Plaintiff-appellant Gary Bryant appeals from the decision of the district court granting summary judgment in favor of defendant-appellee Ford Motor Company. We conclude that the district court lacked jurisdiction over this action because of the presence of Doe defendants at the time of removal from state court.

Bryant initiated this action for negligence, breach of warranty, and strict liability in California state court against Ford and Does 1 through 50. Ford removed the action to the United States District Court for the Central District of California based upon diversity of citizenship.1 Bryant did not object to removal. The district court took no action with respect to the Doe defendants.

Bryant seeks recovery for injuries he sustained in an accident while driving a Ford van for United Parcel Service on March 1, 1983. Bryant contends that the passive restraint system in the van was defective because it did not include a shoulder harness. Bryant's complaint alleges that Does 1 through 50 are related to each other and to Ford as "agents, servants, employees and/or joint venturers." Bryant claims that Ford and each of the Doe defendants were involved in the design, production, inspection, and distribution of the van which Bryant was driving at the time of the accident.

A joint inspection of the van by the parties on May 10, 1984 revealed that Ford had manufactured only the chassis of the van. The body and other components, including the passive restraint system, were produced by other companies as part of a joint venture. The companies responsible for producing the component parts could not be identified at the time of removal or the time of inspection because the van was produced in 1968 and Ford destroyed records containing this information after ten years.

Ford moved for summary judgment. In opposing Ford's motion, Bryant noted that he planned to name Doe defendants as soon as he discovered their identities. The district court nonetheless granted summary judgment in favor of Ford, concluding that there were no material facts supporting Ford's liability because of the inspection evidence that Ford was not involved in the production of the passive restraint system.2 Bryant then moved the court to add City Ford Company, the seller of the van, General Seating and Sash Company, the producer of the seats, and Grumman-Olson Company, the producer of the body, to the action and to remand the case to state court. City Ford and Grumman-Olson are California corporations. The district court denied Bryant's motion, finding that the presence of non-diverse parties was not new evidence justifying relief from judgment under Fed.R.Civ.P. 60(b) (1982).

Bryant appealed the grant of summary judgment. We granted a limited remand at Bryant's request for the district court to again reconsider its previous rulings. The district court again refused to join the additional parties, and this appeal of the district court's rulings followed.

Applying Ninth Circuit law, a panel of this court then held that because the Doe defendants in the complaint were real but unidentified people or entities, the district court could not determine whether they would defeat diversity jurisdiction once identified. Bryant v. Ford Motor Co., 794 F.2d 450, 453 (9th Cir.1986). The panel remanded the case to the district court with instructions to remand to the appropriate state court. Id. After a petition for rehearing was filed, the panel requested en banc consideration of this case in order to clarify Doe defendant law in the Ninth Circuit. For the reasons set forth below, we now remand this case to the district court with instructions to remand to the appropriate state court.

California law allows a plaintiff to sue any potential defendant whose name is unknown under a fictitious name (commonly as a Doe defendant). Cal.Civ.Proc.Code Sec. 474 (West 1979).3 A plaintiff who names a Doe defendant in his complaint and alleges that the defendant's true name is unknown has three years from the commencement of the action in which to discover the identity of the Doe defendant, to amend the complaint accordingly, and to effect service of the complaint. Cal.Civ.Proc.Code Sec. 581a (West 1976).4

Up to this point, the general rule in the Ninth Circuit has been that the naming of Doe defendants defeats diversity jurisdiction and, therefore, that district courts should remand cases containing allegations against Doe defendants to state court. See, e.g., Othman v. Globe Indem. Co., 759 F.2d 1458, 1462-63 (9th Cir.1985). This general rule has become riddled with exceptions, however. Under our cases, an action need not be remanded to state court in at least five situations: (1) when named defendants prove that the Doe defendants as described in the complaint are wholly fictitious, see, e.g., Grigg v. Southern Pacific Co., 246 F.2d 613, 619 (9th Cir.1957); (2) when the complaint contains no charging allegations against the Doe defendants, see, e.g., Chism v. National Heritage Life Ins. Co., 637 F.2d 1328, 1330 (9th Cir.1981); (3) when plaintiffs unequivocally abandon their claims against the Doe defendants, see, e.g., Southern Pac. Co. v. Haight, 126 F.2d 900, 905 (9th Cir.), cert. denied, 317 U.S. 676, 63 S. Ct. 154, 87 L. Ed. 542 (1942); (4) when the complaint does not identify the Doe defendants with sufficient specificity, see, e.g., Hartwell Corp. v. Boeing Co., 678 F.2d 842, 843 (9th Cir.1982); and (5) when the Doe defendants are not indispensable parties, see, e.g., Othman, 759 F.2d at 1463.

The numerous exceptions to the general principle that the naming of Doe defendants defeats diversity jurisdiction have led to considerable confusion as we ourselves have recognized. In Othman, 759 F.2d at 1462 & n. 7, we noted that "the circumstances under which an action including 'Doe' defendants may be removed to federal court [are] not entirely clear in this circuit." See also Pelleport Investors, Inc. v. Budco Quality Theatres, Inc., 741 F.2d 273, 279 n. 2 (9th Cir.1984) (describing "the vague contours of when Doe pleading is specific enough to defeat diversity"). District court judges and commentators have also noted the doctrinal disarray in our decisions. See, e.g., Goldberg v. CPC Int'l, Inc., 495 F. Supp. 233, 236 (N.D.Cal.1980) (circumstances under which the presence of Doe defendants destroys diversity "unfortunately remain shrouded in mystery and confusion"); Note, Doe Defendants and Other State Relations Back Doctrines in Federal Diversity Cases, 35 Stan.L.Rev. 297, 308 n. 38 (1982) (noting inconsistency in Ninth Circuit case law).

At the request of the three-judge panel, this court agreed to hear this case en banc in order to develop a coherent standard in the Doe defendant area. We now hold that the presence of Doe defendants under California Doe defendant law destroys diversity and, thus, precludes removal. The nature of the allegations against such Doe defendants is irrelevant for federal removal purposes. See CTS Printex, Inc. v. American Motorists Ins. Co., 639 F. Supp. 1272, 1277 (N.D.Cal.1986). We overrule all of our cases creating exceptions to this general rule. See, e.g., Grigg, 246 F.2d at 619; Chism, 637 F.2d at 1330; Hartwell, 678 F.2d at 843; Othman, 759 F.2d at 1463. Under our new rule district courts will no longer have to make the near-impossible determination of when the allegations against Doe defendants are "specific" enough to defeat diversity. Instead, the 30-day time limit for removal contained in 28 U.S.C. Sec. 1446 (b) will not commence until all Doe defendants are either named, unequivocally abandoned by the plaintiff, or dismissed by the state court.5 If a defendant attempts to remove a case prior to this time, the district court must remand the case to state court.6 This new rule accommodates both a plaintiff's right under California law to a three-year extension of the statute of limitations and a defendant's statutory right to removal under 28 U.S.C. Sec. 1441.7III.

Because the complaint in this case contained Doe defendants as parties, removal was premature. We REMAND to the district court with instructions to remand to the appropriate state court. Each party shall bear its own costs on this appeal.

I concur in the judgment because I agree that the district court lacked jurisdiction at the time of removal. Bryant's complaint alleged that Ford and each of the 50 Doe defendants were involved in the design, production and distribution of the Ford van which Bryant claimed had a defective passive restraint system because it did not include a shoulder harness. At the time of removal, Ford made no effort to show that no California resident could have committed acts within the charging allegations of the complaint and thus failed to carry its burden of establishing complete diversity. See Pullman Co. v. Jenkins, 305 U.S. 534, 540, 59 S. Ct. 347, 350, 83 L. Ed. 334 (1939) (named defendant bound to show unnamed defendant a nonresident to justify removal). It made no effort, for example, to show that no one who had any involvement in the design, production or distribution of the van was a resident of California. Consequently, I believe we should decide this case on the narrow ground that the district court lacked jurisdiction because of Ford's failure to show complete diversity at the time of removal.

Although I concur in the judgment, I cannot join the court's opinion because I agree with Judge Kozinski that the court writes too broadly. Dissent at 1085. The court says that there can be no removal until all Does are either named, unequivocally abandoned, or dismissed in the state court action. Opinion at 1083. This necessarily means that an action may not be removed as long as a Doe defendant continues to be named in the complaint and remains unidentified. Thus, the court effectively adopts a rule that Doe defendants are conclusively presumed to be real, not phantom or sham, and are conclusively presumed to be nondiverse. I believe that such a conclusive presumption is inconsistent with Pullman which I read as requiring that a nonresident defendant must be given the opportunity at the time of removal to show that no legitimate defendant is a resident. 305 U.S. at 541, 59 S.Ct. at 350 ("It is always open to the nonresident defendant to show that the resident defendant has not been joined in good faith and for that reason should not be considered in determining the right to remove.") (emphasis provided).1

Suppose hypothetically that Ford had filed affidavits with its removal petition showing that the van had been designed and assembled in Michigan, and that all components, including the passive restraint system, had been produced by companies outside of California, and that Bryant's employer purchased the van from a dealer in Detroit, took delivery in Detroit, and drove the van to California. Suppose further that Bryant was unable to come up with any evidence to controvert the facts set forth in Ford's affidavits. It seems to me that had that been the state of the record at the time the district court acted on Ford's removal petition, removal would have been proper and subject matter jurisdiction would have vested in the district court. A defendant's burden to show complete diversity when faced with Doe allegations may be a heavy one, but I see no justification for denying defendants the opportunity to try.

KOZINSKI, Circuit Judge, with whom Judge O'SCANNLAIN joins, dissenting.

The court has taken this case en banc to resolve a problem that has vexed our district courts for some time: how to treat fictitious parties--so-called Doe defendants--when a case is removed from state court on the basis of diversity of citizenship. That the problem is real and serious is without doubt.1 Far more in doubt is the court's solution. The court does not explain why it has chosen that particular approach to the problem, nor does it consider alternatives that might better reconcile the relevant state and federal interests.

The court's lack of analysis reflects, perhaps, the dearth of briefing and argument on the issue. The case was presented to the en banc panel on the basis of the petition for rehearing, the opposition thereto and the reply. No additional briefing was called for. The rule the court now adopts as the law of the circuit--the so-called CTS Printex rule, named after the district court case where it was first announced, CTS Printex, Inc. v. American Motorists Ins. Co., 639 F. Supp. 1272 (N.D.Cal.1986)--was mentioned only in passing in the opposition to the petition for rehearing and the reply. It was not discussed at any detail at oral argument. No one argued against its adoption or presented any argument that might expose its flaws.2

I respectfully suggest that the court's new rule does, in fact, have serious flaws. As I discuss at greater length below, today's decision will have very serious effects on the operation of the removal statute in any state that allows Doe pleadings, as does California. Moreover, the court's bright-line rule is in large part dicta and its binding force is therefore highly questionable.

The court's approach is particularly troubling because orders remanding cases to the state courts are ordinarily not appealable. 28 U.S.C. Sec. 1447(d) (1982); Thermtron Prods., Inc. v. Hermansdorfer, 423 U.S. 336, 351-52, 96 S. Ct. 584, 593, 46 L. Ed. 2d 542 (1976). Insofar as the court today defers or entirely forecloses a large number of removals, there may be little or no opportunity to correct any error or fine-tune the rule of decision. This should not deter us from doing what we think is necessary. But it does counsel making very sure that we are right. I am just not sure that we are.

A. The Rationale of the Cases the Court Is Overruling Continues to Have Vitality

The court overrules four lines of closely related authority that attempt to distinguish between two types of Doe defendants: those that are real, but whose precise identities are unknown, and those that are procedural fictions, named only for the purpose of preserving the plaintiff's right to add defendants that he might learn about later. The cases the court overrules today stand for the proposition that where the Does are of the second variety--phantoms named solely to toll the running of the statute of limitations--they will not interfere with the exercise of federal jurisdiction. No one has made the point so eloquently as our venerable Chief Judge Emeritus in Grigg v. Southern Pacific Co., 246 F.2d 613 (9th Cir.1957), the earliest of the cases the court is overruling:

Perhaps these Does have some proper place under California state practice. But it is hard to believe they serve any purpose when they are included superstitiously and without reason. Certainly their phantoms, when Does live not and are accused of nothing, should not divert the course of justice.

Id. at 620 (Chambers, J.). Similar sentiments are expressed in other cases. See, e.g., Hartwell Corp. v. Boeing Co., 678 F.2d 842, 843 (9th Cir.1982); Chism v. National Heritage Life Ins. Co., 637 F.2d 1328, 1330 (9th Cir.1981); Asher v. Pacific Power & Light Co., 249 F. Supp. 671, 676 (N.D.Cal.1965).

This concern, it seems to me, is a legitimate one. While there may be a small number of cases where defendants are known to exist but their names are not available, there are many more cases where Doe defendants alleged in the complaint are simply procedural fictions. My understanding is that nearly every civil case now filed in California contains Doe allegations.3 The reason is simple: It would be malpractice for a lawyer to omit such allegations if there is even the remotest chance that other defendants might turn up. This means that there is a fair number of cases--perhaps the overwhelming majority--where all the parties that will ever turn up are already before the court and the presence of Doe allegations impairs removal for absolutely no good reason.

The decisions the court overrules today attempted the difficult but important task of distinguishing the cases where Does are real but unidentified parties from those where they are nothing more than procedural fictions. What the court's opinion documents quite clearly, and what our district judges and lawyers have been telling us, is that the fine line-drawing required to separate the goats from the sheep in this fashion is too burdensome and results in too much uncertainty. See, e.g., CTS Printex, 639 F. Supp. at 1277.

Fair enough. The need for a bright-line rule, capable of more predictable application, is squarely presented to us. Yet this does not nullify the reasons we adopted the more complex rule in the first place. Grigg and its progeny pursued an important objective: preserving the authority of the federal courts to exercise jurisdiction conferred on them by an Act of Congress. That interest ought not be defeated by a state procedural device that in most instances is invoked "superstitiously and without reason." 246 F.2d at 620. We should fashion a rule that takes these concerns into account and minimizes the conflict between the federal interest in the proper exercise of removal jurisdiction and the state interests represented by the Doe pleading practice.

The rule the court adopts fails to do this. Without any discussion, the court abandons the federal interest Grigg and its progeny sought to protect. In Judge Chambers' words, the phantoms of Does that "live not and are accused of nothing" are now given substance for the purpose of deferring or defeating federal jurisdiction.

If the court is intent on adopting a bright line rule, there are two from which it can choose: one that treats Doe defendants as always destroying diversity, and one that treats them as mere procedural fictions, never destroying diversity. I fear that the court has selected the wrong bright-line rule, the one that least well reconciles the relevant state and federal interests.

B. The Court's New Rule Seriously Interferes with the Federal Court's Exercise of Jurisdiction Under the Removal Statute

The court recognizes that its rule "may lead to removal at a late stage in the proceedings." Majority op. at 1083, n. 5. Probably more accurate is the observation in CTS Printex that, under the new rule, removal "may occur on the eve of trial (if trial occurs within three years of filing of the complaint)...." 639 F. Supp. at 1277. Even if trial is delayed, removal will not always be available after three years. While California law requires that defendants be served within three years, the rule is subject to a variety of exceptions and exclusions. See Cal.Civ.Proc.Code Secs. 583.220-240 (West Supp.1987).4 In addition, the three-year rule is also subject to judicial interpretation that provides further refinements and exceptions. See, e.g., Barrington v. A.H. Robins Co., 39 Cal. 3d 146, 157, 216 Cal. Rptr. 405, 411, 702 P.2d 563, 569 (1985).

Thus, under the majority's rule, removal will be delayed to the eve of trial,5 to three years after filing, or to some far later time. This is very, very late indeed to be bringing cases to federal court under a statute that, by its terms, directs that removals occur within thirty days after the filing of the complaint. This wholesale delay of removals in virtually all diversity cases filed in California (and perhaps other states allowing Doe pleadings)6 is not only at odds with the plain statutory language, it defeats longstanding congressional policy with respect to removals and will result in a variety of practical problems.

1. The Majority's Approach Conflicts with the Policy of the Federal Removal Statute

The federal removal statute, 28 U.S.C. Secs. 1441-1452 (1982 & Supp. III 1985), embodies congressional policy that cases brought in state court be transferred to federal court if they could have been brought there in the first place. See Grubbs v. General Elec. Credit Corp., 405 U.S. 699, 92 S. Ct. 1344, 31 L. Ed. 2d 612 (1972). This provision dates back to the Judiciary Act of 1789, 1 Stat. 73, 79 (1789), and, except for a short period during the 19th century, has reflected two policies: First, cases should be removed as early as possible in the litigation, Powers v. Chesapeake & Ohio Ry., 169 U.S. 92, 100, 18 S. Ct. 264, 267, 42 L. Ed. 673 (1898); second, federal removal law should be consistent nationwide, unaffected by differences in state law. Grubbs, 405 U.S. at 705, 92 S.Ct. at 1348; Chicago, R.I. & Pac. R.R. v. Stude, 346 U.S. 574, 580, 74 S. Ct. 290, 294, 98 L. Ed. 317 (1954); Shamrock Oil & Gas Corp. v. Sheets, 313 U.S. 100, 104, 61 S. Ct. 868, 870, 85 L. Ed. 1214 (1941). The rule the court adopts today offends both of these policies.

Legislative histories were not in the early days of the Republic what they are today (perhaps fortunately so). We therefore have little to tell us why the drafters of the first Judiciary Act required removal early in the litigation. What we do have, however, is an experiment during the Reconstruction era when removals were allowed any time before trial. This innovation was short-lived. In 1875, the date was moved back closer to the commencement of the litigation. 18 Stat. 471 (1875). Such legislative history as we have shows the obvious: The rule change was occasioned by dissatisfaction with removals so close to trial. In the course of debates on the floor of the House, Congressman Poland declared:

The third section of this bill provides that in all actions hereafter brought the application for removal shall be made at the first term of the court thereafter. That was the old law of 1789, that when the defendant sought a removal of the cause he must do so at the time of entering his appearance. But in later statutes, especially those passed during and since the war, we have provided that applications for removal may be made at any time before trial; which I think, and the committee think [sic], is very mischievous in its consequences. A party will let his case run on in a State court until he is satisfied that he will be beaten, and then dodge off into a Federal court. The third section of this bill provides that in all suits hereafter commenced the practice shall go back to the act of 1789, and parties seeking the removal of causes shall do so at the very outset.

3 Cong.Rec. 4302 (1874).

This has been the law ever since, although the precise formula has varied from time to time.7 As the Supreme Court stated in Powers: "This provision clearly manifests the intention of Congress that the petition for removal should be filed at the earliest possible opportunity." 169 U.S. at 100, 18 S.Ct. at 267. There is good reason for this. Late removals can be extremely disruptive, lead to serious duplication of effort and be manipulated for tactical advantage. See pp. 1089-90 infra.

To be sure, the rule is not absolute; it does allow for slippage where the case becomes removable later in the litigation. See 28 U.S.C. Sec. 1446(b). However, it is important to recognize that the exception is just that, an exception. It should not be allowed to swallow up the basic policy of early removal. The court's new rule thwarts this settled policy by postponing removal based on diversity of citizenship for at least three years in practically every case coming out of California, a state that is responsible for well over half of the diversity filings in the nation's largest circuit.8 I can think of no more effective way of undermining the congressional policy of early removal.9

The court's rule also undermines the congressional policy that removal procedures be uniform in the federal courts, unaffected by the vagaries of state law. The court's rule gives California--and every other state that has or can adopt a California-style Doe pleading rule--the authority to defer federal removal for years, perhaps indefinitely.10 Congress did not intend that the right to remove be manipulated by state law in this fashion. As the Supreme Court said in Shamrock, "[t]he removal statute, which is nationwide in its operation, was intended to be uniform in its application, unaffected by local law definition or characterization...." 313 U.S. at 104, 61 S.Ct. at 870.

Moreover, deferring removals on a wholesale basis is more than a question of timing; it significantly affects the very availability of removal, impairing the exercise of a right guaranteed by Congress. Many cases are terminated early in the litigation by dismissal or summary judgment on the basis of procedural default or for some other reason. As to those cases, removal is effectively denied under the court's rule even if there is not the slightest possibility that any of the Does will materialize or that they would destroy diversity.

Equally troubling, delaying removal to the eve of trial, or for three or more years into the litigation, changes significantly the incentives for and against removal, subjecting it to the kind of manipulation Congress condemned when it amended the removal statute in 1875. See p. 1088 supra. The decision to remove at the start of litigation is based principally, perhaps entirely, on the choice of forum, the choice Congress meant for defendants to make. On the eve of trial, a decision to remove is based on far different considerations. For one thing, the possible advantages of federal pretrial practice and case management would no longer be relevant. Moreover, a defendant would have to consider whether to remove the case and possibly expose himself to another round of discovery in district court or, conversely, whether to remove in order to subject the plaintiff thereto.

Also relevant at that point would be the outcome of various pretrial rulings. A defendant who found himself on the losing side of such rulings might view removal as an opportunity to relitigate those issues. See p. 1090 infra. Perhaps most obviously--and least appropriately--removals on the eve of trial can be used as devices for oppression by further delaying a plaintiff's day in court,11 precisely what Congress worried about when it moved the removal date back to the start of the litigation.12 If there are sufficient reasons for adopting a rule that so runs against the grain of the removal statute, the court does not explain what they are.

2. The Court's Rule Will Create Serious Practical Problems

Deferring removal of all diversity cases for three years will create a number of procedural problems. For starters, this will generate a substantial duplication of effort. When a case is removed to federal court, all interlocutory rulings of the state court are subject to reconsideration by the district judge. 28 U.S.C. Sec. 1450; Granny Goose Foods, Inc. v. Brotherhood of Teamsters, 415 U.S. 423, 435-37, 94 S. Ct. 1113, 1122-23, 39 L. Ed. 2d 435 (1974); Preaseau, 591 F.2d at 79; 1A J. Moore & B. Ringle, Moore's Federal Practice p 0.168[4.-1], at 640 (2d ed. 1987). This would include discovery rulings, decisions on summary judgment motions and interlocutory stays and injunctions. The later the removal, the more interlocutory rulings there are likely to be, and the greater the opportunity for relitigation. While federal judges may generally defer to their state colleagues on these matters, losing litigants have every incentive to seek reconsideration in the new forum. In Preaseau, for example, they did so and were successful. 591 F.2d at 79-83. But win or lose, the attempt will consume the time of the court and the money of the parties. When state and federal courts are barely able to handle their burgeoning caseloads, and there are frequent and justified complaints about the cost and delay of litigation, the court should be reluctant to fashion a rule that will cause such wholesale duplication of effort.

We also should not underestimate the disruption caused by plucking cases from the state courts many years after the start of litigation. Discovery may be discontinued or disrupted; pending motions would have to be withdrawn from the state court and refiled or reargued before the district court; injunctions, stays, receiverships and other equitable remedies would have to be transferred and conformed to federal procedural and bonding requirements. Various rulings may be in the process of interlocutory appeal to the state appellate courts; such appeals would be short-circuited by removal. Any trial date set by the state court would have to be cancelled, with the attendant disruption to parties and witnesses, and the case would normally take its place at the bottom of the district court's civil docket. And, all of this assumes both parties are acting in the best of faith; if one side wants to use the removal as an excuse for dragging its feet, it will find plenty of reasons for doing so. While it is difficult to predict with any accuracy what procedural problems this will spawn, it is a fair guess that there will be many more than if the cases were removed thirty days after filing of the complaint, as Congress intended.13 Some of these problems may occur now when the right to remove does not arise until late in the litigation. But, again, these are exceptions, rare situations that may involve an occasional disruption. It is quite a different matter when removal is delayed in virtually every diversity case coming out of California (and perhaps other Doe states) for three years or more.

All that having been said, I might nevertheless be willing to go along with the court's rule if I thought it represented a tenable reading of the removal statute. It does not. Indeed, in order to make the rule work at all, the court has to address situations not presented by this record; much of the court's supposedly bright-line rule is therefore dicta.

While the court's rule appears to be simple enough, it in fact deals with two rather distinct issues: (1) whether inclusion of Doe parties in the complaint destroys diversity; and (2) when Does originally pled may be deemed eliminated so that an originally non-diverse case becomes diverse. The court is certainly in a position to address the first question: We have here Doe allegations that, arguably, were sufficiently specific to destroy diversity. If the court wants to say that Doe allegations, no matter how unspecific, will always destroy diversity, it certainly may do so, although I would question the wisdom of the rule.

But the court goes much further than this. In an effort to adopt a simple rule that will solve all problems, the court goes on and addresses the second question: when Doe allegations disappear from a case by abandonment or otherwise. But the court is in no position to speak on this issue because the case before us does not present the issue of abandonment: No one has claimed that the Doe defendants have been abandoned and every indication is that the plaintiff fully intends to rely on the Doe allegations. Under these circumstances, that part of the court's bright-line rule dealing with this issue is, quite simply, dicta, and very mischievous dicta at that. Perhaps because the issue is not presented to us in a concrete controversy, the rule the court adopts runs rough-shod over the statutory language and demonstrably excludes a variety of situations where federal jurisdiction is authorized by the removal statute.

The removal statute provides that if the case is originally not removable, the defendant may remove within 30 days after he receives "a copy of an amended pleading, motion ... or other paper from which it may first be ascertained that the case is one which is or has become removable." 28 U.S.C. Sec. 1446(b). The court here reformulates this statutory standard as follows: "The 30-day time limit for removal ... will not commence until all Doe defendants are either named, unequivocally abandoned by the plaintiff, or dismissed by the state court." Majority op. at 1083 (footnote omitted). The first of these possibilities is tantalizing but meaningless. The typical complaint filed in the California courts pleads Does by the dozens, sometimes by the hundreds, far outnumbering the defendants that could conceivably be brought into the case. There are fifty Does in this case alone. See p. 1092 n. 14 infra. The last possibility--dismissal of the Does--is real enough, but only materializes after the period for serving Does has expired.

It is the second possibility--"unequivocal abandonment"--that is the most troublesome, however. In a footnote, the court explains that "[u]nequivocal abandonment occurs in only two situations: (1) where the plaintiff drops the Doe defendants from the complaint or (2) where the trial commences without service of the Doe defendants." Majority op. at 1083 n. 5 (citation omitted). The court's restrictive approach is difficult to reconcile with statutory language which provides that a case may be removed within 30 days after the defendant receives "a copy of an amended pleading, motion ... or other paper from which it may first be ascertained that the case is one which is or has become removable." This language, I submit, covers a wide variety of situations, not just the two enumerated in footnote 5 of the court's opinion. Consider the following:

Plaintiff writes defendant a letter advising him that he is abandoning the unserved Doe defendants. Defendant writes back requesting that plaintiff formally amend the complaint to drop the Does but plaintiff refuses to do so. This situation clearly seems to fit the statutory language as an event triggering the right to remove, yet, under the court's new rule, this would not be considered "unequivocal abandonment."

Defendant propounds requests for admission asking plaintiff to admit that there are no unknown parties left unserved. Plaintiff does not deny. Ditto.

Plaintiff files an at-issue memorandum containing the following certification: "Plaintiff is aware of no unserved parties that will be brought in the case and abandons its claims against any parties who have not been served as of this date." Ditto.

These are not wild examples; they, and countless others like them, happen every day. Our district courts and litigants will be required to confront them. The court's opinion will no doubt create a serious dilemma to those seeking to apply its teachings. On the one hand, the opinion speaks with the authority of the full court, and its rulings must be taken very seriously. On the other hand, this aspect of its ruling is so obviously dicta, and so obviously fails to follow the clear statutory language, that courts and litigations may well be tempted to shrug it off as not really meaning what it says. While defiance of binding authority is never appropriate, blindly following dicta also has its hazards.

That the court must resort to dicta to make its rule workable should be a further indication that its rule may be flawed. The difficulty, as I see it, is that the court is the captive of precedent that it is authorized to overrule but fails to reconsider. Specifically, the court fails to consider the fundamental question of the proper status of Doe defendants when a case is removed to federal court. I attempt to do so below.

There are two diametrically opposed models of how Doe defendants should be viewed by the federal courts. On the one hand, they could be treated like real parties--actual flesh and blood people--whose names happen to be unknown. On the other hand, Does could be viewed as procedural fictions, magic words used by lawyers in pleadings for the sole purpose of tolling the statute of limitations against parties that might conceivably turn up. While there may be a small number of cases falling into the former category, the vast majority of Does that populate the state courts of California are of the latter type. Doe allegations, of which those here are typical,14 generally do not represent real people, diverse or otherwise. They should not, therefore, automatically be treated as if they do. That decision should be made on the basis of whether to do so is consistent with the Federal Rules of Civil Procedure and other relevant federal interests, including the unimpaired operation of the federal removal statute.

The Federal Rules of Civil Procedure do not provide for suing fictitious parties. Indeed, the practice is inconsistent with many of the rules and incompatible with federal procedure. See, e.g., Fed.R.Civ.P. 10(a) ("[i]n the complaint the title of the action shall include the names of all the parties...."); pp. 1094-96 infra. We have repeatedly held that a suit naming Doe defendants may not be maintained in federal court. See, e.g., Fifty Assocs. v. Prudential Ins. Co. of Am., 446 F.2d 1187, 1191 (9th Cir.1970); Craig v. United States, 413 F.2d 854, 856 (9th Cir.), cert. denied, 396 U.S. 987, 90 S. Ct. 483, 24 L. Ed. 2d 451 (1969); Molnar v. National Broadcasting Co., 231 F.2d 684, 686-87 (9th Cir.1956). Nevertheless, in a more recent case, we held, that a plaintiff is entitled to the benefit of the California Doe pleading practice when his case is removed to federal court. Lindley, 780 F.2d 797.

This is a critical assumption. If plaintiffs may, as a matter of right, substitute the names of real people for the Does, then the Does must be treated as real people for purposes of determining diversity of citizenship, for they may well become real at the plaintiff's option. But, it seems to me, Doe defendants are more readily treated as nullities for purposes of federal practice, seeing as they are not authorized or contemplated by our Federal Rules of Civil Procedure. If that were the case, the problems with which we have been grappling would disappear; the court could look only at the named parties for purposes of determining diversity at the time of removal; additional parties could be added later under the Federal Rules, just as in any other diversity case brought in district court.15 While the court does not discuss Lindley, the Lindley opinion plainly is the predicate for its ruling today.16 It becomes crucial, therefore, to consider whether Lindley was correctly decided.

California's Doe pleading practice addresses a specific issue that arises during the course of litigation: How does the filing of the lawsuit affect the running of the statute of limitations as to potential defendants whose identity is unknown at the time suit is brought? The California rule is that, for three years after filing, the plaintiff may amend the complaint by adding any such parties as a matter of right. Cal.Civ.Proc.Code Secs. 474, 583.210 (West 1979 & Supp.1987). In federal court, the same problem is addressed and resolved by Fed.R.Civ.P. 15(c). Under the Federal Rule, an amendment to the complaint adding a party relates back to the date of the original pleading only if, inter alia, the party to be added: (1) has received notice of the action; and (2) knew or should have known that, but for the mistake in naming the original party, the action would have been brought against him.

When a diversity case is initially filed in federal court, Rule 15(c) clearly supersedes state law on relation back. See, e.g., Santana v. Holiday Inns, Inc., 686 F.2d 736, 738-39 (9th Cir.1982); Britt v. Arvanitis, 590 F.2d 57, 61-62 (3d Cir.1978). However, Lindley holds that when the case is first filed in state court and then removed, the Doe pleading rule is imported into the case as a matter of state substantive law, trumping Rule 15(c). 780 F.2d at 800-01.

The Lindley approach is foreclosed by the Supreme Court's decision in Hanna v. Plumer, 380 U.S. 460, 85 S. Ct. 1136, 14 L. Ed. 2d 8 (1965). Hanna addressed the question of what happens when there is a conflict between state law and the Federal Rules of Civil Procedure. It holds that if there is a direct conflict between the Federal Rules and state law, the Federal Rule takes precedence unless, of course, the Federal Rule is invalid. The following passage from Hanna could not be clearer on this point:

It is true that both the [Rules] Enabling Act [28 U.S.C. Sec. 2072] and the Erie rule say, roughly, that federal courts are to apply state "substantive" law and federal "procedural" law, but from that it need not follow that the tests are identical. For they were designed to control very different sorts of decisions. When a situation is covered by one of the Federal Rules, the question facing the court is a far cry from the typical, relatively unguided Erie choice: the court has been instructed to apply the Federal Rule, and can refuse to do so only if the Advisory Committee, this Court, and Congress erred in their prima facie judgment that the Rule in question transgresses neither the terms of the Enabling Act nor constitutional restrictions.

380 U.S. at 471, 85 S. Ct. at 1144 (emphasis added; footnote omitted). See generally 3 J. Moore, Moore's Federal Practice p 15.15, at 15-142 (2d ed. 1987) ("Hanna v. Plumer is dispositive of the issue and ... the matter is ... one solely of federal practice under Rule 15(c)") (footnote omitted).

Earlier this year, the Supreme Court unanimously reaffirmed the vitality of Hanna and further explained its scope and rationale. See Burlington Northern R.R. v. Woods, --- U.S. ----, 107 S. Ct. 967, 94 L. Ed. 2d 1 (1987). Burlington Northern dealt with an Alabama statute requiring that a 10 percent penalty be assessed against an unsuccessful appellant who had obtained a stay of the judgment pending appeal. The appellee in a federal diversity case removed from state court claimed the benefit of this rule, arguing that he was entitled to it as a matter of substantive state law. Speaking for the Court, Justice Marshall held that the application of the Alabama statute was foreclosed under Hanna by Federal Rule of Appellate Procedure 38, which provided that "[i]f a court of appeals shall determine that an appeal is frivolous, it may award just damages and single or double costs to the appellee."

The Court first noted that "[t]he cardinal purpose of Congress in authorizing the development of a uniform and consistent system of rules governing federal practice and procedure suggests that Rules which incidentally affect litigants' substantive rights do not violate [the Rules Enabling Act's requirement that the rule not abridge, enlarge or modify a substantive right] if reasonably necessary to maintain the integrity of that system of rules." 107 S. Ct. at 970 (emphasis added). The Court then noted that Rule 38's "discretionary mode of operation unmistakably conflicts with the mandatory provision of Alabama's affirmance penalty statute. Moreover, the purposes underlying the Rule are sufficiently coextensive with the asserted purpose of the Alabama statute to indicate that the Rule occupies the [state] statute's field of operation so as to preclude its application in federal diversity actions." Id. at 970-71 (emphasis added).17 The Court rejected the argument that the state and Federal Rules were not truly in conflict because "a federal court sitting in diversity could impose the mandatory penalty and likewise remain free to exercise its discretionary authority under Federal Rule 38." Id. at 971. The Court stated:

This argument ... ignores the significant possibility that a Court of Appeals may, in any given case, find a limited justification for imposing penalties in an amount less than 10% of the lower court's judgment. Federal Rule 38 adopts a case-by-case approach to identifying and deterring frivolous appeals; the Alabama statute precludes any exercise of discretion within its scope of operation.

Id. (emphasis original).

Burlington Northern puts an important gloss on Hanna. It stands for the proposition that there may be a conflict between the Federal Rules and state law even if there is no direct contradiction. The relevant questions are whether the Federal Rule "is reasonably necessary to maintain the integrity of [the federal] system of rules," and whether it "occupies the [state] statute's field of operation." Id. at 970.

Under this standard, it is clear that California's Doe pleading practice is supplanted by the Federal Rules of Civil Procedure once the case is removed to federal court. Lindley's contrary conclusion, 780 F.2d at 800-01, reached without the benefit of the Supreme Court's discussion in Burlington, is no longer plausible. We not only have direct conflicts between the operation of the state statute and the Federal Rules, see pp. 1095-96 infra, we also have serious disruptions of federal procedures and, as of today, a wholesale subservience to state law of an Act of Congress dating back to the earliest days of the Republic. Compared to this, the conflict between state and federal law in Burlington Northern seems like a trifle.

The conflict between state and federal procedure in this case is clear. California's Doe pleading rule has two components. First, Cal.Civ.Proc.Code Sec. 474 provides that a defendant may be designated by a fictitious name; the real name may be substituted whenever the true identity is discovered. Next, Cal.Civ.Proc.Code Sec. 583.210 provides that a plaintiff has up to three years after filing to serve the summons and complaint on any defendant. California courts have interpreted this to extend the time available to replace Doe defendants with named parties. Lesko v. Superior Court, 127 Cal. App. 3d 476, 481-82, 179 Cal. Rptr. 595, 598 (1982). The analogous Federal Rules are Fed.R.Civ.P. 4(j), 15(c), 19, 20 and 21.

Rule 4(j) provides that the defendant must be served within 120 days of the filing of the complaint; absent good cause, failure to serve within this time renders the complaint subject to dismissal. This provision cannot be reconciled with Cal.Civ.Proc.Code Sec. 583.210, which allows three years for service.18

Rules 19, 20 and 21 prescribe when new parties may or must be added to complaints filed in federal court. The rules are comprehensive and give the district court broad authority to accept or reject new parties. By contrast, Cal.Civ.Proc.Code Sec. 474 has no restrictions whatsoever; the court has no discretion; the plaintiff can bring in newly-identified parties at will. The Federal Rules and California statutes thus address exactly the same problem and resolve it in different ways, clearly occupying the same "field of operation."

Finally, Federal Rule 15(c) determines the extent to which parties added by amendment are subject to the complaint's original filing date. Under the Federal Rule, an amendment to the complaint adding a party relates back to the date of the original complaint only if, inter alia, the party to be added: (1) has received notice of the action; and (2) knew or should have known that, but for the mistake in naming the original party, the action would have been brought against him. The state rule requires no such notice to the new party; the amendment relates back in all cases. Here again, the Federal Rules and state law reach inconsistent solutions to the same problem.

When federal law and state law address precisely the same issues and resolve them in different and inconsistent ways, the state statute would seem to be clearly preempted under the Supreme Court's analysis in Burlington Northern. Lindley, however, rejects this conclusion. 780 F.2d at 800-01. Not surprisingly, this leads to a variety of anomalies. Perhaps the most serious of these is that if--as Lindley holds--Doe pleadings must be permitted in federal court, this must surely apply to all diversity cases, not merely those initiated in state court and then removed to federal district court. Fed.R.Civ.P. 4(j) then would become inapplicable to all California diversity cases, plaintiffs in all such cases having three years, not 120 days, to serve the complaint.

Similarly, if Lindley has identified a rule of substantive state law that we must apply under Erie, all California diversity plaintiffs, not merely those where the case is removed from state court, should be entitled to file a complaint in federal court naming Doe defendants, and have an absolute right to bring in real parties within three years, regardless of whether they would otherwise be permitted to do so under Fed.R.Civ.P. 19, 20 and 21. Circuit authority is squarely to the contrary. In Molnar, 231 F.2d at 686-87, we held that a district court diversity complaint containing Doe allegations is subject to dismissal. Accord Fifty Assocs., 446 F.2d at 1191. Based on these decisions, the local rules of at least three of our district courts prohibit Doe pleadings. D.Ariz.R. 10(d); C.D.Cal.R. 3.7.2.1; S.D.Cal.R. 200-6. If Lindley's rationale is accepted, it overrules Molnar and Fifty Associates, and invalidates these local rules.

Lindley creates yet another conflict with one of our cases interpreting the Federal Rules. Santana, 686 F.2d 736, considered whether Rule 15(c) could be applied in a diversity case to extend the state statute of limitations where state law had no equivalent relation-back doctrine. Analyzing the situation under Hanna, we held in Santana that the Federal Rule trumped state law: "We conclude that Hanna commands application of Rule 15(c) in the face of a contrary state rule, and is thus applicable in the present case." Id. at 740.

Lindley attempts to distinguish Santana by arguing that Rule 15(c) trumps state law when it extends the state statute of limitations but not when it shortens it. 780 F.2d at 801. This takes an incongruous view of state law. Statutes of limitations provide rights for plaintiffs (to bring suit within a specified time) and for defendants (repose after that time runs). Lindley and Santana can only be reconciled under the theory that plaintiffs' state law rights are more important than those of defendants. The Federal Rules embody no such one-sided principle. It seems to me that either Santana is right or Lindley is, but not both.

Finally, we ought not overlook the confusion Doe pleadings have caused in the district courts, confusion so severe that we convened an en banc panel to deal with the problem. It is also relevant that the en banc panel is so troubled by the problem that it has adopted a mechanical rule that will have a major impact on the operation of the removal statute, a statute that has reflected congressional policy going back to the first Judiciary Act. Before taking this step, it is worth considering an alternative.

Overruling Lindley would do much to resolve the confusion in this area of the law. If a plaintiff imports with him no right to substitute real parties for fictitious ones in federal court, the fictitious defendants have no meaning and can be disregarded in determining diversity of citizenship.19 If and when new parties are later discovered, they may be added in accordance with the Federal Rules of Civil Procedure. For the great majority of cases, where all the parties are known and the Does are procedural fictions, exclusion of the Does from the federal litigation would have no effect whatsoever.

This resolution of the problem has the benefit of both simplicity and completeness. It would eliminate all the problems associated with the addition of Doe defendants, not just some of them. It would not matter whether the Doe allegations were specific or general; nor would a defendant have to guess whether a particular statement or pleading by the plaintiff constitutes an abandonment or waiver of his right to pursue Does, triggering the right to removal. Moreover, the decision as to removal under this rule would be made when Congress intended: very early in the litigation, before the case had an opportunity to grow roots on the state court's docket. Finally, it would make removals consistent nationwide and avoid giving states and litigants the power to manipulate the timing of federal removals. See p. 1090 n. 13 supra.

This approach would not necessarily deny plaintiffs the right to pursue unknown parties as they are permitted by state law. In the first place, the Federal Rules contain liberal joinder provisions and in many cases--probably most cases--a plaintiff may be able to join a newly discovered party after the litigation has begun. In any event, the rule would only apply to the federal courts. It does not, and cannot, speak to what rights the plaintiffs may have under state law. That, it seems to me, is a problem for the state courts and the state legislature to work out. Finally, if, in individual cases, the district court is concerned that application of the federal rule will work undue hardship on a plaintiff, it can remand the Doe allegations and allow the plaintiff to pursue the case against the Does in state court.

The court continues to take the law of removal in the wrong direction. Convened to solve a problem, the court only exacerbates it. While the rule it adopts may initially lessen somewhat the burdens on the district courts, it surely will not solve all the problems and may create many more. And it may well be a false economy; duplication of effort resulting from late removals may create as much work as it saves. Moreover, the court's rule sacrifices, unnecessarily I submit, important federal rights. While the decision may please those unsympathetic to diversity jurisdiction, it is inconsistent with the law as Congress has written it. I respectfully, but firmly, dissent.