The question before us is whether federal safety regulations preempt a state law claim, asserted in a federal diversity action, that a motor vehicle was defective because it lacked air bags.

This is an interlocutory appeal in a product liability diversity action brought in the United States District Court for the District of Massachusetts by plaintiff-appellee Patricia Wood against defendant-appellant General Motors Corporation. Wood has alleged that under Massachusetts law a vehicle manufactured by General Motors was defective because it was equipped with seat belts rather than air bags or some other type of "passive restraint." General Motors moved for summary judgment on the ground that Wood's state claim is preempted by the National Traffic and Motor Vehicle Safety Act, 15 U.S.C. Secs. 1381 et seq. (1982), and Federal Motor Vehicle Safety Standards promulgated under the Act. The district court denied the motion, Wood v. General Motors Corp., 673 F. Supp. 1108 (D.Mass.1987), and we have agreed to hear, in part, General Motors' interlocutory appeal from the court's action. 28 U.S.C. Sec. 1292(b) (1982).

On May 19, 1981, appellee Patricia Wood was on her way home from school, riding in the front passenger seat of a 1976 Chevrolet Blazer. The Blazer, manufactured by appellant General Motors, was equipped with seat belts and complied with all applicable federal motor vehicle safety regulations. Wood was not, however, wearing her seat belt. For some reason other than a defect in the vehicle, the Blazer left the road and collided with a tree. Suffering severe injuries, Wood was rendered quadriplegic.

In May 1984, Wood brought an action for damages in the United States District Court for the District of Massachusetts, claiming that General Motors was liable for her injuries under Massachusetts state law theories of negligent design, negligent manufacture, and breach of implied and express warranty. One paragraph of the complaint alleged that

Defendant negligently failed to provide reasonably safe and adequate safety devices, which include but are not limited to "air bag" devices, to protect passengers and minimize the seriousness of injuries in reasonably foreseeable circumstances which include collisions[.]

General Motors moved for summary judgment on all claims. It characterized Wood's complaint as being wholly dependent on the theory that the Blazer lacked passive restraints (i.e., safety devices that do not require any action by the passenger, e.g., air bags or automatic seat belts), and was therefore defectively designed. Such a theory, General Motors argued, was both preempted by federal safety regulations and invalid under the Massachusetts law of product liability. After briefing and argument, the district court denied General Motors's motion for summary judgment.

General Motors then moved for immediate interlocutory appeal under 28 U.S.C. Sec. 1292(b). The district court found that its denial of the motion for summary judgment met the statutory requirements, and authorized an immediate appeal. We granted in part General Motors's petition for immediate appeal, limiting our review to

the question whether federal law preempts a state law product liability claim against a motor vehicle manufacturer based on its installing seat belts, rather than airbags, in a motor vehicle.1

The National Traffic and Motor Vehicle Safety Act of 1966 (the "Safety Act" or "Act") was enacted by Congress in response to the "soaring rate of death and debilitation on the Nation's highways," S.Rep. No. 1301, 89th Cong., 2d Sess. 1, reprinted in 1966 U.S.Code Cong. & Admin.News 2709, 2709 [hereinafter "Senate Report"]. The Safety Act sought to increase automotive safety by authorizing the promulgation of federal "motor vehicle safety standards," hereinafter referred to as "FMVSS." 15 U.S.C. Secs. 1391(2), 1392(a) (1982). The FMVSS, to be written and administered by the Secretary of Transportation,2 were to be mandatory standards which would apply to all new motor vehicles. The delegation was broad; the new law stated only that each FMVSS "be practicable, shall meet the need for motor vehicle safety, and shall be stated in objective terms." 15 U.S.C. Sec. 1392(a).

Congress envisioned that the FMVSS would be addressed to two types of dangers: 1) vehicle defects which caused accidents, and 2) vehicle defects that aggravated injuries to the occupants once an accident had occurred. The latter problem, sometimes labeled "crashworthiness," received special attention from Congress. Senate Report, 1966 U.S.Code Cong. & Admin.News at 2712. Within the field of crashworthiness, Congress focused especially on the problem of the "second collision"--the potentially devastating impact between the vehicle's occupants and the vehicle's interior.3 The Senate Committee noted that the " 'second collision' ... has been largely neglected," and that "[r]ecessed dashboard instruments and the use of seat belts can mean the difference between a bruised forehead and a fractured skull." Id. at 2710-11. This appeal concerns FMVSS 208, which is the most important FMVSS addressed to the problem of the second collision.

When framing the Safety Act, Congress indicated clearly its intention that the primary responsibility for setting standards regulating the national automobile manufacturing industry rested upon the federal government, not the states. The Senate Report stated,While the contribution of the several States to automobile safety has been significant, and justifies to the States a consultative role in the setting of standards, the primary responsibility for regulating the national automotive manufacturing industry must fall squarely upon the Federal Government.

Id. at 2712 (emphasis added). The limited, consultative role of the states is reflected in two provisions of the Safety Act. First, 15 U.S.C. Sec. 1392(f) states that in prescribing FMVSS, the Secretary "shall ... consult with the Vehicle Equipment Safety Commission, and such other State or interstate agencies (including legislative committees) as he deems appropriate...." Second, the Safety Act contains a preemption provision which explicitly preempts any state safety standard--even state standards which are more stringent than the federal standards--which covers "the same aspect of performance" as a federal standard but which is "not identical" to the federal standard. 15 U.S.C. Sec. 1392(d).

With respect to safety standards implied under state tort law, however, the Safety Act's allocation of exclusive federal authority to establish automotive safety standards is less clear. Notwithstanding the above preemption provision, the Safety Act also contains section 1397, with the following general "savings clause":

Compliance with any Federal motor vehicle safety standard issued under this subchapter does not exempt any person from any liability under common law.

15 U.S.C. Sec. 1397(c). The Safety Act is thus ambiguous as to the current issue, to wit, whether a state product liability claim alleging improper design--which seeks to hold an automotive manufacturer liable for failing to adopt a standard "not identical" to a federal safety standard covering the "same aspect of performance"--is preserved by section 1397(c) and not preempted by section 1392(d), as would be the case were a Massachusetts law or regulation to have required adoption of the very same standard (i.e., provision of an air bag).

FMVSS 208, entitled "Occupant crash protection," has an intricate and contentious history of over 20 years. See generally State Farm Mutual Automobile Insurance Co. v. Department of Transportation, 680 F.2d 206, 209-18 (D.C.Cir.1982) (hereinafter "State Farm I "), vacated and remanded sub nom. Motor Vehicle Manufacturers Association v. State Farm Mutual Automobile Insurance Co., 463 U.S. 29, 34-38, 103 S. Ct. 2856, 2862-2864, 77 L. Ed. 2d 443 (1973) (hereinafter "State Farm II "); State Farm Mutual Automobile Insurance Co. v. Dole, 802 F.2d 474, 477-78 (D.C.Cir.1986), cert. denied, 480 U.S. 951, 107 S. Ct. 1616, 94 L. Ed. 2d 800 (1987) (hereinafter "State Farm III "). In this period, all three branches of the federal government have considered and, at times, forced changes in various aspects of the regulation. On at least three occasions, federal courts of appeals have considered the legality of FMVSS 208 under the Administrative Procedure Act, 5 U.S.C. Secs. 551 et seq. See Pacific Legal Foundation v. Department of Transportation, 593 F.2d 1338 (D.C.Cir.1979) (upheld); Chrysler Corp. v. Department of Transportation, 472 F.2d 659 (6th Cir.1972) (upheld in part and vacated in part); State Farm III, 802 F.2d at 474 (upheld). The Supreme Court, see State Farm II, 463 U.S. at 57, 103 S.Ct. at 2874, has vacated and remanded the standard. Congress, in overseeing its original delegation of authority in the 1966 Safety Act, has struck down a portion of FMVSS 208. See Motor Vehicle and Schoolbus Safety Amendments of 1974, Pub.L. No. 93-492, Sec. 109, 88 Stat. 1482 (codified at 15 U.S.C. Sec. 1410b) (instructing the Secretary to amend FMVSS 208, and forbidding the Secretary from requiring, or permitting compliance by means of, ignition interlocks). In addition, Congress has amended the Safety Act so that the Secretary may not require any restraint system other than a seat and shoulder belt system without giving Congress the opportunity to exercise a legislative veto. See 15 U.S.C. Sec. 1410b(b). Finally, the Department of Transportation has drastically reversed the course of FMVSS 208 on at least two occasions,4 and has made numerous less radical amendments.5

Most of this national controversy has concerned the very issue which underlies Wood's state law claim, namely, whether seat belts are adequate to protect occupants from the harmful effect of the "second collision," or whether vehicles should be equipped with "passive restraints," i.e., safety devices such as air bags that operate without the occupants' participation. Until recently, however, the controversy has not dealt with the role of state product liability law. In fact, the long, convoluted history of FMVSS 208 contains only a single, unhelpful reference to state design defect suits predicated on the theory that a car is defective if it lacks air bags.6

Despite this complex regulatory history, the portion of FMVSS 208 applicable to the 1976 Chevrolet Blazer has, for relevant purposes, remained unchanged since 1976. For the purpose of our preemption analysis, FMVSS 208 is the same whether one takes the regulation at the time of the Blazer's manufacture, at the time of the accident, at the time Wood's suit was filed, or at the time of decision. Under FMVSS 208, the 1976 Chevrolet Blazer is classified as a "multipurpose passenger vehicle, with GVWR [gross vehicle weight rating] of 10,000 pounds or less." 49 C.F.R. Sec. 571.208, S4.2.2 (1987). For the protection of front seat passengers, manufacturers of such vehicles are given three separate options, compliance with any one of which will satisfy FMVSS 208:

1) passive protection from frontal and angular collisions;

2) passive protection from head-on collisions, supplemented by seat belts and a belt warning system; or

3) lap and shoulder belts, plus a belt warning system.

49 C.F.R. Sec. 571.208, S4.2.2 (1987) (referring to S4.1.2.1, S4.1.2.2, and S4.1.2.3); see also 49 C.F.R. Sec. 571.208, S4.2.2 (Oct.1976).

It is uncontroverted that the Blazer complied with the third option (lap belts, shoulder belts, and a warning system) and, for that reason, complied with FMVSS 208. Wood's suit alleges, in effect, that the Blazer was defectively designed because it complied only with this third option, rather than with the first or second option, which calls for one or another form of passive restraint, e.g., an air bag.

The district court denied General Motors's motion for summary judgment based on federal preemption. General Motors had argued that under theories of "express preemption" or "implied preemption," or both, the Safety Act and FMVSS 208 preempted Wood's Massachusetts product liability claim. First, General Motors argued that section 1392(d) expressly preempted the state claim because it would, in the words of section 1392(d), "establish ... [a] safety standard" which was not identical to FMVSS 208. The district court rejected this argument, reasoning that 1) the language of section 1392(d) seemed to apply to state regulations, and not to actions at law; 2) if Congress had meant to preempt defective design actions, it would explicitly have mentioned them in section 1392(d); 3) the savings clause, section 1397(c), refuted any express congressional intent to preempt product liability actions; and 4) the existence of a presumption against preemption.

Second, General Motors argued that a state action based on a failure to install air bags would produce an irreconcilable conflict with FMVSS 208 and would frustrate the objectives of the Safety Act. The court rejected the former argument, explaining that General Motors could comply with both FMVSS 208 and the state standard simply by installing air bags. The district court also used a broader rationale purportedly based on the recent case of Silkwood v. Kerr-McGee Corp., 464 U.S. 238, 248, 104 S. Ct. 615, 621, 78 L. Ed. 2d 443 (1984): a product liability action does not conflict with federal safety regulations, the court thought, because even if a defendant manufacturer has to pay a damages award in this case, it need not alter its behavior in other instances. (Thus General Motors could, if it wished, continue to sell vehicles without passive restraints, taking its chances on future lawsuits similar to this one.) The court held further that Wood's claim would not frustrate the objectives of the Safety Act nor of FMVSS 208 because if it did have a regulatory effect, the result would be safer cars. The district court also recognized a "subsidiary purpose" of the Safety Act, namely, "Congress's intention that the safety standards be uniform throughout the country." It noted that the product liability suit might create some "tension" with this subsidiary purpose of the Safety Act, but held that under Silkwood v. Kerr-McGee Corp., this tension was acceptable. It explained,

Congress passed the Safety Act fully aware that damage suits would be initiated against parties who complied with the federal regulations. [Citations omitted.] This Court therefore concludes, as the Supreme Court did in Silkwood, that Congress recognized and sanctioned this "tension" between compliance with federal regulations and state common law claims. See Palmer [v. Liggett Group, Inc.], 633 F.Supp. [1171] at 1179 [ (D.Mass.1986) ] (citing and following Silkwood for similar reasons in not finding preemption by the Federal Cigarette Labeling and Advertising Act, 15 U.S.C. Secs. 1331-41).

673 F.Supp. at 1117. The district court then elaborated on its reasoning that both the majority and dissenting opinions in Silkwood establish that damages awards have very little regulatory effect.

In addition to the present action, about two dozen other suits have been recently filed claiming that an automobile was defectively designed because it lacked passive restraints. As yet, none of these cases has reached a federal appellate court or state supreme court. Courts have found both for and against preemption, with the majority ruling in favor of preemption.7

Against this backdrop, we now turn to the question certified to us.II. OUR CONCLUSIONS SUMMARIZED

In this interlocutory appeal, General Motors presses the argument that Wood's claim is both expressly and impliedly preempted by FMVSS 208 and section 1392(d) of the Safety Act. While we do not find an express preemption, we agree that Wood's claim is impliedly preempted.

Preemption is a matter of congressional intent. As a unanimous Supreme Court recently said,

A pre-emption question requires an examination of congressional intent.... Of course, Congress explicitly may define the extent to which its enactments preempt state law. See, e.g., Shaw v. Delta Air Lines, Inc., 463 U.S. 85, 95-96 [103 S. Ct. 2890, 2898-2899, 77 L. Ed. 2d 490] (1983). In the absence of explicit statutory language, however, Congress implicitly may indicate an intent to occupy a given field to the exclusion of state law. Such a purpose properly may be inferred where the pervasiveness of the federal regulation precludes supplementation by the States, where the federal interest in the field is sufficiently dominant, or where "the object sought to be obtained by the federal law and the character of obligations imposed by it ... reveal the same purpose." Rice v. Santa Fe Elevator Corp., 331 U.S. 218, 230 [67 S. Ct. 1146, 1152, 91 L. Ed. 1447] (1947). Finally, even where Congress has not entirely displaced state regulation in a particular field, state law is preempted when it actually conflicts with federal law. Such a conflict will be found " 'when it is impossible to comply with both state and federal law, Florida Lime & Avocado Growers, Inc. v. Paul, 373 U.S. 132, 142-143 [83 S. Ct. 1210, 1217-1218, 10 L. Ed. 2d 248] (1963), or where the state law stands as an obstacle to the accomplishment of the full purposes and objectives of Congress, Hines v. Davidowitz, 312 U.S. 52, 67 [61 S. Ct. 399, 404, 85 L. Ed. 581] (1941).' " California Coastal Comm'n v. Granite Rock Co. [480 U.S. 572,] 107 S. Ct. 1419, 1425 [94 L. Ed. 2d 577] (1987), quoting Silkwood v. Kerr-McGee Corp., 464 U.S. 238, 248 [104 S. Ct. 615, 621, 78 L. Ed. 2d 443] (1984).

Schneidewind v. ANR Pipeline Co., --- U.S. ----, 108 S. Ct. 1145, 1150-51, 99 L. Ed. 2d 316 (1988) (citations modified).

We find the Safety Act facially ambiguous as to Congress's intent in the present situation. The difficulty arises because section 1392(d) (the preemption clause) and section 1397(c) (the savings clause) send conflicting messages in these particular circumstances. For plaintiff Wood to win on her design claim, it must be judicially determined under state law that vehicles like hers are unsafe unless equipped with air bags. That finding is tantamount to announcing a state safety standard (vehicles must have air bags) that differs from the federal safety standard covering the same aspect of performance (viz., the relevant federal standard permits seat belts in lieu of air bags). 15 U.S.C. Sec. 1392(d). Such a safety standard, if promulgated by state statute or regulation, would be preempted by the clear commands of section 1392(d). Is the same standard preempted even though created by lawsuit? Section 1392(d) says nothing about being limited to legislatively established state standards. However, section 1397(c) of the Safety Act provides that compliance with a federal safety standard shall not be a defense to common law liability. Arguably this "preserves" a state or federal diversity court's right to create such a standard notwithstanding its conflict with the relevant federal standard.8

Broadly speaking, two analyses are possible. The one adopted by the court below--which has the virtue of simplicity--is to harmonize sections 1392(d) (the preemption clause) and 1397(c) (the savings clause) by ascribing to Congress an intent to preempt contradictory state safety standards when promulgated by statute or regulation, but not to preempt such standards when established in the course of a lawsuit. Congress is said to welcome, for some reason, the "tension" between the federal safety standards and any dissimilar state standards created by litigation--even though Congress strictly forbade the states to create dissimilar standards by statute or regulation.

The alternative analysis--which we prefer as being a more plausible view of Congress's intentions--begins with an examination of the historical setting within which Congress, in 1966, wrote the Safety Act. At that time, the only kind of legal claim which could give rise to the present dilemma--a cause of action based upon alleged automobile design defects--had yet to take its place in the arsenal of the plaintiffs' bar. We infer from this, as well as from the total silence of the legislative record concerning the present dilemma, that Congress simply did not anticipate the situation that now confronts us.9 While Congress intended that federal safety standards would not interfere with ongoing state litigation as then understood, it did not foresee the possibility of litigation that could, in practical effect, impose a new and conflicting state safety standard on national automobile manufacturers. Had Congress done so, we think, the same logic that dictated the insertion of section 1392(d) would inescapably have dictated that section 1392(d) extend to this situation.

Nonetheless, given Congress's failure to foresee this problem, we are not persuaded that section 1392(d) can be construed to manifest an express intention to preempt state design lawsuits having the present effect. By the same token, we do not construe the general language of section 1397(c) as intended to preserve the narrow class of design lawsuits that give rise to a conflicting state safety standard. We, therefore, reject General Motors's argument that the Safety Act works an express preemption, but we also reject Wood's contention that Congress meant by the savings clause to express approval of the "tension" that would ensue if federal safety standards and contradictory litigation-created state design standards were to co-exist. Rather, we think Congress simply overlooked the possibility of the present dilemma, resulting in a statute that lacks clear and express direction on the subject.

While we, therefore, do not find express preemption, we are convinced that Congress's purposes, as revealed in the Safety Act and in the legislative history, plainly imply a preemptive intent. The instant product liability claim alleging that the absence of an air bag rendered the vehicle's design faulty would, if upheld, clearly "stand as an obstacle" to the regulatory scheme of the Safety Act. A state common law action sustaining the theory that a vehicle was defective because it lacked an air bag would, in effect, create a state safety standard related to the same aspect of performance of FMVSS 208 but not identical to FMVSS 208. Such an action is, in our view, impliedly preempted because it would effectively circumvent section 1392(d)'s prohibition of nonidentical state standards covering the same aspect of performance as a federal safety standard. Allowing a common law action holding manufacturers liable for failing to install air bags in motor vehicles would be tantamount to establishing a conflicting safety standard that necessarily encroaches upon the goal of uniformity specifically set forth by Congress in this area.10

While our approach, as just stated, is clear, the arguments pro and con involve countless details and byways. In the following sections of this opinion, we explore these matters extensively.

The language of the Safety Act sends conflicting signals on the preemption of state actions for faulty design that could establish a safety standard not identical to a federal standard. The potential for ambiguity in special circumstances like the present is reflected in the preemption clause and the savings clause. The preemption clause provides:

Whenever a Federal motor vehicle safety standard established under this subchapter is in effect, no State or political subdivision of a State shall have any authority either to establish, or to continue in effect, with respect to any motor vehicle or item of motor vehicle equipment any safety standard applicable to the same aspect of performance of such vehicle or item of equipment which is not identical to the Federal standard. Nothing in this section shall be construed as preventing any State from enforcing any safety standard which is identical to a Federal safety standard. Nothing in this section shall be construed to prevent the Federal Government or the government of any State or political subdivision thereof from establishing a safety requirement applicable to motor vehicles or motor vehicle equipment procured for its own use if such requirement imposes a higher standard of performance than that required to comply with the otherwise applicable Federal standard.

15 U.S.C. Sec. 1392(d). The savings clause provides:

Compliance with any Federal motor vehicle safety standard issued under this subchapter does not exempt any person from any liability under common law.

15 U.S.C. Sec. 1397(c).

General Motors and Wood each use the provision favorable to its or her viewpoint to construct arguments that Congress expressly meant either to preempt or to preserve the instant tort action. General Motors argues that as "no State ... shall have any authority ... to establish ... any safety standard applicable to the same aspect of performance ... which is not identical to the Federal standard," id. Sec. 1392(d), and as an air bag suit would create a state air bag standard nonidentical to FMVSS 208, Congress expressly meant to preempt this type of action. Wood argues that preempting her theory of liability would effectively "exempt" General Motors from liability, and that Congress has expressly forbidden such a result: "compliance with any Federal motor vehicle safety standard ... does not exempt any person from any liability under common law." 15 U.S.C. Sec. 1397(c). The district court accepted this latter reading of congressional intent. It expressed the view that

Congress passed the Safety Act fully aware that damage suits would be initiated against parties who complied with the federal regulations.... This Court therefore concludes, as the Supreme Court did in Silkwood, that Congress recognized and sanctioned this "tension" between compliance with federal regulations and state common law claims.

673 F.Supp. at 1117.

We agree with neither General Motors's reading of Congress's express intent nor with that of Wood and the district court. We instead reach a third conclusion: Congress in 1966 did not contemplate the likelihood that there would be a state tort action that would effectively create a state design standard conflicting with a federal safety standard. As it did not envisage this peculiar type of lawsuit, and as no reason appears in the legislative history or comes to mind as to why Congress would have meant to endorse a state claim having the same effect as a forbidden state regulation, we do not find that Congress meant to sanction the "tension" between the FMVSS and state common law existing in the current situation.11 This interpretation of the Safety Act and of Congress's intent is first suggested by the very fact that the two provisions, read alone, yield such different results. If Congress had considered the instant type of tort claim in these facts, it would scarcely have left unexplained such a glaring ambiguity. Our reading is supported by the state of tort law when the Safety Act was passed in 1966 and the legislative history of the Act.

In determining questions of preemption, a court "must examine the [act's] language against the background of its legislative history and historical context." California Federal Savings & Loan Association v. Guerra, 479 U.S. 272, 107 S. Ct. 683, 691, 93 L. Ed. 2d 613 (1987). Examining the historical context of the Safety Act's enactment in 1966, we find it unlikely that anyone in Congress would have considered the possible preemption of certain defective design actions because there was, up to 1966, hardly any experience at all with cases of that type. Until the Eighth Circuit's seminal decision in Larsen v. General Motors Corp., 391 F.2d 495 (8th Cir.1968) (allowing the jury to decide whether a vehicle with a rigid steering linkage which caused severe injuries to the driver in a front-end collision was defectively designed), persons injured in automobile accidents seldom sued manufacturers on a theory of design defect, and with the exception of one decision involving a bus, see Carpini v. Pittsburgh & Weirton Bus Co., 216 F.2d 404 (3d Cir.1954), they did not prevail. As one commentator noted in 1956, "[t]here is a dearth of reported cases where negligence in design was urged by plaintiff as a basis for liability." Katz, Liability of Automobile Manufacturers for Unsafe Design of Passenger Cars, 69 Harv.L.Rev. 863, 863 n. 5 (1956). The commentator concluded that "[i]t is indeed striking to observe how ineffective the law has been to date in doing anything about effectuating improvements in design to alleviate the daily 'blood bath of automobile inspired tragedy.' " Id. at 873 (citation omitted).

Courts were hardly more active in the ten years between the writing of that comment and the passage of the Safety Act. In 1967, an article co-authored by Ralph Nader decried the failure of courts to promote automotive safety:

However, this trend [toward increased liability] has not extended to automobile manufacturers' liability for unsafe design of passenger cars. While a number of suits involving unsafe design have been settled, plaintiffs have had little luck on the trial court level. On appeals, plaintiffs are batting zero. The appellate courts have yet to reverse a judgment for a manufacturer or affirm a judgment for a plaintiff in a case involving a traffic accident allegedly caused by the unsafe design of an American passenger car.

Nader & Page, Automobile Design and the Judicial Process, 55 Calif.L.Rev. 645, 645-46 (1967). See also Recent Cases, 80 Harv.L.Rev. 688 (1967) (case comment on Evans v. General Motors Corp., 359 F.2d 822 (7th Cir.1966)); Note, Manufacturer's Liability for an "Uncrashworthy" Automobile, 52 Cornell L.Q. 444 (1967). Seen in this context, it would not be surprising for the Congress which enacted the Safety Act to have failed to perceive the possible conflict between state defective design suits and the new federal motor vehicle safety standards.

We do not mean to suggest that broadened design liability, as reflected by Larsen v. General Motors Corp., was entirely unforeseeable. At least since 1916 manufacturers had been held liable for manufacturing defects which caused accidents. See MacPherson v. Buick Motor Co., 217 N.Y. 382, 111 N.E. 1050 (1916). Furthermore, the Restatement of Torts contained a general principle of liability for defective design. See Restatement of Torts Sec. 398 (1938).12

However, despite the theoretical possibility of bringing them, defective design actions against automobile manufacturers were an unpromising and for that reason, virtually non-existent breed in 1966. First, it was not at all clear just how the law relative to design would be applied. In an early article advocating increased liability for defective design, Professor Noel noted that

in the design cases, particularly those involving widely-used products made by established manufacturers, judges and juries have been understandably hesitant to impose liability. This hesitation results partly from a reluctance to let a jury pass on a product prepared by experts in the field, and partly from the realization that a judgment for a particular plaintiff may open the door to many additional claims and suits.

Noel, Manufacturer's Negligence of Design or Directions for Use of a Product, 71 Yale L.J. 816, 816 (1962). See also Nader & Page, 55 Calif.L.Rev. at 653-54 (noting that although courts recognize "some sort of duty with respect to design," "a duty with respect to design which is defined in terms so narrow as to eliminate virtually any possibility of finding breach of duty differs little from a judicial determination that no duty exists."). Indeed, as of 1966, no plaintiff had yet prevailed on a claim that an automobile was defectively designed. In the legal climate of 1966 one could not tell to what degree, if any, design defect suits would ever have an effect on automobile safety.

Furthermore, prior to the Larsen decision, plaintiffs had an especially tough burden if they alleged not that the design defect caused the accident but that a design defect made a car less safe in an accident or in other situations of misuse--types of defects which, as noted above, were a special concern of the Safety Act. Under either the theory of "intended use" or "patent danger," see Note, 52 Cornell L.Q. at 447-49, a plaintiff would likely lose because the automobile was not intended to be misused or because plaintiff should have been aware of the alleged defect and was thus solely at fault. See, e.g., Amason v. Ford Motor Co., 80 F.2d 265 (5th Cir.1935) (alleged defective design was passenger door hinged at the rear and a door handle which could catch a passenger's hand, resulting in the death of plaintiff when he tried to open and shut the door while the car was moving at high speed; summary judgment affirmed for defendant because decedent's actions were not "ordinary use" and decedent should have known of the danger); Poore v. Edgar Brothers Co., 33 Cal. App. 2d 6, 90 P.2d 808 (1939) (alleged defective design was the use of plain glass rather than safety glass in automobile door; demurrer upheld for defendant because "the glass in [automobile] windows is not intended to withstand blows as part of its ordinary use"); Muncy v. General Motors Corp., 357 S.W.2d 430 (Tex.Civ.App.1962) (alleged defective design was ignition key switch which allowed key to be removed while car was running and in gear, resulting in accidental injury to a pedestrian; plea of privilege upheld for defendant because driver was "well aware" of the design of the key switch and because driver failed to use the car for its intended purpose).13

In fact, the very year that Congress passed the Safety Act, a federal appeals court ruled that a plaintiff had failed to state a claim by alleging that a car was uncrashworthy. See Evans v. General Motors Corp., 359 F.2d 822 (7th Cir.), cert. denied, 385 U.S. 836, 87 S. Ct. 83, 17 L. Ed. 2d 70 (1966) (alleged design defect was the use of an X-shaped frame which lacked protective side rails). The court invoked the doctrines of intended use--"The intended purpose of an automobile does not include its participation in collisions"--and patent danger--"A manufacturer is not under a duty to make his automobile accident-proof or fool-proof; nor must he render the vehicle 'more' safe where the danger to be avoided is obvious to all." Id. at 824-25. The Evans court explained that requiring manufacturers to design crashworthy cars might be a desirable goal, but that "that would be a legislative function, not an aspect of judicial interpretation of existing law." Id. at 824.

Congress may or may not have known of the Evans decision when it passed the Safety Act, but Evans plainly shows that in the historical context of 1966, persons then concerned with automobile safety would likely have attached little practical importance to claims of design defects alleged to have rendered a car uncrashworthy.

Even if Congress foresaw defective design lawsuits, a further reason Congress would not have anticipated the current preemption problem is that, even among the body of lawsuits claiming defective design, only a few would be likely, as here, to establish what is in practical effect a new safety standard. This is so because the Safety Act authorizes promulgation of "performance" rather than design standards as such. 15 U.S.C. Sec. 1391(2). Although, as this case shows, there can be a direct clash between a standard for performance and for design, see Section IVC, D, supra, it will be rare that a design standard established in a lawsuit will overlap and conflict with a performance standard established under the Safety Act.

We conclude that it is unrealistic to ascribe to the authors of the Safety Act, and to Congress generally, an awareness that in the years ahead a new breed of state tort actions would be developed from which design standards might emerge that, on some rare occasion, might create a direct conflict with a particular FMVSS. Thus when Congress inserted the savings clause, it did not contemplate that lawsuits would be brought with the potential to give rise to the current dilemma.

The legislative history of the Safety Act likewise indicates that neither the preemption clause nor the savings clause specifically addressed the status of conflicting design standards established through the mechanism of common law suits for defective automobile design. Neither party has directed us to a single legislative reference to actions for design defects. Instead, the legislative history, like the savings clause itself, employs more general terms such as "common law standards of care," Senate Report, 1966 U.S.Code Cong. & Admin.News at 2720, "rights of parties under common law particularly those relating to warranty, contract, and tort liability," H.Rep. No. 1776, 89th Cong., 2d Sess. 24 (1966), "common-law remedy," 112 Cong.Rec. 19,663 (1966) (remarks of Representative Dingell), and "common law on product liability," 112 Cong.Rec. 14,230 (1966) (remarks of Senator Magnuson). These, as indicated above, would not before 1966 have been based upon claims of inadequate design.

The fullest discussion in the legislative history of the Safety Act's impact on state law leaves the exact same ambiguity as the Act's conflicting preemption and savings clauses. The Senate Report has a brief section with the promising title, "Effect on State Law":

The centralized, mass production, high volume character of the motor vehicle manufacturing industry in the United States requires that motor vehicle safety standards be not only strong and adequately enforced, but that they be uniform throughout the country. At the same time, the committee believes that the States should be free to adopt standards identical to the Federal standards, which apply only to the first sale of a new vehicle, so that the States may play a significant role in the vehicle safety field by applying and enforcing standards over the life of the car. Accordingly, State standards are preempted only if they differ from Federal standards applicable to the particular aspect of the vehicle or item of vehicle equipment (sec. 104).

The States are also permitted to set more stringent requirements for purposes of their own procurement. Moreover, the Federal minimum safety standards need not be interpreted as restricting State common law standards of care. Compliance with such standards would not necessarily shield any person from product liability at common law.

Senate Report, 1966 U.S.Code Cong. & Admin.News at 2720. On the one hand, "State standards are preempted only if they differ from Federal standards." On the other hand, the Safety Act "need not be interpreted as restricting State common law standards of care." The dichotomy, we believe, reflects a Congress that never had occasion to consider what would occur if a state product liability action itself created an overlapping automobile design standard so similar to FMVSS as to raise the question of the former's preemption by the latter.14 Instead Congress had in mind the then customary "State common law standards of care" pertaining to the negligent operation or manufacture of vehicles, coupled, no doubt, with warranty claims then customary, none of which had the potential to create the present problem.

We accordingly divine no specific congressional intent in section 1392(d) expressly to preempt an action of the present type. Nor do we construe the savings clause (section 1397(c)) as reflecting any specific congressional concern to preserve design actions such as the present which would, in effect, create standards different from federal safety standards promulgated under the Safety Act.

Having found that Congress did not, in the language of the Safety Act, make an explicit statement regarding preemption of state claims that create, in effect, a state design standard not identical to a federal standard, we turn next to whether the Safety Act impliedly preempts the present action. Cf. International Paper Co. v. Ouellette, 479 U.S. 481, 107 S. Ct. 805, 812, 93 L. Ed. 2d 883 (1987) ("Given that the Act itself does not speak directly to the issue, the Court must be guided by the goals and policies of the Act in determining whether it in fact preempts [state law].").15 The kind of implied preemption that would be applicable here is "where the state law stands as an obstacle to the accomplishment of the full purposes and objectives of Congress." Schneidewind v. ANR Pipeline Co., --- U.S. ----, 108 S. Ct. 1145, 1151, 99 L. Ed. 2d 316 (1988) (citing Hines v. Davidowitz, 312 U.S. 52, 67, 61 S. Ct. 399, 404, 85 L. Ed. 581 (1941)). See, e.g., Hyde Park Partners, L.P. v. Connolly, 839 F.2d 837 (1st Cir.1988) (holding that the Williams Act preempted the Massachusetts anti-takeover statute; the state law stood as an obstacle to Congress's preferred balance between management and shareholders); Hernandez-Colon v. Secretary of Labor, 835 F.2d 958 (1st Cir.1988) (holding that the Job Training Partnership Act preempted a Puerto Rico law giving the Governor veto power over the formation of intermunicipal bodies; the state law stood as an obstacle to the federal law's equilibrium between state and local interests).

In a case decided after the district court had ruled on General Motors's motion for summary judgment, the Supreme Court made clear that in implied preemption analysis, one must look to interference with Congress's chosen method as well as to the ultimate goal of the statute. See International Paper Co. v. Ouellette, 479 U.S. 481, 107 S. Ct. 805, 93 L. Ed. 2d 883 (1987). Ouellette dealt with the implied preemption of common law water pollution suits by the Clean Water Act ("CWA"). The Court explained,

In determining whether Vermont nuisance law "stands as an obstacle" to the full implementation of the CWA, it is not enough to say that the ultimate goal of both federal and state law is to eliminate water pollution. A state law also is preempted if it interferes with the methods by which the federal statute was designed to reach this goal.

Id. 107 S.Ct. at 813 (citing Michigan Canners & Freezers Association v. Agricultural Marketing & Bargaining Board, 467 U.S. 461, 477, 104 S. Ct. 2518, 2527, 81 L. Ed. 2d 399 (1984)).

We believe that this "stand as an obstacle" type of preemption preempts Wood's design defect claim based on the absence of an air bag. Even though the goal of Wood's design defect action might be the same as that of the Safety Act--that is, to increase automobile safety--Wood's theory of recovery is preempted by FMVSS 208 and the Safety Act because it interferes with the method by which Congress intended to meet this goal.

Our analysis, set out below, has three steps. (1) A state regulation requiring passive restraints would be expressly preempted by section 1392(d) of the Safety Act. (2) Wood's state law tort action would have an effect similar to such a prohibited state regulation. (3) As Wood's design defect claim, if successful, would have the same effect as an impermissible state regulation, it is preempted because it stands as an obstacle to Congress's chosen method for achieving auto safety.

1. A State Air Bag Regulation Would Be Preempted

There is no question but that a nonconforming air bag or other passive restraint regulation issued by the Massachusetts legislature or state or local agency would be expressly preempted by section 1392(d) of the Safety Act. The state regulation might provide, for example, that "all vehicles registered in this State must be equipped with passive restraints." The federal standard, as applicable to multipurpose vehicles like the Blazer, has three optional modes of compliance, any one of which would be acceptable: vehicles must be equipped with seat belts and shoulder harnesses, with passive restraints, or with some combination of lap belts and passive restraints. 49 C.F.R. Sec. 571.208, S4.1.2. Comparing the federal standard and hypothetical state regulation, the state regulation is "applicable to the same aspect of performance" as the federal standard, but is "not identical" to the federal standard. Whereas the federal standard allows seat belts and shoulder harnesses alone, the hypothetical state regulation requires passive restraints. Thus, the state standard is expressly preempted by section 1392(d), which provides that "no State or political subdivision of a State shall have any authority either to establish, or to continue in effect, with respect to any motor vehicle or item of motor vehicle equipment any safety standard applicable to the same aspect of performance of such vehicle or item of equipment which is not identical to the Federal standard."

In the hypothetical regulation presented, which is directly applicable to the present case, application of section 1392(d)'s preemption of standards "applicable to the same aspect of performance" as a federal standard is neither difficult nor controversial. However, in order to show the high degree of conflict between such a state standard and FMVSS 208, we elaborate on just why a state air bag regulation is preempted by section 1392(d) and FMVSS 208.

The key issue in deciding whether a state regulation is preempted by section 1392(d) is defining the "aspect of performance" addressed by the applicable FMVSS. The question is difficult and important: depending on how broadly one defines the "aspect of performance," the federal regulation will have either a broad or narrow scope of preemption. Fortunately, in Chrysler Corp. v. Rhodes, 416 F.2d 319 (1st Cir.1969), this court already has outlined the procedure for deciding the "aspect of performance" of a FMVSS.

In Chrysler Corp. v. Rhodes, the question was whether a New Hampshire regulation prohibiting the sale of cars equipped with a supplementary light for night driving (the "Super Lite") was preempted by FMVSS 108, which regulated "lamps, reflective devices, and associated equipment." FMVSS 108 had no provisions directly addressed to the Super Lite. Nonetheless, Chrysler claimed that the New Hampshire regulation was preempted by FMVSS 108 and section 1392(d). Chrysler argued that the "purpose and scope" section of FMVSS 108 established the "aspect of performance" for purposes of preemption under section 1392(d). FMVSS 108, section 1, provided:

Purpose and scope. This standard specifies requirements for lamps, reflective devices, and associated equipment, for signalling and to enable safe operation in darkness and other conditions of reduced visibility.

416 F.2d at 322. Relying on this section, Chrysler argued that the "aspect of performance" addressed by FMVSS 108 was "safe operation in darkness," and that a state regulation covering the Super Lite was preempted because it too related to safe operation in darkness. We rejected this argument.

While "the purpose and scope" section of a federal standard may well be the starting point in defining "aspect of performance," the inquiry cannot end there. In our view, resort must be had to the specific requirements and categories of the standard....

Id. at 325. As FMVSS 108 had no specific requirements addressed to items like the Super Lite, we held that the New Hampshire regulation was not preempted.

In this case, the state air bag regulation is not preempted solely because it falls within the broad purpose and scope provisions of the federal standard.16 Rather, the air bag regulation is preempted because it is addressed to "the same aspect of performance," no matter how narrowly one defines that term, as the federal regulation. First, by including passive restraints as an option for complying with FMVSS 208, NHTSA has left no doubt that passive restraints are within the "aspect of performance" addressed by the federal standard. Furthermore, the history of FMVSS 208 shows irrefutably that passive restraints are within the "aspect of performance" covered by the standard. All of the regulatory changes, lawsuits, and congressional actions have focused on the same question as that presented by Wood's lawsuit: are seat belts adequate to protect occupants in crashes, or are passive restraints required? See pages 398-399 supra. In these circumstances, we have no doubt that a hypothetical state air bag regulation would be addressed to the same "aspect of performance" as FMVSS 208, and would thus be expressly preempted by section 1392(d) of the Safety Act.

2. Wood's Tort Action Would Have An Effect Similar To A

If Wood prevailed in her defective design claim, the effect would be similar to state promulgation of the hypothetical state air bag regulation discussed above. The Supreme Court has recognized the regulatory effect of damages awards.

[R]egulation can be as effectively exerted through an award of damages as through some form of preventive relief. The obligation to pay compensation can be, indeed is designed to be, a potent method of governing conduct and controlling policy.

San Diego Building Trades Council v. Garmon, 359 U.S. 236, 247, 79 S. Ct. 773, 780-781, 3 L. Ed. 2d 775 (1959). In Palmer v. Liggett Group, Inc., 825 F.2d 620 (1st Cir.1987), this court recently held that a state tort suit was preempted because it had a regulatory effect. Palmer involved the preemptive scope of the Federal Cigarette Labeling and Advertising Act as applied to a state tort suit based on a theory of inadequate warning. Like the Safety Act, the Labeling Act forbade states from imposing different labeling requirements, but did not expressly mention state tort suits. We thus accepted plaintiff's argument that the tort suit was not expressly preempted. Nonetheless, we ruled that the suit was impliedly preempted because it had a similar effect to direct state regulation. We explained that if this suit were not preempted, it would "arrogate[ ] to a single jury the regulatory power explicitly denied to all fifty states' legislative bodies." Id. at 628.17

The Massachusetts Supreme Judicial Court has similarly acknowledged that the purpose of automobile design defect suits is to set state design standards, and that such suits have much the same effect as legislative standards. In Smith v. Ariens, 375 Mass. 620, 377 N.E.2d 954 (1978), the Supreme Judicial Court considered whether it would follow Evans v. General Motors Corp. (not allowing suits based on uncrashworthiness) or Larsen v. General Motors Corp. (seminal case allowing suits based on uncrashworthiness, see supra ). The court decided to follow Larsen, explaining as follows:

The major argument against the imposition of liability for negligent design which results in enhanced injury is that the Legislature, rather than the judiciary, should determine design standards. See Evans v. General Motors Corp. ... However, as noted in Larsen, "[t]he common law is not sterile or rigid and serves the best interests of society by adapting standards of conduct and responsibility that fairly meet the emerging and developing needs of our time. The common law standard of a duty to use reasonable care in light of all the circumstances can at least serve the needs of our society until the legislature imposes higher standards."

Smith v. Ariens, 375 Mass. at 624-25, 377 N.E.2d at 957 (citing Larsen v. General Motors Corp., 391 F.2d at 506).

Wood argues that her suit will have no regulatory effect because General Motors has the choice of paying the damages award without modifying the design of its vehicles. This argument might have some force in certain cases, if, for example, the claimed design defect was highly particular to the vehicle involved, or the accident was in some way unusual. With respect to a claim that a vehicle is defective because it lacks air bags, however, it seems obvious that a damages award will have a pronounced effect on the manufacturer's future conduct. The same theory of recovery could be pursued by every front seat occupant injured in a multipurpose vehicle. It is well known that injuries to front seat occupants are frequent; well over 20 lawsuits have been filed on the theory of absence of air bags in the last two years. See section I.E supra. In these circumstances, General Motors's choice to avoid modification of its design "seems akin to the free choice of coming up for air after being underwater." Palmer, 825 F.2d at 627.

We also disagree with the district court's statement that in Silkwood v. Kerr-McGee Corp., 464 U.S. 238, 104 S. Ct. 615, 78 L. Ed. 2d 443 (1983), the Supreme Court "found that the impact of such [tort] awards should be considered narrowly."18 673 F.Supp. at 1117. Silkwood, in fact, affirms that tort awards have regulatory effect. The issue in Silkwood was the preemptive scope of federal nuclear safety regulations on state tort law. Nuclear safety was a difficult area because Congress had enacted two potentially conflicting sets of laws. On the one hand, the Supreme Court had recently decided that federal safety regulations occupied the field of nuclear safety, precluding any direct state regulation. 464 U.S. at 240-41, 104 S.Ct. at 617-18. On the other hand, Congress had enacted the Price-Anderson Act, 42 U.S.C. Sec. 2210 (1982), which limited the aggregate liability of nuclear facilities resulting from accidents. The very existence of the Price-Anderson Act implied that Congress intended that nuclear facilities be liable for damages caused by nuclear accidents.

The specific issue in Silkwood was whether a plaintiff could receive punitive damages; her right to compensatory damages was unquestioned. The defendant argued that as punitive damages are imposed solely to change conduct, these types of damages were impliedly preempted. In a 5-4 decision, the Court found that punitive damages awards were not preempted.

The Court's reasoning, however, affirmed that compensatory damages awards have a regulatory effect. It agreed with the defendant that "there is a tension between the conclusion that safety regulation is the exclusive concern of the federal law and the conclusion that a State may nevertheless award damages based on its own law of liability." 464 U.S. at 256, 104 S.Ct. at 625-26. The "tension" was that "the award of damages based on state law is regulatory in the sense that a nuclear plant will be threatened with damages liability if it does not conform to state standards." Id. Silkwood does not support the proposition that damages awards lack regulatory effect.19

We recognize that in Silkwood, after thoroughly examining the legislative history of the Price-Anderson Act, the Court held that Congress intended that there be a "tension" between the Price-Anderson Act and the exclusive federal control of nuclear safety. As the Court explained, since Congress decided to tolerate this tension, "[w]e can do no less." Wood insists that the present case is just like Silkwood--Congress intended there to be a tension between the savings clause (section 1397(c)) and the preemption provision of the Safety Act, and it is the courts' duty to abide by this congressional decision favoring tension. We find no parallel, however, between the circumstances here and in Silkwood. Wood is unable to suggest any convincing reason why Congress would want to encourage states to impose an inconsistent safety standard of this nature by lawsuits but not by regulations. We shall discuss this further below.

We think it plain that Wood's action, if successful, would stand as an obstacle to Congress's chosen method of increasing automobile safety. See Ouellette, 107 S. Ct. at 813. While "courts should not lightly infer preemption," Wood's product liability action must bow to the supremacy of federal law because it disturbs the Safety Act's allocation of authority between the federal government and the states. Id. at 811.

Congress, in enacting the Safety Act, set up a clear division of authority between the states and the federal government. If the federal government has not issued a safety standard on a certain aspect of performance, the states are allowed to set their own standards in these areas. However, Congress decided that once the federal government had promulgated a standard, the states' usual role in setting safety standards was subordinated in the interest of national uniformity. When a federal standard is in effect, the states are prohibited from establishing a nonidentical standard, see 15 U.S.C. Sec. 1392(d); and their role is then limited to enforcing the federal standard, id., and to issuing higher standards for vehicles procured for their own use. Id. The states are also free to consult with federal authorities. Id. at Sec. 1392(f).

This division of authority between state and federal government was part of Congress's chosen method for implementing the Safety Act. Congress believed that for the federal standards to be effective, they had to be uniform throughout the country. See 15 U.S.C. Sec. 1392(d); Senate Report, 1966 U.S.Code Cong. & Admin.News at 2720 ("The centralized, mass production, high volume character of the motor vehicle manufacturing industry in the United States requires that motor vehicle safety standards be not only strong and adequately enforced, but that they be uniform throughout the country.").

Congress thus intended that all of the FMVSS, including FMVSS 208, be uniform national standards. A state regulation requiring passive restraints would be expressly preempted because it would destroy the uniformity of the federal standard, and exceed the Safety Act's prescribed role for state regulation. By the same token, a defective design action which, if successful, effectively would require the installation of passive restraints, would likewise destroy the national uniformity of the federal standard and exceed the state's authority. Thus, although Wood's air bag suit is not expressly preempted by the Safety Act, it is impliedly preempted because it presents an "actual conflict" with the Safety Act--specifically because it "stands as an obstacle" to Congress's determination that safety is best served by having uniform national standards.

Wood's response, as already indicated, is that Congress manifested its intent via the savings clause to allow Wood's air bag suit notwithstanding its direct impact on the uniformity of federal standards, and notwithstanding Congress's refusal to allow the states to mandate air bags through direct regulation. As preemption is always a question of congressional intent, we would, of course, find no preemption if we were satisfied that Congress did, in fact, intend to allow the courts to accomplish what a state was otherwise prohibited from doing. We are not, however, so convinced.

The Supreme Court in Silkwood did not lay down some kind of general rule that "tension" between federal regulation and clashing state product liability law is generally favored. Silkwood merely held, on its own facts, that such tension was there envisaged by Congress. In Silkwood, the Court had the entire Price-Anderson Act as evidence that Congress intended that nuclear facilities would be subject to common law liability. The availability of state actions for compensatory damages was, indeed, uncontested. The only issue was the availability of punitive damages, and the Court observed that "punitive damages have long been a part of traditional state tort law." In contrast, the only evidence here of a possible congressional intent to sanction the handful of design defect actions that might create standards which clash directly with federal standards is section 1397(c), a general savings clause which applies to the universe of state tort actions. As earlier explained, automobile design defect actions, particularly ones based on a theory of uncrashworthiness, were not in 1966 an established "part of traditional tort law," and the legislative history gives no evidence that Congress anticipated the present dilemma. For this and other reasons discussed below, the language in the savings clause refusing to exempt persons from liability under common law because of compliance with a safety standard does not reflect Congress's intention to avoid preemption of those few and unique design actions that would have the effect of creating a state safety standard in direct conflict with an existing FMVSS.

In Silkwood, moreover, regulations implementing the Price-Anderson Act specifically mentioned punitive damages, indicating that the federal agency believed that punitive damages were allowed. Id. at 255, 104 S.Ct. at 625. In this case, there is no comparable signal from NHTSA indicating that state product liability actions should be allowed to establish state safety standards that conflict directly with relevant federal standards.

Less obvious, but even more important, the preemption claimed by the defendant in Silkwood was more akin to the "occupy the field variety" than the "actual conflict" type (which applies in this case). In Silkwood, the theory of liability did not focus on any specific portion of the nuclear facility's design nor on any particular procedure of the defendant. Accordingly, the suit had little potential for establishing a state safety standard; the opinion never discusses a single federal safety standard that was potentially threatened by a judicially created state standard. The court was careful to state that if a common law suit did present an actual conflict with federal law, the suit would be preempted.

We do not suggest that there could never be an instance in which the federal law would pre-empt the recovery of damages based on state law. But insofar as damages for radiation injuries are concerned, pre-emption should not be judged on the basis that the Federal Government has so completely preoccupied the field of safety that state remedies are foreclosed but on whether there is an irreconcilable conflict between the federal and state standards or whether the imposition of a state standard in a damage action would frustrate the objectives of federal law. We perceive no such conflict or frustration in the circumstances of this case.

Id. at 256, 104 S. Ct. at 625-26. We believe the Court would be much less willing to find that Congress created a "tension" between federal law and state common law when the state law, as here, evidences an actual conflict. Where Wood's action creates an actual conflict, Silkwood offers little support for Wood's contention that Congress intended state product liability suits to impose standards that would disrupt the uniformity of federal safety standards.

We conclude that the air bag claim, being in direct conflict with federal safety standards, is impliedly preempted by the Safety Act, and that the savings clause does not evidence a contrary congressional intention.

A. The FMVSS are Not Merely Minimum Standards

Wood argues that because the Safety Act defines the FMVSS as minimum standards, they cannot preempt more stringent state standards.20 This argument has no merit. Although the standards are "minimum" in the sense that a manufacturer may make a vehicle safer than required by federal law, the standards are not "minimum" in relation to state law. Section 1392(d) provides that states are forbidden from establishing standards not "identical" to a federal standard. Thus section 1392 preempts state standards which are either more or less stringent than the federal standard.

B. The Savings Clause Does Not Foreclose Implied Preemption

Wood argues that two recent Supreme Court cases establish the following "principle": when a statute contains a savings clause, "it is not within the province" of a court to engage in implied preemption analysis. Applying this principle, Wood argues that "Congress has expressly stated its intention" to preserve the state air bag claim, and that we should thus not look at implied preemption.

We disagree both with Wood's "principle" and with the proposed application of the principle to this case. With regard to the latter point, the Safety Act, even on its face, does not express an unambiguous intent to preserve a state suit that conflicts with a federal safety standard. As noted in Section IIIA, above, the Safety Act contains both a savings clause which would preserve the claim and a preemption provision that would preclude the claim. Neither provision can be read in isolation, and thus we have no choice but to look beyond the face of the statute to find Congress's intent with respect to preemption of the air bag claim. Thus even if Wood's principle were correct, it would not apply in the present case.

Moreover, Wood's proposed principle is not supported by the Supreme Court cases on preemption. Wood mainly relies on California Federal Savings & Loan Association v. Guerra, 479 U.S. 272, 107 S. Ct. 683, 93 L. Ed. 2d 613 (1987), which addressed the preemptive scope of the pregnancy discrimination provisions of the Civil Rights Act. Wood in particular focuses on Justice Marshall's statement21 that because the Civil Rights Act has two savings clauses, "there is no need to infer congressional intent to pre-empt state laws from the substantive provisions of the [act]." Id. at 690. Wood fails to mention, however, that one of the savings clauses in the Civil Rights Act provides, in effect, that implied preemption analysis must be applied to state laws. See 42 U.S.C. Sec. 2000h-4 (preempting state laws that are "inconsistent with any purposes of this Act, or any provision thereof"). Accordingly, Justice Marshall stated that part of his inquiry was whether the state law was "inconsistent with the purposes of the statute." Thus Wood's proposed principle--which applies in situations where a savings clause arguably forecloses the use of implied preemption analysis--has absolutely nothing to do with Guerra--a situation where the "savings clause" mandates implied preemption analysis. In fact, it is ironic that Wood cites Justice Marshall's opinion to argue against implied preemption--both Justice Scalia, concurring separately, and Professor Tribe have criticized Justice Marshall's opinion for ranging too far afield in its implied preemption analysis. See id. 107 S.Ct. at 697 (Scalia, J., concurring in the judgment); L. Tribe, American Constitutional Law Sec. 6-26, at 482-83 n. 8 (2d ed. 1988).

The other case cited by Wood, California Coastal Commission v. Granite Rock Co., 480 U.S. 572, 107 S. Ct. 1419, 94 L. Ed. 2d 577 (1987), also fails to support the principle that a savings clause prevents a court from applying implied preemption analysis. Granite Rock involved a facial challenge to a system of state mining permits on the basis of preemption by several federal laws. One of these laws, the Coastal Zone Management Act ("CZMA"), had a broadly worded savings clause, and the Court relied on this clause to hold that the CZMA did not preempt the state law in question. Id. 107 S.Ct. at 1430. However, as Granite Rock involved a facial challenge to the state law, the Court only held that the CZMA did not "occupy the field"; the Court also stated that state law might be preempted if there was an actual conflict with the CZMA. The Court wrote that the CZMA did not preempt state law "except in cases of actual conflict," and that "the CZMA does not automatically preempt all state regulation." Id. 107 S.Ct. at 1430-31 (emphasis added).

If there is a general principle with respect to the application of implied preemption analysis to statutes with general savings clauses, it may be closer to that advocated by amicus curiae Product Liability Advisory Council: "general savings clauses may not be read literally to permit common law actions that contradict and subvert a statutory scheme." There are Supreme Court cases that support this view. In Texas & Pacific Railway Co. v. Abilene Cotton Oil Co. (Abilene), 204 U.S. 426, 27 S. Ct. 350, 51 L. Ed. 553 (1907), the question was whether a shipper's common law suit to recover an unreasonably high shipping charge was preempted because the railroad had filed its rate schedule with the Interstate Commerce Commission ("ICC"). The shipper, arguing against preemption, relied on a savings clause in the "act to regulate commerce." The clause provided

Nothing in this act contained shall in any way abridge or alter the remedies now existing at common law or by statute, but the provisions of this act are in addition to such remedies.

Id. at 446, 27 S.Ct. at 357-58. In spite of this clause, the Court proceeded to engage in implied preemption analysis:

[A]s it is conceded that the act to regulate commerce did not in so many words abrogate such [common law] right, it follows that the contention that the right was taken away by the act to regulate commerce rests upon the proposition that such result was accomplished by implication.

Id. at 436, 27 S.Ct. at 353. The Court went on to hold that the shipper's common law suit was "wholly inconsistent" with the administrative scheme of the Commerce Act, and that the shipper could not proceed with its suit.

In the recent Supreme Court decision addressing the preemptive scope of the Clean Water Act ("CWA"), the Court again used implied preemption analysis despite the presence of a broad savings clause. The issue in Ouellette was whether interstate nuisance suits (i.e., suits brought by a plaintiff in one state against a polluter in another state) were preempted by the CWA. The CWA contains two broad savings clauses, including the following provision:

Nothing in this section shall restrict any right which any person (or class of persons) may have under any statute or common law to seek enforcement of any effluent standard or limitation or to seek any other relief....

33 U.S.C. Sec. 1365(e). Despite this provision, the Court held that the CWA did "not speak directly to the issue" of preempting interstate nuisance suits, because the provision started with the words "nothing in this section" (which was the only section dealing with private causes of action) rather than the words "nothing in this law." 107 S. Ct. at 812. The Court then proceeded to apply implied preemption analysis, and found that interstate nuisance suits were preempted unless they employed the law of the source state. Otherwise, the Court explained, the common law action would disturb the balance between source states and affected states, and would interfere with the CWA's policy of having predictable standards. Id. 107 S.Ct. at 814.

Although the Court in Ouellette found some ambiguity in the CWA's savings clause, the decision shows a general reluctance by the Court to follow a savings clause if state law will actually conflict with a federal regulatory scheme. The Court explained that "we do not believe that Congress intended to undermine this carefully drawn statute through a general saving clause." Id. at 812. The Court also implied that in certain statutes the savings clause might be a boilerplate provision that Congress did not mean to apply in cases of actual conflict. The Court wrote:

The fact that the language of [the saving clause] is repeated in haec verba in the citizen suit provisions of a vast array of environmental legislation ... indicates that it does not reflect any considered judgment about what other remedies were previously available or continue to be available under any particular statute.

Id. at 812 n. 14 (quoting Milwaukee v. Illinois, 451 U.S. 304, 329 n. 22, 101 S. Ct. 1784, 1798-99 n. 22, 68 L. Ed. 2d 114 (1981)).

Abilene and Ouellette belie the supposed principle that a general savings clause precludes a court from looking into implied preemption. On the other hand, we do not adopt the opposite conclusion--that implied preemption always applies. Rather, in each case the question requires a close reading of the statute and a thorough examination of congressional intent. In this case, the ambiguity created by section 1392(d) in combination with the savings clause, combined with the paucity of design defect suits at the time the Safety Act was passed, leads us to conclude that Congress would have intended to preempt any state standard like this (even though established through the medium of a design lawsuit rather than by a state regulation) that, in the parlance of implied preemption analysis, "actually conflict[s]" with the Safety Act.

C. Performance Standards v. Design Standards

Plaintiff argues that a state product liability action can never conflict with a FMVSS because product liability actions establish design standards, while under the Safety Act, NHTSA may only promulgate performance standards. See 15 U.S.C. Sec. 1391(2) (" 'Motor vehicle safety standards' means a minimum standard for motor vehicle performance, or motor vehicle equipment performance...."). A performance standard establishes a test for a certain aspect of a vehicle's performance, without mandating how the vehicle should be designed to comply with the test. As explained by the Senate Report

The Secretary would thus be concerned with the measurable performance of a braking system, but not its design details. Such standards will be analogous to a building code which specifies the minimum load-carrying characteristics of the structural members of a building wall, but leaves the builder free to choose his own materials and designs.

Senate Report, 1966 U.S.Code Cong. & Admin.News at 2714.22 In contrast, a design standard mandates how a vehicle or item of vehicle equipment should be designed, and does not dictate how the product should perform in response to certain tests. We agree with Wood that the usual common law action for design defect, like the present one claiming that the Blazer was defectively designed because it lacked air bags, creates a design standard rather than a performance standard.

Merely stating the difference between design and performance standards, however, does not lead to the conclusion that a state design standard will never conflict with a federal performance standard. First, although design and performance standards are analytically distinct, in practice the line is not so clear. For example, a performance standard requiring that a vehicle's driver should be able to see in the dark is not very different from a design standard requiring that a vehicle have lights. Similarly, although FMVSS 208 is a performance standard dealing with the protection of occupants from the second collision, it specifies that vehicles must either have seat belts meeting certain criteria or have other safety equipment (passive restraints) that will protect occupants from the second collision.23 By requiring seat belts or passive restraints, FMVSS 208 has elements of a design standard.

Second, although the Safety Act provides that NHTSA should develop performance standards, the Safety Act's preemption provision is not limited to any one type of standard. Section 1392(d) provides that "any safety standard applicable to the same aspect of performance" as a federal standard is preempted. The above discussed hypothetical state regulation--requiring all vehicles registered in the state to be equipped with air bags--is a design standard. Yet, as we discussed above, such a state standard is unquestionably related to the same aspect of performance as FMVSS 208, and is thus expressly preempted by section 1392(d). Similarly, if a state product liability action would establish a design standard related to the "same aspect of performance," as that term is defined narrowly in Chrysler v. Rhodes, as a federal standard but not identical to the federal standard, then it is impliedly preempted by the Safety Act.

D. Preemption v. Raising FMVSS as a Defense to a Design Defect Claim

Wood claims that "General Motors' preemption argument has been specifically addressed and rejected in other Circuits." The cases cited by Wood fail to support this assertion. Not a single appellate decision addresses the question of the implied preemption of a state law cause of action that would effectively create a state standard, not identical to a federal standard, in violation of section 1392(d).

Instead, the cited cases hold that compliance with a federal safety standard does not serve as a defense on the merits of a defective design claim. See, e.g., Shipp v. General Motors Corp., 750 F.2d 418 (5th Cir.1985); Sours v. General Motors Corp., 717 F.2d 1511, 1516-17 (6th Cir.1983); Dawson v. Chrysler Corp., 630 F.2d 950, 957-58 (3d Cir.1980), cert. denied, 450 U.S. 959, 101 S. Ct. 1418, 67 L. Ed. 2d 383 (1981). In each of these cases, the defendant manufacturer argued that the vehicle could not have been defectively designed because it complied with applicable FMVSS. This argument is summarily rejected, usually in a single paragraph. See Shipp, 750 F.2d at 421 ("Of course, compliance with such minimum safety standards does not exempt or immunize a manufacturer from common law strict liability"); Sours, 717 F.2d at 1517 ("[T]he very federal safety statute upon which General Motors relies makes it abundantly clear that compliance with the regulations promulgated thereunder does not immunize a manufacturer from common law liability."); Dawson, 630 F.2d at 958 ("Compliance with the safety standards promulgated pursuant to the National Traffic Motor Vehicle Safety Act, however, does not relieve Chrysler of liability in this action."). We have no quarrel with our sister circuits' holdings that compliance with a FMVSS is not a defense on the merits to a claim of design defect.

The issue in this case--implied preemption of a state product liability suit that would stand as an obstacle to the uniformity of federal standards--is distinct from a defense on the merits. To be sure, the defendant manufacturer would be happy to win on any theory. However, the road to a successful defense based on preemption is entirely different, and is more unlikely to be found, than that to a defense on the merits. The latter is much broader--the manufacturer just claims compliance with a standard, and argues that compliance proves that the car could not have been defectively designed. To prove preemption, on the other hand, the manufacturer must show 1) that the theory of recovery advanced by the product liability suit is directed at the same "aspect of performance," as that term is narrowly defined in Chrysler Corp. v. Rhodes, 416 F.2d 319 (1st Cir.1969), as the federal safety standard, and 2) that the theory of recovery is of sufficiently general applicability that a successful tort suit would be equivalent to a state regulation.

We cannot say whether implied preemption would have been a viable defense in any of the cited cases where FMVSS were invoked as a defense on the merits. However, we do not think it coincidental that defendants first raised, and courts have first accepted, the preemption argument in the unusual context of an air bag suit. Air bags suits are peculiarly well suited to implied preemption. First, as discussed in Section IVC above, the applicable federal regulation has moved beyond a pure performance standard toward a design standard. Product liability actions based on a theory of defective design have a much greater potential to conflict with a design-type standard because such an action will allege that some specific aspect of the vehicle's design was defective. Defective design actions will not ordinarily allege that a vehicle should be manufactured to meet some hypothetical performance standard. Second, air bag suits are of the broadest general applicability--potentially affecting every vehicle on the road--and thus are most similar to a state regulation.

However, regardless of the applicability of implied preemption to the defense on the merits cases cited by plaintiff, these cases are not controlling as the issue of implied preemption was never argued or addressed.

In contrast to our reading of the conflicting preemption and savings clauses, plaintiff argues that the two provisions embody Congress's well thought out intent that the Safety Act would create a dual system of regulating automobile safety, where both state courts and a federal agency would be free to extend the standards of care of the safe design of automobiles. We find this argument utterly lacking in historic or statutory support. However, we address it fully because the policy implications of the argument have undoubted appeal, and may have influenced the district court.

Congress, plaintiff argues, believed that two institutions--both state courts and a federal agency--were required because the federal agency would be subject to "capture" by the automotive industry. "Agency capture," as explained by legal scholars, is the undesirable scenario where the regulated industry gains influence over the regulators, and the regulators end up serving the interests of the industry, rather than the general public. See generally Wiley, A Capture Theory of Antitrust Federalism, 99 Harv.L.Rev. 713, 724-26 (1986); Stewart, The Reformation of American Administrative Law, 88 Harv.L.Rev. 1667, 1684-87, 1713-15 (1975). This theory was applied to the Safety Act's savings clause in a 1967 article co-authored by Ralph Nader:

While there is much to be said for consistency and uniformity, achieving these ends by making the agency the sole and ultimate arbiter of design standards ignores the political aspect of design decisions. When the standard-setting function is centered exclusively in a single agency, that agency becomes the sole target for outside pressures. There is only one battle to be won. The considerable power of the automobile industry, brought to bear on the agency, will inevitably affect the standard-setting process. Therefore, a decentralization of the decision-making function allows the creation of countervailing pressures which can support or prod the agency. This is perhaps what Congress intended to be the thrust of section [1397(c) ].

Nader & Page, Automobile Design and the Judicial Process, 55 Calif.L.Rev. 645, 676 (1967). Several explanations are advanced for the considerable influence which industries can exert on their regulators, including the following which, arguably, could apply to NHTSA and the automotive industry.The division of responsibility between the regulated firms, which retain primary control over their own affairs, and the administrator, whose power is essentially negative and who is dependent on industry cooperation in order to achieve his objectives, places the administrator in an inherently weak position. The administrator will, nonetheless, be held responsible if the industry suffers serious economic dislocation. For both of these reasons he may pursue conservative policies.

Stewart, The Reformation of American Administrative Law, 99 Harv.L.Rev. at 1685. The district court, in fact, seemed influenced by the "agency capture" argument--at the hearing on the motion for summary judgment, the court stated

If the [Safety Act] reads the way you [General Motors] say it is, this statute, which is supposed to be a safety statute, is an incredible piece of self-interest legislation on the part of the auto industry, isn't it?

Although plaintiff's "agency capture" argument may have force in the field of public policy, it lacks persuasive force in this case. Our task is to divine the legislative intent behind the preemption and savings clauses, and neither the text nor legislative history of the Safety Act indicate that Congress was concerned with the "capture" of federal regulators by the automobile industry. As noted above, the legislative history of the Safety Act fails to even mention suits for unsafe designs, no less the plaintiff's sophisticated dual system of safety regulation based on state court defective design suits and federal motor vehicle safety standards. When Congress did mention another source of design regulation--state government--it explicitly provided that these institutions would only have a consultative role in areas covered by federal standards. We find no indication that Congress intended the dual regulatory scheme which now seems so reasonable to plaintiff.

Furthermore, if Congress intended the savings clause to create this dual system of regulation, plaintiff fails to explain why Congress forbade state agencies from setting standards nonidentical to the federal standards. Although state agencies are also subject to "agency capture," the automobile industry would certainly have more difficulty capturing 50 states than one federal agency. And as state tort law was ineffective in setting design standards prior to 1966, state agencies were the only other source of regulation when the Safety Act was passed in 1966. Thus if Congress really feared agency capture, and wanted a dual system of safety regulation, it would not have preempted state regulation.

We thus answer the question certified on interlocutory appeal that federal law does preempt the Massachusetts product liability claim in question against General Motors based on its installing seat belts rather than air bags in the motor vehicle in which plaintiff was injured.

Remanded for further proceedings not inconsistent herewith.

SELYA, Circuit Judge (dissenting).

My brethren today espouse a vision of preemption that I find needlessly wide and a conception of the common law that seemingly misperceives tort as a regulatory system. What is more, this espousal occurs in an arena fraught with political drama--automobile safety regulation--where we should be especially reluctant to insulate administrative decisions from the prophylaxis of the civil jury, thereby placing common law protections beyond the reach of the motoring public. I am not so naive as to question that Congress may, if it chooses, shield its legislative product from unwanted state-law infringement. Short of that, however, the regulatory schematas of administrative agencies should be no less subject to the full panoply of democratic institutions than are the overt acts of other governmental instrumentalities.

While jury verdicts have an admittedly anecdotal quality, they are a core ingredient of our representative democracy, signifying the considered judgment of citizens forced to confront and define the scope of law in action. Because I can discern neither a clear expression that this tort action has been preempted nor any sufficient reason to imply so drastic a result, I would answer the certified question in the negative and affirm the decision below.

* We have been asked, in effect, to police the border between two subsections of the National Traffic and Motor Vehicle Safety Act of 1966, specifically 15 U.S.C. Sec. 1392(d) (preempting state automobile safety standards) and 15 U.S.C. Sec. 1397(c) (preserving common law remedies).24 With respect, the majority does not so much police the border as displace it, claiming more territory for federal exclusivity than the statute warrants.

My colleagues and I are not entirely at odds. I agree with them that "[p]reemption is a matter of congressional intent," ante at 401, and that the statute sub judice is "facially ambiguous". Id. at 401.25 I depart from their analysis, however, when they purport to resolve that ambiguity with a sojourn through the unknown and the unknowable. I subscribe instead to the Court's teaching that "[t]he essence of our federal system is that within the realm of authority left open to them under the Constitution, the States must be equally free to engage in any activity that their citizens choose for the common meal, no matter how unorthodox or unnecessary anyone else--including the federal judiciary--deems state involvement to be." Garcia v. San Antonio Metropolitan Transit Auth., 469 U.S. 528, 546, 105 S. Ct. 1005, 1015, 83 L. Ed. 2d 1016 (1985). I am unwilling to find preemption in a case as charged with ambiguity as this one. See Maryland v. Louisiana, 451 U.S. 725, 746, 101 S. Ct. 2114, 2128-29, 68 L. Ed. 2d 576 (1981); Rice v. Santa Fe Elevator Corp., 331 U.S. 218, 230, 67 S. Ct. 1146, 1152, 91 L. Ed. 1447 (1947). Rather, I believe, as did the court below, that Massachusetts remains free to impose its common law concepts of tort liability in this area.

The point of reasoned beginning in interpreting a law is the language of the statute itself. United States v. James, 478 U.S. 597, 604, 106 S. Ct. 3116, 3121, 92 L. Ed. 2d 483 (1986). In construing statutes, our role is to give effect to them as written and as a whole, not to enforce one section at the expense of another. We must "assume that the legislative purpose is expressed by the ordinary meaning of the words used." American Tobacco Co. v. Patterson, 456 U.S. 63, 68, 102 S. Ct. 1534, 1537, 71 L. Ed. 2d 748 (1982). That canon of construction has particular pertinency here, where both the preemption and savings clauses use the word "any." The Supreme Court has recently placed special emphasis on that term, remarking that "[i]t is difficult to imagine broader language." James, 478 U.S. at 604, 106 S.Ct. at 3121 (footnote omitted). Applying its analysis to statutory language dealing with liability, the Court continued: "Congress' choice of the language 'any damage' and 'liability of any kind' further undercuts a narrow construction." Id. at 605, 106 S.Ct. at 3121. (emphasis supplied by the Court).

A savings clause exists as a specific exception to a general preemption clause: it is designed to save a specific subset of legal expressions that would otherwise be preempted.26 Thus, while we are generally obligated to give a broad construction to the word "any" as used in the preemption clause, we are specifically obligated to give great breadth to its use in the savings clause.

Even if these canons of construction do not, in and of themselves, suggest the proper resolution of the facial ambiguity in the statute--and I believe they do--they at least inform one's reading of the legislative history. The majority concludes that "Congress in 1966 did not contemplate the likelihood that there would be a state tort action that would effectively create a state design standard conflicting with a federal safety standard." Ante at 403. In arriving at this point, my confreres first examine the "historical context of the Safety Act's enactment in 1966...." Id. at 404. That approach reverses the proper analytic sequence in two ways. For one thing, the legislative history comprises the statements which Congress itself deemed most relevant to statutory understanding. Both legal and historical research techniques suggest that the most relevant materials be analyzed first.27 See, e.g., Blum v. Stenson, 465 U.S. 886, 896, 104 S. Ct. 1541, 1547-48, 79 L. Ed. 2d 891 (1984). It is only if a reviewing court needs enlightenment that "outside" historical sources must be plumbed to illuminate the language of the legislative materials.

The second deviation from the proper norm is more subtle. My brethren radically alter the inquiry's focus by defining the peek into the past in the narrowest conceivable terms (e.g., the historical question deals with "possible preemption of certain defective design actions ... up to 1966," ante at 404), unabashedly presuming Congress to have been nescient in the face of the dynamic evolution of tort law. Given the level of generality in both the statutory language and the legislative history--a level of generality which appears conceded, see ante at 406--I think that the better approach is to ask what one means by the general language in the context of legal history circa 1966. That approach grounds the interpretative query in what Congress said, rather than in the "precarious business" of attempting to draw inferences from congressional silence. Symons v. Chrysler Loan Guarantee Board, 670 F.2d 238, 242 (D.C.Cir.1981); see also Capitol Technical Services, Inc. v. FAA, 791 F.2d 964, 971 & n. 42 (D.C.Cir.1986) (same); cf. D. Fischer, Historians' Fallacies 15-24, 47-48 (1970) (discussing fallacies of fictional question, semantical question, and negative proof).

Try as I might, I can fathom no implicit exception for design defects in the general statements contained in the legislative history.

To the exact contrary, an unbiased glimpse of tort law circa 1966 indicates to me that design defect litigation, although then a relatively recent phenomenon, was not so new as to catch the Congress unawares. Let me illustrate briefly. By 1966, the judiciary had amassed a great deal of experience with design defects in respect to a host of products. When one looks to the intellectual world of tort at that time and seeks to ascertain what was meant by generic references to products liability, the sensible approach is to assume that broad language derives its meaning from the commensurately broad universe of legal practice which informs it. When my colleagues focus on the date of what they term the "seminal decision" in Larsen v. General Motors Corp., 391 F.2d 495 (8th Cir.1968), ante at 404, they imply that the liability it imposed for negligent automotive design sprang full-blown from the collective brows of the Eighth Circuit at that moment. But Larsen did not emerge ex nihilo; the Larsen court simply took what seemed the next logical step in the evolution of tort law. See, e.g., 391 F.2d at 504 (automotive industry not "being singled out for any special adverse treatment; liability for design defects predicated upon established "general negligence principles").

Fairly read, the Larsen rationale derived from cases, annotations, legislative materials, and scholarly commentary which, by and large, were available in 1966. Moreover, the decision itself took note of the heavy criticism trailing in the wake of Evans v. General Motors Corp., 359 F.2d 822, (7th Cir.), cert. denied, 385 U.S. 836, 87 S. Ct. 83, 17 L. Ed. 2d 70 (1966), a pre-Safety Act case holding that an automobile manufacturer could not be held liable for negligence in design. Larsen, 391 F.2d at 505 n. 9. Even if we accept arguendo the majority's isthmian frame of reference, there is a great deal of evidence that when the general words "common law liability" or "products liability" were used in 1966, they could easily have embraced what the Evans dissent characterized as a "broad national issue"--automobile design defects. Evans, 359 F.2d at 828 (Kiley, J., dissenting). Significantly, Judge Kiley drew attention to the extensive literature then available concerning automobile "safety design and production," id., and noted that the law was clearly pointing "to[ward] greater responsibility of manufacturers in designing ... products...." Id. at 826.

Indeed, Larsen itself addressed the relationship between NHTSA standards and common law liability with some particularity. Discussing the savings clause, 15 U.S.C. Sec. 1397(c), in the context of a design defect action, the court invoked the general language of common law liability:

It is apparent that the National Traffic Safety Act is intended to be supplementary of and in addition to the common law of negligence and product liability. The common law is not sterile or rigid and serves the best interests of society by adapting standards of conduct and responsibility that fairly meet the emerging and developing needs of our time. The common law standard of a duty to use reasonable care in light of all the circumstances can at least serve the needs of our society until the legislature imposes higher standards or the courts expand the doctrine of strict liability for tort. The Act is a salutary step in this direction and not an exemption from common law liability.

Larsen, 391 F.2d at 506. While I do not regard the Larsen court's discussion of the interplay among products liability, tort, and regulation as dispositive, such expressions, nearly contemporaneous with passage of the Safety Act, are extremely suggestive. At a distance of some twenty years, I fail to see how we can construe the ordinary language of a statute enacted in 1966 better than a court not twenty months distant.

Given that courts both before and closely after the passage of the Safety Act addressed the question of automobile design defects in general terms, I believe we ought to give the legislative history's general language the overall meaning that contemporary courts gave it. This is especially so since Congress could hardly have been blind and deaf to the transformation that tort law had undergone in the preceding decades. Congress made no specific exception for products liability for design defects, and we disserve the proper performance of our role by carving such an exception against the grain of history. At the very least, the usage of the times suggests sufficient ambiguity in legislative history and historical context to require us to resolve the amphiboly based on the canons of statutory interpretation earlier discussed. See supra at 420-421 & n. 26. While I am not an historian, I believe that the mood in 1966 was both different and more fluid than my colleagues portray when they selectively scrutinize the history of design defect actions generally28 and then proceed to conflate that history with developments in the automotive industry.

The history of tort law over the period in question has been the subject of much scholarly attention in recent years. The sum of the relevant scholarship makes me less sanguine than my brethren that Congress, in 1966, was ignorant of the impact of design defect actions. The general consensus seems clear: "Since 1960, our modern civil liability regime has experienced a conceptual revolution that is among the most dramatic ever witnessed in the Anglo-American legal system." Priest, The Invention of Enterprise Liability: A Critical History of the Intellectual Foundations of Modern Tort Law, 15 J.Leg.Stud. 461, 461 (1985); G. White, Tort Law in America: An Intellectual History 168-73, 197-208 (1980) (similar). In Professor Priest's view, product defect claims, based on the theory that an enterprise should internalize the costs of accidents resulting from failings in its wares, were fully "synthesized" and generally accepted by 1965. Priest, op. cit. supra, at 505-18. And by then, although the definition of defective design was not firmly established, the commentators had begun to urge, and some courts had begun to adopt, expansive definitions. Id. at 521-23; see also Noel, Recent Trends in Manufacturer's Negligence as to Design, Instructions or Warnings, 19 Sw.L.J. 43, 47-48 51-60 (1965); Noel, Manufacturer's Negligence of Design or Directions for Use of a Product, 71 Yale L.J. 816 (1962). The evolving scholarship itself acted as a stimulus for rapid doctrinal change in the area of products liability. See White, op. cit. supra, at 168. The Safety Act was passed in the midst of just such turbulent change. Had Congress wanted to separate the spheres of common law and federal regulation completely, it knew how to do so. It chose instead repeatedly to use the general language of tort, fashioning no such clear dichotomy as my colleagues now divine. The statutory framework, the language of the Act, the historical antecedents which pertain, and the tenor of the legal times coalesce to convince me that Congress, when enacting this law, did not intend to restrict the evolution of tort, much less to restrict it by way of preemption.

Notwithstanding that the savings clause deserves fuller effect, it can still be preempted if the cause of action which plaintiffs endeavor to assert "stands as an obstacle to the accomplishment of the full purposes and objectives of Congress." Schneidewind v. ANR Pipeline Co., --- U.S. ----, 108 S. Ct. 1145, 1151, 99 L. Ed. 2d 316 (1988); see also San Diego Building Trades Council v. Garmon, 359 U.S. 236, 247, 79 S. Ct. 773, 780-81, 3 L. Ed. 2d 775 (1959); Palmer v. Liggett Group, Inc., 825 F.2d 620, 626 (1st Cir.1987). Yet, mindful that the savings clause must, for the reasons stated in Part I hereof, preserve the widest possible range of causes of action not in conflict with the safety standard, I believe the plaintiff's complaint passes muster.

Let me start this segment of the analysis by noting my frank disagreement with the majority when it professes to see no difference between direct regulation and the indirect regulatory effect of tort law vis-a-vis this case. See ante at 410-14. While I have no wish to indulge--let alone sponsor--an arid jurisprudential debate, I think there is a meaningful distinction. Direct regulation comprises a set of commands which all must obey, and thereby overrides both economic and noneconomic considerations. Tort law regulates not by obligation, but by suasion; though it lacks the power to command behavioral modifications, it affects the economics of production and consumption, thereby acquiring considerable persuasive impact. The spectre of potential tort liability forces a firm to choose how best to respond to liability costs and, depending upon market conditions, how to allocate those costs among consumers and shareholders. Put another way, compliance with a course of action mandated by actual regulation is a given; compliance with a course of action suggested by the spectre of tort liability, or not, is a choice constrained by economic factors.

In that sense, tort operates to change the background costs of the manufacturer's design decisions. And, the resultant choice--say, to install airbags or forgo them--is not a pure one-for-one choice, for all background costs, not just tort liability, influence design decisions. So viewed, every election which a manufacturer makes among the options available under the safety standard is governed by a variegated economic mix.

Because a manufacturer's choice is in this sense never a "free" choice, the analogy which the majority draws to the colorful rhetoric of Palmer seems badly skewed.29 Palmer was a cigarette warning case where, absent preemption, defendant would have been faced with a Hobson's choice: it could avoid future liability only by adding a new warning label, one not within the prescription of 15 U.S.C. Sec. 1333. This case does not pose that sort of dilemma for General Motors. Here, defendant can avoid liability by choosing an option which NHTSA (and even Congress, see Pacific Legal Foundation v. Dep't of Transportation, 593 F.2d 1338, 1342 n. 29 (D.C.Cir.) (Congress did not exercise its option to veto gradual introduction of passive restraints, including airbags) cert. denied, 444 U.S. 830, 100 S. Ct. 57, 62 L. Ed. 2d 38 (1979)), has approved. Unlike cigarette warnings--where the labels required would have "contravene[d] the Act's policy of uniform labelings," 825 F.2d at 627--the statute here in play provides only for "minimum" safety standards. See 15 U.S.C. Sec. 1391(2). I fully agree that any state holding that covered the same aspect of performance as the safety standard in question and that would have the effect of requiring something other than a permitted statutory option would--and should--be preempted as non-identical. But that is some other case--not this one. As the majority concedes, "the standards are 'minimum' in the sense that a manufacturer may make a vehicle safer than required by federal law...." Ante at 414. The instant case is thus at a sizable remove from Palmer. Absolutely no conflict exists between federal law and state tort law when the latter allows for the imposition of damages; the economic calculus changes, to be sure--but the change does little more than counsel that the wiser course for the manufacturer may be to alter or expand its selection among the choices legitimately available to it.

Perhaps the root cause of my fellow panelists' objection to allowing a suit like this one to go forward is the concern that different juries in different locales may choose differently among the statutory options, see ante at 400, thus "requiring" a manufacturer to implement all three of the permissible safety systems. This, it is intimated, would frustrate NHTSA's apparent purpose of providing manufacturers a choice. But that objection is something of a heuristic. In the first place, NHTSA's method must be in service to the aim of the statute. In the second place, the contention that variation in verdicts would destroy uniformity holds very little water. And finally, the effect of requiring any particular option may be such as to allow a manufacturer to absorb or pass along liability costs from suits based on lack of the remaining options. Let me amplify on these three lines of response.

1. Aim of the Statute. In line with the enabling legislation itself, NHTSA's decision to allow manufacturers to exercise an option must be designed to save lives and reduce injuries. The regulation itself tells the tale:

The purpose of this standard is to reduce the number of deaths of vehicle occupants, and the severity of injuries ... by specifying equipment requirements for active and passive restraint systems.

49 C.F.R. Sec. 571.208 (1976). One searches in vain for anything in the statute or the standard intimating that the provision of choice to the manufacturer is an "objective of federal law." Silkwood v. Kerr-McGee Corp., 464 U.S. 238, 256, 104 S. Ct. 615, 625-26, 78 L. Ed. 2d 443 (1984). If no such objective can be discerned, then a fortiori, the statute cannot be frustrated by allowing state tort laws to operate. Given that the option is simply in service to the statutory and regulatory objective actually stated, that is, safety, I can perceive no irreconcilable conflict which would stem from court decisions in cases like this one. Damage suits will tend to encourage safety, not countervail it. Adjudications adverse to manufacturers would likely have the effect of implementing the safety standard in toto.

Texas & Pacific Ry Co. v. Abilene Cotton Oil Co., 204 U.S. 426, 27 S. Ct. 350, 51 L. Ed. 2d 553 (1907), is not to the contrary. Abilene arose out of a scenario wherein the Interstate Commerce Commission had been empowered to set "reasonable" freight rates and railroads were required to charge accordingly. Id. at 437-38, 27 S.Ct. at 354-55. To allow juries to find other rates "reasonable" would have resulted in the "absolute destruction of the act." Id. at 440, 27 S.Ct. at 355. But in the case at bar, unlike in Abilene, the suggested tort action is not "wholly inconsistent with the administrative power." Id. at 441, 27 S.Ct. at 355. As I have pointed out, all three options contained in FMVSS 208 are statutorily and administratively legitimate and, as the majority concedes, ante at 414, could be utilized coactively by defendant if it chose to do so.

I am also unpersuaded that International Paper v. Ouellette, 479 U.S. 481, 107 S. Ct. 805, 93 L. Ed. 2d 883 (1987), compels the majority's result. Ouellette is dependent upon the wording of the idiosyncratic savings clause there at issue and presents, at bottom, a question of which state-law actions survive in the face of federal regulation--not a question of whether any state-law actions survive at all. Not only did the Ouellette Court act to preserve certain state-law claims, but it also decreed Vermont law preempted solely because its application would "allow respondents to circumvent the NPDES permit system, thereby upsetting the balance of public and private interests so carefully addressed by the [Clean Water] Act." Id. at 494, 107 S.Ct. at 813. In this case, General Motors cannot so facilely escape the administrative scheme; all of its choices stem from acceptable options. The interest balance--public/private--remains unaffected. Under the Clean Water Act, after all, a state was given the right to exceed federal standards for pollution sources within its jurisdiction. Id. at 489-90, 107 S.Ct. at 810-11. Under the Safety Act, the states lack even that power; the sovereignty of each state, the implicit concern of the Ouellette Court, id. at 495, 107 S.Ct. at 813, remains uncompromised by validating the tort law of Massachusetts.

2. Uniformity. I find my colleagues' fear that maintenance of this action will destroy uniformity, see, e.g., ante at 406-407, 412-413, to be unfounded. They point to nothing in the voluminous record which so much as suggests that installation of any one or more of the proffered options makes installation of any of the others impossible. Unless a manufacturer wilfully attempted to engineer different modes of compliance for each jurisdiction, an act which would be tantamount to economic self-immolation, nothing about the "centralized, mass production, high volume character" of the industry....," ante at 412 (citation omitted), would be compromised. A manufacturer would never be in the position of having to implement contradictory safety features, nor could it ever be faced with one state's safety "maximum" and another's conflicting "minimum." In no meaningful sense would uniformity be threatened; in the worst-case scenario--"worst" from the manufacturer's viewpoint--self-interest would require either concurrent implementation of the FMVSS 208 options or payment of whatever damage awards (if any) might accrue. Such a result does not create an irreconcilable conflict between the federal regulatory scheme and state common law. See Wood v. General Motors Corp., 673 F. Supp. 1108, 1114-16 (D.Mass.1987) (explaining why potential for damage award under state law is not incompatible with FMVSS 208; citing relevant authority).

3. Cost Considerations. In the last analysis, the likelihood that all three options could ever be required, in the sense that tort law ever "requires" anything to be done, seems remote. The synergy inherent in, for example, adding airbags to an automobile that already has manual belts with warning systems, would likely be enormous. Under such circumstances, neglect to furnish the incremental protection of the additional (third) option might well yield such small liability losses that manufacturers could logically rate the game as eminently worth the candle. This assertion, to be sure, has a speculative quality. But that, perhaps, is the point: it is no less problematic than the chain of conjectural inferences put together by my colleagues, to the effect that a contrary decision (1) will open the door to statelaw claims, which (2) will be realized in this and countless other actions, (3) yielding damage awards so large as to force car manufacturers to alter their behavior.30

There is, perhaps, one other observation which should be made. The majority apparently agrees that "compliance with a federal safety standard does not serve as a defense on the merits of a defective design claim." Ante at 417 (emphasis in original). Yet the abundant caselaw supporting that proposition cannot be reconciled with this court's opinion today. If compliance is not a defense on the merits, then almost by definition, a successful common law action for inadequate design could force manufacturers to go beyond the safety standards in place under any imaginable federal law. The effect, therefore, would be the same as requiring a non-identical standard covering the same aspect of performance. It would be a strange "savings clause" indeed which could salvage an action on the merits in this fashion but be impuissant to stop preemption, when the effect on uniformity of regulation and manufacture would be precisely the same in either case. Cf. Dawson v. Chrysler Corp., 630 F.2d 950, 953 (3d Cir.1980) (holding that compliance with NHTSA safety standards does not relieve an automobile manufacturer of liability and noting "the troubling public policy dilemma ... that under existing federal law individual juries in the various states are permitted, in effect, to establish national automobile safety standards"), cert. denied, 450 U.S. 959, 101 S. Ct. 1418, 67 L. Ed. 2d 383 (1981).

I need go no further. I recognize the closeness of the question and the imprecisions of the statutory scheme. I understand full well that the balance of policy considerations may have altered since the Safety Act of 1966 was passed. And it can, I suppose, be hypothesized that automobile safety is not today the pressing concern it was in 1966 (although that is not a theorem to which I can willingly subscribe). But the crux of the matter is not whether the majority's vision may best serve an era in which the regulated are free both to choose safety options which exempt them from the imposed wisdom of juries under the common law and to safeguard their profit margins wherever Congress's speech is less than explicit. What counts, I suggest, is that Congress did not originally, and has yet to, superimpose that entrepreneurial vision on the Safety Act. Nor do I understand the safety standard promulgated by NHTSA to have done so. Because of those parallel truths, I would affirm the district court and allow this suit to proceed. The plaintiff's cause of action in tort, held by the trial judge to state a viable claim under the common law of Massachusetts, 673 F. Supp. at 1120, is not preempted.31

I respectfully dissent.