This case presents complex questions about the law of trademark and the law of sovereign immunity, as the latter applies to a state university. The contending parties are Phoenix International Software, Inc., a small software developer, and the Board of Regents of the University of Wisconsin System, which is an arm of the state of Wisconsin. Their dispute centers around two computer programs, each of which holds the registered trademark “CONDOR.” We delve into the details of the case below. For now, it is enough to say that two central issues have occupied us on appeal: first, the question whether the likelihood of confusion between Wisconsin’s CONDOR mark and Phoenix’s identical mark could be ascertained in summary judgment proceedings; and second, whether Wisconsin is entitled to immunity from Phoenix’s federal counterclaims. When we first heard this case, the panel unanimously concluded that summary judgment on the trademark dispute was inappropriate and thus further proceedings were needed, and a majority ruled that the university was entitled to immunity from Phoenix’s counterclaims. The panel granted rehearing limited to the immunity questions. We now reaffirm our ruling rejecting summary judgment; this portion of our opinion draws heavily on Judge Tinder’s original opinion. After the benefit of the arguments on rehearing, we conclude that the state is not entitled to assert sovereign immunity over the counterclaims, and so we reverse that part of the district court’s judgment as well.
I
Phoenix registered the CONDOR trademark in 1997 for software that runs on mainframe computers and provides online programming development, library management, and systems development; four years later, Wisconsin registered the identical mark, but for software that takes advantage of unused processing power *451across a network of computers. (For convenience, we occasionally refer to CONDOR-Phoenix and CONDOR-Wisconsin, to keep clear which product we are addressing.) In 2004, Phoenix filed a petition to cancel Wisconsin’s registration with the Trademark Trial and Appeal Board (TTAB), arguing that the state’s mark was likely to cause confusion in trade. See 15 U.S.C. §§ 1052(d) and 1064. Concluding that Phoenix had shown that confusion between the marks was likely, the TTAB granted the petition and cancelled Wisconsin’s registration. Phoenix Software Int’l v. Board of Regents of the Univ. of Wis. Sys., Cancellation No. 92042881 (T.T.A.B. Sept. 26, 2007).
Wisconsin decided to challenge the TTAB’s decision through a suit in the federal district court. This was one option that federal trademark law made available to it. See 15 U.S.C. § 1071(b). Another option would have been to appeal the TTAB’s decision to the U.S. Court of Appeals for the Federal Circuit. See 15 U.S.C. § 1071(a). Phoenix responded to Wisconsin’s action both by defending the TTAB’s decision and by asserting counterclaims for trademark infringement and false designation of origin under the Lanham Act, see 15 U.S.C. §§ 1114 and 1125(a). The district court dismissed Phoenix’s federal counterclaims on the ground that they were barred by the state’s sovereign immunity. The parties then filed cross-motions for summary judgment. The district court granted Wisconsin’s motion, denied Phoenix’s, and thereby reversed the TTAB’s decision to cancel Wisconsin’s registration.
As we have noted, our first opinion concluded unanimously that Phoenix was entitled to a trial on its confusion-in-trade allegation. The question whether Wisconsin was entitled to immunity, however, proved more difficult to resolve, and we thank the parties for their patience and their additional attention to this important issue. We now hold that the Supreme Court’s decision in Lapides v. Board of Regents of the University System of Georgia, 535 U.S. 613, 122 S.Ct. 1640, 152 L.Ed.2d 806 (2002), requires us to find that Wisconsin waived its sovereign immunity when it filed suit in the federal district court seeking to overturn the decision of the TTAB. To maximize its chances of reversing the agency’s decision, the state availed itself of the advantages of a fresh lawsuit, choosing that path over a number of others available. It would be anomalous if, after invoking federal jurisdiction, the state could declare that the federal court has no authority to consider related aspects of the case. Cf. Lapides, 535 U.S. at 619, 122 S.Ct. 1640. Phoenix’s counterclaims are compulsory in nature and thus lie well within the scope of Wisconsin’s waiver of immunity. Accordingly, we reverse the district court’s grant of summary judgment, reinstate Phoenix’s federal counterclaims, and remand for further proceedings.
II
A
Before turning to the merits, it is necessary to say a word about the standard of review. We have mentioned that trademark law provides two avenues for review of TTAB decisions. The road not taken by Wisconsin was a direct appeal to the Federal Circuit, which would have been restricted to the record developed before the TTAB and would have focused on whether substantial evidence supported the agency’s decision. CAE, Inc. v. Clean Air Eng’g, Inc., 267 F.3d 660, 675 & n. 9 (7th Cir.2001). The option Wisconsin chose was a new action in the district court. One advantage of this path for the state was the ability to expand the record *452by offering new evidence to fend off Phoenix’s cancellation claim. A challenge to the TTAB’s decision in a district court is “both an appeal and a new action, which allows the parties to request additional relief and to submit new evidence.” Id. at 673. In such an action, the district court wears two hats: “[it] is an appellate reviewer of facts found by the TTAB and is also a fact-finder based on new evidence introduced to the court.” Id. at 674. The court here properly followed this approach. It applied a deferential standard of review to the TTAB’s findings, in keeping with Dickinson v. Zurko, 527 U.S. 150, 119 S.Ct. 1816, 144 L.Ed.2d 143 (1999), and CAE, 267 F.3d at 675, and for summary judgment purposes the court viewed new evidence in the light most favorable to the nonmoving party.
This standard of review, in combination with the posture of the case and the issues presented, presents a real obstacle to summary judgment in Wisconsin’s favor. The central issue here — the likelihood of confusion between the parties’ trademarks — is a question of fact for the jury. AutoZone, Inc. v. Strick, 543 F.3d 923, 929 (7th Cir.2008). It was Phoenix that prevailed before the TTAB, and so the district court could not have ruled in Wisconsin’s favor without concluding either that no finder of fact could have thought that substantial evidence supported the TTAB’s determination, or that a legal error clouded the agency’s understanding of the likelihood-of-confusion issue. A party is entitled to summary judgment only if there exists no genuine issue of material fact and the moving party is entitled to judgment as a matter of law. FED. R. CIV. P. 56. Because Wisconsin was the moving party, it would have to point to compelling facts that it neglected to bring to the TTAB’s attention, and those facts had to be enough — viewed in the appropriate light— to require judgment in Wisconsin’s favor, even if credit was given to all the facts the TTAB found during its proceedings.
B
1. Findings of the Trademark Trial and Appeal Board
Before the TTAB was testimony from Wisconsin’s mainframe coordinator that “an organization that did not have a mainframe or is not involved in developing software application^] for mainframe computers” would have no use for Phoenix’s software, as well as testimony from Wisconsin’s software creator stating that the state’s software was not used on mainframe computers. Phoenix, however, countered with evidence that its software was not limited to mainframes. The company’s sole shareholder, Fred Hoschett, pointed out that many of his CONDOR customers did not have mainframes; he testified that “effectively we can run our software, unchanged, unaltered on a workstation, on someone’s desktop, as if it were on a mainframe” and that the software “often” operates on a network of workstations, which he defined as a “LAN, WAN or some other network that allows the interconnection of these workstations.” Hoschett also read the description of Wisconsin’s CONDOR software posted on the University’s website:
Condor is a specialized workload management system for computer-intensive jobs. Like other full-featured batch systems, Condor provides a job queuing mechanism, scheduling policy, priority scheme, resource monitoring, and resource management. Users submit their serial and parallel jobs to Condor. Condor places them into a queue, chooses when and where to run the jobs based upon a policy, carefully monitors their progress, and ultimately informs the user upon completion.
*453This language seriously concerned Hoschett; initially, he thought that it was describing his product. Phoenix described its software as “a toolbox of functionality to be used essentially by anyone who uses a computer to assist them in doing their jobs, whether it be programming software, submitted batch jobs and queuing batch jobs, or managing the environment or managing the resources.”
The TTAB found that there was “at least some evidence in the record that the parties’ respective software performs the same general functions and the evidence does not demonstrate the goods are used in distinctly different fields,” and that “there is no clear division between the parties’ software that would cause us to conclude that these products are not related.” The agency found the biggest difference between the two CONDORs to be the fact that Phoenix’s version was “used in a mainframe environment while [Wisconsin’s] goods are used in a network of individual computer workstations.” That distinction was “not necessarily significant,” however, because as one of Wisconsin’s witnesses conceded, “there might be [s]ome incentive” to operate “in both environments.” The agency made a number of additional significant findings. Both programs, it concluded, were “downloadable”; consumers of either program were sophisticated and could exercise special care with their purchases; there was no evidence that any consumer was actually confused as to the source of either product; and both parties’ marketing practices were “relatively limited,” though Wisconsin told the TTAB that it was expanding its operations, which the agency thought made the chances of confusion more likely.
Based on the record before it, the TTAB concluded that Phoenix had successfully proven the likelihood of confusion. It thus granted Phoenix’s petition to cancel Wisconsin’s registration of the CONDOR mark.
2. Findings of the District Court
At the district court, Phoenix attempted to supplement the record with evidence bolstering its position that the CONDOR software products performed overlapping functions. According to Hoschett, Wisconsin had struck deals with IBM to make Wisconsin’s version of CONDOR available on PC-based mainframes. Hoschett also provided a list of customers operating Phoenix’s version of CONDOR on PC-based mainframes. The district court, however, rejected the proffered evidence on the ground that it was filed too late. The account that follows is based on evidence that the district court found to be undisputed, for summary judgment purposes.
While Phoenix’s software cannot run on a network of workstations that are unconnected to a mainframe system, it can function on non-mainframe computers if emulation software is used. The Phoenix software allows “users to submit batch jobs to local and remote computers through a network of computers to more effectively utilize and balance the available computing cycles.” Phoenix’s customers must be specialized, because mainframe computer systems “are generally expensive computing systems that are extremely reliable and secure and capable of enormous throughput,” they are “centrally managed and maintained,” and a choice of software for use on a mainframe “requires careful consideration.” The end-users of software like Phoenix’s are “mainframe systems administrators and mainframe systems application developers.” These end-users form a tight-knit group that learns about products through word-of-mouth advertising, mainframe trade shows and conferences, and the advice of con-*454suiting firms. Advertising of CONDOR-Phoenix is done at trade shows, on the Internet, and through brochures. In 2000, the company spent approximately $65,000 on marketing; that number was virtually unchanged in 2003.
Although distribution and customer overlap between the two parties’ programs began at low levels, it has been growing. While CONDOR-Wisconsin does not run on mainframes, a mainframe might be part of the network of computers on which Wisconsin’s software is operating. Because the state distributes its software under an open software license, anyone may download and use the program for free. This means that users are hard to identify; Wisconsin estimates that the total number is in the tens of thousands. A person generally must have a “systems-level understanding” of computers to make CONDOR-Wisconsin work. Thus, typical users are systems operators of scientific research groups, such as the “high energy physics community, the DOE [Department of Energy] National Labs, biology and computer science departments, and industrial groups.” The evidence indicated that 3,738 copies of Wisconsin’s software were downloaded in 2000; by 2004, the number of downloads grew to 15,155, an increase of more than 400 percent. A promotional program offered by the University of Wisconsin, “CONDOR Week,” grew over the same time from a one-day event attracting 20 participants to a four-day event with more than 150 participants.
C
We come, then, to the central question on the trademark part of the appeal: whether consumers were likely to be confused by Wisconsin’s and Phoenix’s concurrent use of the CONDOR marks. The TTAB said yes, while the district court said no. The court found that the TTAB “erred when it considered the actual nature of the parties’ goods and misapplied the burden of proof to its determination of a likelihood of confusion.”
There are a number of multiple-factor tests that are used across the circuits to determine the likelihood of confusion. These are useful insofar as they operate as a checklist to ensure that we do not overlook relevant evidence, but they are a means to an end, not an end in themselves. This court has identified the following points as especially important:
1. the similarity between the marks in appearance and suggestion;
2. the similarity of the products;
3. the area and manner of concurrent use;
4. the degree of care likely to be exercised by consumers;
5. the strength of the plaintiffs mark;
6. any actual confusion; and
7. the intent of the defendant to “palm off’ his product as that of another.
AutoZone, 543 F.3d at 929. In reaching its decision, the lower court emphasized the similarity of the products and the area and manner of their use (factors 2 and 3). No one disputes that the marks are identical (factor 1, favoring Phoenix), nor that the proof of current actual confusion is weak (factor 6, favoring Wisconsin). There also is no accusation that “palming off’ was taking place (which removes factor 7 from the discussion). Customer sophistication (factor 4) appears to be a wash on these facts, and neither party has made anything of the strength of the mark (factor 5). We therefore may safely confíne our attention to the two points on which the district court focused: similarity of products and manner of use.
The district court identified two findings of the TTAB that it believed were in error. First, it thought that the TTAB should not *455have considered the “actual nature of the parties’ goods” but instead should have limited itself to the descriptions of the goods contained in the parties’ respective trademark registrations. In examining similarity of products, the district court— urged on by Wisconsin — emphasized similarity of function rather than actual or potential use. Its functional inquiry, moreover, was confined to an examination of the registration materials each party had filed. Second, the court believed that the TTAB should not have placed the burden on Wisconsin to prove that the parties’ goods were distinct in the way they were used or sold. Phoenix, it held, should have borne this burden, since Phoenix was trying to cancel the registration of a presumptively valid mark. After rejecting the TTAJB’s finding that the products were similar, the court concluded that Phoenix failed to meet its burden.
The district court erred by placing so much weight on the parties’ registration statements. To decide whether there is a likelihood of confusion between the two CONDOR products, a court must ask whether consumers, and specifically consumers who would use either product, would be likely to attribute them to a single source. AutoZone, 543 F.3d at 931; see also McGraw-Edison Co. v. Walt Disney Prod., 787 F.2d 1163, 1169 (7th Cir.1986). “[Dissimilarity is not dispositive of the likelihood of confusion inquiry. A likelihood of confusion may exist even if the parties are not in direct competition, ... or their products and services are not identical____Rather, because the rights of an owner of a registered trademark extend to any goods that might be, in the minds of consumers, ‘related,’ i.e., put out by a single producer, the more accurate inquiry is whether the public is likely to attribute the products ... to a single source.” CAE, 267 F.3d at 679 (citations omitted); see also Eli Lilly & Co. v. Natural Answers, Inc., 233 F.3d 456, 463 (7th Cir.2000). The TTAB’s opinion shows that it asked the right question and applied the correct standard.
The descriptions in the trademark registry are thus of little help. Products do not even have to perform similar functions, much less be described identically, for a likelihood of confusion to exist. We have found that a likelihood of confusion can exist between a mark for electrical fuses and the mark on Disney’s merchandise for the movie Tron, McGraw-Edison, 787 F.2d at 1169; between a mark registered by a company that designed and manufactured sophisticated measuring equipment and a mark registered by a company that tested facilities for compliance with pollution laws, CAE, 267 F.3d at 679; or between an auto-parts retailer’s mark and the mark of an oil change and carwash operator, AutoZone, 543 F.3d at 931-32. All of these cases featured products displaying less similarity than those at issue here. In most of the cases, moreover, the products had identical marks. While more than the same mark is needed to show confusion, see, e.g., M2 Software, Inc. v. M2 Communications, Inc., 450 F.3d 1378, 1385 (Fed.Cir.2006), the presence of such identity often creates triable issues of fact regarding the various ways a product is marketed.
The Federal Circuit case on which the district court relied, Octocom Systems, Inc. v. Houston Computer Services, Inc., 918 F.2d 937, 942 (Fed.Cir.1990), is not to the contrary. There a registrant whose mark was challenged tried to supplement the registration to show that it intended its mark to cover a narrower set of goods than those described in the registration. The Federal Circuit rejected this attempt, saying that the court should consider the goods as described in the registration. A *456likelihood of confusion existed because the registrant’s original application “encompassed modems and computer programs” and thus conflicted with the petitioner’s registration of a similar mark for computer programs. Furthermore, the record showed that modems and computer programs are used together in networking; they can come from a single source; and they may be identified with the same mark. Id. at 943. The critical fact in Octocom was thus not that the registrations were identical; it was that they covered similar products.
Similarity may be reflected “expressly or inherently.” Id. at 942. While a court should consider the marks as they are described in the registration, it is error to bar any evidence of their actual use as irrelevant. One of the factors in the test for likelihood of confusion, after all, is the area and manner of concurrent use. The actual use of a product is important in its own right and information about use is relevant to explain the meaning of the terms used in the registration. Id. at 943; see also Forum Corp. of N. Am. v. Forum, Ltd., 903 F.2d 434, 442 (7th Cir.1990). The TTAB committed no error when it declared that it was considering the nature of the parties’ goods “[t]o the extent that these facts provide some information about the market and purchasers of these goods.”
Our disagreement with the district court’s rationale, however, does not dispose of the case. Our review is de novo, and we may affirm on any ground supported by the record. To defeat Wisconsin’s motion for summary judgment, Phoenix had to produce enough evidence to create an issue for the trier of fact. Wisconsin would still be entitled to summary judgment if there were no evidence that users would be likely to confuse a product running on a mainframe system (CONDOR-Phoenix) with a product that runs on a network of individual computer workstations (CONDOR-Wisconsin).
The TTAB offered three reasons for cancelling Wisconsin’s registration. First, and most importantly, it relied on the conceded fact that the marks are identical. Second, the TTAB found that the two software products perform similar functions and thus cannot be said to occupy unrelated fields. Third, it found that “sophisticated purchasers would likely believe that there is some relationship or association between the sources of the goods under these circumstances.” These findings tipped the balance on the questions of similarity and manner of use in favor of Phoenix.
The TTAB credited Phoenix’s evidence that its mainframe software can operate on a network; it found that the two CONDORS perform similar functions, noting that Wisconsin’s description of its product sounded very much like Phoenix’s; there was evidence that the products were delivered the same way; the same customers were likely to encounter both products, particularly in light of Wisconsin’s expanded marketing efforts; and there was some incentive to operate in both mainframe and network environments. Wisconsin offered new evidence in the district court to rebut these findings, and it argued that the facts should be interpreted differently. It relied most heavily on the sophistication of consumers and the theory that mainframe purchasers take care when choosing what product to buy. This is relevant to, but not dispositive of the likelihood of confusion issue, see 4 McCarthy on Trademarks and Unfair Competition § 23:103 (4th ed.2011). But the question is not whether purchasers of Phoenix’s CONDOR product would accidentally buy Wisconsin’s product; it is whether those consumers would likely attribute them to a single source. *457Moreover, the TTAB credited Phoenix’s witness Hoschett, who testified that he was confused by Wisconsin’s description of its product, and we are bound to give this important credibility finding deference. Wisconsin’s evidence of sophistication is not compelling enough to eliminate any issue of fact.
The question in the end is not whether the evidence compelled a finding in favor of Phoenix. Wisconsin has pointed to a number of facts in its favor: there was no actual confusion; the “downloadability” of both programs is not dispositive of whether the products were sold in similar trade channels; Phoenix may not have been diligent about protecting its mark; and any confusion might be quickly rectified. But the record includes enough evidence supporting Phoenix that further proceedings are necessary. Accordingly, we must reverse the district court and remand for a trial on the likelihood of confusion issue.
Ill
Phoenix also asks that we reinstate on remand the federal counterclaims that it asserted against Wisconsin. Behind Phoenix’s request is a difficult question of constitutional law: are Phoenix’s counterclaims against Wisconsin barred by the sovereign immunity doctrine that the Supreme Court has found reflected in the Eleventh Amendment to the U.S. Constitution? The district court thought so. Congress, the court said, has not abrogated Wisconsin’s immunity in this area; the state’s decision to participate in the federal trademark system did not effect a waiver of immunity; and the state had done nothing to voluntarily invoke federal jurisdiction. The district court regarded Wisconsin’s appearance in federal court as nothing more than the involuntary appeal of an unfavorable agency decision. In many respects, the district court’s analysis of Wisconsin’s immunity was correct. In our view, however, Wisconsin’s litigation conduct in this case was sufficient to waive its sovereign immunity with respect to the counterclaims Phoenix has asserted.
While the language of the Eleventh Amendment literally says only that federal jurisdiction is limited where a state is sued by citizens of another state or foreign country, the Supreme Court has “understood the Eleventh Amendment to stand not so much for what it says, but for the presupposition of our constitutional structure which it confirms.” Blatchford v. Native Vill. of Noatak and Circle Vill., 501 U.S. 775, 779, 111 S.Ct. 2578, 115 L.Ed.2d 686 (1991). The power established in Article III does not supersede “the sovereign immunity that the States possessed before entering the Union.” College Sav. Bank v. Florida Prepaid Postsecondary Educ. Expense Bd., 527 U.S. 666, 669, 119 S.Ct. 2219, 144 L.Ed.2d 605 (1999) (“College Savings ”). So interpreted, the Eleventh Amendment guarantees that “an unconsenting State is immune from suits brought in federal courts by her own citizens as well as by citizens of another State.” Edelman v. Jordan, 415 U.S. 651, 662-63, 94 S.Ct. 1347, 39 L.Ed.2d 662 (1974).
This robust immunity from suit, however, is not absolute. Two exceptions are potentially relevant to this case. First, Congress can authorize suits against the states by exercising its power to enforce the Fourteenth Amendment to the Constitution, see College Savings, 527 U.S. at 670, 119 S.Ct. 2219 (citing Fitzpatrick v. Bitzer, 427 U.S. 445, 96 S.Ct. 2666, 49 L.Ed.2d 614 (1976)); this exception recognizes that the Fourteenth Amendment, ratified more than 70 years after the Eleventh Amendment, represents a fundamental reorganization of our federal system and a limitation on the sovereign power of *458the states. Second, a state may voluntarily waive its sovereign immunity by consenting to federal jurisdiction explicitly or by invoking that jurisdiction through its behavior. See id. (citing Clark v. Barnard, 108 U.S. 436, 447-48, 2 S.Ct. 878, 27 L.Ed. 780 (1883)). Phoenix takes the position that its counterclaims should move forward under either exception.
A
We can be brief on the subject of congressional abrogation. After Wisconsin initiated its action in the district court, Phoenix counterclaimed for infringement and false designation of origin under 15 U.S.C. §§ 1114 and 1125(a). Both of these statutes demonstrate Congress’s intention to subject the state to liability in trademark actions brought by those injured by a state’s acts. Nonetheless, we doubt that either provision would survive a constitutional challenge in the Supreme Court. College Savings held that a provision of the Trademark Remedy Clarification Act (TRCA) making states liable for false advertising (one form that a claim under 15 U.S.C. § 1125(a) may take) violated the Constitution. 527 U.S. at 691, 119 S.Ct. 2219. In addition, Florida Prepaid Post-secondary Education Expense Board v. College Savings Bank (“Florida Prepaid ”), found that the TRCA’s sister statute, which established state liability for patent infringement was similarly unconstitutional. 527 U.S. 627, 647-48, 119 S.Ct. 2199, 144 L.Ed.2d 575 (1999). As the district court noted, these decisions appear to foreclose any argument that Congress has properly abrogated Wisconsin’s immunity from Phoenix’s federal counterclaims. We leave closer examination of the issue for another day, however, because in our view the state has waived its immunity from suit.
B
The Supreme Court has long recognized that a state may waive its sovereign immunity. E.g., Clark, 108 U.S. 436, 2 S.Ct. 878. “Generally, we will find a waiver either if the State voluntarily invokes our jurisdiction, ... or else if the State makes a clear declaration that it intends to submit itself to our jurisdiction.” College Savings, 527 U.S. at 675-76, 119 S.Ct. 2219 (internal quotation marks and citations omitted). One way in which a state may submit itself to federal jurisdiction is through its conduct during litigation. In 1883, the Supreme Court held that when a state makes a “voluntary appearance” in federal court as an intervenor, that participation amounts to a waiver of immunity. Clark, 108 U.S. at 447, 2 S.Ct. 878. To the same effect, it held in Gunter v. Atlantic Coast Line R.R. Co., 200 U.S. 273, 284, 26 S.Ct. 252, 50 L.Ed. 477 (1906), that “where a state voluntarily become a party to a cause, and submits its rights for judicial determination, it will be bound thereby, and cannot escape the result of its own voluntary act by invoking the prohibitions of the 11th Amendment.” Later, it wrote in Gardner v. New Jersey that, in the context of a bankruptcy dispute, a state that voluntarily files in federal court “waives any immunity ... respecting the adjudication of the claim.” 329 U.S. 565, 574, 67 S.Ct. 467, 91 L.Ed. 504 (1947). Most recently, the Court unanimously reaffirmed this principle in Lapides: “The Court has long accepted this statement of the law as valid, often citing with approval the cases embodying that principle.” 535 U.S. at 619, 122 S.Ct. 1640.
Lapides confirms that the Court did not eliminate the doctrine of waiver by litigation conduct in Ford Motor Co. v. Department of Treasury of Indiana, 323 U.S. 459, 65 S.Ct. 347, 89 L.Ed. 389 (1945). Ford held that a state could assert its *459sovereign immunity for the first time in the Supreme Court, despite the state attorney general’s defense on the merits in lower courts. Id. at 467-69, 65 S.Ct. 347. For a time, it seemed Ford was in tension with the view that voluntary invocation of federal jurisdiction waives immunity. But the Court’s decision in Wisconsin Department of Corrections v. Schacht, 524 U.S. 381, 118 S.Ct. 2047, 141 L.Ed.2d 364 (1998), resolved any such tension. There, the Court wrote that “[t]he Eleventh Amendment ... does not automatically destroy original jurisdiction. Rather, the Eleventh Amendment grants the State a legal power to assert a sovereign immunity defense should it choose to do so. The State can waive the defense.... Nor need a court raise the defect on its own. Unless the State raises the matter, a court can ignore it.” Id. at 389, 118 S.Ct. 2047 (citations omitted). The year after Schacht, the Court reaffirmed “the unremarkable proposition that a State waives its sovereign immunity by voluntarily invoking the jurisdiction of the federal courts.” College Savings, 527 U.S. at 681 n. 3, 119 S.Ct. 2219.
The question in Lapides was “whether a state waive[s] its Eleventh Amendment immunity by its affirmative litigation conduct when it removes a case to federal court....” 535 U.S. at 617, 122 S.Ct. 1640 (internal quotation marks omitted). Lapides, a professor employed by the Georgia state university system, had sued the Board of Regents in state court in both their personal and official capacities. He asserted that they had violated both state law and his federal constitutional rights by placing allegations of sexual harassment in his personnel file. The state defendants removed the case to federal court, where they promptly sought dismissal of the official-action claims on state sovereign immunity grounds. The Court concluded that it would adhere to the rule in Gunter, quoted above, and stressed that a state cannot use the Eleventh Amendment as a get-out-of-court-free card-when it voluntarily submits to a federal tribunal for a judicial determination of its rights. Id. at 619, 122 S.Ct. 1640 (quoting Gunter, 200 U.S. at 284, 26 S.Ct. 252). It offered a number of reasons for its endorsement of waiver by litigation conduct:
[A]n interpretation. of the Eleventh Amendment that finds waiver in the litigation context rests upon the Amendment’s presumed recognition of the judicial need to avoid inconsistency, anomaly, and unfairness, and not upon a State’s actual preference or desire, which might, after all, favor selective use of “immunity” to achieve litigation advantages.... The relevant “clarity” here must focus on the litigation act the State takes that creates the waiver. And that act — removal-—-is clear.
Id. at 620, 122 S.Ct. 1640 (citations omitted). Central to the holding in Ford, the Court said, was the fact that it “involved a State that a private plaintiff had involuntarily made a defendant in federal court.” Id. at 622, 122 S.Ct. 1640. Concluding “that Clark, Gunter, and Gardner represent the sounder line of authority,” the Court “[found] Ford inconsistent with the basic rationale of that line of cases” and “overr.ule[d] Ford insofar as it would otherwise apply.” Id. at 623, 122 S.Ct. 1640.
The Court could not have expressed itself more plainly. Ford is limited to its facts; states can waive their immunity by voluntary conduct in particular cases; and the potential sovereign immunity of a state does not implicate the federal court’s subject-matter jurisdiction. The question is how to apply these broad principles to the case before us.
1. Constructive Waiver
It is important for purposes of the waiver inquiry to be precise about what *460aspect of Wisconsin’s conduct we are talking about. Although as a theoretical matter one might consider whether Wisconsin’s decision to participate in the federal trademark system at all amounts to constructive waiver, we put that possibility to one side. College Savings rejected the same argument in the patent context, and the Court’s decision appears to eliminate the doctrine of constructive waiver outright. See 527 U.S. at 680-84, 119 S.Ct. 2219. This rules out any possibility that Wisconsin subjected itself to suit by registering its trademark with the Patent and Trademark Office. Participation in the trademark system is not the sort of conduct that voluntarily invokes the jurisdiction of the federal courts. Importantly, however, the Court took care in Lapides to recall that “College Savings Bank distinguished the kind of constructive waivers repudiated there from waivers effected by litigation conduct.” Lapides, 535 U.S. at 620, 122 S.Ct. 1640.
2. Waiver by Litigation Conduct
Quite separate from its general participation in the trademark system is Wisconsin’s specific conduct in this lawsuit. The question is whether the state’s decision to challenge the TTAB’s adverse decision by filing a lawsuit in federal district court effected the type of waiver of immunity that the Supreme Court discussed in Lapides.
We must consider two preliminary questions about the scope of the Lapides rule before analyzing what Wisconsin did here. The first is whether Lapides applies to all instances in which a state removes a case to federal court; the second is whether the removal mechanism is central to the Court’s holding in Lapides, or whether other paths to federal court cause a similar waiver of the immunity defense. To answer both, it is helpful to take a closer look at the Lapides litigation.
The dispute in Lapides began as a lawsuit against the State of Georgia in a Georgia court under a state law that explicitly waived Georgia’s immunity to damages in state court. 535 U.S. at 617, 122 S.Ct. 1640. In addition to the state-law claim, the plaintiff in Lapides asserted a claim against Georgia based on § 1983. The state removed both claims to federal court. The latter claim, the Court noted, could not go forward because a state is not a “person” for purposes of § 1983. Lapides, 535 U.S. at 617, 122 S.Ct. 1640 (citing Will v. Michigan Dep’t of State Police, 491 U.S. 58, 66, 109 S.Ct. 2304, 105 L.Ed.2d 45 (1989)). That is why the Court was careful to say that its conclusion that the state’s act of removing the case to federal court led to a waiver of its sovereign immunity was reached in “the context of state-law claims, in respect to which the State has explicitly waived immunity from state-court proceedings.” Id. The Court’s rationale, however, was not so limited:
It would seem anomalous or inconsistent for a State both (1) to invoke federal jurisdiction, thereby contending that the “Judicial power of the United States” extends to the case at hand, and (2) to claim Eleventh Amendment immunity, thereby denying that the “Judicial power of the United States” extends to the case at hand. And a Constitution that permitted States to follow their litigation interests by freely asserting both claims in the same case could generate seriously unfair results.
Id. at 619, 122 S.Ct. 1640. The Court added that “[a] benign motive [for removing to federal court] ... cannot make the critical difference for which Georgia hopes. Motives are difficult to evaluate, while jurisdictional rules should be clear.... To adopt the State’s Eleventh Amendment po*461sition would permit States to achieve unfair tactical advantages, if not in this case, in others.” Id. at 621, 122 S.Ct. 1640 (internal citations omitted). We observed in our en banc decision in United States v. Skoien that “[t]his is the sort of message that, whether or not technically dictum, a court of appeals must respect, given the Supreme Court’s entitlement to speak through its opinions as well as through its technical holdings.” 614 F.3d 638, 641 (7th Cir.2010) (en banc). That principle applies with equal force to the logic applied by the Court in Lapides.
Reflecting that spirit, most Courts of Appeals have applied the rule of Lapides to all instances of removal initiated by a state. The Fifth Circuit’s decision in Meyers ex rel. Benzing v. Texas explains why this is the proper result:
[I]n formulating its rationale, the Court did not restrict itself to facts, rules, or reasons peculiar to the Lapides case. Rather, throughout its opinion, the Court’s reasoning, rule-making, and choice of precepts were derived from generally applicable principles serving “the judicial need to avoid inconsistency, anomaly, and unfairness” in states’ claims of immunity in all types of federal litigation.
410 F.3d 236, 244 (5th Cir.2005) (quoting Lapides, 535 U.S. at 620, 122 S.Ct. 1640). See also Lombardo v. Pennsylvania Dep’t of Public Welfare, 540 F.3d 190, 198 (3d Cir.2008) (“We hold that the [Statej’s removal of federal-law claims to federal court effected a waiver of immunity from suit in federal court.”); Embury v. King, 361 F.3d 562, 564 (9th Cir.2004) (finding that Lapides applies to action removed by the state to federal court based on either state or federal law); Estes v. Wyoming Dep’t of Transp., 302 F.3d 1200, 1206 (10th Cir.2002) (same). In fact, only the Fourth Circuit has concluded that Lapides should be limited to its facts. Stewart v. North Carolina, 393 F.3d 484, 488-90 (4th Cir.2005); but see Watters v. Washington Metro. Area Transit Auth., 295 F.3d 36, 42 n. 13 (D.C.Cir.2002) (suggesting that the holding of Lapides is narrow but refusing to consider the issue because the parties had not raised it).
To date, this court has had the opportunity to apply Lapides only in circumstances functionally equivalent to those that were at issue in Lapides itself. See Omosegbon v. Wells, 335 F.3d 668 (7th Cir.2003) (considering a claim based on a state law waiving immunity that was removed to federal court by a defendant state). Contrary to Wisconsin’s suggestion, however, none of our past decisions confínes Lapides to those limited circumstances. Like the Fifth Circuit, we regard the Fourth Circuit’s Stewart decision as an outlier that “misconstrues important principles animating Lapides,” Meyers, 410 F.3d at 249, and we join the majority of our other sister circuits in reading Lapides to state a more general rule.
That brings us to the question whether anything in Lapides turned on the fact that the case reached the federal court through removal. We think not. As the Lapides Court explained, “In large part the rule governing voluntary invocations of federal jurisdiction has rested upon the problems of inconsistency and unfairness that a contrary rule of law would create.” 535 U.S. at 622, 122 S.Ct. 1640. “[Removal is a form of voluntary invocation of a federal court’s jurisdiction sufficient to waive the State’s otherwise valid objection to litigation of the matter (here of state law) in a federal forum.” Id. at 624, 122 S.Ct. 1640. But it is, in the end, just a mechanism for invoking the federal court’s jurisdiction. There is no reason to think that the state’s use of any other mechanism — -such as filing an original action in *462federal court — carries less force for waiver purposes.
The Supreme Court took care in Lapides to point out that “the State was brought involuntarily into the case as a defendant” but then “voluntarily agreed to remove the case to federal court.” Id. at 620, 122 S.Ct. 1640. Waivers by litigation conduct depend on whether the state has made a voluntary change in behavior that demonstrates it is no longer defending the lawsuit and is instead taking advantage of the federal forum. This is not a question of what label the state assumes in litigation: the analysis of waiver by litigation conduct should not turn on whether the state is a defendant, as in Lapides, an intervening claimant, as in Clark, 108 U.S. at 447, 2 S.Ct. 878, a plaintiff, as Wisconsin is here, or even an appellant in this court, see Indiana Prot. and Advocacy Servs. v. Indiana Family and Soc. Services Admin., 603 F.3d 365, 370-71 (7th Cir.2010) (en banc). Instead, the crucial considerations are the voluntariness of the state’s choice of forum and the functional consequences of that choice.
The Federal Circuit has had occasion to consider how the voluntary invocation principles in Lapides apply outside of the removal context. It decided that a state that initiates and prevails in a patent interference proceeding against a competing applicant cannot use sovereign immunity to block an appeal to federal court of the agency’s decision. Vas-Cath, Inc. v. Curators of Univ. of Missouri, 473 F.3d 1376, 1385 (Fed.Cir.2007). In addition, the Federal Circuit has held that a state that files suit in federal court to enforce a patent claim consents to all compulsory counterclaims that arise from the same transaction or occurrence. Regents of the Univ. of New Mexico v. Knight, 321 F.3d 1111, 1125-26 (Fed.Cir.2003). In contrast, it has refused to find waiver where a state is sued and the plaintiff claims that the state has waived its immunity because the state had waived its immunity in an earlier lawsuit involving the same parties. See Biomedical Patent Mgmt. Corp. v. California Dep’t of Health Servs., 505 F.3d 1328, 1334-41 (Fed.Cir.2007). Similarly, that court has decided that a state that files a lawsuit in one district court does not waive its immunity in a related lawsuit filed by a party in another district court, at least in situations where that party could have intervened in the action filed by the state. Tegic Communications Corp. v. Board of Regents of the Univ. of Texas Sys., 458 F.3d 1335, 1342 (Fed.Cir.2006). Lastly, the Federal Circuit has found no waiver when a state simply defends against a Lanham Act claim in federal court. See State Contracting & Eng’g Corp. v. Florida, 258 F.3d 1329, 1336 (Fed.Cir.2001).
Phoenix’s case presents an additional wrinkle: Wisconsin did not initiate the proceeding before the administrative agency, but it did choose to go to court after Phoenix prevailed in the TTAB. Cases in the First and Eighth Circuits shed light on this scenario. In New Hampshire v. Ramsey, 366 F.3d 1, 16-17 (1st Cir.2004), the First Circuit held that when a state voluntarily participates in proceedings before a federal arbitration panel without raising a sovereign immunity defense, the state cannot challenge the arbitral decision in federal district court by claiming it was entitled to immunity from suit. The court reached this conclusion “even though [the state] was not formally the plaintiff in the administrative proceeding.” Id. at 16. Any other decision, it reasoned, would allow the state to gain an unfair advantage. Id. at 16-17. Earlier, the same court had determined that the state’s participation as a defendant in administrative proceedings did not waive a sovereign immunity defense in federal court when the state had “consistently asserted its sovereign immu*463nity, both [in federal court] and in the administrative proceeding.” Rhode Island Dep’t of Envtl. Mgmt. v. United States, 304 F.3d 31, 49 (1st Cir.2002).
The First Circuit’s most recent decision involving state sovereign immunity and administrative proceedings concluded that a state maintained its sovereign immunity even though the state initiated proceedings before an Administrative Law Judge (ALJ). Taylor v. U.S. Dep’t of Labor, 440 F.3d 1 (1st Cir.2005). While at first glance Taylor seems to contradict the First Circuit’s earlier decisions, the case involved a “slightly different” situation, id. at 5, and its unusual circumstances provide little help in evaluating Wisconsin’s litigation conduct in our case. The state in Taylor faced a situation where the sovereign immunity defense “was not available at the investigatory stage of the administrative proceedings.” Id. at 8. As a result, the state requested a proceeding before an ALJ precisely for the purpose of asserting its sovereign immunity, and it claimed its immunity at the first opportunity once that proceeding began. Id. at 7-8. Together, these cases suggest that a state may be required to assert its immunity at the first opportunity in administrative proceedings or risk being deemed to have waived its immunity by its litigation conduct. Such an approach would be consistent with the Supreme Court’s recognition that state sovereign immunity extends to administrative proceedings initiated by private parties. See Federal Maritime Comm’n v. South Carolina State Ports Auth., 535 U.S. 743, 760, 122 S.Ct. 1864, 152 L.Ed.2d 962 (2002).
Most recently, the Eighth Circuit held in United States v. Metropolitan St. Louis Sewer Dist., 578 F.3d 722 (8th Cir.2009), that the State of Missouri had waived its sovereign immunity by litigation conduct in a case that is strikingly similar to this one. Missouri joined the United States as a plaintiff in an enforcement action based on the Clean Water Act, filing a complaint against a local water and sewage district that had discharged untreated wastewater. The district counterclaimed, arguing that the state was required (by federal law and because of equitable considerations) to indemnify it for any costs that it might incur as a result of the suit. In response to the counterclaims, Missouri asserted its sovereign immunity defense. It pointed to section 309(e) of the Clean Water Act, see 33 U.S.C. § 1319(e), which provides that whenever a municipality is sued by the United States under the statute, “the State in which such municipality is located shall be joined as a party.” Because the statute compelled its participation as a party, Missouri argued, it had not taken any step during the litigation that could be seen as inconsistent with an assertion of sovereign immunity. The Eight Circuit disagreed. Pointing to Lapides, the court said, “The filing of a complaint in a federal district court is the quintessential means of invoking its jurisdiction. There is no indication in the record that Missouri was reluctant to proceed as a coplaintiff....” Metropolitan St. Louis Sewer Dist., 578 F.3d at 725. The state’s litigation conduct was thus sufficient to waive its immunity, and the water district’s counterclaims were allowed to proceed.
We need not explore these cases in any more detail to resolve the dispute between Wisconsin and Phoenix. The distinction between a voluntary, active decision by the state to entrust a matter to federal court and involuntary, defensive measures is reflected in the Supreme Court’s decisions addressing waiver by litigation conduct. When a state chooses to intervene in a federal case, it waives its immunity for purposes of those proceedings. Clark, 108 U.S. at 447-48, 2 S.Ct. *464878. If a state voluntarily files a claim in federal court, waiver once again occurs. Gardner, 329 U.S. at 574, 67 S.Ct. 467. Moreover, a waiver of immunity in an initial proceeding extends to all ancillary proceedings that follow. Gunter, 200 U.S. at 281-82, 289-90, 26 S.Ct. 252. Here, Wisconsin not only declined to raise its immunity during the administrative proceedings, it also decided to challenge the TTAB’s determination by initiating a civil action in the federal district court. As we have noted, this type of civil action is “both an appeal and a new action,” CAE, 267 F.3d at 673, and it reflects exactly the sort of affirmative decision to place a dispute in the federal court’s hands that effects a waiver of immunity.
Wisconsin insists that the necessary element of voluntary behavior is missing here, but none of its arguments withstands close examination. The state complains that it was forced to bring this lawsuit in the district court once the TTAB ruled adversely to it. We disagree, for the simple reason that Wisconsin was not compelled to do anything at all. Just as Georgia in Lapides had the option of litigating in its home court rather than removing to federal court, Wisconsin here enjoyed a number of options, and each one carried a different implication for sovereign immunity-
Wisconsin chose to challenge the TTAB’s decision granting Phoenix’s petition and cancelling the state’s registration by filing a complaint in the district court. But resort to that court was far from Wisconsin’s only choice. In fact, we can think of at least five options that Wisconsin had: (1) the state could have done nothing and let the TTAB’s decision stand; (2) it might have refused to acquiesce in the agency’s decision or the TTAB proceedings in the first place; (3) it could have taken action against Phoenix in state court before the agency proceedings began; (4) it could have appealed the TTAB’s decision directly to the Federal Circuit, 15 U.S.C. § 1071(a); or (5) it could have filed (as it did) a civil action in district court challenging the agency’s decision, 15 U.S.C. § 1071(b). An exploration of these paths and the consequences each carried illustrates why Wisconsin’s choice can only be seen as a voluntary invocation of federal jurisdiction.
a. Do Nothing. Wisconsin could have acquiesced in the TTAB’s cancellation of its mark and found a new trademark for its software. Nothing forced it to spend a minute in federal court. The state may object that it is unfair to make it choose between maintaining its immunity and challenging an adverse agency decision. We disagree. Although College Savings forecloses a theory of constructive waiver based solely on the fact of the state’s participation in the trademark system, there is no reason to expand this principle to its outer limits. (As we explain in the final section of this opinion, the evolution of the sovereign immunity doctrine suggests that there is every reason to be cautious about expansions when we are dealing with the states’ commercial activities.) Administrative agencies routinely resolve the rights of parties before them, and Congress does not always see fit to provide for recourse to the courts. Indeed, we are not familiar with any administrative scheme in which the party who loses before the agency is compelled to appeal the adverse decision. Only if a challenge to the TTAB decision in federal district court is a necessary facet of any party’s participation in the trademark system would College Savings dictate the conclusion that the state’s invocation of federal jurisdiction is involuntary and thus is inconsistent with a waiver of sovereign immunity.
*465Nor is there anything particularly offensive about requiring a state to accept the TTAB’s decision without further recourse. Other regimes pass constitutional muster where federal court review of agency adjudication is curtailed, and they concern rights at least as precious as a party’s interest in a trademark. In 2010, the Supreme Court reaffirmed that judicial review of the Attorney General’s decision to deport a person from the United States may be limited, see Kucana v. Holder, — U.S. —, 130 S.Ct. 827, 831, 175 L.Ed.2d 694 (2010), and we have recognized in at least two en banc decisions that review of an agency’s entitlement determinations may be foreclosed, see Czerkies v. U.S. Dep’t of Labor, 73 F.3d 1435, 1437-39 (7th Cir.1996) (en banc) (recognizing that a federal court may not review a challenge to a benefits determination of the Office of Workers’ Compensation Programs in the Department of Labor); Marozsan v. United States, 852 F.2d 1469, 1473 n. 10 (7th Cir.1988) (en banc) (discussing how veterans may not obtain federal-court review of an individual claims determination by the Veterans Administration). In many circumstances, Congress has seen fit to limit the role of federal courts in disputes about important rights, and courts do not regard the limitation as impermissible. Consider, for example, the jurisdiction-limiting provisions in the Prison Litigation Reform Act of 1996, Pub.L. No. 104-134,110 Stat. 1321 (Apr. 26, 1996), and the Antiterrorism and Effective Death Penalty Act of 1996, Pub.L. No. 104-132, 110 Stat. 1214 (Apr. 24, 1996). If there is nothing wrong with foreclosing review of the TTAB’s decision altogether, then there is nothing unfair about compelling a state to choose between its sovereign immunity and a second bite at the apple after the agency has spoken.
b. Refuse to Acquiesce in the TTAB’s Decision or the Agency Proceedings. Wisconsin also could have refused to participate at an earlier stage of the TTAB’s proceedings. As we have mentioned, the Supreme Court recognized in Federal Maritime Commission, 535 U.S. at 760, 122 S.Ct. 1864, that state sovereign immunity applies to privately initiated agency proceedings. Similarly, the state might have refused to acquiesce in the TTAB’s determination after the decision issued (though we certainly are not suggesting that this would be a responsible course of action). In either case, Phoenix would have been forced to sue in state court— either for a determination of its rights or for enforcement of the agency’s decision— and the state would either have retained its immunity or enjoyed the benefit of litigating under state law in its home courts.
c. File a Lawsuit in State Court. Wisconsin also ignores that at one point in the history of the case it had another state-court avenue available. Before Phoenix initiated TTAB proceedings, Wisconsin could have filed a suit against Phoenix in state court to resolve who had rights to the CONDOR mark. Although the federal courts have jurisdiction over trademark claims brought under the Lanham Act, that jurisdiction is not exclusive. See 28 U.S.C. § 1338(a); 15 U.S.C. § 1121(a); Alpharma, Inc. v. Pennfield Oil Co., 411 F.3d 934, 938 (8th Cir.2005); Aquatherm Indus., Inc. v. Florida Power & Light Co., 84 F.3d 1388, 1394 (11th Cir.1996). A party alleging a trademark violation under the statute may litigate in state court if it so chooses. Accordingly, if Wisconsin was concerned about Phoenix’s competing CONDOR trademark and wished to preserve its sovereign immunity, it had the option of filing its own infringement action in Wisconsin state court.
d. Appeal to the Federal Circuit. Putting the first three options to one side, Wisconsin’s sovereign immunity claim los*466es any remaining force when one realizes that the state freely chose to challenge the TTAB decision in federal district court rather than in the Federal Circuit. Compare 15 U.S.C. § 1071(a), with 15 U.S.C. § 1071(b). Had Wisconsin opted instead for the Federal Circuit, Phoenix would not have been able to introduce counterclaims into the litigation. (There is no provision in the Federal Rules of Appellate Procedure that would allow an appellee to introduce a new counterclaim.) The Federal Circuit would have been able to focus exclusively on the TTAB’s decision, without any threat to Wisconsin’s sovereign immunity.
Wisconsin argues that the possibility of appealing to the Federal Circuit should not affect our view of whether it waived immunity by its litigation conduct, because even if it had taken such an appeal, Phoenix could have forced it back into the federal district court. The state is referring to 15 U.S.C. § 1071(a), which provides that an adverse party to the TTAB proceedings (like Phoenix) can move a challenge to the agency decision that is filed in the Federal Circuit to a federal district court. But it is hard to see how this helps the state. As Phoenix conceded during oral argument, if it had been the one invoking the district court’s jurisdiction, whether by forcing Wisconsin into district court after the state filed a challenge in the Federal Circuit or by any other means, it could not then claim that Wisconsin had waived its immunity by litigation conduct. Wisconsin would have found itself forced into district court and fully entitled to sovereign immunity.
e. File an Action in District Court. Wisconsin’s decision to file its challenge to the TTAB decision in the district court must be understood against the backdrop of the full range of options it had. The state argues strenuously that the case it filed was nothing more than an appeal of an adverse agency determination in a proceeding in which it was an unwilling party. But even if we put to one side the fact that Wisconsin failed to assert its immunity in the administrative proceeding and we ignore the fact that the district court proceeding is “both an appeal and a new action,” CAE, 267 F.3d at 673, Wisconsin’s argument fails because it misconstrues the relevant inquiry after Lapides. When we judge whether litigation conduct effected a waiver of immunity, we do not focus exclusively on whether the district court proceedings were somehow related to the agency proceedings that came beforehand; instead, we must analyze why Wisconsin chose to proceed in the manner that it did and the consequences that its decision carried. In this case, the question is why Wisconsin chose to attack the agency decision in the district court rather than in the Federal Circuit (or through any of the other options we have identified), given that the Federal Circuit could provide the state with all of the relief it sought and insulation from Phoenix’s counterclaims.
An animating principle of Lapides is that a state should not reap litigation advantages through its selection of a forum and subsequent assertion of sovereign immunity as a defense. The choice of the district court came with at least three advantages for Wisconsin: it provided the opportunity to introduce new evidence that had not been provided to the TTAB (review in the Federal Circuit would have been confined to the agency record); it allowed the state to add supplemental claims to broaden the relief sought; and it provided (at least in part) a de novo standard of review, rather than the more deferential stance that the Federal Circuit would have taken toward agency findings and conclusions. See CAE, 267 F.3d at 673; see also City of Chicago v. Interna*467tional Coll. of Surgeons, 522 U.S. 156, 118 S.Ct. 523, 139 L.Ed.2d 525 (1997).
Wisconsin says that these were not necessarily advantages. To an extent, it is correct: Phoenix took advantage of the forum Wisconsin chose by asserting counterclaims — a development that Wisconsin did not welcome. But in all other respects the state’s argument is thin. Wisconsin says the chance to introduce new evidence cannot be considered an advantage because both parties were allowed to supplement the record. Similarly, it points out that the district court’s less-deferential standard of review applied only to evidence presented for the first time in the district court, not to evidence that had been before the agency. Structurally, however, the features that come with challenging the TTAB’s decision in a district court provide an advantage to the party that has lost in the agency. By choosing to file a new action, Wisconsin gave itself better odds of reversing the agency’s cancellation of its mark. There might also have been pragmatic reasons behind the state’s decision to pursue relief in the federal district court. One could understand if the state found Madison, Wisconsin, where the university is located, to be a more convenient or cost-effective place to litigate than the Federal Circuit in Washington, D.C. (though we note that the Federal Circuit from time to time visits other parts of the country to hear argument, see 28 U.S.C. § 48(a) & (d)). In addition, the state’s lawyers may have been more familiar with litigation practices in that particular district court. Or perhaps the state thought that district judges sitting there would be more sympathetic to its claim than a panel of the Federal Circuit. Wisconsin, of course, does not suggest that any of these is the reason that it turned to the district court. Even if it had provided a pragmatic justification for choosing a federal court close to home, the decision it made went considerably beyond a choice of venue. The significant differences between an appeal to the Federal Circuit and a new action in district court convince us that the state’s choice of courts was driven primarily by its desire to increase its chance of success. Choosing one court over another to increase the chance of victory and then denying the chosen court’s competence to resolve related claims is exactly the. sort of gamesmanship that the-Lapides Court hoped to discourage. Wisconsin’s choice to contest the decision of the TTAB in the district court is thus litigation conduct that is inconsistent with an assertion of sovereign immunity.
C
That brings us to the question whether Wisconsin’s waiver of immunity extends to Phoenix’s federal counterclaims. Wisconsin, unsurprisingly, says no. . It points to In re Friendship Medical Center, Ltd., where we said:
“[T]he waiver of immunity is limited to matters ...- arising out of the same transaction or occurrence which is the subject matter of-the,suit, to the extent of defeating the plaintiffs claim. Waiver does not extend to what federal procedure terms ‘permissive’ counterclaims, ... claims for affirmative relief in excess of or different in' kind from that sought by the plaintiff.”
710 F.2d 1297, 1301 (7th Cir.1983) (quoting Federal Savings & Loan Insurance Corp. v. Quinn, 419 F.2d 1014, 1017 (7th Cir.1969)) (emphasis in original). Wisconsin reads this passage to. mean that it has waived immunity only so far as compulsory counterclaims in recoupment are concerned. It says that Phoenix’s claims are barred by its immunity because they are neither compulsory (because, the state asserts, they arise from a different transac*468tion or occurrence than the state’s claim) nor are they claims for recoupment (because they ask for relief beyond that sought by the state).
One can imagine at least three possibilities when it comes to defining the scope of a state’s waiver by litigation conduct. First, the waiver might expose the state to any additional claims that form part of the same constitutional case or controversy. According to that understanding, Wisconsin’s waiver would permit Phoenix to assert any claims arising from a “common nucleus of operative fact.” United Mine Workers v. Gibbs, 383 U.S. 715, 725, 86 S.Ct. 1130, 16 L.Ed.2d 218 (1966). Only “a loose factual connection” between the claims would be required. Baer v. First Options of Chicago, Inc., 72 F.3d 1294, 1299 (7th Cir.1995) (internal quotation marks and citation omitted). Second, the waiver might permit only counterclaims that are compulsory within the meaning of Federal Rule of Civil Procedure 13(a), which would expose the state to those claims that “arise[ ] out of the same transaction or occurrence” as its own claim. Third, the waiver may be limited to compulsory counterclaims for recoupment— the subset of compulsory counterclaims that seek relief of the same kind or nature, and, if monetary, ask for a recovery that does not exceed the amount the state has asked for. See F.D.I.C. v. Hulsey, 22 F.3d 1472, 1487 (10th Cir.1994). Compare Fed. R.Crv.P. 13(d) (counterclaims against the United States).
The statement in Friendship Medical on which Wisconsin relies comes from Quinn, which involved the amenability of the United States to counterclaims filed by a private party it had sued on a note. As the Quinn court noted, and as Rule 13(d) confirms, the United States is subject only to counterclaims that both arise out of the same transaction or occurrence and are limited to recoupment. See generally 6 Charles Alan Wright, et al., Federal Practice and Procedure § 1427, at 232 (3d ed.2010). Friendship Medical imported to the bankruptcy setting Quinn’s rule limiting waivers to recoupment counterclaims. The Friendship Medical court had to decide whether a state agency that filed a proof of claim in a bankruptcy proceeding had, by so doing, opened itself up to any and all claims that the debtor might have against it. Given the fact that bankruptcy opens up literally every facet of the debt- or’s financial situation, we had no trouble in saying no. We held that a state’s decision to file a claim waived its sovereign immunity only for matters arising out of the same transaction, to the extent that they might defeat the state’s claim. 710 F.2d at 1301. Otherwise, debtors could use the bankruptcy proceeding as a lever to hale the state into court on virtually any commercial dealing they had ever had with it.
We have never had occasion to decide, however, how the compulsory-counterclaim rule found in Rule 13(a) operates in ordinary litigation, when the state voluntarily enters court as the plaintiff. (We recognize that some commentators have assumed that we have adopted a recoupment rule, but they were extrapolating from decisions that did not squarely raise this question. See Richard A. Fallon, et al., Hart and Wechsler’s the Federal Courts and the Federal System 883 (6th ed.2009) (citing 71 U.S.L.W. 2592 (Mar. 18, 2003)).) Guidance from the Supreme Court and the decisions of our sister circuits persuade us that there is no implicit exception to Rule 13(a) that operates as a counterpart to Rule 13(d)’s express provision for the United States. In Gardner, the Supreme Court held that “[wjhen the state becomes the actor and files a claim against the fund it waives any immunity which it otherwise might have had respecting the adjudication *469of the claim.” 329 U.S. at 574, 67 S.Ct. 467. Later, it reaffirmed in College Savings that Gardner “stands for the unremarkable proposition that a State waives its sovereign immunity by voluntarily invoking the jurisdiction of the federal courts,” 527 U.S. at 681 n. 3, 119 S.Ct. 2219. Although Gardner on its facts was a bankruptcy case that did not involve any effort to obtain a judgment against the state, 329 U.S. at 574, 67 S.Ct. 467, the Court did not assign any significance to this fact. See Arecibo Community Health v. Commonwealth of Puerto Rico, 270 F.3d 17, 27 (1st Cir.2001) (“Although the scope of waiver found constitutional in Gardner and affirmed in College Savings was limited to recoupment ... nothing in either decision explicitly precludes a broader rule of waiver.”); Teresa K. Goebel, Comment, Obtaining Jurisdiction over States in Bankruptcy Proceedings after Seminole Tribe, 65 U. Chi. L. Rev. 911, 925 (1998) (“The Gardner Court held that the defensive counterclaim rule was constitutional, but did not foreclose the possibility that a broader test may be constitutional.”). A narrow reading of Gardner would be especially difficult to reconcile with the Lapides Court’s concern about states picking and choosing what aspects of a case the federal court might be competent to consider.
None of our sister circuits has found such a limitation in either Rule 13(a) or any decision of the Supreme Court. They have held instead that a waiver of sovereign immunity encompasses all compulsory counterclaims. See Arecibo, 270 F.3d at 28 (“Where a state avails itself of the federal courts to protect a claim, we think it reasonable to consider that action to waive the state’s immunity with respect to that claim in toto and, therefore, to construe that waiver to encompass compulsory counterclaims, even though they could require affirmative recovery from the state.”); In re Straight, 143 F.3d 1387, 1392 (10th Cir.1998) (finding that a state’s proof of claim in bankruptcy proceedings waives Eleventh Amendment immunity from compulsory counterclaims); In re Creative Goldsmiths, 119 F.3d 1140, 1148 (4th Cir.1997) (suggesting that as long as the claims of the party opposing the plaintiff state “amount to a compulsory counterclaim, a state has waived any Eleventh Amendment immunity against that counterclaim in order to avail itself of the federal forum”); In re 995 Fifth Ave. Associates, L.P., 963 F.2d 503, 509 (2d Cir.1992) (holding that a state waives Eleventh Amendment immunity to claims that arise from the same transaction or occurrence).
Since Lapides, this trend among the circuits has continued. The Ninth Circuit initially declined to choose between the recoupment approach and the broader understanding of waiver relied upon by other circuits, In re Lazar, 237 F.3d 967, 978 (9th Cir.2001), but more recently and after Lapides that court has decided to “follow[ ] a number of sister circuits to hold that by filing a proof of a claim, the state waives its Eleventh Amendment immunity with regard to ... claims that arise from the same transaction or occurrence as the state’s claim,” In re Pegasus Gold Corp., 394 F.3d 1189, 1195 (9th Cir.2005). The Second Circuit’s opinion in In re Charter Oak Associates, 361 F.3d 760 (2d Cir.2004), notes that cases on state waivers of immunity during bankruptcy proceedings are simply an application of the Lapides litigation-conduct rule, id. at 767, and observes that “[m]ost circuits agree ... that when a state files a proof of claim, it waives its immunity as to at least some counterclaims, specifically compulsory counterclaims,” id. at 768. The Second Circuit in that case concluded that even permissive counterclaims capped by a set-off limitation fall within a state’s waiver of immuni*470ty. Id. at 768-69. Finally, the Federal Circuit has decided that “the compulsory counterclaim criterion is superior to the recoupment criterion for determining waiver of immunity with respect to counterclaims,” and held that “when a state files suit in federal court to enforce its claims to certain patents, the state shall be considered to have consented to have litigated in the same forum all compulsory counterclaims.... ” Knight, 321 F.3d at 1126.
Against this weight of authority, Wisconsin submits that its understanding of Friendship Medical would be the preferable rule. It reminds us that the Supreme Court has said that waivers of immunity are to be narrowly construed in favor of the sovereign. E.g., Lane v. Pena, 518 U.S. 187, 192, 116 S.Ct. 2092, 135 L.Ed.2d 486 (1996). But that canon of construction is a tool that federal courts use to interpret a sovereign’s explicit waiver of immunity in a statute (and it is not always relied upon in that context, see Dolan v. U.S. Postal Service, 546 U.S. 481, 491-92, 126 S.Ct. 1252, 163 L.Ed.2d 1079 (2006)); the canon “has little relevance in this context, where the waiver is effected not through the language a state has used, but through the actions it has taken during the course of litigation,” Charter Oak Associates, 361 F.3d at 770. As we have said, Wisconsin enjoyed a number of advantages when it selected the district court to challenge an adverse agency decision; Lapides expresses the Court’s concern with unfair litigation tactics; and it would be manifestly unfair if Wisconsin were allowed to enjoy the advantages of the district court while using sovereign immunity to avoid the disadvantages. We conclude that our sister circuits have articulated the rule that is most consistent with Lapides: when a state waives its sovereign immunity by litigation conduct, that waiver opens the door to counterclaims regarded as compulsory within the meaning of Federal Rule of Civil Procedure 13(a). Whether such a waiver extends more broadly — perhaps to the entire constitutional case or controversy — is a question for another day.
Rule 13(a) defines a compulsory counterclaim as one that “arises out of the transaction or occurrence that is the subject matter of the opposing party’s claim.” Fed.R.CivP. 13(a). We use a “logical relationship” test to decide whether two matters are the same for purposes of Rule 13(a). The approach is necessarily flexible: as we have noted before, “[a] court should consider the totality of the claims, including the nature of the claims, the legal basis for recovery, the law involved, and the respective factual backgrounds.” Burlington Northern R.R. Co. v. Strong, 907 F.2d 707, 711 (1990). Our approach thus focuses on the facts of the case, rather than on the technical elements of the claims in question.
Wisconsin maintains that Phoenix’s claims for infringement and false designation of origin under the Lanham Act fail this test because the facts that Phoenix must prove to succeed are different from those that the state will present in its effort to reverse the TTAB’s decision to cancel its registration. It insists that the state’s claim is about CONDOR-Wisconsin, while Phoenix’s counterclaims are about CONDOR-Phoenix. The first flaw in Wisconsin’s position is the fact that this court (as well as others) does not use a “same evidence” test for purposes of Rule 13(a). See 6 Wright, et al., supra, § 1410, at 60-61 (giving a number of examples in which counterclaims are compulsory even though the evidence differs). Applying the proper test, we have no trouble concluding that there is a logical relationship between Wisconsin’s claim and Phoenix’s counter*471claims. All of the claims focus on a single factual dispute: the conflict between the two CONDOR marks. To make a case of infringement (or false designation of origin), Phoenix will be required to show that it had a protectable mark and that the state’s use of that mark is likely to cause confusion in trade. CAE, 267 F.3d at 673-74. On remand, Wisconsin’s effort to overturn the TTAB’s decision will depend almost entirely on how the jury resolves the question whether Wisconsin’s registration of the CONDOR mark is likely to cause confusion among consumers.
This does not mean that infringement and false designation counterclaims are compulsory as a matter of law. One could imagine a case where the registration dispute turned on an allegation that a party committed fraud, while the counterclaims focused on the likelihood of confusion in trade. It is entirely possible that those counterclaims would not be compulsory. See, e.g., Nasalok Coating Corp. v. Nylok Corp., 522 F.3d 1320, 1326 (Fed.Cir.2008). But in this case, Wisconsin’s challenge to the TTAB decision is part of the same transaction or occurrence that gives rise to Phoenix’s counterclaims. See J. Thomas McCarthy, McCarthy on Trademarks and Unfair Competition § 21:20 & n. 11 (4th ed.2011) (one advantage of challenging a TTAB decision in the district court is that “the case can be expanded to include a prayer for injunctive relief for trademark infringement,” but noting that in such a case the “[djefendant may have a claim for infringement which is a compulsory counterclaim under F.R.C.P. 13(a)”). Accordingly, Wisconsin’s waiver of sovereign immunity is broad enough to encompass the federal counterclaims that Phoenix has asserted here.
D
In summary, we conclude that Wisconsin, by choosing to bring its challenge to the TTAB decision in the district court, waived its sovereign immunity against compulsory counterclaims that meet the requirements of Rule 13(a). On remand, Phoenix’s federal counterclaims are reinstated for further appropriate proceedings.
IV
While this is arguably enough to resolve the appeal before us, it is significant that broader policies underlying the doctrine of sovereign immunity also support our result. To the extent that Wisconsin has argued that our result is in tension with the Supreme Court’s sovereign immunity cases, we explain briefly why we believe that is not so.
A
From the time of the Declaration of Independence until the Constitution of 1787 took effect, the states were fully sovereign in the international sense of the term: they were States, just as modern-day France, Japan, or India are States today. This fact is reflected in Articles II and III of the Articles of Confederation. See Articles of Confederation of 1781, art. II (“Each state retains its sovereignty, freedom, and independence, and every power, jurisdiction and right, which is not by this Confederation expressly delegated to the United States, in Congress assembled.”); id. art. Ill (“The said states hereby severally enter into a firm league of friendship with each other.... ”). It was against this backdrop that the 1787 Constitution was written. Although that Constitution greatly strengthened the powers of the central government, it did not change the fundamental principle under which the states remain sovereign entities to the extent that the Constitution does not provide otherwise. Indeed, that is precisely the point of the Tenth Amendment, which re*472serves to the states or the people the powers not delegated in the Constitution to the United States nor prohibited to the States. U.S. Const, amend. X.
Over the years, the Supreme Court has consistently recognized the sovereign immunity of the states. Even in Chisholm v. Georgia, 2 U.S. (2 Dall.) 419, 1 L.Ed. 440 (1793), better known for its holding that the state of Georgia could be sued than for the majority’s reasoning, all members of the Court understood that they had to decide whether the state’s sovereignty had been overridden by the federal Constitution. Justice Blair thought that Article III of the Constitution provided a clear answer in the affirmative to that question; he saw no need to rely on external sources. Id. at 450. Justice Wilson chose to look more broadly at principles of general jurisprudence, the laws and practices of other nations, as well as the constitutional text. He too concluded that states could constitutionally be brought to account before tribunals. Id. at 453-66.
Justice Iredell, whose view was quickly vindicated by the passage of the Eleventh Amendment, would have found immunity for Georgia. Id. at 449-50. He emphasized the lack of federal legislation permitting the lawsuit before the Court, adding that he could not imagine “any construction of [the Constitution], which will admit, under any circumstances, a compulsive suit against a State for the recovery of money.” Id. at 449. Justice Iredell took this absolute position in reliance on the English traditions about suits against the crown. Id. at 437-38. He drew a direct link between the states’ immunity from suit before the adoption of the Constitution— immunity derived from principles embodied in public international law — and the states’ continuing immunity afterwards. Id.
Throughout the nineteenth century, the Court commonly turned to the same principles of public international law that it was using for foreign relations when it had to decide issues involving interstate relations. Thus, for example, in Bank of the United States v. Donnally, 33 U.S. 361, 372, 8 Pet. 361, 8 L.Ed. 974 (1834), Justice Story wrote that “whatever may be the legislation of a state, as to the obligation or remedy on contracts, its acts can have no binding authority beyond its own territorial jurisdiction. Whatever authority they have in other states, depends upon principles of international comity, and a sense of justice.” More than forty years later, Justice Field used almost the same words in Pennoyer v. Neff, 95 U.S. 714, 24 L.Ed. 565 (1877), the decision that for years was the wellspring of the constitutional law of personal jurisdiction. He wrote that “[t]he several States of the Union are not, it is true, in every respect independent, many of the right and powers which originally belonged to them being now vested in the government created by the Constitution. But, except as restrained and limited by that instrument, they possess and exercise the authority of independent States, and the principles of public law to which we have referred are applicable to them.” Id. at 722.
The Supreme Court has returned repeatedly to this theme in the line of cases that began with Seminole Tribe of Florida v. Florida, 517 U.S. 44, 116 S.Ct. 1114, 134 L.Ed.2d 252 (1996), in which the Court reinvigorated the doctrine of state sovereign immunity:
Although the text of the [Eleventh] Amendment would appear to restrict only the Article III diversity jurisdiction of the federal courts, we have understood the Eleventh Amendment to stand not so much for what it says, but for the presupposition ... which it confirms .... That presupposition, first ob*473served over a century ago in Hans v. Louisiana, 134 U.S. 1 [10 S.Ct. 504, 33 L.Ed. 842] (1890), has two parts: first, that each State is a sovereign entity in our federal system; and second, that “ ‘[i]t is inherent in the nature of sovereignty not to be amenable to the suit of an individual without its consent,’ ” id. at 13 [10 S.Ct. 504] (emphasis deleted), quoting The Federalist No. 81, p. 487 (C. Rossiter ed.1961) (A. Hamilton).
Id. at 54, 116 S.Ct. 1114 (some internal quotation marks and citations omitted). The theme continues in Florida Prepaid, 527 U.S. at 634-35, 119 S.Ct. 2199, College Savings, 527 U.S. at 669-70, 119 S.Ct. 2219 (referring to “the sovereign immunity that the States possessed before entering the Union”), Alden v. Maine, 527 U.S. 706, 713, 119 S.Ct. 2240, 144 L.Ed.2d 636 (1999) (“[T]he States’ immunity from suit is a fundamental aspect of the sovereignty which the States enjoyed before the ratification of the Constitution, and which they retain today (either literally or by virtue of their admission into the Union upon an equal footing with the other States) except as altered by the plan of the Convention or certain constitutional Amendments.”), Kimel v. Florida Board of Regents, 528 U.S. 62, 72-73, 120 S.Ct. 631, 145 L.Ed.2d 522 (2000), Board of Trustees of the University of Alabama v. Garrett, 531 U.S. 356, 363—64, 121 S.Ct. 955, 148 L.Ed.2d 866 (2001), and Federal Maritime Commission, 535 U.S. at 760, 122 S.Ct. 1864 (“The preeminent purpose of state sovereign immunity is to accord States the dignity that is consistent with their status as sovereign entities.”).
B
The fact that the states enjoy “sovereign immunity” is just the beginning of the inquiry. The term requires some understanding both of sovereignty and of the scope of the immunity that attends it. Early theorists thought that sovereignty was necessarily a singular phenomenon: either an entity enjoyed sovereignty, or it did not — there was no such thing as partial sovereignty. See, e.g., J.L. Brierly, The Law of Nations: An Introduction to the International Law of Peace 8 (Sir Humphrey Waldock ed., 6th ed.1963). The Framers of the Constitution, however, had a more nuanced view of the concept of sovereignty. As Justice Kennedy put it in his concurring opinion in US. Term Limits, Inc. v. Thornton, the Framers “split the atom of sovereignty” between the states and the national government. 514 U.S. 779, 838, 115 S.Ct. 1842, 131 L.Ed.2d 881 (1995) (Kennedy, J., concurring). Thus, with reference to state or Native American sovereignty, it is always necessary to consider whether sovereign powers have already been ceded, see, e.g., Central Virginia Community College v. Katz, 546 U.S. 356, 377-78, 126 S.Ct. 990, 163 L.Ed.2d 945 (2006) (concluding that the states agreed in the plan of the constitutional Convention not to assert sovereign immunity in bankruptcy proceedings), or abrogated by valid congressional legislation, see, e.g., City of Boerne v. Flores, 521 U.S. 507, 529-30, 117 S.Ct. 2157, 138 L.Ed.2d 624 (1997). Another established attribute of state sovereign immunity under the federal Constitution is the privilege of the state to choose whether to assert or to waive the defense. Schacht, 524 U.S. at 389, 118 S.Ct. 2047.
Some questions about state sovereign immunity, however, have received less attention. One deals with the scope of the defense. Is the doctrine absolute, or is it qualified or restrictive? And if it has changed over time, should courts look to the version in effect in 1787 (or perhaps the date of each state’s entry into the Union) or to the doctrine as it is generally understood today?
*474While unanswered in the context of state sovereign immunity, these issues have received extensive attention in both Congress and the courts, insofar as foreign states are concerned. From the Founding until 1952, the United States granted foreign nations absolute immunity from suit in U.S. courts as a matter of comity. Verlinden B.V. v. Central Bank of Nigeria, 461 U.S. 480, 486-89, 103 S.Ct. 1962, 76 L.Ed.2d 81 (1983) (discussing The Schooner Exchange v. McFaddon, 11 U.S. (7 Cranch) 116, 3 L.Ed. 287 (1812)). The Supreme Court deferred to the decisions of the political branches on whether immunity existed, and for many years the State Department “ordinarily requested immunity in all actions against friendly foreign sovereigns.” Id. at 486, 103 S.Ct. 1962. Reacting in the wake of World War II to the important role of state-owned enterprises in the Communist countries and reluctant to give a litigation advantage in U.S. courts to those enterprises, the United States shifted its position. Commentators regularly outlined the tension between absolute sovereign immunity and the rise of state trading, particularly in the Soviet Union, and expressed the concern that “private traders will be reluctant to deal with state traders if their legal rights and remedies are greatly curtailed by the principle of sovereign immunity.” Bernard Fensterwald, Jr., Sovereign Immunity and Soviet State Trading, 63 Harv. L.Rev. 614, 614 (1950); see also Sigmund Tim-berg, Sovereign Immunity, State Trading, Socialism and Self-Deception, 56 Nw. U.L.Rev. 109, 111 (1961).
In 1952, Jack Tate, Acting Legal Adviser for the Department of State, announced that the State Department intended henceforth to apply a restrictive theory of foreign sovereign immunity. According to that theory, he explained, “the immunity of the sovereign is recognized with regard to sovereign or public acts (jure imperii) of a state, but not with respect to private acts (jure gestionis).” Letter from Jack B. Tate, Acting Legal Adviser, U.S. Department of State, to Acting U.S. Attorney General Philip B. Perlman (May 19, 1952), reprinted in 26 Dep’t State Bull. 984-85 (1952). Tate noted that “the widespread and increasing practice on the part of governments of engaging in commercial activities [made] necessary a practice which [would] enable persons doing business with them to have their rights determined in the courts.” Id. “The reasons which obviously motivate state trading countries in adhering to the theory [of absolute immunity] with perhaps increasing rigidity,” wrote Tate, “are most persuasive that the United States should change its policy.” Id.
Experience under the Tate Letter revealed problems with the system of relying on the Executive for suggestions of immunity. See Verlinden B.V., 461 U.S. at 487-88, 103 S.Ct. 1962. Thus, in 1976, Congress passed the Foreign Sovereign Immunities Act (FSIA), 28 U.S.C. §§ 1330, 1332, 1391(f), 1441(d), and 1602-1611, “in order to free the Government from the case-by-case diplomatic pressures, to clarify the governing standards, and to ‘assur[e] litigants that ... decisions are made on purely legal grounds and under procedures that insure due process,’ H.R.Rep. No. 94^1487, p. 7 [1976 U.S.C.C.A.N. 6604, 6606] (1976).” Verlinden B.V., 461 U.S. at 488, 103 S.Ct. 1962. Codifying the restrictive theory of sovereign immunity, the FSIA provides that “a foreign state shall be immune from the jurisdiction of the courts of the United States and of the States,” 28 U.S.C. § 1604, except if one of several enumerated exceptions in the statute applies, see id. §§ 1605(a), 1605A (exceptions for waiver, commercial activity, expropriations, certain other property rights, noncommercial torts, maritime *475liens, and acts of terrorism). In addition, in actions brought by foreign States (or in cases where a foreign State intervenes) the foreign State is not accorded immunity with respect to any counterclaim:
(a) for which a foreign state would not be entitled to immunity under section 1605 or 1605A ... had such claim been brought in a separate action against the foreign state; or
(b) arising out of the transaction or ocv currence that is the subject matter of the claim of the foreign state; or
(c) to the extent that the counterclaim does not seek relief exceeding in amount or differing in kind from that sought by the foreign state.
28 U.S.C. § 1607 (emphasis added).
“The most significant of the FSIA’s exceptions ... is the ‘commercial’ exception of § 1605(a)(2)____” Republic of Argentina v. Weltover, Inc., 504 U.S. 607, 611, 112 S.Ct. 2160, 119 L.Ed.2d 394 (1992). When evaluating whether a foreign State has engaged in commercial activity that falls within the statutory exception, a court must evaluate the nature of the act, rather than the purpose for which that act was undertaken. See 28 U.S.C. § 1603(d); Republic of Argentina, 504 U.S. at 614, 112 S.Ct. 2160 (“[T]he issue is whether the particular actions that the foreign state performs (whatever the motive behind them) are the type of actions by which a private party engages in ‘trade or traffic or commerce,’ Black’s Law Dictionary 270 (6th ed.1990).”). “[W]hen a foreign government acts, not as regulator of a market, but in the manner of a private player within it, the foreign sovereign’s actions are ‘commercial’ within the meaning of the FSIA.” Republic of Argentina, 504 U.S. at 614, 112 S.Ct. 2160. Where one of the FSIA’s statutory exceptions, like the one for commercial acts, applies, “the foreign state shall be liable in the same manner and to the same extent as a private individual under like circumstances” (with modifications for punitive damages and wrongful death actions). 28 U.S.C. § 1606.
The Supreme Court has recognized that “the distinction between state sovereign acts, on the one hand, and state commercial and private acts, on the other, [is] not entirely novel to American law.” Republic of Argentina, 504 U.S. at 613, 112 S.Ct. 2160 (discussing Alfred Dunhill of London, Inc. v. Republic of Cuba, 425 U.S. 682, 695-96, 96 S.Ct. 1854, 48 L.Ed.2d 301 (1976)). Indeed, as early as 1823 Chief Justice Marshall noted the importance of the distinction:
It is, we think, a sound principle, that when a government becomes a partner in any trading company, it divests itself, so far as concerns the transactions of that company, of its sovereign character, and takes that of a private citizen. Instead of communicating to the company its privileges and its prerogatives, it descends to a level with those with whom it associates itself, and takes the character which belongs to its associates, and to the business which is to be transacted.
Bank of the United States v. Planters’ Bank of Georgia, 22 U.S. (9 Wheat.) 904, 907, 6 L.Ed. 244 (1823); cf. Sloan Shipyards Corp. v. United States Shipping Bd. Emergency Fleet Corp., 258 U.S. 549, 567-68, 42 S.Ct. 386, 66 L.Ed. 762 (1922). Relying on precedents dealing with state action within the United States, the Dunhill plurality drew a line between commercial and governmental acts, commenting that “[w]hen a state enters the market place seeking customers it divests itself of its Quasi sovereignty Pro tanto, and takes on the character of a trader....” 425 U.S. at 696, 96 S.Ct. 1854.
*476C
Although the Supreme Court has addressed the question of state sovereign immunity on a number of occasions in recent years, it has not given sustained attention to the question whether the default rule ought to be one of absolute or restrictive immunity. It came closest to this issue when it rejected the argument that the market-participant doctrine found in dormant Commerce Clause cases was pertinent to the states’ sovereign immunity. College Savings, 527 U.S. at 685-86, 119 S.Ct. 2219 (citations omitted). As the Court pointed out, the market-participant doctrine was developed to address the risk that the states might construct barriers to trade between themselves, in violation of the principles behind the Commerce Clause. In contrast, it suggested, “a suit by an individual against an unconsenting State is the very evil at which the Eleventh Amendment is directed — and it exists whether or not the State is acting for profit, in a traditionally ‘private’ enterprise, and as a ‘market participant.’ ” Id. at 685, 119 S.Ct. 2219. Aside from a couple of fleeting references in dissenting opinions and an even briefer response by the majority, it does not appear that the rationale underlying the recognition of a commercial-acts exception for purposes of foreign sovereign immunity was considered by the Court.
What attention there was primarily came from Justices Stevens and Breyer. In his dissenting opinion in that case, Justice Stevens remarked:
The procedural posture of this case requires the Court to assume that Florida Prepaid is an “arm of the State” of Florida.... But the validity of that assumption is doubtful if the Court’s jurisprudence in this area is to be based primarily on present-day assumptions about the status of the doctrine of sovereign immunity in the 18th century. Sovereigns did not then play the kind of role in the commercial marketplace that they do today. In future cases, it may therefore be appropriate to limit the coverage of state sovereign immunity by treating the commercial enterprises of the States like the commercial activities of foreign sovereigns under the Foreign Sovereign Immunities Act of 1976.
527 U.S. at 691-92 & n. 1, 119 S.Ct. 2219 (Stevens, J., dissenting) (citing the FSIA’s commercial activity exception, 28 U.S.C. § 1605(a)(2), and its definition of “commercial activity,” id. § 1603(d)). Justice Breyer’s dissent made a comparable point. 527 U.S. at 699, 119 S.Ct. 2219 (Breyer, J., dissenting).
The majority found those allusions to foreign sovereign immunity unhelpful; it dismissed the “suggestion ... that we limit state sovereign immunity to noncommercial state activities because Congress has so limited foreign sovereign immunity” on the assumption that state sovereign immunity is rooted in the Constitution, while foreign sovereign immunity is not. 527 U.S. at 686 n. 4, 119 S.Ct. 2219 (opinion of Scalia, J.). “The text of the Eleventh Amendment,” the majority wrote, “makes no distinction between commercial and noncommercial state activities....” Id. That text, of course, also does not delineate the doctrine of state sovereign immunity that the Supreme Court has recognized. In fact, neither state nor foreign sovereign immunity as understood today is mentioned directly by the Constitution. Both doctrines flow from broader constitutional principles: state sovereign immunity is considered to be inherent in the constitutional plan, while foreign sovereign immunity flows from the interaction between the United States and its fellow sovereign nations, reflected among other places in Article II of the Constitution.
*477We neither have nor need a crystal ball to tell us how these immunity doctrines will develop in the-future. Our only point is that it is worth recalling that at least through the end of the nineteenth century the Court regularly applied principles of public international law to the U.S. states. Those principles permeate this country’s foundational documents, and international analogies abound in the Federalist Papers. Nothing in the Court’s line of state sovereign immunity cases implies that any less dignity is due today to the states. If parity is to be the rule, then the day may return when it exists across-the-board.
Many of the competitive concerns that motivated the move from absolute to restrictive immunity in the foreign setting apply to the states with equal force. There is no apparent reason, for example, why the University of Wisconsin should be immune from lawsuits that Marquette University, a Catholic Jesuit institution located in Milwaukee, would have to defend. Nor is there any apparent reason why a state-owned hospital, or garbage pick-up service, or power plant, should have a competitive edge over a private competitor. And, given the fact that litigation imposes transaction costs — often very high costs— on the parties, the failure to recognize a commercial-act exception for state entities may confer the same kind of competitive advantage on the states that the United States was reluctant to confer on socialist or Communist countries. See, e.g., Joan E. Donoghue, Taking the “Sovereign-’ Out' of the Foreign Sovereign Immunities Act: A Functional Approach to the Commercial Activity Exception, 17 Yale J. Int’l L. 489, 490 (1992); see also 1 Restatement (Third) of the Foreign Relations Law of the United States 390 (1987) (Introductory Note to Part IV, Ch. 5).
Capitalism and private ownership have served the United States well. Even though there is no clause in the Constitution explicitly committing this country to-such an economic system (although the Takings Clause of the Fifth Amendment may come close), the antitrust laws have been called quasi-constitutional, and there seems little doubt that economic freedom is high on the list of cherished rights. See, e.g., United States v. Topco Associates, Inc., 405 U.S. 596, 610, 92 S.Ct. 1126, 31 L.Ed.2d 515 (1972) (“Antitrust laws ... are the Magna Carta of free enterprise.”); Northern Pac. R.R. Co. v. United States, 356 U.S. 1, 4, 78 S.Ct. 514, 2 L.Ed.2d 545 (1958) (“The Sherman Act was designed to be a comprehensive charter of economic liberty!.]”). At a minimum, this public policy suggests caution in extending sovereign immunity to new areas of commercial activity.
V
Far from doing violence to the policies behind the immunity doctrines, the outcome in this case is fully consistent with them. A straightforward application of Lapides and the doctrine of waiver by litigation conduct require us to reinstate Phoenix’s counterclaims on remand. Moreover, there are genuine issues of fact on the question whether there is a likelihood of confusing Phoenix’s CONDOR mark with Wisconsin’s, in light of the TTAB’s factual findings. The decision of the district court is Reversed and the case is Remanded for further proceedings consistent with this opinion.