REINHARDT, Circuit Judge:

I. FACTUAL BACKGROUND

Appellant Irma Molsbergen is the widow of John Martens Molsbergen, a former pilot with the U.S. Navy, who died from cancer on January 8, 1981. Subsequent to Mr. Molsbergen's death, appellant filed administrative claims with the Department of Defense, the Department of the Navy, the Veterans Administration and the Nuclear Regulatory Commission. These claims were denied on January 20, 1983.

On June 16, 1983, appellant filed a complaint against the United States Government, among others, pursuant to the Federal Tort Claims Act ("FTCA"), 28 U.S.C. § 1346(b) (1982), and the United States Constitution. Appellant alleged that Mr. Molsbergen's death resulted from his exposure to radiation subsequent to the 1945 detonation of an atomic bomb over Nagasaki, Japan.

In count one of her complaint, appellant alleged that while Mr. Molsbergen was on active duty in the U.S. Navy, he was ordered to participate in a mission "at Nagasaki, Japan immediately after [the] detonation." Appellant alleged that in the course of the mission, Mr. Molsbergen was exposed to ionizing radiation which caused his eventual death. Appellant further alleged that subsequent to Mr. Molsbergen's discharge from the Navy, the government learned of certain dangers related to radiation exposure. Appellant alleged that by failing to warn Mr. Molsbergen of these dangers once it became aware of them, the government breached a legal duty which resulted in Mr. Molsbergen's death.

In count two of her complaint, appellant alleged that the government intentionally exposed Mr. Molsbergen to ionizing radiation without obtaining his consent. Contrary to the averments contained in count one, appellant alleged that the government was fully aware of the danger involved with such exposure when it ordered Mr. Molsbergen to participate in the mission. Appellant further alleged that the government ordered Mr. Molsbergen to participate in the mission for purposes of human experimentation.

In count three of her complaint, appellant alleged that the government's conduct deprived Mr. Molsbergen of certain constitutional protections. Specifically, appellant alleged that the government violated Mr. Molsbergen's rights to life and liberty, secured by the Fifth Amendment to the United States Constitution.

The district court dismissed all three counts of the complaint for lack of subject matter jurisdiction. Relying principally on Feres v. United States, 340 U.S. 135, 71 S.Ct. 153, 95 L.Ed. 152 (1952), the district court concluded that the FTCA's waiver of sovereign immunity did not extend to appellant's claim for in-service human experimentation (count two). The district court also held that Feres barred appellant's claim for the government's failure to warn Mr. Molsbergen subsequent to its learning that his life or health were in jeopardy (count one). The district court interpreted count two of appellant's complaint as an admission against count one, concluding that since the allegations contained in count one conflicted with those contained in count two, appellant had merely attempted "to divide one in-service wrongful act continuing past discharge into pre- and post-discharge wrongful acts." Finally, the district court concluded that Feres barred appellant's constitutional claim (count three).

In this appeal, appellant contests only the dismissal of count one of her complaint. Because we conclude that the district court erred in its application of the relevant law, we reverse and remand the matter for reinstatement of that count.

II. APPLICABILITY OF THE FERES DOCTRINE [10] A. Appellant's Inconsistent Averments

As a threshold matter, we note that the district court erred in construing count two of appellant's complaint as an admission against count one and in concluding that count one was therefore barred by Feres. Indeed, to permit such a construction would undermine the clear intent of the Federal Rules of Civil Procedure, which explicitly authorize litigants to present alternative and inconsistent pleadings. Pursuant to Rule 8(e)(2), "[a] party may set forth two or more statements of a claim or defense alternatively or hypothetically." The Rule further provides that "[a] party may also state as many separate claims or defenses as he has regardless of consistency." Id. Clearly, a policy which permits one claim to be invoked as an admission against an alternative or inconsistent claim would significantly restrict, if not eliminate, the freedom to plead inconsistent claims provided by Rule 8(e)(2). Thus, courts have been reluctant to permit one pleading to be read as a judicial or evidentiary admission against an alternative or inconsistent pleading. See Douglas Equipment, Inc. v. Mack Trucks, Inc., 471 F.2d 222 (7th Cir. 1972); Continental Insurance Co. v. Sherman, 439 F.2d 1294 (5th Cir. 1971); Giannone v. United States Steel Corp., 238 F.2d 544 (3d Cir. 1956); McCormick, Evidence § 265 (2nd ed. 1972). Cf. Ryan v. Foster and Marshall, 556 F.2d 460, 463 (9th Cir. 1977) (plaintiffs' assertion of inconsistent and alternative claims may not be construed as a waiver by plaintiffs of their rights to recovery under either claim).

In light of the liberal pleading policy embodied in Rule 8(e)(2), we hold that a pleading should not be construed as an admission against another alternative or inconsistent pleading in the same case under the circumstances present here. Shipek v. United States, 752 F.2d 1352, 1356 (9th Cir. 1985). Thus, in this case, the district court should have examined counts one and two of appellant's complaint independently. Properly analyzed, count one of appellant's complaint alleges that subsequent to Mr. Molsbergen's discharge, the government learned of a risk to which it had exposed Mr. Molsbergen and negligently failed to warn him of prospective harm.

B. Applicability of Feres to Appellant's Allegation of Post-Discharge Negligence

In Feres v. United States, 340 U.S. 135, 71 S.Ct. 153, 95 L.Ed. 152 (1950), the Supreme Court held that "the Government is not liable for injuries to servicemen where the injuries arise out of or are in the course of activity incident to service." Id. at 146. However, as we have previously held, Feres does not bar a claim for a post-discharge failure to warn. In Broudy v. United States, 661 F.2d 125 (9th Cir. 1981) (" Broudy I"), this court addressed an issue identical to that raised by appellant in count one of her complaint. At issue in Broudy I was the validity of a cause of action under the FTCA for post-discharge injuries sustained by a former member of the armed forces as a result of his exposure to radiation while on active duty. Although we concluded that a claim predicated on the government's decision to expose the veteran to radiation was barred by Feres, we reasoned that under certain circumstances, a claim for injuries resulting from the government's failure to warn the veteran of the danger associated with radiation exposure could properly be asserted under the FTCA. Specifically, we held that

if the Appellant can allege and prove an independent post-service negligent act on the part of the Government, her claim would be cognizable under the FTCA. The Government's failure to warn Major Broudy of and monitor any possible injuries arising from his exposure to radiation might constitute such an act if the Government learned of the danger after Major Broudy left the service.

Id. at 129.

In count one of her complaint, appellant has alleged that Mr. Molsbergen was exposed to radiation while he was in military service; that the government learned of the potential effects of such exposure subsequent to Mr. Molsbergen's discharge; that the government breached a duty to warn Mr. Molsbergen of these potential effects once it became aware of them; and that Mr. Molsbergen's death resulted from breach of the government's duty to warn. Thus, appellant's allegations are limited to the government's conduct subsequent to Mr. Molsbergen's discharge. Because the allegations contained in count one of appellant's complaint meet the requirements we articulated in Broudy I, we conclude that she has stated a cause of action which is not barred by Feres and which is cognizable under the FTCA. See Shipek v. United States, 752 F.2d 1352, 1356 (9th Cir. 1985); Seveney v. United States, 550 F. Supp. 653, 660 (D.R.I. 1982); Everett v. United States, 492 F. Supp. 318, 325-26 (S.D.Ohio 1980); Thornwell v. United States, 471 F. Supp. 344, 352 (D.D.C. 1979).

III. EXISTENCE OF A DUTY TO WARN

Having concluded that appellant's claim is not barred by Feres, we must next determine whether, upon learning of the dangers associated with radiation exposure, the government incurred a duty to warn Mr. Molsbergen. The government may be held liable for claims brought pursuant to the FTCA only if a private person would, under similar circumstances, be liable pursuant to the law of the place where the act or omission occurred. 28 U.S.C. § 1346(b) (1976). Here, the alleged wrongdoing occurred in California, the state in which appellant and her husband were domiciled. We therefore inquire into whether, under California law, a private party that enjoyed a relationship with Mr. Molsbergen analogous to that which the government enjoyed, would have a duty to warn Mr. Molsbergen of the dangers to which he had been exposed. See Broudy v. United States, 722 F.2d 566, 569 (9th Cir. 1983) ( "Broudy II") (determination of whether a duty exists which is actionable under the FTCA involves application of the law of the state where the act or omission occurred).

In determining whether California law would impose a duty upon the government to warn Mr. Molsbergen of the danger to which he had been exposed, the district court appropriately drew an analogy to the relationship between private employers and their former employees. The district court was unable to identify any controlling law which definitively imposed a duty on private employers to warn former employees where, subsequent to termination of the employment relationship, the employer obtained knowledge of a potential danger to which the former employee had been exposed. Consequently, the district court held that no such duty exists. However, as appellant has pointed out, there exists no case which prohibits the imposition of such a duty. Thus, all that we may reasonably infer from this absence of case law is that California courts have not yet addressed this issue.

When a decision turns upon applicable state law, and the state's highest court has not adjudicated the issue, the district court must make a reasonable determination, based upon such recognized sources as statutes, treatises, restatements and published opinions, as to the result that the highest state court would reach if it were deciding the case. See Commercial Union Insurance Co. v. Ford Motor Co., 640 F.2d 210, 212 (9th Cir.) cert. denied 454 U.S. 858, 102 S.Ct. 310, 70 L.Ed.2d 154 (1981); Associated General Contractors v. San Francisco Unified School District, 616 F.2d 1381, 1384 (9th Cir.), cert. denied, 449 U.S. 1061, 101 S.Ct. 783, 66 L.Ed.2d 603 (1980). See also Scandinavian Air Service v. United Aircraft, 601 F.2d 425, 429 (9th Cir. 1979) (when there has been no definitive adjudication by the state's highest court in a factually similar case, decisions from the state's courts of appeal are data for determining how the highest state court would rule). Here, the district court concluded that the prohibitive social costs involved with imposing a duty on employers to warn former employees of dangers to which they had been exposed counseled against imposition of any such duty. We view this conclusion as the district court's determination of the applicable state law and review the determination de novo. In re McLinn, 739 F.2d 1395, 1400 (9th Cir. 1984) (en banc). Thus, our task is to predict whether, and under what circumstances, the California Supreme Court would impose a duty to warn on a private employer who learned of a danger to which a former employee had been exposed in the course of his employment. We conclude that in the circumstances of this case at least, the California Supreme Court would impose such a duty.

When determining whether a duty exists, the California Supreme Court begins its analysis with the proposition that "legal duties are not discoverable facts of nature, but merely conclusory expressions that, in cases of a particular type, liability should be imposed for damage done." Tarasoff v. Regents of the University of California, 17 Cal.3d 425, 434, 551 P.2d 334, 342, 131 Cal.Rptr. 14, 22 (1976). Thus, "duty is not sacrosanct in itself, but only an expression of the sum total of those considerations of policy which lead the law to say that the particular plaintiff is entitled to protection." Id. (quoting Prosser, Law of Torts, 3d ed. 1964, at 332-33). As the California courts have noted on at least two occasions,

[a]ny number of considerations may justify the imposition of duty in particular circumstances, including the guidance of history, our continually refined concepts of morals and justice, the convenience of the rule, and social judgment as to where the loss should fall.

Weirum v. RKO General, Inc., 15 Cal.3d 40, 46, 123 Cal.Rptr. 468, 471, 539 P.2d 36, 39 (1975); Tresemer v. Barke, 86 Cal.App.3d 656, 671, 150 Cal.Rptr. 384, 393 (1978).

Under California law, the existence of a duty to warn an individual of dangers to which he has been or will be exposed turns largely upon five salient factors: the nature of the relationship between the party who has knowledge of the danger and the party who is threatened with potential injury; the seriousness of the potential injury; the extent to which the potential injury is a foreseeable result of the defendant's conduct; the extent of the burden on the defendant of imposing a duty to warn; and the likelihood that a warning will have a practical effect. California courts weigh all of these factors together when determining whether a duty to warn exists. In order to find such a duty, each factor need not favor the plaintiff. Rather, it is the court's evaluation of all the relevant factors, viewed in the aggregate, which determines the existence of the duty.

In Johnson v. California, 69 Cal.2d 782, 447 P.2d 352, 73 Cal.Rptr. 240 (1968), a foster parent who had been injured in an assault by a recently-paroled foster child sought to recover damages from the state for its failure to warn of the youth's dangerous tendencies. In reversing a grant of summary judgment in favor of the state, the California Supreme Court held that "[a]s the party placing the youth with [the plaintiff], the state's relationship to plaintiff was such that its duty extended to warning of latent, dangerous qualities suggested by the parolee's history or character." Id. at 785, 73 Cal.Rptr. 240, 447 P.2d 352. Upon its review of relevant case law, the court further concluded that the law imposed a duty "upon those who create a foreseeable peril, not readily discernible by endangered persons, to warn them of such potential peril." Id. at 786, 73 Cal.Rptr. 240, 447 P.2d 352 (footnote omitted). Thus, under Johnson, the existence of some sort of relationship between the plaintiff and the defendant as well as the foreseeability of the peril created by the defendant's conduct were the principal factors that led to the imposition of a duty to warn. Cf. Peterson v. San Francisco Community College Dist., 36 Cal.3d 799, 807, 685 P.2d 1193, 1196, 205 Cal.Rptr. 842, 845 (1984).

In Tarasoff v. Regents of the University of California, 17 Cal.3d 425, 551 P.2d 334, 131 Cal.Rptr. 14 (1976), the parents of a murder victim brought suit against a psychologist, inter alia, who had been treating the individual who committed the murder. The plaintiffs alleged that since the offender had confided his intention to kill the victim to the psychologist, the psychologist owed the victim a duty to warn of the impending danger. In reversing the lower court's dismissal of the action, the California Supreme Court held that in view of the defendant's "special relationship" to the offender and the foreseeability of the harm, the law imposed a duty upon the psychologist to warn the victim.

The court first observed that although a number of factors are relevant to determining whether a duty exists, "[t]he most important of these considerations in establishing duty is foreseeability. As a general principle, `a defendant owes a duty of care to all persons who are foreseeably endangered by his conduct, with respect to all risks which make the conduct unreasonably dangerous.'" Id. at 434, 131 Cal.Rptr. 14, 551 P.2d 334 (citations omitted). The court subsequently concluded that the relationship between the psychologist and the patient was sufficiently special to warrant imposition of a duty to warn the victim of foreseeable dangers posed by the patient.

In Tresemer v. Barke, 86 Cal.App.3d 656, 150 Cal.Rptr. 384 (1978), plaintiff brought suit against a physician to recover damages for injuries resulting from plaintiff's use of an intrauterine device. Plaintiff alleged that defendant had breached a duty to warn her of the dangerous effects of the device when, subsequent to its insertion, he obtained actual knowledge of its hazards. The California Court of Appeal upheld plaintiff's cause of action, concluding that a duty to warn arose "by virtue of a confidential relationship between doctor and patient." Id. at 672, 150 Cal.Rptr. 384. In determining that a duty to warn existed, the court observed first that "foreseeability of the risk is a primary consideration in establishing the element of duty." Id. at 671, 150 Cal.Rptr. 384. The court emphasized developments in tort law pursuant to which courts have imposed liability on parties who unknowingly create risks of harm and who fail to take action upon learning of the risks. The court also noted that the law was evolving toward the imposition of the duty to warn on parties who merely become aware of risks to which others have been exposed. Adopting as its own the observations of Prosser, the court said:

Where the original danger is created by innocent conduct, involving no fault on the part of the defendant, it was formerly the rule that no duty to make a reasonable effort to give assistance and avoid any further harm arose; but this appears to have given way, in recent decisions, to a recognition of the duty to take action, both where the prior innocent conduct has created an unreasonable risk of harm to the plaintiff and where it has already injured him. . . This process of extension has been slow, and marked with extreme caution; but there is reason to think that it may continue until it approaches a general holding that the mere knowledge of serious peril, threatening death or great bodily harm to another, which an identified defendant might avoid with little inconvenience, creates a sufficient relation, recognized by every moral and social standard, to impose a duty of action. When the duty is recognized, it is agreed that it calls for nothing more than reasonable care under the circumstances.

Id. at 672, 150 Cal.Rptr. 384 (quoting from Prosser, Law of Torts, § 56, pp. 338-39) (emphasis in original). Under Prosser's analysis, as reiterated in Tresemer, when a party's conduct, though perhaps innocent, gives rise to a serious risk to which another identifiable party has been exposed, and of which the first party could notify him at little cost, the first party has a duty to warn of foreseeable harm resulting from exposure to the risk. See Johnson v. County of Los Angeles, 143 Cal.App.3d 298, 311-12, 191 Cal.Rptr. 704, 712-13 (1983).

Finally, in Thompson v. County of Alameda, 27 Cal.3d 741, 614 P.2d 728, 167 Cal.Rptr. 70 (1980), the court held that the County had no duty to warn the plaintiff, whose child had been murdered by a recently released juvenile offender, where, prior to the offender's release, the County had become aware of only nonspecific threats of harm directed at nonspecific victims. In reaching its holding, the court distinguished Thompson from Johnson and its progeny, reasoning that

[t]hese cases impose a duty upon those who create a foreseeable peril, not readily discoverable by endangered persons, to warn them of such potential peril . . . In Johnson, we emphasized the relationship between the state and plaintiff-victim, and the fact that the state by its conduct placed the specific plaintiff in a position of clearly foreseeable danger. In contrast with the situation in Johnson, in which the risk of danger focused precisely on plaintiff, here County bore no special and continuous relationship with the specific plaintiffs nor did County knowingly place the specific plaintiff's decedent into a foreseeably dangerous position.

Id. at 751, 167 Cal.Rptr. 70, 614 P.2d 728 (emphasis in original). See also Davidson v. City of Westminster, 32 Cal.3d 197, 205-09, 649 P.2d 894, 898-900, 185 Cal.Rptr. 252, 256-58 (1982). In its evaluations of the burden involved with imposing a duty to warn and the resulting benefits to the community, the court further distinguished Thompson from Johnson and Tarasoff. The court concluded that

[u]nlike members of the general public, in Tarasoff and Johnson the potential victims were specifically known and designated individuals. The warnings which we therein required were directed at making those individuals aware of the danger to which they were uniquely exposed. The threatened targets were precise. In such cases, it is fair to conclude that warnings given directly and to a limited number of persons would have a greater effect because they would alert those particular targeted individuals of the possibility of a specific threat of harm pointed at them. In contrast, the warnings sought by plaintiffs would of necessity have to be made to a broad segment of the population and would be only general in nature. In addition to the likelihood that such generalized warnings when frequently repeated would do little as a practical matter to stimulate increased safety measures . . . such extensive warnings would be difficult to give.

Id. 27 Cal.3d at 755, 167 Cal.Rptr. 70, 614 P.2d 728. See also Bill v. Superior Court, 137 Cal.App.3d 1002, 1014, 187 Cal.Rptr. 625, 633 (1982). Thus, the Thompson court concluded that since a duty to warn "would require an expenditure of time and limited resources that [state] agencies cannot spare and would be of questionable value," imposition of such a duty would be unreasonable under the circumstances of that case. Id. 27 Cal. 3d at 757, 167 Cal.Rptr. 70, 614 P.2d 728.

As a result of our study of California law, it is clear to us that in some instances at least, the California Supreme Court would impose a duty on a private employer to warn a former employee of foreseeable dangers to which the employee had been exposed in the course of his employment. While we need not here determine all of the circumstances under which the court would impose such a duty, we have no doubt that it would do so when (1) a party has information relating to a serious risk to the life, safety or health of another; (2) the conduct of the party, though perhaps innocent, gave rise to the risk; (3) the burden resulting from imposition of a duty to warn is not onerous; and (4) there is reason to believe that a warning would have some practical effect.

Accordingly, we conclude that when an employer gains information about a serious danger to which a readily identifiable former employee has been exposed in the course of his employment, the relative cost or inconvenience of warning him is not substantial, and there is reason to believe that the warning might be of some benefit to its recipient, the California Supreme Court would find that a duty to warn exists. If, for example, an employer, who manages a radiation laboratory, learned that a number of former employees had unknowingly been exposed to radiation, and the employer had ready access to the former employees, we are convinced that the California Supreme Court would hold that in view of the high probability of harm and the relative ease with which the employer could issue a warning, the employer owes a duty to the former employees to warn them of the foreseeable harm.

Here, taking appellant's allegations as true, it is clear that Mr. Molsbergen's former employer had information about a serious threat to his life resulting from conditions to which he had been exposed in the course of his employment. In view of the available resources on which his former employer could have drawn in issuing a warning, along with its ready access to former employees of the class to which Mr. Molsbergen belonged (i.e. veterans), it is reasonable to conclude that the burden of notifying him, as well as others similarly situated, of the risk to which he had been exposed would not, relatively speaking, have been substantial. Further, since the warning would have been directed toward making Mr. Molsbergen, and others like him, "aware of the dangers to which they were uniquely exposed," there is reason to believe that the warning would have been of substantial value to its recipients. See Thompson, 27 Cal.3d at 755, 167 Cal.Rptr. 70, 614 P.2d 728. We therefore hold that a private employer would have had a duty to warn Mr. Molsbergen of the foreseeable harm resulting from his exposure to radiation. Since, under California law, a private employer would have had a duty to warn Mr. Molsbergen, the government did also. See 28 U.S.C. § 1346(b) (1976).

IV. CONCLUSION

Because we conclude that appellant has stated a cognizable cause of action under the FTCA, we reverse the district court's dismissal of count one of appellant's complaint and remand the matter to the district court for further proceedings.

REVERSED AND REMANDED.

ANDERSON, J., concurs in the result.

* * * * * *

(d) Allow access by employees or their representatives to accurate records of employee exposures to potentially toxic materials or harmful physical agents.

(e) Notification of any employee who has been or is being exposed to toxic materials or harmful physical agents in concentrations or at levels exceeding those prescribed by an applicable standard, order, or special order, and informing any employee so exposed of corrective action being taken.