These consolidated petitions of International Harvester and the three major auto companies, Ford, General Motors and Chrysler, seek review1 of a decision by the Administrator of the Environmental Protection Agency denying petitioners' applications, filed pursuant to Section 202 of the Clean Air Act,2 for one-year suspensions of the 1975 emission standards prescribed under the statute for light duty vehicles in the absence of suspension.

The tension of forces presented by the controversy over automobile emission standards may be focused by two central observations:

(1) The automobile is an essential pillar of the American economy. Some 28 per cent of the nonfarm workforce draws its livelihood from the automobile industry and its products.3

(2) The automobile has had a devastating impact on the American environment. As of 1970, authoritative voices stated that "[a]utomotive pollution constitutes in excess of 60% of our national air pollution problem" and more than 80 per cent of the air pollutants in concentrated urban areas.4

Congressional concern over the problem of automotive emissions dates back to the 1950's,5 but it was not until the passage of the Clean Air Act in 1965 that Congress established the principle of Federal standards for automobile emissions. Under the 1965 act and its successor, the Air Quality Act of 1967, the Department of Health, Education and Welfare was authorized to promulgate emission limitations commensurate with existing technological feasibility.6

The development of emission control technology proceeded haltingly. The Secretary of HEW testified in 1967 that "the state of the art has tended to meander along until some sort of regulation took it by the hand and gave it a good pull. . . . There has been a long period of waiting for it, and it hasn't worked very well."7

The legislative background must also take into account the fact that in 1969 the Department of Justice brought suit against the four largest automobile manufacturers on grounds that they had conspired to delay the development of emission control devices.8

On December 31, 1970, Congress grasped the nettle and amended the Clean Air Act to set a statutory standard for required reductions in levels of hydrocarbons (HC) and carbon monoxide. (CO) which must be achieved for 1975 models of light duty vehicles. Section 202(b) of the Act added by the Clean Air Amendments of 1970, provides that, beginning with the 1975 model year, exhaust emission of hydrocarbons and carbon monoxide from "light duty vehicles" must be reduced at least 90 per cent from the permissible emission levels in the 1970 model year.9 In accordance with the Congressional directives, the Administrator on June 23, 1971, promulgated regulations limiting HC and CO emissions from 1975 model light duty vehicles to .41 and 3.4 grams per vehicle mile respectively. 36 Fed.Reg. 12,657 (1971).10 At the same time, as required by section 202(b)(2) of the Act, he prescribed the test procedures by which compliance with these standards is measured.11

Congress was aware that these 1975 standards were "drastic medicine,"12 designed to "force the state of the art."13 There was, naturally, concern whether the manufacturers would be able to achieve this goal. Therefore, Congress provided, in Senator Baker's phrase, a "realistic escape hatch": the manufacturers could petition the Administrator of the EPA for a one-year suspension of the 1975 requirements, and Congress took the precaution of directing the National Academy of Sciences to undertake an ongoing study of the feasibility of compliance with the emission standards. The "escape hatch" provision addressed itself to the possibility that the NAS study or other evidence might indicate that the standards would be unachievable despite all good faith efforts at compliance. This provision was limited to a one-year suspension, which would defer compliance with the 90% reduction requirement until 1976. Under section 202(b)(5)(D) of the Act, 42 U. S.C. Sec. 1857f-1(b)(5)(D), the Administrator is authorized to grant a one-year suspension

only if he determines that (i) such suspension is essential to the public interest or the public health and welfare of the United States, (ii) all good faith efforts have been made to meet the standards established by this subsection, (iii) the applicant has established that effective control technology, processes, operating methods, or other alternatives are not available or have not been available for a sufficient period of time to achieve compliance prior to the effective date of such standards, and (iv) the study and investigation of the National Academy of Sciences conducted pursuant to subsection (c) of this section and other information available to him has not indicated that technology, processes, or other alternatives are available to meet such standards.

The statute provides that an application for suspension may be filed any time after January 1, 1972, and that the Administrator must issue a decision thereon within 60 days. On March 13, 1972, Volvo, Inc., filed an application for suspension and thereby triggered the running of the 60 day period for a decision. 37 Fed.Reg. 5766 (March 21, 1972.)14 Additional suspension requests were filed by International Harvester on March 31, 1972, and by Ford Motor Company, Chrysler Corporation, and General Motors Corporation on April 5, 1972. Public hearings were held from April 10-27, 1972. Representatives of most of the major vehicle manufacturers (in addition to the applicants), a number of suppliers of emission control devices and materials, and spokesmen from various public bodies and groups, testified at the hearings and submitted written data for the public record. The decision to deny suspension to all applicants was issued on May 12, 1972.

The Decision began with the statement of the grounds for denial: ". . . I am unable, on the basis of the information submitted by the applicants or otherwise available to me, to make the determinations required, by section 202 (b)(5)(D)(i), (iii), or (iv) of the Act."15 The EPA Decision specifically focused on requirement (iii) that:

the applicant has established that effective control technology, processes, operating methods, or other alternatives are not available or have not been available for a sufficient period of time to achieve compliance prior to the effective date of such standards . . .,

A Technical Appendix, containing the analysis and methodology used by the Administrator in arriving at his decision, was subsequently issued on July 27, 1972.

The data available from the concerned parties related to 384 test vehicles run by the five applicants and the eight other vehicle manufacturers subpoenaed by the Administrator. In addition, 116 test vehicles were run by catalyst and reactor manufacturers subpoenaed by the Administrator. These 500 vehicles were used to test five principal types of control systems: noble metal monolithic catalysts, base metal pellet catalysts, noble metal pellet catalysts, reactor systems, and various reactor/catalyst combinations.

At the outset of his Decision, the Administrator determined that the most effective system so far developed was the noble metal oxidizing catalyst.16 Additionally, he stated that the "most effective systems typically include: improved carburetion; a fast-release choke; a device for promoting fuel vaporization during warm-up; more consistent and durable ignition systems; exhaust gas recirculation; and a system for injecting air into the engine exhaust manifold to cause further combustion of unburned gases and to create an oxidizing atmosphere for the catalyst."17 It was this system to which the data base was initially narrowed: only cars using this kind of system were to be considered in making the "available technology" determination.

The problem the Administrator faced in making a determination that technology was available, on the basis of these data, was that actual tests showed only one car with actual emissions which conformed to the standard prescribing a maximum of .41 grams, per mile, of HC and 3.4 grams per mile of CO.18 No car had actually been driven 50,000 miles, the statutory "useful life" of a vehicle and the time period for which conformity to the emission standards is required.19 In the view of the EPA Administrator, however, the reasons for the high test readings were uncertain or ambivalent.

Instead, certain data of the auto companies were used as a starting point for making a prediction, but remolded into a more useable form for this purpose. As the Administrator put it:20

Much of the data reports emissions measured by test procedures different from the 1975 Federal test procedure and requires conversion to the 1975 procedure by calculations which cannot be regarded as precise. Emission data was frequently submitted without an adequate description of the vehicle being tested, the emission control systems employed, or the purpose of the test. The fuel and oil used in tests were not always specified. Adjustments made to components of the engine or emission control system were frequently made and seldom fully explained. In most cases, tests were not repeated, even where results departed significantly from established trends, and little or no information was submitted to explain the diagnosis of failure, where test results showed poor results. Most important, only a few test cars were driven to 20,000 miles or more, and no vehicle employing all components of any applicant's proposed 1975 control systems has yet been driven to 50,000 miles. In the face of these difficulties, analysis and interpretation of the data required assumptions and analytical approaches which will necessarily be controversial to some degree. (emphasis added)

In light of these difficulties, the Administrator "adjusted" the data of the auto companies by use of several critical assumptions.

First, he made an adjustment to reflect the assumption that fuel used in 1975 model year cars would either contain an average of .03 grams per gallon or .05 grams per gallon of lead.21 This usually resulted in an increase of emissions predicted, since many companies had tested their vehicles on lead-free gasoline.

Second, the Administrator found that the attempt of some companies to reduce emissions of nitrogen oxides below the 1975 Federal standard of 3.0 grams per vehicle mile22 resulted in increased emissions of hydrocarbons and carbon monoxide. This adjustment resulted in a downward adjustment of observed HC and CO data, by a specified factor.23

Third, the Administrator took into account the effect the "durability" of the preferred systems would have on the emission control obtainable. This required that observed readings at one point of usage be increased by a deterioration factor (DF) to project emissions at a later moment of use. The critical methodological choice was to make this adjustment from a base of emissions observed at 4000 miles. Thus, even if a car had actually been tested over 4000 miles, predicted emissions at 50,000 miles would be determined by multiplying 4000 mile emissions by the DF factor.24

Fourth, the Administrator adjusted for "prototype-to-production slippage." This was an upward adjustment made necessary by the possibility that prototype cars might have features which reduced HC and CO emissions, but were not capable of being used in actual production vehicles.25

Finally, in accord with a regulation assumed, as to substance, in the text of the Decision, but proposed after the suspension hearing,26 a downward adjustment in the data readings was made on the basis of the manufacturers' ability, in conformance with certification procedures, to replace the catalytic converter "once during 50,000 miles of vehicle operation," a change they had not used in their testing.27

With the data submitted and the above assumptions, the Administrator concluded that no showing had been made that requisite technology was not available. The EPA noted that this did not mean that the variety of vehicles produced in 1975 would be as extensive as before. According to EPA, "Congress clearly intended to require major changes in the kinds of automobiles produced for sale in the United States after 1974" and there "is no basis, therefore, for construing the Act to authorizing suspension of the standards simply because the range of performance of cars with effective emission control may be restricted as compared to present cars." As long as "basic demand" for new light duty motor vehicles was satisfied, the applicants could not establish that technology was not available.28

For purposes of judicial review, the initial EPA decision rests on the technology determination. The Administrator did state:29

On the record before me, I do not believe that it is in the public interest to grant these applications, where compliance with 1975 standards by application of present technology can probably be achieved, and where ample additional time is available to manufacturers to apply existing technology to 1975 vehicles. (Emphasis added.)

The statute apparently contemplates the possibility of an EPA denial of suspension for failure to meet criterion (i) of Sec. 202(b)(5)(D) ("essential to the public interest") even though criterion (iii) has been satisfied ("applicant has established that effective control technology . . [is] not available").30 It suffices here to say that the EPA's 1972 "public interest" finding was obviously only a restatement of, and dependent on the validity of, the conclusion of a failure to satisfy standard (iii) by showing that effective control technology is not available.

The Administrator also offered some "comments" on issues pertinent to the required "good faith" determination under standard (ii), as guidance to applications who might seek a one year suspension next year of the 1976 oxides of nitrogen standard. But he explictly disclaimed reaching that question in this proceeding. The thrust of his comment was to call into question the rigid "arms length" relationship structure which vehicle manufacturers imposed on their suppliers, as a source of a halter on progress in developing the required technology.31

After oral argument to this court on December 18, 1972, in a per curiam order issued December 19, 1972, we remanded the record to the Administrator, directing him to supplement his May 12, 1972 decision by setting forth:

(a) the consideration given by the Administrator to the January 1, 1972 Semiannual Report on Technological Feasibility of the National Academy of Sciences; and (b) the basis for his disagreement, if any, with the findings and conclusion in that study concerning the availability of effective technology to achieve compliance with the 1975 model year standards set forth in the Act.

Our remand order was not intended to indicate that we had concluded that an EPA conclusion was required as to clause (iv)-concerning the evaluation based on the NAS study and other information (from sources other than applicants)-when the Administrator had determined under (iii) that the auto companies had not shown technology was not available. We were nevertheless troubled by arguments advanced by petitioners that the methodology used by the Administrator in reaching his conclusion, and indeed the conclusion itself, was inconsistent with that of the Academy. It was our view that if and to the extent such differences existed they should be explained by EPA, in order to aid us in determining whether the Administrator's conclusion under (iii) rested on a reasoned basis.

D. Supplement to the Decision of the Administrator

Our remand of the record resulted in a "Supplement to Decision of the Administrator" issued December 30, 1972. The Administrator in his Supplement stated that "In general I consider the factual findings and technical conclusions set forth in the NAS report and in the subsequent Interim Standards Report dated April 26, 1972 . . to be consistent with my decision of May 12, 1972."32

The Report made by the NAS, pursuant to its obligation under 202(b)(5) (D) of the Clean Air Act, had concluded: "The Committee finds that the technology necessary to meet the requirements of the Clean Air Act Amendments for 1975 model year light-duty motor vehicles is not available at this time."33

The Administrator apparently relied, however, on the NAS Report to bolster his conclusion that the applicants had not established that technology was unavailable. The same NAS Report had stated:34

. . . the status of development and rate of progress made it possible that the larger manufacturers will be able to produce vehicles that will qualify, provided that provisions are made for catalyst replacement and other maintenance, for averaging emissions of production vehicles, and for the general availability of fuel containing suitably low levels of catalyst poisons.

The Administrator pointed out that two of NAS's provisos-catalytic converter replacement and low lead levels-had been accounted for in his analysis of the auto company data, and provision therefor had been insured through regulation.35 As to the third, "averaging emissions of production vehicles,"36 the Administrator offered two reasons for declining to make a judgment about this matter: (1) The significance of averaging related to possible assembly-line tests, as distinct from certification test procedure, and such tests had not yet been worked out. (2) If there were an appropriate assembly-line test it would be expected that each car's emissions could be in conformity, without a need for averaging, since the assembly line vehicles "equipped with fresh catalysts can be expected to have substantially lower emissions at zero miles than at 4000 miles."37

The Administrator also claimed that he had employed the same methodology as the NAS used in its Interim Standards Report, evidently referring to the use of 4000 mile emissions as a base point, and correction for a deterioration factor and a prototype-production slippage factor.38 The identity of methodology was also indicated, in his view, by the fact the EPA and NAS both agreed on the component parts of the most effective emission control system.

The Administrator did refer to the "severe driveability problems" underscored by the NAS Report, which in the judgment of NAS "could have significant safety implications,"39 stating that he had not been presented with any evidence of "specific safety hazard" nor knew of any presented to the NAS. He did not address himself to the issue of performance problems falling short of specific safety hazards.

II. REJECTION OF MANUFACTURERS' GENERAL CONTENTIONS

We begin with consideration, and rejection, of the broad objections leveled by petitioners against EPA's over-all approach.

We cannot accept petitioners' arguments that the Administrator's determination whether technology was "available," within the meaning of section 202(b)(5)(D) of the Act, must be based solely on technology in being as of the time of the application, and that the requirement that this be "available" precludes any consideration by the Administrator of what he determines to be the "probable" or likely sequence of the technology already experienced. Congress recognized that approximately two years' time was required before the start of production for a given model year, for the preparation of tooling and manufacturing processes.40 But Congress did not decide-and there is no reason for us to do so-that all development had to be completed before the toolingup period began. The manufacturers' engineers have admitted that technological improvements can continue during the two years prior to production.41 Thus there was a sound basis for the Administrator's conclusion that the manufacturers could "improve, test, and apply" technology during the lead time period.42

The petitioners' references to the legislative history are unconvincing. None of the statements quoted in their briefs specifically states that "available" as used in the statute means "available in 1972." There is even comment that points to a contrary interpretation.43 In any event, we think the legislative history is consistent with the EPA's basic approach and evidences no ascertainable legislative intent to the contrary.

While we reject the contention as broadly stated, principally by General Motors, we hasten to add that the Administrator's latitude for projection is subject to the restraints of reasonableness, and does not open the door to " 'crystal ball' inquiry."44 The Administrator's latitude for projection is unquestionably limited by relevant considerations of lead time needed for production.45 Implicit also is a requirement of reason in the reliability of the EPA projection. In the present case, the Administrator's prediction of available technology was based on known elements of existing catalytic converter systems. This was a permissible approach subject, of course, to the requirement that any technological developments or refinements of existing systems, used as part of the EPA methodology, would have to rest on a reasoned basis.

Chrysler has advanced a due process claim based upon two principal features of the proceeding, the inability to engage in cross-examination and the inability to present arguments against the methodology used in the Technical Appendix of the Administrator, which served as a basis for his decision.

The suspension provision of Section 202(b)(5)(D) does not require a trial type hearing. It provides:

Within 60 days after receipt of the application for any such suspension, and after public hearing, the Administrator shall issue a decision granting or refusing such suspension.

First, this provision for a "public hearing" contrasts significantly with other provisions that specifically require an adjudicatory hearing.46 More importantly, the nonadjudicatory nature of the "public hearing" contemplated is underscored by the 60 day limit for a decision to be made. The procedure contemplated by Congress in its 1970 legislation must be appraised in light of its concern with "avoidance of previous cumbersome and time-consuming procedures," see Kennecott Copper Corp. v. EPA, 149 U.S.App.D.C. 231, 234, 462 F.2d 846, 849 (1972).

As to legislative history of this provision, the starting point is the provision in Senate Bill 4358:47

Upon receipt of such application, the Secretary shall promptly hold a public hearing to enable such manufacturer or manufacturers to present information relevant to the implementation of such standard. The Secretary, in his discretion, may permit any interested person to intervene to present information relevant to the implementation of such standard.

This was dropped in conference, along with a provision permitting six months for a suspension decision. The resulting legislation both expedited the decision-making, and contemplated EPA solicitation of a wide range of views, from sources other than the auto companies, though the companies' applications and presentation would surely be the focus of consideration. Underlying this approach of both shortening time for decision and enlarging input lies, we think, an assumption of an informative but efficient procedure without mandate for oral cross examination.

In context, the "public hearing" provision amounts to an assurance by Congress that the issues would not be disposed of merely on written comments, the minimum protection assured by the Administrative Procedure Act for rulemaking, but would also comprehend oral submissions of a legislative nature. These are required even for rule-making when "controversial regulations governing competitive practices" are involved. American Airlines, Inc. v. CAB, 123 U.S. App.D.C. 310, 317, 359 F.2d 624, 631 (en banc 1966), cert. denied, 385 U.S. 843, 87 S. Ct. 73, 17 L. Ed. 2d 75 (1966); Walter Holm & Co. v. Hardin, 145 U.S.App. D.C. 347, 449 F.2d 1009 (1971). Even assuming oral submission, in a situation where "general policy" is the focal question, a legislative-type hearing is appropriate.48

A complication is presented by the case before us in that the general policy questions became interfused with relatively specific technical issues. Yet within the context of a quasi-legislative hearing and the time constraints of the statute, we do not think the absence of a general right of cross-examination on the part of the companies was a departure from "basic considerations of fairness." Walter Holm & Co. v. Hardin, supra, 145 U.S.App.D.C. at 354, 449 F.2d at 1016. Hearings ran for two weeks and a wide range of participants was included within the proceeding: manufacturers, vendors of the control devices and public interest groups. The auto companies were allowed to submit written questions to the Hearing Panel to be asked to various witnesses. Opportunity to prepare written questions is not as satisfactory to counsel as the opportunity to proceed on oral cross-examination, with questions that develop from previous answers. But examination on interrogatories has long been used in the law when necessary, albeit second best. And interrogatories to a live witness-often arranged in private lawsuits by use of a commission-avoid the peril of "canned" affidavits and counsel-assisted, or even counsel-drafted, responses to interrogatories. Their availability was a reasonable attempt by EPA to elicit the facts and at the same time cope with the time constraints. We do not think more was required. There was a meaningful opportunity to be heard. The specific nature of a "hearing" varies with circumstances. Cafeteria & Restaurant Workers Union v. McElroy, 367 U.S. 886, 895, 81 S. Ct. 1743, 6 L. Ed. 2d 1230 (1961), cited with approval in Goldberg v. Kelly, 397 U.S. 254, 263, 90 S. Ct. 1011, 25 L. Ed. 2d 287 (1970). Whether particular attributes of forensic presentation are not only salutary but also mandatory must also depend on circumstances. The heft of the hearing problem, including the time constraints on decisions, convinces us that the assertion of a broad right of cross-examination cannot be successfully maintained.

We distinguish between the assertion of a broad right of cross-examination, such as that argued to this court, and a claim of a need for cross-examination of live witnesses on a subject of critical importance which could not be adequately ventilated under the general procedures. This is the kind of distinction that this court made in its en banc opinion in American Airlines v. CAB, supra, 123 U.S.App.D.C. at 318-319, 359 F.2d at 632-633. We see no principled manner in which firm time limits can be scheduled for cross-examination consistent with its unique potential as an "engine of truth"-the capacity given a diligent and resourceful counsel to expose subdued premises, to pursue evasive witnesses, to "explore" the whole witness, often traveling unexpected avenues.

Given the variances in counsel, the reality that seasoning and experience are required even for trial judges who seek to avoid repetitive and undue crossexamination, the enhancement of difficulties encountered with the breadth of issues involved in a "public interest" proceeding, the fairly-anticipated problem of provision for redirect (and recross) and the interplay of different cross-examinations, there is not insignificant potential for havoc. What is most significant is that these complications are likely to be disproportionate to the values achieved, in a proceeding focusing on technical matters where other techniques generally are sufficient to adduce the pertinent information as to both what is known and unknown.

In context, we consider that the technique, adopted by EPA, of pre-screening written questions submitted in advance is reasonable and comports with basic fairness as the general procedure. This approach permits screening by the hearing officer so as to avoid irrelevance and repetition, permits a reasonable estimate of the time required for the questioning, and aids scheduling and allocation of available time among various participants and interests.49 The record reveals that the hearing officers did not propound the pre-submitted questions like robots; they were charged with conducting a hearing for the purpose of focusing information needed for decision, and they quite appropriately "followed up" on questions.

We revert to our observation that a right of cross-examination, consistent with time limitations, might well extend to particular cases of need, on critical points where the general procedure proved inadequate to probe "soft" and sensitive subjects and witnesses. No such circumscribed and justified requests were made in this proceeding.

A more serious problem, at least from the point of an informed decision-making process, is posed by the inability of petitioners to challenge the methodology of EPA at the hearing. In other contexts, it is commonplace for administrative proceedings to focus in detail on agency methodology,50 and such elucidation is salutary, of particular aid to a reviewing court. Again, however, we cannot ignore the problem of time. In part, EPA developed its methodology on the basis of submissions made by the companies at the hearings, as to the parameters of its various data. The requirement of submission of a proposed rule for comment does not automatically generate a new opportunity for comment merely because the rule promulgated by the agency differs from the rule it proposed, partly at least in response to submissions.51 Given the circumstances, we cannot hold the absence of the right to comment on the methodology a violation of the statute or due process, though such opportunity would certainly have been salutary.

While the statute makes no express provision therefor, we assume that Congress contemplated a flexibility in the administrative process permitting the manufacturers to present to EPA any comments as to its methodology, in a petition for reconsideration or modification. However, this opportunity does not permit invocation of the doctrine of failure to exhaust administrative remedies as a bar to these appeals, for those petitions could not have affected or deferred the finality of the EPA decision or the time for seeking judicial review. The opportunity is noted to obviate any possibility that the law, or our comments, may be misunderstood to require a rigid procedure of prompt and unshakeable decision-making. Our own December remand requesting clarification of the Decision illustrates that while this statute imposes some unusual time restraints it does not jettison the flexibility and capacity of reexamination that is rooted in the administrative process. American Airlines v. CAB, supra, 123 U.S. App.D.C. at 319; 359 F.2d at 633.

As matters have shaped up, the central technical issue on this appeal concerns the reliability of EPA's methodology. While we do not say that the failure to provide reasonable opportunity to comment on EPA methodology invalidates the EPA Decision for lack of procedural due process, or similar contention, we must in all candor accompany that ruling with the comment that the lack of such opportunity has had serious implications for the court given the role of judicial review.

We shall subsequently develop the legal questions, primarily questions of EPA's burden of proof, that arise with respect to EPA methodology. We preface these with admission of our doubts and diffidence. We are beset with contentions of petitioners that bear indicia of substantiality. Yet we have no EPA comment on the specific questions raised, apart from some discussion by counsel which is not an adequate or appropriate substitute.52 Our December 1972 remand opened the door to a candid discussion of these matters, but EPA fashioned a carefully limited response.

The EPA might have indicated that it desired to take a fresh look at its methodology on the basis of petitioners' criticisms, in which case, on an adaptation of the Smith v. Pollin,53 procedure, this court might have remanded the case to the agency. This remand would come during the course of our judicial review and would not conflict with the 60-day statutory time limit for the hearing and decision on the applications for suspension.

Indeed, the fact that the Administrator issued the Technical Appendix almost three months after his Decision, at a time when judicial review had already begun to run its course, indicates that the agency did not believe that agency consideration was frozen from the moment that the suspension decision was rendered, a view we approve. The EPA had latitude to continue further consideration even without requesting a court remand (under Smith v. Pollin) that would suspend judicial consideration.

This case ultimately involves difficult issues of statutory interpretation, as to the showing required for applicants to sustain their burden that technology is not available. It also taxes our ability to understand and evaluate technical issues upon which that showing, however it is to be defined, must rest. At the same time, however, larger questions are at stake. As Senator Baker put it, "This may be the biggest industrial judgment that has been made in the United States in this century." 116 Cong.Rec. 33,085 (1970). This task of reviewing the suspension decision was not assigned to us lightly. It was the judgment of Congress that this court, isolated as it is from political pressures, and able to partake of calm and judicious reflection would be a more suitable forum for review than even the Congress.54

Two principal considerations compete for our attention. On the one hand, if suspension is not granted, and the prediction of the EPA Administrator that effective technology will be available is proven incorrect, grave economic consequences could ensue. This is the problem Senator Griffin described as the "dangerous game of economic roulette." 116 Cong.Rec. 33,081 (1970). On the other hand, if suspension is granted, and it later be shown that the Administrator's prediction of feasibility was achievable in 1975 there may be irretrievable ecological costs. It is to this second possibility we first turn.

The most authoritative estimate in the record of the ecological costs of a oneyear suspension is that of the NAS Report. Taking into account such "factors as the vehicle-age distribution among all automobiles, the decrease in vehicle miles driven per year, per car as vehicle age increases, the predicted nationwide growth in vehicle miles driven each year" and the effect of emission standards on exhaust control, NAS concluded that:55

. . . the effect on total emissions of a one-year suspension with no additional interim standards appears to be small. The effect is not more significant because the emission reduction now required of model year 1974 vehicles, as compared with uncontrolled vehicles (80 percent for HC and 69 percent for CO), is already so substantial.

Other considerations may diminish the costs even further. There seems to be agreement that there are performance costs for automobiles in employing pollution control devices, even if the effects on performance cannot fairly be characterized as constituting safety hazards. The NAS Report summarized the problem, as follows:56

Three areas of vehicle performance are likely to be adversely affected by the 1975 emission control systems. These are fuel economy, vehicle-acceleration capability, and vehicle driveability (or ability to perform adequately in all normal operating modes and ambient conditions).

The question in this context is not whether these are costs the consumer should rightly bear if ecological damage is to be minimized, but rather the general effect on consumer purchasing of 1975 model year cars in anticipation of lower performance. A drop-off in purchase of 1975 cars will result in a prolonged usage of older cars with less efficient pollution control devices. If the adverse performance effect deterred purchasing significantly enough, resulting in greater retention of "older" cars in the "mix" of cars in use, it might even come to pass that total actual emissions (of all cars in use) would be greater under the 1975 than the 1974 standards.

Many of the anticiapted performance problems are traceable to the systems introduced to conform cars to control of nitrogen oxides to achieve prescribed 1975 standards, by use of exhaust-gas recycle (EGR). Such systems affect vehicle-acceleration capability because the power output for a given engine displacement, engine speed, and throttle setting is reduced.57 The NAS Report indicates that such systems could result in direct fuel-economy penalties of up to 12 percent compared with 1973 prototype vehicles.58

The NAS Report states that the effects of emission controls on vehicle driveability are difficult to quantify, but nevertheless makes the following qualitative evaluation:59

Driveability after a cold-engine start, and especially with cold ambient conditions, is likely to be impaired. To reduce HC and CO emissions during engine warmup, the choke is set to release quickly, and the fuel-air mixture is leaned out as early as possible after engine startup. Under these conditions, problems of engine stall, and vehicle stumble and hesitation on rapid acceleration, have been prevalent.

The willingness of the consumer to buy 1975 model year cars may also be affected, to some degree, by the anticipated significant costs of pollution control devices. The problem is further bedeviled by the possibility that consumers albeit rightly assigned the cost burden of pollution devices, may seek to avoid that burden, however modest,60 and to exercise, at least in some measure, an option to use older cars. Again, this would have the thrust of increasing actual total emissions of cars in use.

We may also note that it is the belief of many experts-both in and out of the automobile industry-that air pollution cannot be effectively checked until the industry finds a substitute for the conventional automotive power plant-the reciprocating internal combustion (i. e., "piston") engine.61 According to this view, the conventional unit is a "dirty" engine. While emissions from such a motor can be "cleaned" by various thermal and catalytic converter devices, these devices do nothing to decrease the production of emissions in the engine's combustion chambers. The automobile industry has a multi-billion-dollar investment in the conventional engine, and it has been reluctant to introduce new power plants or undertake major modifications of the conventional one.62 Thus the bulk of the industry's work on emission control has focussed narrowly on converter devices. It is clear from the legislative history that Congress expected the Clean Air Amendments to force the industry to broaden the scope of its research-to study new types of engines and new control systems.63 Perhaps even a one-year suspension does not give the industry sufficient time to develop a new approach to emission control and still meet the absolute deadline of 1976. If so, there will be ample time for the EPA and Congress, between now and 1976 to reflect on changing the statutory approach. This kind of cooperation, a unique three-way partnership between the legislature, executive and judiciary, was contemplated by the Congress64 and is apparent in the provisions of the Act.65

The NAS estimated that there would be a small environmental cost to suspension of 1975 standards even if 1974 standards were retained, but further recommended intermediate standards that would dilute even such modest environmental cost.66 The following table shows the various standards, and one put forward by Ford for 1975:

Our concern that the 1975 standards may possibly be counter-productive, due to decreased driveability and increased cost, is not to be extrapolated into a caution against any improvement, and concomitant reduction in permitted emissions. In such matters, as the NAS recommendation for interim standards implicitly suggests, a difference in degree may be critical, and the insistence on absolute 1975 standards, without suspension or intermediate level, may stretch for the increment that is essentially counter-productive.

We also observe that Ford Motor Company is on record as to capability of greater emission controls, i. e., lower level of emissions, than those permitted for 1974 model year cars,67 and Ford proposed that, given certain regulatory assumptions,68 the Administrator adopt an interim standard of 1.6 gm/mi HC and 19.0 gm/mi CO levels, about one half those permitted for the 1974 model year cars.

On balance the record indicates the environmental costs of a one-year suspension are likely to be relatively modest. This must be balanced against the potential economic costs-and ecological costs-if the Administrator's prediction on the availability of effective technology is incorrect.

Theoretical possibility of industry shutdown

If in 1974, when model year 1975 cars start to come off the production line, the automobiles of Ford, General Motors and Chrysler cannot meet the 1975 standards and do not qualify for certification, the Administrator of EPA has the theoretical authority, under the Clean Air Act, to shut down the auto industry, as was clearly recognized in Congressional debate.69 We cannot put blinders on the facts before us so as to omit awareness of the reality that this authority would undoubtedly never be exercised, in light of the fact that approximately 1 out of every 7 jobs in this country is dependent on the production of the automobile.70 Senator Muskie, the principal sponsor of the bill, stated quite clearly in the debate on the Act that he envisioned the Congress acting if an auto industry shutdown were in sight.71

The economic consequence of an approach geared to stringency, relying on relaxation as a safety valve

A more likely forecast, and one which enlightens what influenced the EPA decision to deny the suspension, was articulated by George Allen, Deputy Assistant Administrator for General Enforcement and a member of EPA's Hearing Panel:72

The problem really comes down to this: A decision has to be made next month, early next month. If the decision is to suspend the standards and adopt an interim standard . . . and in 1975 it turns out that technology exists to meet the statutory standard, today's decision turns out to be wrong.

******

* * *

If, on the other hand, a decision is made today that the standards cannot lawfully be suspended, and we go down to 1975 and nobody can meet the standard, today's decision was wrong.

In [the first] case, there is not much to do about the wrong decision; it was made, many people relied on it; it turns out the standard could have been met, but I doubt if we could change it.

In the second case, if a wrong decision is made, there is probably a remedy, a re-application and a recognition by the agency that it is not technically feasible to meet the standards. You can correct the one; you probably can't correct the other.

Grave problems are presented by the assumption that if technical feasibility proves to be a "wrong decision" it can be remedied by a relaxation.

Certain techniques available to the Administrator, through changes in the certification procedure, can be used in an even handed manner for all three auto companies to facilitate compliance with the 1975 standards. Already lower lead levels in fuel available for 1975 model year cars have been prescribed to increase the efficiency of the catalytic converter. Similarly certain changes in the regulatory system, through allowable maintenance and permitted change in the catalytic converter, have been made by EPA. These techniques work with reasonable impartiality as to the various auto companies.

However, a relaxation of standards, and promulgation of an interim standard, at a later hour-after the base hour for "lead time" has been passed, and the production sequence set in motion-forebodes quite different consequences. The record before us suggests that there already exists a technological gap between Ford and General Motors,73 in Ford's favor. General Motors did not make the decision to concentrate on what EPA found to be the most effective system at the time of its decision-the noble metal monolithic catalyst. Instead it relied principally on testing the base metal catalyst as its first choice system.74 In predicting that General Motors could meet the 1975 standards, EPA employed a unique methodological approach. Instead of taking emissions at 4000 miles of cars with preferred systems-with which none of the General Motors cars was equipped-and applying against this, adjustments for lead levels and deterioration, as had been done in the case of Ford and Chrysler, EPA took emissions at 4000 miles of GM cars which had no converters of any kind, and predicted how they would function with an Engelhard monolithic catalytic converter, based on auto manufacturers' use of this device in a number of cars-principally Ford's-when testing it for durability.75 In his Supplemental Decision the Administrator recognized that this was a departure from NAS methodology, stating:76

In its Interim Standards Report the National Academy recommended a methodology for predicting the emission levels achievable by manufacturers. This recommended methodology is the same methodology that was employed in the technical appendix to my decision in evaluating the test results of all manufacturers except General Motors. (Emphasis added.)

The case is haunted by the irony that what seems to be Ford's technological lead77 may operate to its grievous detriment, assuming the relaxation-if-necessary approach voiced by Mr. Allen,78 If in 1974, when certification of production vehicles begins, any one of the three major companies cannot meet the 1975 standards, it is a likelihood that standards will be set to permit the higher level of emission control achievable by the laggard. This will be the case whether or not the leader has or has not achieved compliance with the 1975 standards. Even if the relaxation is later made industry-wide, the Government's action, in first imposing a standard not generally achievable and then relaxing it, is likely to be detrimental to the leader who has tooled up to meet a higher standard than will ultimately be required .

In some contexts high achievement bestows the advantage that rightly belongs to the leader, of high quality. In this context before us, however, the high achievement in emission control results, under systems presently available, in lessened car performance-an inverse correlation. The competitive disadvantage to the ecological leader presents a forbidding outcome-if the initial assumption of feasibility is not validated, and there is subsequent relaxation-for which we see no remedy.79

We now take up the serious contention of International Harvester (IH) that the EPA decision effectively rules out the production of 1975 model year IH light weight trucks and multi-purpose passenger vehicles (MPVs). This requires us to focus on the Administrator's conception that the 1970 Clean Air Act envisioned restricting production of vehicles to that necessary to fill "basic demand."80

The Administrator does not dispute International Harvester's claim that it will not be able to produce the vehicles in question, and indeed the limited testing of one of its MPVs showed, even as evaluated by EPA methodology, that such standards could not be achieved.81 Yet a suspension was not granted, presumably for the reasons advanced by EPA to this court, that International Harvester was "required to alter the performance characteristics of its vehicles in the interest of meeting the 1975 emission standards."82 The inability of IH vehicles to meet the standards seems accountable by the uses to which they are put, hauling large loads or towing heavy trailers. To serve this purpose vehicles must be designed with higher than normal axle ratios, thus requiring greater power from the engine and producing higher exhaust gas temperatures in order to attain any given speed.83 Therefore, for all practical purposes a redesign of performance characteristics will preclude the present uses to which IH vehicles are put.

The Administrator, nonetheless, takes the position that International Harvester can be denied a suspension because he has found that "new car demand" will be satisfied by the production of the major auto companies, and thus apparently posits that the absence from the 1975 market of all light weight trucks and MPVs is fully consistent with the Act. We cannot agree.

Section 202(b)(1) of the Act applies its drastic standards to 1975 models of "light duty vehicles." It is our view that the legislative history reveals this term to mean "passenger cars." In the Report of the Senate Committee on Public Works on S.4358,84 the Committee clearly distinguished between the automobile, which must "meet a rigid timetable and a high degree of emission control compliance," and other vehicles, such as "trucks and buses and other commercial vehicles," which are governed by a different authority to promulgate standards. At another point of the Senate Report, the legislative use of the term light duty vehicles, as interchangeable with passenger cars, is made even more clear:85

The authority provided in section 202 (a) would continue to be available to the [Administrator] to establish standards for light duty motor vehicles (passenger cars) during the period prior to and following the effective date of the standards established by subsection (b).

References abound in Congressional debate to the same effect.86 This kind of legislative intent must be given priority, in interpreting this law, over any presumption of continuance of prior administrative definitions of this term87 or to the policy of upholding reasonable interpretations of statutes by administrative agencies88 in the absence of other discernible legislative intent. Volkswagenwerk v. FMC, 390 U.S. 261, 272, 88 S. Ct. 929, 19 L. Ed. 2d 1090 (1967); Greater Boston Television Corp. v. FCC (I), 143 U.S.App.D.C. 383, 392, 444 F.2d 841, 850, cert. denied, 403 U.S. 923, 91 S. Ct. 2229, 29 L. Ed. 2d 701 (1971).

For the above reasons we cannot sustain the definition of "Light duty vehicle" as:89

any motor vehicle either designed primarily for transportation of property and rated at 6,000 pounds GVW or less or designed primarily for transportation of persons and having a capacity of 12 persons or less

to the extent that it includes light weight trucks in the category that must meet the drastic emission reduction standards set for 1975 models. These light weight trucks will be governed by the standards duly promulgated by EPA for "trucks and buses and other commercial vehicles."

This is not to say that the modification of the "light duty vehicles" definition must exclude MPVs, which largely overlap in their usage with passenger cars. We merely hold the present regulation contrary to legislative intent. We have jurisdiction to decide this issue, even though the reasonableness of the regulation could be challenged in a separate proceeding in the District Court,90 because the validity of the regulation is a premise of the refusal to grant suspension. "It would be an empty and useless thing to review an order . . . based on a regulation the validity of which might be subsequently nullified." Doe v. Civil Aeronautics Board, 356 F.2d 699, 701 (10th Cir. 1966).

We decline the proposal of International Harvester, therefore, that only its vehicles be granted a suspension. Light weight trucks of other manufacturers, such as Ford, equally demonstrated an inability to comply with the 1975 standards.91 Under the view taken here, the light weight trucks of all manufacturers are properly exempted from the scope of "light duty vehicles." This comports with competitive as well as statutory considerations, as the Administrator's own brief delineates:92

If International Harvester is granted a suspension, it should be able to sell its vehicles at a lower cost than competitors who met the standards. This is so because International Harvester's 1975 models would not include expensive catalytic devices to control emissions. Also the Company's vehicles would probably perform better for the same reason. Thus, if suspension is granted, it is likely that International Harvester will gain a substantial competitive advantage over manufacturers who sacrificed the performance of their vehicles, and perhaps profits, in order to comply with the 1975 standards.

Assuming light duty vehicles are defined by EPA to include MPVs a question may arise whether they are entitled to a one-year suspension, for lack of feasibility, even though passenger vehicles generally should be denied a suspension. We shall not consider this question unless and until EPA has had an opportunity to address itself to the problems in the light of our opinion herein.

D. The Issue of Feasibility Sufficient for Basic Auto Demand

The foregoing conclusion is not to be misunderstood as amounting to an acceptance of another "basic demand" contention raised by the auto manufacturers. We are inclined to agree with the Administrator that as long as feasible technology permits the demand for new passenger automobiles to be generally met, the basic requirements of the Act would be satisfied, even though this might occasion fewer models and a more limited choice of engine types. The driving preferences of hot rodders are not to outweigh the goal of a clean environment.

A difficult problem is posed by the companies' contention that the production and major retooling capacity does not exist to shift production from a large number of previous models and engine types to those capable of complying with the 1975 standards and meeting the demand for new cars. The Administrator made no finding as to this problem. We believe the statute requires such a finding, explaining how the Administrator estimates "basic demand" and how his definition conforms to the statutory objective. The emission standards set for 1976 cannot be breached, since they represent an absolute judgment of Congress. But as to the decision on a oneyear suspension, and the underlying issue of technological feasibility, Congress intended, we think, that the Administrator should take into account such "demand" considerations.

A significant decrease in auto production will have a major economic impact on labor and suppliers to the companies. We have no reason to believe that "effective technology" did not comport within its meaning sufficient technology to meet a basic level of consumer demand.

This case inevitably presents, to the court as to the Administrator, the need for a perspective on the suspension that is informed by an analysis which balances the costs of a "wrong decision" on feasibility against the gains of a correct one. These costs include the risks of grave maladjustments for the technological leader from the eleventh-hour grant of a suspension, and the impact on jobs and the economy from a decision which is only partially accurate, allowing companies to produce cars but at a significantly reduced level of output. Against this must be weighed the environmental savings from denial of suspension. The record indicates that these will be relatively modest. There is also the possibility that failure to grant a suspension may be counter-productive to the environment, if there is significant decline in performance characteristics.

Another consideration is present, that the real cost to granting a suspension arises from the symbolic compromise with the goal of a clean environment. We emphasize that our view of a one year suspension, and the intent of Congress as to a one year suspension, is in no sense to be taken as any support for further suspensions. This would plainly be contrary to the intent of Congress to set an absolute standard in 1976. On the contrary, we view the imperative of the Congressional requirement as to the significant improvement that must be wrought no later than 1976, as interrelated with the provision for one-year suspension. The flexibility in the statute provided by the availability of a oneyear suspension only strengthens the impact of the absolute standard. Considerations of fairness will support comprehensive and firm, even drastic, regulations, provided a "safety valve" is also provided-ordinarily a provision for waiver, exception or adjustment, in this case a provision for suspension.93 "The limited safety valve permits a more rigorous adherence to an effective regulation." WAIT Radio v. FCC, supra, 135 U.S.App.D.C. at 323, 418 F.2d at 1159. To hold the safety valve too rigidly is to interfere with the relief that was contemplated as an integral part of the firmness of the overall, enduring program.

We approach the question of the burden of proof on the auto companies with the previous considerations before us.

IV. THE REQUIRED SHOWING ON "AVAILABLE TECHNOLOGY"

It is with utmost diffidence that we approach our assignment to review the Administrator's decision on "available technology." The legal issues are intermeshed with technical matters, and as yet judges have no scientific aides. Our diffidence is rooted in the underlying technical complexities, and remains even when we take into account that ours is a judicial review, and not a technical or policy redetermination, our review is channeled by a salutary restraint, and deference to the expertise of an agency that provides reasoned analysis. Nevertheless we must proceed to the task of judicial review assigned by Congress.

The Act makes suspension dependent on the Administrator's determination that:

the applicant has established that effective control technology, processes, operating methods, or other alternatives are not available or have not been available for a sufficient period of time to achieve compliance prior to the effective data of such standards . . . .

A. Requirement of Observed Data From Manufacturers

Clearly this requires that the applicants come forward with data which showed that they could not comply with the contemplated standards. The normal rules place such a burden on the party in control of the relevant information.94 It was the auto companies who were in possession of the data about emission performance of their cars.

The submission of the auto companies unquestionably showed that no car had actually been driven 50,000 miles and achieved conformity of emissions to the 1975 standards. The Administrator's position is that on the basis of the methodology outlined, he can predict that the auto companies can meet the standards, and that the ability to make a prediction saying the companies can comply means that the petitioners have failed to sustain their burden of proof that they cannot comply.

B. Requisite Reliability of Methodology Relied on by EPA To

Predict Feasibility Notwithstanding Lack of Actual

We agree with the Administrator's proposition in general. Its validity as applied to this case rests on the reliability of his prediction, and the nature of his assumptions. One must distinguish between prediction and prophecy. See EDF v. Ruckelshaus, 142 U.S. App.D.C. 74, 89, 439 F.2d 584, 597 (1971). In a matter of this importance, the predictor must make a showing of reliability of the methodology of prediction, when that is being relied on to overcome this "adverse" actual test data of the auto companies. The statute does not contemplate use of a "crystal ball." See National Resources Defense Council, Inc. v. Morton, 148 U.S.App.D.C. 5, 15, 458 F.2d 827, 837 (1972).

The Administrator, however, raises a different issue by contending that the companies, wholly aside from his methodology, did not submit sufficient evidence to enable him to make the required determination as to "available technology." This goes to the standard rather than the burden of proof, and comes close to adoption of "beyond a reasonable doubt" as the required showing. Aside from a possible finding of bad faith, which the Administrator specifically eschews making, this position cannot stand. The companies came forward with all the data that there was to be had, and the Administrator did not specifically ask for more. Additionally, our perspective on the interests furthered by a sound EPA decision, and jeopardized by a "wrong decision," are material to the issue of standard of proof. This is a situation where, as we have stated, the risks of an erroneous denial of suspension outweigh the risks of an erroneous grant. On the issue of burden of proof, the standard adopted must take into account the nature and consequences of risk of error. See In re Winship, 397 U.S. 358, 371-372, 90 S. Ct. 1068, 25 L. Ed. 2d 368 (1970) (Mr. Justice Harlan, concurring); U. S. v. Brown, 155 U.S.App.D.C. -, 478 F.2d 606 (1973). This view of the standard of proof dictates the standard normally adopted in civil matters, a preponderance of the evidence.95

Our approach relates considerations of ecological and economic costs, dealt with above, to the legal issue of burden and standard of proof. Nominally the statute, in Sec. 202(b)(5)(D), sets forth separate criteria as to "public interest," in clause (i), and "available technology," in clause (iii). But the assignment of the burden and standard of proof on "available technology" inescapably involves many of the same considerations as those involved in a "public interest" determination, and it would have been helpful to this court if the Administrator had expressly commented on the public interest in this connection.

The underlying issue is the reasonableness and reliability of the Administrator's methodology, for it alone offsets the data adduced by petitioners in support of suspension. It is the Administrator who must bear the burden on this matter, because the development and use of the methodology are attributable to his knowledge and expertise. When certain material "lies particularly within the knowledge" of a party he is ordinarily assigned the burden of adducing the pertinent information.96 This assignment of burden to a party is fully appropriate when the other party is confronted with the often-formidable task of establishing a "negative averment." United States v. Denver & R.G.R. Co., 191 U.S. 84, 92, 24 S. Ct. 33, 48 L. Ed. 106 (1903). In the context of this proceeding, this requires that EPA bear a burden of adducing a reasoned presentation supporting the reliability of its methodology.

The multiple assumptions used by the Administrator in making his prediction are subject to serious doubts.

The basic formula used to make the prediction that each of the manufacturers could meet the 1975 standards was based on 1975 certification requirements, so that in part it paralleled testing procedures which would be used in 1975 to certify automobiles for sale. The formula is:97

50,000 mile emissions = 4000 mile emissions x deterioration factor

Four kinds of assumptions were used in making the 50,000 mile emission prediction: (1) regulatory, (2) engineering or scientific, (3) techniques of application of basic formula to particular companies, and (4) statistical reliability of the final prediction.

First, EPA assumed that certain types of maintenance would have to be performed on 1975 model year cars, if its 50,000 miles emission predictions were to be meaningful. Subsequent to the issue of its Technical Appendix, a Proposed Rule Making formulated these requirements as part of 1975 certification procedure.98 This assumption was necessary because much of the data supplied by the companies was obtained from cars that were under rigid controls during testing.99 The problem with such maintenance assumptions is whether the ordinary driver will actually pay for this kind of maintenance just to reduce the emission levels of his automobile. It is one thing to build maintenance into the 1975 certification procedure, when fleet samples are durability tested. It is another to posit that such standards will be maintained, or are reasonably likely to be maintained, by consumers. A hard question is raised by the use of a methodological assumption without evidence that it will correspond to reality, or a reasonable and forthright prediction based on expertise.

Secondly, the predicted emission level assumes that there will be one total replacement of the catalytic converter at some time after 25,000 miles. This entered into the formula as an adjustment to the predicted deterioration factor.100 The critical question is how much will the one replacement reduce emissions otherwise obtainable by use of a single catalyst. This relationship had to be assumed because manufacturers had not used catalytic converter replacements in their testing. The Administrator admitted that this factor was imprecise.101 Yet, in the case of General Motors, the use of the assumed value of this factor was critical in allowing the Administrator to make a 50,000 mile emission prediction under the 1975 standards.102

The third regulatory assumption relates to the average lead level which will exist in gasoline available for 1975 model year cars. Lead levels in gasoline contribute to the levels of HC and CO both in terms of normal emission control achievable (the 4000 mile emission) and to the deterioration in emissions over time (deterioration factor). Thus, in the case of the Chrysler car used to predict conformity with the 1975 standards, a .03 lead in gasoline produced 4000 mile emissions of .27 grams HC and 1.51 CO, whereas a .05 level of lead resulted in .29 and 1.66 grams respectively. Similarly .03 lead produced a corrected deterioration factor of .67 HC and 1.5 CO, whereas a .05 level produced .73 HC and 1.65 CO.103

On December 27, 1972, a regulation was promulgated "designed to assure general availability by July 1, 1974, of suitable gasolines containing no more than .05 grams per gallon of lead. . . ."104 It was the assumption of the Administrator that the .05 maximum would result in gas containing on the average .03 grams per gallon of lead. The discrepancy between the maximum and average is accounted for by the contamination of lead free gasoline from its point of production to its marketing outlet. Thus EPA will allow a maximum of .05 but anticipates that on the average fuel will be at .03. This assumption is, however, subject to testimony in the record indicating a difference between companies in their ability to achieve gasoline with a low lead level complying with the proposed regulation. Amoco said that its proposal for a .07 maximum "should result in effective lead levels of .02 to .03 grams of lead per gallon."105 Texaco did not think it could deliver gas to service stations at a lead level below .07.106 We cannot resolve whether a differential ability really exists, but we also have no refinement and resolution by the EPA (as distinguished from the briefs of its counsel). We do not say this matter is a critical defect; still it leaves a residue of uncertainty that beclouds the EPA assumption of a .03 average, needed in its methodology to predict conformity with the 1975 standards.2. Engineering and scientific assumptions

Engineering or scientific assumptions are made in predicting 4000 mile emissions and deterioration factors, and we shall give separate consideration to each independent variable.

a. The 4000 mile emission factor

The use of 4000 mile emissions as a starting point is based on certification procedures.107 No challenge has been made to this mileage as a base point, largely because it appears that at this mileage the engine is broken in and emission levels are relatively stabilized.108 EPA decided to adjust raw data supplied, at least in the case of Ford and Chrysler, of emissions at 4000 miles to take account of a "Lead Adjustment Factor."109 This was done because in most cases emissions data reflected fuels with a close to zero lead level which had been used by the manufacturers in their testing programs.

This Lead Adjustment Factor was calculated using only Ford cars, but the value of the factor was assumed to be the same in adjusting Chrysler 4000 mile emissions with this factor.110 The cars had been tested with a dynamometer, a type of test equipment used for laboratory testing of an engine. A measurement of the efficiency of the catalytic converter at the 4000 mile mark was the critical value which had to be obtained from the dynamometer since this would indicate what the proper lead adjustment factor would be.111

EPA assumed that 200 hours on the dynamometer corresponded to 4000 miles usage, based on a critical and contested EPA assumption that the tests were conducted at 1000 RPM. Petitioners claim that the high temperature readings on the dynamometer reflect a higher RPM, and hence that a testing below 200 hours corresponded to 4000 miles of use. EPA disputes the steps in that chain of reasoning, and argues that a higher temperature may be attributable not to a RPM in excess of 1000, but to a heavy load on the vehicle, and in the alternative contends that even if there was a RPM greater than 1000, the speed may not have increased, due to a shift in gear.

The cause of higher than expected temperature readings cannot be ascertained from the record, and we are left with the alternative contentions of the parties. It is up to EPA, however, to support its methodology as reliable, and this requires more than reliance on the unknown, either by speculation, or mere shifting back of the burden of proof.112b. Deterioration factor

Methodological problems also existed with the calculation of the deterioration factor, which took account of possible deterioration in emission quality from 4000 miles to 50,000 miles. Different questions arose as to the calculation of this factor for Ford and Chrysler.

In the case of Ford, the Administrator predicted that emissions would improve from 4000 to 50,000 miles, and arrived at a deterioration factor of less than 1.113 He calculated average deterioration factors for Ford vehicles of .80 HC and .83 CO. This is to be compared with a deterioration factor of 2.5 used by NAS.114 The Administrator never explained why there should be no deterioration. Nor does EPA explain how this result can be squared with other data on Ford catalyst efficiencies, which was used in the case of the General Motors prediction, showing 50,000 mile catalyst efficiencies ranging from 21% to 53% for HC and 47% to 72% for CO.115

In the case of Chrysler, the deterioration factor was also calculated to be less than 1, but this figure was only arrived at after eliminating some data points from the emission measurement on the tested car #333, due to what EPA claimed were unrepresentative points resulting from non-catalyst malfunctions.116 Although it may be, as EPA argues here, that including the data points would still produce predicted 50,000 emission levels in conformity with the 1975 standard, the fact remains that these data points were removed. Moreover, it is not apparent why one should ignore malfunctions of a car which contribute to high emissions, even if they are not malfunctions of the converter. Malfunctions of cars occur to some degree, and cars operating in 1975 will undoubtedly be subject to them.

A lead adjustment factor is applied to the deterioration factor, as well as to 4000 mile emissions. EPA estimated on the basis of the questionable Ford dynamometer data, that lead levels had no observable effect, which was contrary to industry testimony on the subject.117 The Administrator evidently had doubts as to the dependability of these results as well, and therefore assumed a 10% factor for lead adjustment.118 No explanation is given of the origins of this 10% figure. If the willingness to take some factor evidences distrust in the data, the question then becomes whether 10% is enough.

In the case of General Motors an entirely different methodology from that used for Ford and Chrysler was employed. This was adopted due to limited testing by GM of noble metal catalysts.

The methodology was to take the raw emission values produced by a GM car prior to catalyst treatment of any kind multiplied by a factor representing the efficiency of the catalyst, i. e., the percentage of a given pollutant that the catalyst converts to harmless vapor, in order to obtain the projected overall emission performance at 50,000 miles.119 These methods of calculation were developed by the Administrator and were not used by NAS in their evaluation.120 The catalyst efficiency data were taken from Engelhard converters used principally on Ford cars and applied against the raw emissions of a General Motors engine. This assumed, with no explanation of the validity of such an assumption, that Engelhard catalysts will function as efficiently in General Motors cars as in those of Ford. A prediction was made on the basis of a hypothetical case. One cannot help be troubled by the adoption of this technique for General Motors. It was apparently recognized as at best a second best approach, in terms of the reliability of the prediction, or the same catalyst efficiency procedure would also have been used for Ford and Chrysler.

In this case the Administrator is necessarily making a prediction. No tests exist on whether this prediction is or is not reliable. It would, therefore, seem incumbent on the Administrator to estimate the possible degree of error in his prediction. The NAS, for example, said that the data of the manufacturers were subject to ± 20-30% margin of error,121 and this is separate from any margin of error that may be due to the various assumptions made by the Administrator. It is not decisive to say, as EPA argues in its brief, that this is just a matter of quality control in production. The first issue is whether the automobile built with rigid adherence to specifications will perform as predicted. The issue of quality control, whether cars will indeed be built in accordance with specifications, raises a separate and additional problem.

The possibility of error must take into account that only 1 Ford car, 1 Chrysler car, and 1 hypothetical General Motors car form the foundation for predicted conformity with the 1975 standard.122 The Administrator would say that it is enough to validate the principle of the electric light bulb if only one is seen at work. But we do not yet have one that has worked; instead we have four predictions. Questions like these arise: (1) For how many different types of engines will these predictions be valid? (2) Does it make a difference that the tested cars were experimental and driven under the most controlled conditions? The best car analysis of EPA raises even further doubts when considered alongside the NAS Report which used 55 vehicles in arriving at its recommended interim standard.123

We may sensibly begin our conclusion with a statement of diffidence.124 It is not without diffidence that a court undertakes to probe even partly into technical matters of the complexity of those covered in this opinion. It is with even more diffidence that a court concludes that the law, as judicially construed, requires a different approach from that taken by an official or agency with technical expertise. Yet this is an inescapable aspect of the judicial condition, though we stay mindful of the overarching consideration that a court's role on judicial review embraces that of a constructive cooperation with the agency involved in furtherance of the public interest.125A court does not depart from its proper function when it undertakes a study of the record, hopefully perceptive, even as to the evidence on technical and specialized matters, for this enables the court to penetrate to the underlying decisions of the agency, to satisfy itself that the agency has exercised a reasoned discretion, with reasons that do not deviate from or ignore the ascertainable legislative intent.126

In this case technical issues permeate the "available technology" determination which the Administrator made the focal point of his decision. In approaching our judicial task we conclude that the requirement of a "reasoned decision" by the Environmental Protection Agency means, in present context, a reasoned presentation of the reliability of a prediction and methodology that is relied upon to overcome a conclusion, of lack of available technology, supported prima faciely by the only actual and observed data available, the manufacturers' testing.

The number of unexplained assumptions used by the Administrator, the variance in methodology from that of the Report of the National Academy of Sciences, and the absence of an indication of the statistical reliability of the prediction, combine to generate grave doubts as to whether technology is available to meet the 1975 statutory standards. We say this, incidentally, without implying or intending any acceptance of petitioners' substitute assumptions. These grave doubts have a legal consequence. This is customarily couched, by legal convention, in terms of "burden of proof." We visualize the problem in less structured terms although the underlying considerations, relating to risk of error, are related. As we see it the issue must be viewed as one of legislative intent. And since there is neither express wording or legislative history on the precise issue, the intent must be imputed. The court must seek to discern and reconstruct what the legislature that enacted the statute would have contemplated for the court's action if it could have been able to foresee the precise situation.127 It is in this perspective that we have not flinched from our discussion of the economic and ecological risks inherent in a "wrong decision" by the Administrator. We think the vehicle manufacturers established by a preponderance of the evidence, in the record before us, that technology was not available, within the meaning of the Act, when they adduced the tests on actual vehicles; that the Administrator's reliance on technological methodology to offset the actual tests raised serious doubts and failed to meet the burden of proof which in our view was properly assignable to him, in the light of accepted legal doctrine and the intent of Congress discerned, in part, by taking into account that the risk of an "erroneous" denial of suspension outweighed the risk of an "erroneous" grant of suspension. We do not use the burden of proof in the conventional sense of civil trials, but the Administrator must sustain the burden of adducing a reasoned presentation supporting the reliability of EPA's methodology.

EPA's diligence in this proceeding, fraught with questions of statutory interpretation, technical difficulties and burdensome time constraints placed on the decision-making process, has been commendable. The agency was presented with a prickly task, but has acted expeditiously to carry out what it perceived to be a drastic mandate from Congress. This statute was, indeed, deliberately designed as "shock treatment" to the industry. Our central difference with the Administrator, simply put, stems from our view concerning the Congressional intent underlying the one year suspension provision. That was a purposeful cushion -with the twin purpose of providing "escape hatch" relief for 1975, and thus establishing a context supportive of the rigor and firmness of the basic standards slated for no later than 1976. In our view the overall legislative firmness does not necessarily require a "hardnosed" approach to the application for suspension, as the Administrator apparently supposed, and may indeed be furthered by our more moderate view of the suspension issue, particularly in assigning to the Administrator the burden of producing a reasoned presentation of the reliability of his methodology. This is not a matter of clemency, but rather a benign approach that moderates the "shock treatment" so as to obviate excessive and unnecessary risk of harm.

Our decision is also responsive to the differences between the EPA decision and the NAS Report. Although in some instances "the factual findings and technical conclusions"128 are consistent with those of the Administrator, the NAS conclusion was that technology was not available to meet the standards in 1975. Congress called on NAS, with presumed reliance on the knowledge and objectivity of that prestigious body, to make an independent judgment. The statute makes the NAS conclusion a necessary but not sufficient condition of suspension. While in consideration of the other conditions of suspension, EPA was not necessarily bound by NAS's approach, particularly as to matters interlaced with policy and legal aspects, we do not think that it was contemplated that EPA could alter the conclusion of NAS by revising the NAS assumptions, or injecting new ones, unless it states its reasons for finding reliability-possibly by challenging the NAS approach in terms of later-acquired research and experience.

These factors combine to convince us that, under our view of Congressional intent, we cannot affirm the EPA's denial of suspension as stated. That is not necessarily to assume, as at least some petitioners do, that the EPA's process must be brought to nullity.

The procedures followed in this case, whether or not based on rulings that were "mistaken" when made, have resulted in a record that leaves this court uncertain, at a minimum, whether the essentials of the intention of Congress were achieved. This requires a remand whereby the record as made will be supplemented by further proceedings. In the interest of justice, see 28 U.S.C. Sec. 2106, and mutual regard for Congressional objective, the parties should have opportunity on remand to address themselves to matters not previously put before them by EPA for comment, including material contained in the Technical Appendix filed by EPA in 1972 subsequent to its Decision.

It is contemplated that, in the interest of providing a reasoned decision, the remand proceeding will involve some opportunity for cross-examination. In the remand proceeding-not governed by the same time congestion as the initial Decision process-we require reasonable cross-examination as to new lines of testimony, and as to submissions previously made to EPA in the hearing on a proffer that critical questions could not be satisfactorily pursued by procedures previously in effect. There is, however, still need for expedition, both by virtue of our order and the "lead time" problem, and the EPA may properly confine crossexamination to the essentials, avoiding discursive or repetitive questioning.

Following our suggestion in Environmental Defense Fund, Inc. v. EPA, 150 U.S.App.D.C. 348, 465 F.2d 528 (1972), the Administrator may consider possible use of interim standards short of complete suspension. The statute permits conditioning of suspension on the adoption, by virtue of the information adduced in the suspension proceeding, of interim standards, higher than those set for 1974.129

We cannot grant petitioners' request that this court order a suspension since determinations which Congress made necessary conditions of suspension, as to the public interest and good faith, have not been made by the Administrator. The Administrator's decision did not reach these questions and accordingly we must remand for further consideration. The initial requirement that an EPA decision on the suspension, aye or nay, be made within 60 days of the application, obviously does not preclude further consideration following remand by the court. In the interest of justice, 28 U.S.C. Sec. 2106, and the Congressional intention that decisions be made timely in the light of considerations of "lead time" for 1975 model year production, we require the suspension deliberations by EPA to be completed within 60 days. The Administrator's decision on remand must, of course, be consistent with our legal rulings herein-including the need for redefinition of light duty vehicles, and promulgation of an appropriate regulation.

In conformance to the Congressional contemplation of expedition, and our responsibilities as an appellate court, we further require that the Administrator render a decision, on the basis of the best information available, which extends to all the determinations which the statute requires as a condition of suspension.130 We do not preclude further consideration of the question of "available technology," especially if developments in the art provide enlightenment. Last but not least, especially in view of Ford's submission and the NAS Report concerning interim standards, we reiterate that the EPA's determination may consist of a conditional suspension that results in higher standards than an outright grant of applications for suspension.

The case is remanded for further proceedings not inconsistent with this opinion.

Socrates said that wisdom is the recognition of how much one does not know.1 I may be wise if that is wisdom, because I recognize that I do not know enough about dynamometer extrapolations, deterioration factor adjustments, and the like to decide whether or not the government's approach to these matters was statistically valid. Therein lies my disagreement with the majority.

The court's opinion today centers on a substantive evaluation of the Administrator's assumptions and methodology. I do not have the technical know-how to agree or disagree with that evaluation-at least on the basis of the present record. My grounds for remanding the case rest upon the Administrator's failure to employ a reasonable decision-making process for so critical and complex a matter. At this time I cannot say to what extent I could undertake an evaluation of the Administrator's findings if they were based on an adequate decisional process.

I cannot believe that Congress intended this court to delve into the substance of the mechanical, statistical, and technological disputes in this case. Senator Cooper, the author of the judicial review provision, stated repeatedly that this court's role would be to "determine the question of due process."2 Thus the court's proper role is to see to it that the agency provides "a framework for principled decision-making."3 Such a framework necessarily includes the right of interested parties to confront the agency's decision and the requirement that the agency set forth with clarity the grounds for its rejection of opposing views.

The majority's interpretation of the present statute and the administrative precedents would give us no right to establish these procedural guidelines. Their opinion maintains that the strict deadlines in the Clean Air Act preclude any right to challenge the Administrator until after the decision has been made. It indicates that, since this hearing was "rule-making" rather than "adjudicatory", cross-examination and confrontation are not required under traditional rules of administrative law.

I understand this viewpoint, but I do not share it. I do not think the authors of the Clean Air Act intended to put such strict limits on our review of the Administrator's decision-making process. Further, the interests at stake in this case are too important to be resolved on the basis of traditional administrative labels. We recognized two years ago that environmental litigation represents a "new era" in administrative law.4 We are dealing here not with an airline's fares or a broadcaster's wattage, but with all humanity's interest in life, health, and a harmonious relationship with the elements of nature.

This "new era" does not mean that courts will dig deeper into the technical intricacies of an agency's decision. It means instead that courts will go further in requiring the agency to establish a decision-making process adequate to protect the interests of all "consumers" of the natural environment.5 In some situations, traditional rules of "fairness"-designed only to guard the interests of the specific parties to an agency proceeding-will be inadequate to protect these broader interests. This is such a case. Whether or not traditional administrative rules require it, the critical character of this decision requires at the least a carefully limited right of crossexamination at the hearing and an opportunity to challenge the assumptions and methodology underlying the decision.

The majority's approach permits the parties to challenge the Administrator's methodology only through the vehicle of judicial review. I do not think this is an adequate substitute for confrontation prior to the decision. I reach this position not only out of concern for fairness to the parties (". . . for if a party first learns of noticed facts through the final report . . . the burden of upsetting a decision announced as final is a heavy one."6 ) but also out of awareness of the limits of our own competence for the task. The petitioners' challenges to the decision force the court to deal with technical intricacies that are beyond our ken.7 These complex questions should be resolved in the crucible of debate through the clash of informed but opposing scientific and technological viewpoints.

It is true that courts occasionally find themselves in the thick of technological controversies-e. g., in patent cases. But those are different circumstances. We do not review patent disputes until they have been through a full panoply of procedures involving full rights of confrontation. Further, unlike our decision in a patent case, our decision on the Administrator's action here is sure to be tested by analysis and challenge in Congress, in the scientific community, and among the public.

My brethren and I are reaching for the same end-a "reasoned decision"-through different means. They would have us examine the substance of the decision before us. There are some areas of administrative law-involving issues of liberty and individual rights-where judges are on firm ground in undertaking a substantive review of agency action. But in cases of great technological complexity, the best way for courts to guard against unreasonable or erroneous administrative decisions is not for the judges themselves to scrutinize the technical merits of each decision. Rather, it is to establish a decision-making process which assures a reasoned decision that can be held up to the scrutiny of the scientific community and the public.8 "[T]he best test of truth is the power of the thought to get itself accepted in the competition of the market."9 If we were to require procedures in this case that open the Administrator's decision to challenge and force him to respond, we could rely on an informed "market" rather than on our own groping in the dark to test the validity of that decision.

Candor requires the admission that the process of confrontation and challenge might not be possible within the statutory decision period of 60 days. My response would be to permit an extension of the time limit-perhaps 30 days more. This would put less strain on the overall statutory scheme-and on the manufacturers' lead time-than the months that have been expended in litigation, and now a remand, over the decision. Congress did not intend for us to enforce this relatively minor time restriction so strictly as to do major damage to the statute as a whole.

My brethren argue that the 60-day time limit in the statute precluded any opportunity for cross-examination or confrontation at the time of the original decision. But their opinion would apparently permit these procedural rights on the remand. This bit of judicial legerdemain confounds me. I can find nothing in the statute or common sense to support this distinction. If anything, the statute, with its obvious emphasis on reaching a final decision quickly, would dictate procedures at the original decision which were sufficient to produce a reasoned decision without the need for a remand.

Outside of the foregoing differences, I agree with much of the majority opinion. I would have preferred to make the "public interest" factor-the considerations set forth in Part III of that opinion-an independent ground for suspension. The court today deals with the public interest indirectly, through the device of burden of proof. I do not fully understand this approach, but I suspect it leads to essentially the same result I favor.