dissenting.
The petitioners, two chemical companies, have brought this breach-of-eontract action seeking reimbursement from the Government for their contribution to the settlement of lawsuits brought by Vietnam veterans exposed to their product Agent Orange. The companies argue that their contracts with the Government to produce Agent Orange contain certain promises or warranties that, in effect, hold them harmless. To win this case, as in the most elementary breach-of-contract case, the companies must show that the Government in fact made the warranties or promises, that the Government breached them, and that the Agent Orange settlement contribution was a consequent foreseeable harm. See Restatement (Second) of Contracts §§346, 347, 351 (1979); 5 A. Corbin, Contracts §§997, 1001, 1002 (1964).
The companies concede that the promises, or warranties, are not written explicitly in their contracts; but, the companies intend to prove certain background facts and legal circumstances, which, they say, will show that these promises, or warranties, are an implicit part of the bargain that the parties struck. See 3 id., §§538, 551 (common and trade usage, course of dealings, and existing statutes and rules of law are always probative as to the meaning of the parties).
The background facts alleged include the following:
• In the 1960’s the Government, by exercising special statutory authority, required the companies to enter into the Agent Orange production contracts over the explicit objection of at least one of the companies. See Defense Production Act of 1950 (DPA), 50 U. S. C. App. §2061 et seq. (1988 ed. and Supp. V); App. 8-9, 23-24.
• The Government required the companies to produce Agent Orange according to precise, detailed production specifications. Ibid.
• At that time the Government knew but did not reveal that Agent Orange was defective, or unsafe, to the point *432where its use might lead to plausible tort claims advanced by those who used it. Id., at 10-11, 25.
• The Government specified that the companies could not label Agent Orange in ways that might have promoted its safe use (with, say, dilution instructions), while, at the same time, the Government permitted its soldiers to use Agent Orange in unreasonably risky ways (such as using empty containers for showers or barbecues). Id., at 8-10.
The background (1960’s) legal circumstances include the following:
• United States v. Spearin, 248 U. S. 132 (1918), in which this Court approved the common judicial practice of reading Government contracts that provide detailed “plans and specifications,” as containing an implied warranty that “the contractor will not be responsible for the consequences of defects in the plans and specifications.” Id., at 136.
• Lower court decisions reading Government contracts as containing an implied warranty that performance costs will not increase due to the Government’s superior knowledge of undisclosed “vital information” that causes the cost increase. See Helene Curtis Industries, Inc. v. United States, 312 F. 2d 774, 777-778, and n. 1 (Ct. Cl. 1963) (collecting cases).
• The broad language of the statute that authorized the President to enter into defense procurement contracts— language broad enough to authorize Government promises to indemnify. 50 U. S. C. § 1431 (1988 ed., Supp. V); Exec. Order 10789, 3 CFR 426 (1954-1958 Comp.). See also Exec. Order 11610, 3 CFR 594 (1971-1975 Comp.) (taking view that the statute grants authority to promise indemnification).
• The language of the DPA, which, while permitting the Government to place compulsory defense orders, also says that the compelled firms shall not “be held liable for damages ... for any act or failure to act resulting directly or indirectly from compliance with” such an “order.” 50 U. S. C. App. §2157 (1988 ed.).
*433These background facts and circumstances, say the companies, show that the Government knew far better than they that its Agent Orange contract specifications would force them to produce a risky product for risky use. They add that all parties knew of legal doctrines that made the Government responsible in analogous circumstances for analogous risks. They argue that the Government would not have wanted to force them (under the DPA) to enter into a contract subjecting them (through its specifications) to serious risks of damage, in respect to which (because of the Government’s superior knowledge) they could not bargain for compensation. They conclude that the Government, in the contracts, took responsibility for those risks by implicitly promising to assume responsibility for the consequences of specification defects, including indemnification for the settlement of defect-related tort suits.
The Federal Circuit affirmed a grant of summary judgment against the companies. But, in doing so, it did not decide against the companies in respect to their claimed promises. Instead the Federal Circuit assumed (reluctantly and for argument’s sake) that the companies would be able to prove the existence of the promises, but it went on to hold against them regardless. Even assuming the promises, the Circuit wrote, the companies will not be able to prove causation between promises and damages. The Circuit believed that, had the companies litigated the Agent Orange tort suits instead of settling them, they would have asserted a “government contractor defense,” see Boyle v. United Technologies Corp., 487 U. S. 500 (1988), and thereby won the lawsuits. It concluded that, since the companies could readily have won the suits, the settlement amounts to a “voluntary payment” that cuts any causal link between a broken promise, or warranty, and resulting harm.
The companies, in their petition for certiorari and initial brief on the merits, primarily asked us to review, and to re*434verse, this “no causation” holding. The Court, in today’s opinion, does not discuss that holding. Instead it holds that the companies will not be able to prove the existence of the implicit promises. In my view, however, the record before us now does not permit this latter holding. Rather, this Court should reverse the Court of Appeals’ “no causation” holding and then remand this case for further proceedings.
I need mention only one fatal flaw in the Court of Appeals’ “no causation” holding, that of hindsight. The Court of Appeals, in essence, found the companies’ Agent Orange settlement so obviously unnecessary, so abnormal, so far removed from ordinary litigation behavior, that it could not have been “foreseeable,” see Restatement (Second) of Contracts §351; 5 Corbin, Contracts § 1002, or (if I recast the same point in the Court’s tort-like “causation” language) that it cut the causal link between promise breach and harm. But, viewed without the benefit of legal hindsight, the settlement was neither unforeseeable nor was it an intervening “cause” of the loss.
In 1984, when the companies settled, the settlement was not notably different in terms of reasonableness or motivation from other settlements that terminate major litigation, for at that time the law that might have provided the companies with a defense was far less clear than it is today. I concede that even then some Circuits already had found in the law a “government contractor defense” that, in effect, immunized defense contractors from most suits by servicemen claiming injury from defective products. See McKay v. Rockwell International Corp., 704 F. 2d 444, 448-451 (CA9 1983), cert. denied, 464 U. S. 1043 (1984); Brown v. Caterpillar Tractor Co., 696 F. 2d 246, 249-254 (CA3 1982); Tillett v. J. I. Case Co., 756 F. 2d 591, 599-600 (CA7 1985). But, most of these Circuits had held that the existence of such a defense was a matter of state law, which might differ among the States. See Brown, supra; Tillett, supra; Hansen v. Johns-Manville Products Corp., 734 F. 2d 1036, 1044-1045 (CA5 *4351984), cert. denied, 470 U. S. 1051 (1985). The Second Circuit, the home of the Agent Orange litigation, had not decided the issue. And, the two Agent Orange Second Circuit trial judges who (due to certain here irrelevant procedural considerations) both considered the companies’ “government contractor” defense decided the issue in opposite ways. Compare In re “Agent Orange” Product Liability Litigation, 565 F. Supp. 1263, 1274-1275 (EDNY 1983), with In re “Agent Orange” Product Liability Litigation, 597 F. Supp. 740, 847-850 (EDNY 1984), aff’d, 818 F. 2d 145 (CA2 1987), cert. denied sub nom. Fraticelli v. Dow Chemical Co., 484 U. S. 1004 (1988). This Court did not authoritatively uphold the “government contractor” defense until 1988, four years after the settlement here at issue. Boyle, supra. And, it did so on a ground different from that upon which the Circuit Courts had previously relied. Compare McKay, supra, at 448-451; Tillett, supra, at 596-597 (finding the “government contractor defense” implicit in Feres v. United States, 340 U. S. 135 (1950)), with Boyle, supra, at 509-511 (explicitly rejecting Feres as the basis for a “government contractor defense”).
In light of this contemporaneous legal uncertainty, the settlement, viewed from the companies’ perspective and without benefit of hindsight, seems a reasonable litigation strategy, through which the companies avoided added litigation costs and the threat of significant additional liability while helping to provide the veterans with at least some compensation. See In re “Agent Orange,” 597 F. Supp., at 749 (explaining why Agent Orange District Court approved the settlement). Nothing in the record here suggests the contrary. And, if reasonable at the time, the settlement must have been a “foreseeable” potential consequence of litigation and therefore within the scope of what the companies claim were implicit promises or warranties protecting them against the harms of litigation. See also 24 F. 3d 188, 205-208 (C A Fed. 1994) (Plager, C. J., dissenting). For that reason, this Court *436simply should set aside the Court of Appeals’ determination on the point.
The Court instead decides this case on an alternative basis, namely, that the companies cannot prove the existence of the implicit promises or warranties that they claim. But the existence of a contractual promise implied in fact is very much a creature of particular circumstance — the particular terms, the negotiating circumstances, and the background understandings of law or industry practice. See 3 Corbin, supra, §§562, 566-570. Unlike the majority, which compartmentalizes the companies’ claims into several separate doctrinal categories (a “Spearin” claim, an implied indemnification claim) — each rejected separately for doctrine-specific reasons — I believe the companies’ submissions, fairly read, also set forth a much more general fact-based claim. In essence, the companies say that the parties, when specifying the details of this compulsory defense order, implicitly agreed to allocate to the Government certain risks of defective-government-specification-caused harm — namely, those risks for which each company, because of its inferior knowledge, could not seek compensation in the contract price. And, the companies allege background facts that, if true and complete (as we must assume at this stage of the proceedings), make that implication plausible.
The legal considerations to which the majority points do not answer the companies’ basic implied-in-fact contentions. To do so, the majority would have to argue that the five sets of legal circumstances to which it points, taken separately or together, show that no Government contracting officer would have agreed to a promise or warranty (of the sort claimed); hence, one cannot possibly imply the existence of such a promise “in fact.” See 3 Corbin, supra, §561. The majority cannot argue that, because those five sets of circumstances suggest the contrary.
First, the majority implies that a contracting officer, in all likelihood, would not have agreed to an implicit promise of *437indemnity, for doing so would amount to a bypass of, and “render ... superfluous,” the statutes and “panoply of implementing regulations” that set forth specific procedures that contractors must follow to obtain a promise of indemnity. Ante, at 429. My problem with this argument lies in the fact that, in 1964, the relevant statute, Executive Order, and regulations read very differently. At that time, their language was nonspecific or ambiguous on the procedures required for indemnification. The statute has always been phrased in general language, making no explicit reference to indemnification. See 50 U. S. C. § 1431 (1988 ed., Supp. V); 50 U. S. C. § 1431 (1964 ed.). The portion of the Executive Order that today treats indemnification as special, and sets out procedures for indemnification, did not exist in 1964, and the relevant regulations were also either silent or much more ambiguous than they are today. Compare Exec. Order 11610, 3 CFR 594 (1971-1975 Comp.) (indemnification), with Exec. Order 10789, 3 CFR 426 (1954-1958 Comp.) (no specific reference to indemnification); compare 32 CFR § 17 — 301 et seq. (1975) (implementing today’s indemnification procedures) with 32 CFR § 17-301 et seq. (1964) (no reference to indemnification procedures) and 32 CFR § 17.204-4 (1960) (“Informal commitments may be formalized under certain circumstances to permit payment to persons who have taken action without formal contract [e. g., where a person has furnished property or services to the military in good-faith reliance on the apparent authority of a person giving an oral instruction]. Formalization of commitments under such circumstances normally will facilitate the national defense by assuring such persons that they will be treated fairly and paid expeditiously”); 32 CFR §17.206(i) (1964) (indemnification contracts must subject Government’s obligation to availability of appropriated funds).
Second, the majority points to Comptroller General opinions that say that an “open-ended” agreement to indemnify would violate the Anti-Deficiency Act, 31 U. S. C. § 1341 (1988 *438ed.). Ante, at 427, and n. 10 (citing In re Assumption by Government of Contractor Liability to Third Persons— Reconsideration, 62 Comp. Gen. 361 (1983)). The problem with this argument is that other Comptroller General opinions say that an agreement to indemnify that is not open-ended, but is capped at an amount that a private insurer might have provided, is not improper. See Reimbursement of Costs in Connection with Liabilities to Third Parties for Employees’ Negligence, 22 Comp. Gen. 892 (1943). A capped agreement, which, if reflected in the contract price, makes the Government a kind of self-insurer, is in effect within the appropriation (because the expenditure of assuming the risk of liability will roughly equal the cost of premiums that the Government saves by self-insuring), and may well prove sufficient for the plaintiffs’ purposes. After all, on plaintiffs’ factual allegations, a contractor who was as aware of the plaintiffs’ alleged risks as was the Government would have sensed trouble, wanted insurance, and likely have obtained a premium payment sufficient to buy it. The companies need argue only for a capped implicit warranty that would treat the unknowing contractor similarly. See also In re Government Indemnification of Public Utilities Against Loss Arising Out of Sale of Power to Government, 59 Comp. Gen. 705 (1980) (indemnification in contracts with a “sole source” of a good or service lawful under Anti-Deficiency Act). Whether an agreement to spend money beyond that which was appropriated is in writing or not is irrelevant to the Anti-Deficiency Act.
Third, the majority distinguishes United States v. Spearin, 248 U. S. 132 (1918), on the ground that the implied warranty that Justice Brandéis there discussed protects a contractor from “specifications” that, in the majority’s words, will “frustrate performance or make it impossible,” but does not “extend . . . beyond performance to third-party claims against the contractor.” Ante, at 425. Spearin itself does not make this distinction. Nor have subsequent cases. See *439Michigan Wisconsin Pipeline Co. v. Williams-McWilliams Co., 551 F. 2d 945 (CA5 1977) (allowing recovery against the Government of damages paid by a Government contractor to a third party to which the contractor caused damage by following Government specifications). See also 24 F. 3d, at 197 (Spearin holds contractor harmless if the product proves defective). If the Government must pay, say, for the contractor’s own machinery destroyed by a (defective-specification-caused) explosion when that destruction frustrates performance, see Ordnance Research, Inc. v. United States, 609 F. 2d 462, 479 (Ct. Cl. 1979) (treating explosions causing increased costs as a breach of the warranty of specifications), why should the Government not also have to pay when the explosion takes place just after performance is complete? And, why should it not have to reimburse the contractor’s payment for identical damage caused his next-door neighbor in the same explosion? In any event, whether or not there are good answers to these questions, they are unlikely to answer plaintiffs’ further argument, namely that, even if Spearin does not compel a decision in their favor, it offers indirect support, as background, for implying a promise that would provide (in the particular circumstances) Spearin-like protection.
Fourth, the majority says that the DPA’s “hold harmless” provision (“No person shall be held liable for damages . . . for any act or failure to act resulting directly or indirectly from compliance with [an] order”) does not provide for indemnification. Ante, at 429. The petitioners, however, do not claim the contrary. They state explicitly that they “do not attempt to interpret the DPA’s hold harmless language as an affirmative indemnity.” Reply Brief for Petitioners 2. They add that “an indemnity should be implied from all the circumstances of this case, including the circumstance that petitioners and the Government contracted against the backdrop of the sweeping hold harmless language contained in the DPA.” Ibid. They argue simply that the DPA’s stated *440objective — to relieve them of involuntarily created liability— would have led contracting officers in the 1960’s (given the parties’ uncertainty about future statutory interpretation) to have believed that a contractual “hold harmless” warranty was reasonable in the circumstances, not the contrary. See 3 Corbin, Contracts § 551 (existing statutes and rules of law are always evidence of the meaning of the parties). The relevant point is not whether Congress intended to indemnify, but the likely effect of the DPA’s language (before judicial interpretation limited it to an immunity provision) on what risks contracting officers at the time might have thought the Government was assuming in a forced production contract under the Act.
Fifth, both the Federal Circuit, 24 F. 3d, at 198, n. 8, and the majority, ante, at 425, imply that a 1960’s contracting officer would not have accepted an indemnification provision because of Stencel Aero Engineering Corp. v. United States, 431 U. S. 666 (1977). That case held (in light of the Feres doctrine providing the Government with immunity from armed services personnel tort suits) that Government contractors, whom armed services personnel had sued in tort, could not, in turn, sue the Government for indemnification. Otherwise a soldier, unable (given Feres) to sue the Government for injury caused, say, by a defective rifle, would sue the rifle manufacturer instead, and the rifle manufacturer would then sue the Government for indemnity, thereby, in a sense, circumventing the immunity that Feres promised the Government.
One problem with this argument is that Stencel postdates the formation of the contracts here at issue by about a decade. More importantly Stencel does not involve contractual promises to indemnify a contractor. Rather it concerns an indemnification provided by state tort law. Stencel, supra, at 667-668, nn. 2, 3. And, it nowhere says, or directly implies, that the law prohibits the Government from agreeing, explicitly or implicitly, to indemnify a contractor. Indeed, *441this Court has explicitly written that it “fail[s] to see how the Stencel holding . . . supports the conclusion that if the Tort Claims Act bars a tort remedy, neither is there a contractual remedy. The absence of Government tort liability has not been thought to bar contractual remedies on implied-in-fact contracts, even in those cases also having elements of a tort.” Hatzlachh Supply Co. v. United States, 444 U. S. 460, 465 (1980) (per curiam). I agree with the majority insofar as it warns against a court’s too easily reading an implicit promise to indemnify a contractor’s armed-services-related tort liability; but, then, its words would represent simply a wise caution and not an absolute prohibition.
In sum, the companies argue factual circumstances — compelled production, superior knowledge, detailed specifications, and significant defect — which, if true, suggest that a government, dealing in good faith with its contractors, would have agreed to the “implied” promise, particularly in light of legal authorities, known at the time, that offered somewhat similar guarantees to contractors in somewhat similar circumstances. The validity of their claim is likely to turn on the strength of the companies’ factual case, as supported by evidence, and upon the details of Government contracting practices in the 1960’s — matters not now before us and with which the lower courts are more familiar than are we.
The Court today unnecessarily restricts Spearin warranties, and, lacking particular facts at this stage of the proceeding, it relies on statutory circumstances that are common to many Government contracts. I fear that the practical effect of disposing of the companies’ claim at this stage of the proceeding will be to make it more difficult, in other cases even if not here, for courts to interpret Government contracts with an eye toward achieving the fair allocation of risks that the parties likely intended.
For these reasons, I would remand this case for further proceedings.