RANDALL, Circuit Judge:

In this diversity tort suit, the court below held that the plaintiff's claim was barred by the applicable statute of limitations. Although the ultimate application of the appropriate statute of limitations is a comparatively simple business, the intermediate inquiry of which state's statute of limitations should be applied leads us on a merry chase through the murky area in which the Erie doctrine and the federal change of venue statutes intersect. For the reasons set forth below, we reverse and remand.

I. FACTUAL BACKGROUND LEADING TO THIS APPEAL

On July 29, 1976, Ida B. Ellis allegedly suffered a fall while alighting from a "log ride" at the Six Flags Over Texas amusement park in Arlington, Texas. As a consequence of the fall, Mrs. Ellis allegedly sustained a brain injury that subsequently caused her death. On July 28, 1978 — exactly two years from the time of her fall, less one day — Jeff Ellis, as administrator of her estate, filed this survival and wrongful death action in the United States District Court for the Western District of Arkansas, Hot Springs Division. Ellis claimed that Mrs. Ellis' death was proximately caused by the negligence of the amusement park, and sought damages in the amount of $404,300.

Three corporations were named as defendants in the original complaint. One, Great Southwestern Corporation, was never served with valid process. The other two defendants named were Six Flags, Inc. ("SFI") and Six Flags Over Texas ("SFOT"). The complaint alleged that SFI was a California corporation with no place of business in Arkansas; SFOT was alleged to be a Texas corporation with its principal place of business in Arlington, Texas. Jeff Ellis, the named plaintiff, was alleged to be a citizen of Arkansas. Subject-matter jurisdiction was apparently premised upon diversity of citizenship under 28 U.S.C. § 1332 (1976), although the complaint contained no statement as to either subject-matter jurisdiction or venue.

At the time he filed his original complaint, Ellis had requested that the U.S. Marshal for the Western District of Arkansas serve SFI and SFOT by substituted service on the Arkansas Secretary of State pursuant to the Arkansas long-arm statute, Ark.Stat.Ann. §§ 27-339.1, 27-340 (1979). This request was delivered to the U.S. Marshal on July 31, 1978, and the Arkansas Secretary of State accepted service for these defendants on August 2, 1978.

On August 16, 1978, SFI and SFOT filed a joint motion to dismiss in which they alleged lack of personal jurisdiction, insufficiency of process, and failure to state a claim upon which relief could be granted. This motion was apparently predicated on Fed.R.Civ.P. 12(b)(2), (4), and (6), although that rule was not referred to in the motion. On September 21, 1978, SFI and SFOT filed a supporting affidavit from Gary W. Nielsen, the secretary and corporate counsel of the parent company of both SFI and SFOT. Nielsen averred that SFI and SFOT were both Delaware corporations and that neither had done, was doing, or was authorized to do business in Arkansas.

At the February 20, 1979, docket call, Ellis requested a continuance to develop information relating to jurisdiction, and was given ninety days to respond to the defendants' motion to dismiss. Although he did not conduct any formal discovery or file controverting pleadings or affidavits, Ellis filed a motion to transfer the case on June 11, 1979. The motion alleged only "[t]hat venue in the above-styled and numbered cause properly lies in the United States District Court, for the Northern District of Texas, Ft. Worth Division, and that said cause should be transferred to the aforementioned Court." The motion did not specify whether such transfer was to be under 28 U.S.C. § 1404(a) (1976) or 28 U.S.C. § 1406(a) (1976); neither did it allege that venue was not proper in Arkansas, or that the convenience of the parties or the interests of justice would be served by a transfer to Texas. Nonetheless, on June 14, 1979, the motion to transfer was granted. The transfer order does not indicate whether the transfer was pursuant to section 1404(a) or section 1406(a). The parties and the transferee court, however, have consistently characterized the transfer as having been pursuant to section 1406(a).

On August 6, 1979, SFI and SFOT filed a joint answer, subject to their pending motion to dismiss. The defendants generally denied Ellis' allegations of negligence, and pleaded as an affirmative defense that the suit was barred by the applicable statute of limitations.

Ellis claims to have requested on August 22, 1979, that the U.S. Marshal for the Northern District of Texas serve SFI and SFOT. For reasons not entirely clear from the record or the parties' briefs, however, summons was not issued at that time. The docket sheet in the record on appeal contains only this mysterious notation, dated December 17, 1979:

Re-issued SUMMONS/AMENDED COMPLAINT to C.T. Corporation, Republic National Bank Bldg, Dallas, Texas and Six Flass [ sic], Inc., by Secy of State (Clerk's error this was not issued until this date — also atty had not mailed $10.00 check).

The docket sheet then indicates that SFOT was personally served on December 21, 1979, with the Marshal's return thereupon filed on December 27, 1979; substituted service on SFI through the Texas Secretary of State was executed on January 11, 1980, and the return thereupon was filed on January 16, 1980. The parties agree that the initial substituted service upon SFI and SFOT through the Arkansas Secretary of State was invalid, and that it was not until the U.S. Marshal in Texas served SFI and SFOT that the court obtained personal jurisdiction over those two defendants.

On January 16, 1980, the defendants moved for summary judgment, arguing that Ellis' suit was barred under either the two-year statute of limitations contained in Tex.Rev.Civ.Stat.Ann. art. 5526 (Vernon Supp. 1980), or the three-year statutes of limitations contained in Ark.Stat.Ann. § 27-907 (1979) (wrongful death) and § 37-206 (1962) (survival). The court below granted summary judgment on June 30, 1980, in an unpublished memorandum opinion and accompanying order. The court addressed itself primarily to the Arkansas statutes of limitations, since they were longer than the Texas statute and since Ellis' arguments had focused more specifically on them. The court concluded that it was not until August 22, 1979, that Ellis had completed the actions necessary to toll the running of the Arkansas statutes of limitations: under Arkansas law, according to the court below, Ellis' action was not "commenced" for purposes of tolling the limitations period until summons was issued to the U.S. Marshal in the proper district. Since this date was more than three years after the accrual of the cause of action on July 28, 1976, the court held that Ellis' suit was barred.

II. WHICH STATE'S STATUTE OF LIMITATIONS? [13] A. The Choice of Law Rules to be Applied in Choosing the Proper Statute of Limitations

Our starting point in this case, as in all cases in which subject-matter jurisdiction is premised on diversity of citizenship, is Erie Railroad Co. v. Tompkins, 304 U.S. 64, 58 S.Ct. 817, 82 L.Ed. 1188 (1938), in which the Supreme Court held that there is no federal general common law and that a federal court must apply the laws of the state in which it sits except in those cases governed by the Constitution or laws of the United States. Two of Erie's progeny are particularly relevant. In Guaranty Trust Co. v. York, 326 U.S. 99, 65 S.Ct. 1464, 89 L.Ed. 2079 (1945), the Court confirmed that the Erie doctrine extends to state statutes of limitations. And in Klaxon Co. v. Stentor Electric Manufacturing Co., 313 U.S. 487, 61 S.Ct. 1020, 85 L.Ed. 1477 (1941), the Court confirmed that the Erie doctrine means also that as a general matter, a federal court is bound to apply the choice of law rules of the state in which it sits in determining whether that state's or some different state's substantive law should govern.

Together, these two cases mean that in diversity lawsuits, a federal court is ordinarily bound to look to the choice of law rules of the state in which it sits to determine whether the state courts of that state would apply their own state's statute of limitations or the statute of limitations of some other state. Baron Tube Co. v. Transport Insurance Co., 365 F.2d 858, 860 (5th Cir. 1966) (en banc). This principle may have significant practical effect in those cases in which the state courts of the state in which the federal court sits would apply the substantive law of another state, but would apply their own statute of limitations.

The differences between state and federal procedural rules have from time to time been the source of considerable confusion when federal courts have attempted to apply state statutes of limitations in diversity cases. For example, Fed.R.Civ.P. 4, rather than the analogous state procedural rule, governs the manner in which process is to be served on the defendants. Hanna v. Plumer, 380 U.S. 460, 85 S.Ct. 1136, 14 L.Ed.2d 8 (1965). By contrast, Fed.R.Civ.P. 3 provides that "[a] civil action is commenced by filing a complaint with the court"; yet this rule of federal procedure does not displace state rules which provide, for example, that for purposes of tolling state statutes of limitations, a suit is not "commenced" until service of summons on the defendant. Walker v. Armco Steel Corp., 446 U.S. 740, 100 S.Ct. 1978, 64 L.Ed.2d 659 (1980); Ragan v. Merchants Transfer Warehouse Co., 337 U.S. 530, 69 S.Ct. 1233, 93 L.Ed. 1520 (1949); Anderson v. Papillion, 445 F.2d 841, 842 (5th Cir. 1971).

Once grasped, these propositions are not very difficult to apply in an ordinary case. Complications arise, however, when (as here) the case has been transferred to the court that must apply these propositions from another federal court in a different state. In such a case, the transferee federal court must decide whether to apply the choice of law rules of the state in which it sits, or the choice of law rules of the state in which the transferor court sits. As discussed below, the answer to this question conceivably could turn upon which party requested the transfer, the reasons behind the transfer, and the statute authorizing the transfer.

B. Choosing Between the Change of Venue Statutes

There are two federal statutes of general application under which a change of venue ordinarily may be made. The first provides as follows:

For the convenience of parties and witnesses, in the interest of justice, a district court may transfer any civil action to any other district or division in which it could have been brought.

28 U.S.C. § 1404(a) (1976). In addition to section 1404(a), which modifies the common-law doctrine of forum non conveniens by making transfer an alternative to dismissal, there is section 1406(a), which provides as follows:

The district court of a district in which is filed a case laying venue in the wrong division or district shall dismiss, or if it be in the interest of justice, transfer such case to any district or division in which it could have been brought.

28 U.S.C. § 1406(a) (1976).

If one reads the statutes to apply only to those instances in which the transferor court has personal jurisdiction over the defendant, sections 1404(a) and 1406(a) would appear to apply to two mutually exclusive situations: section 1404(a) would apply when venue is properly laid in the district in which the transferor court sits, and section 1406(a) would apply in all other instances. The statutes are most frequently used in precisely this manner. Both statutes, however, have been read by the courts in such a broad manner that considerations of venue and personal jurisdiction have become almost inextricably intertwined.

The source of the broad reading of both statutes is Goldlawr, Inc. v. Heiman, 369 U.S. 463, 82 S.Ct. 913, 8 L.Ed.2d 39 (1962). Goldlawr was a private antitrust action filed in federal court in Pennsylvania. Both venue and personal jurisdiction in antitrust cases are governed by section 12 of the Clayton Act, 15 U.S.C. § 22 (1976); when that section was applied to the facts in Goldlawr, however, it was clear that both venue was improper and personal jurisdiction was lacking in Pennsylvania. The district court transferred the case under section 1406(a) to a federal court sitting in New York, where both venue was proper and personal jurisdiction could be obtained. The defendants subsequently contended that the transfer was invalid because the transferor court lacked personal jurisdiction; the transferee court and the Second Circuit agreed.

The Supreme Court reversed, holding that section 1406(a) transfers are not limited to those cases in which the transferor court has personal jurisdiction over the defendants:

The language of § 1406(a) is amply broad enough to authorize the transfer of cases, however wrong the plaintiff may have been in filing his case as to venue, whether the court in which it was filed had personal jurisdiction over the defendants or not. The section is thus in accord with the general purpose which has prompted many of the procedural changes of the past few years — that of removing whatever obstacles may impede an expeditious and orderly adjudication of cases and controversies on their merits.

369 U.S. at 466-67, 82 S.Ct. at 915-16. Justice Harlan dissented. He argued that it was

incongruous to consider, as the Court's holding would seem to imply, that in the "interest of justice" Congress sought in § 1406(a) to deal with the transfer of cases where both venue and jurisdiction are lacking in the district where the action is commenced, while neglecting to provide any comparable alleviative measures for the plaintiff who selects a district where venue is proper but where personal jurisdiction cannot be obtained.

Id. at 468, 82 S.Ct. at 916 (Harlan, J., dissenting) (emphasis in original; footnote omitted). In footnote, Justice Harlan illustrated this proposition by postulating an ordinary diversity case in which venue was proper in the district wherein suit was brought because the plaintiff resided in that district, but personal jurisdiction could not be obtained. "Since this would not be 'a case laying venue in the wrong division or district,' § 1406(a) would be inapplicable." Id. at 468 n.[*], 82 S.Ct. at 916 n.[*]

A great many of the lower federal courts that have subsequently interpreted section 1406(a) have avoided the incongruous results suggested by Justice Harlan by holding that a section 1406(a) transfer is proper in exactly the situation in which Justice Harlan assumed it would not be. The leading case is Dubin v. United States, 380 F.2d 813 (5th Cir. 1967). Dubin was a suit for unpaid taxes that was filed in federal court in Ohio. Although venue was proper in Ohio, since that was the state in which the tax return had been filed and in which the liability had accrued, the Ohio court could not obtain personal jurisdiction over the defendant because he had moved to Florida. Because the government was precluded by the applicable statute of limitations from bringing a new suit in Florida, the federal court in Ohio transferred the case under section 1406(a) to a federal court in Florida. This court concluded that the transfer was authorized by section 1406(a) despite the fact that venue was proper in Ohio:

The statute does not refer to "wrong" venue, but rather to venue laid in a "wrong division or district." We conclude that a district is "wrong" within the meaning of § 1406 whenever there exists an "obstacle [to] . . . an expeditious and orderly adjudication" on the merits. Inability to perfect service of process on a defendant in an otherwise correct venue is such an obstacle.

Id. at 815 (quoting Goldlawr, 369 U.S. at 466, 82 S.Ct. at 915; bracketed portion and ellipsis inserted by Dubin court). Accord, Mayo Clinic v. Kaiser, 383 F.2d 653, 655-56 (8th Cir. 1967); Taylor v. Love, 415 F.2d 1118, 1120 (6th Cir. 1969), cert. denied, 397 U.S. 1023, 90 S.Ct. 1257, 25 L.Ed.2d 533 (1970); Martin v. Stokes, 623 F.2d 469, 474 (6th Cir. 1980). See also Atkins v. Schmutz Manufacturing Co., 401 F.2d 731, 736 n. 7 (4th Cir. 1968) (Craven, J., dissenting) (accepting Dubin's rationale), modified on rehearing en banc, 435 F.2d 527 (6th Cir. 1970), cert. denied, 402 U.S. 932, 91 S.Ct. 1526, 28 L.Ed.2d 867 (1971).

Some commentators have criticized the Dubin rationale for blurring the distinction between improper venue and inability to obtain personal jurisdiction; while agreeing that the results reached under the Dubin rationale are sound, the commentators argue that for purposes of analytical clarity, courts should apply section 1404(a) instead of section 1406(a) when venue is proper in the transferor district but personal jurisdiction over the defendant cannot be obtained.

Various courts, including on occasion this one, have done exactly that. Prior to Dubin, in Koehring Co. v. Hyde Construction Co., 324 F.2d 295, 297-98 (5th Cir. 1963), this court reasoned by analogy from Goldlawr in holding that a section 1404(a) transfer could properly be made despite the fact that the transferor court lacked personal jurisdiction over the defendant. Accord, United States v. Berkowitz, 328 F.2d 358, 361 (3d Cir.), cert. denied, 379 U.S. 821, 85 S.Ct. 42, 13 L.Ed.2d 32 (1964); Smith v. Peters, 482 F.2d 799, 802-03 (6th Cir. 1973), cert. denied, 415 U.S. 989, 94 S.Ct. 1587, 39 L.Ed.2d 886 (1974). Dubin did not cite our previous decision in Koehring, but criticized Berkowitz for relying on section 1404(a) since that section "operates when there are two (or more) forums where a suit could be brought and where it could proceed." 380 F.2d at 816 (emphasis added). Under the Dubin rationale, by contrast, section 1406(a) is available "when[ever] there exists an obstacle — either incorrect venue, absence of personal jurisdiction, or both — to a prompt adjudication on the merits in the forum where [the suit was] originally brought." Id.

Faced with this nearly hopeless muddle of conflicting reasoning and precedent, the Second Circuit, in a scholarly and well-researched opinion that is nonetheless somewhat ambiguous, declined to decide which of sections 1404(a) and 1406(a) is the proper basis for transfer in cases in which the transferor court lacks personal jurisdiction over the defendant. The Second Circuit, reading the two sections together, placed "a judicial gloss on the statutory language, thereby curing Congress' defective draftsmanship"; under that gloss, it was sufficient to hold that a transferor court "'has power to transfer the case even if there is no personal jurisdiction over the defendants, and whether or not venue is proper in [the] district, if a transfer would be in the interest of justice.'" Corke v. Sameiet M.S. Song of Norway, 572 F.2d 77, 80 n. 9 (2d Cir. 1978) (quoting Volk Corp. v. ArtPak Clip Art Service, 432 F. Supp. 1179, 1181 (S.D.N.Y. 1977); bracketed portion inserted by Second Circuit).

Nonetheless, it remains the rule in this circuit that a transfer to a district in which personal jurisdiction over the defendant can be obtained may properly be made under either section 1404(a) or section 1406(a). Aguacate Consolidated Mines, Inc. v. Deeprock, Inc., 566 F.2d 523, 524-25 (5th Cir. 1978) (citing both Dubin and Koehring).

C. Choosing Between the Choice of Law Rules of the States in Which the Transferor and Transferee Courts Sit

As pointed out above in part II- A of this opinion, the process of deciding which state's statute of limitations should be applied in a diversity case becomes considerably more complicated when the district court making that decision is not the district court in which the case was originally filed. If the case has been transferred from a district court in one state to a district court in another state, the transferee court must first decide which state's choice of law rules it should apply in determining which state's statute of limitations should be applied. In some cases the outcome of the litigation may well hinge on this inquiry. Unfortunately, there is precious little guidance from the Supreme Court, and very little consistency among the lower courts and the commentators, on this topic.

The Supreme Court has spoken on the choice of law issue that arises in connection with one type of transfer under section 1404(a). In Van Dusen v. Barrack, 376 U.S. 612, 84 S.Ct. 805, 11 L.Ed.2d 945 (1964), the Court held that when a case that had properly been filed in one district court was transferred under section 1404(a) at the motion of the defendant to another district court in a different state, the transferee court was bound to apply the law that would have been applied by the state courts of the state in which the transferor court sat:

[W]e should ensure that the "accident" of federal diversity jurisdiction does not enable a party to utilize a transfer to achieve a result in federal court which could not have been achieved in the courts of the State where the action was filed. This purpose would be defeated in cases such as the present if nonresident defendants, properly subjected to suit in the transferor State (Pennsylvania), could invoke § 1404(a) to gain the benefits of the laws of another jurisdiction (Massachusetts) . . . . [I]n applying the [reasoning of Erie] to § 1404(a), the critical identity to be maintained is between the federal district court which decides the case and the courts of the State in which the action was filed.

We conclude, therefore, that in cases such as the present, where the defendants seek transfer, the transferee district court must be obligated to apply the state law that would have been applied if there had been no change of venue. A change of venue under § 1404(a) generally should be, with respect to state law, but a change of courtrooms.

Id. at 638-39, 84 S.Ct. at 820-21 (emphasis added; footnotes omitted). The Van Dusen Court specifically reserved, however, the question whether "the same considerations would govern if a plaintiff sought transfer under § 1404(a)." Id. at 640, 84 S.Ct. at 821.

On cases whose facts closely parallel those of Van Dusen — i. e., transfers under section 1404(a) upon the motion of a defendant over whom the transferor court has personal jurisdiction — the lower courts have had comparatively little difficulty in determining the proper statute of limitations to apply. Loughan v. Firestone Tire Rubber Co., 624 F.2d 726, 728-30 (5th Cir. 1980) (per Lynne, District Judge, sitting by designation), is a fine example of how this analysis should be conducted and the difference it can make in the outcome of a case. But what of transfers under section 1404(a) when the transferor court lacks personal jurisdiction over the defendant? What of section 1404(a) transfers made upon motion of the plaintiff, or upon the court's own motion? And what of transfers under section 1406(a)?

Some commentators have suggested that the inverse of the Van Dusen rule should apply when a section 1404(a) transfer is on the plaintiff's motion. "The Court in Van Dusen did not decide what law applies after a transfer on plaintiff's motion, but it seems that the law of the transferee state should control." 15 C. Wright, A. Miller E. Cooper, Federal Practice and Procedure § 3846 (1976) (footnotes omitted). The Sixth Circuit appears to have taken this position at one time, Carson v. U-Haul Co., 434 F.2d 916 (6th Cir. 1970), but apparently has abandoned that view, of which more later. Professor Moore suggests that the analysis should be more complex than a simple determination of which party moved for transfer:

When the venue is proper in the transferor district, and service can be obtained there, there appears to be no technical reason why the transfer may not be made, and whatever dissembling is involved in seeking both the most favorable law and the most convenient forum, the plaintiff does not seek the application of the law of a state that he could not obtain. In such a case, then, it appears that the Van Dusen rules should apply after such a transfer . . . .

When the venue in the district in which the action is brought is proper, but service of process cannot be had there, different considerations apply. In such a situation plaintiff could not maintain his action in the district in which he filed it, and therefore could not take advantage of the law governing that district, since he could not obtain jurisdiction of the defendant. Therefore he should not be permitted to file his action there for the purpose of capturing the law of that jurisdiction for transportation to the jurisdiction in which service can be obtained.

1 J. Moore, Federal Practice ¶ 0.145[4. — 5], at 1608-09 (2d ed. 1980) (footnotes omitted). Professor Moore continues by noting that the courts have generally resisted efforts to "capture" the choice of law rules of a state in which the plaintiff could not have prosecuted his suit to a successful completion, citing Parham v. Edwards, 346 F. Supp. 968, 971-73 (S.D.Ga. 1972), aff'd per curiam, 470 F.2d 1000 (5th Cir. 1973) (affirming "for reasons sufficiently stated by the District Court").

A recent Sixth Circuit opinion interprets sections 1404(a) and 1406(a) in such a manner as to produce precisely the result suggested by Professor Moore. In Martin v. Stokes, 623 F.2d 469 (6th Cir. 1980), that court reaffirmed its previous adoption in Taylor v. Love of the Dubin rationale:

The law in this Circuit, therefore, is that § 1406(a) provides the basis for any transfer made for the purpose of avoiding an obstacle to adjudication on the merits in the district court where the action was originally brought. That defect may be either improper venue or lack of personal jurisdiction. This construction of § 1406(a) necessarily limits the application of § 1404(a) to the transfer of actions commenced in a district court where both personal jurisdiction and venue are proper.

623 F.2d at 474. The court then held that the law to be applied following the transfer of an action depends solely upon the nature of the transfer. Id. In this regard, the court discussed the reasoning behind Van Dusen in concluding that "[s]ince a transfer under § 1404(a) represents only a change in courtrooms for the convenience of the litigants and witnesses, it should not affect the state law governing the action." Id. at 471-72.

Once a plaintiff has exercised his choice of forum under § 1404(a), the state law of that forum should govern the action, regardless of the wisdom of the plaintiff's selection. Thus, no matter who seeks to transfer the action to a more convenient forum under § 1404(a), the state law of the forum in which the action was originally commenced remains controlling.

Id. at 473. The court reached the opposite conclusion with respect to transfers under section 1406(a):

A transfer under § 1406(a) is based not on the inconvenience of the transferor forum but on the impropriety of that forum. If the state law of the forum in which the action was originally commenced is applied following a § 1406(a) transfer, the plaintiff could benefit from having brought the action in an impermissible forum. Plaintiffs would thereby be encouraged to file their actions in the federal district court where the state law was the most advantageous, regardless of whether that district court was a proper forum. . . . Accordingly, we conclude, as have the majority of authorities that have considered this question, that following a transfer under § 1406(a), the transferee district court should apply its own state law rather than the state law of the transferor district court.

Id. at 472 (citations omitted).

Were we not bound by our circuit's prior decisions in Koehring and Aguacate Consolidated Mines, we would be inclined to follow the compelling logic of the Sixth Circuit's decision in Martin v. Stokes. But under those prior Fifth Circuit decisions, which a panel of this court is not free to overrule, section 1404(a) is an appropriate means for transfer when the transferor court cannot obtain personal jurisdiction over the defendants. Nonetheless, we believe that practical results identical to those dictated by Martin v. Stokes can be achieved without doing violence to the prior precedent of our circuit.

In the case at bar, it is impossible to tell from the record before us whether the transfer occurred pursuant to section 1404(a) and Koehring or pursuant to section 1406(a) and Dubin. Accordingly, we must determine which state's choice of law rules should be applied under each.

We fully agree with the analysis of Martin v. Stokes on the question of which state's choice of law rules should be applied after a section 1406(a) transfer. For the reasons stated in that case, we hold that following a section 1406(a) transfer, regardless of which party requested the transfer or the purpose behind the transfer, the transferee court must apply the choice of law rules of the state in which it sits.

For substantially the reasons stated by the Sixth Circuit in Martin v. Stokes and by Professor Moore in the excerpt quoted above, we hold that following a section 1404(a) transfer from a district in which personal jurisdiction over the defendant could not be obtained, the transferee court must apply the choice of law rules of the state in which it sits, regardless of which party requested the transfer. This result, as explained by Professor Moore and the Sixth Circuit, is fully consonant with the purposes underlying the Erie doctrine. Indeed, a contrary result may be constitutionally proscribed, although we express no opinion on that question. See Reyno v. Piper Aircraft Co., 630 F.2d 149, 164-65 (3d Cir. 1980), cert. granted, ___ U.S. ___, 101 S.Ct. 1346, 67 L.Ed.2d 333 (1981). The result of our holding will be to "ensure that the 'accident' of federal diversity jurisdiction does not enable a party to utilize a transfer to achieve a result in federal court which could not have been achieved in the courts of the State where the action was filed." Van Dusen, 376 U.S. at 638, 84 S.Ct. at 820. We are not presented with, and need not resolve, the question of which state's choice of law rules should be applied when, upon the motion of a plaintiff a section 1404(a) transfer has been made from a district in which venue was proper and personal jurisdiction over the defendant had been or could have been obtained.

Applying this holding to the case at bar, we note that it is undisputed that the transferor court in Arkansas was unable to obtain personal jurisdiction over any of the defendants. Therefore, regardless of whether the transfer was pursuant to section 1404(a) or section 1406(a), the applicable choice of law rules are those of the state in which the transferee court sits — i. e., Texas.

D. Texas' Choice of Law Rules

When confronted with a lawsuit in which the substantive law of another jurisdiction is to be applied, Texas courts will most often apply their own state's statute of limitations; this general principle is based on the theory that the foreign jurisdiction's statute of limitations is most often part of its procedural, rather than substantive, law. The exception to this general principle arises when the foreign jurisdiction's statute "creates a right and also incorporates a limitation upon the time within which the suit is to be brought"; in this situation, "the limitation qualifies the right so that it becomes part of the substantive law rather than procedural, and . . . unless suit is brought within the time allowed by [the foreign state's] statute no right of action can be maintained even though the law of [Texas] provides for a longer period of limitations." California v. Copus, 158 Tex. 196, 201, 309 S.W.2d 227, 231, cert. denied, 356 U.S. 967, 78 S.Ct. 1006, 2 L.Ed.2d 1074 (1958). See also Franco v. Allstate Insurance Co., 505 S.W.2d 789, 792 (Tex. 1974); Gaston v. B. F. Walker, Inc., 400 F.2d 671, 673 (5th Cir. 1968) (applying Texas law). See generally R. Weintraub, Commentary on the Conflict of Laws 48-53 (1971); 53 C.J.S. Limitations of Actions §§ 27, 29, 30 (1948 1980 Supp.). But of course, this exception to the general rule that the forum state's statute of limitations governs can have no application to the case at bar if Texas courts would have applied their own state's substantive law had the case been brought originally in the Texas courts. That is, before the Texas courts could conceivably apply the Arkansas statute of limitations, they would first have to find that Arkansas' substantive laws governed the contours of the cause of action.

In an opinion written by then-Texas Supreme Court Justice Sam D. Johnson shortly before he became a member of this court, Texas abolished the common law doctrine of lex loci delicti, under which the law of the place where the wrong occurred dictated the substantive rights of the parties. In place of lex loci delicti, Texas explicitly adopted the "most significant relationship test" as enunciated in the Restatement (Second) of Conflicts. Gutierrez v. Collins, 583 S.W.2d 312, 318-19 (Tex. 1979).

Applying this test to the case at bar, we are convinced that the Texas courts would apply their own state's substantive law in defining the relative rights of the parties. In both quantitative and qualitative terms, Texas bears a more significant relationship to this litigation than does any other forum. Both the injury and the conduct that allegedly gave rise to the injury occurred in Texas; the business relationship between the parties was centered in Texas; and the defendants have a place of business there. By contrast, the only contact that Arkansas has with the matter is through the plaintiff's residence there. We hold that Texas' substantive law governs both causes of action; accordingly, there is no question but that Texas courts would apply their own state's statute of limitations.

The district court erred, then, in assuming that Arkansas' three-year statute of limitations applied. We therefore need not decide whether the court was correct in its interpretation of Arkansas law. The question remains, however, whether Ellis' suit is barred under Texas law.

III. APPLYING THE TEXAS STATUTE OF LIMITATIONS

The parties agree that if Texas law applies, the applicable limitations period is that contained in Tex.Rev.Civ.Stat.Ann. art. 5526 (Vernon Supp. 1980), which provides in pertinent part as follows:

There shall be commenced and prosecuted within two years after the cause of action shall have accrued, and not afterward, all actions or suits in court of the following description:

. . . .

4. Action for injury done to the person of another.

5. Action for injury done to the person of another where death ensued from such injury; and the cause of action shall be considered as having accrued at the death of the party injured.

The parties agree that both the wrongful death and survival causes of action accrued on July 29, 1976. Ellis' first complaint was filed on July 28, 1978 — one day short of two years after the causes of action accrued.

It is absolutely clear, however, that under Texas law, "the mere filing of a suit will not interrupt or toll the running of a statute of limitation; . . . to interrupt the statute, the use of diligence in procuring the issuance and service of citation is required." Zale Corp. v. Rosenbaum, 520 S.W.2d 889, 890 (Tex. 1975). Accord, Rigo Manufacturing Co. v. Thomas, 458 S.W.2d 180, 182 (Tex. 1970); Buie v. Couch, 126 S.W.2d 565, 566 (Tex.Civ.App.-Waco 1939, writ ref'd). The diligence that must be shown by the plaintiff is the same as that which would have been exercised by an ordinarily prudent person under the same or similar circumstances. E.g., Hamilton v. Goodson, 578 S.W.2d 448, 449 (Tex.Civ.App. — Houston [1st Dist.] 1979, no writ); McGuire v. Federal Deposit Insurance Corp., 561 S.W.2d 213, 215 (Tex.Civ.App.-Houston [1st Dist.] 1977, no writ); Beavers v. Darling, 491 S.W.2d 711, 714 (Tex.Civ.App.-Waco 1973, no writ).

The reasonableness of a plaintiff's delay in procuring the issuance and service of citation is usually a question of fact. E. g., Hamilton, 578 S.W.2d at 449; Beavers, 491 S.W.2d at 714. Here, the district court has made no finding of fact as to Ellis' diligence. SFI and SFOT, however, urge that under Texas law, the delay in procuring the issuance and service of valid process was such as to make Ellis nondiligent as a matter of law. The key Texas case upon which they rely is Rigo Manufacturing Co. v. Thomas, supra.

The plaintiffs in Rigo first filed their wrongful death and survival action in federal district court on June 24, 1966. That suit, however, was dismissed by the federal court for lack of jurisdiction on October 4, 1966. The plaintiffs then filed in state court on October 14, 1966 — two years, three months, and sixteen days after the causes of action accrued (June 28, 1964). After filing in state court, the plaintiffs served Rigo with nonresident notice under Tex.R. Civ.P. 108 on October 21, 1966; Rigo promptly challenged the sufficiency of service under this rule, but the trial court did not sustain Rigo's plea in abatement until March 28, 1968. Immediately thereafter, the plaintiffs secured the issuance and service of proper citation.

The court of civil appeals held that the limitations period specified in article 5526 had been tolled under the terms of the Texas saving statute, Tex.Rev.Civ.Stat. Ann. art. 5539a (Vernon 1958), which provides that when an action is dismissed from any trial court for want of jurisdiction but is refiled in the proper court within sixty days, the period between the date of the first filing and that of commencement in the proper court shall not be counted as part of the limitations period unless the defendant shows that the first filing was in intentional disregard of jurisdiction. The Texas Supreme Court reversed, holding that as a matter of law, the plaintiffs had been nondiligent in obtaining the issuance and service of citation in the state-court suit, and that therefore that suit had not been "commenced" within the required sixty days under article 5539a. The court first noted that it had previously held that a delay of eight and one-third months demonstrated a lack of diligence as a matter of law. It then noted that in the lawsuit against Rigo,

there was a delay of some seventeen and one-half months [ after filing suit in state court] in securing the issuance and service of proper and effective citation, so that at the time of issuance and service thereof a period of three years and nine months had elapsed since accrual of the causes of action.

. . . [T]he inadequacy of the [citation under Tex.R.Civ.P. 108] was promptly called to [the plaintiffs'] attention by Rigo's plea to the jurisdiction. . . . Thus put on notice of the inadequacy of the Rule 108 citation, [the plaintiffs] nevertheless waited some seventeen months before procuring issuance and service of proper citation.

458 S.W.2d at 182.

SFI and SFOT would have us hold in the case at bar that the motion to dismiss which they filed in the federal court sitting in Arkansas similarly should have put Ellis on notice of the inadequacy of the substituted service on the Arkansas Secretary of State; they would further have us hold that under Rigo, Ellis' delay of over 500 days in obtaining valid service after that motion was filed should demonstrate that as a matter of law, Ellis lacked diligence in obtaining issuance and service of process.

We cannot agree with SFI and SFOT that Rigo and other Texas cases to similar effect directly control the case at bar. These cases, taken together, stand for the proposition that an unexplained delay of as little as six months demonstrates as a matter of law that the plaintiff lacked diligence in obtaining the issuance and service of process. When there is an excuse offered that if proved would negate the inference of nondiligence, however, Texas law is equally clear that nondiligence cannot be determined as a matter of law, but instead must be the subject of factual findings. E. g., Beavers v. Darling, 491 S.W.2d 711, 714 (Tex.Civ.App.-Waco 1973, no writ) (reversing summary judgment, despite delay of over three years in obtaining valid service of process, on grounds that plaintiff's proffered excuse that the defendant had moved from place to place in a large city during this period, and that this had defeated plaintiff's multiple efforts to obtain valid service of process, raised a factual issue as to his diligence). Here, there is at least some hint of an explanation for the delay in the mysterious docket entry that refers to the "clerk's error" in failing to issue process in August. In these circumstances, we are loath to hold that Ellis has been nondiligent as a matter of law. Neither can we say that SFI and SFOT have met their burden under Fed.R.Civ.P. 56 in demonstrating that there is no genuine issue of fact as to Ellis' diligence. Accordingly, we must reverse the summary judgment rendered by the district court and remand for further proceedings on the question of Ellis' diligence in obtaining the issuance and service of process.

The question remains, however, whether and to what extent the district court on remand should consider the time during which Ellis' suit was pending in the federal court in Arkansas. As a practical matter, there can be no direct guidance from the Texas courts on this question, for there is no direct analog in Texas practice to the interstate transfers that are possible in federal practice under sections 1404(a) and 1406(a). Nonetheless, we think that the general rule in Texas — i. e., that to toll the running of the statute of limitations, the plaintiff must demonstrate the diligence in procuring the issuance and service of citation that would have been exercised by an ordinary prudent person under the same or similar circumstances — is flexible enough to encompass the situation presented by the case at bar.

In the case at bar, for example, the "circumstances" to be considered in determining Ellis' diligence would include those surrounding his initial filing of suit in the federal court sitting in Arkansas and the circumstances surrounding the transfer to the federal court in Texas. With respect to the period of time during which his action was pending in the transferor court, it would be appropriate to consider whether Ellis' filing suit in Arkansas rather than elsewhere was in intentional disregard of the fact that he could not obtain personal jurisdiction over the defendants. It would also be appropriate to consider whether Ellis acted promptly in moving for and securing a transfer under section 1404(a) or section 1406(a) once the transferor court had determined that it lacked personal jurisdiction over the defendants. But by the same token, the fact that Ellis may have resisted the defendants' challenge to the transferor court's personal jurisdiction should not be considered as an indication of nondiligence, at least so long as Ellis' resistance was in good faith. With respect to the period of time after the transfer had been completed, the inquiry would shift to whether Ellis was diligent in seeking the valid issuance and service of process in the new forum.

For the foregoing reasons, we reverse and remand. If it is determined that Ellis was not diligent, his suit is time-barred under article 5526; if he was diligent, the litigation may proceed.

REVERSED and REMANDED.