involved an electrical welding process containing a combination of an alkaline earth metal silicate and any other silicate. The patented composition actually used silicates of calcium and magnesium, which were both alkaline earth metal silicates. The accused composition used silicates of calcium and manganese, the latter not being an alkaline earth metal. That difference prevented the finding of literal infringement. Nevertheless, the Supreme Court noted that the two compositions used the same mechanical methods, were identical in operation, and produced the same kind and quality of weld. The key issue became whether the substitution of a nonalkaline earth metal (manganese) for the alkaline earth metal (magnesium) claimed in the patent was, under the circumstances of the case and in view of the prior art and the technology involved, a sufficiently insubstantial change which justified application of the doctrine of equivalents. The Supreme Court concluded, despite the claim language, that the use of manganese was an insubstantial change. In fact, the Court found infringement under the doctrine of equivalents despite several prior art patents which disclosed the use of manganese in welding fluxes, and despite the patent specification itself indicating that the patentee had experimented with using manganese in its fluxes. The current case presents, as did Graver Tank and as does any case involving the doctrine of equivalents, a choice between conflicting policies. On the one hand, there is the historic right of affording the public fair notice of what the patentee regards as his claimed invention in order to allow competitors to avoid actions which infringe the patent and to permit “designing around” the patent. See, e.g., State Indus., Inc. v. A.O. Smith Corp., 751 F.2d 1226, 1236, 224 USPQ 418, 424 (Fed.Cir.1985). On the other hand, equally important to the statutory purpose of encouraging progress in the useful arts, is the policy of affording the patent owner complete and fair protection of what was invented. The resulting tension between the competing policies has been long recognized and the inability to reconcile fully the two views has long been accepted by the courts. In Royal Typewriter Co. v. Remington Rand, Inc., 168 F.2d 691, 692, 77 USPQ 517, 518 (2d Cir.1948), Judge Learned Hand elaborated: [A] patent is like any other legal instrument; but it is peculiar in this, that after all aids to interpretation have been exhausted, and the scope of the claims has been enlarged as far as the words can be stretched, on proper occasions courts make them cover more than their meaning will bear. If they applied the law with inexorable rigidity, they would never do this.... [But] at times they resort to the “doctrine of equivalents” to tern-per unsparing logic and prevent an in-fringer from stealing the benefit of the invention. No doubt, this is, strictly speaking an anomaly; but it is one which courts have frankly faced and accepted almost from the beginning (footnote omitted). All patents are entitled to its benefit to an extent, measured on the one hand by their contribution to the art, and on the other by the degree to which it is necessary to depart from the meaning to reach a just result. The same principles were stated by this court in Thomas & Betts Corp. v. Litton Sys., Inc., 720 F.2d 1572, 1579, 220 USPQ 1, 6 (Fed.Cir.1983): If a patentee were bound by the literal language of his specification and claims, the purpose of the doctrine of equivalents, to give relief against the copier who merely makes insubstantial substitutions in a claimed invention, would be frustrated. Thus, the proposition that the claims, taken in view of the specification, measure the metes and bounds of the invention has been realistically tempered by the judicially-formulated doctrine of equivalents. The equitable doctrine of equivalents has existed since Winans v. Denmead, 56 U.S. (15 How.) 330, 14 L.Ed. 717 (1853), and “has been consistently applied by [the Supreme Court] and the lower federal courts ... when the proper circumstances for its application arise.” Graver Tank, 339 U.S. at 608, 70 S.Ct. at 856, 85 USPQ at 330. In these more than 130 years of jurisprudence there have indeed arisen instances of judicial interpretation of claims beyond their literal boundaries, in the interest of justice. The view that such judicial flexibility should not exist is contrary to these decades of decisions, and appears to be contrary to that good judgment which underlies our precedents. To preserve to the district courts the ability to apply the long-standing equitable doctrine of equivalents in infringement suits that merit such equity does not pose a threat to the statutory role of claims. Even when the doctrine of equivalents is invoked, the claims set forth that against which the equivalency is tested. To require a one-to-one correspondence creates a bright line rule easier to apply, but costly in terms of unfair results in exceptional cases. Such inequity has been rejected by the Supreme Court and, up until today, by our own precedents. The slight gain in predictability resulting from the majority’s opinion will be greatly outweighed by the loss resulting from the inability to remedy inequities resulting from insubstantial changes to and substitutions in patent claims, as the Court recognized in Graver Tank. Had the majority in this case been confronted with the facts of Graver Tank, it is doubtful that it would have ever looked beyond the claim limitation specifically requiring an alkaline earth metal. Yet the Court’s discussion and analysis in Graver Tank illustrates both the shortsightedness of the policy choices and the shortcomings in the analysis adopted by this majority in the current opinion. [C]ourts have also recognized that to permit imitation of a patented invention which does not copy every literal detail would be to convert the protection of the patent grant into a hollow and useless thing. Such a limitation would leave room for — indeed encourage — the unscrupulous copyist to make unimportant and insubstantial changes and substitutions in the patent which, though adding nothing, would be enough to take the copied matter outside the claim, and hence outside the reach of law.... Outright and forthright duplication is a dull and very rare type of infringement. To prohibit no other would place the inventor at the mercy of verbalism and would be subordinating substance to form. It would deprive him of the benefit of his invention and would foster concealment rather than disclosure of inventions, which is one of the primary purposes of the patent system. The doctrine of equivalents evolved in response to this experience. The essence of the doctrine is that one may not practice a fraud on a patent. 339 U.S. at 607-08, 70 S.Ct. at 856, 85 USPQ at 330. Literal differences between the claimed and accused devices should always exist before the issue of infringement under the doctrine is reached. The majority’s holding in this case that the making of the color grade comparisons after the item reaches the weight scale in the Durand device is enough to avoid infringement under the doctrine is nothing more than a conclusion that “any unimportant and insubstantial changes and substitutions” which take the accused device “outside the claim” are also sufficient to show that the accused device does not function in substantially the same way as the claimed invention. Thus, the same features which defeat the possibility of literal infringement are now being used to preclude possible application of the doctrine of equivalents. Up until now, this court has rejected such a limited view of the doctrine. See, e.g., Loctite Corp. v. Ultraseal Ltd., 781 F.2d 861, 228 USPQ 90 (Fed.Cir.1985). Furthermore, the Supreme Court made clear that the doctrine of equivalents was not to be so limited. Equivalence, in the patent law, is not the prisoner of a formula and is not an absolute to be considered in a vacuum. It does not require complete identity for every purpose and in every respect. In determining equivalents, things equal to the same thing may not be equal to each other and, by the same token, things for most purposes different may sometimes be equivalents. Consideration must be given to the purpose for which an ingredient is used in a patent, the qualities it has when combined with the other ingredients, and the function which it is intended to perform. Graver Tank, 339 U.S. at 609, 70 S.Ct. at 856, 85 USPQ at 330-31. The application of the doctrine of equivalents to “the claimed invention as a whole” as expressed by this court in Hughes, 717 F.2d at 1364, 219 USPQ at 482, is inherent in the policy expressed in Graver Tank. In order to determine equivalency in light of “the purpose for which an ingredient is used in a patent” and “the qualities it has when combined with the other ingredients” as mandated by the Supreme Court, it is indeed necessary to view the entire claim as a whole, and not merely to conduct an element by element comparison. It is only after such an analysis that the fact finder can determine, in light of the entire claim and all of its limitations, whether the changes or substitutions made alter substantially the way that the accused device works when compared to the claimed invention as a whole. Thus, the proper inquiry was and should remain whether the devices considered as a whole satisfy the tripartite test of Graver Tank. That question has yet to be properly and completely considered by the trial court in this case. I would vacate the district court’s decision of noninfringement under the doctrine of equivalents, and remand for a proper consideration of that question in light of the law set out in our prior precedents and this opinion. The question of infringement, whether literal or by equivalents, is a question of fact. SRI Int’l v. Matsushita Elec. Corp., 775 F.2d 1107, 1125, 227 USPQ 577, 589 (Fed.Cir.1985). It is a question for the district co