JERRY E. SMITH, Circuit Judge:

Having been granted leave to pursue an interlocutory appeal, see FED.R.CIV.P. 23(f), defendants challenge an order certifying a single class of plaintiffs. Relying largely on Central Bank, N.A. v. First Interstate Bank, N.A., 511 U.S. 164, 114 S.Ct. 1439, 128 L.Ed.2d 119 (1994), and its progeny, we reverse and remand.

I.

The facts are difficult to detail but easy to summarize. Plaintiffs allege that defendants Credit Suisse First Boston ("Credit Suisse"), Merrill Lynch Company, Inc. ("Merrill Lynch"), and Barclays Bank PLC ("Barclays Bank") (collectively "the banks") entered into partnerships and transactions that allowed Enron Corporation ("Enron") to take liabilities off of its books temporarily and to book revenue from the transactions when it was actually incurring debt. The common feature of these transactions is that they allowed Enron to misstate its financial condition; there is no allegation that the banks were fiduciaries of the plaintiffs, that they improperly filed financial reports on Enron's behalf, or that they engaged in wash sales or other manipulative activities directly in the market for Enron securities.

For example, plaintiffs allege that Merrill Lynch engaged in what they dub the "Nigerian Barges Transaction." According to plaintiffs, Enron wanted to "sell" its interest in electricity-generating barges off the coast of Nigeria by the end of 1999 so that it could book revenue and meet stock analysts' estimates for the calendar quarter. It could find no legitimate buyer, so it contacted Merrill Lynch and guaranteed that it would buy the barges back within six months at a premium for Merrill Lynch.

Six months later, Enron made good on its guarantee; an Enron-controlled partnership bought the barges from Merrill Lynch at a premium. When Enron reported its results for 1999, instead of booking the transaction as a loan, the characterization that Enron's outside accountants state would have been appropriate had they known of the side-agreement to buy back the barges, Enron booked the transaction as a sale and accordingly listed the revenue therefrom in its year-end financial statement.

Plaintiffs allege that the banks knew exactly why Enron was engaging in seemingly irrational transactions such as this. They cite certain of the banks' internal communications they characterize as proving that the banks were aware of the personal compensation Enron executives received as a result of inflating their stock price through the illusion of revenue and that the banks intended to profit by helping the executives maintain that illusion. Likewise, plaintiffs allege that, although each defendant may not have been aware of exactly how each other defendant was helping Enron to misrepresent its financial health, the defendants knew in general that other defendants were doing so and that Enron was engaged in a long-term scheme to defraud investors and maximize executive compensation by inflating revenue and disguising risk and liabilities through its partnerships and transactions with the banks.

II.

This suit followed Enron's collapse in 2001. The first action was filed on October 22 of that year; by December 12, 2001, the district court had consolidated over thirty actions relating to Enron securities and had designated the Regents of the University of California as the lead plaintiff. Years of discovery have ensued, and tens of millions of documents have been produced.

Early in the litigation, the banks filed motions to dismiss, but the district court denied them in a December 19, 2002, opinion. The court reconsidered some of the issues relevant to those motions in its opinion regarding class certification, issued on June 5, 2006, in light of intervening developments in appellate caselaw. The court justified its reconsideration, stating that it had

the power to reconsider such interlocutory decisions, especially in light of the limited and much of it recent case law emerging on scheme liability. Moreover . . . at class certification, especially after such substantial discovery as has been done here, the court may look behind the pleadings at evidence to determine whether a class should be certified.

The court determined that a "deceptive act" within the meaning of rule 10b-5(c) includes participating in a "transaction whose principal purpose and effect is to create a false appearance of revenues."

The district court decided that rule 10b-5(a)'s prohibition of any "scheme . . . to defraud" gives rise to joint and several liability for defendants who commit individual acts of deception in furtherance of such a scheme. Implicit in that ruling is the conclusion that plaintiffs have alleged that just such a scheme existed.

The court's theory of scheme liability considerably simplified finding commonality among the plaintiffs with respect to loss causation. The court stated that "a reasonable argument can be made that where a defendant knowingly engaged in a primary violation of the federal securities laws that was in furtherance of a larger scheme, it should be jointly and severally liable for the loss caused by the entire overarching scheme, including conduct of other scheme participants about which it knew nothing."

The district court concluded that plaintiffs are entitled to rely on the classwide presumptions of reliance for omissions and fraud on the market. The court held that the Affiliated Ute presumption applies because the facts indicate that the banks had failed in their "duty not to engage in a fraudulent `scheme.'" The court concluded, with respect to the fraud-on-the-market presumption of reliance, that no preliminary finding of market efficiency or investors' reliance thereon need be made where plaintiffs plead under rule 10b-5(a) (forbidding deceptive devices, schemes, and artifices) and 10b-5(c) (prohibiting deceptive acts, practices, and courses of business) rather than under the more usual cause of action, rule 10b-5(b) (proscribing misrepresentation).

A month after issuing its opinion on class certification, the district court, after reviewing plaintiffs' revised class definition and trial plan, issued its class certification order, dated July 5, 2006. It determined that, although the proportionate liability provisions of the Private Securities Litigation Reform Act ("PSLRA") are generally problematic, there is no necessary conflict between the court's theory of liability and that statute. See 15 U.S.C. § 78u-4(b)(4), (f)(2). The court ordered defendants to prepare a list of non-parties to whom they intend to assign responsibility and declared that defendants will bear the burden to prove non-parties' responsibility by a preponderance of the evidence.

The district court certified a class of all persons who purchased Enron securities between October 19, 1998, and November 27, 2001, and were injured thereby. The class seeks damages of $40 billion, against which losing defendants would be entitled to offset roughly $7 billion obtained by plaintiffs in previous settlements with former co-defendants. On November 1, 2006, a motions panel of this court granted defendants leave to appeal the class certification order, and we sua sponte expedited the appeal.

III.

Plaintiffs point out that we are not bound by the motion panel's decision to grant leave to appeal; they urge that leave to appeal was improvidently granted. We disagree.

This is a legally and practically significant class certification decision, and the motions panel properly allowed the appeal. The commentary to rule 23(f) indicates that it is appropriate to grant leave to appeal an adverse determination where (1) a "certification decision turns on a novel or unsettled question of law" or (2) "[a]n order granting certification . . . may force a defendant to settle rather than incur the costs of defending a class action and run the risk of potentially ruinous liability." FED.R.CIV.P. 23(f) advisory committee note.

Plaintiffs contend that, even if we are entitled to address defendants' merits-based arguments, those arguments are sufficiently intertwined with the facts to counsel against interlocutory appeal before a complete factual record is established. Plaintiffs reason that the banks have not yet been "cowed" into settling, nor are they likely to be; the district court has afforded procedural fairness to all parties; and the panel should defer to its judgment by declining to hear this appeal until after the district court has entered a final judgment.

The fact that the banks have not yet been persuaded to settle is no reason to decline a rule 23(f) appeal; it means only that the litigation continues. As we have recognized, class certification may be the backbreaking decision that places "insurmountable pressure" on a defendant to settle, even where the defendant has a good chance of succeeding on the merits. See Castano v. Am. Tobacco, 84 F.3d 734, 746 (5th Cir. 1996). Here, where plaintiffs seek to hold the banks liable for nearly the entirety of securities losses stemming from the Enron collapse, settlement pressure appears to be particularly acute, so it is appropriate to provide appellate review before settlement may be coerced by an erroneous class certification decision.

Moreover, although the legal issues underlying the certification decision are intertwined with the merit of plaintiffs' theory of liability, these broad legal issues are not especially contingent on particular facts likely to be further developed in the district court. This case gives rise to unsettled questions of law concerning the entanglement of the merits with the class certification decision, as well as the district court's theory of "deceptive act" liability underlying its finding that common issues of reliance predominate. Both of the Advisory Committee criteria are met here, so we proceed to consider the rule 23(f) appeal.

IV.

We review class certification decisions for abuse of discretion in "recognition of the essentially factual basis of the certification inquiry and of the district court's inherent power to manage and control pending litigation. . . . Whether the district court applied the correct legal standard in reaching its decision on class certification, however, is a legal question that we review de novo." Allison v. Citgo Petroleum Corp., 151 F.3d 402, 408 (5th Cir. 1998) (citations omitted). Where a district court premises its legal analysis on an erroneous understanding of governing law, it has abused its discretion. See Unger v. Amedisys Inc., 401 F.3d 316, 320 (5th Cir. 2005). Albeit with the best of intentions and after herculean effort, the district court arrives at an erroneous understanding of securities law that gives rise to its application of classwide presumptions of reliance.

V.

We first consider the scope of our jurisdiction. Plaintiffs accuse the banks of re-packaging this rule 23(f) appeal as a demurrer. They point out that the scope of our review is "bridled by Rule 23(f)." They urge us to refuse to consider the banks' arguments concerning the interpretation of § 10(b) and the prohibition of aiding and abetting liability under that statute. They contend that we should accept the district court's view of the underlying theories of liability as valid for purposes of this appeal.

The scope of our review is limited, but it is not quite so circumscribed as plaintiffs say. Although we may not conduct an independent inquiry into the legal or factual merit of this case as though we were reviewing a motion under Federal Rule of Civil Procedure 12(b)(6) or 56, we may address arguments that implicate the merits of plaintiffs' cause of action insofar as those arguments also implicate the merits of the class certification decision.

Rule 23(f) states that "[a] court of appeals may in its discretion permit an appeal from an order of a district court granting or denying class action certification under this rule if application is made to it within ten days after entry of the order." FED.R.CIV.P. 23(f). The text of the rule makes plain that the sole order that may be appealed is the class certification; "no other issues may be raised." Bell, 422 F.3d at 314. The fact that an issue is relevant to both class certification and the merits, however, does not preclude review of that issue.

After all, the Supreme Court has refused to designate class certification decisions as collateral orders subject to interlocutory appellate review, precisely because "the class determination generally involves considerations that are enmeshed in the factual and legal issues comprising the plaintiff's cause of action." Coopers Lybrand v. Livesay, 437 U.S. 463, 469, 98 S.Ct. 2454, 57 L.Ed.2d 351 (1978) (internal citations omitted). "The analysis under Rule 23 must focus on the requirements of the rule, and if findings made in connection with those requirements overlap findings that will have to be made on the merits, such overlap is only coincidental." Bell, 422 F.3d at 311. See also Unger, 401 F.3d at 320.

Our circuit's conclusion that review of the factual and legal analysis supporting the district court's decision is appropriate on review of class certification enjoys widespread acceptance in the courts of appeals, and neither the Supreme Court authority nor the Fifth Circuit caselaw that plaintiffs cite for the proposition that no merits inquiry is permitted is to the contrary. Miller and Eisen (which cited Miller to establish the "no merits inquiry" rule) addressed cases in which district courts had conducted wide-ranging inquiries into the merits of claims as part of the class certification decision without reference to the criteria for class certification. As the Bell rule cited above suggests, the prohibition against looking into the merits applies only to such inquiries, not to evaluations of the merits that overlap with consideration of the requirements for class certification. Id. In a rule 23(f) appeal, this court can, and in fact must, review the merits of the district court's theory of liability insofar as they also concern issues relevant to class certification.

VI.

Two of the banks' arguments on appeal have considerable implications for the substantive legal merit of plaintiffs' complaint. First, the district court's definition of "deceptive act" underlies its application of the classwide presumption of reliance on fraud on the market. Likewise, its broad theory of "scheme" liability allows it to certify a single class of plaintiffs whose losses were caused in common by the scheme rather than to certify subclasses whose losses were caused by the actions of individual defendants. Both of these arguments are also highly relevant to class certification, but we address only the definition of "deceptive act," because it is dispositive of this appeal.

The district court's definition of "deceptive act" is integral to its conclusion that the requirements for class certification are met. Federal Rule of Civil Procedure 23(a) requires plaintiffs seeking class certification to satisfy four criteria that we have previously summarized as (1) numerosity, (2) commonality, (3) typicality, and (4) representativeness. Because no defendant seriously challenges whether these prima facie requirements are met, we do not discuss them further.

Once the rule 23(a) requirements are satisfied, a class may be certified if "[1] the court finds that the questions of law or fact common to the members of the class pre-dominate over any questions affecting only-individual members, and [2] that a class action is superior to other available methods for the fair and efficient adjudication of the controversy." Rule 23(b)(3). We refer to these two requirements respectively as the "predominance" and "superiority" criteria.

The district court's theory of liability implicates primarily the predominance requirement. To succeed on a claim of securities fraud, a plaintiff must prove "(1) a material misrepresentation or omission by the defendant, (2) scienter on the part of the defendant, (3) reliance, and (4) due diligence by the plaintiff to pursue his or her own interest with care and good faith." linger, 401 F.3d at 322 n. 2 (5th Cir. 2005) (citations omitted). A plaintiff must prove not only that the fraud occurred but that it proximately caused his losses. See Dura Pharms., Inc. v. Broudo, 544 U.S. 336, 346, 125 S.Ct. 1627, 161 L.Ed.2d 577 (2005).

Without its broad conception of liability for "deceptive acts," the district court could not have found that the entire class was entitled to rely on Basic's fraud-on-the-market theory, because the market may not be presumed to rely on an omission or misrepresentation in a disclosure to which it was not legally entitled. The plaintiffs are likely correct that the market for Enron securities was efficient and that inherent in that conclusion is the fact that the market price reflected all publicly available information. But the factual probability that the market relied on the banks' behavior and/or omissions does not mean that plaintiffs are entitled to the legal presumption of reliance.

Market efficiency was not the sole condition that the Court in Basic required plaintiffs to prove existed to qualify for the classwide presumption; the defendant had to make public and material misrepresentations. See Basic, 485 U.S. at 248 n. 27, 108 S.Ct. 978. If the banks' actions were non-public, immaterial, or not misrepresentative because the market had no right to rely on them (in other words, the banks owed no duty), the banks should be able to defeat the presumption. See Gariety, 368 F.3d at 369.

Without a classwide presumption of reliance, plaintiffs would have to prove individual reliance on defendants' conduct. "[A] fraud class action cannot be certified when individual reliance will be an issue." Castano, 84 F.3d at 745. Because, as we will explain, the district court misapplies the Affiliated Ute presumption, the fraud-on-the-market presumption is the only presumption potentially available in this case. Accordingly, the meaning of "deceptive act" is critical to classwide certification; classwide reliance stands or falls with it.

Erroneous presumptions of reliance were at the heart of the Supreme Court's concern when it ruled that § 10(b) does not give rise to aiding and abetting liability. It is essential for us to ensure that the district court does not misapply aiding-and-abetting liability under the guise of primary liability, through an overly broad definition of "deceptive act[s]," and thereby give rise to an erroneous classwide presumption of fraud on the market.

VII.

We proceed to the merits of this limited rule 23(f) appeal. With all respect for the district court's diligent efforts, its determination that the Affiliated Ute presumption applies to the facts of this case is incorrect. In Affiliated Ute, the Court considered whether a group of investors was required to prove reliance affirmatively where it alleged that bank officers bought their restricted stock without disclosing the bank's creation of a secondary market in which the stock would be resold for profit. See 406 U.S. at 133-39, 92 S.Ct. 1456. The Court ruled that the investors' allegations were not based on misrepresentation under what is now rule 10b-5(b), but instead on a "`course of business' or a `device, scheme or artifice' that operated as a fraud" under what are now rule 10b-5(a) and (c). Id. at 153, 92 S.Ct. 1456. The Court determined that, unlike a mere transfer agent, these bankers had a duty to disclose the existence of this secondary market to the plaintiffs. Id. at 152-53, 92 S.Ct. 1456.

Where liability is premised on a failure to disclose rather than on a misrepresentation, "positive proof of reliance is not a prerequisite to recovery. . . . This obligation to disclose and the withholding of a material fact establish the requisite element of causation in fact." Id. at 153-54, 92 S.Ct. 1456. In Basic, 485 U.S. at 243, 108 S.Ct. 978, the Court later summarized the rule of Affiliated Ute thusly: "[W]here a duty to disclose material information had been breached . . . the necessary nexus between the plaintiffs' injury and the defendants' wrongful conduct had been established."

For us to invoke the Affiliated Ute presumption of reliance on an omission, a plaintiff must (1) allege a case primarily based on omissions or non-disclosure and (2) demonstrate that the defendant owed him a duty of disclosure. The case at bar does not satisfy this conjunctive test.

Assuming arguendo that plaintiffs' case primarily concerns improper omissions, the banks were not fiduciaries and were not otherwise obligated to the plaintiffs. They did not owe plaintiffs any duty to disclose the nature of the alleged transactions. The district court agrees that the banks lacked any specific duty, but, citing our caselaw, the court finds that the presumption applies because the banks omitted their duty not to engage in a fraudulent scheme. Neither Smith nor any other of this circuit's cases is authority for that proposition.

As we will explain in more detail, "deception" within the meaning of § 10(b) requires that a defendant fail to satisfy a duty to disclose material information to a plaintiff. Merely pleading that defendants failed to fulfill that duty by means of a scheme or an act, rather than by a misleading statement, does not entitle plaintiffs to employ the Affiliated Ute presumption. In Smith, this court discussed only the first element of the Abell test recounted above; Smith stands for the straightforward proposition that a case brought under rule 10b-5(a) or (c) is more likely to be based primarily on omission than is a case under rule 10b-5(b), which requires that a defendant affirmatively make a misleading representation. In Smith, we never reached the second prong of the Abell test, because we determined that, even if there had been an omission, any presumption of reliance was rebutted (because the plaintiff would have behaved identically had he been aware of the omitted information). See Smith, 845 F.2d at 1364. Abell is the law of this circuit, and Smith is not to the contrary. When it determined (correctly) that the banks owed no duty to the plaintiffs other than the general duty not to engage in fraudulent schemes or acts (that is, the duty not to break the law), the district court should have declined to apply the Affiliated Ute presumption. Instead, it presumed what the plaintiffs had only alleged: that reliance, which is a specific, defining element of the relevant legal violation, had in fact occurred.

The logic of Affiliated Ute is that, where a plaintiff is entitled to rely on the disclosures of someone who owes him a duty, requiring him to prove "how he would have acted if omitted material information had been disclosed" is unfair. Basic, 485 U.S. at 245, 108 S.Ct. 978. It is natural to expect a plaintiff to rely on the candor of one who owes him a duty of disclosure, and it is fair to force one who breached his duty to prove that the plaintiff did not so rely. Here, however, where the plaintiffs had no expectation that the banks would provide them with information, there is no reason to expect that the plaintiffs were relying on their candor. Accordingly, it is only sensible to put plaintiffs to their proof that they individually relied on the banks' omissions.

VIII.

Having determined that the Affiliated Ute presumption is inapplicable, we proceed to review the district court's determination that Basic's fraud-on-the-market presumption applies. It does not; the court predicates its ruling on an erroneous interpretation of § 10(b).

The banks launch a two-pronged attack on the district court's ruling with respect to fraud on the market. First, they argue that the presumption was never properly established: The district court's overly broad definition of "deceptive act" led it inexorably to the mistaken conclusion that the banks' actions constituted "misrepresentations" on which the market was legally presumed to rely. Second, the banks assert that, even if the plaintiffs did establish the presumption, it was rebutted according to this court's standards. Because the district court erred in ruling that the presumption had been established, we do not address whether it was rebutted.

In Basic, 485 U.S. at 245, 108 S.Ct. 978, the Court accepted the fraud-on-the-market theory that courts could presume reliance where individuals who had traded shares did so "in reliance on the integrity of the price set by the market, but because of [defendant's] material misrepresentations, that price had been fraudulently [altered]." The presumption is founded on the economic hypothesis that "the market is transposed between seller and buyer, and ideally, transmits information to the investor in the processed form of the market price. . . . The market is acting as the unpaid agent of the investor, informing him that given all the information available to it, the value of the stock is worth the market price." Id. at 246, 108 S.Ct. 978 (citing In re LTV Sec. Litig., 88 F.R.D. 134, 143 (N.D.Tex.1980)).

To qualify for the presumption, however, a plaintiff must not only indicate that a market is efficient, but also must allege that the defendant made public and material misrepresentations; i.e., the type of fraud on which an efficient market may be presumed to rely. These plaintiffs have not alleged such fraud.

The district court's conception of "deceptive act" liability is inconsistent with the Supreme Court's decision that § 10 does not give rise to aiding and abetting liability. An act cannot be deceptive within the meaning of § 10(b) where the actor has no duty to disclose. Presuming plaintiffs' allegations to be true, Enron committed fraud by misstating its accounts, but the banks only aided an abetted that fraud by engaging in transactions to make it more plausible; they owed no duty to Enron's shareholders.

Section 10(b) does not give rise to aiding and abetting liability. In Central Bank, the Court emphasized that securities fraud liability is an area of the law that demands certainty and predictability. Secondary liability brings neither; instead it gives rise to confusion about the extent of secondary actors' obligations and invites vague and conflicting standards of proof in divers courts. See Cent. Bank, 511 U.S. at 188, 114 S.Ct. 1439. Unfortunately, the Court has left some uncertainty in this regard.

Though the Court conclusively foreclosed the application of secondary liability under § 10(b), it stated that secondary actors such as investment banks and accountants can be liable as primary violators in some circumstances. Id. at 191, 114 S.Ct. 1439. The Court has never, however, precisely delineated the boundary between primary and secondary liability. As the district court noted, the lower courts have struggled to do so, and our circuit has not previously announced a standard that conclusively governs this case.

Although plaintiffs try to reconcile the cases, the Eighth and Ninth Circuits have split with respect to the scope of primary liability for secondary actors. The district court adopts a rule advocated by the Securities and Exchange Commission ("SEC"), in an amicus curiae brief before the Ninth Circuit, under which primary liability attaches to anyone who engages in a "transaction whose principal purpose and effect is to create a false appearance of revenues." We agree with the Eighth Circuit that the SEC's proposed test (by which we are not bound) is too broad to fit within the contours of § 10(b).

The appropriate starting point is the text of the statute. See Cent. Bank, 511 U.S. at 172-73, 114 S.Ct. 1439. Decisions interpreting the statutory text place a limit on the possible definitions that can be ascribed to the words contained in the SEC's rule promulgated thereunder. It is by losing sight of the limits that the statute places on the rule, and by ascribing, natural, dictionary definitions to the words of the rule, that the district court and like-minded courts have gone awry. Central Bank was informed by a series of decisions construing the statute and narrowly defining the scope of "fraud" in the context of securities. In Ernst Ernst, 425 U.S. 185, at 197, 96 S.Ct. 1375, 47 L.Ed.2d 668, for example, the Court rejected the SEC's notion that securities fraud can be committed negligently; it has to be knowing. Even more significantly for purposes of this case, the Court later stated that "the language of § 10(b) gives no indication that Congress meant to prohibit any conduct not involving manipulation or deception." Santa Fe Indus, v. Green, 430 U.S. 462, 473, 97 S.Ct. 1292, 51 L.Ed.2d 480 (1977). The Court had already discussed "manipulation" in Ernst Ernst: "Use of the word `manipulative' is especially significant. It is and was virtually a term of art when used in connection with securities markets. It connotes intentional or willful conduct designed to deceive or defraud investors by controlling or artificially affecting the price of securities." Ernst Ernst, 425 U.S. at 199, 96 S.Ct. 1375.

The Court further refined that definition by stating that "[manipulation] refers generally to practices, such as wash sales, matched orders, or rigged prices, that are intended to mislead investors by artificially affecting market activity." Santa Fe, 430 U.S. at 476, 97 S.Ct. 1292. Finally, when evaluating the scope of liability for deceptive omissions of disclosure in the context of insider trading, the Court stated, "When an allegation of fraud is based upon non-disclosure, there can be no fraud absent a duty to speak." Chiarella v. United States, 445 U.S. 222, 234, 100 S.Ct. 1108, 63 L.Ed.2d 348 (1980).

The Eighth Circuit, unlike the Ninth, has correctly taken these decisions collectively to mean that "`deceptive' conduct involves either a misstatement or a failure to disclose by one who has a duty to disclose." Charter, 443 F.3d at 990. That court quoted the technical definition of "manipulation" from Santa Fe and stated that "any defendant who does not make or affirmatively cause to be made a fraudulent statement or omission, or who does not directly engage in manipulative securities trading practices, is at most guilty of aiding and abetting and cannot be held liable under § 10(b) or any subpart of Rule 10b-5." Id. at 992.

By this holding the court in Charter found that there was no liability against vendors of set-top cable boxes who had sold their boxes to Charter at inflated prices subject to a kickback agreement whereby they would direct the value of the price inflation back to Charter in the form of advertising purchases. See id. at 989-90. The vendors were alleged to have known that Charter was doing this to falsify its accounts by depreciating its expenses, as capital investments, from the purchase of the set-top boxes, but was booking the increased advertising fees as recurring revenue. See id. In other words, the court dismissed the case on facts extraordinarily similar to the facts that are present here.

The Ninth Circuit came to a different conclusion. In Simpson, the defunct company (Homestore.com) "bought revenue" by engaging in the same type of round-trip transactions that took place in Charterand are alleged to have occurred here. Simpson, 452 F.3d at 1043-44. It paid inflated prices for shares or services from thinly capitalized companies looking to generate liquidity so they could go public, in return for which it extracted side agreements that the companies would pay back the value of the inflation by buying advertising from AOL to be displayed at Homestore's AOL-based website. Like Charter, Homestore would then list its payments to the other companies as capital investments but would characterize its advertising income from them as recurring revenue. See id.

Although the defendants were dismissed because they did not meet the standard for liability that the Ninth Circuit announced, the court promulgated a standard very similar to the one the instant plaintiffs urge us to adopt. The court concluded:

[C]onduct by a defendant that had the principal purpose and effect of creating a false appearance in deceptive transactions as part of a scheme to defraud is conduct that uses or employs a deceptive device within the meaning of § 10(b). Furthermore, such conduct may be in connection with the purchase or sale of securities if it is part of a scheme to misrepresent public financial information where the scheme is not complete until the misleading information is disseminated into the securities market. Finally, a plaintiff may be presumed to have relied on this scheme to defraud if a misrepresentation, which necessarily resulted from the scheme and the defendant's conduct therein, was disseminated into an efficient market and was reflected in the market price.

Id. at 1052. See also In re Parmalat Sec. Litig., 376 F.Supp.2d 472, 481-90 (S.D.N.Y.2005).

The Ninth Circuit relied in part on a law review article that had questioned the assertion that a defendant could be liable only for its own statements because § 10(b) forbids the use of a "device" and rule 10b-5 condemns those who act "indirectly." What this reasoning overlooks is that the Supreme Court had appeared to limit the scope of "deception" rather than the scope of "device."

The Supreme Court has defined "device" by referring to a dictionary but has pointedly refused to define "deceptive" in any way except through caselaw: "[D]evice" means "(t)hat which is devised, or formed by design; a contrivance; an invention; project; scheme; often, a scheme to deceive; a stratagem; an artifice," and "contrivance" in pertinent part as "(a) thing contrived or used in contriving; a scheme, plan, or artifice." In turn, "contrive" in pertinent part is defined as "(t)o devise; to plan; to plot . . . (t)o fabricate . . . design; invent . . . to scheme. . . ." Ernst Ernst, 425 U.S. at 199 n. 20, 96 S.Ct. 1375 (citing WEBSTER'S INTERNATIONAL DICTIONARY (2d ed.1934)). Having established the meaning of "device" (and relying on it to hold that § 10(b) requires scienter), the Court, in its other eases interpreting § 10(b), has established that a device, such as a scheme, is not "deceptive" unless it involves breach of some duty of candid disclosure.

For this reason, defining "deceptive" by referring to the same dictionary the Court used to define "device" — the approach taken by the court in Parmalat, 376 F.Supp.2d at 502, and approvingly cited by the district court a quo — is improperly to substitute the authority of the dictionary for that of the Supreme Court. Likewise, plaintiffs' reference to the common law meaning of "deceptive" is fruitless; where the Supreme Court has authoritatively construed the pertinent language of the statute giving rise to the plaintiffs' cause of action, the common law meaning of that language is irrelevant.

Although some of our securities cases have considered the common law where the Supreme Court has placed no gloss on the relevant terms, none of this court's decisions has contradicted either the fundamental principle just stated or the Supreme Court's interpretation of "deceptive." Because "device" is modified by "deceptive," no device can be illegal if it is not deceptive within the meaning of the statute. Similarly, because the rule may not be broader than the statute, this conclusion as to the meaning of "deceptive device" precludes an interpretation of "indirectly" that contradicts the accepted meaning of "deception."

The district court's definition of "deceptive acts" thus sweeps too broadly; the transactions in which the banks engaged were not encompassed within the proper meaning of that phrase. Enron had a duty to its shareholders, but the banks did not. The transactions in which the banks engaged at most aided and abetted Enron's deceit by making its misrepresentations more plausible. The banks' participation in the transactions, regardless of the purpose or effect of those transactions, did not give rise to primary liability under § 10(b).

IX.

Having determined that the banks' alleged actions were not "misrepresentations" in the sense of "deceptive acts" on which an efficient market may be presumed to rely, we proceed to consider whether they constituted manipulation. They did not.

Manipulation requires that a defendant act directly in the market for the relevant security. The Supreme Court has cited a dictionary definition of the word but, at the same time, has attached the caveat that, as used in securities fraud law, it is "virtually a term of art." Ernst Ernst, 425 U.S. at 199 n. 21, 96 S.Ct. 1375. Although the Court has not precisely defined the term beyond providing a few examples such as wash sales, matched orders, and rigged prices, then-District Judge Higginbotham, in an influential opinion issued shortly after Santa Fe, exhaustively analyzed the meaning of "manipulation" and concluded that "[f]rom this study, the following definition emerges: practices in the marketplace which have the effect of either creating the false impression that certain market activity is occurring when in fact such activity is unrelated to actual supply and demand or tampering with the price itself are manipulative." Hundahl v. United Benefit Life Ins. Co., 465 F.Supp. 1349, 1360 (N.D.Tex.1979).

In Hundahl, Judge Higginbotham carefully emphasized that such activity could not take place outside the market for the relevant security and retain the title of manipulation; conduct that affects the marketplace indirectly can violate § 10(b) only if it constitutes deception. Id. at 1359, 1362. Like the Eighth Circuit, we adopt Judge Higginbotham's reasoning and definition in full, and we are aware of no circuit that recognizes a broader definition. See Charter, 443 F.3d at 992 n. 2.

Plaintiffs argue that this course is foreclosed to us by Schreiber v. Burlington Northern Inc., 472 U.S. 1, 6-7, 105 S.Ct. 2458, 86 L.Ed.2d 1 (1985), and Shores. We disagree. In Schreiber, 472 U.S. at 6, 105 S.Ct. 2458, the Court declared only that its definition of manipulation, insofar as it had defined that term in Ernst Ernst, is consistent with both the dictionary and the common law. So is Judge Higginbotham's.

In Hundahl, Judge Higginbotham thoroughly analyzed the common law history of the term and concluded that the "manipulation" cause of action was primarily concerned with keeping free markets clear of interference but does not reach all conduct that might constitute deception or breach of fiduciary duty. See Hundahl, 465 F.Supp. at 1359-62. The Court in Schreiber, 472 U.S. at 7, 105 S.Ct. 2458, citing Santa Fe, adopted a similar limited construction to determine that not all breaches of state law fiduciary duty constituted manipulation for purposes of the federal securities laws. The fact that the Supreme Court's definition of "manipulation" is consistent with the dictionary's does not mean that it is coextensive with it; "manipulation" is a term of art, and it applies only to conduct that takes place directly within the market for the relevant security.

Our holding in Shores requires somewhat more explanation. In that case, we adopted the "fraud-created-the-market" theory, whereby actors who introduced an otherwise unmarketable security into the market by means of fraud are deemed guilty of manipulation, and a plaintiff can plead that he relied on the integrity of the market rather than on individual fraudulent disclosures. Shores, 647 F.2d at 469-70 n. 8. We determined that lawyers and other secondary actors involved in preparing the fraudulent statements that facilitated introduction of the otherwise unmarketable security could be liable for the plaintiff security purchaser's loss. See id. Shores does not preclude the decision we reach in this case. The basis of the fraud-created-the-market theory is that the fraudster directly interfered with the market by introducing something that is not like the others: an objectively unmarketable security that has no business being there. This is qualitatively different from what the banks are alleged to have done, namely engage in transactions elsewhere that gave a misleading impression of the value of Enron securities that were already on the market. Moreover, in Shores the manipulation happened when the bogus security was introduced into the market; lawyers and other secondary actors were rendered liable for having conspired to achieve that end. In the wake of Central Bank, however, conspiracy is no longer a viable theory of § 10(b) liability, so that aspect of Shores has been overruled.

Nothing in today's decision contradicts our precedent. Applying the Hundahl definition of manipulation, we conclude that the banks' actions are not alleged to be the type of manipulative devices on which an efficient market may be legally presumed to rely because the banks did not act directly in the market for Enron securities.

X.

As the Supreme Court did in Central Bank, it may be worth taking into account certain policy considerations to determine whether our interpretation of § 10(b) plausibly accords with the will of Congress. Defendants do, after all, escape liability for alleged conduct that was hardly praiseworthy. According to plaintiffs, defendants could have pulled the plug on the Enron fraud; instead they profited from it while large numbers of people eventually lost an aggregate sum in the tens of billions of dollars.

Ultimately, however, the rule of liability must be either overinclusive or underinclusive so as to avoid what Hundahl called " in terrorem settlements" resulting from the expense and difficulty of, even meritoriously, defending this kind of litigation. Hundahl, 465 F.Supp. at 1363. Strict construction of § 10(b) against inputting aiding and abetting liability for secondary actors under the rubric of "deceptive acts" or "schemes" gives rise to the type of certainty that the Court sought in Central Bank. The banks may exaggerate the length of the parade of horribles they present wherein defendants are continually taken out of and put back into endless securities cases based on, shifting, ad hoc, fact-based perceptions of liability influenced by plaintiffs' skill at artful pleading.

But the fact that the banks may be on to something serious might be best demonstrated by the fact that in Simpson, 452 F.3d at 1050, the court attempted to distinguish Charter as addressing an arms-length transaction not subject to primary liability, i.e., a ruling consistent with its own. If there is a distinct difference between the culpability of defendants' actions based on the pleadings in those two cases, it is not apparent to us and is likely beyond the understanding of good-faith financial professionals who are attempting to avoid liability.

This is not to say that the instant matter should be decided in accord with this court's policy preferences. We mention policy only to demonstrate that, even considering the scope of the Enron disaster, Congress was not irrational to promote plain legal standards for actors in the financial markets by limiting secondary liability. As the Eighth Circuit has said,

To impose liability for securities fraud on one party to an arm's length business transaction in goods or services other than securities because that party knew or should have known that the other party would use the transaction to mislead investors in its stock would introduce potentially far-reaching duties and uncertainties for those engaged in day-to-day business dealings. Decisions of this magnitude should be made by Congress.

Charter, 443 F.3d at 992-93.

XI.

The necessity of establishing a classwide presumption of reliance in securities class actions makes substantial merits review on a rule 23(f) appeal inevitable. A classwide presumption of reliance is not only crucial to class certification, it prima facie establishes a critical element of the substantive tort. Reliance "provides the requisite causal connection between a defendant's misrepresentation and a plaintiff's injury." Basic, 485 U.S. at 243, 108 S.Ct. 978. Where the plaintiffs' several interactions with the market are alleged to supply the basis for their joint reliance on defendants' conduct, we must examine carefully the third leg of that metaphorical triangle: the legal nature of defendants' interactions with the market.

If, as is probably the case here, that legally appropriate examination makes interlocutory appeals in securities cases practically dispositive of the merits, we take comfort in two observations. First, the availability of broad presumptions in this area means that the legal merit of securities cases is somewhat less likely than that of other cases to be contingent on facts that have been only incompletely developed at the time of class certification. Second, as we observed in Castano, 84 F.3d at 746, class certification is often practically dispositive of litigation like the case at bar. If the certification decision is so entangled with the merits as to make interlocutory appeal dispositive of the substantive litigation, it is incidentally but perhaps happily more likely that the legal merit and practical outcome of securities cases will coincide.

We recognize, however, that our ruling on legal merit may not coincide, particularly in the minds of aggrieved former Enron shareholders who have lost billions of dollars in a fraud they allege was aided and abetted by the defendants at bar, with notions of justice and fair play. We acknowledge that the courts' interpretation of § 10(b) could have gone in a different direction and might have established liability for the actions the banks are alleged to have undertaken. Indeed, one of our sister circuits — the Ninth — believes that it did. We have applied the Supreme Court's guidance in ascribing a limited interpretation to the words of § 10, viewing the statute as the result of Congress's balancing of competing desires to provide for some remedy for securities fraud without opening the floodgates for nearly unlimited and frequently unpredictable liability for secondary actors in the securities markets.

In summary, the Affiliated Ute presumption of classwide reliance cannot apply here. Likewise, the district court, albeit with the best of intentions, misapplied the fraud-on-the-market presumption; the facts alleged do not constitute misrepresentations on which an efficient market may be presumed to rely.

Because no class may be certified in a § 10(b) case without a classwide presumption of reliance, our analysis of reliance disposes of this appeal. We decline to address whether, had defendants' actions been misrepresentations on which the market was presumed to rely, they would have been appropriately grouped together as a unitary scheme giving rise to common issues of loss causation among the class members. Likewise, we abstain from addressing the manageability of the district court's plan to implement the proportionate liability provisions of the PSLRA.

The order certifying a class is REVERSED and REMANDED for further-proceedings as appropriate. The motion to stay the trial is DENIED. The mandate shall issue forthwith.