JUSTICE SOUTER delivered the opinion of the Court.

This case turns on the conditions for certifying a mandatory settlement class on a limited fund theory under Federal Rule of Civil Procedure 23(b)(1)(B). We hold that applicants for contested certification on this rationale must show that the fund is limited by more than the agreement of the parties, and has been allocated to claimants belonging within the class by a process addressing any conflicting interests of class members.

I

Like Amchem Products, Inc. v. Windsor, 521 U.S. 591 (1997), this case is a class action prompted by the elephantine mass of asbestos cases, and our discussion in Amchem will suffice to show how this litigation defies customary judicial administration and calls for national legislation. In 1967, one of the first actions for personal asbestos injury was filed in the United States District Court for the Eastern District of Texas against a group of asbestos manufacturers. App. to Pet. for Cert. 252a. In the 1970's and 1980's, plaintiffs' lawyers throughout the country, particularly in East Texas, honed the litigation of asbestos claims to the point of almost mechanical regularity, improving the forensic identification of diseases caused by asbestos, refining theories of liability, and often settling large inventories of cases. See D. Hensler, W. Felstiner, M. Selvin, P. Ebener, Asbestos in the Courts: The Challenge of Mass. Toxic Torts vii (1985); McGovern, Resolving Mature Mass. Tort Litigation, 69 B. U. L. Rev. 659, 660-661 (1989); see also App. to Pet. for Cert. 253a.

Respondent Fibreboard Corporation was a defendant in the 1967 action. Although it was primarily a timber company, from the 1920's through 1971 the company manufactured a variety of products containing asbestos, mainly for high-temperature industrial applications. As the tide of asbestos litigation rose, Fibreboard found itself litigating on two fronts. On one, plaintiffs were filing a stream of personal injury claims against it, swelling throughout the 1980's and 1990's to thousands of new claims for compensatory damages each year. Id., at 265a; App. 1040a. On the second front, Fibreboard was battling for funds to pay its tort claimants. From May 1957 through March, 1959, respondent Continental Casualty Company had provided Fibreboard with a comprehensive general liability policy with limits of $1 million per occurrence, $500,000 per claim, and no aggregate limit. Fibreboard also claimed that respondent Pacific Indemnity Company had insured it from 1956 to 1957 under a similar policy. App. to Pet. for Cert. 267a-268a. Beginning in 1979, Fibreboard was locked in coverage litigation with Continental and Pacific in a California state trial court, which in 1990 held Continental and Pacific responsible for indemnification as to any claim by a claimant exposed to Fibreboard asbestos products prior to their policies' respective expiration dates. Id., at 268a-269a. The decree also required the insurers to pay the full cost of defense for each claim covered. Ibid. The insurance companies appealed.

With asbestos case filings continuing unabated, and its secure insurance assets almost depleted, Fibreboard in 1988 began a practice of "structured settlement," paying plaintiffs 40 percent of the settlement figure up front with the balance contingent upon a successful resolution of the coverage dispute. By 1991, however, the pace of filings forced Fibreboard to start settling cases entirely with the assignments of its rights against Continental, with no initial payment. To reflect the risk that Continental might prevail in the coverage dispute, these assignment agreements generally carried a figure about twice the nominal amount of earlier settlements. Continental challenged Fibreboard's right to make unilateral assignments, but in 1992 a California state court ruled for Fibreboard in that dispute.

Meanwhile, in the aftermath of a 1990 Federal Judicial Center conference on the asbestos litigation crisis, Fibreboard approached a group of leading asbestos plaintiffs' lawyers, offering to discuss a "global settlement" of its asbestos personal-injury liability. Early negotiations bore relatively little fruit, save for the December 1992 settlement by assignment of a significant inventory of pending claims. This settlement brought Fibreboard's deferred settlement obligations to more than $1.2 billion, all contingent upon victory over Continental on the scope of coverage and the validity of the settlement assignments.

In February 1993, after Continental had lost on both issues at the trial level, and thus faced the possibility of practically unbounded liability, it too joined the global settlement negotiations. Because Continental conditioned its part in any settlement on a guarantee of "total peace," ensuring no unknown future liabilities, talks focused on the feasibility of a mandatory class action, one binding all potential plaintiffs and giving none of them any choice to opt out of the certified class. Negotiations continued throughout the spring and summer of 1993, but the difficulty of settling both actually pending and potential future claims simultaneously led to an agreement in early August to segregate and settle an inventory of some 45,000 pending claims, being substantially all those filed by one of the plaintiffs' firms negotiating the global settlement. The settlement amounts per claim were higher than average, with one-half due on closing and the remainder contingent upon either a global settlement or Fibreboard's success in the coverage litigation. This agreement provided the model for settling inventory claims of other firms.

With the insurance companies' appeal of the consolidated coverage case set to be heard on August 27, the negotiating parties faced a motivating deadline, and about midnight before the argument, in a coffee shop in Tyler, Texas, the negotiators finally agreed upon $1.535 billion as the key term of a "Global Settlement Agreement." $1.525 billion of this sum would come from Continental and Pacific, in the proportion established by the California trial court in the coverage case, while Fibreboard would contribute $10 million, all but $500,000 of it from other insurance proceeds, App. 84a. The negotiators also agreed to identify unsettled present claims against Fibreboard and set aside an as-then unspecified fund to resolve them, anticipating that the bulk of any excess left in that fund would be transferred to class claimants. Ahearn v. Fibreboard Corp., 162 F.R.D. 505, 517 (ED Tex. 1995). The next day, as a hedge against the possibility that the Global Settlement Agreement might fail, plaintiffs' counsel insisted as a condition of that agreement that Fibreboard and its two insurers settle the coverage dispute by what came to be known as the "Trilateral Settlement Agreement." The two insurers agreed to provide Fibreboard with funds eventually set at $2 billion to defend against asbestos claimants and pay the winners, should the Global Settlement Agreement fail to win approval. Id., at 517, 521; see also App. to Pet. for Cert. 492a.

On September 9, 1993, as agreed, a group of named plaintiffs filed an action in the United States District Court for the Eastern District of Texas, seeking certification for settlement purposes of a mandatory class comprising three groups: all persons with personal injury claims against Fibreboard for asbestos exposure who had not yet brought suit or settled their claims before the previous August 27; those who had dismissed such a claim but retained the right to bring a future action against Fibreboard; and "past, present and future spouses, parents, children, and other relatives" of class members exposed to Fibreboard asbestos. The class did not include claimants with actions presently pending against Fibreboard or claimants "who filed and, for cash payment or some other negotiated value, dismissed claims against Fibreboard, and whose only retained right is to sue Fibreboard upon development of an asbestos-related malignancy." Id., at 534a-535a. The complaint pleaded personal injury claims against Fibreboard, and, as justification for class certification, relied on the shared necessity of ensuring insurance funds sufficient for compensation. Id., at 552a-569a. After Continental and Pacific had obtained leave to intervene as party-defendants, the District Court provisionally granted class certification, enjoined commencement of further separate litigation against Fibreboard by class members, and appointed a guardian ad litem to review the fairness of the settlement to the class members. See In re Asbestos Litigation, 90 F.3d 963, 972 (CA5 1996).

As finally negotiated, the Global Settlement Agreement provided that in exchange for full releases from class members, Fibreboard, Continental, and Pacific would establish a trust to process and pay class members' asbestos personal injury and death claims. Claimants seeking compensation would be required to try to settle with the trust. If initial settlement attempts failed, claimants would have to proceed to mediation, arbitration, and a mandatory settlement conference. Only after exhausting that process could claimants go to court against the trust, subject to a limit of $500,000 per claim, with punitive damages and prejudgment interest barred. Claims resolved without litigation would be discharged over three years, while judgments would be paid out over a 5- to 10-year period. The Global Settlement Agreement also contained spendthrift provisions to conserve the trust, and provided for paying more serious claims first in the event of a shortfall in any given year. Id., at 973.

After an extensive campaign to give notice of the pending settlement to potential class members, the District Court allowed groups of objectors, including petitioners here, to intervene. After an 8-day fairness hearing, the District Court certified the class and approved the settlement as "fair, adequate, and reasonable" under Rule 23(e). Ahearn, 162 F. R. D., at 527. Satisfied that the requirements of Rule 23(a) were met, id., at 523-526, the District Court certified the class under Rule 23(b)(1)(B), citing the risk that Fibreboard might lose or fare poorly on appeal of the coverage case or lose the assignment-settlement dispute, leaving it without funds to pay all claims. Id., at 526. The "allowance of individual adjudications by class members," the District Court concluded, "would have destroyed the opportunity to compromise the insurance coverage dispute by creating the settlement fund, and would have exposed the class members to the very risks that the settlement addresses." Id., at 527. In response to intervenors' objections that the absence of a "limited fund" precluded certification under Rule 23(b)(1)(B), the District Court ruled that although the subdivision is not so restricted, if it were, this case would qualify. It found both the "disputed insurance asset liquidated by the $1.535 billion Global Settlement," and, alternatively, "the sum of the value of Fibreboard plus the value of its insurance coverage," as measured by the insurance funds' settlement value, to be relevant "limited funds." App. to Pet. for Cert. 491a-492a.

On appeal, the Fifth Circuit affirmed both as to class certification and adequacy of settlement. In re Asbestos Litigation, supra. Agreeing with the District Court's application of Rule 23(a), the Court of Appeals found that there was commonality in class members' shared interest in securing and equitably distributing maximum possible settlement funds, and that the representative plaintiffs were sufficiently typical both in sharing that interest and in basing their claims on the same legal and remedial theories that absent class members might raise. Id., at 975-976. The Fifth Circuit also thought that there were no conflicts of interest sufficiently serious to undermine the adequacy of class counsel's representation. Id., at 976-982. As to Rule 23(b)(1)(B), the court approved the class certification on a "limited fund" rationale based on the threat to "the ability of other members of the class to receive full payment for their injuries from Fibreboard's limited assets." Id. at 982. The Court of Appeals cited expert testimony that Fibreboard faced enormous potential liabilities and defense costs that would likely equal or exceed the amount of damages paid out, and concluded that even combining Fibreboard's value of some $235 million with the $2 billion provided in the Trilateral Settlement Agreement, the company would be unable to pay all valid claims against it within five to nine years. Ibid. Judge Smith dissented, arguing among other things that the majority had skimped on serious due process concerns, had glossed over problems of commonality, typicality, and adequacy of representation, and had ignored a number of justiciability issues. See generally id., at 993-1026.

Shortly thereafter, this Court decided Amchem and proceeded to vacate the Fifth Circuit's judgment and remand for further consideration in light of that decision. 521 U.S. 1114 (1997). On remand, the Fifth Circuit again affirmed, in a brief per curiam opinion, distinguishing Amchem on the grounds that the instant action proceeded under Rule 23(b)(1)(B) rather than (b)(3), and did not allocate awards according to the nature of the claimant's injury. In re Asbestos Litigation, 134 F.3d 668, 669-670 (1998). Again citing the findings on certification under Rule 23(b)(1)(B), the Fifth Circuit affirmed as "incontestable" the District Court's conclusion that the terms of the subdivision had been met. Id., at 670. The Court of Appeals acknowledged Amchem's admonition that settlement class actions may not proceed unless the requirements of Rule 23(a) are met, but noted that the District Court had made extensive findings supporting its Rule 23(a) determinations. Ibid. Judge Smith again dissented, reiterating his previous concerns, and argued specifically that the District Court erred in certifying the class under Rule 23(b)(1)(B) on a "limited fund" theory because the only limited fund in the case was a creature of the settlement itself. Id., at 671-674.

We granted certiorari, 524 U.S. 936 (1998), and now reverse.

II

The nub of this case is the certification of the class under Rule 23(b)(1)(B) on a limited fund rationale, but before we reach that issue, there are two threshold matters. First, petitioners call the class claims nonjusticiable under Article III, saying that this is a feigned action initiated by Fibreboard to control its future asbestos tort liability, with the "vast majority" of the "exposure-only" class members being without injury in fact and hence without standing to sue. Brief for Petitioners 44-50. Ordinarily, of course, this or any other Article III court must be sure of its own jurisdiction before getting to the merits. Steel Co. v. Citizens For Better Environment, 523 U.S. 83, 88-89 (1998). But the class certification issues are, as they were in Amchem, "logically antecedent" to Article III concerns, 521 U.S., at 612, and themselves pertain to statutory standing, which may properly be treated before Article III standing, see Steel Co., supra, at 92. Thus the issue about Rule 23 certification should be treated first, "mindful that [the Rule's] requirements must be interpreted in keeping with Article III constraints. . . ." Amchem, supra, at 612-613.

Petitioners also argue that the Fifth Circuit on remand disregarded Amchem in passing on the Rule 23(a) issues of commonality, typicality, and adequacy of representation. Brief for Petitioners 13-22. We agree that in reinstating its affirmance of the District Court's certification decision, the Fifth Circuit fell short in its attention to Amchem's explanation of the governing legal standards. Two aspects in particular of the District Court's certification should have received more detailed treatment by the Court of Appeals. First, the District Court's enquiry into both commonality and typicality focused almost entirely on the terms of the settlement. See Ahearn, 162 F. R. D., at 524. Second, and more significantly, the District Court took no steps at the outset to ensure that the potentially conflicting interests of easily identifiable categories of claimants be protected by provisional certification of subclasses under Rule 23(c)(4), relying instead on its post hoc findings at the fairness hearing that these subclasses in fact had been adequately represented. As will be seen, however, these points will reappear when we review the certification on the Court of Appeals's "limited fund" theory under Rule 23(b)(1)(B). We accordingly turn directly to that.

III A

Although representative suits have been recognized in various forms since the earliest days of English law, see generally S. Yeazell, From Medieval Group Litigation to the Modern Class Action (1987); see also Marcin, Searching for the Origin of the Class Action, 23 Cath. U. L. Rev. 515, 517-524 (1973), class actions as we recognize them today developed as an exception to the formal rigidity of the necessary parties rule in equity, see Hazard, Gedid, Sowle, An Historical Analysis of the Binding Effect of Class Suits, 146 U. Pa. L. Rev. 1849, 1859-1860 (1998) (hereinafter Hazard, Gedid, Sowle), as well as from the bill of peace, an equitable device for combining multiple suits, see Z. Chafee, Some Problems of Equity 161-167, 200-203 (1950). The necessary parties rule in equity mandated that "all persons materially interested, either as plaintiffs or defendants in the subject matter of the bill ought to be made parties to the suit, however numerous they may be." West v. Randall, 29 F. Cas. 718, 721 (No. 17,424) (CC RI) (1820) (Story, J.). But because that rule would at times unfairly deny recovery to the party before the court, equity developed exceptions, among them one to cover situations "where the parties are very numerous, and the court perceives, that it will be almost impossible to bring them all before the court; or where the question is of general interest, and a few may sue for the benefit of the whole; or where the parties form a part of a voluntary association for public or private purposes, and may be fairly supposed to represent the rights and interests of the whole. . . ." Id., at 722; see J. Story, Commentaries on Equity Pleadings § 97 (J. Gould 10th rev. ed. 1892); F. Calvert, A Treatise upon the Law Respecting Parties to Suits in Equity 17-29 (1837) (hereinafter Calvert, Parties to Suits in Equity). From these roots, modern class action practice emerged in the 1966 revision of Rule 23. In drafting Rule 23(b), the Advisory Committee sought to catalogue in "functional" terms "those recurrent life patterns which call for mass litigation through representative parties." Kaplan, A Prefatory Note, 10 B. C. Ind. Com. L. Rev. 497 (1969).

Rule 23(b)(1)(B) speaks from "a vantage point within the class, [from which the Advisory Committee] spied out situations where lawsuits conducted with individual members of the class would have the practical if not technical effect of concluding the interests of the other members as well, or of impairing the ability of the others to protect their own interests." Kaplan, Continuing Work of the Civil Committee: 1966 Amendments of the Federal Rules of Civil Procedure (I), 81 Harv. L. Rev. 356, 388 (1967) (hereinafter Kaplan, Continuing Work). Thus, the subdivision (read with subdivision (c)(2)) provides for certification of a class whose members have no right to withdraw, when "the prosecution of separate actions . . . would create a risk" of "adjudications with respect to individual members of the class which would as a practical matter be dispositive of the interests of the other members not parties to the adjudications or substantially impair or impede their ability to protect their interests." Fed. Rule Civ. Proc. 23(b)(1)(B). Classic examples of such a risk of impairment may, for example, be found in suits brought to reorganize fraternal-benefit societies, see, e.g., Supreme Tribe of Ben-Hur v. Cauble, 255 U.S. 356 (1921); actions by shareholders to declare a dividend or otherwise to "fix [their] rights," Kaplan, Continuing Work 388; and actions charging "a breach of trust by an indenture trustee or other fiduciary similarly affecting the members of a large class" of beneficiaries, requiring an accounting or similar procedure "to restore the subject of the trust," Advisory Committee's Notes on Fed. Rule Civ. Proc. 23, 28 U.S.C. App., p. 696 (hereinafter Adv. Comm. Notes). In each of these categories, the shared character of rights claimed or relief awarded entails that any individual adjudication by a class member disposes of, or substantially affects, the interests of absent class members.

Among the traditional varieties of representative suit encompassed by Rule 23(b)(1)(B) were those involving "the presence of property which call[ed] for distribution or management," J. Moore J. Friedman, 2 Federal Practice 2240 (1938) (hereinafter Moore Friedman). One recurring type of such suits was the limited fund class action, aggregating "claims . . . made by numerous persons against a fund insufficient to satisfy all claims." Adv. Comm. Notes 697; cf. 1 Newberg § 4.09, at 4-33 ("Classic" limited fund class actions "include claimants to trust assets, a bank account, insurance proceeds, company assets in a liquidation sale, proceeds of a ship sale in a maritime accident suit, and others"). The Advisory Committee cited Dickinson v. Burnham, 197 F.2d 973 (CA2), cert. denied, 344 U.S. 875 (1952), as illustrative of this tradition. In Dickinson, investors hoping to save a failing company had contributed some $600,000, which had been misused until nothing was left but a pool of secret profits on a fraction of the original investment. In a class action, the District Court took charge of this fund, subjecting it to a constructive trust for division among subscribers who demonstrated their claims, in amounts proportional to each class member's percentage of all substantiated claims. 197 F.2d, at 978. The Second Circuit approved the class action and the distribution of the entire pool to claimants, noting that "[a]lthough none of the contributors has been paid in full, no one . . . now asserts or suggests that they should have full recovery . . . as on an ordinary tort liability for conspiracy and defrauding. The court's power of disposition over the fund was therefore absolute and final." Id., at 980. As the Advisory Committee recognized in describing Dickinson, equity required absent parties to be represented, joinder being impractical, where individual claims to be satisfied from the one asset would, as a practical matter, prejudice the rights of absent claimants against a fund inadequate to pay them all.

Equity, of course, recognized the same necessity to bind absent claimants to a limited fund when no formal imposition of a constructive trust was entailed. In Guffanti v. National Surety Co., 196 N.Y. 452, 458, 90 N.E. 174, 176 (1909), for example, the defendant received money to supply steamship tickets and had posted a $15,000 bond as required by state law. He converted to personal use funds collected from more than 150 ticket purchasers, was then adjudged bankrupt, and absconded. One of the defrauded ticket purchasers sued the surety in equity on behalf of himself and all others like him. Over the defendant's objection, the New York Court of Appeals sustained the equitable class suit, citing among other considerations the fact that all recovery had to come from a "limited fund out of which the aggregate recoveries must be sought" that was inadequate to pay all claims, and subject to pro rata distribution. Id., at 458, 90 N.E., at 176. See Hazard, Gedid, Sowle 1915 ("[ Guffanti] explained that when a debtor's assets were less than the total of the creditors' claims, a binding class action was not only permitted but was required; otherwise some creditors (the parties) would be paid and others (the absentees) would not"). See also Morrison v. Warren 174 Misc. 233, 234, 20 N.Y.S.2d 26, 27 (Sup.Ct. N.Y. Cty. 1940) (suit on behalf of more than 400 beneficiaries of an insurance policy following a fire appropriate where "the amount of the claims . . . greatly exceeds the amount of the insurance"); National Surety Co. v. Graves, 211 Ala. 533, 534, 101 So. 190 (1924) (suit against a surety company by stockholders "for the benefit of themselves and all others similarly situate who will join the suit" where it was alleged that individual suits were being filed on surety bonds that "would result in the exhaustion of the penalties of the bonds, leaving many stockholders without remedy").

Ross v. Crary, 1 Paige Ch. 416, 417-418 (N Y Ch. 1829), presents the concept of the limited fund class action in another incarnation. "[D]ivers suits for general legacies," id., at 417, were brought by various legatees against the executor of a decedent's estate. The Ross court stated that where "there is an allegation of a deficiency of the fund, so that an account of the estate is necessary," the court will "direc[t] an account in one cause only" and "stay the proceeding[s] in the others, leaving all the parties interested in the fund, to come in under the decree." Id., at 417-418. Thus, in equity, legatee and creditor bills against the assets of a decedent's estate had to be brought on behalf of all similarly situated claimants where it was clear from the pleadings that the available portion of the estate could not satisfy the aggregate claims against it.

B

The cases forming this pedigree of the limited fund class action as understood by the drafters of Rule 23 have a number of common characteristics, despite the variety of circumstances from which they arose. The points of resemblance are not necessarily the points of contention resolved in the particular cases, but they show what the Advisory Committee must have assumed would be at least a sufficient set of conditions to justify binding absent members of a class under Rule 23(b)(1)(B), from which no one has the right to secede.

The first and most distinctive characteristic is that the totals of the aggregated liquidated claims and the fund available for satisfying them, set definitely at their maximums, demonstrate the inadequacy of the fund to pay all the claims. The concept driving this type of suit was insufficiency, which alone justified the limit on an early feast to avoid a later famine. See, e.g., Guffanti, supra, at 457, 90 N.E., at 176 ("The total amount of the claims exceeds the penalty of the bond. . . .A just and equitable payment from the bond would be a distribution pro rata upon the amount of the several embezzlements. Unless in a case like this the amount of the bond is so distributed among the persons having claims which are secured thereby, it must necessarily result in a scramble for precedence in payment, and the amount of the bond may be paid to the favored, or to those first obtaining knowledge of the embezzlements"); Graves, supra, at 534, 101 So., at 190 ("The primary equity of the bill is the adjustment of claims and the equitable apportionment of a fund provided by law, which is insufficient to pay claimants in full"). The equity of the limitation is its necessity.

Second, the whole of the inadequate fund was to be devoted to the overwhelming claims. See, e.g., Dickinson, 197 F.2d, at 979-980 (rejecting a challenge by holder of funds to the court's disposition of the entire fund); see also United States v. Butterworth-Judson Corp., 269 U.S. 504, 513 (1926) ("Here, the fund being less than the debts, the creditors are entitled to have all of it distributed among them according to their rights and priorities"). It went without saying that the defendant or estate or constructive trustee with the inadequate assets had no opportunity to benefit himself or claimants of lower priority by holding back on the amount distributed to the class. The limited fund cases thus ensured that the class as a whole was given the best deal; they did not give a defendant a better deal than seriatim litigation would have produced.

Third, the claimants identified by a common theory of recovery were treated equitably among themselves. The cases assume that the class will comprise everyone who might state a claim on a single or repeated set of facts, invoking a common theory of recovery, to be satisfied from the limited fund as the source of payment. Each of the people represented in Ross, for example, had comparable entitlement as a legatee under the testator's will. Those subject to representation in Dickinson had a common source of claims in the solicitation of funds by parties whose subsequent defalcation left them without their investment, while in Guffanti the individuals represented had each entrusted money for ticket purchases. In these cases the hope of recovery was limited, respectively, by estate assets, the residuum of profits, and the amount of the bond. Once the represented classes were so identified, there was no question of omitting anyone whose claim shared the common theory of liability and would contribute to the calculated shortfall of recovery. See Railroad Co. v. Orr, 18 Wall. 471, 474 (1873) (reciting the "well settled" general rule "that when it appears on the face of the bill that there will be a deficiency in the fund, and that there are other creditors or legatees who are entitled to a ratable distribution with the complainants, and who have a common interest with them, such creditors or legatees should be made parties to the bill, or the suit should be brought by the complainants in behalf of themselves and all others standing in a similar situation"). The plaintiff appeared on behalf of all similarly situated parties, see Calvert, Parties to Suits in Equity 24 ("[I]t is not sufficient that the plaintiff appear on behalf of numerous parties: the rule seems to be, that he must appear on behalf of all who are interested"); thus, the creditors' bill was brought on behalf of all creditors, cf. Leigh v. Thomas, 2 Ves. Sen. 312, 313, 28 Eng. Rep. 201 (Ch. 1751) ("No doubt but a bill may be by a few creditors in behalf of themselves and the rest . . . but there is no instance of a bill by three or four to have an account of the estate, without saying they bring it in behalf of themselves and the rest of the creditors"), the constructive trust was asserted on behalf of all victims of the fraud, and the surety suit was brought on behalf of all entitled to a share of the bond. Once all similar claims were brought directly or by representation before the court, these antecedents of the mandatory class action presented straightforward models of equitable treatment, with the simple equity of a pro rata distribution providing the required fairness, see 1 J. Pomeroy, Equity Jurisprudence § 407, pp. 764-765 (4th ed. 1918) ("[I]f the fund is not sufficient to discharge all claims upon it in full . . . equity will incline to regard all the demands as standing upon an equal footing, and will decree a pro rata distribution or payment").

In sum, mandatory class treatment through representative actions on a limited fund theory was justified with reference to a "fund" with a definitely ascertained limit, all of which would be distributed to satisfy all those with liquidated claims based on a common theory of liability, by an equitable, pro rata distribution.

C

The Advisory Committee, and presumably the Congress in approving subdivision (b)(1)(B), must have assumed that an action with these characteristics would satisfy the limited fund rationale cognizable under that subdivision. The question remains how far the same characteristics are necessary for limited fund treatment. While we cannot settle all the details of a subdivision (b)(1)(B) limited fund here (and so cannot decide the ultimate question whether settlements of multitudes of related tort actions are amenable to mandatory class treatment), there are good reasons to treat these characteristics as presumptively necessary, and not merely sufficient, to satisfy the limited fund rationale for a mandatory action. At the least, the burden of justification rests on the proponent of any departure from the traditional norm.

It is true, of course, that the text of Rule 23(b)(1)(B) is on its face open to a more lenient limited fund concept, just as it covers more historical antecedents than the limited fund. But the greater the leniency in departing from the historical limited fund model, the greater the likelihood of abuse in ways that will be apparent when we apply the limited fund criteria to the case before us. The prudent course, therefore, is to presume that when subdivision (b)(1)(B) was devised to cover limited fund actions, the object was to stay close to the historical model. As will be seen, this limiting construction finds support in the Advisory Committee's expressions of understanding, minimizes potential conflict with the Rules Enabling Act, and avoids serious constitutional concerns raised by the mandatory class resolution of individual legal claims, especially where a case seeks to resolve future liability in a settlement-only action.

To begin with, the Advisory Committee looked cautiously at the potential for creativity under Rule 23(b)(1)(B), at least in comparison with Rule 23(b)(3). Although the Committee crafted all three subdivisions of the Rule in general, practical terms, without the formalism that had bedeviled the original Rule 23, see Kaplan, Continuing Work 380-386, the Committee was consciously retrospective with intent to codify pre-Rule categories under Rule 23(b)(1), not forward-looking as it was in anticipating innovations under Rule 23(b)(3). Compare Civil Rules Advisory Committee Meeting, Oct. 31-Nov. 2, 1963, Congressional Information Service Records of the U.S. Judicial Conference, Committee on Rules of Practice and Procedure 1935-1988, No. CI-7104-53, p. 11 (hereinafter Civil Rules Meeting) (comments of Reporter Kaplan) (Rule 23(b)(3) represents "the growing point of the law"); id., at 16 (comments of Committee Member Prof. Albert M. Sacks) (Rule 23(b)(3) is "an evolving area"). Thus, the Committee intended subdivision (b)(1) to capture the "`standard'" class actions recognized in pre-Rule practice, Kaplan, Continuing Work 394.

Consistent with its backward look under subdivision (b)(1), as commentators have pointed out, it is clear that the Advisory Committee did not contemplate that the mandatory class action codified in subdivision (b)(1)(B) would be used to aggregate unliquidated tort claims on a limited fund rationale. See Monaghan, Antisuit Injunctions and Preclusion Against Absent Nonresident Class Members, 98 Colum. L. Rev. 1148, 1164 (1998) ("The `framers' of Rule 23 did not envision the expansive interpretations of the rule that have emerged. . . .No draftsmen contemplated that, in mass torts, (b)(1)(B) `limited fund' classes would emerge as the functional equivalent to bankruptcy by embracing `funds' created by the litigation itself"); see also Schwarzer, Settlement of Mass. Tort Class Actions: Order Out of Chaos, 80 Cornell L. Rev. 837, 840 (1995) ("The original concept of the limited fund class does not readily fit the situation where a large volume of claims might eventually result in judgments that in the aggregate could exceed the assets available to satisfy them"); Marcus, They Can't Do That, Can They? Tort Reform Via Rule 23, 80 Cornell L. Rev. 858, 877 (1995). None of the examples cited in the Advisory Committee Notes or by Professor Kaplan in explaining Rule 23(b)(1)(B) remotely approach what was then described as a "mass accident" case. While the Advisory Committee focused much attention on the amenability of Rule 23(b)(3) to such cases, the Committee's debates are silent about resolving tort claims under a mandatory limited fund rationale under Rule 23(b)(1)(B). It is simply implausible that the Advisory Committee, so concerned about the potential difficulties posed by dealing with mass tort cases under Rule 23(b)(3), with its provisions for notice and the right to opt out, see Rule 23(c)(2), would have uncritically assumed that mandatory versions of such class actions, lacking such protections, could be certified under Rule 23(b)(1)(B). We do not, it is true, decide the ultimate question whether Rule 23(b)(1)(B) may ever be used to aggregate individual tort claims, cf. Ticor Title Ins. Co. v. Brown, 511 U.S. 117, 121 (1994) (per curiam). But we do recognize that the Committee would have thought such an application of the Rule surprising, and take this as a good reason to limit any surprise by presuming that the Rule's historical antecedents identify requirements.

The Rules Enabling Act underscores the need for caution. As we said in Amchem, no reading of the Rule can ignore the Act's mandate that "rules of procedure `shall not abridge, enlarge or modify any substantive right,'" Amchem, 521 U.S., at 613 (quoting 28 U.S.C. § 2072(b)); cf. Guaranty Trust Co. v. York, 326 U.S. 99, 105 (1945) ("In giving federal courts `cognizance' of equity suits in cases of diversity jurisdiction, Congress never gave, nor did the federal courts ever claim, the power to deny substantive rights created by State law or to create substantive rights denied by State law"). Petitioners argue that the Act has been violated here, asserting that the Global Settlement Agreement's priorities of claims and compromise of full recovery abrogated the state law that must govern this diversity action under 28 U.S.C. § 1652. See Brief for Petitioners 31-36. Although we need not grapple with the difficult choice-of-law and substantive state-law questions raised by petitioners' assertion, we do need to recognize the tension between the limited fund class action's pro rata distribution in equity and the rights of individual tort victims at law. Even if we assume that some such tension is acceptable under the Rules Enabling Act, it is best kept within tolerable limits by keeping limited fund practice under Rule 23(b)(1)(B) close to the practice preceding its adoption.

Finally, if we needed further counsel against adventurous application of Rule 23(b)(1)(B), the Rules Enabling Act and the general doctrine of constitutional avoidance would jointly sound a warning of the serious constitutional concerns that come with any attempt to aggregate individual tort claims on a limited fund rationale. First, the certification of a mandatory class followed by settlement of its action for money damages obviously implicates the Seventh Amendment jury trial rights of absent class members. We noted in Ross v. Bernhard, 396 U.S. 531 (1970), that since the merger of law and equity in 1938, it has become settled among the lower courts that "class action plaintiffs may obtain a jury trial on any legal issues they present." Id., at 541. By its nature, however, a mandatory settlement-only class action with legal issues and future claimants compromises their Seventh Amendment rights without their consent.

Second, and no less important, mandatory class actions aggregating damages claims implicate the due process "principle of general application in Anglo-American jurisprudence that one is not bound by a judgment in personam in a litigation in which he is not designated as a party or to which he has not been made a party by service of process," Hansberry v. Lee, 311 U.S. 32, 40 (1940), it being "our `deep-rooted historic tradition that everyone should have his own day in court,'" Martin v. Wilks, 490 U.S. 755, 762 (1989) (quoting 18 C. Wright, A. Miller, E. Cooper, Federal Practice and Procedure § 4449, p. 417 (1981)); see Richards v. Jefferson County, 517 U.S. 793, 798-799 (1996). Although "`[w]e have recognized an exception to the general rule when, in certain limited circumstances, a person, although not a party, has his interests adequately represented by someone with the same interests who is a party," or "where a special remedial scheme exists expressly foreclosing successive litigation by nonlitigants, as for example in bankruptcy or probate," Martin, supra, at 762, n. 2 (citations omitted), the burden of justification rests on the exception.

The inherent tension between representative suits and the day-in-court ideal is only magnified if applied to damages claims gathered in a mandatory class. Unlike Rule 23(b)(3) class members, objectors to the collectivism of a mandatory subdivision (b)(1)(B) action have no inherent right to abstain. The legal rights of absent class members (which in a class like this one would include claimants who by definition may be unidentifiable when the class is certified) are resolved regardless of either their consent, or, in a class with objectors, their express wish to the contrary. And in settlement-only class actions the procedural protections built into the Rule to protect the rights of absent class members during litigation are never invoked in an adversarial setting, see Amchem, supra, at 620.

In related circumstances, we raised the flag on this issue of due process more than a decade ago in Phillips Petroleum Co. v. Shutts, 472 U.S. 797 (1985). Shutts was a state class action for small sums of interest on royalty payments suspended on the authority of a federal regulation. Id., at 800. After certification of the class, the named plaintiffs notified each member by first-class mail of the right to opt out of the lawsuit. Out of a class of 33,000, some 3,400 exercised that right, and another 1,500 were excluded because their notices could not be delivered. Id., at 801. After losing at trial, the defendant, Phillips Petroleum, argued that the state court had no jurisdiction over claims of out-of-state plaintiffs without their affirmative consent. We said no and held that out-of-state plaintiffs could not invoke the same due process limits on personal jurisdiction that out-of-state defendants had under International Shoe Co. v. Washington, 326 U.S. 310 (1945), and its progeny. 472 U.S., at 806-808. But we also saw that before an absent class member's right of action was extinguishable due process required that the member "receive notice plus an opportunity to be heard and participate in the litigation," and we said that "at a minimum . . . an absent plaintiff [must] be provided with an opportunity to remove himself from the class." Id., at 812.

IV

The record on which the District Court rested its certification of the class for the purpose of the global settlement did not support the essential premises of mandatory limited fund actions. It failed to demonstrate that the fund was limited except by the agreement of the parties, and it showed exclusions from the class and allocations of assets at odds with the concept of limited fund treatment and the structural protections of Rule 23(a) explained in Amchem.

A

The defect of certification going to the most characteristic feature of a limited fund action was the uncritical adoption by both the District Court and the Court of Appeals of figures agreed upon by the parties in defining the limits of the fund and demonstrating its inadequacy. When a district court, as here, certifies for class action settlement only, the moment of certification requires "heightene[d] attention," Amchem, 521 U.S., at 620, to the justifications for binding the class members. This is so because certification of a mandatory settlement class, however provisional technically, effectively concludes the proceeding save for the final fairness hearing. And, as we held in Amchem, a fairness hearing under Rule 23(e) is no substitute for rigorous adherence to those provisions of the Rule "designed to protect absentees," ibid., among them subdivision (b)(1)(B). Thus, in an action such as this the settling parties must present not only their agreement, but evidence on which the district court may ascertain the limit and the insufficiency of the fund, with support in findings of fact following a proceeding in which the evidence is subject to challenge, see In re Bendectin Products Liability Litigation, 749 F.2d 300, 306 (CA6 1984) ("[T]he district court, as a matter of law, must have a fact-finding inquiry on this question and allow the opponents of class certification to present evidence that a limited fund does not exist"); see also In re Temple, 851 F.2d 1269, 1272 (CA11 1988) ("Without a finding as to the net worth of the defendant, it is difficult to see how the fact of a limited fund could have been established given that all of [the defendant's] assets are potentially available to suitors"); In re Dennis Greenman Securities Litigation, 829 F.2d 1539, 1546 (CA11 1987) (discussing factual findings necessary for certification of a limited fund class action).

We have already alluded to the difficulties facing limited fund treatment of huge numbers of actions for unliquidated damages arising from mass torts, the first such hurdle being a computation of the total claims. It is simply not a matter of adding up the liquidated amounts, as in the models of limited fund actions. Although we might assume, arguendo, that prior judicial experience with asbestos claims would allow a court to make a sufficiently reliable determination of the probable total, the District Court here apparently thought otherwise, concluding that "there is no way to predict Fibreboard's future asbestos liability with any certainty." 162 F. R. D., at 528. Nothing turns on this conclusion, however, since there was no adequate demonstration of the second element required for limited fund treatment, the upper limit of the fund itself, without which no showing of insufficiency is possible.

The "fund" in this case comprised both the general assets of Fibreboard and the insurance assets provided by the two policies, see 90 F.3d, at 982 (describing the fund as Fibreboard's entire equity and $2 billion in insurance assets under the Trilateral Settlement Agreement). As to Fibreboard's assets exclusive of the contested insurance, the District Court and the Fifth Circuit concluded that Fibreboard had a then-current sale value of $235 million that could be devoted to the limited fund. While that estimate may have been conservative, at least the District Court heard evidence and made an independent finding at some point in the proceedings. The same, however, cannot be said for the value of the disputed insurance.

The insurance assets would obviously be "limited" in the traditional sense if the total of demonstrable claims would render the insurers insolvent, or if the policies provided aggregate limits falling short of that total; calculation might be difficult, but the way to demonstrate the limit would be clear. Neither possibility is presented in this case, however. Instead, any limit of the insurance asset here had to be a product of potentially unlimited policy coverage discounted by the risk that Fibreboard would ultimately lose the coverage dispute litigation. This sense of limit as a value discounted by risk is of course a step removed from the historical model, but even on the assumption that it would suffice for limited fund treatment, there was no adequate finding of fact to support its application here. Instead of undertaking an independent evaluation of potential insurance funds, the District Court (and, later, the Court of Appeals), simply accepted the $2 billion Trilateral Settlement Agreement figure as representing the maximum amount the insurance companies could be required to pay tort victims, concluding that "[w]here insurance coverage is disputed, it is appropriate to value the insurance asset at a settlement value." App. to Pet. for Cert. 492a.

Settlement value is not always acceptable, however. One may take a settlement amount as good evidence of the maximum available if one can assume that parties of equal knowledge and negotiating skill agreed upon the figure through arms-length bargaining, unhindered by any considerations tugging against the interests of the parties ostensibly represented in the negotiation. But no such assumption may be indulged in this case, or probably in any class action settlement with the potential for gigantic fees. In this case, certainly, any assumption that plaintiffs' counsel could be of a mind to do their simple best in bargaining for the benefit of the settlement class is patently at odds with the fact that at least some of the same lawyers representing plaintiffs and the class had also negotiated the separate settlement of 45,000 pending claims, 90 F.3d, at 969-970, 971, the full payment of which was contingent on a successful Global Settlement Agreement or the successful resolution of the insurance coverage dispute (either by litigation or by agreement, as eventually occurred in the Trilateral Settlement Agreement), id., at 971, n. 3; App. 119a-120a. Class counsel thus had great incentive to reach any agreement in the global settlement negotiations that they thought might survive a Rule 23(e) fairness hearing, rather than the best possible arrangement for the substantially unidentified global settlement class. Cf. Cramton, Individualized Justice, Mass. Torts, and "Settlement Class Actions": An Introduction, 80 Cornell L. Rev. 811, 832 (1995) ("[S]ide settlements suggest that class counsel has been laboring under an impermissible conflict of interest and that it may have preferred the interests of current clients to those of the future claimants in the settlement class"). The resulting incentive to favor the known plaintiffs in the earlier settlement was, indeed, an egregious example of the conflict noted in Amchem resulting from divergent interests of the presently injured and future claimants. See 521 U.S., at 626-627 (discussing adequacy of named representatives under Rule 23(a)(4)).

We do not, of course, know exactly what an independent valuation of the limit of the insurance assets would have shown. It might have revealed that even on the assumption that Fibreboard's coverage claim was sound, there would be insufficient assets to pay claims, considered with reference to their probable timing; if Fibreboard's own assets would not have been enough to pay the insurance shortfall plus any claims in excess of policy limits, the projected insolvency of the insurers and Fibreboard would have indicated a truly limited fund. (Nothing in the record, however, suggests that this would have been a supportable finding.) Or an independent valuation might have revealed assets of insufficient value to pay all projected claims if the assets were discounted by the prospects that the insurers would win the coverage cases. Or the court's independent valuation might have shown, discount or no discount, the probability of enough assets to pay all projected claims, precluding certification of any mandatory class on a limited fund rationale. Throughout this litigation the courts have accepted the assumption that the third possibility was out of the question, and they may have been right. But objecting and unidentified class members alike are entitled to have the issue settled by specific evidentiary findings independent of the agreement of defendants and conflicted class counsel.

B

The explanation of need for independent determination of the fund has necessarily anticipated our application of the requirement of equity among members of the class. There are two issues, the inclusiveness of the class and the fairness of distributions to those within it. On each, this certification for settlement fell short.

The definition of the class excludes myriad claimants with causes of action, or foreseeable causes of action, arising from exposure to Fibreboard asbestos. While the class includes those with present claims never filed, present claims withdrawn without prejudice, and future claimants, it fails to include those who had previously settled with Fibreboard while retaining the right to sue again "upon development of an asbestos related malignancy," plaintiffs with claims pending against Fibreboard at the time of the initial announcement of the Global Settlement Agreement, and the plaintiffs in the "inventory" claims settled as a supposedly necessary step in reaching the global settlement, see 90 F.3d, at 971. The number of those outside the class who settled with a reservation of rights may be uncertain, but there is no such uncertainty about the significance of the settlement's exclusion of the 45,000 inventory plaintiffs and the plaintiffs in the unsettled present cases, estimated by the Guardian Ad Litem at more than 53,000 as of August 27, 1993, see App. in No. 95-40635 (CA5), 6 Record, Tab 55, p. 72 (Report of the Guardian Ad Litem). It is a fair question how far a natural class may be depleted by prior dispositions of claims and still qualify as a mandatory limited fund class, but there can be no question that such a mandatory settlement class will not qualify when in the very negotiations aimed at a class settlement, class counsel agree to exclude what could turn out to be as much as a third of the claimants that negotiators thought might eventually be involved, a substantial number of whom class counsel represent, see App. to Pet. for Cert. 321a (noting that the parties negotiating the global settlement agreed to use a negotiating benchmark of 186,000 future claims against Fibreboard).

Might such class exclusions be forgiven if it were shown that the class members with present claims and the outsiders ended up with comparable benefits? The question is academic here. On the record before us, we cannot speculate on how the unsettled claims would fare if the global settlement were approved, or under the trilateral settlement. As for the settled inventory claims, their plaintiffs appeared to have obtained better terms than the class members. They received an immediate payment of 50 percent of a settlement higher than the historical average, and would get the remainder if the global settlement were sustained (or the coverage litigation resolved, as it turned out to be by the Trilateral Settlement Agreement); the class members, by contrast, would be assured of a 3-year payout for claims settled, whereas the unsettled faced a prospect of mediation followed by arbitration as prior conditions of instituting suit, which would even then be subject to a recovery limit, a slower payout, and the limitations of the trust's spendthrift protection. See supra, at 827. Finally, as discussed below, even ostensible parity between settling nonclass plaintiffs and class members would be insufficient to overcome the failure to provide the structural protection of independent representation as for subclasses with conflicting interests.

On the second element of equity within the class, the fairness of the distribution of the fund among class members, the settlement certification is likewise deficient. Fair treatment in the older cases was characteristically assured by straightforward pro rata distribution of the limited fund. See supra, at 841. While equity in such a simple sense is unattainable in a settlement covering present claims not specifically proven and claims not even due to arise, if at all, until some future time, at the least such a settlement must seek equity by providing for procedures to resolve the difficult issues of treating such differently situated claimants with fairness as among themselves.

First, it is obvious after Amchem that a class divided between holders of present and future claims (some of the latter involving no physical injury and attributable to claimants not yet born) requires division into homogeneous subclasses under Rule 23(c)(4)(B), with separate representation to eliminate conflicting interests of counsel. See Amchem, 521 U.S., at 627 (class settlements must provide "structural assurance of fair and adequate representation for the diverse groups and individuals affected"); cf. 5 J. Moore, T. Chorvat, D. Feinberg, R. Marmer, J. Solovy, Moore's Federal Practice § 23.25[5][e], p. 23-149 (3d ed. 1998) (an attorney who represents another class against the same defendant may not serve as class counsel). As we said in Amchem, "for the currently injured, the critical goal is generous immediate payments," but "[t]hat goal tugs against the interest of exposure-only plaintiffs in ensuring an ample, inflation-protected fund for the future." 521 U.S. 626. No such procedure was employed here, and the conflict was as contrary to the equitable obligation entailed by the limited fund rationale as it was to the requirements of structural protection applicable to all class actions under Rule 23(a)(4).

Second, the class included those exposed to Fibreboard's asbestos products both before and after 1959. The date is significant, for that year saw the expiration of Fibreboard's insurance policy with Continental, the one provided the bulk of the insurance funds for the settlement. Pre-1959 claimants accordingly had more valuable claims than post-1959 claimants, see 90 F.3d, at 1012-1013 (Smith, J., dissenting), the consequence being a second instance of disparate interests within the certified class. While at some point there must be an end to reclassification with separate counsel, these two instances of conflict are well within the requirement of structural protection recognized in Amchem.

It is no answer to say, as the Fifth Circuit said on remand, that these conflicts may be ignored because the settlement makes no disparate allocation of resources as between the conflicting classes. See 134 F.3d, at 669-670. The settlement decides that the claims of the immediately injured deserve no provisions more favorable than the more speculative claims of those projected to have future injuries, and that liability subject to indemnification is no different from liability with no indemnification. The very decision to treat them all the same is itself an allocation decision with results almost certainly different from the results that those with immediate injuries or claims of indemnified liability would have chosen.

Nor does it answer the settlement's failures to provide structural protections in the service of equity to argue that the certified class members' common interest in securing contested insurance funds for the payment of claims was so weighty as to diminish the deficiencies beneath recognition here. See Brief for Respondent Class Representatives Ahearn et al. 31 (discussing this issue in the context of the Rule 23(a)(4) adequacy of representation requirement); id., at 35-36 (citing, e.g., In re "Agent Orange" Product Liability Litigation, 996 F.2d 1425, 1435-1436 (CA2 1993); In re "Agent Orange" Product Liability Litigation, 800 F.2d 14, 18-19 (CA2 1986)). This argument is simply a variation of the position put forward by the proponents of the settlement in Amchem, who tried to discount the comparable failure in that case to provide separate representatives for subclasses with conflicting interests, see Brief for Petitioners in Amchem Products, Inc. v. Windsor, O. T. 1996, No. 96-270, p. 48 (arguing that "achieving a global settlement" was "an overriding concern that all plaintiffs [held] in common"); see also id., at 42 (arguing that the requirement of Rule 23(b)(3) that there be predominance of common questions of law or fact had been met by shared interest in "the fairness of the settlement"). The current position is just as unavailing as its predecessor in Amchem. There we gave the argument no weight, see 521 U.S., at 625-628, observing that "[t]he benefits asbestos-exposed persons might gain from the establishment of a grand-scale compensation scheme is a matter fit for legislative consideration," but the determination whether "proposed classes are sufficiently cohesive to warrant adjudication" must focus on "questions that preexist any settlement," id., at 622-623. Here, just as in the earlier case, the proponents of the settlement are trying to rewrite Rule 23; each ignores the fact that Rule 23 requires protections under subdivisions (a) and (b) against inequity and potential inequity at the precertification stage, quite independently of the required determination at postcertification fairness review under subdivision (e) that any settlement is fair in an overriding sense. A fairness hearing under subdivision (e) can no more swallow the preceding protective requirements of Rule 23 in a subdivision (b)(1)(B) action than in one under subdivision (b)(3).

C

A third contested feature of this settlement certification that departs markedly from the limited fund antecedents is the ultimate provision for a fund smaller than the assets understood by the Court of Appeals to be available for payment of the mandatory class members' claims; most notably, Fibreboard was allowed to retain virtually its entire net worth. Given our treatment of the two preceding deficiencies of the certification, there is of course no need to decide whether this feature of the agreement would alone be fatal to the Global Settlement Agreement. To ignore it entirely, however, would be so misleading that we have decided simply to identify the issue it raises, without purporting to resolve it at this time.

Fibreboard listed its supposed entire net worth as a component of the total (and allegedly inadequate) assets available for claimants, but subsequently retained all but $500,000 of that equity for itself. On the face of it, the arrangement seems irreconcilable with the justification of necessity in denying any opportunity for withdrawal of class members whose jury trial rights will be compromised, whose damages will be capped, and whose payments will be delayed. With Fibreboard retaining nearly all its net worth, it hardly appears that such a regime is the best that can be provided for class members. Given the nature of a limited fund and the need to apply its criteria at the certification stage, it is not enough for a District Court to say that it "need not ensure that a defendant designate a particular source of its assets to satisfy the class' claims; [but only that] the amount recovered by the class [be] fair." A hearn, 162 F. R. D., at 527.

The District Court in this case seems to have had a further point in mind, however. One great advantage of class action treatment of mass tort cases is the opportunity to save the enormous transaction costs of piecemeal litigation, an advantage to which the settlement's proponents have referred in this case. Although the District Court made no specific finding about the transaction cost saving likely from this class settlement, estimating the amount in the "hundreds of millions," id., at 529, it did conclude that the amount would exceed Fibreboard's net worth as the Court valued it, ibid. (Fibreboard's net worth of $235 million "is considerably less than the likely savings in defense costs under the Global Settlement"). If a settlement thus saves transaction costs that would never have gone into a class member's pocket in the absence of settlement, may a credit for some of the savings be recognized in a mandatory class action as an incentive to settlement? It is at least a legitimate question, which we leave for another day.

V

Our decision rests on a different basis from the ground of JUSTICE BREYER's dissent, just as there was a difference in approach between majority and dissenters in Amchem. The nub of our position is that we are bound to follow Rule 23 as we understood it upon its adoption, and that we are not free to alter it except through the process prescribed by Congress in the Rules Enabling Act. Although, as the dissent notes, post, at 882, the revised text adopted in 1966 was understood (somewhat cautiously) to authorize the courts to provide for class treatment of mass tort litigation, it was also the Court's understanding that the Rule's growing edge for that purpose would be the opt-out class authorized by subdivision (b)(3), not the mandatory class under subdivision (b)(1)(B), see supra, at 843-844. While we have not ruled out the possibility under the present Rule of a mandatory class to deal with mass tort litigation on a limited fund rationale, we are not free to dispense with the safeguards that have protected mandatory class members under that theory traditionally.

Apart from its effect on the requirements of subdivision (a) as explained and held binding in Amchem, the dissent would move the standards for mandatory actions in the direction of opt-out class requirements by according weight to this "unusual limited fund['s] . . . witching hour," post, at 877, in exercising discretion over class certification. It is on this belief (that we should sustain the allowances made by the District Court in consideration of the exigencies of this settlement proceeding) that the dissent addresses each of the criteria for limited fund treatment (demonstrably insufficient fund, intraclass equity, and dedication of the entire fund, see post, at 873-883).

As to the calculation of the fund, the dissent believes an independent valuation by the District Court may be dispensed with here in favor of the figure agreed upon by the settling parties. The dissent discounts the conflicts on the part of class counsel who negotiated the Global Settlement Agreement by arguing that the " relevant" settlement negotiation, and hence the relevant benchmark for judging the actual value of the insurance amount, was the negotiation between Fibreboard and the insurers that produced the Trilateral Settlement Agreement. See post, at 876. This argument, however, minimizes two facts: (1) that Fibreboard and the insurers made this separate, backup agreement only at the insistence of class counsel as a condition for reaching the Global Settlement Agreement; (2) even more important, that "[t]he Insurers were . . . adamant that they would not agree to pay any more in the context of a backup agreement than in a global agreement," a principle "Fibreboard acceded to" on the day the Global Settlement Agreement was announced "as the price of permitting an agreement to be reached with respect to a global settlement," Ahearn, 162 F. R. D., at 516. Under these circumstances the reliability of the Trilateral Settlement Agreement's figure is inadequate as an independent benchmark that might excuse the want of any independent judicial determination that the Global Settlement Agreement's fund was the maximum possible. In any event, the dissent says, it is not crucial whether a $30 claim has to settle for $15 or $20. But it is crucial. Conflict-free counsel, as required by Rule 23(a) and Amchem, might have negotiated a $20 figure, and a limited fund rationale for mandatory class treatment of a settlement-only action requires assurance that claimants are receiving the maximum fund, not a potentially significant fraction less.

With respect to the requirement of intraclass equity, the dissent argues that conflicts both within this certified class and between the class as certified and those excluded from it may be mitigated because separate counsel were simply not to be had in the short time that a settlement agreement was possible before the argument (or likely decision) in the coverage case. But this is to say that when the clock is about to strike midnight, a court considering class certification may lower the structural requirements of Rule 23(a) as declared in Amchem, and the parallel equity requirements necessary to justify mandatory class treatment on a limited fund theory.

Finally, the dissent would excuse Fibreboard's retention of virtually all its net worth, and the loss to members of the certified class of some 13 percent of the fund putatively available to them, on the ground that the settlement made more money available than any other effort would likely have done. But even if we could be certain that this evaluation were true, this is to reargue Amchem: the settlement's fairness under Rule 23(e) does not dispense with the requirements of Rules 23(a) and (b).

We believe that if an allowance for exigency can make a substantial difference in the level of Rule 23 scrutiny, the economic temptations at work on counsel in class actions will guarantee enough exigencies to take the law back before Amchem and unsettle the line between mandatory class actions under subdivision (b)(1)(B) and opt-out actions under subdivision (b)(3).

VI

In sum, the applicability of Rule 23(b)(1)(B) to a fund and plan purporting to liquidate actual and potential tort claims is subject to question, and its purported application in this case was in any event improper. The Advisory Committee did not envision mandatory class actions in cases like this one, and both the Rules Enabling Act and the policy of avoiding serious constitutional issues counsel against leniency in recognizing mandatory limited fund actions in circumstances markedly different from the traditional paradigm. Assuming, arguendo, that a mandatory, limited fund rationale could under some circumstances be applied to a settlement class of tort claimants, it would be essential that the fund be shown to be limited independently of the agreement of the parties to the action, and equally essential under Rules 23(a) and (b)(1)(B) that the class include all those with claims unsatisfied at the time of the settlement negotiations, with intraclass conflicts addressed by recognizing independently represented subclasses. In this case, the limit of the fund was determined by treating the settlement agreement as dispositive, an error magnified by the representation of class members by counsel also representing excluded plaintiffs, whose settlements would be funded fully upon settlement of the class action on any terms that could survive final fairness review. Those separate settlements, together with other exclusions from the claimant class, precluded adequate structural protection by subclass treatment, which was not even afforded to the conflicting elements within the class as certified.

The judgment of the Court of Appeals, accordingly, is reversed, and the case is remanded for further proceedings consistent with this opinion.

It is so ordered.