Both parties appeal aspects of the judgment of the United States District Court for the District of Massachusetts,1 wherein Micron Separations, Inc. (“MSI”) was adjudged *1215to have infringed a patent owned by Pall Corporation (“Pall”) relating to polyamide membranes that are used in microfiltration. We affirm the judgment of infringement, reverse as to willful infringement, modify the damages award, and remand for recalculation of damages.
Background
Microfiltration membranes have an extremely small pore size, and are used to remove unwanted microscopic substances such as bacteria from fluids, or to separate and retain desired microscopic substances such as antibodies. Microfiltration membranes are used, for example, by pharmaceutical companies in bacteria-free drug production, by electronics manufacturers to purify water for chip manufacture, by hospitals to filter contaminants from fluids that are administered to patients, in diagnostics to detect the presence of antibodies or antigens, and as transfer membranes for identification of genetic characteristics. A particularly useful pore size for microfiltration membranes is about 0.22 micron, for this pore size yields the highest flow rate consistent with retention of the 0.3 micron Pseudomonas diminutia bacteria; removal of these bacteria is the standard criterion for sterilization by microfiltration.
Microfiltration membranes are not new, although membranes of the prior art were generally subject to deficiencies such as brittleness, excessive shrinkage, poor solvent resistance, limited flow rate, excessive clogging, and limited sterilizing ability. Microfiltration membranes had previously been made from organic polymers, but those membranes generally required the use of wetting agents for filtration, for water does not flow well through mieroporous membranes that are not instantly wettable. The addition of wetting agents to facilitate flow is undesirable because it adds impurities to the system.
In accordance with United States Patent No. 4,840,479, invention of Dr. David B. Pall (“the Pall patent”), microfiltration membranes are made from organic polymers called polyamides. The particular polyam-ides of this invention are sometimes called “nylons,” the generic term for linear aliphatic polyamides. The polyamide membranes at issue were made from the commercially available polymers nylon 66 and nylon 46.2 The method of making these membranes and the membranes themselves are described in the Pall patent and in the district court’s opinion, and will be repeated only as necessary to explain our decision.
Membranes had been made of polyamides before Dr. Pall’s invention, but the prior membranes were either soluble in alcohol or required the use of wetting agents for efficient filtration. The membranes of the prior art also had a surface skin of reduced porosity that retarded wettability of the membrane material and increased the flow time. The Pall patented membranes are naturally and instantaneously wettable and alcohol-insoluble, with uniform pore size throughout the membrane. They achieve efficient flow without the use of wetting agents. The Pall membranes, due to their improved properties and increased effectiveness in microfiltration, achieved not only commercial success but also scientific recognition. The trial court found:
Dr. Pall and his associates were pioneers both in the invention, development, and commercialization of functional nylon mi-crofiltration membranes and in the theory of microfiltration.
The Pall microfiltration membranes were made of nylon 66, and were introduced into commerce in 1978. The Pall patent issued in 1982. In 1983 MSI introduced a microfiltration membrane made of nylon 66. Pall filed suit in 1986. In 1989 MSI converted most but not all of its nylon 66 membrane production to nylon 46. The case came to trial in 1991.
The district court held that the Pall patent was valid, that MSI’s nylon 66 membranes *1216literally infringed the patent’s product claims, that MSI’s nylon 46 membranes infringed the product claims under the doctrine of equivalents, and that MSI’s process for producing the nylon membranes infringed the patent’s process claims under the doctrine of equivalents. The court held that MSI’s infringement was willful to the limited extent of MSI’s continued manufacture of some nylon 66 membranes after it had shifted most of its production to nylon 46. Although MSI declared bankruptcy after the adverse district court judgment,3 it remains operating under a stay of injunction granted by the district court as to MSI’s nylon 46 membranes.4
MSI appeals the rulings of infringement and willful infringement. Pall cross-appeals aspects of the damage award.
INFRINGEMENT
Nylon 66 Membranes
A representative product claim of the Pall patent is:
34. A hydrophilic skinless alcohol-insoluble polyamide resin membrane sheet of alcohol-insoluble hydrophobic polyamide resin selected from the group consisting of polyhexamethylene adipamide, polyhexam-ethylene sebacate, and poly-e-eaprolactam, and capable when completely immersed in water of being wetted through within no more than one second, and reverting when heated to a temperature just below the softening temperature of the membrane to a hydrophobic material which is no longer wetted by water.
Nylon 66 is polyhexamethylene adipamide, named in claim 34. It was not disputed that all of the limitations of claim 84 were met by MSI’s nylon 66 membranes, except for the term “skinless,” the meaning of which received extensive attention at trial. MSI argued that the plain meaning of “skinless” eliminated the MSI products from infringement, because a photomicrograph showed a “skin” on the MSI membranes. Pali’s position was that “skinless” as used in the specification and claims should be understood as a performance characteristic, for a skin is a nonporous layer that impedes filtration.
Claim interpretation is a question of law. Markman v. Westview Instruments, Inc., 52 F.3d 967, 34 USPQ2d 1321 (Fed.Cir.1995) (en banc). We review the construction of the claims without deference to that of the district court. Id. at 979, 34 USPQ2d at 1329 (“Because claim construction is a matter of law, the construction given the claims is reviewed de novo on appeal.”) In construing the claims we look to the language of the claims, the specification, and the prosecution history. Id. Extrinsic evidence may also be considered, if needed to assist in determining the meaning or scope of technical terms in the claims. Id. at 980, 34 USPQ2d at 1330.
At trial both sides presented support for their conflicting positions concerning the term “skinless.” Pall presented data showing “bubble point,” “flow time,” and “KL curve” measurements of the membranes in suit. These measurements were described in the Pall patent specification as tests for determining the presence or absence of a skin on a membrane. The bubble point test measures the pressure applied to a membrane whose pores are filled with water in order to force the water through the pores; the larger the bubble point value, the smaller the pores in the membrane. Flow time is the time required for a certain amount of water to pass through a given area of membrane; the smaller the pores the longer the flow time. The KL curve depicts the flow of water through a membrane at constant pressure; skinless membranes characteristically show an initially level KL curve followed by a sharp rise in the curve as flow commences, whereas KL curves for skinned membranes show a gradual upward-sloping curve. These are generally accepted tests of whether a membrane is skinned or skinless, for a nonporous skin impedes the passage of water through the membrane, and is detected by all three tests. The district court defined “skinless” as a performance characteristic in accordance with the parameters of these tests. We too conclude, based on the specification *1217and the data in evidence, that “skinless” is properly construed as a performance characteristic, and that a surface that does not impede flow is “skinless” as that term is used in the art of filtration membranes.
The district court then applied the construed claims to MSI’s membranes. The court found that the MSI membranes had bubble points, flow times, and KL curves characteristic of skinless membranes. The court also observed that MSI’s product literature described its membranes as of uniform pore size or having precise porosity, which the court found were necessary characteristics of skinless membranes. The court thus found, on the totality of the evidence, that the MSI membranes did not have a barrier layer that blocked fluid flow, and that the MSI membranes were skinless as the Pall patent used that term. We discern no clear error in this finding.
Literal infringement is found when every limitation of a claim is met in the accused structure. Applying its interpretation of the term “skinless,” which we have confirmed, the district court found that the MSI nylon 66 membranes literally infringed the Pall patent. The judgment of infringement by MSI’s nylon 66 membranes is affirmed.
Nylon 46 Membranes
Three years after Pall filed this suit, MSI converted most of its membrane manufacture from nylon 66 to nylon 46. Nylon 46 is made not from hexamethylenediamine and adipic acid, but from tetramethylenediamine and adipic acid. Claim 34, quoted supra, does not name polytetramethylene adipamide (nylon 46). However, claim 116 of the Pall patent describes the polyamide resins in terms of their methylene to amide ratio:
116. A hydrophilic skinless alcohol-insoluble polyamide resin membrane sheet of alcohol-insoluble hydrophobic polyamide resin having a ratio CH2:NHCO of methylene CH% to amide NHCO groups within the range of about 5:1 to about 7:1; capable when completely immersed in water of being wetted through within no more than one second, and reverting when heated to a temperature just below the softening temperature of the membrane to a hydrophobic material which is no longer wetted by water. [Emphasis added.]
Nylon 66 has a ratio of methylene to amide groups of 5:1. Nylon 46 has a ratio of 4:1. Pall asserts that the term “about 5:1 to about 7:1” is infringed by the ratio of 4:1, either literally or by the doctrine of equivalents.
A. Literal Infringement
The district court, construing the term “about 5:1 to about 7:1,” observed that the word “about” does not have a universal meaning in patent claims, and that the meaning depends on the technological facts of the particular case. We have so held. E.g., Andrew Corp. v. Gabriel Electronics, Inc., 847 F.2d 819, 821-22, 6 USPQ2d 2010, 2013 (Fed.Cir.), cert. denied, 488 U.S. 927, 109 S.Ct. 312, 102 L.Ed.2d 330 (1988); W.L. Gore & Assoc. Inc. v. Garlock, Inc., 842 F.2d 1275, 1280, 6 USPQ2d 1277, 1282 (Fed.Cir.1988).
The determination of whether the literal meaning or scope of “about 5:1 to about 7:1” includes 4:1 is a matter of claim construction, a question of law for decision de novo by this court. The use of the word “about,” avoids a strict numerical boundary to the specified parameter. Its range must be interpreted in its technologic and stylistic context. We thus consider how the term “about 5:1 to about 7:1” was used in the patent specification, the prosecution history, and other claims. It is appropriate to consider the effects of varying that parameter, for the inventor’s intended meaning is relevant. Extrinsic evidence of meaning and usage in the art may be helpful in determining the criticality of the parameter, and may be received from the inventor and others skilled in the field of the invention. See Markman, 52 F.3d at 980, 34 USPQ2d at 1330 (“The court may, in its discretion, receive extrinsic evidence in order ‘to aid the court in coming to a correct conclusion’ as to the ‘true meaning of the language employed’ in the patent.”) (quoting Seymour v. Osborne, 78 U.S. (11 Wall.) 516, 546, 20 L.Ed. 33 (1871)).
Dr. Pall, the inventor, explained his usage of “about 5:1 to about 7:1” as deriving from *1218his tests of the performance of various nylon resin membranes. He explained that the ratios change in half integers, e.g., 7:1, 6.5:1, 6:1, etc., depending on the total number of methylene groups in the acid and amine components of the polymer. For example, each recurring unit of nylon 66 contains 10 methylene groups and 2 amide groups, for a ratio of 10:2, which reduces to 5:1. Each recurring unit of nylon 46 contains 8 methylene groups and 2 amide groups, for a ratio of 4:1. The Pall patent also illustrated a mixture having the ratio of 5.3:1.
Dr. Pall testified that he conducted experiments with several commercially available nylon resins. He found that a nylon resin having a methylene:amide ratio higher than 7:1 produced membranes that were less readily wettable, such that use of a wetting-agent was required for optimum results. He therefore placed the upper limit of “about 7:1” in the claims. Dr. Pall stated that the only commercially available nylon resin with a ratio lower than 5:1 was a nylon with the ratio of 3:1, and that he made a membrane of the 3:1 nylon and found it to be soluble in alcohol and therefore unacceptable. No experiments were conducted with a nylon between 3:1 and 5:1, he testified, because none was commercially available. He placed the lower limit at “about 5:1.” The reasons for these claim limitations also appear in the patent specification.
The district court found that Dr. Pali’s use of the word “about” was “appropriate.” We agree that the evidence showed that an exact limitation would have been inappropriate. However, the evidence showed that while a ratio of 7:1 was satisfactory, higher ratios were not, suggesting that the upper limit was close to 7:1, and did not extend to, for example, 8:1. At the lower ratios, although Dr. Pall conducted no tests with resins between 3:1 and 5:1, the 3:1 resin was clearly unsatisfactory. Reviewing all the evidence, the district court held that a literal reading of “about 5:1 to about 7:1” did not include the ratio of 4:1. On plenary review we reach the same conclusion.
Since the claim is construed more narrowly than would literally encompass the ratio of 4:1, the district court’s finding that MSI’s nylon 46 membranes do not literally infringe the Pall patent is affirmed.
B. Infringement by Equivalents
When literal infringement is not established, infringement may be proved under the doctrine of equivalents when there is not a substantial difference between the claimed invention and the accused product. 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); Hilton Davis Chemical Co. v. Warner-Jenkinson Co., 62 F.3d 1512, 35 USPQ2d 1641 (Fed.Cir.1995) (en banc). The determination of whether the accused product is substantially the same as the claimed invention is a question of fact, and the district court’s determination thereof is reviewed for clear error. Id. at 1521, 35 USPQ2d at 1647.
The district court found that MSI’s nylon 46 membranes had substantially the same chemical and physical structure, performed the same function in the same way, and achieved the same result, as Pall’s claimed membranes. The court thus found that claim 116 was infringed under the doctrine of equivalents.
MSI does not argue on this appeal that the nylon 46 and nylon 66 membranes are not equivalent. Instead, MSI argues that Pall is estopped to establish infringement based on equivalency because during patent prosecution Pall voluntarily gave up claim scope that would have literally included nylon 46. MSI asserts that Pall is estopped from obtaining that scope under the doctrine of equivalents, even if the products are in fact equivalent.
Prosecution history estoppel limits infringement by otherwise equivalent structures, by barring recapture by the patentee of scope that was surrendered in order to obtain allowance of the claims. Mannesmann Demag Corp. v. Engineered Metal Products Co., 793 F.2d 1279, 1285, 230 USPQ 45, 48 (Fed.Cir.1986); Thomas and Betts Corp. v. Litton Sys., Inc., 720 F.2d 1572, 1579, 220 USPQ 1, 6 (Fed.Cir.1983). Thus, by actions taken during patent prosecution the patentee can be estopped from reaching *1219subject matter that otherwise meets the criteria of equivalency.
Pall states that there is no estoppel with respect to nylon 46, because no claim scope covering nylon 46 was yielded due to prior art or based on any requirement of patent examination. We outline the circumstances: Pali’s initial patent application contained broad claims, wherein the only limit placed on the nylon polyamide resin was that it be insoluble in alcohol. These claims stated no upper or lower limit to the methylene:amide ratio. Dr. Pall testified that upon continuing his research while his patent application was pending, he discovered that his view of the nylon resins usable in his invention was too broad. He eventually refiled' the patent application with additional data, and claims to nylon resins “within the range of about 5:1 to about 7:1.” Dr. Pall stated, and the prosecution record is in accord, that the refiling was based not on prior art or any requirement by the patent examiner, but on his own research.
The examiner for the continuation-in-part application at first rejected the new ratio claims under 35 U.S.C. § 112, stating that the claimed range was too broad and unsupported by the disclosure. No prior art was cited. Dr. Pall responded that the claimed range was “actually rather narrow” and that the claims “clearly exclude the vast majority of polyamide resins.” MSI relies on this statement as the basis of estoppel. MSI states that whether polyamide resins having a ratio below about 5:1 were required by the examiner to be surrendered, or were voluntarily not claimed by Pall, resins having a 4:1 ratio are excluded by estoppel because of this response to the examiner. Pall states that “about 5:1” was simply descriptive of the inventor’s knowledge at the time the continuation-in-part application was filed, and that there is no estoppel against the admittedly equivalent 4:1 resin.
Prosecution history estoppel normally arises when a change of claim scope is made in order to overcome an examiner’s rejection based on prior art. Estoppel may arise whether the change is made by amendment of the claims during prosecution, or by refiling the patent application with changed claims. Thus a patentee is estopped from recovering through equivalency that which was deemed unpatentable in view of the prior art. However, when a rejection based on prior art did not dictate the claim change that was made, it is necessary to look at the specific change and the reason, in ascertaining whether an estoppel has arisen by virtue of the change. Diversitech Corp. v. Century Steps, Inc., 850 F.2d 675, 681, 7 USPQ2d 1315, 1320 (Fed.Cir.1988). See Bayer Aktiengesellschaft v. Duphar Int’l Research B.V., 738 F.2d 1237, 1243, 222 USPQ 649, 653 (Fed.Cir.1984) (“a close examination must be made as to not only what was surrendered, but also the reason for such a surrender”); Hughes Aircraft Co. v. United States, 717 F.2d 1351, 1363, 219 USPQ 473, 481 (Fed.Cir. 1983) (an amendment to a claim may “have a limiting effect within a spectrum ranging from great to small to zero”). We take note that in the course of patent examination claims are often amended and rewritten and added and subtracted. A non-substantive change or a change that did not in fact determine patentability does not create an estoppel. Laitram Corp. v. NEC Corp., 952 F.2d 1357, 1361, 21 USPQ2d 1276, 1279 (Fed.Cir.1991).
The estoppel asserted by MSI arose not from prior art but from Pali’s statement to the examiner, in response to the rejection that the claims were too broad and unsupported by the disclosure, that the claims were “actually rather narrow.” There were no changes made in the claims in response to this rejection. - A rejection for lack of support in the specification is deemed a rejection under 35 U.S.C. § 112, first paragraph. In re Borkowski, 422 F.2d 904, 909, 164 USPQ 642, 646 (CCPA 1970). Section 112, first paragraph, states in part:
The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise, and exact terms as to enable any person skilled in the art to which it pertains, to make and use the same....
Whether amendment or argument made in response to a rejection under § 112 produces *1220an estoppel, as does an amendment made to obtain allowance in view of cited references, is dependent on the particular facts. There is no all-encompassing rule that estoppel results from all claim changes, or all arguments, whatever their cause or purpose. Laitram v. NEC, 952 F.2d at 1361, 21 USPQ2d at 1280; Hi-Life Products, Inc. v. American National Water-Mattress Corp., 842 F.2d 323, 325, 6 USPQ2d 1132, 1134 (Fed.Cir.1988).
As we have observed, a concession made or position taken to establish patentability in view of prior art on which the examiner has relied, is a substantive position on the technology for which a patent is sought, and will generally generate an estoppel. In contrast, when claim changes or arguments are made in order to more particularly point out the applicant’s invention, the purpose is to impart precision, not to overcome prior art. Such prosecution is not presumed to raise an estoppel, but is reviewed on its facts, with the guidance of precedent. See, e.g., Sun Studs, Inc. v. ATA Equipment Leasing, Inc., 872 F.2d 978, 990, 10 USPQ2d 1338, 1348 (Fed.Cir.1989) (rejection for lack of enablement, withdrawn based on the patentee’s arguments, did not serve as a basis for estoppel “in these events”); Moeller v. Ionetics, Inc., 794 F.2d 653, 659-60, 229 USPQ 992, 996-97 (Fed.Cir.1986) (“on this record” prosecution history estoppel not applicable to amendment to more particularly point out the invention); Mannesmann Demag, 793 F.2d at 1285, 230 USPQ at 48 (estoppel not necessarily created by an amendment designed only to remove a § 112 indefiniteness rejection); Caterpillar Tractor Co. v. Berco, S.p.A, 714 F.2d 1110, 1115, 219 USPQ 185, 187-88 (Fed.Cir.1983) (prosecution history estoppel not found where rejection based on indefiniteness under § 112, since prior art did not dictate the limitation at issue).
MSI argues that even if prosecution history estoppel does not limit Pall’s claims to precisely the 5:1 to 7:1 range, the doctrine of equivalents can be invoked only for subject matter that is disclosed and enabled in the patent specification but not claimed. That is incorrect. The doctrine serves to guard against “fraud on a patent,” Graver Tank, 339 U.S. at 608, 70 S.Ct. at 856, 85 USPQ at 330, by enabling fair protection of the paten-tee’s contribution. It is not controlling whether the inventor foresaw and described this potential equivalent at the time the patent application was filed. See, e.g., Atlas Powder Co. v. E.I. du Pont De Nemours & Co., 750 F.2d 1569, 1581, 224 USPQ 409, 417 (Fed.Cir.1984).
The district court applied these principles. The court found that Pall had not surrendered or abandoned patent protection outside of the 5:1 to 7:1 range, and that Pall’s voluntary restriction of the claims in the continuation-in-part application, and the response to the examiner’s rejection under § 112, did not produce an estoppel with respect to the 4:1 product. The court found that although nylon 46 had been known chemically since 1942, nylon 46 was not commercially available at the time of Dr. Pall’s invention. The court stated: “I do not find that nylon 46 was known in the sense that there was any requirement that Pall experiment with nylon 46 to understand its properties.” Observing the close similarity in chemical structure and the identity of function, way, and result to the claimed membranes, the district court found infringement by MSI’s nylon 46 membranes under the doctrine of equivalents.
We discern no reversible error in the application of the law of estoppel to the facts of this case, or in the court’s finding that the nylon 46 membranes infringed under the doctrine of equivalents. The judgment of infringement is affirmed.
The Process Claims
In view of our affirmance of infringement of the product claims for the nylon 66 and nylon 46 membranes, we need not reach the process claims. A patent is infringed if any claim is infringed, Intervet America, Inc. v. Kee-Vet Lab., Inc., 887 F.2d 1050, 1055, 12 USPQ2d 1474, 1477-78 (Fed.Cir.1989), for each claim is a separate statement of the patented invention. 35 U.S.C. § 282; Jones v. Hardy, 727 F.2d 1524, 1528, 220 USPQ 1021, 1024 (Fed.Cir.1984).
*1221 Willfulness
In the context of patent infringement, a distinction is drawn between infringement that is not deemed “willful” because the accused infringer had a reasonable basis for believing that its actions did not infringe the patent, and infringement that deliberately disregarded the property rights of the patentee. See Stickle v. Heublein, Inc., 716 F.2d 1550, 1565, 219 USPQ 377, 388 (Fed.Cir.1983).
Willfulness of infringement is a question of fact, for it includes elements of intent, reasonableness, and belief. See Richardson v. Suzuki Motor Co., 868 F.2d 1226, 1250, 9 USPQ2d 1913, 1932 (Fed.Cir.1989); King Instrument Corp. v. Otari Corp., 767 F.2d 853, 867, 226 USPQ 402, 412 (Fed.Cir.1985); Underwater Devices, Inc. v. Morrison-Knudsen Co., 717 F.2d 1380, 1390, 219 USPQ 569, 576-77 (Fed.Cir.1983). The boundary between unintentional and culpable acts is not always bright, see Rite-Hite Corp. v. Kelley Co., 819 F.2d 1120, 1125-26, 2 USPQ2d 1915, 1919 (Fed.Cir.1987), for the facts often include subjective as well as objective elements. Thus willful infringement must be established by clear and convincing evidence, E.I. du Pont de Nemours & Co. v. Phillips Petroleum Co., 849 F.2d 1430, 1440, 7 USPQ2d 1129, 1137 (Fed.Cir.), cert. denied, 488 U.S. 986, 109 S.Ct. 542, 102 L.Ed.2d 572 (1988), for it is a punitive finding, and can have the consequence of multiplication of damages.
The district court found that MSI’s infringement was not willful at its inception and for several years thereafter. The court found that although MSI failed to obtain an infringement opinion of counsel before producing its nylon 66 membranes, MSI made “significant efforts to avoid the claims of the patent.” Thus the district court found that MSI’s infringing production of nylon 66 membranes from 1983 until October 1989 was not willful. The court also found that MSI’s production of nylon 46 membranes after October 1989 did not constitute willful infringement. Pall does not appeal these findings.
In October 1989 MSI shifted most of its production to nylon 46 membranes. However, MSI continued to manufacture some membranes using nylon 66. The district court referred to MSI’s shift to nylon 46 as evidence that MSI knew that it had an alternative to its nylon 66 product, and found that MSI should have known by that time that its nylon 66 product infringed the Pall patent. The court stated that it assumed that MSI had acquired competent patent counsel, after three years of litigation. On these grounds the district court found that MSI’s failure to shift to nylon 46 for all of its production showed the requisite culpability to sustain willfulness, from that time forward, with respect to MSI’s continuing nylon 66 membrane production. The court doubled the damages for MSI’s infringing sales of nylon 66 after October 1989. MSI appeals these rulings.
MSI argues that the district court violated Federal Rule of Evidence 407 in finding that MSI’s continued production of nylon 66 membranes, after MSI switched to nylon 46, was willful infringement. Rule 407 bars evidence of subsequent remedial action in proving culpability for a prior act or event.5 The policy implemented by Rule 407 is to avoid inhibiting post-accident repair, lest additional injury occur. Herndon v. Seven Bar Flying Service, Inc., 716 F.2d 1322, 1327 (10th Cir.1983), cert. denied, 466 U.S. 958, 104 S.Ct. 2170, 80 L.Ed.2d 553 (1984). Here MSI’s conversion to nylon 46 was not relied on as evidence of culpability for the prior infringement by nylon 66 membranes, and the court declined to find willfulness for that prior infringement. Thus by its terms Rule 407 was not violated.
However, patent infringement is a continuing tort, and an action even if inno*1222cently begun does not automatically retain its purity as circumstances change. The filing of a lawsuit does not stop the clock insofar as culpability may arise from continuing disregard of the legal rights of the patentee. The district court relied on the fact that MSI switched to an alternative product three years into the litigation, yet continued to produce some of the nylon 66 membranes.
Willfulness of infringement after October 1989 is determined on the totality of the evidence. We do not hold that because the infringement was found not to be willful when begun, there is a greater burden on the patentee to prove willfulness as circumstances change. The requirement of law-abiding respect for the property of others is not reduced simply because lesser transgressions were overlooked. However, upon the ruling that willfulness was not established for MSI’s activities during the initial six years of infringement, the partial conversion to the nylon 46 product while continuing some of the nylon 66 production is not probative of willfulness for the nylon 66 continuing production. Attempts to avoid or mitigate infringement, whether or not successful, do not of themselves enlarge the culpability of the continuing activity.
The finding of willful infringement for the continuing production of nylon 66 is clearly in error. Absent other grounds for the finding of willfulness, this ruling and the award of damages attributed thereto are reversed.
DAMAGES
A
The district court awarded damages measured by Pali’s lost profits on 25% of MSI’s infringing sales, and a royalty of 8% on the remaining 75% of MSI’s infringing sales. The district court relied on the presence in the marketplace of nylon membranes sold by Cuno Corporation, with whom Pall had settled worldwide litigation with respect to the patent here in suit and several Cuno patents, each having charged the other with infringement. Pall states that it is entitled to lost profits on all of MSI’s infringing sales, at least until May 9, 1990, when Pal1 granted immunity to Cuno under the Pall patent.
In order to recover lost profits damages the patentee “must show a reasonable probability that, ‘but for’ the infringement, it would have made the sales that were made by the infringer.” Rite-Hite Corp. v. Kelley Co., 56 F.3d 1538, 1545, 35 USPQ2d 1065, 1069 (Fed.Cir.1995) (en banc); Yarway Corp. v. Eur-Control USA, Inc., 775 F.2d 268, 275, 227 USPQ 352, 357 (Fed.Cir.1985); King Instrument, 767 F.2d at 863, 226 USPQ at 409. In a market with only two suppliers, the patentee and the infringer, this requirement is readily met, for example by applying the guideline set forth in Panduit Corp. v. Stahlin Bros. Fibre Works, Inc., 575 F.2d 1152, 1156, 197 USPQ 726, 730 (6th Cir.1978). To establish lost profits by applying the evidentiary guideline of Panduit the patentee must show (1) that there was a demand for the patented product, (2) the absence of acceptable noninfringing substitutes, (3) that the patentee was capable of meeting the demand, and (4) the amount of profits lost. This guideline facilitates the determination of damages by stating conditions which allow the inference that the patentee would reasonably have made the infringer’s sales. See Del Mar Avionics, Inc. v. Quinton Instrument Co., 836 F.2d 1320, 1327, 5 USPQ2d 1255, 1260 (Fed.Cir.1987).
The district court, applying this guideline, held that the Cuno membranes were “acceptable noninfringing substitutes” for the Pall membranes. Pall states that this is incorrect, that the Cuno membranes were the subject of ongoing infringement litigation until May 1990, when they were licensed. The district court should have recognized the distinction between the legal and market situation before and after the licensing of the Cuno products. During the period before the Cuno products were licensed their presence in the marketplace did not defeat Pall’s entitlement to lost profits damages for all of MSI’s infringing sales, for the Cuno products were not “noninfringing substitutes.”
It was not necessary for Pall to continue the Cuno litigation to judicial decision of the issue of infringement, as MSI argues. The voluntary settlement of litigation does not retrospectively transform an accused in*1223fringing product into a “noninfringing substitute.” See State Indus., Inc. v. Mor-Flo Indus., Inc., 883 F.2d 1573, 1578, 12 USPQ2d 1026, 1029 (Fed.Cir.1989) (if the other suppliers “were likely infringers,” State would have been entitled to their shares of the market in determining damages based on lost sales), cert. denied, 493 U.S. 1022, 107 L.Ed.2d 744 (1990); Micro Motion, Inc. v. Kane Steel Co., 894 F.2d 1318, 1322, 13 USPQ2d 1696, 1699 (Fed.Cir.1990) (there is precedent for finding causation for lost profits damages if the alternative source of supply is an infringer); Datascope Corp. v. SMEC, Inc., 879 F.2d 820, 824-25, 11 USPQ2d 1321, 1323-24 (Fed.Cir.1989) (the finding that there was an acceptable noninfringing alternative can not be supported by the ongoing litigation between Datascope and a third party), cert. denied, 493 U.S. 1024, 110 S.Ct. 729, 107 L.Ed.2d 747 (1990).
However, after Pall settled with Cuno, the district court correctly held that Cuno’s presence in the marketplace could not be ignored, and limited the award of lost profits to the share of MSI’s sales that Pall would reasonably have made. For the remaining share the court viewed Pall as a hypothetical willing licensor, and accordingly assessed damages as an 8% royalty. This holding implements the reasoning that the purpose of compensatory damages is not to punish the in-fringer, but to make the patentee whole. See Aro Mfg. Co. v. Convertible Top Replacement Co., 377 U.S. 476, 507, 84 S.Ct. 1526, 1543, 12 L.Ed.2d 457, 141 USPQ 681, 694 (1964) (the question to be asked is “Had the infringer not infringed, what would the patent holder ... have made?”); State Industries, 883 F.2d at 1577, 12 USPQ2d at 1028 (damages shall compensate the patentee for its pecuniary loss because of the infringement). Pall argues that this result provides a windfall to the infringer, for when there are multiple infringers the infringement will be profitable to the remaining infringers as soon as the patentee settles with any one of them. We doubt that injustice is inevitable, for the setting of the royalty rate and the discretion to multiply damages can assist in achieving a just remedy.
In summary, for MSI’s sales after Pall settled with Cuno, we affirm the damages award based on recovery of Pall’s lost profits for 25% of MSI’s infringing sales, and a royalty of 8% for the remaining 75% of MSI’s infringing sales. For the period before the grant of immunity to Cuno, Pall shall recover damages measured as its lost profits for all of MSI’s infringing sales. We remand for the recalculation of damages.
B
The district court selected 45% as the rate of Pall’s lost profits. MSI does not challenge this rate on appeal. However, Pall appeals this determination, stating that the court selected an arbitrary rate that was lower than the actual rate that Pall had proved.
Among its reasons for reducing the rate from that presented by Pall’s accountants, the district court considered that MSI’s prices were lower than Pali’s, concluded that some of MSI’s customers would have shifted to membranes other than nylon, and factored this into its reduction of Pall’s estimated profits. Pall disputes the presence as well as the significance of asserted price differentials, and challenges the other premises of the district court’s decision.
In general, an infringer’s sales at a lower price do not defeat the patentee’s recovery of its losses at the patentee’s price, for the principle of patent damages is to return the patentee to the pecuniary position it would have been in but for the infringement. Aro Manufacturing, 377 U.S. at 507, 84 S.Ct. at 1543, 141 USPQ at 694. See also Rite-Hite, 56 F.3d at 1544, 35 USPQ2d at 1068 (“The statute thus mandates that a claimant receive damages ‘adequate’ to compensate for infringement.”) However, in view of the complex of financial factors presented at trial, we discern no clear error in the district court’s finding of the rate of Pall’s lost profits. The district court’s ruling on this aspect is affirmed.
Motions
MSI moves to strike the opinion of the district court that was written on April *122414,1992. That opinion was an edited form of the court’s prior opinion from the bench, the district court having reserved the right to present a written opinion. Although MSI objected to Pall’s proposed substitute opinion, and to the transmittal to this court of a “substitute opinion, of any kind,” it made no objection to that opinion after it was forwarded by the district court and did not move to strike it until two years after the district court’s edited opinion was issued. MSI offers no explanation for the delay. The motion is denied.
Pall in turn states that the court should strike MSI’s argument on the merits, raised in MSI’s motion to strike the opinion of the district court. In view of our disposition herein, the motion is dismissed as moot.
Injunction
The stay of injunction pending appeal is vacated.
Costs
Each party shall bear its costs.
AFFIRMED IN PART, REVERSED IN PART, MODIFIED IN PART, AND REMANDED. STAY VACATED.
. Pall Corp. v. Micron Separations, Inc., No. 86-1427-Y (D.Mass. June 24, 1991). On Nov. 5, 1993 the court decided to hear this case en banc in order to consider the questions set forth in the court's order of Jan. 7, 1994. Having resolved those questions in the case of Markman v. Westview Instruments, Inc., 52 F.3d 967, 34 USPQ2d 1321 (Fed.Cir.1995) (en banc), the case has been returned to the panel for disposition in accordance therewith.
. In this common nomenclature for nylon po-lyamides, integers signify the number of carbon atoms in the monomers from which the polymer is made. For example, in nylon 66 the first 6 represents the number of carbon atoms derived from the diamine monomer, and the second 6 represents the number of carbon atoms in the diacid monomer. Nylon 46 has 4 carbon atoms in the diamine and 6 carbon atoms in the diacid.
. In re Micron Separations, Inc., No. 91-41885-JFQ (Bankr.D.Mass.).
. Pall Corp. v. Micron Separations, Inc., No. 86-1427-Y, slip op. at 1.
. Rule 407. When, after an event, measures are taken which, if taken previously, would have made the event less likely to occur, evidence of the subsequent measures is not admissible to prove negligence or culpable conduct in connection with the event. This rule does not require the exclusion of evidence of subsequent measures when offered for another purpose, such as proving ownership, control, or feasibility of precautionary measures, if controverted, or impeachment.