Opinion by Judge McKEOWN; Dissent by Judge FERGUSON.

This tort case is a suit stemming from a personal tragedy. Nollie Costo and Christopher Graham drowned during an employer-sponsored rafting trip, and their estates sued their employer for negligence. But their employer is not an ordinary one. It is the United States Navy. Thus, the suit is barred unless the United States has waived its sovereign immunity. To determine whether the suit can properly proceed, we must confront — yet again — the Feres doctrine, which limits the United States’ waiver of sovereign immunity. We conclude that this suit falls within the doctrine’s ever-expanding reach. We reach this conclusion only reluctantly, bound by circuit precedent to apply this doctrine to yet another case that seems far removed from its original purposes.

I. Factual and Procedural Background

Nollie Costo and Christopher Graham were sailors in the United States Navy, stationed at Naval Air Station Whidbey Island, in Oak Harbor, Washington. On July 1, 1995, they participated in a Navy-led recreational rafting trip on the Nook-sack River in Whatcom County, Washington. Both were off duty and on liberty1 at the time. The trip, which included three rafts, was led by Brian Benjamin, a civilian in charge of the base’s rafting program.

The rafting program was operated within the command structure of the military. *865The Navy sponsors various recreational programs that are intended to "effectively contribute to the morale, well-being and quality of life of naval personnel and their family members." Department of the Navy, Bureau of Personnel Instruction (BUPERINST) 1710.11B ¶ 3 (July 1, 1994), available at http://www.bu-pers.navy.mil. Among these are the Morale, Welfare and Recreation (MWR) programs.

According to Navy regulations, the "administration, supervision, and operation of local MWR programs supporting all eligible personnel is a command function and is the responsibility of cognizant commanding officers." Id. at Authorities and Re-sponsibifities ¶ 3. Here, the commanding officer was Captain John Schork. Underneath Schork was the MWR Director, Thomas Lindscott, a civilian. Lindscott was accountable to Schork "for the program content, financial integrity, and health and successful accomplishment of the MWR mission." Id. at Authorities and Responsibilities ¶ 4.

Beneath Lindscott in the chain of command was Richard Score, also a civilian, who headed the recreation division of MWR. Score, in turn, supervised Edward Dunning, a civilian, who managed the Outdoor Recreation Center. It was Dunning who implemented the rafting program, He advertised in local papers for a lead raft guide, and eventually hired Brian Benjamin to head the rafting program. Benjamin hired Tim Herron, first as a guide, then eventually to handle logistics and training.

Prior to the tragic trip, Benjamin and Cathy Crouch-a civilian guide trained by Benjamin-scouted the route. When they did, they observed a log blocking the river, and determined that the rafts would have to pass through a narrow channel to avoid the log.

On the trip itself, the three rafts reached the log shortly after the trip had begun. The first raft negotiated the narrow channel without difficulty. As the remaining two boats prepared to negotiate the channel, they pulled too close to one another for both to pass safely, and one boat hit the log and flipped over. All of the boat's passengers fell into the water. Costo and Graham were trapped beneath the water in the log's submerged branches and drowned.

Costo's parents and personal representative and Graham's personal representative (collectively referred to as "the estates") brought suit against the United States in federal court in Washington.2 In their Complaint, brought under the Federal Tort Claims Act, 28 U.S.C. § 1346, they alleged that MWR "breached its duty to the plaintiffs" by failing to obtain a rafting permit; failing to hire trained guides; and failing to properly supervine those guides. They further alleged that MWR breached its duty by faffing to scout out the river, to warn the rafters of the river's condition, to properly equip the rafts, to properly instruct the rafters, to rescue the rafters, and to administer life saving aid.

The United States moved to dismiss for lack of jurisdiction under Federal Rule of Civil Procedure 12(b)(1). The district court granted the motion, holding that, because the estates' claims fell within the Feres doctrine, the suit was barred by sovereign immunity.

Whether the Feres doctrine applies to the facts in the record is reviewed *866de novo. Dreier v. United States, 106 F.3d 844, 847 (9th Cir. 1996). Factual findings are reviewed de novo, with all disputed facts resolved in favor of the non-moving party. Id. We have jurisdiction under 28 U.S.C. § 1291, and we affirm.

II. TIlE FERES DOCTRINE-BACKGROUND

The passage of the Federal Tort Claims Act (FTCA) in 1948 resulted in a broad waiver of the Federal Government's sovereign immunity: "The United States shall be liable, respecting the provisions of this title relating to tort claims, in the same manner and to the same extent as a private individual under like circumstances, but shall not be liable for interest prior to judgment or for punitive darn-ages." 28 U.S.C. § 2674. However, this blanket waiver contained an exception, by which the Government withheld consent to be sued for "[a]ny claim arising out of the combatant activities of the military or naval forces, or the Coast Guard, during time of war." Id. § 2680(j). Only two years later, this exception was broadened significantly by the Supreme Court, which held in Feres v. United States that "the Government is not liable under the Federal Tort Claims Act for injuries to servicemen where the injuries arise out of or are in the course of activity incident to service." 340 U.S. 135, 146, 71 S.Ct. 153, 95 L.Ed. 152 (1950). This broad exception has been labeled "the Feres doctrine."

The Supreme Court has enunciated three policy rationales for the Feres doctrine: 1) the distinctively federal nature of the relationship between the Government and the armed forces requires a uniform system of compensation for soldiers stationed around the country and around the world; 2) a generous compensation scheme for soldiers (the Veterans' Benefits Act) serves as an ample alternative to tort recovery; and 3) permitting military personnel to sue the armed forces would endanger discipline. See United States v. Johnson, 481 U.S. 681, 684 n. 2, 107 S.Ct. 2063, 95 L.Ed.2d 648 (1987).

These policy justifications and the doctrine itself have been heavily criticized by commentators and by this Court. See, e.g., Estate of McAllister v. United States, 942 F.2d 1473, 1480 (9th Cir. 1991) ("In [affirming the district court], we follow a long tradition of reluctantly acknowledging the enormous breadth of a troubled doctrine."). The goal of uniformity has been criticized as textually unsupported, Johnson, 481 U.S. at 696 (Scalia, J., dissenting), and illogical, id. at 695-96, 107 S.Ct. 2063 (Scalia, J., dissenting) ("nonuniform recovery cannot possibly be worse than (what Feres provides) uniform nonrecovery"). Further, it has been observed that if uniformity is the goal, it makes just as much sense to establish a federal common law of torts as it does to bar all tort suits. Taber v. Maine, 67 F.3d 1029, 1040 (2d Cir. 1995). The second rationale for the bar to tort suits-the existence of the Veterans' Benefits Act-has been criticized as incoherent, given the fact that in certain cases, soldiers have been permitted to recover under both the VBA and the FTCA. Johnson, 481 U.S. at 697-98, 107 S.Ct. 2063 (Scalia, J., dissenting).

The third policy rationale-the danger to discipline-has been identified as the best explanation for Feres. United States v. Shearer, 473 U.S. 52, 57, 105 S.Ct. 3039, 87 L.Ed.2d 38 (1985); Johnson v. United States, 704 F.2d 1431, 1436 (9th Cir. 1983).3 This rationale has not, however, escaped criticism. If the danger to discipline is *867inherent in soldiers suing their commanding officers, then no such suit should be permitted, regardless of whether the “injuries arise out of or are in the course of activity incident to service.” But Feres itself imposes this limitation. If the fear is that civilian courts will be permitted to second-guess military decisions, then even civilian suits that raise such questions should be barred. But they are not. See Johnson, 481 U.S. at 699-700, 107 S.Ct. 2063 (Scalia, J., dissenting).

Perhaps because of these criticisms, circuit courts — including ours — have shied away from attempts to apply these policy rationales. See Taber, 67 F.3d at 1043 (citing cases from the Fifth and Eleventh Circuits). Instead, we have outlined four factors to consider in determining whether a particular suit should be barred by the Feres doctrine:

1) the place where the negligent act occurred;

2) the duty status of the plaintiff when the negligent act occurred;

3) the benefits accruing to the plaintiff because of his status as a service member; and

4) the nature of the plaintiffs activities at the time the negligent act occurred.

Dreier, 106 F.3d at 848 (citations omitted).

Even this four-factor test must be qualified. First, none of these factors is dispositive. Rather than seizing on any particular combination of factors, we have focused on “the totality of the circumstances.” See, e.g., Dreier, 106 F.3d at 852; Millang v. United States, 817 F.2d 533, 535 (9th Cir. 1987).4 Second, we have reached the unhappy conclusion that the cases applying the Feres doctrine are irreconcilable, and thus, “ ‘comparison of fact patterns to outcomes in cases that have applied the Feres doctrine’ is the most appropriate way to resolve Feres doctrine cases.” Dreier, 106 F.3d at 848 (citing Estate of McAllister, 942 F.2d at 1477). With these competing considerations as well as the four factors in mind, we consider the case at hand.

III. Application of the Feres Doctrine

Our inquiry begins — and, in large measure, ends — with Bon v. United States, 802 F.2d 1092 (9th Cir. 1986), a closely analogous Ninth Circuit case. In Bon, we considered claims brought by a Navy servicewoman who rented a canoe for recreational purposes from a Navy-run recreational center, and was hit and injured by a serviceman driving a recreational motorboat also rented from the base. Id. at 1093. We noted that both parties to the accident were on active duty when it occurred; that the activity was provided as a benefit of Bon’s military service; and that the activity was “incident to military service” because the boat rental was governed by military regulations, and the program was under the command of the base’s commanding officer. Id. at 1095. Thus, we held that the Feres doctrine barred suit.

Bon resembles the case at bar in each of these particulars. Like Janice Bon, Costo and Graham were on active duty but on liberty at the time of the accident. Like the canoe rental in Bon, the rafting trip was provided as a benefit of military service. And, as in Bon, the MWR program here was under the command of the base’s commanding officer. On these facts alone, Bon would be dispositive.

The estates’ efforts to distinguish Bon are unavailing. It is of no consequence that the alleged tortfeasors were civilians. This much is clear from the *868Supreme Court’s decision in Johnson, 481 U.S. at 691 & n. 11 (applying the Feres doctrine, in a case involving the alleged negligence of civilian Federal Aviation Administration employees). Nor is our decision affected by the fact that the accident occurred off-base. The appropriate consideration is the “situs of the negligence,” not the location of the accident. Johnson v. United States, 704 F.2d at 1436. Here, much of the alleged negligence was supervisory, and occurred on base. Moreover, this factor is not dispositive, id. at 1436-37, and does not overcome the factors in this case that weigh to the contrary.

Beyond just Bon, it has long been recognized — in our court, at least — that military-sponsored activities fall within the Feres doctrine, regardless of whether they are related to military duties. Thus, in Uptegrove v. United States, we held that the family of Uptegrove (a soldier) could not sue the military for Uptegrove’s death in a military plane crash, despite the fact that Uptegrove was on leave, on his way to vacation with his family, and aboard the military aircraft only as a “space available” passenger. 600 F.2d 1248, 1249 (9th Cir. 1979). Likewise, in the context of medical malpractice suits, we have consistently barred suit — even when the injury did not arise out of the course of duty. See, e.g., Estate of McAllister, 942 F.2d at 1474; Persons v. United States, 925 F.2d 292, 296 (9th Cir. 1991) (“Although he was off-duty, Kelly Persons enjoyed the use of the naval hospital ‘solely by virtue of his status as a serviceman,’ Millang, 817 F.2d at 535, and the doctors who treated him were subject to military orders.”); Atkinson v. United States, 825 F.2d 202 (9th Cir. 1987) (barring a medical malpractice suit arising from the prenatal care given to a soldier in a military hospital). Indeed, the Feres case itself involved medical malpractice.

Cases from outside the Ninth Circuit mirror this approach. In a range of factual situations, the courts of appeals have held that recreational activities sponsored by the military fall within the Feres doctrine. See, e.g., Pringle v. United States, 208 F.3d 1220, 1227 (10th Cir. 2000) (soldier beaten by gang members after being ejected from military MWR club; “The relationship between the Army and service personnel engaged in recreational activities under the Army’s MWR program is ‘distinctively federal’ in character.”); Jones v. United States, 112 F.3d 299, 301 (7th Cir. 1997) (medical malpractice suit against military physicians arising from an injury suffered while training for the Olympics; “In fact, courts have often concluded military personnel acted ‘incident to service’ and applied the Feres bar in cases arising from seivicemembers taking advantage of recreational military activities or other military perquisites because their use of the facilities was a consequence solely of their status as members of the military.”); Walls v. United States, 832 F.2d 93 (7th Cir. 1987) (crash of airplane belonging to recreational Aero Club); Rayner v. United States, 760 F.2d 1217 (11th Cir. 1985) (elective surgery); Woodside v. United States, 606 F.2d 134 (6th Cir. 1979) (crash of airplane belonging to recreational Aero Club); Hass ex rel. United States v. United States, 518 F.2d 1138, 1141 (4th Cir. 1975) (injury while riding a horse rented from a Marine Corps-operated stable; “Recreational activity provided by the military can reinforce both morale and health and thus serve the overall military purpose.”); Chambers v. United States, 357 F.2d 224, 229 (8th Cir. 1966) (death in on-base swimming pool; “As a matter of fact, Airman Chambers’ use of the pool, which was a part of the base, was related to and dependent upon his military service; otherwise, he would not have been privileged to use it.”).

*869In citing this litany of cases, it bears note that the Supreme Court has not had occasion to apply Feres nearly so broadly as have the circuit courts. Indeed, it has been suggested that the Supreme Court’s cases could be interpreted to fall within the very limitation — activity based on military duty — that the estates urge us to apply here. Cf. Taber, 67 F.3d at 1044-45, 1049-50 & n. 21 (analogizing Feres to workers’ compensation systems, and suggesting that Feres be interpreted to bar suit “... for personal injuries sustained ‘in the performance of [the soldier’s] duty.’ ” (quoting 5 U.S.C. § 8102(a))). As the estates contend, the Supreme Court cases typically involve “activities] incident to service” that implicate military duty, see, e.g., Johnson, 481 U.S. 681, 107 S.Ct. 2063, 95 L.Ed.2d 648, or situations where military discipline was important precisely because it so fundamentally implicated the functioning of the military, see, e.g., Shearer, 473 U.S. at 58, 105 S.Ct. 3039. None of these cases involve military-sponsored re-creatioxl. But whatever the original scope of the Feres doctrine, it is clear that it has been interpreted throughout the lower courts — and, specifically, by our court — to include military-sponsored recreational programs. Therefore, we are compelled to hold that the estates’ suit is barred.

IV. Conclusion

As we noted at the outset, we apply the Feres doctrine here without relish. Nor are we the first to reluctantly reach such a conclusion under the doctrine. Rather, in determining this suit to be barred, we join the many panels of this Court that have criticized the inequitable extension of this doctrine to a range of situations that seem far removed from the doctrine’s original purposes. See, e.g., McAllister, 942 F.2d at 1480; Persons, 925 F.2d at 299; Atkinson, 825 F.2d at 206; Monaco v. United States, 661 F.2d 129, 134 (9th Cir. 1981). But until Congress, the Supreme Court, or an en banc panel of this Court reorients the doctrine, we are bound to follow this well-worn path.

AFFIRMED.

dissenting:

I begin this dissent with the understanding that I and this Court of Appeals must abide by the decisions of the Supreme Court. I write to demonstrate that the Feres doctrine is unconstitutional and to present the Supreme Court with a case and controversy in order for it to be able to review the doctrine. See United States v. Johnson, 481 U.S. 681, 692, 107 S.Ct. 2063, 95 L.Ed.2d 648 (1987) (Scalia, J., dissenting).

I believe that the Feres doctrine'violates the equal protection rights of military service men and women. I also believe that Feres violates our constitutional separation of powers.

In passing the Federal Tort Claims Act (FTCA), Congress chose to place all Americans on an equal footing in litigating the civil liability of the federal government for claims of tort injuries. 28 U.S.C. § 1346. From this unprecedented, albeit limited, waiver of sovereign immunity, Congress excluded claims arising out of a number of government activities, including “[a]ny claim arising out of the combatant activities of the military or naval forces, or the Coast Guard, during time of war.” 28 U.S.C. § 2680(j). This exception of liability for injuries in times of war was expanded by the Feres doctrine to all injuries of military personnel “where the injuries arise out of or are in the course of activity incident to service.” Feres v. United States, 340 U.S. 135, 146, 71 S.Ct. 153, 95 L.Ed. 152 (1950). This doctrine represents judicial legislation, effectively negating the Congressional limitations that the *870excluded claims must arise from “combatant activities ... during time of war.” 28 U.S.C. § 2680(j). The doctrine effectively declares that the members of the United States military are not equal citizens, as their rights against their government are less than the rights of their fellow Americans.1

This judicially-created classification runs afoul of the Equal Protection clause of the 14th and 5th Amendments. Clearly, had Congress barred all claims by members of the military it would have been within its constitutional powers to do so. See, e.g., Rostker v. Goldberg, 453 U.S. 57, 66-67, 83, 101 S.Ct. 2646, 69 L.Ed.2d 478 (1981) (holding that Congress acted within its constitutional authority in authorizing the military registration of men but not women). Indeed, as the Feres court noted, Congress was cognizant of potential military claims when drafting the FTCA and, had it chosen to do so, could have explicitly excluded them. See Feres, 340 U.S. at 138-39, 71 S.Ct. 153 (noting that all but two of the eighteen drafts of the FTCA considered by Congress barred suits by members of the military); 28 U.S.C. § 2671 (specifically including “members of the military or naval forces of the United States” “acting in the line of duty” in its enumeration of individuals and acts for which the United States may be liable). Likewise, as Congress enacted the FTCA, its decision to bar suits by military members arising in combat during times of war easily passes equal protection scrutiny. Because service men and women are not a “suspect class,” and their access to federal court under the FTCA is not a “fundamental right,” see Miller v. United States, 73 F.3d 878, 881 (9th Cir. 1995), Congress’ classification needed merely to be rationally related to a legitimate government interest. See Vacco v. Quill, 521 U.S. 793, 799-801, 117 S.Ct. 2293, 138 L.Ed.2d 834 (1997). This standard “does not allow us to substitute our personal notions of good public policy for those of Congress.” Schweiker v. Wilson, 450 U.S. 221, 234, 101 S.Ct. 1074, 67 L.Ed.2d 186 (1981). It is not for “the judiciary [to] sit as a super-legislature to judge the wisdom or desirability of legislative policy determinations made in areas that neither affect fundamental rights nor proceed along suspect lines.” City of New Orleans v. Dukes, 427 U.S. 297, 303, 96 S.Ct. 2513, 49 L.Ed.2d *871511 (1976). See also United States v. Johnson, 481 U.S. 681, 702, 107 S.Ct. 2063, 95 L.Ed.2d 648 (1987) (Scalia, J., dissenting) (arguing that the only possible justification for the chaos created by the Feres doctrine would be if it reflected a decision grounded in the democratic process).

When considering the Feres doctrine, however, we are not dealing with a legislative action, but rather with a judicial rewriting of an unambiguous and constitutional statute. Even to the courts that have considered it, the Feres decision stands not for an interpretation of statute but rather a “judicially created exception” to the FTCA. Stencel Aero Eng’g Corp. v. United States, 431 U.S. 666, 674, 97 S.Ct. 2054, 52 L.Ed.2d 665 (1977) (Marshal, J., dissenting); Schoemer v. United States, 59 F.3d 26, 28 (5th Cir. 1995); Pringle v. United States, 208 F.3d 1220, 1223 (10th Cir. 2000); Romero by Romero v. United States, 954 F.2d 223, 224 (4th Cir. 1992). This judicial re-writing runs against our basic separation of powers principles and complicates a equal protection analysis of the resulting Feres doctrine. See THE FEDERALIST NO. 47 (James Madison) (“Were the power of judging joined with the legislative, the life and liberty of the subject would be exposed to arbitrary controul, for the judge would then be the legislator”) (quoting Montesquieu); City of Cleburne v. Cleburne Living Ctr., 473 U.S. 432, 441, 105 S.Ct. 3249, 87 L.Ed.2d 313 (1985) (“Courts have been very reluctant, as they should be in our federal system and with our respect for the separation of powers, to closely scrutinize legislative choices”); Tenn. Valley Auth. v. Hill, 437 U.S. 153, 194-195, 98 S.Ct. 2279, 57 L.Ed.2d 117 (1978) (“Our individual appraisal of the wisdom or unwisdom of a particular course consciously selected by the Congress is to be put aside in the process of interpreting a statute. Once the meaning of an enactment is discerned and its constitutionality determined, the judicial process comes to an end. We do not sit as a committee of review, nor are we vested with the power of veto.”)

I do not suggest that judges never alter legislation, but that they should never overrule the plain language of Congress unless there is a constitutional violation. See Marbury v. Madison, 5 U.S. (1 Cranch) 137, 177-79, 2 L.Ed. 60 (1803). Although the Constitution clearly delegates law-making activities to the legislature and law-interpreting activities to the judiciary, the reality of our judicial system is that courts must occasionally take more active roles in order to interpret an ambiguous statute, Comm’r of Internal Revenue v. Asphalt Prods. Co., Inc., 482 U.S. 117, 121, 107 S.Ct. 2275, 96 L.Ed.2d 97 (1987) (“Judicial perception that a particular result would be unreasonable may enter into the construction of ambiguous provisions, but cannot justify disregard of what Congress has plainly and intentionally provided”), or to prevent a statute from running afoul of the Constitution, Elfbrandt v. Russell, 384 U.S. 11, 14-19, 86 S.Ct. 1238, 16 L.Ed.2d 321 (1966). Legislative silence on an issue may also require judges to determine, through history or analogy, the most appropriate legal rule in a particular situation. See, e.g., Georgia-Pacific Corp. v. United States Envtl. Prot. Agency, 671 F.2d 1235, 1240 (9th Cir. 1982); Farouki v. Emirates Bank Int’l, Ltd., 14 F.3d 244, 250 n. 17 (4th Cir. 1994).

Feres presented neither ambiguity nor constitutional violation nor legislative silence.2 Instead, Feres took a fairly small, *872clearly defined, legislatively-created classification and broadened it considerably. As the Supreme Court has noted repeatedly, this kind of classification “inevitably requires that some persons who have an almost equally strong claim to favored treatment be placed on different sides of the line, and the fact [that] the line might have been drawn differently at some points is a matter for legislative, rather than judicial, consideration." FCC v. Beach Communications, Inc., 508 U.S. 307, 315-16, 113 S.Ct. 2096, 124 L.Ed.2d 211 (1993) (quoting United States R.R. Ret. Bd. v. Fritz, 449 U.S. 166, 179, 101 S.Ct. 453, 66 L.Ed.2d 368 (1981)) (emphasis added). The result raises both equal protection and separation of powers concerns and, I believe, violates our Constitution.

There is no authority suggesting a standard of review for an act of judicial legislation, and I hesitate to propose one. Surely, however, a more stringent standard than “rational review” applies to classifications created by the judiciary rather than the legislature, as the constitutional implications are greater. Cf. Bearden v. Georgia, 461 U.S. 660, 665, 103 S.Ct. 2064, 76 L.Ed.2d 221 (1983) (noting that in cases considering the constitutionality of court fees the Court applies a hybrid of its “due process” and “equal protection” analyses). The Supreme Court, in protecting another balance of constitutional powers, that between state and federal governments, has created an exacting standard requiring Congress to convincingly articulate concerns of sufficient urgency before encroaching on state immunity. See, e.g., Board of Trustees of Univ. of Alabama v. Garrett, 531 U.S. 356, -, 121 S.Ct. 955, 964, 148 L.Ed.2d 866 (2001); Fla. Prepaid Postsecondary Educ. Expense Bd. v. College Sav. Bank, 527 U.S. 627, 639, 119 S.Ct. 2199, 144 L.Ed.2d 575 (1999). Even if the relationship between the ends and means need merely be rational, the judiciary should articulate a reason that it is necessary to ignore the balance of federal powers. This Feres failed to do.

A brief history of the fate of the four Feres rationales illustrates my point.3 The *873Feres Court initially proffered three rationales for the Feres decision. It subsequently rejected the first rationale, Indian Towing Co. v. United States, 350 U.S. 61, 66-69, 76 S.Ct. 122, 100 L.Ed. 48 (1955), and noted that the other two rationales were “no longer controlling,” United States v. Shearer, 473 U.S. 52, 58 n. 4, 105 S.Ct. 3039, 87 L.Ed.2d 38 (1985). In the meantime, the Court developed a fourth rationale, not considered by Feres, and labeled it the “best” justification for the Feres decision. Id. at 57. This fourth rationale — the negative effect of lawsuits on military discipline and decision-making — did not do much to support a Feres bar in the Court’s most recent Feres case, which involved a civilian tortfeasor. United States v. Johnson, 481 U.S. 681, 107 S.Ct. 2063, 95 L.Ed.2d 648 (1987). The Court instead revived the second and third rationales from Feres and stated that all three surviving rationales applied. Id. at 688-91. If Congress had engaged in such inconsistent and post hoc rationalization to invade the spheres of executive, judicial, or state powers, the Constitution would not have permitted it. Cf. Northern Pipeline Const. Co. v. Marathon Pipe Line Co., 458 U.S. 50, 83-84, 102 S.Ct. 2858, 73 L.Ed.2d 598 (1982) (finding that the Bankruptcy Act of 1978 was invalid because it threatened to encroach upon Article III courts despite characterization to the contrary by Congress); Metro. Washington Airports Auth. v. Citizens for Abatement of Aircraft Noise, Inc., 501 U.S. 252, 276, 111 S.Ct. 2298, 115 L.Ed.2d 236 (1991) (finding Congressional oversight committee effectively encroached on executive branch); Alden v. Maine, 527 U.S. 706, 731-35, 119 S.Ct. 2240, 144 L.Ed.2d 636 (1999) (finding that Congressional legislation impermissibly encroached on state immunity).

We can speculate forever upon reasons why Feres refused to apply a law written by Congress. It is clear that Feres recognized that the direct and unambiguous command of Congress created liability for claims caused by members of the military or naval forces of the United States, Feres, 340 U.S. at 138, 71 S.Ct. 153; 28 U.S.C. § 2671, and that the direct and unambiguous command of Congress exempted only claims arising out of combatant activities during time of war, Feres, 340 U.S. at 138, 71 S.Ct. 153; 28 U.S.C. § 2680(j). If a statute is ambiguous then it is a necessary judicial function to clarify it. But Feres did not take this approach in reaching its conclusion. Instead, it appears to have reevaluated the law. For example, the Court declared as one reason for its rejection of a Congressional mandate “[t]hat the geography of an injury should select the law to be applied to [a soldier’s] tort claims makes no sense.” Feres, 340 U.S. at 143, 71 S.Ct. 153. This language reveals that the Court simply did not agree with Congress and searched in puzzling ways to declare that military personnel are not equal to civilians.

The Court recently had occasion to evaluate a judicial classification in Bush v. Gore, 531 U.S. 98, 121 S.Ct. 525, 148 L.Ed.2d 388 (2000). There, the Court evaluated a decision by the Florida Supreme Court that .allowed disputed ballots to be counted according to an “intent of the voter” standard. Putting aside the separation of powers question, the Court stated that the “problem [with the judicially-created classification] inheres in the absence of specific standards to ensure its equal application.” Id. at 530. As many commentators have noted, one of the hall*874marks of the Feres eases is the ill-defined standard “incident to service” that has lead to very unequal applications even within the military. See, e.g., Anne R. Riley, United States v. Johnson: Expansion of the Feres Doctrine to Include Service-members’ FTCA Suits Against Civilian Government Employees, 42 Vand. L. Rev. 233, 266-67 (1989) (“Consistent in denying recovery, the Court is embarrassingly inconsistent in how it arrives at this result”); Martha J. Burns, They Fight to Protect Our Rights; Shouldn’t we do the Same for Them? Intramilitary Immunity in Light of United States v. Stanley, 38 Depaul L. Rev. 127, 154 (Fall, 1988) (“If the Supreme Court had established a definition for 'incident to service’ in this case, it would have negated the problems that result from the inconsistent definitions and applications of the term”). Compare, e.g., Ricks v. United States, 842 F.2d 300 (11th Cir. 1988) (Feres barred claim for death of serviceman on temporary disability retired list), and Cortez v. United States, 854 F.2d 723 (5th Cir. 1988) (Feres did not bar claim for death of serviceman on temporary disability retired list).

Even if we were to apply a rational basis analysis — the most lenient of equal protection inquiries — to the Feres doctrine, the Court’s jurisprudence in this area over the past fifty years defies this approach. In the last case to directly address Feres, United States v. Johnson, 481 U.S. 681, 107 S.Ct. 2063, 95 L.Ed.2d 648 (1987), Justice Scalia, in a dissent joined by Justices Brennan, Marshall, and Stevens, thoroughly and articulately rebutted the four most common justifications for the continued application of the Feres doctrine. Johnson, 481 U.S. at 692-703, 107 S.Ct. 2063 (Scalia, J., dissenting). Not only is each justification belied by reality and the language of the FTCA, but each justification has, at various points, been discredited by the Court.

The first justification proffered by Feres was that there was no parallel private right of action whereby military members could sue their employer. Feres, 340 U.S. at 141-42, 71 S.Ct. 153. This ignores other provisions of the FTCA, however, which opened to liability a number of areas where parallel private rights of action did not previously exist, including the “trans-mifssion of] postal matter, 28 U.S.C. § 2680(b), collection of] taxes or custom duties, § 2680(c), imposition of] quarantines, § 2680(f), [and regulation of] the monetary system, § 2680(f).” Johnson, 481 U.S. at 694, 107 S.Ct. 2063 (Scalia, J., dissenting). This rationale was also rejected by the Court in subsequent FTCA cases. Rayonier Inc. v. United States, 352 U.S. 315, 319, 77 S.Ct. 374, 1 L.Ed.2d 354 (1957); Indian Towing, 350 U.S. at 66-69, 76 S.Ct. 122.

The second justification was the “distinctively federal” relationship between the Government and the military that risked being supplanted by local tort law. Putting aside the Court’s preference for “uniform nonrecovery” over “nonuniform recovery,” Johnson, 481 U.S. at 695-96, 107 S.Ct. 2063, Justice Scalia observed that “we have effectively disavowed this ‘uniformity’ justification ... by permitting servicemen to recover under the FTCA for injuries suffered not incident to service, and permitting civilians to recover for injuries caused by military negligence.” Id. at 696, 107 S.Ct. 2063 (emphasis in original). See also Hunt v. United States, 636 F.2d 580, 598 (D.C.Cir. 1980) (“[G]iven the Supreme Court’s inconsistent treatment of this factor, it cannot be said that the presence of an alternative compensation system either explains or justifies the Feres doctrine; it only makes the effect of the doctrine more palatable.”). Cf. City of Cleburne v. Cleburne Living Center, 473 U.S. 432, 446, 449-50, 105 S.Ct. 3249, 87 *875L.Ed.2d 313 (1985) (finding that an ordinance did not pass rational basis scrutiny where the proffered justification was too attenuated from the classification). The Court stated that this rationale of the distinctively federal relationship was “no longer controlling” in Shearer, 473 U.S. at 58 n. 4, 105 S.Ct. 3039, before reviving it in Johnson, 481 U.S. at 688-90, 107 S.Ct. 2063.

The Court gave the third Feres rationale the same treatment. See Shearer, 473 U.S. at 58 n. 4, 105 S.Ct. 3039; Johnson, 481 U.S. at 688-90, 107 S.Ct. 2063. This rationale reasoned that, despite the plain language of the FTCA, Congress did not intend to allow military personnel to recover under the FTCA when they were guaranteed recovery under the Veterans Benefit Act (“VBA”), 38 U.S.C. § 301. Feres, 340 U.S. at 144. In fact, however, “both before and after Feres [the Court] permitted injured servicemen to bring FTCA suits, even though they had been compensated under the VBA.” Johnson, 481 U.S. at 697, 107 S.Ct. 2063 (citing Brooks v. United States, 337 U.S. 49, 53, 69 S.Ct. 918, 93 L.Ed. 1200 (1949) and United States v. Brown, 348 U.S. 110, 111, 75 S.Ct. 141, 99 L.Ed. 139 (1954)). See also Sidley v. United States Dep’t of Navy, 861 F.2d 988 (6th Cir. 1988) (holding that Feres barred claim even though serviceman had been denied veteran’s benefits), Cf. Cleburne, 473 U.S. at 449-50, 105 S.Ct. 3249 (holding that an ordinance failed the rational basis test where the stated policy objectives were not met by the challenged classification).

It was only in cases after Feres that the Court articulated its dominant explanation for the Feres doctrine — the possibly negative effect on military discipline and decision-making. Brown, 348 U.S. at 112, 75 S.Ct. 141; Shearer, 473 U.S. at 57, 105 S.Ct. 3039. The Court’s holding in Johnson — barring a military widow from suing a civilian tortfeasor — calls this rationale into doubt. Furthermore, the text of the FTCA already accommodates the Court’s concerns about interfering in “military judgments and decision that are inextricably intertwined with the conduct of the military mission.” Johnson, 481 U.S. at 691, 107 S.Ct. 2063 (majority opinion). The Act itself prohibits everyone, military and civilian, from suing for discretionary actions that implicate policy judgments, 28 U.S.C. § 2680(a), or for “claims arising in foreign countries, § 2680(k), intentional torts, § 2680(h), and claims based on the execution of a statute or regulation, § 2680(a).” Id. at 699-700, 107 S.Ct. 2063 (Scalia, J., dissenting). The Feres decision was not necessary to preserve our tradition of judicial deference to the “complex subtle, and professional decisions as to the compositing, training, equipping, and control of a military force,” Gilligan v. Morgan, 413 U.S. 1, 10, 93 S.Ct. 2440, 37 L.Ed.2d 407 (1973). Perhaps this is why Feres did not seek to rely on this rationale. I am left to conclude that the only conceivable reason for the Court to engage in rewriting of such a momentous statute was that it believed that Congress had not given enough protection to the government against the men and women in the armed forces. It is curious that the Court thought that it had to protect the government against those persons whose sole duty was to protect the government.

The articulated “rational bases” for the Feres doctrine lead in this case, as in many cases, to inconsistent results that have no relation to the original purpose of Feres. Less than half of the persons on the rafting trip that claimed the lives of Costo and Graham were identified as members of the armed services. The holding today would have allowed any of the civilians injured or killed on the trip to sue, but barred such recourse to the military personnel, despite *876the fact that the two suits would have implicated virtually identical policy concerns regarding the law of the situs and military decision-making. On the other hand, had Costo and Graham participated in a similar rafting trip run entirely by civilians, they may have been able to sue, yet still collect veteran’s benefits. I cannot find a rational basis for the court to engage in such line-drawing on the basis of an “incident to service” test. Under the exceptions Congress has already placed in the FTCA, if the negligence leading to Costo and Graham’s death implicated protected policy judgments, their executors would not be able to sue the government regardless of Feres. 28 U.S.C. 2680(a).

Although the Court’s efforts to provide a “rational basis” for the classifications created by Feres have not been successful, the actions of this Circuit have been even worse. We have recognized the impossibility of applying the Feres rationales and instead retreated to the four-prong factual inquiry described by the majority in this case. Majority op. at 866-67. See also Dreier v. United States, 106 F.3d 844, 848-49 (9th Cir. 1996); Bon v. United States, 802 F.2d 1092, 1094 (9th Cir. 1986). We have, in short, abandoned any pretense that there is a rational basis for the classifications drawn in the original Feres opinion, and yet we have continued to apply the “incident to service” test with little thought to the constitutional principles at stake.4 Nor have we been the only circuit to take this approach. See Pringle v. U.S., 208 F.3d 1220, 1224 (10th Cir. 2000); Schoemer v. U.S., 59 F.3d 26, 28-29 (5th Cir. 1995). See also Maas v. U.S., 94 F.3d 291, 295 (7th Cir. 1996) (“Application of the Feres doctrine does not depend on the extent to which its rationales are present in a particular ease. Rather, the test is whether the injuries are based on ‘service-related activities’ ”) (citation omitted). This blind adherence has proved virtually unworkable, a result that only underscores the wisdom of our founding fathers and the fragile complexities of our system of government.5 See THE FEDERALIST NO. 78 (Alexander Hamilton) (“[L]iberty can have nothing to fear from the judiciary alone, but would have every thing to fear from its union with either of the other departments”).

It is time for the Supreme Court to revisit the Feres doctrine. If it is to be applied, the equal protection issues it raises must be reconciled with the constitutional principles that courts have so often articulated.