In what may be the largest class action ever attempted in federal court, the district court in this case embarked “on a road certainly less traveled, if ever taken at all,” Castano v. American Tobacco Co., 160 F.R.D. 544, 560 (E.D.La.1995) (citing EDWARD C. Latham, The Poetry of Robert Frost, “The Road Not Taken” 105 (1969)), and entered a class certification order. The court defined the class as:

(a) All nicotine-dependent persons in the United States ... who have purchased and smoked cigarettes manufactured by the defendants;

(b) the estates, representatives, and administrators of these nicotine-dependent cigarette smokers; and

(e) the spouses, children, relatives and “significant others” of these nicotine-dependent cigarette smokers as their heirs or survivors.

Id. at 560-61. The plaintiffs limit the claims to years since 1943. ^[1]

This matter comes before us on interlocutory appeal, under 28 U.S.C. § 1292(b), of the class certification order. Concluding that district court abused its discretion in certifying the class, we reverse.

I.

A. The Class Complaint

The plaintiffs ^[2] filed this class complaint against the defendant tobacco companies ^[3] and the Tobacco Institute, Inc., seeking compensation solely for the injury of nicotine addiction. The gravamen of their complaint is the novel and wholly untested theory that the defendants fraudulently failed to inform consumers that nicotine is addictive and manipulated the level of nicotine in cigarettes to sustain their addictive nature. The class complaint alleges nine causes of action: fraud and deceit, negligent misrepresentation, intentional infliction of emotional distress, negligence and negligent infliction of emotional distress, violation of state consumer protection statutes, breach of express warranty, breach of implied warranty, strict product liability, and redhibition pursuant to the Louisiana Civil Code.

The plaintiffs seek compensatory ^[4] and punitive damages ^[5] and attorneys’ fees. ^[6] In[*738] addition, the plaintiffs seek equitable relief for fraud and deceit, negligent misrepresentation, violation of consumer protection statutes, and breach of express and implied warranty. The equitable remedies include a declaration that defendants are financially responsible for notifying all class members of nicotine’s addictive nature, a declaration that the defendants manipulated nicotine levels with the intent to sustain the addiction of plaintiffs and the class members, an order that the defendants disgorge any profits made from the sale of cigarettes, restitution for sums paid for cigarettes, and the establishment of a medical monitoring fund.

The plaintiffs initially defined the class as “all nicotine dependent persons in the United States,” including current, former and deceased smokers since 1943. Plaintiffs conceded that addiction would have to be proven by each class member; the defendants argued that proving class membership will require individual mini-trials to determine whether addiction actually exists.

In response to the district court’s inquiry, the plaintiffs proposed a four-phase trial plan. ^[7] In phase 1, a jury would determine common issues of “core liability.” Phase 1 issues would include ^[8] (1) issues of law and fact relating to defendants’ course of conduct, fraud, and negligence liability (including duty, standard of care, misrepresentation and concealment, knowledge, intent); (2) issues of law and fact relating to defendants’ alleged conspiracy and concert of action; (3) issues of fact relating to the addictive nature/dependency creating characteristics and properties of nicotine; (4) issues of fact relating to nicotine cigarettes as defective products; (5) issues of fact relating to whether defendants’ wrongful conduct was intentional, reckless or negligent; (6) identifying which defendants specifically targeted their advertising and promotional efforts to particular groups (e.g. youths, minorities, etc.); (7) availability of a presumption of reliance; (8) whether defendants’ misrepresentations/suppression of fact and/or of addictive properties of nicotine preclude availability of a “personal choice” defense; (9) defendants’ liability for actual damages, and the categories of such damages; (10) defendants’ liability for emotional distress damages; and (11) defendants’ liability for punitive damages.

Phase 1 would be followed by notice of the trial verdict and claim forms to class members. In phase 2, the jury would determine compensatory damages in sample plaintiff eases. The jury then would establish a ratio of punitive damages to compensatory damages, which ratio thereafter would apply to each class member.

Phase 3 would entail a complicated procedure to determine compensatory damages for individual class members. The trial plan envisions determination of absent class members’ compensatory economic and emotional distress damages on the basis of claim forms, “subject to verification techniques and assertion of defendants’ affirmative defenses under grouping, sampling, or representative procedures to be determined by the Court.”

The trial plan left open how jury trials on class members’ personal injury/wrongful death claims would be handled, but the trial plan discussed the possibility of bifurcation. In phase 4, the court would apply the punitive damage ratio based on individual damage awards and would conduct a review of the reasonableness of the award.

B. The Class Certification Order

Following extensive briefing, the district court granted, in part, plaintiffs’ motion for class certification, concluding that the prerequisites of Fed.R.Civ.P. 23(a) had been met. ^[9] The court rejected certification, under[*739] Fed.R.Civ.P. 23(b)(2), of the plaintiffs’ claim for equitable relief, including the claim for medical monitoring. 160 F.R.D. at 552. Ap-pellees have not cross-appealed that portion of the order.

The court did grant the plaintiffs’ motion to certify the class under Fed.R.Civ.P. 23(b)(3), ^[10] organizing the class action issues into four categories: (1) core liability; (2) injury-in-fact, proximate cause, reliance and affirmative defenses; (3) compensatory damages; and (4) punitive damages. Id. at 553-58. It then analyzed each category to determine whether it met the predominance and superiority requirements of rule 23(b)(3). Using its power to sever issues for certification under Fed.R.Civ.P. 23(c)(4), the court certified the class on core liability and punitive damages, and certified the class conditionally pursuant to Fed.R.Civ.P. 23(c)(1).

1. Core Liability Issues

The court defined core liability issues as “common factual issues [of] whether defendants knew cigarette smoking was addictive, failed to inform cigarette smokers of such; and took actions to addict cigarette smokers. Common legal issues include fraud, negligence, breach of warranty (express or implied), strict liability, and violation of consumer protection statutes.” 160 F.R.D. at 553..

The court found that the predominance requirement of rule 23(b)(3) was satisfied for the core liability issues. Without any specific analysis regarding the multitude of issues that make up “core liability,” the court found that under Jenkins v. Raymark Indus., 782 F.2d 468 (5th Cir.1986), common issues predominate because resolution of core liability issues would significantly advance the individual cases. The court did not discuss why “core liability” issues would be a significant, rather than just common, part of each individual trial, nor why the individual issues in the remaining categories did not predominate over the common “core liability” issues.

The only specific analysis on predominance analysis was on the plaintiffs’ fraud claim. The court determined that it would be premature to hold that individual reliance issues predominate over common issues. Relying on Eisen v. Carlisle & Jacquelin, 417 U.S. 156, 94 S.Ct. 2140, 40 L.Ed.2d 732 (1974), the court stated that it could not inquire into the merits of the plaintiffs’ claim to determine whether reliance would be an issue in individual trials. 160 F.R.D. at 554. Moreover, the court, recognized the possibility that under state law, reliance can be inferred when a fraud claim is based on an omission. Accordingly, the court was convinced that it could certify the class and defer the consideration of how reliance would affect predominance.

The court also deferred substantial consideration of how variations in state law would affect predominance. Relying on two district court opinions, ^[11] the court concluded that issues of fraud, breach of warranty, negligence, intentional tort, and strict liability do not vary so much from state to state as to cause individual issues to predominate. The court noted that any determination of how state law variations affect predominance was premature, as the court had yet to make a choice of law determination. As for the consumer protection claims, the court also deferred analysis of state law variations, because “there has been no showing that the[*740] consumer protection statutes differ so much as to make individual issues predominate.” Id.

The court also concluded that a class action is superior to other methods for adjudication of the core liability issues. Relying heavily on Jenkins, the court noted that having this common issue litigated in a class action was superior to repeated trials of the same evidence. Recognizing serious problems with manageability, it determined that such- problems were outweighed by “the specter of thousands, if not millions, of similar trials of liability proceeding in thousands of courtrooms around the nation.” Id. at 555-56.

2. Injury-in-faet, Proximate Cause, Reliance, Affirmative Defenses, and Compensatory Damages

Using the same methodology as it did for the core liability issues, the district court refused to certify the issues of injury-in-fact, proximate cause, reliance, affirmative defenses, and compensatory damages, concluding that the “issues are so overwhelmingly replete with individual circumstances that they quickly outweigh predominance and superiority.” Id. at 556. Specifically, the court found that whether a person suffered emotional injury from addiction, whether his addiction was caused by the defendants’ actions, whether he relied on the defendants’ misrepresentations, and whether affirmative defenses unique to each class member precluded reeoveiy were all individual issues. As to compensatory damages and the claim for medical monitoring, the court concluded that such claims were so intertwined with proximate cause and affirmative defenses that class certification would not materially advance the individual cases.

3. Punitive Damages

In certifying punitive damages for class treatment, the court adopted the plaintiffs’ trial plan for punitive damages: The class jury would develop a ratio of punitive damages to actual damages, and the court would apply that ratio in individual cases. As it did with the core liability issues, the court determined that variations in state law, including differing burdens of proof, did not preclude certification. Rather than conduct an independent review of predominance or superiority, the court relied on Jenkins and on Watson v. Shell Oil Co., 979 F.2d 1014 (5th Cir.1992), vacated for rehearing en banc, 990 F.2d 805 (5th Cir.1993), appeal dismissed, 53 F.3d 663 (5th Cir.1994), for support of its certification order. ^[12]

II.

A district court must conduct a rigorous analysis of the rule 23 prerequisites before certifying a class. General Tel. Co. v. Falcon, 457 U.S. 147, 161, 102 S.Ct. 2364, 2372, 72 L.Ed.2d 740 (1982); Applewhite v. Reichhold Chems., 67 F.3d 571, 573 (5th Cir.1995). The decision to certify is within the broad discretion of the court, but that discretion must be exercised within the framework of rule 23. Gulf Oil Co. v. Bernard, 452 U.S. 89, 100, 101 S.Ct. 2193, 2200, 68 L.Ed.2d 693 (1981). The party seeking certification bears the burden of proof. Horton v. Goose Creek. Ind. Sch. Dist., 690 F.2d 470, 486 (5th Cir.1982), ce rt. denied, 463 U.S. 1207, 103 S.Ct. 3536, 77 L.Ed.2d 1387 (1983); In re American Medical Sys., 75 F.3d 1069, 1086 (6th Cir.1996) (concluding that district court reversed the proper burden of proof by asking defendants to show cause why the court should not certify the class).

The district court erred in its analysis in two distinct ways. First, it failed to consider how variations in state law affect predominance and superiority. Second, its predominance inquiry did not include consideration of how a trial on the merits would be conducted.

Each of these defects mandates reversal. Moreover, at this time, while the tort is immature, the class complaint must be dis[*741] missed, as class certification cannot be found to be a superior method of adjudication. ^[13]

A. Variations in State Law

Although rule 28(c)(1) requires that a class should be certified “as soon as practicable” and allows a court to certify a conditional class, it does not follow that the rule’s requirements are lessened when the class is conditional. As a sister circuit explained:

Conditional certification is not a means whereby the District Court can avoid deciding whether," at that time, the requirements of the Rule have been substantially met. The purpose of conditional certification is to preserve the Court’s power to revoke certification in those cases wherein the magnitude or complexity of the litigation may eventually reveal problems not theretofore apparent. But in this case the District Court seemed to brush aside one of the requirements of Rule 23(b)(3) by stating that at this time “analysis of the individual versus common questions would be for the Court to act as a seer.” However difficult it may have been for the District Court to decide whether common questions predominate over individual questions, it should not have sidestepped this preliminary requirement of the Rule by merely stating that the problem of indi-, vidual questions “lies far beyond the horizon in the realm of speculation.”

In re Hotel Tel. Charges, 500 F.2d 86, 90 (9th Cir.1974).

In a multi-state class action, variations in state law may swamp any common issues and defeat predominance. See Georgine v. Amchem Prods., 83 F.3d 610, 618 (3d Cir.1996) (decertifying class' because legal and factual differences in the plaintiffs’ claims “when exponentially magnified by choice of law considerations, eclipse any common issues in this case”); American Medical Sys., 75 F.3d at 1085 (granting mandamus in a multi-state products liability action, in part because “[t]he district court ... failed to consider how the law of negligence differs from jurisdiction to jurisdiction”).

Accordingly, a district court must consider how variations in state law affect predominance and superiority. Walsh v. Ford Motor Co., 807 F.2d 1000 (D.C.Cir.1986) (Ruth Bader Ginsburg, J.), cert. denied, 482 U.S. 915, 107 S.Ct. 3188, 96 L.Ed.2d 677 (1987). The Walsh court rejected the notion that a district court may defer considering variations in state law:

Appellees see the “which law” matter as academic. They say no variations in state warranty laws relevant to this case exist. A court cannot accept such an assertion “on faith.” Appellees, as class action proponents, must show that it is accurate. We have made no inquiry of our own on this score and, for the current purpose, simply note the general unstartling statement made in a leading treatise: “The Uniform Commercial Code is not uniform.”

Id. at 1016-17 (footnotes omitted).

A district court’s duty to determine whether the plaintiff has borne its burden on class certification requires that a court consider variations in state law when a class action involves multiple jurisdictions. “In order to make the findings required to certify a class action under Rule 23(b)(3) ... one must initially identify the substantive law issues which will control the outcome of the litigation.” Alabama v. Blue Bird Body Co., 573 F.2d 309, 316 (5th Cir.1978).

A requirement that a court know which law will apply before making a predominance determination is especially important when there may be differences in state law. See In re Rhone-Poulenc Rorer, Inc. (“Rhone-Poulenc ”), 51 F.3d 1293, 1299-1302 (7th Cir.) (mandamus) (comparing differing state pattern instructions on negligence and differing formulations of the meaning of negligence), cer t. denied, — U.S. -, 116 S.Ct. 184, 133 L.Ed.2d 122 (1995); In re [*742] “Agent Orange” Prod. Liability Litig., 818 F.2d 145, 165 (2d Cir.1987) (noting possibility of differences in state products liability law), cert. denied, 484 U.S. 1004, 108 S.Ct. 695, 98 L.Ed.2d 647 (1988). Given the plaintiffs’ burden, a court cannot rely on assurances of counsel that any problems with predominance or superiority can be overcome. Windham v. American Brands, Inc., 565 F.2d 59, 70 (4th Cir.1977), cert. denied, 435 U.S. 968, 98 S.Ct. 1605, 56 L.Ed.2d 58 (1978).

The able opinion in School Asbestos demonstrates what is required from a district court when variations in state law exist. There, the court affirmed class certification, despite variations in state law, because:

To meet the problem of diversity in applicable state law, class plaintiffs have undertaken an extensive analysis of the variances in products liability among the jurisdictions. That review separates the law into four categories. Even assuming additional permutations and combinations, plaintiffs have made a creditable showing, which apparently satisfied the district court, that class certification does not present insuperable obstacles. Although we have some doubt on this score, the effort may nonetheless prove successful.

789 F.2d at 1010; see also Georgine, 83 F.3d at 627 & n. 13 (distinguishing School Asbestos because it involved few individualized questions, and class counsel had made a credible argument that the applicable law of the different states could be categorized into four patterns); Walsh, 807 F.2d at 1017 (holding that “nationwide class action mov-ants must creditably demonstrate, through an ‘extensive analysis’ of state law variances, ‘that class certification does not present insuperable obstacles’ ”).

A thorough review of the record demonstrates that, in this case, the district court did not properly consider how variations in state law affect predominance. The court acknowledged as much in its order granting class certification, for, in declining to make a choice of law determination, it noted that “[t]he parties have only briefly addressed the conflict of laws issue in this matter.” 160 F.R.D. at 554. Similarly, the court stated that “there has been no showing that the consumer protection statutes differ so much as to make individual issues predominate.” Id. 14

The district court’s review of state law variances can hardly be considered extensive; it conducted a cursory review of state law variations and gave short shrift to the defendants’ arguments concerning variations. In response to the defendants’ extensive analysis of how state law varied on fraud, products liability, affirmative defenses, negligent infliction of emotional distress, consumer protection statutes, and punitive damages, ^[15] the court examined a sample phase 1[*743] jury interrogatory and verdict form, a survey of medical monitoring decisions, a survey of consumer fraud class actions, and a survey of punitive damages law in the defendants’ home states. The court also relied on two district court opinions granting certification in multi-state class actions.

The district court’s consideration of state law variations was inadequate. The surveys provided by the plaintiffs failed to discuss, in any meaningful way, how the court could deal with variations in state law. The consumer fraud survey simply quoted a few state courts that had certified state class actions. The survey of punitive damages was limited to the defendants’ home states. Moreover, the two district court opinions on which the court relied did not support the proposition that variations in state law could be ignored. ^[16] Nothing in the record demonstrates that the court critically analyzed how variations in state law would affect predominance.

The court also failed to perform its duty to determine whether the class action would be manageable in light of state law variations. The court’s only discussion of manageability[*744] is a citation to Jenkins and the claim that “[w]hile manageability of the liability issues in this case may well prove to be difficult, the Court finds that any such difficulties pale in comparison to the specter of thousands, if not millions, of similar trials of liability proceeding in thousands of courtrooms around the nation.” Id. at 555-56.

The problem with this approach is that it substitutes case-specific analysis with a generalized reference to Jenkins. The Jenkins court, however, was not faced with managing a novel claim involving eight causes of action, multiple jurisdictions, millions of plaintiffs, eight defendants, and over fifty years of alleged wrongful conduct. Instead, Jenkins involved only 893 personal injury asbestos cases, the law of only one state, and the prospect of trial occurring in only one district. Accordingly, for purposes of the instant case, Jenkins is largely inapposite.

In summary, whether the specter of millions of cases outweighs any manageability problems in this class is uncertain when the scope of any manageability problems is unknown. Absent considered judgment on the manageability of the class, a comparison to millions of individual trials is meaningless.

B. Predominance

The district court’s second error was that it failed to consider how the plaintiffs’ addiction claims would be tried, individually, or on a class basis. See 160 F.R.D. at 554. The district court, based on Eisen v. Carlisle & Jacquelin, 417 U.S. 156, 177-78, 94 S.Ct. 2140, 2152-53, 40 L.Ed.2d 732 (1974), and Miller v. Mackey Int'l, 452 F.2d 424 (5th Cir.1971), believed that it could not go past the pleadings for the certification decision. The result was an incomplete and inadequate predominance inquiry.

The crux of the court’s error was that it misinterpreted Eisen and Miller. Neither case suggests that a court is limited to the pleadings when deciding on certification. Both, instead, stand for the unremarkable proposition that the strength of a plaintiffs claim should not affect the certification decision. In Eisen, the Court held that it was improper to make a preliminary inquiry into the merits of a case, determine that the plaintiff was likely to succeed, and consequently shift the cost of providing notice to the defendant. 417 U.S. at 177, 94 S.Ct. at 2152. In Miller, this court held that a district court could not deny certification based on its belief that the plaintiff could not prevail on the merits. 452 F.2d at 427.

A district court certainly may look past the pleadings to determine whether the requirements of rule 23 have been met. ^[17] Going beyond the pleadings is necessary, as a court must understand the claims, defenses, relevant facts, and applicable substantive law in order to make a meaningful determination of the certification issues. See Manual FOR Complex Litigation § 30.11 (3d ed. 1995).

The district court’s predominance inquiry demonstrates why such an understanding is necessary. The premise of the court’s opinion is a citation to Jenkins and a conclusion that class treatment of common issues would significantly advance the individual trials.[*745] Absent knowledge of how addietion-as-injury cases would actually be tried, however, it was impossible for the court to know whether the common issues would be a “significant” portion of the individual trials. The court just assumed that because the common issues would play a part in every trial, they must be significant. ^[18] The court’s synthesis of Jenkins and Eisen would write the predominance requirement out of the rule, and any common issue would predominate if it were common to all the individual trials. ^[19]

The court’s treatment of the fraud claim also demonstrates the error inherent in its approach. ^[20] According to both the advisory committee’s notes to Rule 23(b)(3) and this court’s decision in Simon v. Merrill Lynch, Pierce, Fenner & Smith, Inc., 482 F.2d 880 (5th Cir.1973), a fraud class action cannot be certified when individual reliance will be an issue. The district court avoided the reach of this court’s decision in Simon by an erroneous reading of Eisen; the court refused to consider whether reliance would be an issue in individual trials.

The problem with the district court’s approach is that after the class trial, it might have decided that reliance must be proven in individual trials. The court then would have been faced with the difficult choice of decerti-fying the class after phase 1 and wasting judicial resources, or continuing with a class action that would have failed the predominance requirement of rule 23(b)(3). ^[21]

[*746] III.

In addition to the reasons given above, regarding the district court’s procedural errors, this class must be decertified because it independently fails the superiority requirement of rule 23(b)(3). In the context of mass tort class actions, certification dramatically affects the stakes for defendants. Class certification magnifies and strengthens the number of unmeritorious claims. Agent Orange, 818 F.2d at 165-66. Aggregation of claims also makes it more likely that a defendant will be found liable and results in significantly higher damage awards. MaNual por Complex Litigation § 33.26 n. 1056; Kenneth S. Bordens and Irwin A. Horowitz, Mass Tort Civil Litigation: The Impact of Procedural Changes on Jury Decisions, 73 JudicatüRE 22 (-1989).

In addition to skewing trial outcomes, class certification creates insurmountable pressure on defendants to settle, whereas individual trials would not. See Peter H. Schuck, Mass Torts: An Institutional Evolutionist Perspective, 80 CORNELL L.Rev. 941, 958 (1995). The risk of facing an all-or-nothing verdict presents too high a risk, even when the probability of an adverse judgment is low. Rhone-Poulenc, 51 F.3d at 1298. These settlements have been referred to as judicial blackmail. ^[22]

It is no surprise then, that historically, certification of mass tort litigation classes has been disfavored. ^[23] The traditional concern[*747] over the rights of defendants in mass tort class actions is magnified in the instant case. Our specific concern is that a mass tort cannot be properly certified without a prior track record of trials from which the district court can draw the information necessary to make the predominance and superiority analysis required by rule 23. This is because certification of an immature tort results in a higher than normal risk that the class action may not be superior to individual adjudication.

We first address the district court’s superiority analysis. The court acknowledged the extensive manageability problems with this class. Such problems include difficult choice of law determinations, subclassing of eight claims with variations in state law, Erie guesses, notice to millions of class members, further subclassing to take account of transient plaintiffs, and the difficult procedure for determining who is nicotine-dependent. Cases with far fewer manageability problems have given courts pause. See, e.g., Georgine, 83 F.3d at 632; In re Hotel Tel., 500 F.2d at 90.

The district court’s rationale for certification in spite of such problems — i.e., that a class trial would preserve judicial resources in the millions of inevitable individual trials— is based on pure speculation. Not every mass tort is asbestos, and not every mass tort will result in the same judicial crises. ^[24] The judicial crisis to which the district court referred is only theoretical.

What the district court failed to consider, and what no court can determine at this time, is the very real possibility that the judicial crisis may fail to materialize. ^[25] The plain[*748] tiffs’ claims are based on a new theory of liability and the existence of new evidence. Until plaintiffs decide to file individual claims, a court cannot, from the existence of injury, presume that all or even any plaintiffs will pursue legal remedies. ^[26] Nor can a court make a superiority determination based on such speculation. American Medical Sys., 75 F.3d at 1085 (opining that superiority is lacking where judicial management crisis does not exist and individual trials are possible).

Severe manageability problems and the lack of a judicial crisis are not the only reasons why superiority is lacking. The most compelling rationale for finding superiority in a class action — the existence of a negative value suit — is missing in this case. Accord Phillips Petroleum Co. v. Shutts, 472 U.S. 797, 809, 105 S.Ct. 2965, 2973, 86 L.Ed.2d 628 (1985); Rhone-Poulenc, 51 F.3d at 1299.

As he stated in the record, plaintiffs’ counsel in this ease has promised to inundate the courts with individual claims if class certification is denied. Independently of the reliability of this self-serving promise, there is reason to believe that individual suits are feasible. First, individual damage claims are high, and punitive damages are available in most states. The expense of litigation does not necessarily turn this ease into a negative value suit, in part because the prevailing party may recover attorneys’ fees under many consumer protection statutes. See Boggs v. Alto Trailer Sales, 511 F.2d 114, 118 (5th Cir.1975) (acknowledging that the availability of attorneys’ fees is a common basis for finding non-superiority).

In a ease such as this one, where each plaintiff may receive a large award, and fee shifting often is available, we find Chief Judge Posner’s analysis of superiority to be persuasive:

For this consensus or maturing of judgment the district judge proposes to substitute a single trial before a single jury.... One jury ... will hold the fate of an industry in the palm of its hand.... That kind of thing can happen in our system of civil justice.... But it need not be tolerated when the alternative exists of submitting an issue to multiple juries constituting in the aggregate a much larger and more diverse sample of decision-makers. That would not be a feasible option if the stakes to each class member were too slight to repay the cost of suit.... But this is not the case.... Each plaintiff if successful is apt to receive a judgment in the millions. With the aggregate stakes in the tens or hundreds of millions of dollars, or even in the billions, it is not a waste of judicial resources to conduct more than one trial, before more than six jurors, to determine whether a major segment of the international pharmaceutical industry is to follow the asbestos manufacturers into Chapter 11.

Rhone-Poulenc, 51 F.3d at 1300. So too here, we cannot say that it would be a waste to allow individual trials to proceed, before a district court engages in the complicated predominance and superiority analysis necessary to certify a class.

Fairness may demand that mass torts with few prior verdicts or judgments be litigated first in smaller units — even single-plaintiff, single-defendant trials — until general causation, typical injuries, and levels of damages become established. Thus, “mature” mass torts like asbestos or Daikon Shield may call for procedures that are not appropriate for incipient mass tort eases, such as those involving injuries arising[*749] from new products, chemical substances, or pharmaceuticals.

Manual foe Complex Litigation § 38.26.

The remaining rationale for superiority— judicial efficiency ^[27] — is also lacking. In the context of an immature tort, any savings in judicial resources is speculative, and any imagined savings would be overwhelmed by the procedural problems that certification of a sui generis cause of action brings with it.

Even assuming arguendo that the tort system will see many more addiction-as-injury claims, a conclusion that certification will save judicial resources is premature at this stage of the litigation. Take for example the district court’s plan to divide core liability from other issues such as comparative negligence and reliance. The assumption is that after a class verdict, the common issues will not be a part of follow-up trials. The court has no basis for that assumption.

It may be that comparative negligence will be raised in the individual trials, and the evidence presented at the class trial will have to be repeated. The same may be true for reliance. ^[28] The net result may be a waste, not a savings, in judicial resources. Only after the courts have more experience with this type of case can a court certify issues in a way that preserves judicial resources. See Jenkins, 782 F.2d 468 (certifying state of the art defense because experience had demonstrated that judicial resources could be saved by certification).

Even assuming that certification at this time would result in judicial efficiencies in individual trials, certification of an immature tort brings with it unique problems that may consume more judicial resources than certification will save. These problems are not speculative; the district court faced, and ignored, many of the problems that immature torts can cause.

The primary procedural difficulty created by immature torts is the inherent difficulty a district court will have in determining whether the requirements of rule 23 have been met. We have already identified a number of defects with the district court’s predominance and manageability inquires, defects that will continue to exist on remand because of the unique nature of the plaintiffs’ claim.

The district court’s predominance inquiry, or lack of it, squarely presents the problems associated with certification of immature torts. Determining whether the common issues are a “significant” part of each individual case has an abstract quality to it when no court in this country has ever tried an injury-as-addiction claim. As the plaintiffs admitted to the district court, “we don’t have the learning curb [sic] that is necessary to say to Your Honor ‘this is precisely how this case can be tried and that will not run afoul of the teachings of the 5th Circuit.’ ”

Yet, an accurate finding on predominance is necessary before the court can certify a class. It may turn out that the defendant’s conduct, while common, is a minor part of each trial. Premature certification deprives the defendant of the opportunity to present that argument to any court and risks decerti-fication after considerable resources have been expended.

The court’s analysis of reliance also demonstrates the potential judicial inefficiencies in immature tort class actions. Individual trials will determine whether individual reliance will be an issue. Rather than guess that reliance may be inferred, a district court should base its determination that individual reliance does not predominate on the wisdom of such individual trials. The risk that a district court will make the wrong guess, that the parties will engage in years of litigation, and that the class ultimately will be decerti-fied (because reliance predominates over common issues) prevents this class action from being a superior method of adjudication.

The complexity of the choice of law inquiry also makes individual adjudication superior[*750] to class treatment. The plaintiffs have asserted eight theories of liability from every state. Prior to certification, the district court must determine whether variations in state law defeat predominance. While the task may not be impossible, its complexity certainly makes individual trials a more attractive alternative and, ipso facto, renders class treatment not superior. See Georgine, 83 F.3d at 634 (recommending that Congress solve the problems inherent in multi-state class actions by federalizing choice of law rules, but rejecting, such legislation when it masquerades as judicial innovation).

Through individual adjudication, the plaintiffs can winnow their claims to the strongest causes of action. ^[29] The result will be an easier choice of law inquiry and a less complicated predominance inquiry. State courts can address the more novel of the plaintiffs’ claims, making the federal court’s Erie guesses less complicated. It is far more desirable to allow state courts to apply and develop their own law than to have a federal court apply “a kind of Esperanto [jury] instruction.” Rhone-Poulenc, 51 F.3d at 1300; MaNual for Complex Litigation § 33.26 (discussing the full cycle of litigation necessary for a tort to mature).

The full development of trials in every state will make subclassing an easier process. The result of allowing individual trials to proceed is a more accurate determination of predominance. We have already seen the result of certifying this class without individual adjudications, and we are not alone in expressing .discomfort with a district court’s certification of a novel theory. See Rhone-Poulenc, 51 F.3d at 1300.

Another factor weighing heavily in favor of individual trials is the risk that in order to make this class action manageable, the court will be forced to bifurcate issues in violation of the Seventh Amendment. This class action is permeated with individual issues, such as proximate causation, comparative negligence, reliance, and compensatory damages. In order to manage so many individual issues, the district court proposed to empanel a class jury to adjudicate common issues. A second jury, or a number of “second” juries, will pass on the individual issues, either on a case-by-case basis or through group trials of individual plaintiffs.

The Seventh Amendment entitles parties to have fact issues decided by one jury, and prohibits a second jury from reexamining those facts and issues. ^[30] Thus, Constitution allows bifurcation of issues that are so separable that the second jury will not be called upon to reconsider findings of fact by the first:

[T]his Court has cautioned that separation of issues is not the usual course that should be followed, and that the issue to be tried must be so distinct and separable from the others that a trial of it alone may be had without injustice. This limitation on the use of bifurcation is a recognition of the fact that inherent in the Seventh Amendment guarantee of a trial by jury is the general right of a litigant to have only one jury pass on a common issue of fact. The Supreme Court recognized this principle in Gasoline Products [Co., Inc. v. Champlin Refining Co., 283 U.S. 494, 51 S.Ct. 513, 75 L.Ed. 1188 (1931) ].... The Court explained ... that a partial new trial may not be “properly resorted to unless it clearly appears that the issue to be retried is so distinct and separable from the others that a trial of it alone may be had without injustice.” Such a rule is dictated for the very practical reason that if separate juries are allowed to pass on issues involving overlapping legal and factual questions the[*751] verdicts rendered by each jury could be inconsistent.

Alabama v. Blue Bird Body Co., 573 F.2d 309, 318 (5th Cir.1978) (citations and footnotes omitted).

The Seventh Circuit recently addressed Seventh Amendment limitations to bifurcation. In Rhone-Poulenc, 51 F.3d at 1302-03, Chief Judge Posner described the constitutional limitation as one requiring a court to “carve at the joint” in such a way so that the same issue is not reexamined by different juries. “The right to a jury trial ... is a right to have juriable issues determined by the first jury impaneled to hear them (provided there are no errors warranting a new trial), and not reexamined by another finder of fact.” Id. at 1303.

Severing a defendant’s conduct from comparative negligence results in the type of risk that our court forbade in Blue Bird. Comparative negligence, by definition, requires a comparison between the defendant’s and the plaintiffs conduct. Rhone-Poulenc, 51 F.3d at 1303 (“Comparative negligence entails, as the name implies, a comparison of the degree of negligence of plaintiff and defendant.”). At a bare minimum, a second jury will rehear evidence of the defendant’s conduct. There is a risk that in apportioning fault, the second jury could reevaluate the defendant’s fault, determine that the defendant was not at fault, and apportion 100% of the fault to the plaintiff. In such a situation, the second jury would be impermissibly reconsidering the findings of a first jury. The risk of such reevaluation is so great that class treatment can hardly be said to be superior to individual adjudication. ^[31]

The plaintiffs’ final retort is that individual trials are inadequate because time is running out for many of the plaintiffs. ^[32] They point out that prior litigation against the tobacco companies has taken up to ten years to wind through the legal system. While a compelling rhetorical argument, it is ultimately inconsistent with the plaintiffs’ own arguments and ignores the realities of the legal system. First, the plaintiffs’ reliance on prior personal injury cases is unpersuasive, as they admit that they have new evidence and are pursuing a claim entirely different from that of past plaintiffs.

Second, the plaintiffs’ claim that time is running out ignores the reality of the class action device. In a complicated case involving multiple jurisdictions, the conflict of law question itself could take decades to work its way through the courts. ^[33] Once that issue has been resolved, discovery, subclassing, and ultimately the class trial would take place. Next would come the appellate process. After the class trial, the individual trials and appeals on comparative negligence and damages would have to take place. The net result could be that the class action device would lengthen, not shorten, the time it takes for the plaintiffs to reach final judgment.

[*752] IV.

The district court abused its discretion by ignoring variations in state law and how a trial on the alleged causes of action would be tried. Those errors cannot be corrected on remand because of the novelty of the plaintiffs’ claims. Accordingly, class treatment is not superior to individual adjudication.

We have once before stated that “traditional ways of proceeding reflect far more than habit. They reflect the very culture of the jury trial....” In re Fibreboard Corp., 893 F.2d 706, 711 (5th Cir.1990). The collective wisdom of individual juries is necessary before this court commits the fate of an entire industry or, indeed, the fate of a class of millions, to a single jury. For the forgoing reasons, we REVERSE and REMAND with instructions that the district court dismiss the class complaint.