LINN, Circuit Judge.

Net MoneyIN, Inc. (“NMI”) appeals from a final judgment of the United States District Court for the District of Arizona, which held the asserted claims of U.S. Patents No. 5,822,737 (“the '737 patent”) and No. 5,963,917 (“the '917 patent”) invalid. NMI also appeals from the district court’s denial of its motion for leave to amend its complaint to assert a claim for inducement of infringement. Because the district court correctly found claims 1, 13, and 14 of the '737 patent and claim 1 of the '917 patent, which contain limitations in means-plus-function format, invalid under 35 U.S.C. § 112 ¶2 as lacking corresponding structure, we affirm that portion of the judgment. Because the district court did not abuse its discretion in denying NMI’s motion to amend, we also affirm that ruling. Because the district court applied an incorrect standard of law in finding claim 23 of the '737 patent invalid as anticipated under 35 U.S.C. § 102(a), however, we reverse the grant of summary judgment of anticipation. Thus, we affirm-in-part, reverse-in-part, and remand for proceedings consistent with this opinion.

I. BACKGROUND

This case involves systems for processing credit card transactions over the Internet and for addressing security concerns not present in direct retail transactions. In the early days of Internet commerce, merchants recognized that one key to the *1363success of Internet sales would be the ability to provide customers with assurances of security in the processing of financial transactions over the Internet using credit cards, bank accounts; and other means of electronic payment. Responding to that need, the industry investigated encryption techniques and architectures to protect sensitive data. One such effort is reflected in a 1995 working document entitled “Internet Keyed Payments Protocol” (“the iKP reference”), published by the Internet Engineering Task Force and IBM. That document sets forth standards on “how payments may be accomplished efficiently, reliably[,] and securely.” J.A. at 1375. The iKP reference explains that its goal was “to enable Internet-based secure electronic payments while utilizing the existing financial infrastructure for payment authorization and clearance. The intent is to avoid completely, or at least minimize, changes to the existing financial infrastructure outside the Internet.” Id. To that end, the iKP reference suggests two standard models, or protocols.1

In the first protocol, (1) the customer selects one or more items to purchase from the merchant’s website; (2) the customer sends credit card information to the merchant; (3) the merchant sends the credit card information and amount of the purchase to the merchant’s bank; (4) the merchant’s bank seeks authorization for the purchase from the issuing bank over the existing banking network; and (5) the merchant’s bank notifies the merchant (but not the customer) of transaction approval. See id. at 1381 (flow diagram); Appellant’s Br. at 7.

In the second protocol, (1) the customer selects one or more items to purchase on the merchant’s website; (2) the customer sends an authorization request, along with its credit card information and the amount of the purchase, to the merchant’s bank; (3) the merchant’s bank seeks authorization from the issuing bank over the existing banking network; (4) the merchant’s bank notifies the customer of transaction approval; and (5) the customer sends the authorization response to the merchant. See J.A. at 1342, 1394; Appellant’s Br. at 8-9.

Unsatisfied with the early approaches taken by others, Mark Ogram, an inventor and patent attorney, set out to create a new payment model to remedy what he perceived as two deficiencies in the prior art protocols: “the fact that the customer had to send confidential information over the Internet to an unknown merchant; and the fact -that credit card issuers imposed onerous financial requirements on merchants.” Appellant’s Br. at 10. Ogram’s idea was to add a fifth entity, a “payment processing” or “financial processing” entity, to supplement the conventional model with four entities: the customer, merchant, merchant’s bank, and issuing bank. According to Ogram, the new financial processing entity would: “(1) receive credit card account information and an amount to be charged from the customer when the customer placed the order; (2) seek authorization from the card issuer over the existing banking network; and (3) notify both the customer and the merchant of authorization.” Id.

On February 5, 1996, Ogram filed a patent application directed to a payment model utilizing a financial processing entity. He formed NMI shortly thereafter to *1364implement the model as a business for processing credit card transactions over the Internet. Ogram’s patent application resulted in the '737 and '917 patents, both of which are assigned to NMI. Claim 1 of the '737 patent is illustrative of the invention claimed:

1. A financial transaction system comprising:

a) a first bank computer containing financial data therein, said financial data including customer account numbers and available credit data, said first bank computer including means for generating an authorization indicia in response to queries containing a customer account number and amount;

b) a merchant computer containing promotional data;

c) a customer computer being linked with said merchant computer and receiving said promotional data; and,

d) a financial processing computer remote from said merchant computer and having means for:

1) receiving customer account data and amount data from said customer computer,

2) querying said first bank computer with said customer account data and said amount data,

3) receiving an authorization indicia from said first bank computer,

4) communicating a self-generated transaction indicia to said customer computer, and,

5) communicating the self-generated transaction indicia to said merchant computer.

According to their abstracts, the '737 and '917 patents relate to “[a]n automated payment system particularly suited for purchases over a distributed computer network such as the Internet.”

In 2001, NMI filed suit for infringement of the '737 and '917 patents against a number of parties alleged to compete in the Internet credit card processing field, including VeriSign, Inc. and eProcessing-Network (collectively, “VeriSign”). Following a claim construction hearing, the district court construed a number of terms in dispute. Net MoneyIN, Inc. v. VeriSign, Inc., No. 01-CV-441, 2005 WL 5960650 (D.Ariz. Oct. 19, 2005) (“Claim Construction Decision”). As part of its construction of the claim terms, the district court invalidated claims 1, 13, and 14 of the '737 patent and claim 1 of the '917 patent, which contain limitations in means-plus-function format, as lacking corresponding structure and thus indefinite under 35 U.S.C. § 112 ¶ 2.

Following construction of the claims, the district court entertained two motions for summary judgment that are relevant to this appeal. First, VeriSign moved for summary judgment that it did not induce infringement of NMI’s patents. In response to that motion, NMI moved for leave to amend its complaint to add a claim for inducement of infringement. The district court granted VeriSign’s motion for summary judgment and denied NMI’s motion for leave to amend. Net MoneyIN, Inc. v. VeriSign, Inc., No. 01-CV-441 (D. Ariz. June 8, 2006) (“Amendment Decision ”). Second, VeriSign moved for summary judgment of invalidity, arguing that the iKP reference anticipated claim 23 of the '737 patent under 35 U.S.C. § 102(a). The district court granted VeriSign’s motion. Net MoneyIN, Inc. v. VeriSign, Inc., No. 01-CV-441 (D.Ariz. July 13, 2007) (“Summary Judgment Decision”). The district court then entered final judgment in favor of VeriSign.

NMI timely appealed. We have jurisdiction under 28 U.S.C. § 1295(a)(1).

*1365II. DISCUSSION

A. Standard of Review

Claim construction is a question of law, Markman v. Westview Instruments, Inc., 52 F.3d 967, 970-71 (Fed.Cir. 1995) (en banc), aff'd, 517 U.S. 370, 116 S.Ct. 1384, 134 L.Ed.2d 577 (1996), over which we exercise plenary review. Cybor Corp. v. FAS Techs., Inc., 138 F.3d 1448, 1456 (Fed.Cir.1998) (en banc). Indefiniteness under 35 U.S.C. § 112 ¶2 is also a question of law subject to plenary review. SmithKline Beecham Corp. v. Apotex Corp., 403 F.3d 1331, 1338 (Fed.Cir.2005).

We review a grant of summary judgment de novo, reapplying the standard that the district court employed. Rodime PLC v. Seagate Tech., Inc., 174 F.3d 1294, 1301 (Fed.Cir.1999). Drawing all reasonable inferences in favor of the nonmovant, “[sjummary judgment is appropriate only when ‘there is no genuine issue as to any material fact and ... the moving party is entitled to a judgment as a matter of law.’ ” Id. (quoting Fed.R.Civ.P. 56(c)).

The denial of a motion to amend is a procedural question not unique to patent law and thus is reviewed under the law of the regional circuit. Kalman v. Berlyn Corp., 914 F.2d 1473, 1480 (Fed. Cir.1990). In the Ninth Circuit, the denial of a motion to amend is reviewed for abuse of discretion. Chappel v. Lab. Corp. of Am., 232 F.3d 719, 725 (9th Cir.2000).

B. Analysis

1. Indefiniteness

The district court concluded that claims 1, 13, and 14 of the '737 patent and claim 1 of the '917 patent were indefinite under 35 U.S.C. § 112 ¶2, and thus invalid. Because each of these patents raises different issues, we address them separately.

a. the '737 patent

Claim 1 of the '737 patent recites a financial transaction system comprising, among other things, “a first bank computer containing financial data therein, said financial data including customer account numbers and available credit data, said first bank computer including means for generating an authorization indicia in response to queries containing a customer account number and amount” (emphasis added).2 The district court construed the generating means element in claim 1 as a means-plus-function element. The parties agreed that the function of the claimed means was “generating an authorization indicia in response to queries containing a customer account number and amount.” The district court found, however, that the specification failed to disclose any corresponding structure to perform the claimed function. Accordingly, it deemed the claim invalid under 35 U.S.C. § 112 ¶ 2.

NMI argues that the generating means element is not a means-plus-function element under 35 U.S.C. § 112 ¶ 6 “because the claim itself discloses sufficient structure which performs the function of ‘generating an authorization indicia in response to queries containing a customer account number and amount.’ ” Appellant’s Br. at 21 (emphasis omitted). Alternatively, NMI contends that if the generating means claim element is properly construed as a means-plus-funetion element, then the specification recites sufficient structure to make the claim definite. VeriSign counters that the district court correctly concluded both that the claim does not recite sufficient structure to rebut the means-plus-function presumption and that the specification contains insufficient structure to perform the claimed function.

*1366Section 112, paragraph 6, of title 35 provides that:

An element of a claim for a combination may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure, material, or acts described in the specification and equivalents thereof.

A claim element that contains the word “means” and recites a function is presumed to be drafted in means-plus-function format under 35 U.S.C. § 112 ¶ 6. Envirco Corp. v. Clestra Cleanroom, Inc., 209 F.3d 1360, 1364 (Fed.Cir.2000). The presumption is rebutted, however, “if the claim itself recites sufficient structure to perform the claimed function.” M; see also Sage Prods., Inc. v. Devon Indus., Inc., 126 F.3d 1420, 1427-28 (Fed.Cir.1997) (“[W]here a claim recites a function, but then goes on to elaborate sufficient structure, material, or acts within the claim itself to perform entirely the recited function, the claim is not in means-plus-function format.”).

We first address NMI’s contention that the presumption triggered by the presence of the word “means” in claim 1 is rebutted by the recitation of sufficient structure for performing entirely the recited function of “generating an authorization indicia.” NMI contends that the language, “first bank computer containing financial data therein, said financial data including customer account numbers and available credit data, said first bank computer ... generating an authorization indicia in response to queries containing a customer account number and amount,” is sufficient structure to rebut the means-plus-function presumption. NMI argues that an ordinary artisan would understand the “bank computer” “to be a commonly known structure for generating authorization indi-cia in response to queries containing a custom account number and amount.” Appellant’s Br. at 21, 22. VeriSign responds that the claim does not recite sufficient structure to rebut the presumption “because of the wide variety of types and classes of computers in existence, each being configurable in a variety of different ways using many different programming languages.” Appellees’ Br. at 28 (internal quotation marks and citations omitted).

We agree with VeriSign that the recitation in claim 1 of the “bank computer” is not sufficient to rebut the means-plus-function presumption. The bank computer is not linked in the claim as the “means” for generating an authorization indicia. Rather, the bank computer is recited as “including” those means. NMI’s argument that the first bank computer constitutes sufficient structure would require the first bank computer to include a first bank computer, which is both redundant and illogical. Because the claimed generating means is a subset of the bank computer, there must be a recitation of structure that is a component of the bank computer to rebut the presumption. The claim contains no such recitation. As a result, the district court correctly concluded that the presumption of means-plus-function treatment had not been overcome.

Having concluded that the generating means recited in claim 1 is drafted in means-plus-function format, we turn to whether the specification includes a disclosure of structure sufficient to accomplish the recited function. NMI argues that “the specification does disclose a ‘bank computer’ and this Court’s precedents do not require a description of the ‘internal structure’ of the ‘bank computer.’ ” Appellant’s Br. at 27 (emphasis omitted); see also id. at 31 (“Here, claim 1(a) itself states that the ‘bank computer’ contains ‘financial data’ including ‘customer account numbers and available credit data.’ A per*1367son skilled in the art would know that such a computer would be programmed to compare account data and amount data to those data structures and generate an authorization indicia if credit were available”). VeriSign counters that the district court correctly determined that the '737 patent specification fails to disclose the “structure corresponding to what, in the claimed first bank computer, performs the claimed generating function.” Appellees’ Br. at 51 (internal quotation marks omitted).

A patent applicant who employs means-plus-function language “must set forth in the specification an adequate disclosure showing what is meant by that language. If an applicant fails to set forth an adequate disclosure, the applicant has in effect failed to particularly point out and distinctly claim the invention as required by the second paragraph of section 112.” In re Donaldson Co., 16 F.3d 1189, 1195 (Fed.Cir.1994) (en banc). To avoid purely functional claiming in cases involving computer-implemented inventions, ‘ we have “consistently required that the structure disclosed in the specification be more than simply a general purpose computer or microprocessor.” Aristocrat Techs. Austl. Pty Ltd. v. Int’l Game Tech., 521 F.3d 1328, 1333 (Fed.Cir.2008). “Because general purpose computers can be programmed to perform very different tasks in very different ways, simply disclosing a computer as the structure designated to perform a particular function does not limit the scope of the claim to ‘the corresponding structure, material, or acts’ that perform the function, as required by section 112 paragraph 6.” Id. “Thus, in a means-plus-function claim ‘in which the disclosed structure is a computer, or microprocessor, programmed to carry out an algorithm, the disclosed structure is not the general purpose computer, but rather the special purpose computer programmed to perform the disclosed algorithm.’ ” Id. (quoting WMS Gaming, Inc. v. Int’l Game Tech., 184 F.3d 1339, 1349 (Fed.Cir.1999)). Consequently, a means-plus-function claim element for which the only disclosed structure is a general purpose computer is invalid if the specification fails to disclose an algorithm for performing the claimed function. See id. at 1337-38.

There is no dispute in this case that the specification fails to disclose an algorithm by which a general purpose bank computer “generates] an authorization indicia.”3 As a result, the district court correctly concluded that claims 1, 13, and 14 are indefinite under 35 U.S.C. § 112 ¶2. We therefore affirm that part of the judgment.

b. the '917 patent

Claim 1 of the '917 patent recites a financial transaction system comprising, among other things, “a financial processing computer ... having automatic means responsive to [the] order for ... receiving customer account data and amount data from [the] customer computer and [the] merchant computer.” The parties do not dispute the district court’s construction of this claim element as a means-plus-function element. The parties do dispute, however, the nature of the function. The district court construed the function as “the financial processing computer receives both the customer account data and amount data from both the customer computer and the merchant computer.” Claim Construction Decision at 11. NMI argues that the district court’s construc*1368tion of the function is erroneous. According to NMI, the ordinary meaning of the claim language requires that the function be construed more broadly: “[I]n response to an order, the financial processing computer: (1) receives customer account data from the customer computer, the merchant computer, or both; and (2) it also receives amount data from the customer computer, the merchant computer, or both.” Appellant’s Br. at 46-47. VeriSign counters that the district court correctly construed the function according to the ordinary meaning of the claim language.

The language of the function at issue was construed by the district court as specifying that both the amount data and the account data must come from both the customer computer and the merchant computer. That construction comports with and is fully supported by the language of the claim itself. NMI argues that the function is subject to a different construction, which would permit the amount data and the account data to come from the merchant computer, the customer computer, or both. The problem with NMI’s proffered construction, however, is that it is different from, and broader in scope than, the construction it asserted in the district court. See J.A. at 1046 (arguing to the district court that “the meaning of this element is: ‘the financial processing computer receives the customer account data from the customer computer and the amount data from the merchant computer via' the customer computer’ ”). This is not merely a new argument in support of a previously presented construction, but instead is a new and more expansive construction, which may not properly be asserted on appeal. See Interactive Gift Express, Inc. v. Compuserve Inc., 256 F.3d 1323, 1347 (Fed.Cir.2001). Because NMI’s new construction is not proper on appeal, and because we see no basis on which to overturn the district court’s construction, that construction is affirmed.

NMI concedes that under the district court’s construction, no structure is disclosed in the specification to perform the claimed function. Appellant’s Br. at 46. As a result, the claim is indefinite under 35 U.S.C. § 112 ¶ 2. See Donaldson, 16 F.3d at 1195. Consequently, we affirm the district court’s determination that claim 1 of the '917 patent is invalid.

2. Anticipation

Claim 23 of the '737 patent recites an Internet payment system comprising five “links”:

a) a first link between a customer computer and a vending computer for communicating promotional information from said vending computer to said customer computer;

2) a second link, initiated by said customer computer, between said customer computer and a payment processing computer, remote from said vending computer, for communicating credit card information and amount from said customer computer to said payment processing computer;

3) a third link, initiated by said payment processing computer with a credit card server computer for communicating said credit card information and said amount from said payment processing computer to said credit card server computer, and for communicating, in response, an authorization indicia from said credit card server computer to said payment processing computer; [ ]

d) a fourth link between said payment processing computer and said customer computer for communicating a transactional indicia[;]

*1369[e)] a fifth link between the payment processing computer and said vending computer for communicating said transactional indicia.

The district court, after finding all five of these links in the iKP reference, albeit in two separate disclosed examples, concluded that claim 23 was anticipated under 35 U.S.C. § 102(a) and therefore invalid. Specifically, the district court concluded:

All of the limitations of claim 23 can be found within the iKP reference. A simple combination would produce the system described in claim 23 of the '737 patent. That no specific example within iKP contains all five links does not preclude a finding of anticipation.

Summary Judgment Decision at 3. NMI contends that the district court’s combination of two disclosed examples in order to find all elements of the claim was erroneous.4 VeriSign responds that the district court did not improperly rearrange the links in the iKP reference, but rather “merely relied on various express teachings from a single document that together completely disclose the five claimed links.” Appellees’ Br. at 61. Under VeriSign’s theory, this was sufficient to establish anticipation, because all that is required is “that the four corners of a single, prior art document describe every element of the claimed invention.” Id. at 61-62 (quoting Xerox Corp. v. 3Com Corp., 458 F.3d 1310, 1322 (Fed.Cir.2006)). We disagree with VeriSign, and take this opportunity to clarify what a reference must show in order to anticipate a claimed invention.

Section 102(a) provides that an issued patent is invalid if “the invention [therein] was ... described in a printed publication ... before the invention thereof by the applicant.” Section 102 embodies the concept of novelty — if a device or process has been previously invented (and disclosed to the public), then it is not new, and therefore the claimed invention is “anticipated” by the prior invention. As we have stated numerous times (language on which VeriSign relies), in order to demonstrate anticipation, the proponent must show “that the four corners of a single, prior art document describe every element of the claimed invention.” Xerox, 458 F.3d at 1322 (quoting Advanced Display Sys., Inc. v. Kent State Univ., 212 F.3d 1272, 1282 (Fed.Cir.2000)). This statement embodies the requirement in section 102 that the anticipating invention be “described in a printed publication,” and is, of course, unimpeachable. But it does not tell the whole story. Because the hallmark of anticipation is prior invention, the prior art reference — in order to anticipate under 35 U.S.C. § 102 — must not only disclose all elements of the claim within the four corners of the document, but must also disclose those elements “arranged as in the claim.” Connell v. Sears, Roebuck & Co., 722 F.2d 1542, 1548 (Fed.Cir.1983).5

*1370The meaning of the expression “arranged as in the claim” is readily understood in relation to claims drawn to things such as ingredients mixed in some claimed order. In such instances, a reference that discloses all of the claimed ingredients, but not in the order claimed, would not anticipate, because the reference would be missing any disclosure of the limitations of the claimed invention “arranged as in the claim.” But the “arranged as in the claim” requirement is not limited to such a narrow set of “order of limitations” claims. Rather, our precedent informs that the “arranged as in the claim” requirement applies to all claims and refers to the need for an anticipatory reference to show all of the limitations of the claims arranged or combined in the same way as recited in the claims, not merely in a particular order. The,test is thus more accurately understood to mean “arranged or combined in the same way as in the claim.”

For example, in Lindemann Maschinenfabrik GMBH v. American Hoist & Derrick Co., 730 F.2d 1452 (Fed.Cir.1984), we reviewed a district court’s determination that a patent directed to a hydraulic scrap shearing machine was anticipated by a prior patent directed to a method for shearing spent nuclear fuel bundles. Because the district court had “treated the claims as mere catalogs of separate parts, in disregard of the part-to-part relationships set forth in the claims and that give the claims their meaning,” we reversed. Id. at 1459. Although the prior art reference could be said to contain all of the elements of the claimed invention, it did not anticipate under 35 U.S.C. § 102 because it “disclose[d] an entirely different device, composed of parts distinct from those of the claimed invention, and operating in a different way to process different material differently.” Id. at 1458. The reference thus was deficient because it did not disclose the elements of the claimed invention “arranged as in the claim” as required by 35 U.S.C. § 102. Id.

In Ecolochem, Inc. v. Southern California Edison Co., 227 F.3d 1361 (Fed.Cir. 2000), we reviewed a district court’s decision that a prior art reference directed to “Saving Energy by Catalytic Reduction of Oxygen in Feedwater” anticipated a claim reciting the use of hydrazine with a mixed resin bed to deoxygenate water. In finding that the reference anticipated the claim, the district court considered a figure and accompanying text, which taught the use of hydrogen with a mixed bed to deox-ygenate water, in conjunction with a separate passage discussing deoxygenating water with, among other things, hydrazine. Id. at 1369. We reversed. After determining that the relevant figure and accompanying text described only the use of hydrogen to deoxygenate water, we concluded that the reference could not anticipate the claimed invention because there was no link between that figure and the general discussion of hydrazine as a deoxy-genating agent. Id. In other words, we concluded that although the reference taught all elements of the claim, it did not contain a discussion suggesting or linking hydrazine with the mixed bed in the figure, and thus did not show the invention arranged as in the claim.

Recently, in Finisar Corp. v. DirecTV Group, Inc., 523 F.3d 1323 (Fed.Cir. 2008), we reversed a district court’s denial of a motion for judgment as a matter of law because the jury could not have reasonably concluded that the prior art reference relating to the Videotex architecture did not anticipate the claimed invention directed to systems and methods for scheduling transmission of database tiers on demand at varying repetition rates. Although the anticipation issue dealt largely with the interpretation of the prior art reference, id. at 1335-37, we reemphasized *1371the importance of the requirement that the reference describe not only the elements of the claimed invention, but also that it describe those elements “arranged as in the claim”:

To anticipate a claim, a single prior art reference must expressly or inherently disclose each claim limitation.... But disclosure of each element is not quite enough — this court has long held that “[a]nticipation requires the presence in a single prior art disclosure of all elements of a claimed invention arranged as in the claim.”

Id. at 1334 (quoting Connell, 722 F.2d at 1548). In all of these cases, the prior art reference had to show the claimed invention arranged or combined in the same way as recited in the claim in order to anticipate. We thus hold that unless a reference discloses within the four corners of the document not only all of the limitations claimed but also all of the limitations arranged or combined in the same way as recited in the claim, it cannot be said to prove prior invention of the thing claimed and, thus, cannot anticipate under 35 U.S.C. § 102.

Here, the iKP reference discloses two separate protocols for processing an Internet credit card transaction. Neither of these protocols contains all five links arranged or combined in the same way as claimed in the '737 patent. Thus, although the iKP reference might anticipate a claim directed to either of the two protocols disclosed, it cannot anticipate the system of claim 23. The district court was wrong to conclude otherwise.

The district court was also wrong to combine parts of the separate protocols shown in the iKP reference in concluding that claim 23 was anticipated. Granted, there may be only slight differences between the protocols disclosed in the iKP reference and the system of claim 23. But differences between the prior art reference and a claimed invention, however slight, invoke the question of obviousness, not anticipation. See 35 U.S.C. § 103(a) (“A patent may not be obtained though the invention is not identically disclosed or described as set forth in section 102 of this title, if the differences between the subject matter sought to be patented and the prior art are such that the subject matter as a whole would have been obvious at the time the invention was made to a person having ordinary skill in the art to which said subject matter pertains.” (emphasis added)); see also In re Arkley, 59 C.C.P.A. 804, 455 F.2d 586, 587 (1972) (“[R]ejections under 35 U.S.C. .§ 102 are proper only when the claimed subject matter is identically disclosed or described in the prior art.” (emphasis and internal quotation marks omitted)). Thus, it is not enough that the prior art reference discloses part of the claimed invention, which an ordinary artisan might supplement to make the whole, or that it includes multiple, distinct teachings that the artisan might somehow combine to achieve the claimed invention. See Arkley, 455 F.2d at 587 (“[T]he [prior art] reference must clearly and unequivocally disclose the claimed [invention] or direct those skilled in the art to the [invention] without any need for picking, choosing, and combining various disclosures not directly related to each other by the teachings of the cited reference.”).

Because the parties do not contend that the iKP reference discloses all of the limitations recited in claim 1 arranged or combined in the same way as in the claim, and because it was error for the district court to find anticipation by combining different parts of the separate protocols in the iKP reference simply because they were found within the four corners of the document, we reverse the district court’s grant of summary judgment of invalidity.

*13723. Motion to Amend

During the course of this litigation, NMI filed a Second Amended Complaint in which it clarified that it was asserting, among other things, a claim for inducement of infringement under 35 U.S.C. § 271(b). In a Third Amended Complaint filed in August 2003, however, NMI abandoned its claim for inducement of infringement, stating that it had “elected not to assert a cause of action for inducement.” In answering NMI’s Third Amended Complaint, although some of the defendants reasserted counterclaims for declaratory judgment of noninfringement by inducement, VeriSign did not. J.A. at 1251.

In May 2005, VeriSign moved for partial summary judgment on inducement of infringement. In response, NMI moved, pursuant to Federal Rule of Civil Procedure 15(b), for leave to file a Fourth Amended Complaint to add a claim for inducement of infringement. According to NMI, VeriSign had consented to litigate the issue by moving for summary judgment on that basis. The district court granted VeriSign’s motion for partial summary judgment and denied NMI’s motion to amend.

NMI argues that the district court abused its discretion by denying the motion to amend. According to NMI, the district court had no discretion to deny amendment under Rule 15(b) because Ver-iSign consented to litigate the issue. Veri-Sign argues that, while it did seek partial summary judgment on the issue of inducement, it did so on the ground of waiver, not on the merits. Thus, according to VeriSign, it was within the district court’s discretion to deny amendment.

A district court generally enjoys broad discretion when assessing the propriety of a motion to amend. See Chappel, 232 F.3d at 725. It does not enjoy such discretion, however, and amendment is mandatory, when an issue is tried with the express or implied consent of the parties. See Fed.R.Civ.P. 15(b) (“When an issue not raised by the pleadings is tried by the parties’ express or implied consent, it must be treated in all respects as if raised in the pleadings.”); cf. Wallin v. Fuller, 476 F.2d 1204, 1210 (5th Cir.1973) (“Amendment is thus not merely discretionary but mandatory in such a case.”).

Thus, the first issue we must address is whether VeriSign consented, either expressly or impliedly, to litigate inducement. We agree with the district court that it did not. VeriSign’s motion for partial summary judgment stated, in pertinent part,

NMI’s failure to assert any claim of contributory infringement under Section 271(c) in its Third Amended Complaint, its express disavowal in that pleading of any claim of inducement under Section 271(b), its failure to disclose any indirect infringement theories or supporting evidence in its Supplemental Disclosure to VeriSign, and its failure to disclose any evidence that would establish indirect infringement, including its failure to identify any alleged direct infringer or any acts by VeriSign alleged to constitute contributory infringement, compels entry of partial summary judgment in favor of VeriSign.

J.A. at 7330. This is not an attempt by VeriSign to litigate induced infringement on the merits. Given NMI’s repeated amendment of its complaint, including its history of dropping inducement claims only to later add them, as well as VeriSign’s understanding that NMI planned to resurrect the claim at trial, it is apparent that VeriSign’s motion was made to foreclose NMI’s ability to later raise inducement (again). Notably, this is precisely how the district court construed NMI’s motion:

Plaintiff uses the argument that by filing [a] motion for summary judgment on *1373this issue, the Defendants are consenting to its litigation. This is not the case. Defendants are merely attempting to formally discharge this theory as a claim (as has already been indicated by Plaintiffs counsel) so that the case can be focused on the theory of direct infringement.

Amendment Decision at 7. Thus, the district court was not without discretion to deny the requested amendment.

The question thus becomes whether the district court’s denial of the motion was an abuse of that discretion. In denying the motion, the district court observed that NMI was requesting leave “to amend [its] Complaint for a fourth time in order to allege a claim (inducement of infringement) which [it] ha[s] expressly disavowed, twenty months after the deadline to amend pleadings and four months after the close of discovery.” Id. at 6-7. It also observed that granting NMI’s motion would result in “extreme delay,” id. at 10, and severe prejudice to VeriSign, id. at 11. Under these circumstances, we cannot find that the district court’s denial was an abuse of discretion. See Chappel, 232 F.3d at 725-26 (“A district court acts within its discretion to deny leave to amend when amendment would be futile, when it would cause undue prejudice to the defendant, or when it is sought in bad faith.”).

III. CONCLUSION

For the foregoing reasons, we AFFIRM-IN-PART, REVERSE-IN-PART, and REMAND for proceedings consistent with this opinion.

COSTS

No costs.

*57Justice Stevens

delivered the opinion of the Court.

Section 102(b) of the Patent Act of 1952 provides that no person is entitled to patent an “invention” that has been “on sale” more than one year before filing a patent application.1 We granted certiorari to determine whether the commercial marketing of a newly invented product may mark the beginning of the 1-year period even though the invention has not yet been reduced to practice.2

rH

On April 19, 1982, petitioner, Wayne Pfaff, filed an application for a patent on a computer chip soeket. Therefore, April 19, 1981, constitutes the critical date for purposes of the on-sale bar of 35 TJ. S. C. § 102(b); if the 1-year period *58began to run before that date, Pfaff lost his right to patent his invention.

Pfaff commenced work on the socket in November 1980, when representatives of Texas Instruments asked him to develop a new device for mounting and removing semiconductor chip carriers. In response to this request, he prepared detailed engineering drawings that described the design, the dimensions, and the materials to be used in making the socket. Pfaff sent those drawings to a manufacturer in February or March 1981.

Prior to March 17,1981, Pfaff showed a sketch of his concept to representatives of Texas Instruments. On April 8, 1981, they provided Pfaff with a written confirmation of a previously placed oral purchase order for 30,100 of his new sockets for a total price of $91,155. In accord with his normal practice, Pfaff did not make and test a prototype of the new device before offering to sell it in commercial quantities.3

The manufacturer took several months to develop the customized tooling necessary to produce the device, and Pfaff did not fill the order until July 1981. The evidence therefore indicates that Pfaff first reduced his invention to practice in the summer of 1981. The socket achieved substantial com*59mercial success before Patent No. 4,491,377 (’377 patent) issued to Pfaff on January 1, 1985.4

After the patent issued, petitioner brought an infringement action against respondent, Wells Electronics, Inc., the manufacturer of a competing socket. Wells prevailed on the basis of a finding of no infringement.5 When respondent began to market a modified device, petitioner brought this suit, alleging that the modifications infringed six of the claims in the ’377 patent.

After a full evidentiary hearing before a Special Master,6 the District Court held that two of those claims (1 and 6) were invalid because they had been anticipated in the prior art. Nevertheless, the court concluded that four other claims (7,10,11, and 19) were valid and three (7,10, and 11) were infringed by various models of respondent’s sockets. App. to Pet. for Cert. 21a-22a. Adopting the Special Master’s findings, the District Court rejected respondent’s § 102(b) defense because Pfaff had filed the application for the ’377 patent less than a year after reducing the invention to practice.

The Court of Appeals reversed, finding all six claims invalid. 124 F. 3d 1429 (C A Fed. 1997). Four of the claims (1, 6,7, and 10) described the socket that Pfaff had sold to Texas Instruments prior to April 8,1981. Because that device had been offered for sale on a commercial basis more than one *60year before the patent application was filed on April 19,1982, the court concluded that those claims were invalid under § 102(b). That conclusion rested on the court’s view that as long as the invention was “substantially complete at the time of sale,” the 1-year period began to run, even though the invention had not yet been reduced to practice. Id., at 1434. The other two claims (11 and 19) described a feature that had not been included in Pfaff’s initial design, but the Court of Appeals concluded as a matter of law that the additional feature was not itself patentable because it was an obvious addition to the prior art.7 Given the court’s § 102(b) holding, the prior art included Pfaff’s first four claims. •

Because other courts have held or assumed that an invention cannot be “on sale” within the meaning of § 102(b) unless and until it has been reduced to practice, see, e. g., Timely Products Corp. v. Arron, 523 F. 2d 288, 299-302 (CA2 1975); Dart Industries, Inc. v. E. I. Du Pont de Nemours & Co., 489 F. 2d 1359, 1365, n. 11 (CA7 1973), cert. denied, 417 U. S. 938 (1974), and because the text of § 102(b) makes no reference to “substantial completion” of an invention, we granted certiorari. 523 U. S. 1003 (1998).

r-t í — Í

The primary meaning of the word invention” m the Patent Act unquestionably refers to the inventor’s conception rather than to a physical embodiment of that idea. The statute does not contain any express requirement that an invention must be reduced to practice before it can be patented. *61Neither the statutory definition of the term in § 1008 nor the basic conditions for obtaining a patent set forth in §1019 make any mention of "reduction to practice.” The statute’s only specific reference to that term is found in § 102(g), which sets forth the standard for resolving priority contests between two competing claimants to a patent. That subsection provides:

"In determining priority of invention there shall be considered not only the respective dates of conception and reduction to practice of the invention, but also the reasonable diligence of one who was first to conceive and last to reduce to practice, from a time prior to eoneep - tion by the other.”

Thus, assuming diligence on the part of the applicant, it is normally the first inventor to conceive, rather than the first to reduce to practice, who establishes the right to the patent.

It is well settled that an invention may be patented before it is reduced to practice. In 1888, this Court upheld a patent issued to Alexander Graham Bell even though he had filed his application before constructing a working telephone. Chief Justice Waite’s reasoning in that case merits quoting at length:

"It is quite true that when Bell applied for his patent he had never actually transmitted telegraphically spoken words so that they could be distinctly heard and understood at the receiving end of his line, but in his specification he did describe accurately and with admirable clearness his process, that is to say, the exact *62electrical condition that must be created to accomplish his purpose, and he also described, with sufficient precision to enable one of ordinary skill in such matters to make it, a form of apparatus which, if used in the way pointed out, would produce the required effect, receive the words, and carry them to and deliver them at the appointed place. The particular instrument which he had, and which he used in his experiments, did not, under the circumstances in which it was tried, reproduce the words spoken, so that they could be clearly understood, but the proof is abundant and of the most convincing character, that other instruments, carefully constructed and made exactly in accordance with the specification, without any additions whatever, have operated and will operate successfully. A good mechanic of proper skill in matters of the kind can take the patent and, by following the specification strictly, can, without more, construct an apparatus which, when used in the way pointed out, will do all that it is claimed the method or process will do ....

“The law does not require that a discoverer or inventor, in order to get a patent for a process, must have succeeded in bringing his art to the highest degree of perfection. It is enough if he describes his method with sufficient clearness and precision to enable those skilled in the matter to understand what the process is, and if he points out some practicable way of putting it into operation.” The Telephone Cases, 126 U. S. 1, 535-536 (1888).10

When we apply the reasoning of The Telephone Cases to the facts of the case before us today, it is evident that Pfaff *63could have obtained a patent on his novel socket when he accepted the purchase order from Texas Instruments for 30,100 units. At that time he provided the manufacturer with a description and drawings that had “sufficient clearness and precision to enable those skilled in the matter” to produce the device. Id., at 536. The parties agree that the sockets manufactured to fill that order embody Pfaff’s conception as set forth in claims 1,6,7, and 10 of the ’377 patent. We can find no basis in the text of § 102(b) or in the facts of this case for concluding that Pfaff’s invention was not “on sale” within the meaning of the statute until after it had been reduced to practice.

Ill

Pfaff nevertheless argues that longstanding precedent, buttressed by the strong interest in providing inventors with a clear standard identifying the onset of the 1-year period, justifies a special interpretation of the word “invention” as used in § 102(b). We are persuaded that this nontextual argument should be rejected.

As we have often explained, most recently in Bonito Boats, Inc. v. Thunder Craft Boats, Inc., 489 U. S. 141, 151 (1989), the patent system represents a carefully crafted bargain that encourages both the creation and the public disclosure of new and useful advances in technology, in return for an exclusive monopoly for a limited period of time. The balance between the interest in motivating innovation and enlightenment by rewarding invention with patent protection on the one hand, and the interest in avoiding monopolies that unnecessarily stifle competition on the other, has been a feature of the federal patent laws since their inception. As this Court explained in 1871:

“Letters patent are not to be regarded as monopolies... but as public franchises granted to the inventors of new and useful improvements for the purpose of securing to them, as such inventors, for the limited term therein *64mentioned, the exclusive right and liberty to make and use and vend to others to be used their own inventions, as tending to promote the progress of science and the useful arts, and as matter of compensation to the inventors for their labor, toil, and expense in making the inventions, and reducing the same to practice for the public benefit, as contemplated by the Constitution and sanctioned by the laws of Congress.” Seymour v. Osborne, 11 Wall. 516, 533-534.

Consistent with these ends, § 102 of the Patent Act serves as a limiting provision, both excluding ideas that are in the public domain from patent protection and confining the duration of the monopoly to the statutory term. See, e.g., Frantz Mfg. Co. v. Phenix Mfg. Co., 457 P. 2d 314, 320 (CA7 1972).

We originally held that an inventor loses his right to a patent if he puts his invention into public use before filing a patent application. “His voluntary act or acquiescence in the public sale and use is an abandonment of his right.” Pennock v. Dialogue, 2 Pet. 1, 24 (1829) (Story, J.). A similar reluctance to allow an inventor to remove existing knowledge from public use undergirds the on-sale bar.

Nevertheless, an inventor who seeks to perfect his discovery may conduct extensive testing without losing his right to obtain a patent for his invention — even if such testing occurs in the public eye. The law has long recognized the distinction between inventions put to experimental use and products sold commercially. In 1878, we explained why pat-entability may turn on an inventor’s use of his product.

“It is sometimes said that an inventor acquires an undue advantage over the public by delaying to take out a patent, inasmuch as he thereby preserves the monopoly to himself for a longer period than is allowed by the policy of the law; but this cannot be said with justice when the *65delay is occasioned by a bona fide effort to bring his invention to perfection, or to ascertain whether it will answer the purpose intended. His monopoly only continues for the allotted period, in any event; and it is the interest of the public, as well as himself, that the invention should be perfect and properly tested, before a patent is granted for it. Any attempt to use it for a profit, and not by way of experiment, for a longer period than two years before the application, would deprive the inventor of his right to a patent.” Elizabeth v. Pavement Co., 97 U. S. 126, 137 (emphasis added).

The patent laws therefore seek both to protect the public’s right to retain knowledge already in the public domain and the inventor’s right to control whether and when he may patent his invention. The Patent Act of 1836, 5 Stat. 117, was the first statute that expressly included an on-sale bar to the issuance of a patent. Like the earlier holding in Pen-nock, that provision precluded patentability if the invention had been placed on sale at any time before the patent application was filed. In 1839, Congress ameliorated that requirement by enacting a 2-year grace period in which the inventor could file an application. 5 Stat. 353.

In Andrews v. Hovey, 123 U. S. 267, 274 (1887), we noted that the purpose of that amendment was “to fix a period of limitation which should be certain”; it required the inventor to make sure that a patent application was filed “within two years from the completion of his invention,” ibid. In 1939, Congress reduced the grace period from two years to one year. 53 Stat. 1212.

Petitioner correctly argues that these provisions identify an interest in providing inventors with a definite standard for determining when a patent application must be filed. A rule that makes the timeliness of an application depend on the date when an invention is “substantially complete” se*66riously undermines the interest in certainty.11 Moreover, such a rule finds no support in the text of the statute. Thus, petitioner’s argument calls into question the standard applied by the Court of Appeals, but it does not persuade us that it is necessary to engraft a reduction to practice element into the meaning of the term “invention” as used in § 102(b).

The word “invention” must refer to a concept that is complete, rather than merely one that is “substantially complete.” It is true that reduction to practice ordinarily provides the best evidence that an invention is complete. But just because reduction to practice is sufficient evidence of completion, it does not follow that proof of reduction to practice is necessary in every case. Indeed, both the facts of The Telephone Cases and the facts of this case demonstrate that one can prove that an invention is complete and ready for patenting before it has actually been reduced to practice.12

*67We conclude, therefore, that the on-sale bar applies when two conditions are satisfied before the critical date.

First, the product must be the subject of a commercial offer for sale. An inventor can both understand and control the timing of the first commercial marketing of his invention. The experimental use doctrine, for example, has not generated concerns about indefiniteness,13 and we perceive no reason why unmanageable uncertainty should attend a rule that measures the application of the on-sale bar of § 102(b) against the date when an invention that is ready for patenting is first marketed commercially. In this ease the acceptance of the purchase order prior to April 8, 1981, makes it clear that such an offer had been made, and there is no question that the sale was commercial rather than experimental in character.

Second, the invention must be ready for patenting. That condition may be satisfied in at least two ways: by proof of reduction to practice before the critical date; or by proof that prior to the critical date the inventor had prepared drawings or other descriptions of the invention that were sufficiently specific to enable a person skilled in the art to *68practice the invention.14 In this case the second condition of the on-sale bar is satisfied because the drawings Pfaff sent to the manufacturer before the critical date fully disclosed the invention.

The evidence in this case thus fulfills the two essential conditions of the on-sale bar. As succinctly stated by Learned Hand:

“[I]t is a condition upon an inventor's right to a patent that he shall not exploit his discovery competitively after it is ready for patenting; he must content himself with either secrecy, or legal monopoly.” Metallizing Engineering Co. v. Kenyon Bearing & Auto Parts Co., 153 F. 2d 516, 520 (CA2 1946).

The judgment of the Court of Appeals finds support not only in the text of the statute but also in the basic policies underlying the statutory scheme, ineluding § 102(b). When Pfaff accepted the purchase order for his new sockets prior to April 8,1981, his invention was ready for patenting. The fact that the manufacturer was able to produce the socket using his detailed drawings and specifications demonstrates this fact. Furthermore, those sockets contained all the elements of the invention claimed in the ’377 patent. Therefore, Pfaff’s ’377 patent is invalid because the invention had *69been on sale for more than one year in this country before he filed his patent application. Accordingly, the judgment of the Court of Appeals is affirmed.

It is so ordered.

BRYSON, Circuit Judge.

This appeal involves a challenge to the validity of two plant patents for varieties of table grapes developed by the U.S. Department of Agriculture (“USDA”) and licensed to the California Table Grape Commission. The plaintiffs filed suit against the USDA and the California Table Grape Commission, seeking to invalidate the patents on the ground that the two grape varieties were in public use more than one year before the applications for both plant patents were filed, and that the patents are therefore invalid under the public use bar of 35 U.S.C. § 102(b) (2006).

The United States District Court for the Eastern District of California initially ruled that sovereign immunity barred this action against the USDA, and that the case could not go forward without the USDA as a party. On appeal, we reversed the district court’s ruling as to the sovereign immunity issue, holding that the Administrative Procedure Act waives sovereign immunity for purposes of an action such as this one. Delano Farms Co. v. *1245 Cal. Table Grape Comm’n, 655 F.3d 1337 (Fed.Cir.2011).

On remand, the district court granted partial summary judgment for the defendants on matters that are not at issue on appeal. The court held a bench trial on the remaining question whether the actions of two individuals who obtained samples of the two patented plant varieties in an unauthorized manner and planted them in their own fields constituted an invalidating public use of the plant varieties. The district court found that the actions of those individuals did not constitute a public use of the two plant varieties and therefore rejected the plaintiffs’ challenge to the patents. We affirm.

I

The patented table grape varieties at issue in this case are known as Scarlet Royal (U.S. Patent No. PP16,229), and Autumn King (U.S. Patent No. PP16,284). The USDA, as owner of the patents, has exclusively licensed the patents to the California Table Grape Commission, an agency of the State of California. The Commission sublicenses the patents to grape growers in California and collects royalties that are shared by the Commission and the USDA. The licensing agreements with the growers require the growers to pay a royalty on the grapes produced by plants of the patented varieties, and they prohibit the growers from propagating the plants.

The three plaintiffs are all California grape growers who purchased grapevines covered by the patents, signed license agreements with the California Table Grape Commission, and paid the Commission’s licensing fee. They brought this action challenging the validity and enforceability of the plant patents, as well as the conduct of the Commission and the USDA in licensing and enforcing the patents.

Following the bench trial, the district court made detailed findings of fact. The court’s findings, and the evidence at trial that supported those findings, are summarized below.

The applications that resulted in the plant patents covering Scarlet Royal and Autumn King were filed on September 28, 2004. The “critical date” for the public use bar of 35 U.S.C. § 102(b) (2006), i.e., the beginning date of the one-year period before which a public use would invalidate the patents, is therefore September 28, 2003. Both varieties were made commercially available on July 13, 2005.

Jim Ludy, a California grape grower, cultivated table grapes at J & J Ludy Farms, in Delano, California, with his brother, Jack Ludy, from 1976 to 2003. Larry Ludy, Jim Ludy’s first cousin, also cultivated table grapes at various properties in the area.

On August 22, 2001, the USDA held an experimental variety open house at the California State University, Fresno. At the open house, USDA representatives displayed several unreleased table grape varieties, including Scarlet Royal and Autumn King. Only the mature fruit from the varieties was on display, not other plant material such as the vines or wood. Visiting growers were not permitted to take any plant material relating to the unreleased varieties or even view the plants in the field.

Both Jim Ludy and Larry Ludy attended the August open house. While they were there, the Ludy cousins spoke with Rodney Klassen, who was employed by the USDA at the facility where the Scarlet Royal and Autumn King varieties were being developed. Jim Ludy asked Mr. Klassen if Mr. Klassen could give him some of the plant material for the Scarlet Royal and Autumn King varieties. Mr. Klassen had previously given Jim Ludy plant material for other unreleased table grape varieties. Mr. Klassen was not au*1246thorized to provide Jim Ludy with any unreleased plant material.

Despite his lack of authority, Mr. Klas-sen informed Jim Ludy that he would “take care” of him. Subsequently, in early 2002, Mr. Klassen met with Jim Ludy and gave him plant material for multiple, unreleased varieties, including Scarlet Royal and Autumn King. Mr. Klassen instructed Jim Ludy not to let the material “get away from [him]” and not to “put them in a box,” which Jim Ludy understood to mean that he should not sell the resulting grapes until the varieties were commercially released. Jim Ludy understood that he was to keep the plant material secret. He did not want other growers to obtain the same unreleased plant material he had, and he knew that Mr. Klassen would be in “big trouble” if his actions were discovered. Jim Ludy subsequently testified falsely under oath to protect Mr. Klassen’s identity as the person from whom he had obtained the plant material for the unreleased varieties.

After receiving the plant material from Mr. Klassen, Jim Ludy grafted fewer than 50 vines of each of the Scarlet Royal and Autumn King varieties in early 2002. Jim Ludy also provided “a few buds” of Scarlet Royal and Autumn King to his cousin, Larry Ludy. Larry Ludy knew that the material had originally come from a USDA facility and had not yet been released. He admitted that Jim Ludy told him that they should “keep it to ourselves.” Jim Ludy testified that it was understood between the two of them that their possession of the Scarlet Royal and Autumn King varieties “was supposed to be a secret” and that Larry Ludy would keep his possession of the plant material confidential.

Larry Ludy subsequently grew eight plants of Scarlet Royal and 25 plants of Autumn King. In 2003, Larry Ludy grew additional plants by grafting wood from the original group of plants, producing a total of 108 vines of Scarlet Royal and 650 vines of Autumn King.

Most of the plantings by both Ludys prior to the patents’ critical date bore no usable fruit, and the Ludys sold no grapes from those plantings prior to the critical date. Neither of the Ludys provided plant material to any other persons until after the critical date. Although the various plantings were visible from publicly accessible roads, none of the vines were marked or labeled in any way, and the evidence showed that the particular variety of the grapes could not be readily ascertained from simply viewing the vines.

The only other person who was informed of the Ludys’ possession of the unreleased plant material was Richard Sandrini, who had long served as a table grape marketer for Jim and Larry Ludy.

Larry Ludy showed Mr. Sandrini vines of the unreleased varieties at least twice prior to the patents’ critical date. After the critical date, Mr. Sandrini sold Larry Lud/s 2004 harvest of Autumn King (the first sale of either unreleased variety), but he labeled the grapes as “Thompson Seedless” to avoid detection. Additionally, after the patents’ critical date Larry Ludy provided Mr. Sandrini with wood from which Mr. Sandrini could graft Autumn King on his own property. Larry Ludy testified that he knew it was a “huge competitive advantage” to have grapes that sold at a significant premium before other growers. Mr. Sandrini likewise recognized the competitive advantage inherent in possessing the unreleased plants and intended to grow his own Autumn King grapes.

. Based on the findings summarized above, the district court held that the plaintiffs failed to meet their burden of showing, by clear and convincing evidence, that the Ludys’ use of the unreleased varieties constituted a public use that invali*1247dated the patents under 35 U.S.C. § 102(b) (2006). The plaintiffs appeal.

II

A

An applicant may not be granted a patent for an invention that was “in public use ... in this country, more than one year prior to the date of the application for patent in the United States.” 35 U.S.C. § 102(b) (2006). Following a bench trial, as in this case, “the district court’s conclusion on public use under § 102(b) is subject to review as a question of law while the facts underlying the conclusion on public use are subject to the clearly erroneous standard of review.” Moleculon Research Corp. v. CBS, Inc., 793 F.2d 1261, 1266 (Fed.Cir.1986). “The proper test for the public use prong of the section 102(b) statutory bar is whether the purported use was accessible to the public or was commercially exploited.” Invitrogen Corp. v. Biocrest Mfg., L.P., 424 F.3d 1374, 1380 (Fed.Cir.2005).

Our case law defines the contours of what it means to be “accessible to the public.” The principal policy underlying the statutory bar is to prevent “the removal, from the public domain, of inventions • that the public reasonably has come to believe are freely available.” Tone Bros, v. Sysco Corp., 28 F.3d 1192, 1198 (Fed.Cir.1994). The question in a case such as this one is thus whether the actions taken by the inventor (or, as in this case, a third party) create a reasonable belief as to the invention’s public availability.

Factors that we have previously identified as being helpful in analyzing that question include “the nature of the activity that occurred in public; the public access to and knowledge of the public use; [and] whether there was any confidentiality obligation imposed on persons who observed the use.” Bernhardt, L.L.C. v. Collezione Europa USA, Inc., 386 F.3d 1371, 1379 (Fed.Cir.2004), quoting Allied Colloids Inc. v. Am. Cyanamid Co., 64 F.3d 1570, 1574 (Fed.Cir.1995). The last factor captures “the commonsense notion that whether an invention is ‘accessible to the public’ ... depends, at least in part, on the degree of confidentiality surrounding its use: ‘[A]n agreement of confidentiality, or circumstances creating a similar expectation of secrecy, may negate a public use where there is not commercial exploitation.’ ” Dey, L.P. v. Sunovion Pharm., Inc., 715 F.3d 1351, 1355 (Fed.Cir.2013), quoting Invitrogen, 424 F.3d at 1382.

The analysis is similar when the allegedly public use is performed by an unaffiliated third party rather than the inventor. “Third party prior use accessible to the public is a section 102(b) bar.” Eolas Techs. Inc. v. Microsoft Corp., 399 F.3d 1325, 1334 (Fed.Cir.2005). In order to be invalidating, such use must still be publicly accessible; “secret or confidential third-party uses do not invalidate later-filed patents.” Dey, 715 F.3d at 1355. The adequacy of any confidentiality guarantees are measured in relation “to the party in control of the allegedly invalidating prior use.” Id. at 1358. The actions of an unaffiliated third party acting in secret are evaluated as if he stood in the place of the inventor.

With these principles in mind, we turn to their application to the facts found by the district court.

B

The appellants argue that the cultivation of the unreleased varieties by Jim Ludy and Larry Ludy constituted public use. They point to previous opinions of this court in which the court has held third-party uses of an invention to be invalidating when, for example, “the third party ‘made no attempt to maintain confi*1248dentiality or to deliberately evade disclosure’; made no ‘discernible effort to maintain the [invention] as confidential’; or ‘made no efforts to conceal the device or keep anything about it secret.’ ” Dey, 715 F.3d at 1355 (citations omitted). As evidence of the Ludys’ lack of effort to maintain secrecy regarding then- possession of the Scarlet Royal and Autumn King plant varieties, the appellants point to Jim Ludy’s provision of plants to Larry Ludy, Larry Ludy’s sharing of information with Mr. Sandrini, and the lack of concealment of the vines at either of the Ludys’ farms.

The appellants first argue that Jim Ludy’s provision of plant material to his cousin Larry Ludy resulted in public use. This argument relies primarily on the Supreme Court’s nineteenth century decision in Egbert v. Lippmann, 104 U.S. 333, 26 L.Ed. 755 (1881). In that case, the Court ruled that the inventor of a type of corset steel who gave two of the corset steels to a friend for use 11 years before applying for a patent on the device had put the invention into public use, resulting in the invalidation of the patent. The Court held in that case that “[i]f an inventor, having made his device, gives or sells it to another, to be used by the donee or vendee, without limitation or restriction, or injunction of secrecy, and it is so used, such use is public, even though the use and knowledge of the use may be confined to one person.” Id. at 336.

Although the inventor of the plant varieties in this case did not give or sell the invention to anyone, Jim Ludy obtained control over the unreleased varieties. The appellants argue that for purposes of the public use doctrine, Jim Ludy therefore stands in place of the inventor. They contend that if Jim Ludy gave Larry Ludy the unreleased plant material “without limitation or restriction, or injunction of secrecy,” Larry Ludy’s subsequent cultivation of the plants would be an invalidating public use of the inventions.

The problem with the appellants’ argument is that it is squarely contrary to the district court’s findings of fact. Larry Ludy was present during and participated in Jim Ludy’s conversation with Mr. Klas-sen and knew that Mr. Klassen did not have the authority to provide the Ludys with unreleased varieties. When Jim Ludy gave Larry Ludy the plants, Jim Ludy explicitly told his cousin to “keep [knowledge of the plants] to ourselves” and expected the fact of their possession of the plants to remain private. After the critical date, Larry Ludy allowed Mr. Sandrini to sell the fruit of the unreleased vines under a different name to avoid detection. Moreover, during a deposition in this case, Larry Ludy refused to identify Mr. Klas-sen as the source of the Ludys’ unreleased plants; he acknowledged at that time that the information “should be confidential and not out in the public domain.” The findings of the district court clearly establish, therefore, that both Ludys knew that they were not authorized to have the plants and that they needed to conceal their possession of the plants.

To the extent that the appellants’ argument as to Larry Ludy is based on the lack of an explicit confidentiality agreement between the cousins, “[w]e have never required a formal confidentiality agreement to show non-public use.” Dey, 715 F.3d at 1357; see also Moleculon, 793 F.2d at 1266 (“[T]he presence or absence of [an express confidentiality] agreement is not determinative of the public use issue.”). Instead, we evaluate whether there were “circumstances creating a similar expectation of secrecy.” Invitrogen, 424 F.3d at 1382.

The Supreme Court’s decision in Egbert turned on the inventor’s lack of any effort to maintain control over the use of his *1249invention. The facts of this case, by contrast, show that Jim Ludy sought to maintain control of the plants he obtained from Mr. Klassen. Although Jim Ludy shared the plants with his cousin, the evidence showed that Larry Ludy was aware of the need to keep the plants secret, and at Jim Ludy’s urging, Larry Ludy continued to treat his possession of the unreleased varieties as confidential and non-public. This case is therefore wholly different from the Supreme Court’s decision in Egbert v. Lippmann, where the inventor himself gave the invention to a third party with no understanding or expectation that the third party would maintain secrecy as to the invention.

The appellants’ second argument is that the Ludy cousins’ disclosure of the unreleased plants to Mr. Sandrini constituted public use. Unlike the Ludys, however, Mr. Sandrini could not practice the inventions because he did not possess plant material until after the critical date. Instead, the appellants argue that the disclosure of the plants’ existence to Mr. Sandri-ni demonstrates the lack of confidentiality with which the Ludys treated the unreleased varieties.

The circumstances under which the disclosure to Mr. Sandrini occurred weigh against the application of the public use bar. In similar cases, we have held that the nature of the disclosure did not give rise to a “public use.” For example, in American Seating Co. v. USSC Group, Inc., 514 F.3d 1262 (Fed.Cir.2008), we affirmed the district court’s decision that demonstration of a prototype to “friends and colleagues” was not invalidating because the evidence supported the existence of “a general understanding of confidentiality.” Id. at 1267-68. Similarly, in Bernhardt, L.L.C. v. Collezione Europa USA Inc., we vacated a district court decision finding an invalidating public use where “access was tightly controlled,” “there was an industry-wide understanding [of confidentiality],” “a breach of confidence could have serious consequences for an attendee,” and “there was no effective means for the attendees to divulge the designs they viewed.” 386 F.3d at 1380-81.

In this case, the district court found that Mr. Sandrini was a friend, business partner, and mentor of the Ludys. The court also found that “[e]ach [of the Ludys and Mr. Sandrini] had incentives to keep the Ludys’ possession secret, creating an environment of confidentiality, [and] [e]ach maintained tight control over who knew about the Scarlet Royal and Autumn King vines and their use.” We have no reason to overturn these findings. Based on the district court’s findings and our case law, the Ludys’ disclosure to Mr. Sandrini that they were in possession of the unreleased plants does not qualify as an invalidating public use of the patented plant varieties.

Finally, the appellants argue that the lack of secrecy with which the Ludys cultivated the unreleased varieties mandates a finding of public use. The appellants are correct that the district court found that both Ludys grafted the plants and grew them in locations that were visible from public roads. However, the appellants ignore the district court’s finding that grape varieties cannot be reliably identified simply by viewing the growing vines alone. The plantings of the unreleased varieties were extremely limited in comparison to the total cultivation of the Ludys’ farms. The unreleased varieties were not labeled in any way, and the appellants introduced no evidence that any person other than the Ludys and Mr. Sandrini had ever recognized the unreleased varieties. As this court explained in the Dey case, 715 F.3d at 1359, “a reasonable jury could conclude that if members of the public are not informed of, and cannot readily discern, the *1250claimed features of the invention in the allegedly invalidating prior art, the public has not been put in possession of those features.” In this case, the district court, sitting as the trier of fact, came to exactly that conclusion, and the evidence supports the court’s conclusion.

C

As án alternative ground of decision, the district court found that the plaintiffs failed to satisfy their burden of proof on public use because they failed to corroborate the testimony of the Ludys and Mr. Sandrini. Because we uphold the district court’s determination that the plaintiffs failed to demonstrate a public use even after admitting the testimony of the Ludys and Mr. Sandrini, it is unnecessary for us to address whether that evidence was inadmissible for want of corroboration.

In addition, because the evidence at trial was sufficient to support the district court’s finding that the patented plant varieties were not in public use prior to the critical date, we need not address the question whether use of invention by one who has misappropriated that invention (or obtained it through other improper means) can ever qualify as an invalidating public use. The district court properly ruled that the Ludys’ use of the plant varieties at issue in this case was not public, even apart from the fact that the Ludys obtained the plant material in an unauthorized manner.

Accordingly, we affirm the district court’s decision rejecting the appellants’ challenge to the validity of the Scarlet Royal and Autumn King patents.

AFFIRMED

PAULINE NEWMAN, Circuit Judge.

Continental Can Company USA and Continental PET Technologies (collectively “Continental”) appeal the partial summary judgment of the United States District Court for the Southern District of Ohio, holding that United States Patent No. 4,108,324 (the Conobase or ’324 patent) is invalid.1 Final judgment was entered on this issue, for the purpose of appeal.

Summary Judgment

An issue may be decided on motion for summary judgment when there is no genuine issue of material fact, and the movant is entitled to judgment as a matter of law. Fed.R.Civ.P. 56(c); Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986); Celotex Corp. v. Catrett, 477 U.S. 317, 325-26, 106 S.Ct. 2548, 2554, 91 L.Ed.2d 265 (1986); Scripps Clinic & Research Foundation v. Genentech, Inc., 927 F.2d 1565, 1571, 18 U.S.P.Q.2d 1001, 1005 (Fed.Cir.1991). The movant’s burden is to show that no fact material to the issue is in dispute, that even if all material factual inferences are drawn in favor of the non-movant the movant is entitled to judgment as a matter of law. Id. Summary judgment is as available in patent cases as in other areas of litigation. Chore-Time Equipment, Inc. v. Cumberland Corp., 713 F.2d 774, 778-79, 218 U.S.P.Q. 673, 675. (Fed.Cir.1983)

The purpose of the summary process is to avoid a clearly unnecessary trial, Matsushita Elec. Industrial Co. v. Zenith Radio Corp., 475 U.S. 574, 587, 106 S.Ct. 1348, 1356, 89 L.Ed.2d 538 (1986); it is not designed to substitute lawyers’ advocacy for evidence, or affidavits for examination before the fact-finder, when there is a genuine issue for trial. As stated in Adickes v. S.H. Kress & Co., 398 U.S. 144, 176, 90 S.Ct. 1598, 1618, 26 L.Ed.2d 142 (1970) (Black, J., concurring), “[t]he right to confront, cross-examine and impeach adverse witnesses is one of the most fundamental rights sought to be preserved by the Seventh Amendment”. See also Poller v. Columbia Broadcasting System, Inc., 368 U.S. 464, 473, 82 S.Ct. 486, 491, 7 L.Ed.2d 458 (1962).

While facilitating the disposition of legally meritless suits, when summary judg*1266ment is improvidently granted the effect is to prolong litigation and increase its burdens. This is of particular concern in patent disputes, where the patent property is a wasting asset, and justice is ill served by delay in final resolution. In the case at bar, although some issues could be resolved on the law and undisputed facts, other issues require trial.

The Patented Invention

The ’824 patent, entitled “Ribbed Bottom Structure for Plastic Container”, inventors Suppayan M. Krishnakumar, Siegfried S. Roy, John F.E. Pocock, Salil K. Das, and Gautam K. Mahajan, is directed to a plastic bottle whose bottom structure has sufficient flexibility to impart improved impact resistance, combined with sufficient rigidity to resist deformation under internal pressure. The patented bottle is said to provide a superior combination of these properties. The bottom structure is illustrated as follows:

Claim 1 is the broadest claim of the ’324 patent:

1. A container having a sidewall and a bottom structure closing the container at an end portion of the sidewall,

the outer surface of the bottom structure comprising a central concavity,

a convex heel surrounding the concavity and merging therewith and with the sidewall end portion, the lowermost points of the heel lying in a common plane,

and a plurality of ribs interrupting the outer surface of the concavity and distributed in a symmetrical array,

each rib extending longitudinally in the direction of the heel and downwardly from an inner portion of the concavity, whereby the outer end portion of each rib is lower than the inner end portion thereof,

characterized by the feature that the ribs are hollow.

Claims 2 through 5 include additional limitations, described as contributing to the structure’s rigidity, flexibility, or both. Claim 2 specifies the ratios of thickness of the walls of the bottom structure to the thickness of the sidewall end portions. Claim 3 specifies that the margins of each rib merge smoothly with adjacent portions of the bottom structure. Claim 4 specifies that each rib is convex relative to the bottom structure. Claim 5 specifies that each rib is of fusiform (a gently tapered shape at the ends) configuration. Each claim carries an independent presumption of va*1267lidity, 35 U.S.C. § 282, and stands or falls independent of the other claims. Altoona Publix Theatres, Inc. v. American Tri-Ergon Corp., 294 U.S. 477, 487, 55 S.Ct. 455, 459, 79 L.Ed. 1005 (1935).

Continental brought suit for patent infringement against Monsanto Company and Monsanto’s successor in this business, Hoover Universal, Inc. and Hoover’s parent company, Johnson Controls (collectively “Monsanto”). Monsanto moved for partial summary judgment based on issues of validity under 35 U.S.C. §§ 102 and 103.

I

35 U.S.C. § 102(a)

The statutory requirement that a patented invention be “new” is tested in accordance with 35 U.S.C. § 102(a), which provides that:

§ 102. A person shall be entitled to a patent unless—

(a) the invention was known or used by others in this country, or patented or described in a printed publication in this or a foreign country, before the invention thereof by the applicant for patent....

The district court found that all the claims of the ’324 patent were anticipated by U.S. Patent No. 3,468,443 (the Marcus patent). We conclude that the district court erred in claim interpretation, and also found disputed facts adversely to the nonmovant, thus inappropriately deciding the issue summarily.

Anticipation under § 102(a) requires that the identical invention that is claimed was previously known to others and thus is not new. Scripps Clinic, 927 F.2d at 1576, 18 U.S.P.Q.2d at 1010; Titanium Metals Corp. of Am. v. Banner, 778 F.2d 775, 780, 227 U.S.P.Q. 773, 777-78 (Fed.Cir.1985); Lindemann Maschinenfabrik GmbH v. American Hoist and Derrick Co., 730 F.2d 1452, 1458, 221 U.S.P.Q. 481, 485 (Fed.Cir.1984). When more than one reference is required to establish unpatentability of the claimed invention anticipation under § 102 can not be found, and validity is determined in terms of § 103.

It was Monsanto’s burden to show that every element of the several claims of the ’324 patent was identically described in the asserted anticipating reference, the Marcus patent. The district court focused on the term “characterized by the feature that the ribs are hollow”, which limits all of the ’324 patent claims. Continental argues that the district court incorrectly construed this term, as a matter of law, and that the Marcus patent shows ribs that are not hollow, as that term is used in the ’324 patent. Continental also points to other differences between the ’324 claims and the description in the Marcus patent.

The Marcus patent rib structure is illustrated in Figure 5 and in cross-section in Figure 6:

*1268The Marcus patent does not state that its ribs are “hollow”, or use a similar term. Continental’s witnesses testified by deposition that the Marcus patent shows solid, not hollow, ribs. A witness (Adomaitis) had stated in an internal memorandum written at Continental in 1969, well before this litigation arose, that “the ribs of their [Marcus’] web can be made of solid beams only.” Another witness, ’324 co-inventor Pocock, testified that:

It seems evident to me that he [Marcus] was trying to produce some kind of container integrity by the production of essentially solid ribs on the bottom of the bottle. It seems to go to great length here to illustrate them as such.

Krishnakumar, another co-inventor, testified that it “is very obvious the ribs are shown solid”, and that Figures 5 and 6 as well as Figures 7 through 12 of the Marcus patent all show solid ribs. However, Marcus, testifying for Monsanto, testified that his ribs were hollow, and that conventional blow molding would inherently produce hollow ribs.

The district court defined “hollow” as meaning that “the inside contour of the ribs generally follows the outside contour thereof”, a definition on which the parties agreed. Continental, 11 U.S.P.Q.2d at 1764. See the court’s opinion, 11 U.S.P.Q.2d at 1764-68, for various sketches made by the witnesses. Continental states that the district court erred in construing “hollow”, and that the phrase “characterized by the feature that the ribs are hollow” must be construed in terms of the patent in which it appears. See, e.g., Tandon Corp. v. United States Int’l Trade Comm’n, 831 F.2d 1017, 1021, 4 U.S.P.Q.2d 1283, 1286 (Fed.Cir.1987). The ’324 patent explicitly distinguished the Marcus patent teachings, stating that the ’324 ribs are, unlike Marcus, not filled with plastic. The ’324 specification uses the term “hollow”, as do the prosecution history and the claims, for this purpose. The ’324 patent’s usage of “hollow” is illustrated in the rib cross-section in Figure 5A:

The Marcus patent’s rib structure thus was explicitly differentiated by the term “hollow” as used in the ’324 specification, drawings, and prosecution history. Since the claim term must be construed as used by the patentee, the district court erred in its construction of the ’324 claim term “hollow”. On correct claim construction, the factual question of anticipation must be decided.

Monsanto’s argument is that hollow ribs were inherently produced by Marcus. Monsanto thus argues that anticipation lies because the Marcus patent’s ribs are “inherently” hollow, regardless of how they are shown in the Marcus patent. Monsanto argues that because the Marcus ribs are formed by injection blow molding, which is the same process described for the Cono-base ’324 ribs, hollow ribs are inherently disclosed in the Marcus patent.

To serve as an anticipation when the reference is silent about the asserted inherent characteristic, such gap in the reference may be filled with recourse to extrinsic evidence. Such evidence must make clear that the missing descriptive matter is necessarily present in the thing described in the reference, and that it would be so recognized by persons of ordinary skill. In re Oelrich, 666 F.2d 578, 581, 212 U.S.P.Q. 323, 326 (CCPA 1981) (quoting Hansgirg v. Kemmer, 102 F.2d 212, 214, 40 U.S.P.Q. 665, 667 (CCPA 1939)) provides:

*1269Inherency, however, may not be established by probabilities or possibilities. The mere fact that a certain thing may result from a given set of circumstances is not sufficient. [Citations omitted.] If, however, the disclosure is sufficient to show that the natural result flowing from the operation as taught would result in the performance of the questioned function, it seems to be well settled that the disclosure should be regarded as sufficient.

This modest flexibility in the rule that “anticipation” requires that every element of the claims appear in a single reference accommodates situations where the common knowledge of technologists is not recorded in the reference; that is, where technological facts are known to those in the field of the invention, albeit not known to judges. It is not, however, a substitute for determination of patentability in terms of § 103.

Continental does not dispute the applicability of the injection blow molding process. However, Continental disputes the material fact of whether this process necessarily produced “hollow” ribs in the Marcus base structure, as the term “hollow” is used in the ’324 patent. Resolution of this disputed fact adversely to Continental was improper on summary judgment. The grant of summary judgment of anticipation under § 102(a) is vacated. The issue requires trial.

II

35 U.S.C. § 102(b)

The district court also held that the Marcus bottle was on sale, 35 U.S.C. § 102(b). Section 102(b) bars entitlement to a patent when:

(b) the invention was ... in public use or on sale in this country, more than one year prior to the date of the application for patent in the United States....

The Marcus bottle was developed some ten years before the filing date of the ’324 patent, during a project wherein Marcus’ employer, Admiral Plastics or APL Corporation, entered into agreements with the Coca-Cola Company for the development of a suitable plastic bottle. The agreements provided that Admiral Plastics would make and Coca-Cola would test the bottles, and that if a satisfactory bottle was developed it would be manufactured by Admiral and purchased by Coca-Cola. Minimum commercial quantities and maximum commercial prices were stated in an agreement, and costs were a matter of discussion. Admiral produced a variety of bottle shapes, including the Marcus bottle. The project was terminated after about two years, because the “mechanical performance” requirements were not met, as Coca-Cola wrote at the time.

The district court reasoned that this project “called for the eventual marketing of the Marcus bottles once all technical difficulties were resolved”, Continental, 11 U.S.P.Q.2d at 1766, and on this basis held that the Marcus bottles were on sale. This holding was in error, for the “on sale” bar of § 102(b) does not arise simply because the intended customer was participating in development and testing. See Great Northern Corp. v. Davis Core & Pad Co., 782 F.2d 159, 164-65, 228 U.S.P.Q. 356, 358 (Fed.Cir.1986). In Baker Oil Tools, Inc. v. Geo Vann, Inc., 828 F.2d 1558, 1563-65, 4 U.S.P.Q.2d 1210, 1213-15 (Fed.Cir.1987), this court summarized various factors pertinent to the “on sale” bar when there is an issue concerning the relationship between the patentee and the customer: for example, whether there was a need for testing by other than the patentee; the amount of control exercised; the stage of development of the invention; whether payments were made and the basis thereof; whether confidentiality was required; and whether technological changes were made. All of the circumstances attending the relationship must be considered in light of the public policy underlying § 102(b). UMC Electronics Co. v. United States, 816 F.2d 647, 656, 2 U.S.P.Q.2d 1465, 1471-72 (Fed.Cir.1987), cert. denied, 484 U.S. 1025, 108 S.Ct. 748, 98 L.Ed.2d 761 (1988).

The district court acknowledged that all technical difficulties were not resolved and that no sales were ever made. *1270Although Admiral Plastics' hope was surely commercial sales, and the record shows that prices and quantities were discussed, this does not of itself place the subject matter “on sale” in the sense of § 102(b). The Marcus bottle was part of a terminated development project that never bore commercial fruit and was cloaked in confidentiality. While the line is not always bright between development and being on sale, see generally UMC Electronics, supra, in this case the line was not crossed. The “on sale” bar is measured by “the time the public came into possession of the invention”, id. at 655, 2 U.S.P.Q.2d at 1471 (quoting In re Foster, 343 F.2d 980, 987-88, 145 U.S.P.Q. 166, 173 (CCPA 1965), cert. denied, 383 U.S. 966, 86 S.Ct. 1270, 16 L.Ed.2d 307 (1966) (“What starts the period running is clearly the availability of the invention to the public through the categories of disclosure enumerated in 102(b)_” (emphasis in original))). We conclude that the district court erred in holding that the circumstances that here existed placed the Marcus bottles “on sale” in terms of § 102(b). We therefore reverse and direct that on remand judgment on this issue shall be entered in favor of Continental, as a matter of law.

Ill

35 U.S.C. § 103

Obviousness, 35 U.S.C. § 103, is reviewed as a legal conclusion based upon underlying facts of four general categories, viz. the scope and content of the prior art, the differences between the prior art and the claimed invention, the level of ordinary skill at the time the invention was made, and any objective considerations that may be present. Graham v. John Deere Co., 383 U.S. 1, 17, 86 S.Ct. 684, 693-94, 15 L.Ed.2d 545 (1966); Interconnect Planning Corp. v. Feil, 774 F.2d 1132, 1137-38, 227 U.S.P.Q. 543, 547 (Fed.Cir.1985).

The parties agreed that the scope and content of the prior art was adequately represented by four references: the Marcus patent discussed in Part I ante, a patent to Colombo (U.S. Patent No. 3,403,804), and two patents owned by Continental, U.S. Patent No. 3,598,270 (the Petaloid patent), and No. 3,935,955 (the Decaloid patent). They agreed on little else. In granting summary judgment of invalidity for obviousness, the district court found certain disputed material facts and misapplied certain precepts of law. We conclude that the issue was not amenable to summary resolution. Although it is not entirely clear how the references were combined by the court, we shall review the references briefly, in order to explain our conclusion.

The Petaloid Patent

The district court referred to the deposition testimony of Siegfried Roy, one of the co-inventors of the '324 patent, that the Petaloid base, inverted, was similar to the Conobase. Continental points out that neither Roy nor any other deponent suggested that the Petaloid base could be or should be inverted, or that inversion would provide an improved base structure. In In re Gordon, 733 F.2d 900, 902, 221 U.S.P.Q. 1125, 1127 (Fed.Cir.1984) this court held that although a prior art device could have been turned upside down, that did not make the modification obvious unless the prior art fairly suggested the desirability of turning the device upside down.

Continental points out that the Petaloid description differs in several other ways from the '324 invention. In the '324 structure the outer end of each rib is lower than the inner end, whereas in the Petaloid structure the outer ends of the ribs are higher than the inner ends; that is, the ribs in the Petaloid base extend upward from the center to the sidewall. The Petaloid bottle is supported on feet extending between the ribs, such feet being the locations for stress concentrations. The following drawing is from the Petaloid patent:

*1271

Continental states that the ’324 Conobase is not only different, but avoids the stress concentrations of the Petaloid device, thus enhancing impact resistance. Monsanto argues that Continental simply used the Pe-taloid hollow ribs in combination with the Marcus patent. This requires determination of whether there was something in the prior art as a whole to suggest the desirability, and thus the obviousness, of making the combination, in a way that would produce the ’324 structure. See, e.g., Uniroyal, Inc. v. Rudkin-Wiley Corp., 837 F.2d 1044, 1051, 5 U.S.P.Q.2d 1434, 1438 (Fed. Cir.), cert. denied, 488 U.S. 825, 109 S.Ct. 75, 102 L.Ed.2d 51 (1988). Continental argues that it is not apparent, even with hindsight, how any combination of the Pe-taloid and Marcus patents or other references lead to the ’324 base. The Petaloid patent shows concave ribs that extend all the way to the sidewall, while the Marcus ribs extend “from the heel” toward an annular central ring. The Petaloid base has wide, petal-like, open ribs, while Marcus shows narrow, beam-like ribs. The deposition testimony was in conflict as to the inferences drawn from the references.

On this disputed issue, drawing reasonable inferences in favor of the non-movant, it has not been established that one skilled in the art would be motivated to select and combine features from each source in order to make the '324 base. Interconnect Planning, 774 F.2d at 1143, 227 U.S.P.Q. at 551 (“When prior art references require selective combination by the court to render obvious a subsequent invention, there must be some reason for the combination other than the hindsight gleaned from the invention itself”).

The Decaloid Patent

The district court also referred to combination of the Decaloid base with the Marcus base. The Decaloid base has ten hollow ribs that extend to the sidewall, and ten feet between the ribs:

*1272

Monsanto does not explain, and we can not discern, how the combination with Marcus would have led a person of ordinary skill to the ’324 base. The court’s summary holding of obviousness based on these references, separately or in combination, can not be sustained.

The Colombo Patent

The Colombo base, like the Petaloid and Decaloid bases, has hollow ribs that extend to the sidewall, in a still different structure from that of Marcus and also from that of the ’324 patent. Colombo describes his ribs as inverted U-shapes, concave, located on the outer surface of the central concavity:

Again, drawing reasonable factual infer-enees in favor of Continental, and in the absence of any suggestion or motivation in the prior art as a whole to make a selective combination of the Colombo and Marcus *1273structures along with other changes needed to obtain the ’824 structure, summary judgment of obviousness was inappropriate.

The district court found that there was no substantial difference between the ’324 invention and the combined teachings of the prior art:

As obviousness can be established on the basis of the combined teachings of references, we think it is clear that simple enhancements of existing prior art, i.e. inverting the ’270 petaloid base, do not constitute a substantial difference between the subject matter claimed in the ’324 patent and that of the prior art. Thus, the facts of this case reveal no substantial difference between ’324 and the prior art.

Continental, 11 U.S.P.Q.2d at 1769 (citation omitted). However, as we have discussed, the criterion of § 103 is not whether the differences from the prior art are “simple enhancements”, but whether it would have been obvious to make the claimed structure.

Objective Indicia

The district court concluded that the structure in suit is simply a variation on known themes. It is in such circumstance that the objective indicia — the so-called secondary considerations — are most useful to the decision-maker. The significance of a new structure is often better measured in the marketplace than in the courtroom.

Thus when differences that may appear technologically minor nonetheless have a practical impact, particularly in a crowded field, the decision-maker must consider the obviousness of the new structure in this light. Such objective indicia as commercial success, or filling an existing need, illuminate the technological and commercial environment of the inventor, and aid in understanding the state of the art at the time the invention was made. See In re Piasecki, 745 F.2d 1468, 1475, 223 U.S.P.Q. 785, 790 (Fed.Cir.1984) (secondary considerations “often establish that an invention appearing to have been obvious in light of the prior art was not” (quoting Stratoflex, Inc. v. Aeroquip Corp., 713 F.2d 1530, 1538-39, 218 U.S.P.Q. 871, 879 (Fed.Cir.1983))).

Continental licensed the ’324 counterpart Japanese patent to a Japanese company, Yoshino, that we are told had been unable to develop a plastic bottle for hot-fill applications. A witness for Toyo Seikan, another Japanese licensee, testified that the Co-nobase “sustains itself in higher temperatures, and it does not cause buckling after you fill [the bottle]”, as compared with previously available plastic bottles. Continental asserts that Monsanto had been unable to develop a satisfactory bottle for hot-fill applications, and had therefore obtained this technology from Yoshino.

The district court acknowledged the commercial success of the Conobase, but stated that “we are not convinced that the cono-base alone accounts for any of the success.” 11 U.S.P.Q.2d at 1770 (emphasis in original). The court suggested that the commercial success in Japan was due to the market strength of the Japanese licensees, and held that there is no nexus between the merits of the product and its commercial success. It is not necessary, however, that the patented invention be solely responsible for the commercial success, in order for this factor to be given weight appropriate to the evidence, along with other pertinent factors. See generally Demaco Corp. v. F. Von Langsdorff Licensing Ltd., 851 F.2d 1387, 1392-94, 7 U.S.P.Q.2d 1222, 1226-28 (Fed.Cir.), cert. denied, 488 U.S. 956, 109 S.Ct. 395, 102 L.Ed.2d 383 (1988); Rosemount, Inc. v. Beckman Instruments, Inc., 727 F.2d 1540, 1546, 221 U.S.P.Q. 1, 7 (Fed.Cir.1984). Monsanto also states that the Conobase is different from the ’324 invention, so that even were the Conobase successful, this does not inure to the benefit of the ’324 patent. It is apparent that the factual issues surrounding the objective indicia were disputed, and material.

In view of the material facts requiring resolution, the issue of obviousness was not properly decided on motion for sum*1274mary judgment. We vacate the grant based on 35 U.S.C. § 103, and remand for trial of this issue and the other issues remaining in the case.

Costs

Costs in favor of Continental.

REVERSED IN PART, VACATED IN PART, and REMANDED.