delivered the opinion of the Court.

This case requires that we decide a peculiar jurisdictional battle between the Court of Appeals for the Federal Circuit and the Court of Appeals for the Seventh Circuit. Each court has adamantly disavowed jurisdiction over this case. Each has transferred the case to the other. And each insists that the other’s jurisdictional decision is “clearly wrong.” *804798 F. 2d 1051, 1056-1057 (CA7 1986); 822 F. 2d 1544, 1551, n. 7 (CA Fed. 1987). The parties therefore have been forced to shuttle their appeal back and forth between Chicago and the District of Columbia in search of a hospitable forum, ultimately to have the merits decided, after two years, by a Court of Appeals that still insists it lacks jurisdiction to do so.

HH

Respondent Colt Industries Operating Corp. is the leading manufacturer, seller, and marketer of M16 rifles and their parts and accessories. Colt’s dominant market position dates back to 1959, when it acquired a license for 16 patents to manufacture the M16’s precursor. Colt continued to develop the rifle, which the United States Army adopted as its standard assault rifle, and patented additional improvements. Through various devices, Colt has also maintained a shroud of secrecy around certain specifications essential to the mass production of interchangeable M16 parts. For example, Colt’s patents conceal many of the manufacturing specifications that might otherwise be revealed by its engineering drawings, and when Colt licenses others to manufacture M16 parts or hires employees with access to proprietary information, it contractually obligates them not to disclose specifications.

Petitioner Christianson is a former Colt employee who acceded to such a nondisclosure agreement. Upon leaving respondent’s employ in 1975, Christianson established petitioner International Trade Services, Inc. (ITS), and began selling M16 parts to various customers domestically and abroad. Petitioners’ business depended on information that Colt considers proprietary. Colt expressly waived its proprietary rights at least as to some of petitioners’ early transactions. The precise scope of Colt’s waiver is a matter of considerable dispute. In 1983, however, Colt joined petitioners as defendants in a patent-infringement lawsuit against two companies that had arranged a sale of M16’s to El Salvador. *805Evidence suggested that petitioners supplied the companies with certain M16 specifications, and Colt sought a court order enjoining petitioners from any further disclosures. When the District Court declined the motion, Colt voluntarily dismissed its claims against petitioners. In the meantime, Colt notified several of petitioners’ current and potential customers that petitioners were illegally misappropriating Colt’s trade secrets, and urged them to refrain from doing business with petitioners.

Three days after their dismissal from the lawsuit, petitioners brought this lawsuit in the District Court against Colt “pursuant to Section 4 ... (15 U. S. C. § 15) and Section 16 of the Clayton Act (15 U. S. C. § 26) for damages, injunctive and equitable relief by reason of its violations of Sections 1 and 2 of the Sherman Act (15 U. S. C. §§ 1 & 2). . . .” App. 7. The complaint alleged that Colt’s letters, litigation tactics, and “[o]the[r]. . . conduct” drove petitioners out of business. In that context, petitioners included the following obscure passage:

“18. The validity of the Colt patents had been assumed throughout the life of the Colt patents through 1980. Unless such patents were invalid through the wrongful retention of proprietary information in contravention of United States Patent Law (35 U. S. C. § 112), in 1980, when such patents expired, anyone ‘who has ordinary skill in the rifle-making art’ is able to use the technology of such expired patents for which Colt earlier had a monopoly position for 17 years.

“19. ITS and anyone else has the right to manufacture, contract for the manufacture, supply, market and sell the M-16 and M-16 parts and accessories thereof at the present time.” Id., at 9.

Petitioners later amended their complaint to assert a second cause of action under state law for tortious interference with their business relationships. Colt interposed a defense that *806its conduct was justified by a need to protect its trade secrets and countersued on a variety of claims arising out of petitioners’ alleged misappropriation of M16 specifications.

Petitioners’ motion for summary judgment raised only a patent-law issue obliquely hinted at in the above-quoted paragraphs — that Colt’s patents were invalid from their inception for failure to disclose sufficient information to “enable any person skilled in the art. . . to make and use the same” as well as a description of “the best mode contemplated by the inventor of carrying out his invention.” 35 U. S. C. § 112. Since Colt benefited from the protection of the invalid patents, the argument continues, the “trade secrets” that the patents should have disclosed lost any state-law protection. Petitioners therefore argued that the District Court should hold that “Colt’s trade secrets are invalid and that [their] claim of invalidity shall be taken as established with respect to all claims and counterclaims to which said issue is material.” App. 58.

The District Court awarded petitioners summary judgment as to liability on both the antitrust and the tortious-interference claims, essentially relying on the § 112 theory articulated above. In the process, the District Court invalidated nine of Colt’s patents, declared all trade secrets relating to the M16 unenforceable, enjoined Colt from enforcing “any form of trade secret right in any technical information relating to the M16,” and ordered Colt to disgorge to petitioners all such information. 613 P. Supp. 330, 332 (CD Ill. 1985).

Respondent appealed to the Court of Appeals for the Federal Circuit, which, after full briefing and argument, concluded that it lacked jurisdiction and issued an unpublished order transferring the appeal to the Court of Appeals for the Seventh Circuit. See 28 U. S. C. § 1631. The Seventh Circuit, however, raising the jurisdictional issue sna sponte, concluded that the Federal Circuit was “clearly wrong” and transferred the case back. 798 F. 2d, at 1056-1057. 1062. *807The Federal Circuit, for its part, adhered to its prior jurisdictional ruling, concluding that the Seventh Circuit exhibited “a monumental misunderstanding of the patent jurisdiction granted this court,” 822 F. 2d, at 1547, and was “clearly wrong,” id., at 1551, n. 7. Nevertheless, the Federal Circuit proceeded to address the merits in the “interest of justice,” id., at 1559-1560, and reversed the District Court. We granted certiorari, 484 U. S. 985 (1987), and now vacate the judgment of the Federal Circuit.

hH

As relevant here, 28 U. S. C. § 1295(a)(1) grants the Court of Appeals for the Federal Circuit exclusive jurisdiction over “an appeal from a final decision of a district court of the United States ... if the jurisdiction of that court was based, in whole or in part, on [28 U. S. C.] section 1338 . . . ,”1 Section 1338(a), in turn, provides in relevant part that “[t]he district courts shall have original jurisdiction of any civil action arising under any Act of Congress relating to patents . . . .” Thus, the jurisdictional issue before us turns on whether this is a case “arising under” a federal patent statute, for if it is then the jurisdiction of the District Court was based at least “in part” on § 1338.

A

In interpreting § 1338’s precursor, we held long ago that in order to demonstrate that a case is one “arising under” federal patent law “the plaintiff must set up some right, title or interest under the patent laws, or at least make it appear that some right or privilege will be defeated by one construc*808tion, or sustained by the opposite construction of these laws.” Pratt v. Paris Gas Light & Coke Co., 168 U. S. 255, 259 (1897). See Henry v. A. B. Dick Co., 224 U. S. 1, 16 (1912). Our cases interpreting identical language in other jurisdictional provisions, particularly the general federal-question provision, 28 U. S. C. § 1331 (“The district courts shall have original jurisdiction of all civil actions arising under the Constitution, laws, or treaties of the United States”), have quite naturally applied the same test.2 See Gully v. First National Bank in Meridian, 299 U. S. 109, 112 (1936) (the claim alleged in the complaint “must be such that it will be supported if the Constitution or laws of the United States are given one construction or effect, and defeated if they receive another”) (citations omitted). A district court’s federal-question jurisdiction, we recently explained, extends over “only those cases in which a well-pleaded complaint establishes either that federal law creates the cause of action or that the plaintiff’s right to relief necessarily depends on resolution of a substantial question of federal law,” Franchise Tax Board of California v. Construction Laborers Vacation Trust, 463 U. S. 1, 27-28 (1983), in that “federal law is a necessary element of one of the well-pleaded . . . claims,” id., at 13. Linguistic consistency, to which we have historically adhered, demands that § 1338(a) jurisdiction likewise extend *809only to those cases in which a well-pleaded complaint establishes either that federal patent law creates the cause of action or that the plaintiff’s right to relief necessarily depends on resolution of a substantial question of federal patent law, in that patent law is a necessary element of one of the well-pleaded claims. See 822 F. 2d, at 1553-1556; 798 F. 2d, at 1059-1061.

The most superficial perusal of petitioners’ complaint establishes, and no one disputes, that patent law did not in any sense create petitioners’ antitrust or intentional-interference claims. Since no one asserts that federal jurisdiction rests on petitioners’ state-law claims, the dispute centers around whether patent law “is a necessary element of one of the well-pleaded [antitrust] claims.” See Merrell Dow Pharmaceuticals Inc. v. Thompson, 478 U. S. 804, 813 (1986). Our cases, again mostly in the § 1331 context, establish principles for both defining the “well-pleaded . . . claims” and discerning which elements are “necessary” or “essential” to them. Under the well-pleaded complaint • rule, as appropriately adapted to § 1338(a), whether a claim “arises under” patent law “‘must be determined from "what necessarily appears in the plaintiff’s statement of his own claim in the bill or declaration, unaided by anything alleged in anticipation or avoidance of defenses which it is thought the defendant may interpose.’” Franchise Tax Board, supra, at 10 (quoting Taylor v. Anderson, 234 U. S. 74, 75-76 (1914)). See Louisville & Nashville R. Co. v. Mottley, 211 U. S. 149 (1908). Thus, a case raising a federal patent-law defense does not, for that reason alone, “arise under” patent law, “even if the defense is anticipated in the plaintiff’s complaint, and even if both parties admit that the defense is the only question truly at issue in the case.” Franchise Tax Board, supra, at 14.3 See also Merrell Dow, supra, at 808.

*810Nor is it necessarily sufficient that a well-pleaded claim alleges a single theory under which resolution of a patent-law question is essential. If “on the face of a well-pleaded complaint there are . . . reasons completely unrelated to the provisions and purposes of [the patent laws] why the [plaintiff] may or may not be entitled to the relief it seeks,” Franchise Tax Board, 463 U. S., at 26 (footnote omitted), then the claim does not “arise under” those laws. See id., at 26, n. 29. Thus, a claim supported by alternative theories in the complaint may not form the basis for § 1338(a) jurisdiction unless patent law is essential to each of those theories.

B

Framed in these terms, our resolution of the jurisdictional issue in this case is straightforward. Petitioners’ antitrust count can readily be understood to encompass both a monopolization claim under §2 of the Sherman Act and a group-boycott claim under § 1. The patent-law issue, while arguably necessary to at least one theory under each claim, is not necessary to the overall success of either claim.

Section 2 of the Sherman Act condemns “[e]very person who shall monopolize, or attempt to monopolize . . . .” 15 U. S. C. § 2. The thrust of petitioners’ monopolization claim is that Colt has “embarked on a course of conduct to illegally extend its monopoly position with respect to the described patents and to prevent ITS from engaging in any business with respect to parts and accessories of the M-16.” App. 10. The complaint specifies several acts, most of which relate either to Colt’s prosecution of the lawsuit against petitioners or to letters Colt sent to petitioners’ potential and existing customers. To make out a § 2 claim, petitioners would *811have to present a theory under which the identified conduct amounted to a “willful acquisition or maintenance of [monopoly] power as distinguished from growth or development as a consequence of a superior product, business acumen, or historic accident.” United States v. Grinnell Corp., 384 U. S. 563, 570-571 (1966). Both the Seventh Circuit and Colt focus entirely on what they perceive to be “the only basis Christianson asserted in the complaint for the alleged antitrust violation,” 798 F. 2d, at 1061; see Brief for Respondent 32 — namely, that Colt made false assertions in its letters and pleadings that petitioners were violating its trade secrets, when those trade secrets were not protected under state law because Colt’s patents were invalid under § 112. Thus, Colt concludes, the validity of the patents is an essential element of petitioners’ prima facie monopolization theory and the case “arises under” patent law.

We can assume without deciding that the invalidity of Colt’s patents is an essential element of the foregoing monopolization theory rather than merely an argument in anticipation of a defense. But see 822 F. 2d, at 1547. The well-pleaded complaint rule, however, focuses on claims, not theories, see Franchise Tax Board, supra, at 26, and n. 29; Gully, 299 U. S., at 117, and just because an element that is essential to a particular theory might be governed by federal patent law does not mean that the entire monopolization claim “arises under” patent law.

Examination of the complaint reveals that the monopolization theory that Colt singles out (and on which petitioners ultimately prevailed in the District Court) is only one of several, and the only one for which the patent-law issue is even arguably essential. So far as appears from the complaint, for example, petitioners might have attempted to prove that Colt’s accusations of trade-secret infringement were false not because Colt had no trade secrets, but because Colt authorized petitioners to use them. App. 9-10 (“Contrary to the permission extended to ITS to sell Colt parts and accessories *812and in violation of the anti-trust laws . . . Colt has embarked upon a course of conduct... to prevent ITS from engaging in any business with respect to parts and accessories of the M-16”). In fact, most of the conduct alleged in the complaint could be deemed wrongful quite apart from the truth or falsity of Colt’s accusations. According to the complaint, Colt’s letters also (1) contained “copies of inapplicable court orders” and “suggested] that these court orders prohibited [the recipients] from doing business with” petitioners; and (2) “falsely stat[ed] that ‘Colt’s right’ to proprietary data had been ‘consistently upheld in various courts.’” Id., at 10. Similarly, the complaint alleges that Colt’s lawsuit against petitioners (1) was designed “to contravene the permission previously given”; (2) was “[p]ursued ... in bad faith by subjecting [petitioners] to substantial expense in extended discovery procedures”; and (3) was brought only to enable Colt “to urge customers and potential customers of [petitioners] to refrain from doing business with them.” Id., at 10-11. Since there are “reasons completely unrelated to the provisions and purposes” of federal patent law why petitioners “may or may not be entitled to the relief [they] see[k]” under their monopolization claim, Franchise Tax Board, supra, at 26 (footnote omitted), the claim does not “arise under” federal patent law.

The same analysis obtains as to petitioners’ group-boycott claim under §1 of the Sherman Act, which provides that “[e]very contract, combination ... , or conspiracy, in restraint of trade or commerce ... is declared to be illegal,” 15 U. S. C. § 1. This claim is set forth in the allegation that “virtually all suppliers of ITS and customers of ITS have agreed with Colt to refrain from supplying and purchasing M-16 parts and accessories to or from ITS, which has had the effect of requiring ITS to close its doors and no longer transact business.” App. 11. As this case unfolded, petitioners attempted to prove that the alleged agreement was unreasonable because its purpose was to protect Colt’s trade secrets from petitioners’ infringement and, given the patents’ *813invalidity under § 112, Colt had no trade secrets to infringe. Whether or not the patent-law issue was an “essential” element of that group-boycott theory, however, petitioners could have supported their group-boycott claim with any of several theories having nothing to do with the validity of Colt’s patents. Equally prominent in the complaint, for example, is a theory that the alleged agreement was unreasonable not because Colt had no trade secrets to protect, but because Colt authorized petitioners to use them. Once again, the appearance on the complaint’s face of an alternative, non-patent theory compels the conclusion that the group-boycott claim does not “arise under” patent law.

h — I I — I

Colt offers three arguments for finding jurisdiction in the Federal Circuit, notwithstanding the well-pleaded complaint rule. The first derives from congressional policy; the second is based on Federal Rule of Civil Procedure 15(b); and the third is grounded in principles of the law of the case. We find none of them persuasive.

A

Colt correctly observes that one of Congress’ objectives in creating a Federal Circuit with exclusive jurisdiction over certain patent cases was “to reduce the widespread lack of uniformity and uncertainty of legal doctrine that exist[ed] in the administration of patent law.” H. R. Rep. No. 97-312, p. 23 (1981). Colt might be correct (although not clearly so) that Congress’ goals would be better served if the Federal Circuit’s jurisdiction were to be fixed “by reference to the case actually litigated,” rather than by an ex ante hypothetical assessment of the elements of the complaint that might have been dispositive. Brief for Respondent 31. Congress determined the relevant focus, however, when it granted jurisdiction to the Federal Circuit over “an appeal from ... a *814district court... if the jurisdiction of that court was based ... on section 1338.” 28 U. S. C. § 1295(a)(1) (emphasis added). Since the district court’s jurisdiction is determined by reference to the well-pleaded complaint, not the well-tried case, the referent for the Federal Circuit’s jurisdiction must be the same. The legislative history of the Federal Circuit’s jurisdictional provisions confirms that focus. See, e. g., H. R. Rep. No. 97-312, supra, at 41 (cases fall within the Federal Circuit’s patent jurisdiction “in the same sense that cases are said to ‘arise under’ federal law for purposes of federal question jurisdiction”). In view of that clear congressional intent, we have no more authority to read § 1295(a)(1) as granting the Federal Circuit jurisdiction over an appeal where the well-pleaded complaint does not depend on patent law, than to read § 1338(a) as granting a district court jurisdiction over such a complaint. See Pratt, 168 U. S., at 259.

B

Colt suggests alternatively that under Federal Rule of Civil Procedure 15(b)4 we should deem the complaint amended to encompass a new and independent cause of action — “an implied cause of action under section 112 of the patent laws.” Brief for Respondent 28. Such a cause of action, which Colt finds in petitioners’ summary judgment papers, would plainly “arise under” the patent laws, regardless of its merit. See 822 F. 2d, at 1566 (Nichols, J., concurring and dissenting).

We need not decide under what circumstances, if any, a court of appeals could furnish itself a jurisdictional basis unsupported by the pleadings by deeming the complaint *815amended in light of the parties’ “express or implied consent” to litigate a claim. Fed. Rule Civ. Proc. 15(b). In this case there is simply no evidence of any consent among the parties to litigate the new patent-law claim that Colt imputes to petitioners. Colt points to nothing in petitioners’ summary judgment motion expressly raising such a new cause of action, much less anything in its own motion papers suggesting consent to one. See App. 57-58. True, the summary judgment papers focused almost entirely on the patent-law issues, which petitioners deemed “[bjasic and fundamental to the subject lawsuit.” Id., at 57. But those issues fell squarely within the purview of the theories of recovery, defenses, and counterclaims that the pleadings already encompassed. Petitioners recognized as much when they moved the District Court to hold that their “claim of [patent] invalidity shall be taken as established with respect to all claims and counterclaims to which said issue is material.” Id., at 58. Thus, the patent-law focus of the summary judgment papers hardly heralded the assertion of a new patent-law claim. See, e. g., Quillen v. International Playtex, Inc., 789 F. 2d 1041, 1044 (CA4 1986); 6 C. Wright & A. Miller, Federal Practice and Procedure § 1493, p. 466 (1971). Moreover, the District Court never intimated that the patent issues were relevant to any cause of action other than the antitrust and intentional-interference claims raised expressly in the complaint; the court four times linked its judgment to “liability on Counts I and II,” without any reference to the hypothetical Count III that Colt imputes to petitioners. 609 F. Supp. 1174, 1185 (CD Ill. 1985) See also 613 F. Supp., at 332.

C

Colt’s final argument is that the Federal Circuit was obliged not to revisit the Seventh Circuit’s thorough analysis of the jurisdictional issue, but merely to adopt it as the law of the case. See also 822 F. 2d, at 1565 (Nichols, J., concurring and dissenting). “As most commonly defined, the doctrine *816[of the law of the case] posits that when a court decides upon a rule of law, that decision should continue to govern the same issues in subsequent stages in the same case.” Arizona v. California, 460 U. S. 605, 618 (1983) (dictum). This rule of practice promotes the finality and efficiency of the judicial process by “protecting against the agitation of settled issues.” IB J. Moore, J. Lucas, & T. Currier, Moore’s Federal Practice ¶0.404[1], p. 118 (1984) (hereinafter Moore’s).

Colt is correct that the doctrine applies as much to the decisions of a coordinate court in the same case as to a court’s own decisions. See, e. g., Kori Corp. v. Wilco Marsh Buggies & Draglines, Inc., 761 F. 2d 649, 657 (CA Fed.), cert. denied, 474 U. S. 902 (1985); Perkin-Elmer Corp. v. Computervision Corp., 732 F. 2d 888, 900-901 (CA Fed.), cert. denied, 469 U. S. 857 (1984). Federal courts routinely apply law-of-the-case principles to transfer decisions of coordinate courts. See, e. g., Hayman Cash Register Co. v. Sarokin, 669 F. 2d 162, 164-170 (CA3 1982) (transfer under 28 U. S. C. § 1406(a)); Skil Corp. v. Millers Falls Co., 541 F. 2d 554, 558-559 (CA6) (alternative holding) (transfer under 28 U. S. C. § 1404(a)), cert. denied, 429 U. S. 1029 (1976); IB Moore’s ¶¶0.404[4.-5], 0.404[8]. Cf. Hoffman v. Blaski, 363 U. S. 335, 340-341, n. 9 (1960) (res judicata principles did not limit power of Court of Appeals to reconsider transfer decision not upset by coordinate court). Indeed, the policies supporting the doctrine apply with even greater force to transfer decisions than to decisions of substantive law; transferee courts that feel entirely free to revisit transfer decisions of a coordinate court threaten to send litigants into a vicious circle of litigation. See Hayman, supra, at 169; Chicago & N. W. Transp. Co. v. United States, 574 F. 2d 926, 930 (CA7 1978). Cf. Blaski, supra, at 348-349 (Frankfurter, J., dissenting).5

*817Colt’s conclusion that jurisdiction therefore lay in the Federal Circuit is flawed, however, for three reasons. First, the Federal Circuit, in transferring the case to the Seventh Circuit, was the first to decide the jurisdictional issue. That the Federal Circuit did not explicate its rationale is irrelevant, for the law of the case turns on whether a court previously “decide[d] upon a rule of law” — which the Federal Circuit necessarily did — not on whether, or how well, it explained the decision. Thus, the law of the case was that the Seventh Circuit had jurisdiction, and it was the Seventh Circuit, not the Federal Circuit, that departed from the law of the case. Second, the law-of-the-case doctrine “merely expresses the practice of courts generally to refuse to reopen what has been decided, not a limit to their power.” Messenger v. Anderson, 225 U. S. 436, 444 (1912) (Holmes, J.) (citations- omitted). A court has the power to revisit prior decisions of its own or of a coordinate court in any circumstance, although as a rule courts should be loathe to do so in the absence of extraordinary circumstances such as where the initial decision was “clearly erroneous and would work a manifest injustice.” Arizona v. California, supra, at 618, n. 8 (citation omitted). Thus, even if the Seventh Circuit’s decision was law of the case, the Federal Circuit did not exceed its power in revisiting the jurisdictional issue, and once it concluded that the prior decision was “clearly wrong” it was obliged to decline jurisdiction. Most importantly, law of the case cannot bind this Court in reviewing decisions below. A petition for writ of certiorari can expose the entire case to review. Panama R. Co. v. Napier Shipping Co., 166 U. S. 280, 283-284 (1897). Just as a district court’s adherence to law of the case cannot insulate an issue from appellate review, a court of appeals’ adherence to the law of the case cannot insulate an issue from this Court’s review. See Mes*818senger, supra, at 444; Hamilton-Brown Shoe Co. v. Wolf Brothers & Co., 240 U. S. 251, 257-259 (1916).

HH <1

Our agreement with the Federal Circuit’s conclusion that it lacked jurisdiction, compels us to disapprove of its decision to reach the merits anyway “in the interest of justice.” 822 F. 2d, at 1559. “Courts created by statute can have no jurisdiction but such as the statute confers.” Sheldon v. Sill, 8 How. 441, 449 (1850). See also Firestone Tire & Rubber Co. v. Risjord, 449 U. S. 368, 379-380 (1981). The statute confers on the Federal Circuit authority to make a single decision upon concluding that it lacks jurisdiction — whether to dismiss the case or, “in the interest of justice,” to transfer it to a court of appeals that has jurisdiction. 28 U. S. C. § 1631.

The age-old rule that a court may not in any case, even in the interest of justice, extend its jurisdiction where none exists has always worked injustice in particular cases. Parties often spend years litigating claims only to learn that their efforts and expense were wasted in a court that lacked jurisdiction. Even more exasperating for the litigants (and wasteful for all concerned) is a situation where, as here, the litigants are bandied back and forth helplessly between two courts, each of which insists the other has jurisdiction. Such situations inhere in the very nature of jurisdictional lines, for as our cases aptly illustrate, few jurisdictional lines can be so finely drawn as to leave no room for disagreement on close cases. See, e. g., K mart Corp. v. Cartier, Inc., 485 U. S. 176 (1988); United States v. Hohri, 482 U. S. 64 (1987).

That does not mean, however, that every borderline ease must inevitably culminate in a perpetual game of jurisdictional ping-pong until this Court intervenes to resolve the underlying jurisdictional dispute, or (more likely) until one of the parties surrenders to futility. Such a state of affairs would undermine public confidence in our judiciary, squander *819private and public resources, and commit far too much of this Court’s calendar to the resolution of fact-specific jurisdictional disputes that lack national importance. “Surely a seemly system of judicial remedies . . . regarding controverted transfer provisions of the United States Code should encourage, not discourage, quick settlement of questions of transfer . . . Blaski, 363 U. S., at 349 (Frankfurter, J., dissenting). The courts of appeals should achieve this end by adhering strictly to principles of law of the case. See supra, at 816. Situations might arise, of course, in which the transferee court considers the transfer “clearly erroneous.” Arizona v. California, 460 U. S., at 618, n. 8. But as “[t]he doctrine of the law of the case is ... a heavy deterrent to vacillation on arguable issues,” IB Moore’s ¶0.404[1], at 124, such reversals should necessarily be exceptional; courts will rarely transfer cases over which they have clear jurisdiction, and close questions, by definition, never have clearly correct answers. Under law-of-the-case principles, if the transferee court can find the transfer decision plausible, its jurisdictional inquiry is at an end. See Fogel v. Chestnutt, 668 F. 2d 100, 109 (CA2 1981) (“The law of the case will be disregarded only when the court has ‘a clear conviction of error’ ”) (citation omitted), cert. denied, 459 U. S. 828 (1982). While adherence to the law of the case will not shield an incorrect jurisdictional decision should this Court choose to grant review, see supra, at 817-818, it will obviate the necessity for us to resolve every marginal jurisdictional dispute.

We vacate the judgment of the Court of Appeals for the Federal Circuit and remand with instructions to transfer the case to the Court of Appeals for the Seventh Circuit. See 28 U. S. C. § 1631.

It is so ordered.

with whom Justice Blackmun joins, concurring.

In a seminal case construing federal-question jurisdiction, Justice Cardozo wrote that “[w]hat is needed is something of *820that common-sense accommodation of judgment to kaleidoscopic situations which characterizes the law in its treatment of problems of causation ... a selective process which picks the substantial causes out of the web and lays the other ones aside.” Gully v. First National Bank in Meridian, 299 U. S. 109, 117-118 (1936). Although I agree with the Court’s conclusion in this case that appellate jurisdiction is in the Seventh Circuit rather than the Federal Circuit, I write separately to emphasize that a common-sense application of Justice Cardozo’s dictum requires that the answer to the question whether a claim arises under the patent laws may depend on the time when the question is asked. More specifically, if the question is asked at the end of a trial in order to decide whether the Federal Circuit has appellate jurisdiction, the answer may be different than if it had been asked at the outset to decide whether a federal district court has jurisdiction to try the case.

When Congress passed the Federal Courts Improvement Act in 1982 and vested exclusive jurisdiction in the Court of Appeals for the Federal Circuit to resolve appeals of claims that had arisen under the patent laws in the federal district courts, it was responding to concerns about both the lack of uniformity in federal appellate construction of the patent laws and the forum-shopping that such divergent appellate views had generated. Nonetheless, its definition of the Federal Circuit’s jurisdiction did not embrace all cases in which a district court had decided a patent-law question. Instead, it adopted a standard that requires the appellate court to decide whether the jurisdiction of the district court was based, in whole or in part, on a claim “arising under” the patent laws.1

*821The question whether a claim arises under the patent laws is similar to the question whether a claim arises under federal law. Although there is no single, precise, all-embracing definition of either body of law, the “vast majority” of cases that come within either “grant of jurisdiction are covered by Justice Holmes’ statement that a ‘suit arises under the law that creates the cause of action.’ Thus, the vast majority of cases brought under the general federal-question jurisdiction of the federal courts are those in which federal law creates the cause of action.” Merrell Dow Pharmaceuticals Inc. v. Thompson, 478 U. S. 804, 808 (1986) (citation omitted). In this case it is clear that the causes of action asserted by petitioners were created by the antitrust laws and not the patent laws. Congress did not create an express cause of action to enforce § 112 of the patent laws, and I find no merit in respondent’s suggestion that we should recognize an implied cause of action under § 112. Accordingly, I agree with the Court’s conclusion that the issue of wrongful retention of proprietary information that became the focus of this case under § 112 of the patent laws could not confer appellate jurisdiction in the Federal Circuit, because the issue arose as a defense rather than as a claim.2

*822To the extent that Part III-A of the Court’s opinion does nothing more than abjure the notion that the Federal Circuit has jurisdiction over patent-law issues as well as claims, I am thus in complete agreement. However, in rejecting respondent’s contention that “Congress’ goals would be better served if the Federal Circuit’s jurisdiction were to be fixed ‘by reference to the case actually litigated,’ rather than by an ex ante hypothetical assessment of the elements of the complaint that might have been dispositive,” ante, at 813, the Court’s opinion might be read as suggesting that whether patent claims are properly before the Federal Circuit on appeal should be determined by examining only the initial complaint and not by ascertaining whether a patent claim in fact was litigated in the case. Such an approach would assume that whether a case “arises under” the patent laws turns on the same considerations whether one is determining the Federal Circuit’s appellate jurisdiction or a federal district court’s original jurisdiction. But although 28 U. S. C. *823§ 1338(a) provides the basis for both types of jurisdictional assessment, I think it clear that Congress could not have intended precisely the same analysis in both instances. Two simple examples will illustrate the point.

If a patentee should file a two-count complaint seeking damages (1) under the antitrust laws and (2) for patent infringement, the district court’s jurisdiction would unquestionably be based, at least in part, on § 1338(a). If, however, pretrial discovery convinced the plaintiff that no infringement had occurred, and Count 2 was therefore dismissed voluntarily in advance of trial, the case that would actually be litigated would certainly not arise under the patent laws for purposes of appellate jurisdiction. Even though the district court’s original jurisdiction when the complaint was filed had been based, in part, on § 1338(a), the case would no longer be one arising under the patent laws for purposes of Federal Circuit review when the district court’s judgment was entered. Conversely, if an original complaint alleging only an antitrust violation should be amended after discovery to add a patent-law claim, and if the plaintiff should be successful in proving that its patent was valid and infringed but unsuccessful in proving any basis for recovery under the antitrust laws, the district court’s judgment would sustain a claim arising under the patent laws even though the complaint initially invoking its jurisdiction had not mentioned it, and an appeal would properly lie in the Federal Circuit.

Whether the complaint is actually amended, as in the previous example, or constructively amended to conform to the proof, see Fed. Rule Civ. Proc. 15(b),3 Congress’ goal of en*824suring that appeals of patent-law claims go to the Federal Circuit would be thwarted by determining that court’s appellate jurisdiction only through an examination of the complaint as initially filed. That approach would enable an unscrupulous plaintiff to manipulate appellate court jurisdiction by the timing of the amendments to its complaint. The Court expressly leaves open the question whether a constructive amendment could provide the foundation for Federal Circuit patent-law jurisdiction, see ante, at 814-815,4 and says nothing on the subject whether actual amendments to the complaint can so suffice. But since respondent has asked us to rule in its favor on the ground that petitioners’ complaint added a patent-law claim through constructive amendment, I think we should make it perfectly clear that even though respondent’s approach to the jurisdictional question is sound, its application of that approach to this case fails because the claim that was actually litigated did not arise under the patent laws. Nevertheless, since what the Court has written is not inconsistent with this view, I join its opinion.