dissenting:
This case presents an important issue of the proper procedural approach to take in deciding § 301 preemption issues under the special circumstances presented by an unsettled condition of state law. That problem has not been definitively addressed either by the Supreme Court or this court, and it is a problem that will undoubtedly recur. For that reason and because I disagree with the majority’s approach and consequently with the result it reaches, I respectfully dissent.
In general terms I disagree with the majority’s perception that in such a situation the federal courts should resolve the unsettled question of the pendent state-law claim’s cognizability under state law “before resolving the § 301 preemption inquiry.” At 958. For reasons that follow, I think this is not the appropriate procedure; that instead the federal courts may and should first decide the preemption issue, reserving the question of how and by what court the state-law claim should be resolved if it is found not preempted.
I
The issue posed for a federal court by the assertion that a union employee’s state-law claim against his or her employer is preempted by § 301 of the LMRA is whether “resolution of [the] state-law claim depends upon the meaning of a collective-bargaining agreement.” Lingle v. Norge Div. of Magic Chef, Inc., — U.S. at-, 108 S.Ct. 1877, 1881, 100 L.Ed.2d 410 (1988). In the course of clarifying the proper way to address that issue by focussing upon the state claim’s “independence,” id. at 1882 & n. 7, the Lingle Court illustrated the obvious point that to answer the ultimate issue the federal court has first to identify the substantive elements of the state-law claim. Id. Only so can the ultimate issue whether that claim can be resolved independently of the collective bargaining agreement’s meaning be addressed.
When, as in Lingle itself, and as will ordinarily be the case, the specific state-law claim is one that has been clearly recognized, fleshed out, and applied in state law, that first step is simple. As did the Lingle Court, a court merely looks to the relevant sources — state statutes and judicial decisions — defining and applying the elements of the claim. See id. at 1881-82. It can then address the next, and ultimate, legal question, whether proof of any of those elements will depend upon what a collective bargaining agreement might mean in relevant part.1
*965The problem here is of course that we do not have that ordinary case. Here there is a prior question — and a genuine one— whether the state-law claim as pleaded is one cognizable under state law. The problem this presents is one of first impression with us and there is no direct Supreme Court precedent to guide us.2 It is upon the way that this unusual problem should be handled that I differ from the majority and would reach a different result than the majority reaches.
The majority’s solution is for the federal court to decide the open question of state law — whether the state-law claim as pleaded is one cognizable under state law — as an appropriate exercise of its “discretion.” At 958.3 On this view of things, if the determination is that no such claim exists under state law, two things result. First, the decision operates as a final judgment dismissing the state claim on the merits. Second, it avoids any need to address the ultimate preemption issue since it has now been determined that there is no state claim to be preempted. That of course is the result in fact reached by the majority here. Presumably, under this approach, if the determination were to the contrary— that the state-law claim was one recognized in state law — the normal preemption inquiry into the “independence” of the claim as pleaded would then proceed, with preemption turning on the outcome.
As the majority tacitly recognizes, this approach is not legally compelled.4 As indicated, I do not think it is the most appropriate and legally sound approach available. The more appropriate approach where, as here, the state-law claim is a pendent claim, would be to reserve its resolution while deciding the preemption issue, reserving the option to dismiss the claim without prejudice under the traditional principles of United Mine Workers v. Gibbs, 383 U.S. 715, 86 S.Ct. 1130, 16 L.Ed. 2d 218 (1966), if it be found not preempted.5 To follow this course requires no more than a simple adaptation of the § 301 preemption inquiry to the special problem presented when the state-law claim whose preemption is in issue is one whose existence or exact scope is not established under state law.
The first thing to realize is that though a federal court in that circumstance cannot avoid deciding the preemption issue, it does not have to decide the open question of *966state law and resolve the state claim to do so. All of the purposes of the § 301 preemption inquiry can be achieved and all the relevant federal interests can be served by simply assuming without deciding that the claim as pleaded is a viable one under state law. The inquiry can then proceed in the normal way, looking to the claim as pleaded for its elements, and asking whether proof of any of those elements will depend on an interpretation of the collective bargaining agreement. If the answer is “yes,” the state-law claim as pleaded is preempted whether or not cognizable under state law, and that open question therefore need never be addressed. If the answer is “no,” so that preemption does not result, the court is simply then in the position, frequently encountered, of having before it a state-law pendent claim which in its discretion it may decide on the merits or dismiss without prejudice under guidance of the Gibbs factors.
The virtue of this approach is simply that it preserves the option of abstaining from resolution of a pendent state-law claim found not to be preempted rather than cutting off that option by undertaking federal resolution of the claim under circumstances where federal abstention is the more appropriate course. Because, as indicated, threshold federal resolution is not forced by the need to decide the preemption issue, the option to abstain if preemption is not found obviously may and should be preserved in order to serve the interests of comity and federalism that underlie Gibbs.
Though, as indicated, neither the Supreme Court nor this court has directly addressed this problem, several other courts of appeals recently have confronted it. The approaches taken by those courts, though differing in detail, generally support that one here proposed. None of the courts, so far as I can tell, has taken the approach of the majority here which, in effect, finesses the preemption issue entirely by rejecting the state claim on the merits at the very threshold, “before resolving the § 301 preemption inquiry.” At 958. All, instead, complete the preemption inquiry without purporting to decide the state-law claim on the merits.
Three different approaches have been taken by these courts. In one, the general cognizability of the claim as pleaded under state law has been assumed, but its unsettled contours in state law have been “hypothesized” by a detailed analysis of relevant state law sources. See, e.g., Jackson v. Liquid Carbonic Corp., 863 F.2d 111 (1st Cir.1988) (“independence” of novel state-law claim challenging employee drug testing analyzed by “hypothesizing” “parameters” of state privacy law).
In another, the cognizability of the claim as pleaded has been assumed, without any effort to confirm either its existence or its contours by “finding” or even “hypothesizing” the likely state of state law. See, e.g., United Ass’n . of Journeymen and Apprentices of the Plumbing and Pipe Fitting Industry, Local No. 57 v. Bechtel Power Corp., 834 F.2d 884, 889 (10th Cir.1987) (where state courts had not addressed either the existence or the elements of a “novel” state-law claim, preemption inquiry conducted as “if such a cause of action does exist”).
Under both of these approaches, resolution of the state-law claim on the merits has been avoided pending outcome of the preemption inquiry. Where preemption of the claim as assumed or hypothesized has been found, that has of course disposed of the state-law claim. See Jackson, 863 F.2d at 121. Where preemption has not been found, the possibility of deference to state court resolution has been preserved. See Bechtel, 834 F.2d at 890 (removed state claim remanded after being found not preempted).
In a third approach — the most significant because it involves a situation in which the federal court actually believed, as does the majority here, that the state-law claim as pleaded was not one cognizable under state law — the court nevertheless did not dismiss it on the merits, but instead found it necessarily preempted as being only cognizable under § 301. See, e.g., Laws v. Calmat, 852 F.2d 430, 434 n. 5 (9th Cir.1988) (because novel state-law claim not “an estab*967lished or recognized state-law claim,” necessarily preempted by § 301).
The common perception underlying all these approaches is that advanced here as the appropriate one. It is this. A federal court confronted with the contention that a removed or pendent state-law claim is preempted by § 301 should decide the preemption issue without regard to the perceived merits of the state-law claim, the matter of that claim’s proper disposition under state law being presented only if it is first found not preempted. Where the state-law claim is clearly recognized and its contours established in state law, established state law is the proper reference for making the preemption inquiry. Where the contours or very existence of the state-law claim is not clearly established in state law sources, either the claim as pleaded or as “hypothesized” from available state law sources is the proper reference. Even where the state-law claim is found to be one not recognized in state law, it should nevertheless not be rejected on the merits under state law, but found perforce preempted.6
II
It remains to indicate how I would apply this approach here.
A
I would proceed on a hybrid of the approaches taken by the Tenth Circuit in Bechtel and the First Circuit in Jackson. That is to say, I would simply assume for purposes of the preemption inquiry that the state-law claim here, though “novel” in the sense that it has not been specifically recognized in state law, is a cognizable one. As the majority’s extensive effort to find state law (or non-law) on the subject demonstrates, the existence of a general tort claim for retaliatory discharge is well-established in West Virginia decisional law. See at 962-63. The only open question (and of course it is potentially dispositive on the merits of the state claim if it is not preempted) is whether the West Virginia courts would recognize the specific “public policy principle” whose violation is alleged as an essential element of the claim here in issue. For perfectly obvious reasons, that is a question that should if at all possible be left for first instance resolution by the state courts. As indicated above, the option to do so can readily be preserved while making the preemption inquiry on the assumption of the specific claim’s cognizability-
Where, as here, one can indeed identify the exact contours of the state-law claim, assuming only its cognizability under the particular facts alleged, the preemption inquiry into that claim’s “independence” under Lingle is relatively easy and straightforward. If this claim is cognizable, its proof would require a showing (1) that the claimant was discharged (2) in retaliation (3) for conduct protected by a “substantial public policy principle” expressed in state law. See, e.g., Collins v. Elkay Mining Co., 371 S.E.2d 46 (W.Va.1988); McClung v. Marion County Comm’n, 360 S.E.2d 221 (W.Va.1987); Cordle v. General Hugh Mercer Corp., 325 S.E.2d 111 (W.Va.1984); Shanholtz v. Monongahela Power Co., 165 W.Va. 305, 270 S.E.2d 178 (1980); Harless v. First National Bank in Fairmont, 162 W.Va. 116, 246 S.E.2d 270 (1978).
*968With the elements of the claim as pleaded so identified, it is plain that under the Lingle analysis, any particular claim of retaliatory discharge under West Virginia law may be resolved independently of and without reference to any relevant collective bargaining agreement. The elements of the state tort could in no way be dependent upon or determined by any rights or duties imposed by such an agreement. Any duty not to discharge under state law will be found (or found not to exist) by reference to a specific public policy principle grounded exclusively in state law sources. See Lingle, 108 S.Ct. at 1881-82 (comparable analysis of independence of Illinois retaliatory discharge claim). The fact that resolution of the state-law claim (including determination of its existence vel non) may overlap some of the factual inquiry conducted in resolution of a parallel grievance under the collective bargaining agreement is irrelevant. See id. at 1883.
B
On this analysis, I would therefore find the state claim as pleaded not preempted by § 301, and turn to the question of how we should now deal with that sole remaining claim.
Despite the determination that it is not preempted, we of course would still have jurisdiction to decide it as a pendent state claim.7 I would not do that, however, because it is classically the kind of pendent state claim that should be dismissed without prejudice, leaving its resolution (if the claimant desires to pursue it) to the state courts. All the Gibbs criteria for this disposition are present. As the Supreme Court has recently emphasized:
Under Gibbs, a federal court should consider and weigh in every case, and at every state of the litigation, the values of judicial economy, convenience, fairness, and comity in order to decide whether to exercise jurisdiction over a case brought in that court involving pendent state-law claims. When the balance of these factors indicates that a case properly belongs in state court, as when the federal-law claims have dropped out of the lawsuit in its early stages and only state-law claims remain, the federal court should decline the exercise of jurisdiction by dismissing the case without prejudice.
Carnegie-Mellon University v. Cohill, 484 U.S. 343, 108 S.Ct. 614, 618-19, 98 L.Ed.2d 720 (1988).
Here, all the federal claims have been finally dismissed by the district court. The pendent state claim alone remains. Dismissal of the federal claim came early in the litigation, by summary judgment. Compelling concerns of comity counsel abstention by the federal courts from deciding the important, unsettled issue of state law on which this claim is based.8 The claimant *969faces no statute of limitations problems because of the saving provisions of W.Va. Code § 55-2-18, interpreted in Stare v. Pearcy, 617 F.2d 43, 46 (4th Cir.1980). No federal interest, other than general interests in repose and finality, counsel federal resolution of this state-law question. Here those interests simply do not outweigh the more specific interests of the claimant and the state courts in having that decision made in the first instance by the state courts.
I would reverse the district court’s dismissal of the state-law claim on the merits, and remand with directions to dismiss the claim without prejudice.
. A process that could, but need not necessarily, also require looking at the collective bargaining agreement far enough to determine how it might define or give scope to rights or duties integral to the state claim. This would seem to be required only when that necessity is manifest from an analysis of the essential elements of the state-law claim. See Lingle, 108 S.Ct. at 1881 (describing this as the analytical process required in Allis-Chalmers Corp. v. Lueck, 471 U.S. 202, 105 S.Ct. 1904, 85 L.Ed.2d 206 (1985)).
. In each of the principal Supreme Court cases dealing with the preemption of removed or pendent state-law claims, Lingle; Caterpillar, Inc. v. Williams, 482 U.S. 386, 107 S.Ct. 2425, 96 L.Ed.2d 318 (1987); IBEW, AFL-CIO v. Heckler, 481 U.S. 851, 107 S.Ct. 2161, 95 L.Ed.2d 791 (1987); Allis-Chalmers Corp. v. Lueck, 471 U.S. 202, 105 S.Ct. 1904, 85 L.Ed.2d 206 (1985), the state-law claim directly in issue was one whose existence and essential elements were indisputably established under state law and could simply be referred to by the federal court as there defined.
. There is no doubt of the federal court’s raw power to decide the claim on the merits where, as here, the state-law claim is a pendent claim in an original federal action. As to such a state-law claim jurisdiction would exist irrespective of the preemption decision: under the pendent jurisdiction if not preempted; under federal question jurisdiction if transmuted into a federal claim by preemption. (The same of course would be true with respect to a state-law claim removed on diversity grounds, see, e.g., Lingle, 108 S.Ct. 1877.) The question is not whether there is jurisdiction to resolve the state-law claim — there is, one way or the other — but whether the power should be exercised here. The majority is simply mistaken in saying, slip op. at 13, that this dissenting opinion "asserts that it is beyond the discretion of a federal district court to dismiss the state-law claim with prejudice.” Central to the whole argument here is the recognition that the power to do so exists, but this is coupled with the contention that the option of exercising it should simply be reserved pending prior resolution of the preemption inquiry. Indeed, the majority seems largely to miss the significance of my obvious recognition that such "discretion" exists throughout its extended discussion of Gibbs discretion and the proper time to exercise it.
. See note 3 supra.
. The same approach would be appropriate in the case of a removed state-law claim, with remand to the state court being the option retained in case the claim is found not preempted. See, e.g., Caterpillar, 107 S.Ct. 2425; United Ass’n of Journeymen and Apprentices of the Plumbing and Pipe Fitting Industry, Local No. 57 v. Bechtel Power Corp., 834 F.2d 884 (10th Cir.1987); Paige v. Henry J. Kaiser Co., 826 F.2d 857 (9th Cir.1987); Baldracchi v. Pratt & Whitney Aircraft Div., 814 F.2d 102 (2d Cir.1987). See also Jackson v. Liquid Carbonic Corp., 863 F.2d 111, 124 (1st Cir.1988) (Bownes, J., dissenting).
. My own strong preference in general would be for the Bechtel approach, which avoids any effort to resolve unsettled questions of state law going either to the existence vel non or the exact contours of the state-law claim as pleaded. Where, however, the claim’s general cognizability is established in state law and the only unsettled questions relate to its applicability to the particular circumstances pleaded or to its exact contours, reference to relevant aspects of established state law might certainly be appropriate and helpful in conducting the preemption inquiry, the course taken in Jackson. I would depart from the general approach of assuming general cognizability, and possibly "hypothesizing" unsettled contours, only in the face of a clear indication that the state-law claim had been affirmatively rejected in state law, i.e., I would never find it not cognizable merely for lack of affirmative recognition as did Calmat. And even where it had been affirmatively rejected, I would follow the Calmat approach of finding the state-law claim thereby perforce preempted rather than rejecting it on the merits. The ordering principle is to leave the state law question as open and undisturbed as is possible while necessarily deciding the federal preemption issue.
. See note 3 above.
. Because it bears upon the propriety of Gibbs abstention, I should note that on my assessment of West Virginia law, it is by no means as clear as the majority asserts that Washington “fail[s] to state a valid cause of action for retaliatory discharge under West Virginia law." Slip op. at 14. First, the Supreme Court of West Virginia has on occasion indicated that in searching for articulations of "public policy" as an essential element of "retaliatory discharge” claims, the courts are not confined to statutory sources.
"The sources determinative of public policy are, among others, our federal and state constitutions, our public statutes, our judicial decisions, the applicable principles of the common law, the acknowledged prevailing concepts of the federal and state governments relating to and affecting the safety, health, morals and general welfare of the people for whom government — with us — is factually established."
Cordle v. General Hugh Mercer Corp., 325 S.E.2d 111, 114 (WIVa.1984) (quoting Allen v. Commercial Casualty Ins. Co., 37 A.2d 37, 38-39 (N.J.1944)). In Cordle, for example, the court ruled that it was against state public policy for an employer to require an employee to submit to a polygraph test as a condition of employment. Id. at 117. The court noted that a statute recently had been passed to similar effect, see W.Va. Code § 21-5-5b, but observed that this statute was not dispositive as it was promulgated after the period in question before the court. Cordle, 325 S.E.2d at 113 n. 6. Instead, the court held that the statute itself was "the embodiment of a 'recognized facet of public policy’ in this State.” Id. at 117 (citation omitted).
Second, it is not plain, in any event, that no statutory source exists in West Virginia law to support Washington’s claim. The majority fails to note the existence of West Virginia's statute on employment safety, which provides in relevant part:
*969Every employer shall furnish employment which shall be reasonably safe for the employees therein engaged and shall furnish and use safety devices and safeguards, and shall adopt and use methods and processes reasonably adequate to render employment and the place of employment safe, and shall do every other thing reasonably necessary to protect the life, health, safety, and welfare of such employees....
W.Va.Code § 21-3-1. Federal and West Virginia state courts have several times recognized and applied this statute as one imposing, as a matter of state policy, an obligation upon employers to maintain reasonably safe employment conditions. See, e.g., Hamrick v. Aerojet-General Corp., Indus. Systems Div., 528 F.2d 65 (4th Cir.1975); Davis v. Monsanto Co., 627 F.Supp. 418, 422 (S.D.W.Va.1986); Pack v. Van Meter, 354 S.E.2d 581, 585 (W.Va.1986) (citing cases). It seems to me no great leap to find in it a public policy that would be violated by an employer’s conduct that retaliated for an employee’s effort to enforce that obligation.
That the question of the claim’s cognizability is therefore not only unsettled but obviously arguable under extant state law is, as indicated, irrelevant to a proper resolution of the preemption issue. But it is highly relevant to the question of abstention from federal resolution of that unsettled issue of state law, and militates heavily in favor of abstention under Gibbs.